See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/334285852

Remote sensing of European spruce (Ips typographus, L.) bark beetle green
attack

Thesis · June 2019

DOI: 10.3990/1.9789036547956

CITATIONS READS

6 2,007

1 author:

Haidi Abdullah
University of Twente
25 PUBLICATIONS 447 CITATIONS

SEE PROFILE

All content following this page was uploaded by Haidi Abdullah on 07 July 2019.

The user has requested enhancement of the downloaded file.

 Remote sensing of European spruce
(Ips typographus, L.) bark beetle green attack

Haidi Jamal Abdullah

 REMOTE SENSING OF EUROPEAN SPRUCE
(IPS TYPOGRAPHUS, L.) BARK BEETLE GREEN
ATTACK

DISSERTATION

to obtain
the degree of doctor at the University of Twente,
on the authority of the rector magnificus,
prof.dr. T.T.M. Palstra,
on account of the decision of the Doctorate Board,
to be publicly defended
on Thursday 27 June 2019 at 16:45 hrs

by

Haidi Jamal Abdullah

born on 05 January 1986
in Sulaymaniyah, Iraq
This thesis has been approved by
Prof.dr. A.K. Skidmore, supervisor
Dr. R. Darvishzadeh, co-supervisor

ITC dissertation number 357

ITC, P.O. Box 217, 7500 AE Enschede, The Netherlands

ISBN 978-90-365-4795-6
DOI 10.3990/1.9789036547956

Cover photo by Lio Khoshnaw

Cover designed by Benno Masselink
Printed by ITC Printing Department
Copyright © 2019 by Haidi Abdullah
Graduation committee:

Chairman/Secretary
Prof.dr.ir. A. Veldkamp University of Twente

Supervisor
Prof.dr. A.K. Skidmore University of Twente

Co-supervisor
Dr. R. Darviszadeh Varchehi University of Twente

Members
Prof.dr.ing. W. Verhoef University of Twente
Prof.dr. K. Pfeffer University of Twente
Prof.dr.ir. P.H. Verburg Free University Amsterdam
Prof.dr. O. Niemann University of Victoria, Canada
This thesis is dedicated to my family
for their love, endless support
and encouragement.
Acknowledgements
I would not be able to complete this thesis without the encouragement and
support of a number of people and organisations that I would like to sincerely
thank and acknowledge:

Firstly, I would like to express my appreciation and sincere gratitude to my

promoter, Prof. Andrew K. Skidmore. I feel very fortunate and honoured to
have worked with him; he was always there to motivate, support and guide
with his wisdom throughout my PhD. He taught me very important research
skills and problem-solving techniques, and he kindly guided and supported me
during the time of researching and writing this thesis. I will always remember
the early days when I have struggled writing my proposal, but you have always
supported and helped me, to stay on track. It has been an incredible honour
and privilege to do my PhD under your supervision.

I am also greatly indebted to the excellent supervision and advisory of Dr

Roshanak Darvishzadeh, who has become my supervisor after four months of
my PhD. The proposal development and field data collection phase of this work
would have not been possible without the precious support from Dr Roshanak
Darvishzadeh. Her guidance and suggestions during the proposal development
were pivotal to wipe out the early stage confusions I encountered. Moreover, I
am very grateful for the many inspiring scientific discussions we shared. Her
comments were always to the point that had the power to sharpen this work.
Above all, she is sincere, friendly, and a very easy person to communicate
with. Many thanks go to Dr.Thomas A. Groen, who had a great contribution to
my knowledge on the subject of this study during writing my proposal, and he
continued supporting me during writing my first scientific papers.

Thanks also go to the Bavarian Forest National Park particularly to Dr Marco

Heurich and Prof. Joerg Mueller for approving access to the study area, other
field and logistical facilities. I would like to thank the NRs department staffs,
especially Esther Hondebrink for her enormous assistance and support, ITC
student affairs, Geoscience laboratory team, ITC library, ICT help desk, travel
unit and finance department staffs who helped me during my stay at ITC. My
special thanks go to Loes Colenbrander, Marga Koelen, Carla Gerritsen,
Theresa van den Boogaard and Roelof Schoppers for their excellent services.

My deep gratitude goes to the PhD community at ITC. I have been fortunate
to come across wonderful colleagues at NRS department. It’s hard to mention
all the friends and colleagues. However among them, Dr. Elnaz Neinavaz, Dr.
Abebe Ali, Dr. Xi Zhu, Dr. Wanderi Festus, Dr. Maria Fenanda, Dr. Jing Liu,
Gara Tawanda, Alby Duarte Rocha, Nina Amiri, Yifang Shi, Trini Del Rio, and
Sugandh Chauhan, I would like to thank them for their friendship and valuable
support. It has been a great pleasure working with you and discussing different
scientific topics during our special PhD tutorial every month. Special thanks
also go to my relatives and friends back home and abroad for their relentless
support and encouragement.

Most of all I would like to thank all my family in particular my parents who, as
always, has given me endless love and support throughout my PhD journey.
In particular, I would like to thank my wife Jwan Anwer, for her love,
understanding and inspiration. To my sons Hasti and Hema, I am full of
apologies for being away from all of you.

Finally, I would like to acknowledge the European Commission’s Erasmus

Mundus program for awarding me PhD scholarship and Faculty of Geo-
information Science and Earth Observation (ITC) for the financial support to
pursue this study. Without this financial support, my dream would not have
been accomplished. My appreciation and gratitude go to Salahaddin University-
Erbil as well for offering me a sabbatical to conduct my PhD study.

ii
Table of contents
Acknowledgements ............................................................................. i
List of figures .................................................................................... vi
List of tables ...................................................................................... x
Chapter 1 : General Introduction ....................................................... 1
1.1 Biology and ecology of European spruce bark beetle
(Ips typographus, L.) ...............................................................2
1.2 Host tree response and control strategies ...................................3
1.3 Remote sensing of Ips typographus, L. green attack .....................6
1.4 Research objectives ............................................................... 10
1.5 Thesis outline ....................................................................... 10
Chapter 2 : European spruce bark beetle (Ips typographus, L.) green
attack affects foliar reflectance and biochemical properties ............ 13
Abstract ......................................................................................... 14
2.1 Introduction ......................................................................... 15
2.2 Material and Methods ............................................................. 18
2.2.1 Study area ........................................................................... 18
2.2.2 Data acquisition .................................................................... 19
2.2.3 Chemical analyses ................................................................. 20
2.2.4 Reflectance measurements ..................................................... 20
2.2.5 Data preprocessing and statistical analysis ................................ 21
2.2.5.1 Partial least square regression model (PLSR) ............................ 22
2.3 Results ................................................................................ 23
2.3.1 Spectral reflectance variation due to bark beetle green attack...... 23
2.3.2 Impact of bark beetle green attacks on foliar biochemical
properties ............................................................................ 23
2.3.3 Effects of bark beetle green attack on the retrieval accuracy
of leaf biochemical properties. ................................................. 24
2.4 Discussion ............................................................................ 27
2.5 Implications for remote sensing applications ............................. 30
2.6 Conclusion ........................................................................... 31
Chapter 3: Timing of red-edge and shortwave infrared reflectance
critical for early stress detection induced by bark beetle (Ips
typographus, L.) attack .................................................................... 33
Abstract ......................................................................................... 34
3.1 Introduction ......................................................................... 35
3.2 Material and methods ............................................................ 38
3.2.1 Study area ........................................................................... 38
3.2.2 Field data collection ............................................................... 38
3.2.3 Satellite imagery ................................................................... 40
3.2.3.1 Ancillary data (reference disturbance data) .............................. 40
3.2.3.2 Spectral vegetation indices .................................................... 41
3.2.4 Data analysis ........................................................................ 42

iii
 3.3 Results ................................................................................ 43
3.3.1 Temporal response of measured leaf traits under bark beetle....... 43
3.3.2 Temporal response of canopy spectral data to bark beetle
infestation ............................................................................ 45
3.4 Discussion ............................................................................ 49
3.5 Conclusion ........................................................................... 51
Chapter 4: Sensitivity of Landsat-8 OLI and TIRS Data to Foliar
Properties of Early Stage Bark Beetle (Ips typographus, L.) Infestation
........................................................................................................ 53
Abstract ......................................................................................... 54
4.1 Introduction ......................................................................... 55
4.2 Material and Methods ............................................................. 58
4.2.1 Study site and in situ data collection ........................................ 58
4.2.2 Measurement of leaf properties ............................................... 60
4.2.3 Landsat-8 imagery and pre-processing ..................................... 60
4.2.3.1 Spectral vegetation indices .................................................... 61
4.2.3.2 Canopy surface temperature (CST) ......................................... 63
4.2.4 Statistical Analysis................................................................. 66
4.3 Mapping bark beetle green attack infestation............................. 67
4.3 Results ................................................................................ 68
4.3.1 The importance of CST versus SVIs to estimate measured
leaf properties ...................................................................... 68
4.3.2 Temporal response of CST and SVIs under spruces bark
beetle infestation .................................................................. 69
4.3.3 Mapping bark beetle green attack and validation ........................ 72
4.4 Discussion ............................................................................ 74
4.5 Conclusion ........................................................................... 77
Chapter 5: Sentinel-2 accurately maps green attack stage of European
spruce bark beetle (Ips typographus, L.) compared to Landsat-8 .... 79
Abstract ......................................................................................... 80
5.1 Introduction ......................................................................... 81
5.2 Material and Methods ............................................................. 83
5.2.1 Study area and field data collection.......................................... 83
5.2.2 Satellite imagery ................................................................... 85
5.2.3 Spectral vegetation indices calculation...................................... 87
5.2.4 Statistical analysis ................................................................. 87
5.2.5 Ancillary data and accuracy assessment ................................... 89
5.3 Results ................................................................................ 91
5.3.1 Impact of bark beetle green attack on measured leaf traits ......... 91
5.3.2 Leaf and canopy spectral variations.......................................... 92
5.3.3 Principal component analysis (PCA) and ANOVA test ................... 94
5.3.4 Mapping bark beetle green attack and validation ........................ 95
5.4 Discussion ............................................................................ 99
5.5 Conclusion ......................................................................... 102

iv
Chapter 6 ....................................................................................... 105
Synthesis: Remote sensing of European spruce (Ipstypographus, L.)
bark beetle green attack ................................................................ 105
6.1 Introduction ....................................................................... 106
6.2 Effects of bark beetle green attack on foliar reflectance
and biochemical properties ................................................... 107
6.3 Understanding dynamic changes of Norway spruce trees at
both leaf and canopy level under bark beetle infestation during
green to red attack stage ..................................................... 109
6.4 Sensitivity of Landsat-8 OLI and TIRS Data to Foliar Properties
of Early Stage Bark Beetle (Ips typographus, L.) Infestation ...... 111
6.5 Detection and mapping bark beetle green attack using
spectral vegetation indices from Landsat-8 and Sentinel-2 data . 113
6.6 Practical relevance .............................................................. 115
6.7 The relative importance of ground-based measurements
for bark beetle green attack detection .................................... 117
6.8 Implications for commercial forestry and protected
area management: .............................................................. 118
6.9 Future of remote-sensing platforms in monitoring bark
beetle infestation: UAV and space-borne hyperspectral satellites 119
6.10 Future research Avenues ...................................................... 122
Bibliography ................................................................................... 123
Summary ....................................................................................... 147
Samenvatting ................................................................................. 149
Author's Biography ........................................................................ 151

v
List of figures
Figure 1.1: The generalised life cycle of the European spruce bark beetle and
the process of killing host trees. ................................................................... 5

Figure 2.1: The location of Bavarian Forest National park in Central Europe.
...................................................................................................................... 19
Figure 2.2: Dry dust produced through boring by the bark beetle Ips
typographus, L.; used to identify freshly infested trees in the field........ 20
Figure 2.3: Mean reflectance spectra of healthy and infested leaves at the
green attack stage. Gray areas depict the location of wavebands displaying
is a significant difference between healthy and infested spectra. ........... 24
Figure 2.4: Distribution of measured chlorophyll and nitrogen concentration
for healthy and infested needles. There is a significant difference (p <
0.05) in chlorophyll and foliar nitrogen concentration between healthy and
infested sample. ........................................................................................... 24
Figure 2.5: Measured versus predicted foliar chlorophyll (a) and nitrogen (b)
concentration for healthy and infested samples derived from the PLSR
analysis. ........................................................................................................ 25
Figure 2.6: Correlation between chlorophyll and nitrogen and the reflectance
of individual wavebands. The bold lines represent the wavebands at which
the reflectance spectra correlated significantly with the chlorophyll and
nitrogen concentrations. The arrows show the wavebands in which
recorded the highest correlation with chlorophyll content in both healthy
and infested samples. .................................................................................. 26
Figure 2.7: Importance of wavelengths corresponding to the highest value of
variable importance in the projection scores of partial least squares
regression in healthy and infested samples, used for chlorophyll and
nitrogen estimation ...................................................................................... 27
Figure 2. 8: Standardized residuals of predicted total chlorophyll and nitrogen
concentration for healthy and infested samples. ...................................... 30

Figure 3.1: Distribution of sample plots in the southern part of Bavarian Forest
National Park, Germany. ............................................................................. 41
Figure 3.2: Temporal variation of the measured leaf traits for healthy and
infested samples. Blue and black lines represent healthy and infested
plots, respectively. T1, T,2 and T3 represent first, second, and third
measurements, respectively. (*) Hardly significant, (**) significant, (***)
Strongly significant. ..................................................................................... 44
Figure 3.3: Mean reflectance spectra of healthy and green attacked leaves at
three consecutive repeated time measurements in the summer of 2015.
Blue and black lines represent healthy and infested leaves, respectively.
Wavebands in which there is a significant difference between healthy and
green attacked leaves are presents in Gray. ............................................. 45
Figure 3.4: (A) Temporal variation of canopy reflectance for healthy and
infested plots in the Bavarian Forest National Park using RapidEye satellite
data. Red and Green boxes represent infested and healthy plots,
respectively (B) Unpaired t-test canopy reflectance between healthy and
infested samples. Dark squares indicate spectral wavebands that were

vi
 significantly different (P ≤ 0.05). The red box shows the spectral region
that was significantly different overall temporal data considered in this
study ............................................................................................................. 46
Figure 3.5: (A) Temporal variation of reflectance spectra for healthy and
infested plots in the Bavarian Forest National Park using SPOT-5 satellite
data. Red and Green boxes represent infested and healthy plots,
respectively (B) Unpaired t-test of canopy reflectance between healthy and
infested samples. Dark squares indicate spectral wavebands that were
significantly different (P ≤ 0.05). The red box shows the spectral region
that was significantly different overall temporal data considered in this
study ............................................................................................................. 47
Figure 3.6: Temporal variation of studied spectral vegetation indices for
healthy and infested plots in the Bavarian Forest National Park using
RapidEye satellite data. Green and black lines represent healthy and
infested plots, respectively. ........................................................................ 48
Figure 3.7: Temporal variation of studied spectral vegetation indices for
healthy and infested plots in the Bavarian Forest National Park using SPOT-
5 satellite data. Green and black lines represent healthy and infested plots,
respectively. ................................................................................................. 49

Figure 4.1: Studies that have used remote sensing data to attempt detecting
bark beetle infestation at the green attack stage. .................................... 56
Figure 4.2: The location of Bavarian Forest National Park in Central Europe.
...................................................................................................................... 59
Figure 4.3: The framework of the research methodology. ............................ 68
Figure 4.4: Distribution of measured leaf traits for healthy and infested
needles, in the Bavarian Forest National Park, July 2016. There is a
significant difference (p < 0.05) in all measured leaf traits between healthy
and infested samples. .................................................................................. 69
Figure 4.5: PLSR variable importance in the projection (VIP) for predictor
(CST and SVIs) and response variables (measured leaf traits ) in the
Bavarian Forest National Park, July 2016. (vertical black line represents
VIP ≥ 1) ........................................................................................................ 69
Figure 4.6: Temporal variation of the SVIs and CST (VIP >1 ) for healthy and
green attacked sample plots in the Bavarian Forest National Park. Blue and
red boxes represent healthy and infested plots, respectively. A * indicates
the significant different................................................................................ 71
Figure 4.7: Cluster plots based on the first two principal components (PCs),
blue squares and red circles represent healthy and infested plots,
respectively. (A) Cluster plots based on all spectral indices VIP >1
including CST (B) Cluster plots based all spectral indices VIP >1 excluding
CST. ............................................................................................................... 72
Figure 4.8: The map showing the distribution of Norway spruce stands under
different stress levels in the Bavarian Forest National Park, July 2016. The
map was produced using canopy surface temperature and measured leaf
traits (foliar stomatal conductance, chlorophyll fluorescence, and water
content) ........................................................................................................ 74
Figure 4.9: Mean canopy surface temperature over different habitat classes
in the Bavarian Forest National Park – 2016. The graph shows that

vii
 standing dead-wood, lying dead-wood, clear-cut and rocky areas recorded
high surface temperature comparing to other forest classes................... 77

Figure 5.1: Location of healthy and green attacked sample plots in the
Bavarian Forest National Park in July 2016....................................... 85
Figure 5.2:The flow chart of the methodology used in this study .............. 86
Figure 5.3: Mean and standard deviation for measured leaf traits at both leaf
(A) and plot level (B). Leaf level data is obtained from the average
measurements per tree and plot level data is obtained from the average
measurements of representative trees within each plot. ..................... 92
Figure 5.4: Mean reflectance spectra of healthy and infested foliar at the
green attack stage. Grey areas depict the location of wavebands displaying
is a significant difference between healthy and infested spectra. ......... 93
Figure 5.5: (A) foliar and canopy reflectance using Sentinel-2 data (B) foliar
and canopy reflectance using Landsat-8 data. .................................. 93
Figure 5.6: Comparison of p-values from one-way ANOVA analysis between
SVIs for healthy and infested samples (red dash line represents the
significant level (p < 0.05)). .......................................................... 94
Figure 5.7: Cluster plots based on the first two PCs: (A) Cluster plots based
on all spectral indices including raw spectral bands (B) Cluster plots based
on pigment indices and (C) cluster plots based on water-related indices.
................................................................................................. 95
Figure 5.8: Selection probability value of SVIs obtained from PLS-DA Random
frog algorithm. ............................................................................ 96
Figure 5.9: The box plot shows the variation in SVI values calculated from
Landsat-8 and Sentinel-2, between healthy and infested plots. The black
and hollow boxes represent healthy and infested plots, respectively. The
red box shows the selected threshold value of each SVI where there is no
overlap between the two sample groups (healthy and infested) .......... 98
Figure 5.10: Map showing the spruce cover under green attack stress in the
Bavarian Forest National Park in July 2016 based on spectral vegetation
indices from Landsat-8 and Sentinel-2 selected through the Random frog
algorithm. ................................................................................... 99

Figure 6.1: Research framework and logical established links between

chapters and utilised data. .......................................................... 107
Figure 6.2: Importance of wavelengths corresponding to the highest value of
variable importance in the projection scores of partial least squares
regression in healthy and infested samples. ................................... 108
Figure 6.3: Measured versus predicted foliar chlorophyll and nitrogen
concentration for healthy (a) and infested (b) samples derived from the
PLSR analysis. ........................................................................... 109
Figure 6.4: (A) Temporal variation of studied spectral vegetation indices for
healthy and infested plots in the Bavarian Forest National Park using
RapidEye and SPOT-5 satellite data. Green and black lines represent
healthy and infested plots, respectively; (B) Unpaired t-test of canopy
reflectance between healthy and infested samples. Dark squares indicate
spectral wavebands that were significantly different (P ≤ 0.05). The red
box shows the spectral region that was significantly different overall
temporal data considered in this study. ......................................... 111

viii
Figure 6.5:Temporal variation of CST for healthy and green attacked sample
plots in the Bavarian Forest National Park. Green and red boxes represent
healthy and infested plots, respectively. A * indicates the significant
difference between healthy and infested plots, respectively. A * indicates
the significant difference between healthy and infested plots obtained
using Student t-tests, and a blue o indicates an outlier. .................. 112
Figure 6.6: Correlation between studied leaf traits (chlorophyll fluorescence,
leaf water content, and stomatal conductance) and SVIs, in the Bavarian
Forest National Park, July 2016. ................................................... 113
Figure 6.7: foliar and canopy reflectance using (A) Sentinel-2 data (B)
Landsat-8 data (C & D) correlation between canopy reflectance and foliar
spectra (interpolated from Field ASD spectrometer to Sentinel-2 and
Landsat-8 spectral bands) for both healthy and infested plots. .......... 114
Figure 6.8: The main planned hyperspectral sensors to be launched in the
next five years........................................................................... 120

ix
List of tables
Table 1.1: Existing literature that utilised various remote sensing data
(including aerial photography) to detect bark beetles at the green attack
stage over the last 50 years. ............................................................ 9

Table 2.1: Number of latent variables, R2 and nRMSEp between measured

and predicted chlorophyll and nitrogen in healthy and green attacked
needles. ...................................................................................... 25

Table 3.1: The leaf traits measured and overview of satellite data acquisition
in the Bavarian Forest National Park, 2015. The () are the SPOT-5 data
and (•) are the RapidEye data. T1, 2, and 3 represent first, second, and
third field measurements, respectively. ........................................... 40
Table 3.2: List of spectral vegetation indices calculated from SPOT-5 and
RapidEye data for both healthy and infested sample plots. ................. 42

Table 4.1: The Landsat-8 sensors, the operational land imager (OLI), and the
thermal infrared sensor (TIRS) spectral bands and their spatial resolution.
.................................................................................................. 62
Table 4.2:List of spectral vegetation indices calculated from Landsat-8 optical
bands. ......................................................................................... 62
Table 4.3:Stress level categories classified using the range of the leaf traits
(leaf water content, stomatal conductance, and chlorophyll fluorescence )
from the field, in the Bavarian Forest National Park, July 2016. .......... 68
Table 4.4:Regression equations between canopy surface temperature (CST)
and measured leaf properties. ........................................................ 73
Table 4.5:Assessment of the generated map using reference data obtained
from Aerial photography. ............................................................... 73

Table 5.1: The leaf traits measured for healthy and infested samples in this
study. Fw, Dw and A represent fresh leaf weight (g), dry leaf weight (g),
leaf dry mass per unit area (Cm), and leaf area (cm2 ), respectively. .. 85
Table 5.2: Spectral vegetation indices applied to leaf reflectance
measurements, Sentinel-2 and Landsat-8 in the study area. .............. 90
Table 5.3: One-way ANOVA test between healthy and infested reflectance
data at both leaf (simulated) and canopy level. ................................ 93
Table 5.4: Assessment of the generated maps from Landsat-8 and Sentinel-2
SVIs using the reference data obtained from Aerial photography. ....... 97
Table 5.5: Assessment of the mismatched pixels from Landsat-8 and Sentinel-
2 SVIs using previous years’ (2014 and 2015) infestation data (obtained
from Aerial photography). ............................................................ 102

Table 6.1:Comparison between UAV and other remote-sensing platforms.

................................................................................................ 121

x
Chapter 1 : General Introduction

1
Introduction

1.1 Biology and ecology of European spruce bark

beetle (Ips typographus, L.)
The eight-spined spruce bark beetle Ips typographus L. is one of the most
important and serious insect pests in Central Europe which threaten the
Norway spruce tree (Picea abies) (Lausch et al., 2011; Wermelinger, 2004).
The flight activity of the European spruce bark beetle depends strongly on the
air temperature as the beetles start to swarm and attack new host trees during
the European springtime (between the end of April and the beginning of May)
when the air temperature reaches 16.5 °C and maximum temperature does
not exceed 26 °C (Lobinger, 1994a; Wermelinger, 2004). The male initiates
gallery construction and then emits aggregation pheromones which attract
other beetles to the host tree (Zuber and Benz, 1992). The spruce bark beetle
feeds and reproduces within the phloem tissue under the bark of the infested
trees.

Moreover, in the process of the infestation, the beetle also carries pathogenic
fungi such as blue stain fungi (Ophiostoma and Cerato-cystis species) (Krokene
and Solheim, 1998). This combination of the beetle larva and the fungi may
cut the translocation of water, sugar and other nutrients within the bole of the
tree when both began to penetrate through the living phloem and xylem cells
(Paine et al., 1997b; Wermelinger, 2004). This feeding behaviour impairs the
transport of nutrients from the leaves massively to the roots (Rohde et al.,
1996). From the middle of June through to the end of July, the imagoes then
leave their host tree and attack new trees during their second stage of
development. The development of the European spruce beetle, from laying the
eggs to adulthood, lasts 7-11 weeks. During winter time, all the beetles at
different development stages (larva, pupa and adult) hibernate in the galleries;
the adults can also hibernate in stumps, litter or mineral soil over the winter.
In general, the European spruce beetle Ips typographus, L. has one to two
generations per year, depending on weather conditions, in particular,
temperature. The first generation are active from April to May, while the second
generation is active from July to August (Nikolov et al., 2014).

Generally, the bark beetle infestation occurs on two levels; (i) on an endemic
level (non-outbreak) (small bark beetle population) and (ii) on an epidemic
level (outbreak level). During the endemic level (non-outbreak), old trees with
big diameter (DBH), weakened by storms and drought, are susceptible to be
attacked by bark beetles (Allain et al., 2011; Latifi et al., 2014; Lausch et al.,
2011; Lausch et al., 2013b; Wermelinger, 2004). Therefore, the structure and
composition of a forest may not be greatly affected on top of the damage
already inflicted by storms or drought. This is because healthy trees may erect
defences by producing resin or latex which may contain a number of
insecticidal and fungicidal compounds that can kill or injure attacking insects,

2
 Chapter 1

or the sticky fluid simply immobilises and suffocates them. However, on an

epidemic level (outbreak), the large number of beetles can overwhelm a tree's
defences, resulting in the death of even healthy trees (Christiansen and Bakke,
1988a; Sambaraju et al., 2012; Skakun et al., 2003).

Historically, in central Europe, outbreaks of bark beetle have often coincided

with certain abiotic disturbances such as windthrows trees, heat waves, snow
damage (Öhrn, 2012; Seidl et al., 2017b). However, recent outbreaks are
different, as climate moderation has acted as their initiating factor. Dry, warm
summers have led to the desiccation of host trees, which then become stressed
and attractive for the bark beetle. As a result, an increase in temperature may
lead to an increase in beetle population size, as their eggs hatching and
developing into adults sooner than usual. In addition, an increase in the beetle
population size may have an impact on existence healthy trees, because the
trees’ defence system cannot cope with an attack from a large number of
beetles (Wermelinger, 2004). An increase in temperature will also stress
healthy trees and increase the probability of an attack by bark beetles.
Moreover, there are a number of abiotic factors that have impact on bark
beetles and host trees including elevation, slope and aspect. These factors
influence the interaction process between trees and bark beetle activity
(Jönsson et al., 2009; Simard et al., 2012). Trees at a higher elevation are less
susceptible to attack by bark beetles because elevation is a proxy for
temperature. Thus the higher the elevation, the lower the temperature and
that will limit the bark beetle activities. On the other hand, trees on a steep
slope suffer more from water stress and are therefore more susceptible to
attacks by the bark beetles (Chinellato et al., 2014). Furthermore, factors such
as stand characteristics of the forest affect the process of bark beetle attacks.
For example, areas of high stand density, trees with large diameters are more
susceptible to the attack by bark beetles (Cudmore et al., 2010; Safranyik,
1974; Skakun et al., 2003). In addition, forests with mixed tree species are
less susceptible to attack by the bark beetles compared to pure spruce stands
(Faccoli and Bernardinelli, 2014).

1.2 Host tree response and control strategies

Norway Spruce (Picea abies) trees possess defence mechanisms on several
levels that can prevent a bark beetle attack. The first level of defence (the so-
called primary defence) is activated when the beetle attempts to tunnel under
the bark of the trees, and the attacked tree attempts to release stored resin.
When the first resistance is exhausted by the beetle, the second stage of the
tree’s defence will start. During this stage, the local metabolism around the
entrance hole changes. As such the trees produce procyanidins as a defensive
chemical to prevent the entrance of bark beetle and to affect the establishment
of a brood by impairing the food quality. (Rohde et al., 1996; Wermelinger,

3
Introduction

2004). At the third defence level, a systemic change in the whole tree’s
metabolism occurs. This leads to the production of fewer carbohydrates but
more proteins, which are needed for constitutional resistance. Thus, the
successful colonisation of the bark beetle within the tree is considered to occur
in three successive steps. Firstly, the tree’s defences are exhausted by pioneer
beetles that chemically emit aggregation pheromones to attract a large number
of conspecifics for a mass attack (Wood, 1982). Secondly, mutualistic
relationships with certain fungi play an active part in killing the trees. Thirdly,
the ability to tolerate the defensive secretions of the attacked trees allows for
the final colonisation of the tree (Lieutier, 2002).

The infested tree exhibits three different stages of damage when it is a host of
the bark beetle (Coulson et al., 1985; Raffa et al., 2008; Wermelinger, 2004)
(Fig. 1.1). These stages are commonly known as green, red, and grey attacks.
In the first stage (the green attack), the beetle starts to lay its eggs under the
bark of the tree, and the infested trees have yet to show distinct symptoms
observable by the human eye (Niemann and Visintini, 2005a; Wermelinger,
2004). However, the subsequent degradation of the needles can be noted by
regular field observations. In addition, one of the earliest signs of bark beetle
infestation in trees is a drop in sapwood moisture, as a result of the inoculation
of blue fungi by the beetles that bore through the bark. The spores of the fungi
penetrate the living cells in the phloem and xylem, resulting in a disruption of
water and nutrient flow. A decrease in the transpiration cooling process of the
infested tree results, causing an increase in leaf and canopy temperature
(Christiansen and Bakke, 1988a; Sprintsin, 2011). In the red attack stage, the
needles of the tree turn from green to yellow to red-brown. Finally, the needles
fall off, and only the grey bark remains, hence the last stage is referred to as
a grey attack. This discolouration of the attacked trees is evident at the canopy
level (Coulson et al., 1985).

It is important to note that the length (specification) of these stages varies

depending on the type of beetles. For example, visible symptoms of an
attacked Norway spruce tree by the European Spruce Bark Beetle (Ips
typographus, L.) were evident in the crown in late August (red attack).
However, for North American and Canadian Mountain Pine Beetles
(Dendroctonus ponderosae, Hopkins) visible symptoms of an attack were
evident in the crown from late May to early June following the initial attack
(Safranyik and Carroll, 2006; Sprintsin, 2011).

Bark beetle outbreaks are an important factor in the ecological development of

the forest landscape affecting forest structure and composition (Pfeifer et al.,
2011; Schowalter, 2012; Simard et al., 2012; Zeppenfeld et al., 2015), as well
as biodiversity and ecosystem services (Thom and Seidl, 2016). Bark beetles
can affect forest ecosystems both directly and indirectly. Direct impacts include

4
 Chapter 1

an increase in tree mortality rates and a reduction in forest stand densities

(Bright et al., 2013; Eitel et al., 2011; Filchev, 2012a; Schowalter, 2012;
Vanderhoof et al., 2013; Verbesselt et al., 2009). Indirect impacts include a
reduction in carbon uptake, changes in tree species’ distribution, changes in
erosion processes, increasing fire frequency, and nutrient cycling (Beudert et
al., 2015; Kurz et al., 2008; Lehnert et al., 2013). Recent studies have shown
that bark beetle infestation leads to the enhancement of biodiversity by
opening the canopy layers and altering the microclimate conditions in the
forest. This alteration provides the essential habitats and sources of energy for
various organisms, allowing for species to persist in areas disturbed by the
bark beetle (Beudert et al., 2015; Lehnert et al., 2013; Müller et al., 2008).

Figure 1.1: The generalised life cycle of the European spruce bark beetle and the
process of killing host trees.

When forest management aims to control or preclude a mass outbreak and

minimise economic loss, two control approaches are commonly practised:
indirect control through long-term forest management and direct control
management. Indirect control approaches are essential to forest health,
improving the vitality of trees and the stability of stands; therefore, decreasing
the possibility of bark beetle infestation and outbreak. The long-term
management practices of indirect control include maintaining soil quality,
encouraging mixed wood stands, and addressing thinning and harvesting time
considering the life cycle of the bark beetle (Wermelinger, 2004). Previous
studies have shown that thinning or lowering tree densities reduced tree
mortality rate caused by bark beetle infestation (Egan et al., 2010; Negrón et
al., 2017). Direct control can take many forms, such as salvage and biological
control (trapping). Salvage includes the harvesting of storm-damaged timber
and freshly infested trees (green attack) (Göthlin et al., 2000; Stadelmann et
al., 2013). Harvesting of infested trees is the most common measure for

5
Introduction

controlling I. typographus L. However; this action task must be completed

before a new brood emerges and migrates (Wulder et al., 2009). Therefore,
the removal of fallen and freshly infested trees reduces the emergence of new
infestation locations, and it also moderates the beetle population. Biological
control strategies, such as pheromone trapping, can be used to lure beetles.
The main goal of this approach is to reduce bark beetle infestation in the
remaining stands. Thus, the traps are more often used to prevent attacks on
the remaining healthy trees than to diminish beetle populations (Lobinger and
Skatulla, 1996).

In general, the logistics of the direct (salvage) control of bark beetles require
a mechanism for the early detection of infested trees and monitoring known
infestation locations in the forest (Wermelinger, 2004). Traditionally, foresters
had to look for early signs of infestation by searching for a dry brown powder
that was produced by the bark beetles during the colonisation process when
the beetles tunnel under the bark of the trees. Such a technique, however, is
not practical and is inefficient for application in large areas because it is
significantly laborious and costly.

1.3 Remote sensing of Ips typographus, L. green

attack
Remote sensing has the potential to monitor and detect forest stress induced
by pest infestations over large areas in relatively short periods of time. The
premise of employing remotely sensed data to detect insect-infested forests is
that distinct symptoms are observable by different portions of the
electromagnetic spectrum, which can be acquired from remote sensing
platforms (Chen and Meentemeyer, 2016). Depending on the stage of the
attack, the symptoms indicate the decline in biochemical and biophysical
properties, such as chlorophyll and water content, defoliation, treefall gaps,
and changes in leaf area index (LAI) (Wulder et al., 2009). For example, during
the advanced stages of bark beetle infestation (i.e. red attack), needles on
Norway Spruce trees turn from green to yellow to red-brown, which can be
detected by the human eye at both leaf and canopy levels (Coops et al., 2006;
Franklin et al., 2003b; Niemann and Visintini, 2005a). As a result, these
symptoms can be observed using the visible portion of the electromagnetic
spectrum (400 to 700 nm), which appear similar in the human visual system.
It is important to note that the stress exhibited by the trees mediated by the
bark beetle is brought on gradually, and therefore, it is difficult to detect early-
stage symptoms using the visible portion of the electromagnetic spectrum.
Instead, sensors with the capacity to detect the near-infrared spectrum (700
to 1300 nm) are more sensitive to such physiological stress. Likewise, the
amount of energy reflected in the short-wave infrared range (1300 to 2500
nm) was correlated with vegetation moisture; in particular, leaf water content

6
 Chapter 1

(Aldakheel and Danson, 1997; Laurent et al., 2005). In general, the length of
time required for a moisture deficit to produce visible symptoms in foliage can
vary from weeks to several months. However, these changes are detectable
with spectral data, particularly if acquired in the shortwave and thermal
infrared spectra, where water absorption features in vegetation are well
documented (Berni et al., 2009; Buitrago Acevedo et al., 2017; Jang et al.,
2006; Sepulcre-Cantó et al., 2006).

In general, the usefulness of remotely sensed data to provide forest health-

relevant information (stress signal) depends on the spectral, spatial, temporal
and radiometric resolution of remote sensing instrument (Chen and
Meentemeyer, 2016; Wulder et al., 2009). Recent studies have discovered that
the subtle changes in spectral discrepancies induced by insect infestation can
be better detected by fine-spectral and spatial resolution data (Coops et al.,
2006; Hatala et al., 2010; Ortiz et al., 2013). For example, previous studies
confirmed the potential of high spatial resolution data for fine-scale forest
stress detection (Filchev, 2012a; Immitzer and Atzberger, 2014; Latifi et al.,
2014). Moreover, many forest diseases or insect infestations do not cause tree
death instantly; therefore, continuous monitoring is essential for investigating
and understanding the dynamic characteristics of leaf properties and canopy
reflectance during infestation events. Such observation provides detailed
information regarding the impact on infested trees, and it also allows for the
exploration of diagnostic effects at any stage to detect insect infestation as
early as possible.

Many studies have attempted to use remote sensing data and techniques for
the detection of bark beetle infestation at an early (green attack) stage. As
Table 1 shows, most studies performed on the detection of green attacks were
conducted at the canopy level. The literature review revealed that no previous
studies focused on both the leaf level and the canopy level. Furthermore, the
effect of an Ips typographus green attack on host tree biochemical and spectral
reflectance properties has not yet been investigated in detail. To date, the
successful detection of bark beetle green attack using remote sensing data has
not been documented in the literature.

The little attention that has been paid to the changes in biochemical and
biophysical properties of Norway Spruce trees under bark beetle Ips
typographus infestation and the link with variation in spectral reflectance
behaviour may be due to several issues. First, there are biological limitation
factors, such as the flight activity of the bark beetle, the time lapse for the
colonisation process, and the period during which trees are yet to show visual
signs of stress. As mentioned in section 1.1, the development of the bark beetle
is dependent on temperature, as an increase in temperature may lead to an
increase in the beetle population, due to eggs hatching and developing into

7
Introduction

adults sooner (Bentz et al., 2010; Wermelinger, 2004). Therefore, the fading
of the infested tree is not a consistent or linear process, but it instead depends
on the local environment and tree conditions (Safranyik, 1974). Second, there
are logistical and technological limitation factors. In remotely sensed and field
survey data of bark beetle green attacks, it is evident that time plays a critical
role (Wulder et al., 2009). For example, during a field survey, the temporal
observation associated with beetle biology and the appearance of the
symptoms in the tree foliage should be considered. Likewise, the flight time of
the beetle must be considered in the case of remotely sensed surveys. For
example, the adult bark beetle starts to attack new host trees during the
European springtime (between the end of April and the beginning of May).

This thesis is concerned with the detection of the European Spruce bark beetle
green attack at both leaf and canopy levels using various types of remotely
sensed data in parallel with field measurements. It also considers the impact
of bark beetle green attack on the variation of biochemical and biophysical
properties of the infested leaves. The relationship between bark beetle
infestation and changes in foliar biochemical properties, such as chlorophyll
content, chlorophyll fluorescents, nitrogen content, stomatal conductance,
water content, and dry matter content, are examined and interpreted.

8
 Chapter 1

Table 1.1: Existing literature that utilised various remote sensing data (including aerial
photography) to detect bark beetles at the green attack stage over the last 50 years.
Beetle
No Author Data used level
species
1 (Murtha, 1972b) aerial photography MPB canopy

2 (Murtha and Wiart,

aerial photography MPB canopy
1989)
aerial photography + field leaf and
3 (Runesson, 1991) MPB
spectrometer canopy
(Gimbarzevsky et al., multi-spectral scanner
4 MPB canopy
1992a) imagery + aerial photography
5 (Schweigler,2007) aerial photography NSB canopy
6 (Ortiz et al., 2013) TerraSAR-X and RapidEye NSB canopy
field spectrometer
7 (Ahern, 1988b) MPB leaf
(350 -1100nm)
field spectrometer
8 (Cheng et al., 2010) MPB leaf
(350-2500nm)
9 (Sprintsin, 2011) Landsat ETM+ MPB canopy
Compact Airborne
10 (Heath, 2001a) NSB canopy
Spectrographic imager (CASI)
(Lawrence and Labus,
11 hyperspectral imagery DFB canopy
2003)
(Marx, 2010; Marx
12 and an der Havel, RapidEye NSB canopy
2010a)
13 (Filchev, 2012a) worldview-2 NSB canopy
14 (Lausch et al., 2013b) HyMAP NSB canopy
15 (Latifi et al., 2014) Landsat MSS + SPOT2&4 NSB canopy
(Fassnacht et al.,
16 HyMAP NSB canopy
2014)
(Immitzer and
17 WorldView-2 NSB canopy
Atzberger, 2014)
18 (Näsi et al., 2018) Hypex UAV NSB canopy
field spectrometer + Landsat
19 (Foster et al., 2017) ESB canopy
TM
VNIR-SWIR and discrete
20 (Niemann et al., 2015) MPB canopy
LiDAR
21 (Mullen, 2016) Hypex & WorldView-2 MPB canopy
22 (Roberts et al., 2005) aerial photography MPB canopy
ESB = Engelmann spruce beetle (Dendroctonus rufipennis), MPB = mountain pine
beetle (Dendroctonus ponderosae), NSB = Norway spruce beetle (Ips typographus)

9
Introduction

1.4 Research objectives

The main objective of this research is to utilise different remote sensing data
and approaches at different scales (leaf and canopy) to detect early stage of
bark beetle infestation. This includes four sub-objectives:

First objective: To investigate the impact of an Ips typographus L. green

attack on the foliar properties (i.e. foliar reflectance and biochemical
properties) of Norway Spruce trees. In this objective, the following research
questions are addressed
 How does the early stage bark beetle infestation affect foliar spectral and
biochemical properties?
 Does bark beetle infestation affect the accuracy of foliar biochemical
estimations when using hyperspectral measurements?

Second objective: To study the dynamics of leaf traits and canopy reflectance
spectra of Norway Spruce trees from early to advanced stages (green to red)
of spruce bark beetle Ips typographus, L. infestation. In this objective, the
following research question is addressed
 What are the spectral regions and leaf traits that are affected by infestation
over time and how would this impact the discrimination between healthy
and infested plots in the early stages of an attack?

Third objective: To examine the sensitivity of thermal and visible-short wave

infrared data to foliar properties under bark beetle Ips typographus, L.
infestation. In this objective the following research question is addressed:
 How accurate are thermal and visible-short wave infrared data from
Landsat-8 at detecting an early stage Ips typographus, L. green attack and
estimating foliar properties?

Fourth objective: To evaluate the potential of different spectral vegetation

indices from multispectral satellites to detect and map bark beetle infestation
at the green-attack stage. In this objective the following research question is
addressed:
 To what extent can spectral vegetation indices from multispectral satellite
imagery be utilised to map and detect canopy stress induced by a bark
beetle green attack?

1.5 Thesis outline

This PhD thesis consists of six chapters, four of which are standalone papers.
Three of them have been already published in peer-review ISI journals, and
the other one are under review. The structure of the chapters is as follows:

10
 Chapter 1

Chapter 2 is about the impact of the early stage of bark beetle infestation on
foliar properties (chlorophyll and nitrogen concentration) and reflectance
spectra.

Chapter 3 focuses on the temporal dynamics of foliar properties and canopy

reflectance using temporally dense time series of six RapidEye senses and five
SPOT-5 senses in parallel with field data collection.

Chapter 4 examined the sensitivity of optical and thermal infrared data to

foliar properties (chlorophyll fluorescence, stomatal conductance and water
content) at early stage bark beetle (Ips typographus, L.) infestation.

Chapter 5 deals with using a set of spectral vegetation indices calculated from
Sentinal-2 and Landsat-8 images to map bark beetle infestation at the green
attack stage. Moreover, in this chapter, foliar properties such as reflectance
spectra, chlorophyll content, water content, leaf dry matter content and
nitrogen content were studied.

Chapter 6 provides a synthesis of the results of this research, including the

answers to the research questions.

11
Introduction

12
Chapter 2 : European spruce bark beetle (Ips
typographus, L.) green attack affects foliar
reflectance and biochemical properties


This chapter is based on:

Abdullah, H., Darvishzadeh, R., Skidmore, A.K., Groen, T.A. and Heurich, M., 2018.
European spruce bark beetle (Ips typographus, L.) green attack affects foliar reflectance
and biochemical properties. International Journal of Applied Earth Observation and
Geoinformation, 64: 199-209.

Abdullah, H., Darvishzadeh, R., Skidmore, A.K., Groen, T.A. and Heurich, M. Detecting
the effect of (Ips typographus, L.) green attack on leaf reflectance. The Netherlands
Annual Ecology Meeting, 14-15 February 2017, Lunteren, Netherlands.

Abdullah, H., Darvishzadeh, R., Skidmore, A.K., Groen, T.A. and Heurich, M. Effects of
(Ips typographus, L.) green attack on leaf properties. 10th EARSel SIG Imaging
Spectroscopy Workshop, 19-21 April 2017, Zurich, Switzerland.

13
Effects (Ipstypographus, L.) green attack on leaf properties

Abstract
The European spruce bark beetle Ips typographus, L. (hereafter bark beetle),
causes major economic loss to the forest industry in Europe, especially in
Norway Spruce (Picea abies). To minimise economic loss and preclude a mass
outbreak, early detection of bark beetle infestation (so-called “green attack”
stage – a period at which trees are yet to show visual signs of infestation
stress) is, therefore, a crucial step in the management of Norway spruce
stands. It is expected that a bark beetle infestation at the green attack stage
affects a tree’s physiological and chemical status. However, the concurrent
effect on key foliar biochemical such as foliar nitrogen and chlorophyll as well
as spectral responses are not well documented in the literature. Therefore, in
this study, the early detection of bark beetle green attacks is investigated by
examining foliar biochemical and spectral properties (400 – 2000 nm). We also
assessed whether bark beetle infestation affects the estimation accuracy of
foliar biochemicals. An extensive field survey was conducted in the Bavarian
Forest National Park (BFNP), Germany, in the early summer of 2015 to collect
leaf samples from 120 healthy and green attacked trees. The spectra of the
leaf samples were measured using an ASD FieldSpec3 equipped with an
integrating sphere. Significant differences (p < 0.05) between healthy and
infested needle samples were found in the mean reflectance spectra, with the
most pronounced differences being observed in the NIR and SWIR regions
between 730 and 1370 nm. Furthermore, significant differences (p < 0.05)
were found in the biochemical compositions (chlorophyll and nitrogen
concentration) of healthy versus green attacked samples. Our results further
demonstrate that the estimation accuracy of foliar chlorophyll and nitrogen
concentrations, utilising partial least square regression model, was lower for
the infested compared to the healthy trees. We show that early stage of
infestation reduces not only foliar biochemical content but also their retrieval
accuracy. Our results further indicate that remote sensing measurements can
be successfully used for the early detection of the bark beetle infestation. We
demonstrated that bark beetle infestation at the green attack stage effects leaf
spectral response as well as leaf biochemical properties and their retrievals
from hyperspectral measurements.

14
 Chapter 2

2.1 Introduction
Bark beetles (Ips typographus, L., and Dendroctonus spp.) are important biotic
disturbance agents in the coniferous forests of Europe and North America,
respectively (Christiansen and Bakke, 1988b; Fahse and Heurich, 2011; Raffa
et al., 2008; Seidl et al., 2011; Seidl et al., 2014). In the past decades, an
increasing number of severe bark beetle outbreaks has led to an extensive
economic loss in the forest industry (Goheen and Hansen, 1993; Waring et al.,
2009). The economic impacts include a reduction in the commercial value of
the infested trees and increased management costs (Schowalter, 2012).
Besides causing economic losses, outbreaks of the bark beetle form an
important factor in the ecological development of the forest landscape, in terms
of forest structure and composition (Pfeifer et al., 2011; Schowalter, 2012;
Simard et al., 2012; Zeppenfeld et al., 2015), as well as biodiversity and
ecosystem services (Thom and Seidl, 2015). Bark beetles can affect forest
ecosystems both directly and indirectly. Direct impacts include an increase in
tree mortality rates and a reduction in forest stand densities (Bright et al.,
2013; Eitel et al., 2011; Filchev, 2012b; Hais and Kučera, 2008; Schowalter,
2012; Vanderhoof et al., 2013; Verbesselt et al., 2009). Other indirect impacts
are; reduction in carbon uptake, changes in tree species’ distribution, as well
as changes in erosion processes, fire frequency, and nutrient cycling (Beudert
et al., 2015; Kurz et al., 2008; Lehnert et al., 2013; Lindenmayer and Franklin,
2002; Mikkelson et al., 2013). Additionally, recent studies showed that the
infestation of bark beetle leads to biodiversity enhancement by opening the
canopy layers and altering microclimate condition in the forest. This alteration
will provide the essential habitats and sources of energy for various organisms,
and allow them to persist in the disturbed areas by bark beetle (Beudert et al.,
2015; Lehnert et al., 2013; Müller et al., 2008).

The phenology of bark beetle and the associated host responses are well
described in Wermelinger, (2004). The infested tree goes through three stages
of attack (Coulson et al., 1985; Sprintsin et al., 2011; Wermelinger, 2004).
These stages are termed green, red and grey attacks, respectively. During the
green attack stage, the foliage remains green (hence the name green attack),
and therefore, it is difficult to detect this stage by the human eye at leaf and
canopy levels (Niemann and Visintini, 2005b; Wulder et al., 2006b). However,
the subsequent degradation of the needles can be noted by regular field
observations, as, during the red attack stage, the needles of the tree turn from
green to yellow and red-brown. Finally, the needles fall off, and only the grey
bark will remain, hence the last stage being called a grey attack. This
discolouration of the attacked trees is evident at canopy level (Coulson et al.,
1985). It is expected that bark beetle infestation induces changes in the
spectral response of the infested trees (Filchev, 2012b; Meddens et al., 2013),
as the biophysical characteristics of the entire tree, and very likely the

15
Effects (Ipstypographus, L.) green attack on leaf properties

biochemical features of the needles, change. Therefore, during the infestation

period, the trees are subjected to increasing stress and face physiological
change (Heath, 2001b). This is due to the interruption of the water flow and
the deterioration of chloroplasts as the beetle drills into the tree’s cambium
tissue (Yamaoka et al., 1990). The fungi carried by the beetles penetrate the
living phloem and xylem cells, hampering the translocation of water, sugar and
other nutrients within the bole of the tree (Paine et al., 1997a; Rohde et al.,
1996; Safranyik et al., 2007; Wermelinger, 2004). This leads to a gradual
change in biochemical and water content in the attacked tree, thus inducing
alterations to its spectral characteristics over the course of the infestation
(Deshayes et al., 2006; Lawrence and Labus, 2003; Marx and an der Havel,
2010b; Reid, 1961; Sprintsin et al., 2011; Yamaoka et al., 1990).

A considerable increase in unplanned harvesting of European forests following

bark beetle infestation has led to increased research interest in understanding
the dynamics and improving the management of Ips typographus, L.
outbreaks (Seidl et al., 2011). Furthermore, an increase in the frequency and
severity of bark beetle outbreaks is expected due to global climate change
(Bentz et al., 2010). Consequently, more attention is devoted to this topic. It
has also been evoked to increase discussion regarding salvage logging in
protected areas in Europe (Lehnert et al., 2013). Early detection of Ips
typographus, L. outbreaks at the green attack stage may prove an important
step, as management aims to control this species and preclude a mass
outbreak. At the green attack stage, the trees hold the next generation of
beetles. Management intervention to prevent further outbreaks may,
therefore, involve the removal of infested trees before the new brood emerges
and migrates (Wermelinger, 2004; Wulder et al., 2009). Traditionally, foresters
have performed field surveys to identify infested trees; such surveys are very
laborious, costly, and therefore it is inefficient and hard to apply for
management purposes in large areas. Remote sensing has the potential to
detect pest infestations over large areas in relatively short periods of time.
Employing remotely sensed data allows monitoring of the changes in leaf and
canopy properties before and after insect infestation (Bentz and Endreson,
2003; Deshayes et al., 2006; Dye et al., 2008; Jensen, 2009). To date, the
utilisation of remote sensing for the monitoring and detection of bark beetles
by forest managers has mainly focused on the last two attack stages (i.e., the
red and grey stage) and has achieved high degrees of accuracy. During the
last two stages of the attack, the changes in canopy colour effect the spectral
reflectance signature which are mainly been used as an indicator to detect
infestations (Carter et al., 1998; Franklin et al., 2003a; Heurich et al., 2010b;
Latifi et al., 2014; Meddens, 2012; Nikolov et al., 2014; Skakun et al., 2003;
Wulder et al., 2006b). However, detecting the infestation in the last two stages
is not sufficient for appropriate management, as phenological research proved
that during the red attack stage the newly developed beetles have already left

16
 Chapter 2

their host trees and started to attack new trees. Therefore, the continuing of
an outbreak cannot be prevented by salvage logging during this stage.
Consequently, the detection of the bark beetle at the green attack stage by
means of remote sensing is necessary to have a meaningful effect on the
spreading of the beetle, but is challenging due to the lack of apparent visual
symptoms in needles.

Early detection of infestations by Dendroctonus spp in lodgepole pine trees has

been investigated at canopy level (Gimbarzevsky et al., 1992b; Heath, 2001b;
Klein, 1973; Murtha and Wiart, 1989; Murtha, 1972a). Similarly, detection of
a bark beetle green attack (Ips typographus, L.) at canopy level in Norway
spruce trees has been investigated by (Immitzer and Atzberger, 2014; Lausch
et al., 2013a; Marx and an der Havel, 2010b; Ortiz et al., 2013). However,
these studies did not succeed in discriminating healthy from green attacked
trees. More recently, Niemann et al. (2015) used LIDAR and hyperspectral data
to examine the spectral properties of healthy trees and those under mountain
pine beetle green attack and demonstrated that the most promising
wavelengths, for the detection of mountain pine beetle at green attack stage,
is located in the shortwave infrared region. Furthermore, Näsi et al. (2015)
used a hyperspectral sensor (500-900 nm) mounted on an unmanned aerial
vehicle (UAV) to map bark beetle damage at the tree level, by dividing tree
stands into three different classes (healthy including trees with potential early
infestation stage, red attack; and dead). They found that the healthy and dead
trees can be classified with 90% over-all accuracy, however, when all classes
were considered (healthy, red and dead), the overall accuracy dropped to 76%.
At the leaf level, there are very few studies, although with different beetle and
tree species, that have examined the differences in spectral reflectance
between healthy needles and those that have been under bark beetle green
attack (Ahern, 1988a; Cheng et al., 2010; Foster et al., 2017). In addition to
a change in spectral reflectance properties, Cheng et al. (2010) observed that
there were differences in the water content of healthy and infested needles,
particularly, between 1318 – 1322 nm.

These results show that remote sensing has the potential to detect early stages
of bark beetle attacks. However, these studies have mostly considered the
effect of early infestation on reflectance spectra and not on the biochemical
properties of the needles. Therefore, it is highly important to understand
whether, in addition to the spectral reflectance, the biochemical properties are
also affected at the bark beetle early infestation stage. Consequently, the topic
warrants further investigation.

As mentioned earlier, it is expected that the infested tree will exhibit a change
in terms of its biochemical and spectral properties, due to the beetle larva and
blue stain fungi such as (Ophiostoma and Cerato-cystis species) carried by the

17
Effects (Ipstypographus, L.) green attack on leaf properties

beetles starts to penetrate the living phloem and xylem cells, hampering the
translocation of water, sugar and other nutrients within the bole of the tree
(Paine et al., 1997a; Rohde et al., 1996; Safranyik et al., 2007; Wermelinger,
2004).Therefore, in this study, we aim to investigate the possible early
detection of a bark beetle green attack by examining and comparing the foliar
biochemical (chlorophyll and nitrogen) and spectral properties (400 – 2000
nm) of both healthy and green attacked trees. Chlorophyll and nitrogen are
two important elements that have a key role to play in plant life and status and
can be considered as indicators that reflect the status of plant growth and
health (Heinze and Fiedler, 1976; van Maarschalkerweerd and Husted, 2015;
Wang et al., 2015a; Wang et al., 2015b). As such changes, initiated by bark
beetle attack cause stress during the infestation time, we hypothesise that
chlorophyll and nitrogen concentrations are reduced during such an attack.
These can thus provide suitable proxies for detecting the presence of Ips
typographus L. during a green attack stage. Specifically, there are three main
objectives: (a) to investigate the impact of an Ips typographus L. green attack
on folair spectral reflectance; (b) to examine the changes in folair biochemical
properties due to the Ips typographus L. green attack ; (c) to explore the
impact of the Ips typographus L. green attack on the estimation accuracy of
folair biochemical properties utilizing hyperspectral measurements.

2.2 Material and Methods

2.2.1 Study area
The study area is the Bavarian Forest National Park (BFNP) in south-east
Germany, which covers an area of 240 km2 between 13°12'9" E (longitude)
and 49°3'19" N (latitude) (Fig. 2.1). The elevation in the BFNP ranges from
600 m to 1453 m. This region is characterised as having a temperate climate
with a total annual precipitation along the gradient between 900 and 1800mm
as well as a mean annual temperature that varies between 3.5 and 7.2°C
(Bässler et al., 2008b; Lehnert et al., 2013). The area is divided into three
ecological zones: high elevations, hillsides, and valley bottoms. Around 90%
of the tree stands in high elevations are Norway spruce (Picea abies (L.) Karst),
2% are beech (Fagus sylvatica L.), and the remaining 8% are covered by other
broad-leaves trees, mainly Common Rowan (Sorbus aucuparia). While on the
hillsides, around 58% is occupied by Norway spruce, and the rest is a mixture
of European silver fir (Abies alba) and beech. In the valley bottoms, 83% of
the trees are Norway spruce, and the rest is a mix of species (Cailleret et al.,
2014b; Heurich et al., 2010a). Multiple storm events in the 1980s and a series
of hot summers in the 1990s have led to an extensive mortality of canopy trees
by bark beetles on about 8 000 ha (Lausch et al., 2013a).

18
 Chapter 2

Figure 2.1: The location of Bavarian Forest National park in Central Europe.

2.2.2 Data acquisition

The adult bark beetle starts to attack new host trees during the European
springtime (between the end of April and the beginning of May). This process
strongly depends on the air temperature as the beetles start their swarming
when the air temperature reaches 16.5°C (Lobinger, 1994a; Wermelinger,
2004). Therefore, the field data were collected in the early summer of 2015
within 15 to 30 days from the early stages of infestation. In total 120 trees
were measured, of which 66 were healthy, and 54 were freshly infested trees.
An extensive field survey was conducted to identify the trees under bark beetle
green attack by spotting piles of dry, boring dust, which had been pushed out
onto the bark surface (Fig. 2.2). An Average of 2 – 3 branches were taken from
each tree. Needle samples from each branch were collected separately. All the
branches were taken from the upper part of the trees, which was exposed to
the sunlight. The heights of the trees varied between approximately 25 and 30
meters. A crossbow was used to shoot an arrow with a fishing line attached to
a branch with sunlit leaves (Ali et al., 2016). The fishing line was used to feed
a rope over the targeted branch. Once the rope was in place, the branch was
pulled down gently until it broke off. Needles were immediately removed from
the fallen branches and placed in a labelled plastic zip-locked bag, which was
then covered with wet pulp paper and subsequently transported to the
laboratory in a portable cooling box filled with frozen ice packs to keep the
sample cool. The aim was to retard, as much as possible, any changes in the
needles’ reflectance spectra and biochemical characteristics (Malenovský et al.,
2006). Once the spectral measurements were completed, the same sample
branch was transported to the laboratory to measure the fresh weight, leaf
area and the biochemical properties of the foliar samples.

19
Effects (Ipstypographus, L.) green attack on leaf properties

Figure 2.2: Dry dust produced through boring by the bark beetle Ips typographus, L.;
used to identify freshly infested trees in the field.

2.2.3 Chemical analyses

The concentration of Chlorophyll a and b in the collected fresh leaf samples
was determined destructively by a spectrophotometer following the
Lichtenthaler (1987) method using acetone (v 100%). In addition, to
determine foliar nitrogen, the needles were dried for 72 hours using an oven
dryer at 60º C. The dried needles were properly grounded using mortar and
pestle until the ground leaves became a soft powder and were passed through
a 0.25 mm mesh screen. Subsequently, 2 mg of powdered leaves was
transferred to a small aluminium capsule to measure the nitrogen content,
using an organic elemental analyser (FLASH 2000).

2.2.4 Reflectance measurements

The leaf directional hemispherical reflectance from 350 to 2500 nm was
measured for collected samples, using an ASD FieldSepc-3 Pro FR spectrometer
equipped with an ASD RT3-3ZC integrating sphere (Analytical Spectral
Devices, Inc., Boulder, Co, USA). The integrating sphere had a port diameter
of 15 mm for a reflectance measurement. The Norway spruce needles did,
however, not cover the entire sample port of the integrating sphere. Therefore,
the method proposed by Daughtry et al. (1989) was applied, where a universal
sample holder, designed by Malenovský et al. (2006), was used to
accommodate all sizes and lengths of Norway spruce needles. For every
measurement five to six needles were secured on the sample holder with
cellophane tape, leaving a gap of one needle’s width between needles to avoid
multiple reflectances of the adjusted needles (Ali et al., 2016; Daughtry et al.,

20
 Chapter 2

1989). To minimise the effect of spectral signal noise, for every needle sample
two hundred scans were averaged to present a single spectrum. The
integrating sphere has an internal light source, and therefore, the sun angle
and cloudiness does not affect the spectral measurement (Foster et al., 2017).
However, to prevent possible light entering into the integrating sphere ports,
covering the integrating sphere with a black piece of cloth is recommended by
Mesarch et al. (1999).

The measured spectral reflectance from the sample holder contained the
effects of the gap between the attached needles to the sample holder;
therefore, the spectral measurements were corrected by calculation of the gap
fraction. To do so, a black painted paper mask with a 15-mm diameter circular
aperture was precisely superimposed onto the samples and photographs were
taken using a 16.1 megapixel Panasonic LUMIX camera (Ali et al., 2016).
Then, the images were imported into ImageJ software (Schneider et al., 2012),
to compute the gap fraction, based on the illuminated area of the sample port.
The following equation was adapted from Mesarch et al. (1999) for the Field
ASD spectrometer and to compute the hemispherical reflectance of the sample
needles:

Reflectance = [((𝜌 – Rd) / (1 – Rd)) × Rr] / (1 – GF); (2.1)

where: 𝜌 is the measured reflectance from the sample holder; Rd is stray light
(ambient light) inside the integrating sphere, measured as a radiation flux of
the empty illuminated sample port in reflectance mode; Rr, is the reference of
sample reflectance; and GF is the gap fraction of the sample.

2.2.5 Data pre-processing and statistical analysis

A series of pre-processing steps were performed on the measured reflectance
data. First, the spectral reflectance was examined, and “noisy” bands (in the
spectral ranges 350 – 399 nm and 2000 – 2500 nm) were excluded from the
analysis. Secondly, to eliminate and reduce the sensor noise, a Savitzky –
Golay smoothing filter with a frame size of 15 (2nd-degree polynomial) was
applied to the reflectance spectra (Savitzky and Golay, 1964).

The significance of differences (p ≤ 0.05) in foliar reflectance as well as

chlorophyll and nitrogen concentration between leaf samples from green
attacked and healthy trees, was examined using Student t-tests, in order to
determine whether there is spectral variation at any spectral band. Because
the considered adjacent wavelengths are highly correlated, we corrected the
p-value using Holm (1979) procedure, next the significant wavebands with
mean reflectance spectra of both healthy and infested samples were plotted
following the technique used by Schmidt and Skidmore (2003).

21
Effects (Ipstypographus, L.) green attack on leaf properties

To investigate the relationship between the spectral reflectance of the healthy

and infested samples and their biochemical parameters, the Pearson’s
correlation coefficients were calculated between spectral reflectance (400 -
2000 nm) and chlorophyll as well as nitrogen concentration for both healthy
and infested samples to identify the most sensitive wavebands in healthy and
infested samples.

2.2.5.1 Partial least square regression model (PLSR)

Partial least squares regression (PLSR) was used to investigate the impact of
the bark beetle green attack stage on the retrieval accuracy of the chlorophyll
and nitrogen concentrations in needles. PLSR is a regression method that takes
into account both the variance of the explanatory and the dependent variables.
It specifies a linear relationship between a set of dependent (Y) variables and
a set of predictor (X) variables (Wold et al., 2001). PLSR performs particularly
well when the various X-variables have strong correlations, as is normally the
case with hyperspectral data. (Carrascal et al., 2009; Nicolaï et al., 2007).
Further details on PLSR can be found in Geladi and Kowalski (1986).

PLSR was used to predict the foliar chlorophyll and nitrogen concentrations
(dependent variables) of the healthy and infested needles from spectral
reflectance (predictor variables). PLSR models were built independently for the
healthy and infested samples. For the chlorophyll concentration, PLSR was
fitted to the VNIR range (400 - 790 nm), as this spectral region mainly
contributes to chlorophyll estimation in plants (Curran, 1989; Yoder and
Pettigrew-Crosby, 1995). However, as nitrogen is associated with many other
leaf chemical compounds, the spectral wavebands corresponding to nitrogen
are distributed over the whole spectrum (VIR, NIR, SWIR) (Curran, 1989;
Peterson et al., 1988) and consequently, for the nitrogen concentration, PLSR
was fitted to the entire spectrum (400 - 2000nm). To determine the number
of components needed to build the PLSR model, Monte Carlo cross-validation
(MCCV) was applied (Xu and Liang, 2001). The procedure involves adding an
extra component to the model when the value of the root mean square error
of cross-validation (RMSEcv) is reduced by ≥ 2 % (Darvishzadeh et al., 2008;
Geladi and Kowalski, 1986). The performance of the PLSR models were
assessed by computing the coefficient of determinations (R2), the normalised
root mean square error of predictions (nRMSEP) (Kvalheim,1987), and
normalised root mean square error of cross-validations (nRMSEcv) of predicted
versus measured values.

We further examined whether the most informative wavelengths for the

estimation of chlorophyll and nitrogen concentrations in both healthy and
infested samples correspond. To achieve this, the variable importance in the
projection (VIP) was calculated for all wavebands from the PLSR models. VIP
scores summarise the influence of individual X variables (reflectance) in the

22
 Chapter 2

PLSR model. A variable with a VIP score close to, or greater than 1, can be
considered significant in a given model (Chong and Jun, 2005; Tenenhaus,
1998; Wold et al., 1993). A detailed description of the calculation of VIP scores
can be found in Farrés et al. (2015); Wold et al. (1993). TOMCAT and libPLS
toolbox was used within MATLAB to establish PLSR analysis and calculate VIP
score, respectively (Daszykowski et al., 2007; Li et al., 2014).

2.3 Results
2.3.1 Spectral reflectance variation due to bark beetle green
attack
The results demonstrated that the mean reflectance spectra of healthy and
green attacked foliage were statistically different (p ≤ 0.05) for 917 wavebands
out of the 1600 wavebands used in this study (Fig. 2.3). As can be seen from
Figure 2.3, a clear distinction was observed in the mean reflectance between
healthy and infested foliage in the visible and near-infrared regions. In the
visible region, the mean reflectance of the infested foliage was higher than of
the healthy leaves, in conjunction with chlorophyll degradation (Fig. 2.4).
However, there was a larger difference (p ≤ 0.05) between infested and
healthy needles in the wavelength range 730 – 1370 nm, as the healthy
needles had higher reflectance compared to mean infested spectra.

2.3.2 Impact of bark beetle green attacks on foliar biochemical

properties
A Student t-test demonstrated a significant difference between total
chlorophyll and foliar nitrogen concentrations of healthy and infested leaves (p
< 0.05). In healthy foliage, the mean and ± of the total chlorophyll
concentration were 0.766 ± 0.140 mg/g, respectively, whereas for the green
attacked leaves they were 0.657 ± 1.62 mg/g, respectively. Furthermore, the
concentration of nitrogen was 1.25 ± 0.21 % within healthy leaves, while for
the infested leaves they were 1.13 ± 0.18 %, respectively (Fig. 2.4). The
correlation between total chlorophyll and nitrogen concentration was higher in
needles from healthy trees (r = 0.68, p < 0.05) than in needles from green
attacked trees (r = 0.57, p < 0.05) .

23
Effects (Ipstypographus, L.) green attack on leaf properties

Figure 2.3: Mean reflectance spectra of healthy and infested leaves at the green attack
stage. Gray areas depict the location of wavebands displaying is a significant difference
between healthy and infested spectra.

Figure 2. 4: Distribution of measured chlorophyll and nitrogen concentration for healthy

and infested needles. There is a significant difference (p < 0.05) in chlorophyll and foliar
nitrogen concentration between healthy and infested sample.

2.3.3 Effects of bark beetle green attack on the retrieval

accuracy of leaf biochemical properties.
The impact of bark beetle green attack on the retrieval accuracy of foliar
chlorophyll and nitrogen concentrations was assessed using PLSR. In general,
higher accuracies were observed when healthy samples were assessed (Table
2.1). For healthy foliage, the coefficients of determination for chlorophyll and
nitrogen concentrations were R2 = 0.64 and R2 = 0.76, respectively. However,

24
 Chapter 2

these coefficients decreased for models fitted to data from infested foliage,
both for chlorophyll (R2 = 0.55) and for nitrogen (R2 = 0.68) (Table 2.1).
Similar results were discerned with the Pearson correlation coefficient between
chlorophyll and the reflectance data, as well as between nitrogen and the
reflectance data. As can be observed from Figure 6, significantly higher
correlations were observed for healthy leaves than for infested leaves. In
addition, the prediction error for chlorophyll in healthy leaves (nRSMSEp =
0.20) was much lower than for chlorophyll in infested leaves (nRMSEp = 0.62)
(Fig. 2.5). However, this variation in nRMSEp for nitrogen content was much
lower (0.12 and 0.22) in both healthy and infested samples respectively.

Figure 2.5: Measured versus predicted foliar chlorophyll (a) and nitrogen (b)
concentration for healthy and infested samples derived from the PLSR analysis.

Table 2. 1: Number of latent variables, R2 and nRMSEp between measured and predicted
chlorophyll and nitrogen in healthy and green attacked needles.

Nr. Of
Sample
Variables latent nRMSEp nRMSECV R2
Condition
variables
Chl 9 0.20 0.24 0.64
Healthy
N 10 0.12 0.20 0.76
Chl 10 0.62 0.75 0.55
Infested
N 8 0.22 0.35 0.68

25
Effects (Ipstypographus, L.) green attack on leaf properties

Figure 2. 6: Correlation between chlorophyll and nitrogen and the reflectance of

individual wavebands. The bold lines represent the wavebands at which the reflectance
spectra correlated significantly with the chlorophyll and nitrogen concentrations. The
arrows show the wavebands in which recorded the highest correlation with chlorophyll
content in both healthy and infested samples.

As shown in Fig. 2.6 and 2.7, the spectral regions, in particular, the green and
the red edge (536 - 559nm and 732 - 790nm, respectively), demonstrated the
highest VIP (VIP > 1) and a moderately negative correlation (r = -0.41 and -
0.32, for the green and red edge, respectively; P < 0.05) between chlorophyll
and the reflectance spectra of healthy foliage. The high VIP value indicated
that the spectral feature was of major importance in estimating the chlorophyll
concentration. A weaker negative correlation with a lower VIP score was
observed for the infested samples (r = -0.12, P < 0.05).

The spectral regions that were most sensitive to changes in nitrogen content
were in the range of 980 - 1000 nm and 1448 - 1469 nm, where the infested
leaves had a lower VIP score (Fig. 2.7). Furthermore, the correlation
coefficients between foliar nitrogen concentration and the reflectance spectra
show a distinct variation between healthy and infested foliage, especially in the
spectral region between 720 - 1400 nm. The healthy foliage showed a
moderately negative correlation (r = -0.30, P < 0.05), while for the infested
foliage a positive correlation was found (Fig. 2.6).

26
 Chapter 2

Figure 2. 7: Importance of wavelengths corresponding to the highest value of variable

importance in the projection scores of partial least squares regression in healthy and
infested samples, used for chlorophyll and nitrogen estimation

2.4 Discussion
It has long been believed that the detection of bark beetle green attack using
remote sensing data is a challenging task (Wulder et al., 2009; Niemann and
Visintini, 2005). However, recent studies have shown that remote sensing data
obtained across a number of wavelengths have the potential for the detection
of bark beetle green attack (Cheng et al., 2010; Foster et al., 2017; Niemann
et al., 2015). In this study, we investigated whether bark beetle green attack
has a concurrent effect on key foliar biochemical properties (i.e. chlorophyll
and nitrogen concentration) as well as spectral responses (400–2000 nm). The
results demonstrate that the foliar reflectance of needles from green attacked
trees differs significantly from healthy needles (p ≤ 0.05), in particular
between wavelengths of 730 and 1370nm. Furthermore, the infestation
significantly (p ≤ 0.05) affected the biochemical concentrations of total
chlorophyll and foliar nitrogen and reduced their retrieval accuracy using PLSR
(Fig. 2.3 and 2.4).

The measured reflectance spectra of healthy needles matched those of similar

studies which measured needle reflectance of Norway Spruce (Ali et al., 2016;
Atzberger and Werner, 1998; Kováč et al., 2012). However, there were
significant differences observed between the reflectance spectra of the healthy
and the green attacked samples. This can be attributed to differences in their
foliar biochemical properties, especially to their chlorophyll concentration
which is known to be effected by vegetation stress. In our study, the
reflectance increased in the visible region, and decreased in the NIR, for the
green attack leaves. This is in line with the findings of Ahern (1988a), who

27
Effects (Ipstypographus, L.) green attack on leaf properties

studied the spectral range of 400- 1100 nm and showed that the spectral band
at the green peak increased for the lodgepole pine needles infested by similar
beetle species (Mountain pine beetle) at the green attack stage. In the visible
region, the concentration of pigments such as total chlorophyll is the main
factor for determining leaf spectral variation and absorption peaks (Carter and
Knapp, 2001; Demetriades-Shah et al., 1990; Feret et al., 2008; Zhang et al.,
2008). The infested trees had significantly (p < 0.05) lower chlorophyll and
foliar nitrogen concentrations than the healthy ones, resulting in lower
absorption and higher scattering (Fig. 2.3 and 2.4). The observed behaviour of
the reflectance spectra in the visible region in this study is a clear indication of
stress caused by early infestation and is in agreement with the results of earlier
studies focused on plant stress (Carter, 1993; Carter and Knapp, 2001; Zarco-
Tejada and Sepulcre-Cantó, 2007). Furthermore, wavelength ranges between
730-790 nm that contained the most informative spectral region regarding
variations in chlorophyll concentration, demonstrated the greatest amount of
dissimilarity (58 wavebands) between the healthy and infested leaf samples
(Fig. 2.7). This spectral region is often referred as red-edge, and it responds
quickly to changes in foliar chlorophyll (Carter, 1993; Carter and Knapp, 2001;
Smith et al., 2004). This is in good agreement with finding by Lottering et al.
(2016), as they studied the utility of spectral vegetation indices derived from
WorldView-2 data in detecting and mapping G. scutellatus induced vegetation
defoliation. Their result showed that the combination between red-edge and
NIR region has the potential to detect stress induced by G. scutellatus insect.

Moreover, as noted in Figure 2.3, the more pronounced changes in the

reflectance spectra were observed at the wavelength region between the NIR
and SWIR regions (730-1370 nm), where the reflectance of the infested leaves
was distinctively lower than that of the healthy ones. This can probably be
explained by the changes in water content that occur due to the infestation,
which leads to a degeneration of the internal leaf structure at the cellular level
(Miller et al., 1991; Murtha, 1978; Paine et al., 1997a; Slaton et al., 2001a;
Zhang et al., 2012b). This result partially agrees with the findings of Niemann
et al. (2015), who demonstrated that at the wavelength of 970 nm there was
no separation between healthy trees and samples infested by mountain pine
beetle, while at the wavelength of 1200 nm a significant difference existed
between them. Additionally, Ismail and Mutanga (2010) showed the
importance of the wavelength between 900 – 1110nm and identified both
Ratio975 and water index has the ability to assess water stress induced by S.
noctilio in P. patula trees from an early stage of infestation when there is no
sign of infestation at the canopy level.

In addition to the variations in VIS and NIR, the reflectance spectra of healthy
and green attacked samples were also significantly different ( p < 0.05) in two
short-wave infrared regions (1430-1500 nm and 1897-2000 nm) (Fig. 2.3).

28
 Chapter 2

The most likely cause for this variation is the low water and nitrogen content
in the infested samples (Curran et al., 1992; Mirzaie et al., 2014; Peñuelas and
Filella, 1998). Physiological research has shown that the fungi carried by the
beetles penetrate the living phloem, hampering the translocation of water,
sugar and other nutrients within the bole of the tree (Paine et al., 1997a; Rohde
et al., 1996; Safranyik et al., 2007; Wermelinger, 2004). Not only are the fungi
instrumental in the decline of trees, but they also provide a necessary food
source for insects during the infestation period (Runesson, 1991). This leads
to a gradual decrease in needle biochemical content and water content. Foster
et al. (2017) and Cheng et al. (2010) showed that a number of wavebands in
the shortwave infrared region have the potential for the detection of bark
beetle infestation at the green attack stage which is in line with our
observation. Furthermore, in our study, we have verified that the infested trees
had a significantly ( p < 0.05) lower nitrogen concentration than that found in
the healthy trees.

It should also be noted that we found that infestation at the green attack stage
weakened the correlation between chlorophyll and foliar nitrogen
concentrations. The correlation dropped from r = 0.68 to r = 0.57, which may
be partially explained by the reduction in the range of the two parameters (Fig.
2.4). Furthermore, the correlation between reflectance and both total
chlorophyll and foliar nitrogen deteriorated as leaves were stressed by a bark
beetle green attack (Fig. 2.6). The significant correlation (negative) between
foliar nitrogen and spectral reflectance at the 700 and 1200nm (Fig. 2.6) is
due to high concentrations of foliar nitrogen in healthy needles as they also
contain high chlorophyll (and possibly other pigment) concentrations, lowering
the reflectance. Under stress, trees tend to break down their pigments (Carter
and Knapp, 2001), which probably effects the relationship between nitrogen
and reflectance for the infested needles. Therefore, the infestation affects the
selection of wavelengths that were most sensitive to the biochemical
properties, in particular, the chlorophyll concentration, and their retrieval from
spectral reflectance. This is demonstrated in Figure 2.7, where distinctively
lower VIP scores were observed for the infested samples in the 730-790 nm
spectral region. Consequently, the accuracy of estimation of the two
parameters in the infested samples decreased (Table 2.1). To investigate the
effects of bark beetle green attack on the prediction bias for total chlorophyll
and foliar nitrogen we have used standardized residuals. In general, as can be
seen from figure 2.8, the infestation at the green attack stage caused under-
estimation of these two biochemical parameters from hyperspectral
measurements.

29
Effects (Ipstypographus, L.) green attack on leaf properties

Figure 2. 8: Standardized residuals of predicted total chlorophyll and nitrogen

concentration for healthy and infested samples.

2.5 Implications for remote sensing applications

Our study confirms the importance of hyperspectral measurement as well as
foliar biochemical properties (i.e. chlorophyll and nitrogen) for the detection of
Ips typographus, L. green attack. An extension of this finding would be to
investigate the up-scaling of our findings to the canopy level. However, this
forms a challenge as many factors such as logistical and technological aspects
may limit the scaling up to the canopy level (Foster et al., 2017; Wulder et al.,
2009). One of the critical logistical factors that affect a remotely sensed survey
for green attacked trees is the timing of the optimal conditions for image
acquisition. In addition, spatial and spectral resolutions play an important role
in the studies of bark beetle green attack. However, this limitation may be
overcome by utilising a UAV-based application and newly available
multispectral satellite data, such as provided by Sentinel-2. In our study, the
estimation accuracy of foliar chlorophyll and nitrogen concentration using
hyperspectral measurements decreased when the tree was infested by bark
beetle green attack. This might be used as an indicator for the efficient
landscape-wide detection of bark beetle green attack. It is, however, important
to note that retrieval accuracies for the total chlorophyll and foliar nitrogen

30
 Chapter 2

concentrations at the leaf level would probably be different from those at the
canopy level, because structural and external factors such as illumination and
atmospheric conditions may affect the reflectance spectra at the canopy level.
Additionally, we found that the wavelength region between 730 and 790 nm
(red-edge) is the most informative spectral region regarding variations in
chlorophyll concentration due to bark beetle green attack (Fig. 2.6 and 2.7).
This portion of the spectrum can be found in new multispectral satellites such
as Sentinel-2, World View-2&3 RapidEye and therefore it might be possible to
scaling up our findings to the canopy level, particularly considering the high
spectral and spatial resolution data. Further investigation will show how
accurate green attack stages can be detected with different air and spaceborne
sensors.

2.6 Conclusion
This study demonstrates that reflectance properties of healthy and green
attached Norway spruce trees are significantly different in 917 wavebands
between 400-2000 nm. We also observed differences in the biochemical
properties chlorophyll and nitrogen between healthy and green attacked
needles by the bark beetle (Ips typographus, L.). Furthermore, the results
demonstrated that the relationship between reflectance and both chlorophyll
and nitrogen deteriorated as the leaves became stressed by a bark beetle
green attack. Such an infestation affected the estimation accuracy of
chlorophyll and nitrogen concentrations, examined using PLSR and
hyperspectral reflectance data. The research findings indicate that
hyperspectral measurements are promising, and present a powerful tool to
determine the damage caused by bark beetle green attack at the leaf level.
Further research is required to assess whether or not, the findings of the
current study can be verified at the canopy level using different remote sensing
data.

31
Effects (Ipstypographus, L.) green attack on leaf properties

32
Chapter 3: Timing of red-edge and shortwave
infrared reflectance critical for early stress
detection induced by bark beetle (Ips
typographus, L.) attack


This chapter is based on:

Abdullah, H., Darvishzadeh, R., Skidmore, A.K., and Heurich, M. (Under review). Timing
of red-edge and shortwave infrared reflectance critical for early stress detection induced
by bark beetle (Ips typographus,L.) attack. International Journal of Applied Earth
Observation and Geoinformation. Accepted.

Abdullah, H., Darvishzadeh, R., Skidmore, A.K and Heurich, M. Understanding dynamics
of leaf spectral properties under bark beetle (Ips typographus,L.) infestation. 11th
EARSel SIG Imaging Spectroscopy Workshop, 6-8 February 2019, Brno, Czech republic.

33
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

Abstract
Forest disturbance in Europe, induced by European spruce bark beetle Ips
typographus, L., results in regional-scale dieback. Early stress detection in
Norway spruce stands caused by bark beetle infestation at the green attack
stage (when trees are yet to show distinct symptoms observable by the human
eye) is crucial and can lead to effective forest management and reduce
economic losses. This study aims to investigate and understand the dynamics
of leaf traits and reflectance of Norway spruce (Picea abies) trees during bark
beetle green attack. Using temporal high-resolution images from RapidEye and
SPOT-5 in parallel with the collection of field data, we examined which spectral
regions and leaf traits are affected by infestation over time and how they affect
discrimination between healthy and infested plots at the early stage of attack.
To achieve this aim, we used a novel approach by targeting both leaf and
canopy level. We measured leaf reflectance spectra and six leaf traits (water
content, nitrogen, chlorophyll fluorescence, chlorophyll, and stomatal
conductance) from 66 healthy and 54 infested trees at three consecutive
repeated time measurements in the summer of 2015 in the Bavarian Forest
National Park. Concurrently, canopy reflectance and spectral vegetation indices
(SVIs) were extracted from seven RapidEye images and six SPOT-5 images.
Results showed significant differences (p ≤ 0.05) in the studied leaf traits
between healthy and infested samples, and this difference increased with the
progression of infestation. We found that leaf and canopy reflectance were
significantly higher (p ≤ 0.05) for the infested trees by bark beetle than the
healthy ones in the ‘red edge’ (680–790 nm) and ‘shortwave infrared’ (1110–
1490 nm) spectrum throughout the infestation event. Our results further
demonstrated that the spectral vegetation indices calculated from the red-edge
and SWIR spectral bands, such as NDRE, DSWI, LWCI and NDWI, were able to
differentiate between healthy and infested trees earlier than the other SVIs.
The new insight offered by these results is that the red-edge and SWIR spectral
information from multispectral satellites has the potential to considerably
improve monitoring and detection of forest stress and has important
implications for European field bark beetle management and future studies.

34
 Chapter 3

3.1 Introduction
Insect outbreaks are one of the key natural disturbances in conifer forests that
trigger large-scale tree mortality, with noticeable effects on ecosystem services
(Aakala et al., 2011; Raffa et al., 2008; Thom and Seidl, 2016). One of the
insects that can create large-scale disturbances in short periods of time is the
European spruce bark beetle (Ips typographus, L.) (hereafter, referred to as
the bark beetle) (Fahse and Heurich, 2011). During recent decades, bark
beetle frequency and intensity have dramatically increased (Bentz et al.,
2010). For example, in the Norway spruce forests in Central Europe, spruce
bark beetles have killed a large number of Norway spruce trees in the order of
tens of millions of hectares (Lausch et al., 2013a; Meddens et al., 2012; Raffa
et al., 2008). Similarly, in British Columbia, Mountain pine beetles
(Dendroctonus ponderosae) have killed several million hectares of pine trees
since 1999 (Aukema et al., 2008; Westfall and Ebata, 2009). Bark beetles and
their host trees are susceptible to climatic change, in particular, decreasing
precipitation and increasing temperature (Netherer and Schopf, 2010).
Because the global average temperature is predicted to increase by 1.4–5.8°C
by 2100, enhancements of the intensity and incidence of bark beetle outbreaks
are expected (Morris et al., 2018; Netherer and Schopf, 2010; Overbeck and
Schmidt, 2012; Seidl et al., 2011). The bark beetle population has rapidly
changed from one to three generations per year, as regional and global
temperatures have increased (Bentz et al., 2010). Beetle survival rates have
also increased during the winter at northern latitudes, particularly in the old
spruce forests of northern and central Europe that have thus far been spared
from major outbreaks (Morris et al., 2018; Öhrn, 2012). Furthermore,
simulation studies on the bark beetle population have shown an increase in
temperature at higher latitudes and have identified high-risk areas for bark
beetle outbreaks and related tree mortality in the future (Bentz et al., 2010;
Cailleret et al., 2014).

The extensive increment in harvesting of European conifer forests, due to bark

beetle infestation, has resulted in further research that focuses on various
underlying forces to reduce further bark beetle outbreaks (Seidl et al., 2011).
One of the common techniques used to combat this threat is to cut and remove
the infested trees at an early stage to protect neighbouring unaffected trees
(Fahse and Heurich, 2011). This step should be performed during the early
phase of the infestation —the so-called ‘green attack’—and before beetle larvae
are fully developed and able to infest other nearby trees (Wermelinger, 2004).
At this stage, the beetles carry pathogenic blue stain fungus (Ophiostoma and
Ceratocystis species) and will transfer it to the host trees. This fungus affects
the translocation of nutrients and water within the trunk of an infested tree
(Paine et al., 1997; Rohde et al., 1996). As a result, the initial response of the
hosted trees by bark beetles is to close their stomata, a process which

35
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

decreases the water content due to the shutting down of the stem (Edburg et
al., 2012). This behaviour (closing stomata) of the infested trees leads to a
decrease in photosynthesis as a result of CO2 limitation (Flexas et al., 2004)
and eventual death through water stress.

Traditionally, foresters had to look for early signs of infestation by searching

for a dry brown powder that was produced by the bark beetles during the
colonisation process when the beetles tunnel under the bark of the trees
(Abdullah et al., 2018a). Such a technique, however, is not practical and is
inefficient for application in large areas because it is significantly laborious and
costly. Remote sensing data and techniques are a useful alternative for
effective forest management. The premise of employing remotely sensed data
to monitor and identify stressed or insect-infested forests is that the infested
trees show symptoms detectable by remote sensing sensors (Chen and
Meentemeyer, 2016).

Previous studies have shown the significant potential of remote sensing data
for detecting the advanced stages of bark beetle infestation (so-called ‘red-
attack’ and ‘grey-attack’) (Coops et al., 2006; Filchev, 2012; Franklin et al.,
2003; Hais et al., 2009; Havašová et al., 2015; Meddens et al., 2013; Skakun
et al., 2003; White et al., 2007; White et al., 2006; Wulder et al., 2006). A
red-attack is the advanced stage of bark beetle infestation, in which the
attacked trees develop stress symptoms by turning their needles’ colour from
green to yellow to red-brown. Subsequently, the needles fall from the infested
trees, and only the grey bark remains; hence, this stage is termed ‘grey attack’
(Coulson et al., 1985). It is important to note that the early detection (green
attack) of a bark beetle attack is crucial for preventing an outbreak, and
salvage logging which is performed for loss recovery and appropriate bark
beetle management (Fahse and Heurich, 2011).

Timing plays a critical role in remote sensing and field surveys of bark beetle
green attack (Wulder et al., 2009). An operational survey has to be performed
when the infestation is in its early phase and should consider the temporal field
observation associated with beetle biology and the appearance of the
symptoms in the tree foliage. In other words, in remote sensing-based surveys
of green attack, the swarming event of the beetles must be considered. For
example, the European spruce bark beetle (Ips typographus, L.) starts to
swarm when the air temperature reaches 16.5°C (Lobinger, 1994;
Wermelinger, 2004). To date, the identification of green attack has been less
satisfactory due to many biological, technical and logistical
constraints/limitations. These issues include the flight activity of the bark
beetle, the time duration required for the colonisation process and the period
when the infested trees have yet to show distinct symptoms capable of being
observed by human eye. Hence, it is important to consider biological as well

36
 Chapter 3

as logistical factors when using remotely sensed data to detect bark beetle
green attack (Wulder et al., 2009).

Numerous studies have focused on using single-date remote sensing data or

have compared the utility of different remotely sensed data from different
sensors, to detect early infestation stages; hence, very little attention has been
paid to investigating the temporal response of trees under bark beetle attack.
For example, early detection of a different beetle (Dendroctonus southern pine
beetle (spp) in lodgepole pine trees) has been investigated by Gimbarzevsky
et al. (1992), Murtha (1972) and Murtha and Wiart (1989) using multispectral
aerial photographs. However, it has been challenging in these studies to
differentiate between green attacked trees and the healthy ones. Similarly,
Heath (2001) encountered similar challenges when using airborne
hyperspectral data from CASI (Compact Airborne Spectrographic Imager). On
the other hand, a number of studies have attempted to detect the early stage
of Ips typographus, L. green attack using different remote sensing data and
techniques. Ortiz et al. (2013) used TerraSAR-X and single-date high-
resolution data from the RapidEye satellite. Lausch et al. (2013b) used a single
date airborne hyperspectral HyMAP image data for detecting bark beetle green
attack. More recently, Immitzer and Atzberger (2014) and Näsi et al. (2018)
have used WorldView-2, and hyperspectral data obtained from an unmanned
aerial vehicle (UAV) to detect bark beetle green attack.

As mentioned previously, during the colonisation process, the beetles inoculate

the host trees with pathogenic fungi, such as Ophiostoma and Cerato-cystis.
Therefore, the infested trees will develop stress symptoms, due to changes in
biochemical and spectral properties. As a result, the biochemical properties
such as leaf water content, chlorophyll, nitrogen, and stomatal conductance
are expected to decline within trees colonised by bark beetles (Abdullah et al.,
2018a).

Given the above biological and logistical factors, continuous monitoring is

essential for investigating and understanding the dynamic characteristics of
leaf properties and canopy reflectance under bark beetle infestation, which
may help to detect and monitor changes at an early stage. Such observation
provides detailed information regarding the impact of bark beetle on infested
trees and it also allows us to explore the diagnostic effects at any stage and to
achieve and detect the possible effects of bark beetle infestation as early as
possible. Therefore, in this study for the first time, we sought to investigate
the temporal effect of bark beetle attack on infested trees in the early
development phase (green until the red attack stage). To do this, we targeted
both the foliar and canopy levels by continuously monitoring the changes using
temporal field data measurements and temporal remote sensing satellite data
from RapidEye and SPOT-5. The main goals of this study are (i) to analyse the

37
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

temporal dynamics for a number of leaf traits influenced by bark beetle attack
from the beginning of infestation until the advanced stage of attack (red
attack) and (ii) to identify and explore remote sensing time-series indices
effective for explaining changes in leaf traits, as well as for detecting early
stages of bark beetle infestation.

3.2 Material and methods

3.2.1 Study area
The study site is the southern part of Bavarian Forest National Park (BFNP),
which is located in south-eastern Germany along the border of the Czech
Republic, and it lies between 13°12'9" E (longitude) and 49°3'19" N (latitude).
The BFNP established in 1970 and extended in 1977, covering an area of 240
km2. Depending on the elevation in the BFNP, which ranges from 600 m to
1450 m, the mean annual temperature fluctuates between 3.5° and 9°C, and
the total annual precipitation varies from 900 to 1800 mm (Bässler et al.,
2008; Heurich et al., 2010).

In the BFNP three major forest types can be recognised; these are, highlands,
hillsides, and valleys. The highlands above 1100 m are dominated by Norway
spruce (Picea abies), and some Mountain ash (Sorbus aucuparia); in the
hillsides, with elevation between 600 – 1100 m, is mixed forest including
Norway spruce (Picea abies), European beech (Fagus sylvatica), White fir
(Abies alba), and Sycamore maple (Acer pseudoplatanus); in the valleys,
spruce forests exist, including Mountain ash, Norway spruce, and birches
(Betula pendula, Betula pubescens) (Cailleret et al., 2014; Heurich et al.,
2009; Latifi et al., 2018). Since 1984, the forests of BFNP have been affected
by the spruce bark beetle (Ips typographus, L.), which have caused extensive
disturbance and economic losses to timber production in this region (Cailleret
et al., 2014).

3.2.2 Field data collection

From mid-May to mid-July 2015, an extensive field survey was conducted to
collect field measurements, in which the selected sample trees were visited on
three separate occasions (Table 3.1). We divided the study site into healthy
Norway spruce trees stands as well as stands with trees freshly infested by
bark beetles. Considering the nature of the forest heterogeneity in tree species,
tree age and density, 30 healthy plots were randomly selected. To select the
green attacked trees, we conducted an extensive field survey to search for dry
brown powder around the trees. Altogether, we selected 120 trees (66 healthy
and 54 green attacked) to measure foliar properties. The collected tree
samples were divided between 30 healthy and eight infested plots. Bark beetle
outbreaks generally occur over the course of several years, and to avoid mixed

38
 Chapter 3

reflectance from previous years’ attacked trees and green attacked trees, only
those plots that showed freshly infested or green-attacked trees were
considered. A Leica GPS 1200 differential global positioning system (DGPS)
(Leica Geosystems AG, Heerbrugg, Switzerland) was used to locate the centre
of each plot (Abdullah et al., 2018b).

From each tree, an average of two to three branches were taken from the
upper canopy exposed to sunlight. Because the height of Norway spruce trees
reaches 25 to 30 m, we used a crossbow to shoot an arrow with a fishing line
into a branch with sunlit leaves. Full details regarding the use of the crossbow
can be found in Ali et al. (2016). Next, to measure leaf traits, needle samples
were removed from collected branches. In the field, total chlorophyll,
chlorophyll fluorescence and stomatal conductance were measured. A
handheld chlorophyll content meter (CCM) was used to measure both
chlorophyll and chlorophyll fluorescence. From the fallen branches, an average
of ten readings were immediately taken using the CCM. To measure the needle
stomatal conductance, a steady-state instrument (SC-1 Leaf Porometer) was
used. This instrument computes stomatal conductance utilising a leaf clip
chamber that can monitor the relative humidity (RH %) released from the leaf
stomata.

The needle samples were covered by wet paper and placed in a labelled plastic
zip-locked bag in order to transport to the laboratory in a portable cooling box.
In the laboratory, an ASD FieldSepc-3 Pro FR spectrometer equipped with an
ASD RT3-3ZC integrating sphere (Analytical Spectral Devices, Inc., Boulder,
Co, USA) was used to measure the directional hemispherical reflectance from
350 to 2500 nm for the collected samples. The details concerning the
measurements of hemispherical needle reflectance can be found in Abdullah et
al. (2018) and Malenovský et al. (2006). To calculate leaf water content, fresh
weight (Fw/g) and leaf surface area (LA cm2) were determined for the collected
samples using a digital scale and an AMH 350 leaf area meter. The needle
samples were then oven-dried for 72 hours at 60°C until a constant weight was
obtained. Finally, to measure the nitrogen concentration of the dried needles,
an organic elemental analyser (FLASH 2000) was used.
Leaf water content (Cw) was determined using the following equations:

Cw (g/cm2) = (Fw – Dw) / LA (3.1)

where Fw, Dw, and A represent fresh leaf weight (g), dry leaf weight (g). The
studied leaf traits (total chlorophyll, chlorophyll fluorescence, nitrogen
concentration, water content, and stomatal conductance) were continuously
measured at three different times for the representative trees (Table 3.1).

39
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

Table 3. 1: The leaf traits measured and overview of satellite data acquisition in the
Bavarian Forest National Park, 2015. The () are the SPOT-5 data and (•) are the
RapidEye data. T1, 2, and 3 represent first, second, and third field measurements,
respectively.

3.2.3 Satellite imagery

We identified two series of high-resolution multispectral satellite data, namely
RapidEye and SPOT-5 (Table 3.1). The RapidEye orbit system consists of five
satellites, each of them collecting radiation in the five spectral bands blue,
green, red, red edge and near-infrared (NIR). In this study, seven RapidEye
images were captured within the time period May 2015 to September 2015.
They were systematically geo-corrected radiance and orthorectified at 5 m
spatial resolution and matched each other with sub-pixel accuracy (RapidEye,
2011; Tyc et al., 2005).

For the SPOT-5, we used six images from L2A reflectance product, that
provides data with atmospheric and topographic corrections as well as with
corresponding masks, clouds, and shadows. It captures data in 4 spectral
bands (green, red, near infrared and shortwave infrared). The SPOT-5 L2A has
10 m spatial resolution when acquired within the Take-5 initiative (Meygret,
2007).

3.2.3.1 Ancillary data (reference disturbance data)

Vector-based reference data of the green attacked areas in 2015 were obtained
from the Bavarian Forest National Park (BFNP) administration (Abdullah et al.,
2018b). The reference data is produced from aerial colour-infrared (VIS and
NIR) images with 0.1 m spatial resolution. The aerial images were obtained
during the flight campaign that were carried out in September 2016, and for
which the new deadwood (grey attack stage) from the 2015 infestation was
documented by stereo photogrammetry. For more detailed information about

40
 Chapter 3

the processing and interpretation of aerial images in the BFNP see Heurich et
al. (2009) and Lausch et al. (2013a).

The data were rasterised into 5 m × 5m grid cells to match with a spatial
resolution of RapidEye, and 10 m × 10 m grid cells to match with a spatial
resolution of SPOT-5 data. From the rasterised data 299 infested plots were
selected to extract the reflectance value (Fig. 3.1).

Figure 3. 1: Distribution of sample plots in the southern part of Bavarian Forest

National Park, Germany.

3.2.3.2 Spectral vegetation indices

Several spectral vegetation indices exist in the literature that are used for the
estimation of vegetation biochemical properties (Collins and Woodcock, 1996;
Eitel et al., 2006). In this study, spectral vegetation indices linked to measured
leaf traits were calculated from the spectral reflectance data collected from the
SPOT-5 and RapidEye images (Table 3.2). The spectral reflectance and SVIs
value were extracted from satellite images for the selected 30 healthy and 299
infested sample plots. The value extraction was done using ENVI-IDL5.5
spectral analysis toolbox. From the calculated vegetation indices, time series
data stacks were then generated from each sample plot. The time series data
were then subjected to further analysis.

41
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

Table 3. 2: List of spectral vegetation indices calculated from SPOT-5 and RapidEye data
for both healthy and infested sample plots.

3.2.4 Data analysis

To understand the temporal variations in all the measurements (leaf traits and
extracted canopy reflectance data), we first explored the temporal variation
from the time-series data (T1–T3). At the leaf level, mean and standard error
values were calculated for both the healthy and infested sample trees to
determine the temporal variation in the biochemical components. In addition,
we conducted a Student t-test to investigate whether the variations in the
measured leaf traits were significantly different (p < 0.05) between healthy
and green attacked samples. Similarly, an unpaired Student t-test was used to
examine whether the temporal variation in leaf reflectance spectra correlated
with the bark beetle infestation and to identify the wavelength regions which
were significantly different between the healthy and infested samples.

At the canopy level, box plots and Student t-tests were applied. Firstly, a box
plot was used to explore temporal variations in the extracted spectral
reflectance data in the given years. Secondly, a t-test was used to examine

42
 Chapter 3

whether there were significant differences (p < 0.05) in reflectance data

between these two sample plots (healthy and infested). Moreover, from the
calculated spectral vegetation indices, the mean and standard errors were
extracted for each sample plot, and a temporal comparison was made.

3.3 Results
3.3.1 Temporal response of measured leaf traits under bark
beetle
The results of the temporal variation demonstrated that there was a clear
distinction between healthy and infested sample trees from the Time-1 (T1) to
Time-3 (T3) measurements (Fig. 3.2). In the healthy leaves, the mean of the
total chlorophyll concentrations were 337, 403, and 399 mg/m2 at T1 and T3
respectively. However, for the infested samples, the mean of the total
chlorophyll concentrations were lower, viz. 317, 301, and 275 mg/m2,
respectively. Similar trends were observed for the other leaf traits, where all
measured leaf traits values decreased within the infested samples from T1 to
T3. For example, the stomatal conductance which was measured at the
beginning of the infestation at T1 for the infested leaf was higher when
compared to the ones at T3, whereas, for the healthy leaves, the opposite
result was observed.

This was further confirmed by the results of the Student t-test, which showed
a significant difference (p < 0.05) between healthy and infested samples in
all measured leaf traits (chlorophyll fluorescence, chlorophyll, nitrogen
concentration, stomatal conductance, and leaf water content ). At the
beginning of the infestation (T1), the differences were smaller between the two
sample groups (healthy and infested) for all studied leaf traits. However, there
was a larger difference between the infested and healthy needles at T2 and T3,
as the infested trees developed signs of stress (Fig. 3.2).

43
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

Figure 3.2: Temporal variation of the measured leaf traits for healthy and infested
samples. Blue and black lines represent healthy and infested plots, respectively. T1, T,2
and T3 represent first, second, and third measurements, respectively. (*) Hardly
significant, (**) significant, (***) Strongly significant.

44
 Chapter 3

Further analysis of the leaf spectral reflectance spectra values showed that the
mean reflectance spectra of green attacked samples was higher than for the
healthy ones in the visible and shortwave infrared region (Fig.3.3). The
difference (p < 0.05) was stronger during T2 and T3, as the infested trees
showed increased stress symptoms.

Figure 3. 3: Mean reflectance spectra of healthy and green attacked leaves at three
consecutive repeated time measurements in the summer of 2015. Blue and black lines
represent healthy and infested leaves, respectively. Wavebands in which there is a
significant difference between healthy and green attacked leaves are presents in Gray.

3.3.2 Temporal response of canopy spectral data to bark

beetle infestation
We assessed the response of canopy reflectance to bark beetle attack
employing a time series of remotely sensed images from RapidEye and SPOT-
5. The reflectance data were extracted from each image and for both sets of
satellite data (RapidEye and SPOT-5). We found a symmetrical change of
reflectance data for both sets of satellite data (Figs. 3.4 and 3.5). For example,
for the RapidEye imagery, the reflectance of red-edge and NIR bands were
higher for the infested plots than the healthy ones on May 13. The variation of
these two spectral bands increased with the progression of infestation on July
2 and the later images. Moreover, a Student t-test of spectral bands revealed
similar findings, as the red-edge was significantly different (p < 0.05) between
healthy and infested plots for all RapidEye imagery considered in this study,
while the variation in NIR between healthy and infested plots showed a
significant difference (p < 0.05) from July 2 and later (Fig. 3.4). For the SPOT-
5 imagery, the reflectance of SWIR appears to be the most sensitive to the
bark beetle stressor, as it was significantly different (p < 0.05) from May 11
onwards (Fig. 3.5).

45
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

Figure 3. 4: (A) Temporal variation of canopy reflectance for healthy and infested plots
in the Bavarian Forest National Park using RapidEye satellite data. Red and Green boxes
represent infested and healthy plots, respectively (B) Unpaired t-test canopy reflectance
between healthy and infested samples. Dark squares indicate spectral wavebands that
were significantly different (P ≤ 0.05). The red box shows the spectral region that was
significantly different overall temporal data considered in this study

46
 Chapter 3

Figure 3. 5: (A) Temporal variation of reflectance spectra for healthy and infested plots
in the Bavarian Forest National Park using SPOT-5 satellite data. Red and Green boxes
represent infested and healthy plots, respectively (B) Unpaired t-test of canopy
reflectance between healthy and infested samples. Dark squares indicate spectral
wavebands that were significantly different (P ≤ 0.05). The red box shows the spectral
region that was significantly different overall temporal data considered in this study

Further analysis regarding the SVIs shows that there were a big overlap
between healthy and infested plots in May for both RapidEye and SPOT-5.
However, as infestation progresses, this variation becomes larger and more
distinct (Figs. 3.6 and 3.7). For the RapidEye data, the difference between the
healthy and infested plots by means of spectral vegetation indices values
started to change from July 2nd using NDRE and NGRDI, while the other indices
started to enlarge the variation at the later dates considered in this study (Fig.
3.6). A similar result was found using a reflectance difference index, as the
red-edge and NIR bands became statistically different from May, becoming
progressively more pronounced at later dates.

47
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

Using SPOT-5, the spectral vegetation indices calculated from the combination
of SWIR and other spectral bands, such as DSWI, LWCI and NDWI, exhibited
a larger difference between the healthy and infested plots at the earlier stage
of infestation (Fig. 3.7). This is also confirmed in the result of the Student t-
test analysis, where the SWIR was significantly different (P < 0.05) for all dates
considered in this study.

Figure 3. 6: Temporal variation of studied spectral vegetation indices for healthy and
infested plots in the Bavarian Forest National Park using RapidEye satellite data. Green
and black lines represent healthy and infested plots, respectively.

48
 Chapter 3

Figure 3. 7: Temporal variation of studied spectral vegetation indices for healthy and
infested plots in the Bavarian Forest National Park using SPOT-5 satellite data. Green
and black lines represent healthy and infested plots, respectively.

3.4 Discussion
In this study, we demonstrated that the measured leaf traits and leaf
reflectance spectra from infested trees differed significantly (p ≤ 0.05) from
healthy ones during T1, T2, and T3 measurements (Figs. 3.2 and 3.3).
Furthermore, the red-edge and SWIR spectral bands from RapidEye and SPOT-
5 were influential in the separating between healthy and infested plots.
Moreover, we found that red-edge and SWIR bands maintained their sensitivity
for monitoring and detecting bark beetle infestation from the early to the
advanced stages of infestation (Figs. 3.4 and 3.5).

The absolute difference in measured leaf traits between the healthy and
infested trees increased from T1 to T3. In particular, stomatal conductance and
chlorophyll fluorescence exhibited significant differences between the healthy
and infested trees (Fig. 3.2). A possible explanation for this result might be
that the initial impact of bark beetles on the infested trees shuts down the
translocation of water in the tree due to the blue stain fungi introduced to the

49
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

trees during the infestation process. Because the hydraulic systems of plant
and stomatal conductance are closely correlated with each other (Ewers et al.,
2007), infested trees tend to close their stomata to preserve water. Therefore,
the stomatal closure is the first sign of stress that induces the physiological
response of infested trees to preserve water. It is well-known that physiological
factors, such as plant water content, can control the temperature of plants
through the stomatal transpiration process (Oerke et al., 2006). Moreover, the
stomatal closure leads to a drop in photosynthesis activity due to the CO2
limitation, and, therefore, the chlorophyll fluorescence decreases (Flexas et al.,
2004; Zweifel et al., 2009). Hence, the distinct variation observed in stomatal
conductance between healthy and infested trees is attributed to differences in
the measured leaf traits—in particular, leaf water content and chlorophyll
fluorescence (Fig. 3.2). These findings match those observed in earlier studies
that the initial impact of mountain pine beetles on infested trees causes a drop
in the sapwood moisture and, hence, stomatal closure (Yamaoka et al., 1990).
Likewise, our findings are in agreement with those of (Cheng et al., 2010), who
studied the impact of a similar species, the mountain pine beetle
(Dendroctonus spp) on the leaf water content of lodgepole pine using
hyperspectral data.

Another important finding at the leaf level was that the difference in reflectance
spectra between healthy and infested samples increased with the progression
of infestation from T1 to T3 (Fig. 3.3). Noticeable changes were perceived at
the visible wavelengths, in particular, from 680–790 nm and in the shortwave
infrared wavelengths 1110 –1490 nm. The reflectance of the infested leaves in
the visible spectrum from T1 was distinctly higher than that of the healthy
ones. This result was likely due to having significantly (p ≤0.05) lower
chlorophyll in infested leaves than the healthy leaves from T1 to T3, causing
lower absorption and higher scattering in the visible wavelengths (Fig. 3.2)
(Carter and Knapp, 2001; Zhang et al., 2008). Furthermore, we reported that
the reflectance of the healthy leaves was higher in the NIR region and lower in
SWIR than that observed in the green attacked ones. Physically, the
reflectance at the leaf level is a function of dry matter, water content,
chlorophyll and internal leaf structure (leaf thickness) (Ali et al., 2016; Pu and
Gong, 2011; Slaton et al., 2001). Hence, the observed reflectance pattern of
the infested samples may have been due to the decrease in water content (Fig.
3.2) which is caused by bark beetle infestation.

Similarly, at the canopy level, both red-edge and SWIR bands from RapidEye
and SPOT-5, respectively, showed a significant difference between the healthy
and infested sites for all the dates considered in this study (Figs. 3.4 and 3.5).
When implemented as a time series, they showed distinct temporal variation
in reflectance and indices values between the healthy and infested sites,
especially during the later stages of infestation. For example, the NDRE

50
 Chapter 3

calculated based on the combination of the red-edge and NIR bands is shown
to be more sensitive than the other SVIs due to the stress induced by bark
beetle infestation. The NDRE is known to detect stress or any forest health
decrease earlier than NDVI (Eitel et al., 2011). In our study, the NDRE started
to differentiate between the healthy and infested plots from 2nd July, while the
other SVIs showed a difference almost one month later (Fig. 3.6). This result
is in agreement with our earlier findings (Abdullah et al., 2018b) in which the
NDRE calculated from the Sentinel-2 imagery possessed the most sensitive
indices to differentiate between the healthy and green attacked sample plots.
All the indices, such as DSWI, NDWI and LWCI, which employed the SWIR
bands, were mostly sensitive to stress caused by variations in water content
and were able to discriminate healthy from infested sites. For example, the
DSWI and NDWI showed significant differences between the two groups on
June 5th and July 10th, respectively. This is due to the significantly lower water
content (P < 0.05) of the infested samples, and, therefore, their spectral
reflectance in the SWIR region were highly affected (Figs. 3.2 and 3.5). This
also accords with our earlier observation (Abdullah et al., 2018b), which
showed the SWIR bands from Sentinel-2 and Landsat-8 were sensitive to
detecting bark beetle infestation at the early phase of the attack. Likewise,
Foster et al. (2017) identified the shortwave infrared region as key for
detecting the early stages of beetle infestation in Engelmann spruce trees.

Based on the temporal variations in the canopy reflectance values and

measured SVIs, here we identified for the first time in the European situation
(and in contrast to earlier works in the US) that mid-June to the beginning of
July is an appropriate time frame for the early stress detection induced by bark
beetle infestation (Fig. 3.3, 3.4 and 3.5). This is the most initial period when
the spectral difference between infested and healthy plots peaks. Although this
time period may be slightly late for appropriate bark beetle management, our
study has however shown the impending role of multispectral satellite
(RapidEye and SPOT-5) data for monitoring and detecting forest stress induced
by bark beetle attacks, and thus, has important implications for European field
bark beetle management and future studies. However, further research is
essential to explore the potential of other spaceborne sensors in detecting
canopy reflectance changes due to bark beetle infestation during the early
stage of an attack.

3.5 Conclusion
This study presents an innovative approach to investigate the temporal
response of leaf properties and canopy reflectance spectra to European bark
beetle infestation using a temporally dense time series of seven RapidEye
senses and six SPOT-5 senses in parallel with the collection of field data at
three consecutive repeated time measurements. Remote-sensing-based SVIs

51
Dynamics of leaf and canopy properties under (Ips typographus, L.) Infestation

complemented the ground data collected, providing additional information to

help characterise the temporal response of infested trees through time and
across the landscape. In conclusion, our key concept is that we show for the
first time that:
 All measured leaf traits from infested trees differed significantly from
healthy trees during Time-1, -2, and -3 measurements.
 The red-edge and SWIR were important spectral regions at both leaf and
canopy levels for detecting subtle changes in Norway spruce trees due to
bark beetle infestation.
 The earliest period at which the spectral difference between infested and
healthy plots peaks is mid-June to the beginning of July.
We recommend that further studies should investigate different remote
sensing datasets and monitor over a longer period to show the relationship
between bark beetle infestation and the changes in biochemical variables and
spectral reflectance more clearly.

52
Chapter 4: Sensitivity of Landsat-8 OLI and
TIRS Data to Foliar Properties of Early Stage
Bark Beetle (Ips typographus, L.) Infestation


This chapter is based on:

Abdullah, H., Darvishzadeh, R., Skidmore, A.K., and Heurich, M. ,2019. Sensitivity of
Landsat-8 optical and thermal infrared data to foliar properties at early stage bark beetle
(Ips typographus, L.) infestation. Remote sensing, 4, 398.

53
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

Abstract
In this study, the early stage of European spruce bark beetle (Ips typographus,
L.) infestation (so-called green attack) is investigated using Landsat-8 optical
and thermal data. We conducted an extensive field survey in June and the
beginning of July 2016, to collect field data measurements from several
infested and healthy trees in the Bavarian Forest National Park (BFNP),
Germany. In total, 157 trees were selected, and leaf traits (i.e. stomatal
conductance, chlorophyll fluorescence, and water content) were measured.
Three Landsat-8 images from May, July, and August 2016 were studied,
representing an early stage, advanced stage, and post-infestation,
respectively. Spectral vegetation indices (SVIs) sensitive to the measured
traits were calculated from the optical domain (VIS, NIR, and SWIR), and
canopy surface temperature (CST) was calculated from the thermal infrared
band using the mono-window algorithm. The leaf traits were used to examine
the impact of bark beetle infestation on the infested trees and to explore the
link between these traits and remote sensing data (CST and SVIs). The
differences between healthy and infested samples regarding measured leaf
traits were assessed using Student’s t test. The relative importance of the CST
and SVIs for estimating measured leaf traits was evaluated based on the
variable importance in projection (VIP) obtained from the partial least squares
regression (PLSR) analysis. A temporal comparison was then made for SVIs
with a VIP > 1, including CST, using statistical significance tests. The clustering
method using a principal components analysis (PCA) was used to examine
visually how well the two groups of sample plots (healthy and infested) are
separated in 2-D space based on principal component scores. Finally, linear
regression (LR) was used to generate the leaf traits maps using the SVI that
have highest VIP score and then used to produce a stress map for the study
area. The results revealed that all measured leaf traits were significantly
different (p < 0.05) between healthy versus infested samples. Moreover, the
study showed that CST was superior to the SVIs in detecting subtle canopy
changes due to bark beetle infestation for the three months considered in this
study. The results showed that CST is an essential variable for estimating
measured leaf traits with VIP > 1, improving the results of clustering when
used with other SVIs. Likewise, the stress map produced by CST and leaf traits
well presented the infestation areas at the green attacked stage. The new
insight offered by this study is that the stress induced by the early stage of
bark beetle infestation is more pronounced by Landsat-8 thermal bands than
the SVIs calculated from its optical bands. The potential of CST in detecting
the green attack stage would have positive implications for forest practice.

54
 Chapter 4

4.1 Introduction
Forests are important ecosystem with economical, social, and ecological
values. The economic value of timber in a forest is typically threatened by
natural disturbances such as fire, drought, wind, snow, and insect or disease
outbreaks (Morris et al., 2017; Tchakerian and Coulson, 2011). In Europe, the
European spruce bark beetle (Ips typographus, L.) is a common disturbance
agent in forests dominated by Norway spruce (Picea abies) (Schelhaas et al.,
2003). Bark beetle infestation extends across ten million hectares of trees in
Europe. Much public money has been invested to compensate forest owners
for their economic loss and for the cost of reforestation (Eidmann, 1992;
Pasztor et al., 2014; Seidl et al., 2008). Besides the negative impact on timber
production, bark beetles can also have a positive impact on the ecosystem by
providing a suitable habitat in the form of opening forest canopy, increasing
habitat heterogeneity and biodiversity, all of which enhance the survival of
other species (Lehnert et al., 2013; Müller and Bütler, 2010).

The principle of bark beetles killing host trees has been well described in
Wermelinger (2004). After a successful attack, the trees change colour in three
stages; referred to as green, red, and grey attack, respectively (Raffa et al.,
2015; Wermelinger, 2004). The green attack is the early stage of bark beetle
infestation in which the colonised trees are yet to show distinct symptoms
observable by the human eye (Niemann and Visintini, 2005a; Wulder et al.,
2006c). A red-attack is evidence of the advanced stage of bark beetle
infestation, in which the attacked trees develop stress symptoms involving
turning the colour of their needles from green to yellow to red-brown.
Subsequently, the needles fall from the infested trees, and only the grey bark
remains; hence, this stage is termed ‘grey attack’ (Coulson et al., 1985).

Detection of bark beetle infestation at the green attack stage means locating
infested trees at the stage where the beetle larvae are still within the tree.
Effective measures to manage the beetles can then be undertaken. This will
ultimately help reduce the number of infested trees, lessening the economic
loss. Visual inspections during field surveys and pheromone traps have
traditionally been used to detect bark beetle infestations in Norway’s spruce
forests. However, these methods are subjective, very laborious, costly, and
only able to cover relatively small areas. Remote sensing presents an
alternative to existing methods for monitoring and detecting infestations on
large spatial scales (Wulder et al., 2006a). To date, remote sensing data have
been successfully used to detect the advanced stages (red and grey attacks)
of bark beetle infestation.

For example, many studies have employed spectral vegetation indices (SVIs)
from low-to-medium resolution satellite data to detect the advanced stages of

55
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

bark beetle infestation (red and grey attack) (Filchev, 2012a; Franklin et al.,
2003b; Hais et al., 2009; Havašová et al., 2015; Meddens et al., 2013; Wulder
et al., 2006c); however, to date, the identification of European spruce bark
beetle (Ips typographus, L.) green attack has been less satisfactory due to
many biological, technical and logistical constraints/limitations. These issues
include the flight activity of the bark beetle, the time duration required for the
colonisation process and the period when the infested trees have yet to show
distinct symptoms capable of being observed by human eye. Hence, it is
important to consider biological as well as logistical factors when using
remotely sensed data to detect bark beetle green attack (Wulder et al., 2009)

Figure 4. 1: Studies that have used remote sensing data to attempt detecting bark beetle
infestation at the green attack stage.

There are very few studies that have paid particular attention to investigating
the impact of bark beetle green attack on biochemical properties and
physiological status of infested trees. However, they studied different beetle
(mountain pine beetle) and tree (lodgepole pine) species. For example,
Yamaoka et al. (1990) documented a decline in sapwood moisture content
resulting from mountain pine beetle attack. Likewise, Cheng et al. (2010)
revealed water deficit and changes in chlorophyll content of artificially stressed
pine trees by mountain pine beetle. They also identified the wavelength range
between 950 nm and 1390 nm as a good spectral region able to separate
healthy from green attacked trees. In general, moisture stress in vegetation
may result in non-visual symptoms that are detectable with remote sensing
data, particularly in the shortwave and thermal infrared regions, where water
absorption features exist (Berni et al., 2009; Buitrago Acevedo et al., 2017;
Jang et al., 2006; Sepulcre-Cantó et al., 2006). However, the majority of the
studies on bark beetle (either mountain pine beetle or spruce bark beetle)
green attack detection with remotely sensed data have mainly utilised optical
remote sensing data (Fig. 4.1). There are few published examples of studies
which investigate the use of temperature, and none which focus on the impact

56
 Chapter 4

of European spruce bark beetle infestation on biochemical properties and

physiological status of infested trees in the TIR domain. To our knowledge,
only one study, by Hais and Kučera (2008), has investigated the impact of bark
beetle (Ips typographus, L.) infestation at the advanced stages (Gray attack)
on surface temperature using thermal infrared (TIR) data. Their study
examined the effect of spruce bark beetle infestation and clear-cutting on
surface temperature between 1987 and 2002 in the central part of Sumava
National Park in the Czech Republic. Further, Sprintsin et al. (2011) used
surface temperature calculated from the Landsat 7 ETM+ to detect an early
stage of mountain pine beetle infestation in British Columbia. Although
Sprintsin et al. (2011) suggested that temperature condition indices (TCI) used
in their study have the potential to differentiate between healthy and green
attacked pine they could not validate their results due to lack of ground
reference data for their green attacked study areas.

Numerous studies have shown the significant potential of TIR data to elucidate
plant biophysical and biochemical properties (Buitrago et al., 2016; Neinavaz
et al., 2016; Ullah et al., 2014). For example, the primary absorption feature
that is associated with leaf water content can be observed in both shortwave
and thermal infrared regions (Table 4.1) (Fabre et al., 2011; Kümmerlen et
al., 1999). As such, several studies have shown that TIR data have great
potential to detect plant diseases and pathogens before the plants develop
visual stress symptoms (Aldea et al., 2005; Moller et al., 2007; Ni et al., 2015;
Oerke et al., 2006; Xu et al., 2006). Moreover, retrieving the canopy surface
temperature (CST) from TIR data is widely used to track vegetation water
status (Hunt and Rock, 1989). The surface temperature is interrelated with
plant functions, such as evapotranspiration which is controlled by stomatal
conductance. (Kim et al., 2016; Xu et al., 2006). Therefore, alterations in these
processes will lead to changes in the air and the surface temperature of leaves
(Doughty et al., 2011; Gersony et al., 2016; Stoner and Miller, 1975;
Vanderhoof et al., 2013). Furthermore, physiological studies have shown that
the reduction of leaf water content disturbs stomatal conductance and leads to
an increase in leaf surface temperature (Pierce and Congalton, 1988; Pierce et
al., 1990; Schulze and Hall, 1982).

In addition to water content and stomatal conductance, recent studies have

shown that there is a strong relationship between canopy surface temperature
and photosynthetic activity, which can be measured by chlorophyll
fluorescence (Ni et al., 2015). Chlorophyll fluorescence has been found to be
a reliable indicator for examining the impact of different stressors (drought,
insect infestation, and pathogens) on plant photosynthesis and the
physiological state of vegetation (Méthy et al., 1994; Porcar-Castell et al.,
2014). When leaf transpiration rate decreases (i.e., by closing stomata) both
chlorophyll fluorescence and photosynthetic activity will decrease, while heat

57
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

dissipation will increase (McFarlane et al., 1980). Hence, chlorophyll

fluorescence has been used as a reliable indicator for monitoring plant water
stress (Campbell et al., 2007; Yamaoka et al., 1990).

In this study, for the first time, we sought to explore and compare the potential
of both optical and thermal infrared (TIR) data from Landsat-8 for detection of
pre-visual symptoms induced by European spruce bark beetle green attack.
We studied the impact of spruce bark beetle green attack on leaf properties
(foliar stomatal conductance, chlorophyll fluorescence, and water content). To
do this, canopy surface temperature (CST) and spectral vegetation indices
(SVIs) were calculated using three Landsat-8 images for May, July, and August
2016, as these months, represent the time of early, advanced, and after spruce
bark beetle green infestation, respectively. Our study benefits from valuable in
situ measurements of leaf traits at the time of infestation that have been used
to explain the impact of bark beetle attacks on Landsat 8 optical and TIR data.
Our objectives were as follow: (i) to assess the potential of CST and SVIs
calculated from Landsat 8 data to differentiate between healthy and infested
sample groups; (ii) to study the temporal variation of CST and SVIs under
spruce bark beetle infestation; and (iii) to evaluate the utility of CST and a
number of SVIs to estimate measured leaf traits within infested and healthy
samples.

4.2 Material and Methods

4.2.1 Study site and in situ data collection
The field campaign was conducted during June and July 2016 in the Bavarian
Forest National Park (BFNP), which is a 24.222 ha forest located in south-
eastern Germany along the border with the Czech Republic, between 13°12'9"
E (longitude) and 49°3'19" N (latitude) (Fig. 4.2). The BFNP was established
in 1970, and significantly extended in 1997. Depending on the elevation within
the BFNP, which ranges from 600 m to 1453 m, the mean annual temperature
fluctuates between 3.5°C and 9°C, and the total annual precipitation varies
from 900 mm to 1800 mm (Bässler et al., 2008a; Heurich et al., 2010a). The
first bark beetle infestations began in 1984 and, to date, more than 7000 ha
have been affected. The strict non-intervention policy of the National Park
offers the possibility of studying bark beetle dynamics without human
interference (Cailleret et al., 2014b; Heurich et al., 2010a).

58
 Chapter 4

Figure 4. 2: The location of Bavarian Forest National Park in Central Europe.

We sampled trees from healthy stands and stands with trees freshly infested
by the bark beetle. For the healthy stands, we randomly selected plots (30 m
× 30 m) from all over the national park. To select the green attacked trees,
we conducted an extensive field survey to search for dry brown powder around
the trees. Care was taken to only consider those plots with all trees freshly
infested or which were dominated (80%) by freshly green-attack trees. In this
way, we were assured that the extracted remote sensing signature from these
plots would not have mixed effects from the attacked trees of previous years.
In total, 40 healthy and 21 infested plots were sampled. The centre of each
plot was measured using Differential Global Positioning System (DGPS) Leica
GPS 1200 (Leica Geosystems AG, Heerbrugg, Switzerland) with an accuracy of
better than 1 m after post-processing (Abdullah et al., 2018b). Each of 61
selected plots measured 30 × 30 m to encapsulate the spatial resolution of a
Landsat-8 data, which allows for a 30 m radius buffer zone around the ‘central
pixel’ location for uncertainty in spatial registration of image pixels. By doing
this we minimized the potentially confounding influence (mixed pixels) of green
attacked Norway spruce trees.

From each plot, three to five trees were selected as representative for the plot.
Because the height of Norway spruce trees reaches 25 to 30 m, we used a
crossbow to shoot an arrow with a fishing line attached to a branch with sunlit
leaves. The fishing line was used to feed a rope over the targeted branch from
the upper canopy. Once the rope was in place, the branch was pulled down

59
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

gently until it broke off. Full details regarding the use of the crossbow can be
found in Ali et al. (2016) .n. On average two to three branches, exposed to
sunlight, were taken from the upper part of each tree. Next, needle samples
were taken from the collected branches to measures leaf traits.

4.2.2 Measurement of leaf properties

In this study, a steady-state instrument (SC-1 Leaf Porometer) was used to
measure the needle stomatal conductance. This instrument computes stomatal
conductance using a leaf clip chamber that can monitor the relative humidity
(RH%) released from the leaf stomata. To measure the stomata, a number of
needles were gently attached to cover the chamber port. Since stomata are
sensitive to environmental conditions and physical stress, all the
measurements were taken under similar environmental conditions (clear sky,
between 11 am to 3 pm local time). An average of three to four measurements
was taken for each sampled tree within each plot (SC-1 Operator’s Manual,
2016).

Immediately after the stomata measurement, a chlorophyll content meter

CCM-330 was used to measure the chlorophyll fluorescence ratio (CFR). This
instrument uses the emission ratio of fluorescence at both the red (700 nm)
and the far red (735 nm) part of the electromagnetic spectrum, as proposed
by Gitelson et al. (1999). On average ten readings were taken per sample from
each branch. Then, the leaf water content (CW, mg/cm2) was computed using
fresh and dry weight. To do this, three grams of the freshly harvested needles
were weighed from each sample. A portable leaf area meter (AM-350) was
used to measure the total needles’ surface area for each of the three grams
considered. Norway spruce needles are cylindrical and, therefore, this surface
area was multiplied by a universal conversion factor of 2.57 (Waring, 1983).
Then, to acquire the hemispherical-surface projection of the sample surface
needles, the total area obtained was divided by 2. Finally, the needle samples
were oven dried for 72 hours at 75 ºC, and the dried weight of each sample
was measured. The leaf water content (Cw) was computed using the following
equation (Colombo et al., 2008):

Wf Wd
𝐶w
A
Where Wf and Wd are the fresh and dry weight, respectively, and A is the
sample leaf area.

4.2.3 Landsat-8 imagery and pre-processing

Cloud-free Landsat-8 images were obtained from the USGS Global Visualization
Viewer (http://glovis.usgs.gov/) for the months May, July and August 2016.
According to the Bavarian Forest National Park authorities, the bark beetles of

60
 Chapter 4

the infestation in 2016 had started to swarm on May 10, 2016. Therefore, the
May image was designated as the early stage of bark beetle infestation.

Landsat-8 has two sensors called OLI and TIRS. The OLI sensor collects data
from nine spectral bands ranging from 0.43 – 2.29 nm with a 30 m spatial
resolution, while the TIRS sensor collects data from two thermal bands ranging
from 10.6 – 12.5 nm with a 100 m spatial resolution resampled to 30 m in the
delivered data product (Table 4.1).

For both the OLI and the TIRS sensor, radiance values were calculated, using
the coefficient supplied by USGS. Secondly, for OLI data, a radiometric
calibration was applied to convert the radiance value to Top-of-Atmosphere
(TOA) reflectance. Then, MODTRAN4-based atmospheric correction software
(FLAASH) was used to convert the TOA Reflectance to surface reflectance
(Adler-Golden et al., 1999). Full details regarding the use of FLAASH can be
found in (Module). Finally, the reflectance values of the infested and healthy
plots were extracted from the Landsat-8 scenes and were used for further
analysis.

4.2.3.1 Spectral vegetation indices

Several spectral vegetation indices exist in the literature that are used for
estimation of vegetation biochemical properties (Collins and Woodcock, 1996;
Eitel et al., 2006). In this study, spectral vegetation indices linked to measured
leaf traits were calculated from the spectral reflectance data collected from the
Landsat OLI sensor. In previous work, the selected SVIs have shown sensitivity
to stress-induced variations in chlorophyll and also water content, which are
all important indicators of tree health. This sensitivity could potentially be used
to assess stomatal closure and tree water stress due to insect infestation. The
mathematical transformation for computing these vegetation indices is given
in (Table 4.2).

61
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

Table 4.1: The Landsat-8 sensors, the operational land imager (OLI), and the thermal
infrared sensor (TIRS) spectral bands and their spatial resolution.
Wavelength Resoultion
Bands
(micrometers) (meters)
Band-1 (Coastal aerosol) 0.43–0.45 30
Band-2 (Blue) 0.45–0.51 30
Band-3 (Green) 0.53–0.59 30
Band-4 (Red) 0.64–0.67 30
Band-5 (Near Infrared –NIR) 0.85–0.88 30
Band-6 (SWIR-1) 1.57–1.65 30
Band-7 (SWIR-2) 2.11–2.29 30
Band-9 (Panchromatic) 0.50–0.68 15
Band-10 (Cirrus) 1.36–1.38 30
Band-11 (Thermal infrared- 100 (resampled to
10.60–11.19
TIRS1) 30)
Band-12 (Thermal infrared- 100 (resampled to
11.50–12.51
TIRS2) 30)

Table 4.2: List of spectral vegetation indices calculated from Landsat-8 optical bands.
Index Formula Full name Reference
Cigreen (NIR/Green)-1 Chlorophyll index(Hunt et al.,
green 2011)
CVI NIR × (Red / Green2) Chlorophyll vegetation
(Hunt et al.,
index 2011)
CI (Red – Blue) / Red Coloration index (Escadafal et
al., 1994b)
GVI NIR - Green Green difference (Tucker et
vegetation index al., 1979)
DVI 2.4 × (NIR – Red) Difference vegetation (Bannari et
index al., 1995)
GVMI [(NIR+0.1)- Global vegetation (Ceccato et
(SWIR+0.02)]/[(NIR+0.1)+(SWIR+0.02)] moisture index al., 2002)
GARI [NIR-(Green-(Blue-Red))]/[NIR- Green atmospherically (Gitelson et
(Green+(Blue-Red))] resistant vegetation al., 1996)
index
GLI [2×(Green-Red- Green leaf index (Gobron et
Blue)]/[2×(Green+Red+Blue)] al., 2000)
LWCI [log(1-(NIR-SWIR))]/[log(1- Leaf water content (Cohen,
(NIR+SWIR))] index 1991)
NLI (NIR2 – Red) /( NIR2 + Red) Nonlinear vegetation (Goel and
index Qin, 1994)
PVR (Green – Red) / (Green + Red) Normalized Difference (Metternicht,
Photosynthetic vigour 2003)
ratio
SIWSI (NIR-SWIR1) / (NIR+SWIR1) Normalized Difference (Fensholt and
860/1640 Sandholt,
2003)
BGI Costal / Green Blue-green pigment This study
index
RDI SWIR2 / NIR Ratio Drought Index (Pinder and
McLeod,
1999)
NDVI (NIR – Red) /(NIR + Red) Normalized difference (Tucker,
vegetation index 1979)
TNDVI Log [(NIR-Red)/NIR+Red) × 0.5] Transformed NDVI (Bannari et
al., 2002)

62
 Chapter 4

4.2.3.2 Canopy surface temperature (CST)

The Landsat-8 TIR sensor provides data from two bands (band 10 and band
11) in the thermal domain. However, band 11 is no longer operational for
quantitative analysis, as reported by USGS (https://landsat.usgs.gov/using-
usgs-landsat-8-product). Therefore, in our study, band 10 with wavelengths
ranging from 10.60 to 11.19 nm was utilised to retrieve canopy surface
temperature CST.

To minimise and reduce the effects of different climatic conditions on CST

retrieval, the mono-window algorithm, proposed by Qin et al. (2001), was
applied. The core of this method is the equation transferring thermal radiance,
which converts digital satellite values to a radiometric value (Markham and
Barker, 1986). To apply this algorithm, a number of parameters such as
emissivity, air transmittance, and effective mean atmospheric temperature are
required. The air transmittance and mean atmospheric temperature are
estimated using another two additional parameters, namely: air surface
temperature and water vapour content. Water vapour was calculated using
relative humidity ratio and air surface temperature (Liu and Zhang, 2011). The
following equations summarise the steps required when using the Mono-
Window algorithm for retrieving CST.

First, digital values (DVs) of band 10 are converted to TOA spectral radiance
using the equation:

Ly = ML × Qcal + AL (4.1)

where Ly is TOA spectral radiance received by the sensor measured in mW cm-

2 sr-1 μm-1; ML is the band specific multiplicative rescaling factor for band 10
equal to 0.0003342; Qcal is the actual band 10 digital number (DN), and AL is
a band-specific additive rescaling factor equaling 0.1. This information was
obtained from the metadata file and USGS Landsat-8 product use
documentation (https://landsat.usgs.gov/using-usgs-landsat-8-product ).
The next step was to convert spectral radiance to satellite brightness
temperature (blackbody temperatures) by expressing Planck´s function. The
equation used was proposed by Markham and Barker (1986) :

Bt = [K2 / log(1+K1 /Ly)] – 273.15 (4.2)

where Bt is the satellite brightness temperature in Celsius, K1 and K2 are

calibration constants (774.89 and 1321.07, respectively) of Landsat-8
representing at-sensor spectral radiances of Landsat TM Band 10, and Ly is the
satellite spectral radiance retrieved from Equation 1. The brightness
temperature calculated with this second equation is called the top of the

63
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

atmospheric temperature (Zhang et al., 2006). The following additional

parameters were measured: :

1.Land surface emissivity

The brightness temperature derived from Equation 2 refers to a black body

target. Since the target bodies on land surfaces are not perfect black bodies,
the thermal emissivity of the land surface must be retrieved from the area
being studied (Gartland, 2012). Several factors such as plant chemical
composition and surface roughness affect the estimation of canopy surface
emissivity from remote sensing data (Snyder et al., 1998). Several methods
have been proposed to retrieve emissivity from remote sensing data, such as
image classification and the NDVI-based threshold approach. Image
classification based on Landsat imagery is not an appropriate method for the
estimation of land surface emissivity due to the spatial resolution, with one
pixel possibly comprising different land cover classes (Dozier and Warren,
1982). As a result, the NDVI-based threshold is a more appropriate approach
for the calculation of ground emissivity using Landsat imagery (Vlassova et al.,
2014). The NDVI-based approach was first proposed by Van de Griend and
Owe (1993) and was later modified by Valor and Caselles (1996). In this study,
we followed the approach by Valor and Caselles (1996) using NDVITHM to
calculate emissivity for the study area as follows:

ɛ = ɛv × Pv + (1 – Pv) + 4 < dɛ > Pv × (1 – Pv) (4.3)

where ɛ is ground emissivity; ɛv is emissivity for pure vegetation covered area;

dɛ is the constant value of topography factor (0.01), and Pv is the proportion
of vegetation, obtained from NDVI, based on the following equation proposed
by Carlson and Ripley (1997):

Pv = (NDVI – NDVIg) / (NDVIv - NDVIg) (4.4)

where Pv is the proportion of vegetation for each pixel; NDVIg is the NDVI
value in bare soil ( NDVI < 0.2); and NDVIv is the NDVI value for purely
vegetation-covered areas (NDVI > 0.5). Further details regarding emissivity
based NDVI estimation can be found in (Sobrino et al., 2008).

2. Estimation of atmospheric transmittance

The atmospheric transmittance was estimated using water vapour content

based on the formula by Qin et al. (2001):

Wc = 0.0981 × [10×0.6108× exp ((17.27×(T0))×RH]+0.1697 (4.5)

64
 Chapter 4

where Wc is the water vapour content in g/cm2, T0 is the near-surface or air

temperature in degrees Celsius and RH it is the relative humidity.
Both near-surface temperature and relative humidity (RH) data were obtained
from climate-station Waldhäuser located in the BFNP at 950 m above sea-level
on a south-west slope.
Based on the air surface temperature and humidity ratio data from the
Waldhäuser station, the water vapour content ranged between 1.6 and 3.0
g/cm2 with our data being captured in the summertime. Thus, the following
formula was used to estimate atmospheric transmittance:

Ƭ= 1.031412 – 0.11536 × Wc (4.6)

where Ƭ is the atmospheric transmittance, and Wc is the estimated water

vapour content obtained from the previous equation.

3. Mean atmospheric temperature

The final parameter required for the calculation of CST using the mono-window
algorithm is mean atmospheric temperature. The Landsat data were captured
in May, July and August, 2016., which meant summer conditions. Therefore,
the following formula was used to estimate the mean atmospheric temperature
(Qin et al., 2001; Sun et al., 2010) :

Ƭa = [16.0110 + 0.92621 × (T0)]-273.15 (4.7)

where Ta is the mean atmospheric temperature in degrees Celsius, and T0 is
the surface air temperature in Kelvin. Data on air temperature were obtained
from the Waldhäuser weather station.
Finally, the mono-window algorithm was employed to retrieve canopy surface
temperature over the study area, as follows:

CST = (A × (1 – C –D) + [B × (1 – C – D) + C+D] × Bt – D × Ta) / C (4.8)

where CST is the canopy surface temperature in degrees Celsius, Bt which is

the effective temperature viewed by the satellite under an assumption of unity
emissivity in degrees Celsius, and Ta is the mean atmospheric temperature in
degrees Celsius. In addition, the parameters A and B have standard values
proposed by Qin et al. (2001) which are -67.355351 and 0.458606,
respectively, while the parameters C and D are obtained using the following
formulae:

C=ɛ×Ƭ (4.9)

and

65
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

D= (1 – Ƭ ) × [1 + (1- ɛ) × Ƭ ] (4.10)

where ɛ is emissivity estimated from Eq.3, and Ƭ is atmospheric transmittance.

4.2.4 Statistical Analysis

To fulfil the objective of this research, three statistical analyses were employed,
namely, statistical significance test (Student t-tests), partial least squares
regression (PLSR), and principal components analysis (PCA).The Student t-
tests were used to assess the impact of bark beetle infestation at the green
attack stage on the measured leaf traits (foliar stomatal conductance,
chlorophyll fluorescence, and water content) and remote sensing data (SVIs
and CST). Next, to identify the key SVIs for estimating the studied leaf traits
(foliar stomatal conductance, chlorophyll fluorescence, and water content)
Variable Importance in Projection (VIP) was calculated from the PLSR analysis.
The VIP scores are useful in understanding X space predictor variables (SVIs
& CST) that best explain dependent (Y) variables (leaf traits). The ‘VIP method
selects those X variables that contribute most to the underlying variation in
the X variables. This includes the variation not related to y, but describing
interferences, that is so-called orthogonal variation’ (Farrés et al., 2015). To
achieve this, PLSR analysis was employed. The PLSR ‘is a regression method
that takes into account both the variance of the dependent and the predictor
variables. It specifies a linear relationship between a set of dependent (Y)
variables (leaf traits) and a set of predictor (X) variables’ (CST and SVIs) (Wold
et al., 2001). The PLSR was used for the Landsat-8 image captured in July
2016, as it matched the time of field data collection. Three independent PLSR
models were built for each leaf trait separately as well as for each sample group
(green attacked and healthy samples). In general, a VIP score > 1 unit
indicates that the contribution of the variable (CST or SVI) was significant, and
the greater the VIP value, the larger is the contribution of the variable to the
model (Chong and Jun, 2005).

Moreover, to investigate the potential of separating healthy plots from infested

ones in a 2-D space, the clustering method using PCA was employed. In other
words, PCA was used to determine the importance of each SVI for
discriminating between healthy and infested plots during all three months
considered in this study. PCA is an unsupervised technique and has been
successfully used in many remote sensing studies as a data reduction approach
(Maitra and Yan, 2008). The mean centring and unit variance producer were
used to preprocess the data. The SVIs (those having a VIP score higher than
one for both healthy and infested sample plots) were treated as independent
variables, and for each image date considered in this study, two PCA models
were built as follows: (a) a PCA model including all selected SVIs and CST (VIP
> 1); (b) a PCA model including only SVIs. This step is essential because it

66
 Chapter 4

allows us to understand and identify the most effective SVIs in terms of making
a significant contribution to discriminating healthy from freshly infested spruce
plots (Figure 4.3).

From the selected SVIs, those that have VIP > 1, a box plot technique was
used to explore the temporal variation between healthy and infested plots for
all three months considered in this study. The values of SVIs and CST were
normalized between 0 and 1 using the MinMax normalisation method.

4.3 Mapping bark beetle green attack infestation

To generate the map of bark infestation, we considered the SVI with the
highest VIP score as the most important variable to estimate studied leaf traits.
Linear regression was used to generate maps of the measured leaf traits (foliar
stomatal conductance, chlorophyll fluorescence, and water content), and the
generated maps of leaf traits were used as a proxy to produce a map of bark
beetle green attack. To do this, we classified the produced maps using the
range of the leaf traits from the field (Table 4.3). The final map contains three
classes corresponding to the stress intensity in each trait. Following this, to
validate the generated map of infestation, ancillary vector-based reference
data of the green attacked areas in 2016 were obtained from the Bavarian
Forest National Park (BFNP) administration (Abdullah et al., 2018b). The flight
on 11 Jun 2017 documented new dead wood (grey attack stage) from the
previous year 2016, with reference data collected from airborne colour-infrared
(VIS and NIR) aerial images (CIR) with 0.1 m spatial resolution. Full details
about the processing and interpretation of aerial photography in the BFNP can
be found (Heurich et al., 2009; Lausch et al., 2013a). The reference data
(ground truth) were in the form of polygons and were rasterised into 30 m 
30 m grid cells to match with the produced maps from Landsat-8 SVI. In total,
417 pixels were used as the ground truth data. Furthermore, land cover data
obtained from the national park administration were used to mask out non-
spruce stands and young stands. This is because bark beetles only infest old
and mature Norway spruce (Picea abies (L.) Karst). The masked land cover
was overlaid on the classified leaf traits maps in this study (Fig. 4.3). Finally,
the total number of pixels (ground truth) located within each (stress class)
were extracted and calculated.

67
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

Table 4. 3: Stress level categories classified using the range of the leaf traits (leaf water
content, stomatal conductance, and chlorophyll fluorescence ) from the field, in the
Bavarian Forest National Park, July 2016.
Severely Moderately Healthy
stressed stressed
Leaf water content (g/cm )2
< 0.0135 0.135 – 0.0145 > 0.0145
Stomatal conductance < 103 103 - 120 > 120
(mmol/m2s)
Chlorophyll fluorescence ratio < 1.17 1.17 – 1.25 > 1.25

Figure 4. 3: The framework of the research methodology.

4.3 Results
4.3.1 The importance of CST versus SVIs to estimate
measured leaf properties
Infestation by bark beetle caused substantial changes in all studied leaf traits
during the green attack stage. The result of the Student’s t-test demonstrated
a significant difference (p < 0.05) between healthy and infested leaves for all
the studied leaf traits (chlorophyll fluorescence, leaf water content, and
stomatal conductance) (Fig. 4.4). Furthermore, as can be seen from figure 4.5,
the VIP scores show that CST significantly contributes to the estimation of
studied leaf traits (chlorophyll fluorescence, water content, and stomatal
conductance) for both healthy and infested plots. However, among the SVIs
only 5 out of 23 had a VIP score above one for both healthy and infested sample
plots; these five indices are CIgreen, CVI, CI, NLI, and BGI (Fig. 4.5). The
single bands of Landsat-8 showed the lowest VIP score for all measured leaf
traits.

68
 Chapter 4

Figure 4. 4: Distribution of measured leaf traits for healthy and infested needles, in the
Bavarian Forest National Park, July 2016. There is a significant difference (p < 0.05) in
all measured leaf traits between healthy and infested samples.

Figure 4. 5: PLSR variable importance in the projection (VIP) for predictor (CST and
SVIs) and response variables (measured leaf traits ) in the Bavarian Forest National Park,
July 2016. (vertical black line represents VIP ≥ 1)

4.3.2 Temporal response of CST and SVIs under spruces bark

beetle infestation
To assess the temporal variation of the CST and those SVIs that obtained a
VIP score > 1, we studied the boxplots for all three months considered in this
study. Figure 4.6 shows a comparative time-series of CST and SVIs for healthy
and infested sample plots. As can be seen from this figure, the CST in infested
and healthy plots differ from the SVIs by exhibiting increased temperature from
the beginning of the infestation in May, whereas almost all other indices show

69
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

no differences between healthy and infested plots. Similarly, in July, when the
infestation is still at its early stage (green attack), CST shows high potential
for discriminating between healthy and infested sample plots. However, in
August, when the infested plots entered into the advanced stage (in which the
infested trees developed stress symptoms by turning their needles’ colour from
green to yellow to red-brown and can be detected by the human eye), CST as
well as CIgreen, NLI, and BGI are significantly different (p < 0.05) between
healthy and infested plots.

The selected SVIs, which had a VIP value > 1 (CST, CIgreen, CVI, CI, NLI, and
BGI) have been used to build the PCA model. As shown in Figure 4.7, in May,
when the infestation is in its early stage, there is an apparent overlap and
mixed scattering between healthy and infested sample plots when SVIs were
applied in the model. However, in July and August, when CST is included into
the model, the majority of healthy plots tend to be on the negative side of PC1,
while the infested plots are relatively scattered on the positive side of PC1. In
general, we observed distinct improvement to differentiate between healthy
and infested plots when both CST and SVIs were used to build the PCA model.

70
 Chapter 4

May July August

Figure 4. 6: Temporal variation of the SVIs and CST (VIP >1 ) for healthy and green
attacked sample plots in the Bavarian Forest National Park. Blue and red boxes represent
healthy and infested plots, respectively. A * indicates the significant different

71
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

A B

Figure 4. 7: Cluster plots based on the first two principal components (PCs), blue squares
and red circles represent healthy and infested plots, respectively. (A) Cluster plots based
on all spectral indices VIP >1 including CST (B) Cluster plots based all spectral indices
VIP >1 excluding CST.

4.3.3 Mapping bark beetle green attack and validation

As shown in Figure 4.5, CST recorded the highest VIP score in all measured
leaf traits, and therefore we considered it as the most important variable for
predicting studied leaf traits. The CST was then used to estimate leaf traits
(foliar stomatal conductance, chlorophyll fluorescence, and water content) and
generate maps using the SLR model (Table 4.4). Following this, based on the

72
 Chapter 4

defined threshold value (Table 4.3), stress maps were generated using leaf
traits’ data. These maps were then overlaid with ground truth reference data
(417 pixels) to calculate those pixels located within each stress class. Figure
4.8 and Table 5 depict those areas (pixels) that were located in each class. As
can be seen from Table 5, 274 pixels out of 417 (66%) from ground truth data
were located with the severely stressed class; following this, 89 (21%) pixels
were located in the moderately stressed class and the remaining 54 (13%)
pixels where located within the healthy class .

Table 4. 4: Regression equations between canopy surface temperature (CST) and

measured leaf properties.

Equation No. Regression equations

1 Leaf water content = -0.0004 CST + 0.0203

2 Stomatal conductance = -6.4377 CST + 255.66
3 Chlorophyll fluorescence = -0.0264 CST + 1.68

Table 4. 5: Assessment of the generated map using reference data obtained from Aerial
photography.
Forest stress classes Pixels correctly matched
Severely stressed 274
Moderately stressed 89
Healthy 54
Reference pixels (aerial photography) = 417 pixels (30m).

73
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

Figure 4. 8: The map showing the distribution of Norway spruce stands under different
stress levels in the Bavarian Forest National Park, July 2016. The map was produced
using canopy surface temperature and measured leaf traits (foliar stomatal conductance,
chlorophyll fluorescence, and water content)

4.4 Discussion
In the present study, for the first time, we assessed the potential of CST and
SVIs to discriminate between healthy and green-attacked trees by European
spruce bark beetle (Ips typographus, L.). The results confirmed the superiority
of CST in discriminating subtle differences between healthy and infested trees,
compared to other utilised SVIs at different stages of bark beetle attack.
Furthermore, all three considered leaf traits in this study exhibited significant
differences (p < 0.05) between the two sample groups (healthy and infested),
and CST was found to be an essential indicator (VIP > 1) to estimate them.

Of all 23 remote sensing variables (CST, SVIs, and single bands) used, the
variable importance in projection (VIP > 1) showed that CST was by far the
best at estimating studied leaf traits (Fig. 4.5). This result reaffirms findings
by (Berni et al., 2009), who used TIR data obtained fixed-wing UAVs to study
thermal water stress over the peach trees . Moreover, apart from the
importance of CST in estimating leaf properties, we found that CST better
discriminates between infested and healthy trees than other SVIs. In July and
August, CST was significantly higher (p < 0.05) for the infested plots than for
healthy ones (Fig. 4.6).. For example, in May, just at the beginning of the
infestation, a slightly higher CST was observed for the infested plots, whereas

74
 Chapter 4

the SVIs showed no difference between the two groups. According to bark
beetle phenology studies, the flight activity of bark beetles and their successful
attack on living trees starts in late spring when the air temperature reaches
16.5ºC (Lobinger, 1994b; Wermelinger, 2004). The new insight offered by this
result is the potential of CST to detect the subtle changes induced by bark
beetles at this stage is quite important and promising. Likewise, in July, when
the infestation had progressed, and the infested trees were completely
stressed, CST maintained its sensitivity in differentiating between healthy and
green-attacked sample plots (Fig. 4.6). It is well-known that physiological
factors, such as plant water content, can control the temperature of plants
through the stomatal transpiration process (Oerke et al., 2006). Therefore, the
distinct variation in CST between healthy and infested plots is attributed to
differences in the measured leaf traits, in particular, leaf water content. As
depicted in Figure 4.4, the amount of water content was significantly lower (p
< 0.05) in the infested plots than in the healthy ones. This may be attributed
to the spores of fungi and the drilling activities of the beetle itself that affect
the living cells in the phloem and xylem and, therefore, disrupt the flow of
water and cause stomatal closure in infested trees (Yamaoka et al., 1990). This
result further confirms our earlier finding (Abdullah et al., 2018a) showing a
decrease of foliar biochemical properties (chlorophyll and nitrogen
concentration) in trees infested by bark beetles (I. typographus, L.) at the
green attack stage. Likewise, our findings are in agreement with (Cheng et al.,
2010), who studied the impact of a similar species (the mountain pine beetle,
Dendroctonus spp) on leaf water content using hyperspectral data. Infested
trees tend to close their stomata to prevent further water loss and hydraulic
failure (Ewers et al., 2007). This stomatal closure will temporarily offset xylem
cavitation and cause damage to the photosynthetic apparatus (Chaves et al.,
2003). However, this behaviour leads to other physiological changes in the
infested trees, such as a decrease in the transpiration cooling process and,
therefore, an increase in leaf surface temperature. Over time, this will affect
the process of photosynthesis and cause a change in foliar colour (Larcher,
2003). These findings match those observed in earlier studies that the TCI
index retrieved from the Landsat 7 ETM thermal band could detect areas under
mountain pine beetle (Dendroctonus spp) green attack (Sprintsin et al., 2011).

In August, when the infestation had progressed and the subsequent

degradation of the needles could be observed in the field (the infested trees
developed stress symptoms by turning their needles’ colour from green to
yellow to red-brown), the variation of SVIs became more evident between
healthy and infested plots. CST, Cigreen, NLI and BGI were shown to have the
potential to detect the infested trees at that time. However, for effective
management, identifying infested trees at that stage is too late, because the
beetles will have already left the infested trees, and logging operations will not

75
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

have any effect on the population dynamics of the beetles at that time anymore
(Fahse and Heurich, 2011; Wermelinger, 2004; Wulder et al., 2009).

In addition to the temporal variation, we also observed that CST improved

discrimination between healthy and infested plots when using PCA analysis for
all dates considered in this study (Fig. 4.7). In May, when the PCA model was
built based on optical data only, the two groups of sample plots had mixed
scattering and exhibited an apparent overlap, whereas, when CST was included
into the model the healthy plots were separated from the infested ones. This
indicates the increased sensitivity of Landsat-8 TIR data to changes at canopy
level induced by bark beetle at a very early stage in May. One possible reason
for such sensitivity of CST to spruce bark beetle infestation in May could be
due to the nature of spruce bark beetle outbreaks, which generally occur over
the course of several years. In other words, the healthy trees are always
surrounded by other healthy trees, while, for infested trees in most cases, the
infestation occurs close to the previous year’s infestation (Lausch et al., 2011).
Therefore, the open areas and dead trees around newly infested trees can lead
to an increase in surface temperature. To confirm this, we further analysed
and calculated the CST for different habitat types in the landscape. To do this,
a habitat map obtained from the national park administration was used to
extract the CST for each habitat type (Dupke et al., 2017). As can be observed
from Figure 9, the results revealed that clear-cut areas, followed by lying dead-
wood and standing dead-wood areas, had the highest CST for all three months.
In contrast, mixed stands showed the lowest CST. This is supported by
Peterson et al. (1986) who found that forest structure (i.e. canopy closure) is
negatively correlated with simulated thermal data from the Thematic Mapper
Simulator (TMS).

Similarly, Junttila et al. (2016) and Hais and Kučera (2008) have shown that
higher surface temperatures were detected in dead-wood stands and clear-cut
areas, respectively. Furthermore, a water body had the lowest CST, which
supports the hypothesis that more moisture in healthy tree needles decreased
their surface temperature. This also corresponded with our in-situ
measurements that showed green-attacked trees had lower water content and
more closed stomata than healthy trees (Fig. 4.4). The clear-cut area and
dead-wood stands had no foliage and, thus, no leaf surface area for
evapotranspiration. This resulted in higher canopy temperatures in the cleared
and dead-wood stands than in intact stands.

76
 Chapter 4

Figure 4. 9: Mean canopy surface temperature over different habitat classes in the
Bavarian Forest National Park – 2016. The graph shows that standing dead-wood, lying
dead-wood, clear-cut and rocky areas recorded high surface temperature comparing to
other forest classes.

Moreover, an equally significant aspect of CST was found when used to

estimate measured leaf traits (foliar stomatal conductance, chlorophyll
fluorescence, and water content) and producing the stress map for the study
area (Fig. 4.8). The stress map highlighted the sensitivity of CST for detecting
canopies that are stressed by bark beetle green attack. In general, the majority
(66%) of ground truth data were located within the severely stressed class —
however, 13% where falsely located in the healthy class (Table 4.5).
Consequently, CST derived from Landsat-8 TIR data might be used to generate
hotspot maps of stressed areas within the forest that show high potential areas
of bark beetle green attack. Such maps can provide valuable information to
the forest management practice when they aim to control this species and
preclude a mass outbreak. Furthermore, such maps can also be used to
improve the bark beetle modelling community. For example, in our study and
the previous studies by (Hais and Kučera, 2008) the highest temperature was
observed over the dead-wood stands. Such information on the dead-wood
stands combined with the less reliable information from green attacked areas
can be used to improve the accuracy of predisposition models (Netherer and
Nopp-Mayr, 2005)

4.5 Conclusion
The results of this study showed the potential of CST retrieved from the
Landsat 8 TIR band to detect early-stage bark beetle infestation. We further
found that the infestation at the green attack stage affected the leaf stomatal
conductance, chlorophyll fluorescence, and water content. The CST has the
highest VIP value for estimating all three leaf traits measured in this study for
July. In addition, we found that CST maintained its sensitivity for monitoring
and detecting bark beetle infestation before, during and after infestation. In
conclusion, our key concept is that we show for the first time that satellite TIR

77
Optical and thermal infrared data for early detection of (Ips typographus, L.) infestation

data has a high potential to detect bark beetle green attack and examining the
link between the leaf traits and thermal remote sensing data is an important
step that further improves our understanding of the relationship between Earth
Observation data and plant traits for forest areas under bark beetle green
attack. It is important to note that TIR images obtained from Landsat-8 are
the only available satellite images that provide data at 30 m spatial resolution.
However, the original data were acquired at 100 m, so the actual footprint of
the pixel is larger. Therefore, further research investigating different TIR data
with higher spectral and spatial resolution, such as TIR from airborne
hyperspectral measurements, may improve the use of thermal imagery for
bark beetle green attack detection in Norway spruce forests.

78
Chapter 5: Sentinel-2 accurately maps green
attack stage of European spruce bark beetle
(Ips typographus, L.) compared to Landsat-8 


This chapter is based on:

Abdullah, H., Darvishzadeh, R., Skidmore, A.K., and Heurich, M. Sentinel-2 accurately
maps green attack stage of European spruce bark beetle (Ips typographus, L.) compared
to Landsat-8. Remote Sensing in Ecology and Conservation. 2018, 1–21.

79
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

Abstract
Natural disturbances induced by insect outbreaks have increased in forest
ecosystems over the past decades. To minimise economic loss and prevent a
mass outbreak, early detection of bark beetle green attack – a period when
trees have yet to show visual signs of infestation stress – is therefore crucial
to effective and timely forest management. In this study, we evaluated the
ability of spectral vegetation indices extracted from Landsat-8 and Sentinel-2
imagery to map bark beetle green attack using principal component analysis
(PCA) and partial least square discriminate analysis (PLS-DA). A recent
infestation map produced through visual interpretation of high-resolution aerial
photographs validated the final infestation output maps. Leaf spectral
measurements alongside total chlorophyll and nitrogen concentration, leaf
water content, and leaf dry matter content were measured to assess the impact
of bark beetle green attack on foliar properties. We observed that the majority
of Spectral Vegetation Indices (SVIs) calculated from Sentinel-2, particularly
red-edge dependent indices (NDRE 2 and 3) and water-related indices (SR-
SWIR, NDWI, DSWI, and LWCI), were able to discriminate healthy from
infested plots. In contrast, only the water-related indices (NDWI, DSWI, and
RDI) from Landsat-8 were able to discriminate between healthy and infested
plots efficiently. The total number of pixels identified as harbouring a green
attack that matched with ground truth data (aerial photography) was higher
for Sentinel-2 (67%) than for Landsat-8 (36%) SVIs, indicating the elevated
sensitivity of Sentinel-2 imagery to changes induced by bark beetle green
attack. We also determined that foliar chlorophyll and leaf water content were
significantly higher (p < 0.05) in healthy trees than in green attacked trees.
Our study highlights the potential of Sentinel-2 data for the early detection of
bark beetle infestations and the production of reliable infestation maps at the
green attack stage.

80
 Chapter 5

5.1 Introduction
Insect infestations, as by the European spruce bark beetle (Ips typographus
L.), form the main disturbance events in European forests as they destroy more
forested areas than all other natural disturbances together. Historically, storms
and snow breakage events in Europe and North America have provided a large
surplus of suitable breeding material for bark beetles, leading to outbreaks
(Giunta et al., 2016; Seidl et al., 2017a). Recent outbreaks are different,
however, with climate change (more drought and severe storm events)
appearing to be the initiating factor (Marini et al., 2016; Netherer et al., 2015;
Thom and Seidl, 2016). Increasing temperatures may result in an increased
incidence of drought, possibly affecting tree health. Drought facilitates beetle
outbreaks by stressing trees and increasing the frequency and severity of bark
beetle outbreaks (Bentz et al., 2010; Faccoli and Bernardinelli, 2014; Filchev,
2012b; Thom and Seidl, 2016). A warmer climate may increase storm
frequencies, and severity of wind throws, thus providing more breeding
material for bark beetles (Marini et al., 2016). Furthermore, as the
development of the bark beetle is dependent on temperature, an increase in
temperature may thus lead to an increase in the beetle population size, due to
their eggs hatching and developing into adults sooner (Bentz et al., 2010;
Wermelinger, 2004). In recent decades, therefore, increasing disturbances due
to insect outbreaks have been widely documented across different parts of the
world (Lausch et al., 2013a; Seidl et al., 2014; Wulder et al., 2006a).

Timely, accurate and cost-effective information is needed to mitigate and

control bark beetle outbreaks, to guide forest managers in identifying areas
infested by beetles, as well as to define the timing of bark beetle control
activities (Wulder et al., 2009). To obtain this information, the operation survey
needs to take place at the time when the infestation at its early stage (the
green attack). The green attack stage is the first interaction between the beetle
and the host tree, and occurs when the host is being colonized by the bark
beetle. During this stage, the infested tree is still physiologically green and
very much alive, although exhibiting stress in the near infrared (invisible to the
human eye) (Niemann and Visintini, 2005). Furthermore, during the green
infestation, the newly hatched generation of beetles is developing within the
inner bark of the infested trees. Therefore, management intervention to
prevent further outbreaks may involve the removal of infested trees before the
new brood emerges and migrates (Wermelinger, 2004; Wulder et al., 2009).

Traditionally, foresters perform field surveys to identify infested trees during

the early green attack stage using conventional survey methods (looking for
sawdust). Such surveys are very laborious and therefore make screening large
areas for green attack difficult. Nowadays, remote sensing provides new
opportunities to detect and map bark beetle infestation. Remotely sensed data

81
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

rely on spectral signatures from different regions in the electromagnetic

spectrum. Unique spectral signatures have been linked to different functional
and structural plant traits, such as pigments at 400–700 nm, leaf structure at
700-1100 nm and plant water content, nitrogen concentration, Leaf Area Index
and Specific Leaf Area at 1100 to 2400 nm (Ali et al., 2016; Gitelson et al.,
2002; Mahlein et al., 2013; Viña et al., 2011; Wang et al., 2015a). Stress
affects plant biophysical and biochemical properties and therefore affects
spectral signature. For example, chlorophyll degradation and nitrogen
deficiency lead to an increase in reflectance spectra in the visible region (the
red and green bands in particular). As a result, this wavelength region has
been widely used as a stress indicator when utilising remote sensing data
(Hendry et al., 1987). Significantly, the reflectance at the red portion of the
visible region has been shown to be less sensitive to initial loss of chlorophyll
content (Carter, 1993). This is due to the high spectral absorption by
chlorophyll in this spectral region, which saturates the red reflectance at low
chlorophyll content (Jacquemoud and Baret, 1990). In contrast, the reflectance
at the green and red-edge wavelengths (centred in 550 nm and 700 nm,
respectively) of the VIS region are more sensitive to changes in plant
chlorophyll content (Eitel et al., 2011). Furthermore, the red-edge region has
superior sensitivity when detecting changes that are induced in plants by
stressors such as dehydration, disease, and insect attack, and can, therefore,
improve the early detection of plant stress (Ahern, 1988b; Carter and Knapp,
2001; Carter and Miller, 1994; Eitel et al., 2011). Moreover, the NIR and SWIR
regions have been used widely to assess water content and nitrogen
concentration in plants (Ayala-Silva and Beyl, 2005; Carol et al., 2004;
Jackson, 2004; Munoz-Huerta et al., 2013).

Several studies have utilised spectral vegetation indices (SVIs) from low-to-
medium resolution satellite data to study bark beetle infestation. These studies
have mainly focused on the last two infestation stages (red and grey) of attacks
(Filchev, 2012b; Franklin et al., 2003; Hais et al., 2009; Havašová et al., 2015;
Koch et al., 2010; Meddens et al., 2013; Wulder et al., 2006b). However, for
effective and proper forest management, the detection of infestation should be
early enough to allow for timely intervention to minimise the outbreak. Several
studies have explored the use of commercial remote sensing data such as
Worldview-2 (Filchev, 2012b; Immitzer and Atzberger, 2014), RapidEye (Marx
and an der Havel, 2010b; Ortiz et al., 2013) and HyMAP airborne hyperspectral
data (Lausch et al., 2013b) for the early detection of bark beetle attack, but
with very limited success.

The SVIs were introduced to improve interpretation of vegetation signals when

using remote-sensing data and can be used to measure vegetation status while
minimising solar irradiance and soil background effects (Darvishzadeh et al.,
2009; Jackson and Huete, 1991; Moulin, 1999). Additionally, the combination

82
 Chapter 5

of different spectral bands is closely related to biophysical and biochemical

properties of foliage vigour associated with plant health and can be used to
detect morphological and physiological changes caused by insect outbreaks in
the forest canopy (Rullan-Silva et al., 2013). Therefore, SVIs may be expected
to perform better than individual spectral bands when it comes to detecting
stress induced by the bark beetle. This is especially valid when using low-to-
medium resolution data that include the red-edge spectral domain (Jackson
and Huete, 1991; Zhang et al., 2012). This study evaluates the ability of
different SVIs from Sentinel-2 and Landsat-8 imagery to detect and help map
bark beetle infestation at the green attack stage. Furthermore, we also study
the impact of bark beetle green attack on foliar biochemical and biophysical
properties and their spectral reflectance using foliar spectral data collected
from ASD Fieldspec3.

5.2 Material and Methods

5.2.1 Study area and field data collection
The study area is the Bavarian Forest National Park, which is a 24,369 ha forest
located in south-eastern Germany along the broader with the Czech Republic,
between 13°12'9" E (longitude) and 49°3'19" N (latitude). This region is
characterised as having a temperate climate with an annual precipitation of
between 900 and 1800 mm, and a mean annual temperature of between 3.5°
and 9°C (Bässler et al., 2008b). The forest is dominated by Norway spruce
(Picea abies) (67%) and European beech (Fagus sylvatica) (24.5%) (Cailleret
et al., 2014a). Outbreaks of the bark beetle (Ips typographus, L.) began in
1984 and have caused extensive disturbance to this forest. As such, the area
is a suitable study site for research on bark beetle infestations and outbreaks
(Heurich et al., 2010a).

During June and the beginning of July 2016, an extensive campaign was
conducted to collect field measurements. The study area was divided into two
different strata based on their tree condition: stands with healthy trees and
stands with trees freshly infested by the bark beetle. To select the healthy
stands, a stratified random sampling strategy was adopted. While, for the
recently infested stands, the presence of bark beetle green attack had to be
confirmed in an intensive field survey by searching for piles of dry, boring dust
pushed out onto the bark surface of the tree when the beetle tunnels under
the bark. According to the Bavarian Forest National Park authorities and our
field survey, the infestation of bark beetles in 2016 clearly indicated outbreak
conditions.

To avoid mixed reflectance from healthy and green attacked trees, only plots
fully under bark beetle green attack that covered an area of 30 m × 30 m were

83
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

selected. In total, 40 and 21 plots were selected in healthy and infested

stands, respectively (Fig. 5.1). The centre of each plot was measured using
Differential Global Positioning System (DGPS) Leica GPS 1200 (Leica
Geosystems AG, Heerbrugg, Switzerland) with an accuracy of better than 1m
after post-processing. Each plot was designed as a 30 × 30 m square and
within the plot stand characteristics including DBH, canopy cover, height and
tree density were measured. At each plot, three to five trees were selected as
representative and needle samples from each of these trees were collected
separately. All the samples were taken from the trees’ top layer exposed to the
sunlight. A crossbow was used to shoot an arrow attached to a fishing line at
a branch with sunlit leaves (Ali et al., 2016).

Leaf spectral reflectance and traits including total chlorophyll, leaf water
content (Cw), leaf dry matter content (LDMC), and foliar nitrogen
concentration were measured for healthy and infested samples (Table 5.1). In
the field, the concentration of chlorophyll was determined using a CCM
chlorophyll meter. On average ten readings were immediately taken from each
fallen branch with the CCM. The needles were then removed from the fallen
branches, covered with wet pulp paper and placed in a labelled plastic zip-
locked bag. They were transported to the laboratory using a portable cooling
box; this was done to retard possible changes in the needles’ reflectance
spectra and biochemical characteristics. The leaf directional hemispherical
reflectance from 350 to 2500 nm was measured for the collected samples,
using an ASD FieldSepc-3 Pro FR spectrometer equipped with an ASD RT3-3ZC
integrating sphere (Analytical Spectral Devices, Inc., Boulder, Co, USA).
Further details concerning the measurements of hemispherical needle
reflectance can be found in Abdullah et al. (2018), Ali et al. (2016), Daughtry
et al. (1989), and Malenovský et al. (2006). The needle samples were then
dried for 72 h using an oven at 60° C, until they reached constant weight, to
calculate leaf dry matter and water content (see Table 5.1). To determine foliar
nitrogen, the dried needles were ground using mortar and pestle until they
became a soft powder, after which they were passed through a 0.25 mm mesh
screen. Subsequently, 2 mg of powdered leaves were transferred to a small
aluminium capsule to measure the nitrogen content using an organic elemental
analyser (FLASH 2000). The studied leaf traits (chlorophyll, water content, dry
matter content, and nitrogen concentration) from the measured representative
trees in each plot were then averaged to obtain the leaf traits at the plot level,
hereafter referred to as plot level parameters.

84
 Chapter 5

Table 5. 1: The leaf traits measured for healthy and infested samples in this study. Fw,
Dw and A represent fresh leaf weight (g), dry leaf weight (g), leaf dry mass per unit area
(Cm), and leaf area (cm2 ), respectively.

Leaf traits Equation unit Reference

(Danson et al.,
Leaf water per
(Fw-Dw)/A Mg/cm2 1992, Ceccato
mass area
et al., 2001)
Leaf dry matter (Vile et
Cm / (Cm-Cw) Mg/g
content (LDMC) al.,2005)
PerkinElmer
Nitrogen ----------- %
2400 CHN/O
chlorophyll
Chlorophyll ------------ Mg/m2
content meter

Figure 5. 1: Location of healthy and green attacked sample plots in the Bavarian Forest
National Park in July 2016.

5.2.2 Satellite imagery

Open access multispectral satellite imagery from Sentinel-2 and Landsat-8 was
selected for this study. To fulfil the aims of this research, cloud-free satellite
data were acquired on 8 and 10 July 2016 from Sentinel-2 and Landsat-8,
respectively. Landsat-8 data were obtained from USGS Global Visualization
Viewer (http://glovis.usgs.gov/), and the Sentinel-2 data were obtained from
the ESA Scientific Hub (https://scihub.copernicus.eu). Both images were
geometrically corrected in the Universal Transform Mercator (UTM) coordinate
system and matched each other with sub-pixel accuracy. Sentinel-2 delivers
high spatial (10 to 20 m) and spectral (13 bands) image data. Also, it is the
first freely available optical satellite providing three spectral bands in the red-
edge spectral region.

85
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

To prepare the satellite images for further analysis, a series of pre-processing

was applied. First, for Landsat-8, image radiometric calibration was applied to
convert the pixel values to Top-of-Atmosphere (TOA) reflectance. A
MODTRAN4-based atmospheric correction software package (FLAASH)
developed by the Air Force Phillips Laboratory, Hanscom AFB, and Spectral
Sciences, Inc. was used to convert the TOA Reflectance to surface reflectance
(Adler-Golden et al., 1999) using ENVI software. For Sentinel-2, the TOA
reflectance data were corrected to surface reflectance using the SEN2COR
atmospheric correction software developed by ESA
(http://step.esa.int/main/third-party-plugins-2/sen2cor/). Secondly, the
bands with a similar spatial resolution were stacked. In general, six bands from
Landsat-8 were used and stacked together, namely the bands 2,3,4,5,6 and 7,
which have a 30 m spatial resolution. From the Sentinel-2, the bands 2,3,4
and 8, with a 10 m spectral resolution, were first resampled to 20 m and then
stacked with bands 5,6,7,9, 12 and 13. The three bands with a 60 m spatial
resolution (1,10,11) are mainly relevant for atmospheric corrections and were
not used in this study. Finally, the spectral reflectance values of the sample
plots were extracted from the Landsat-8 and Sentinel-2 scenes and used for
further analysis. To synthesise and compare leaf reflectance data with canopy
reflectance data (from Sentinel-2 and Landsat-8 data), the reflectance spectra
collected from ASD FieldSepc-3 Pro FR spectrometer equipped with an ASD
RT3-3ZC integrating sphere were simulated by convolving to the spectral
resolution of Sentinel-2 and Landsat–8 using the linear interpolation method.

Figure 5. 2:The flow chart of the methodology used in this study

86
 Chapter 5

5.2.3 Spectral vegetation indices calculation

Spectral reflectance data from Sentinel-2 and Landsat-8 images were used to
calculate the most widely used vegetation indices to detect changes in plant
photosynthetic activity and biochemical stress, among other forms of
vegetation stress (Collins and Woodcock, 1996; Eitel et al., 2006). Spectral
vegetation indices include optical vegetation canopy (greenness), which is a
combined property of foliar biochemical (chlorophyll, nitrogen, leaf water
content) and other canopy properties (Jiang et al., 2008). To prevent using
multiple copies of similar band combinations such as (red and NIR) in this
study, we attempted to use statistically-independent spectral indices. In
addition, we employed the raw spectral bands as independent indices and only
those bands that were significantly different (p ≤ 0.5) between healthy and
infested sample plots (Table 5.2). The selected SVIs are sensitive to stress-
induced variations in chlorophyll content (VIS), biomass (NIR), and water
content (SWIR). Further, depending on their use, we categorised the
vegetation indices into three groups: (a) chlorophyll and other pigments, (b)
indices used for detection of water stress, and (c) the raw spectral bands from
Sentinel-2 and Landsat-8. ENVI software was used to compute spectral
vegetation indices. The equations for computing the vegetation indices are
provided in (Table 5.2).

5.2.4 Statistical analysis

Three statistical analyses were employed in this study: one-way ANOVA,
principal component analysis (PCA), and partial least square discriminant
analysis (PLS-DA). ANOVA tests were performed to: (a) ascertain the effect of
bark beetle green attack on measured leaf traits, including leaf and canopy
reflectance data, and (b) to evaluate whether SVIs’ between the two sample
groups (healthy and infested) were significantly different. Moreover, to visually
examine how well the two sample plots (healthy and infested) are separated
in 2-D space based on principal component scores, the clustering method using
principal component analysis is employed in this study. As such, PCA was used
to evaluate the potential for different spectral vegetation indices to
differentiate between healthy and infested sample plots. PCA is an
unsupervised technique and remains a popular method used to reduce the
dimension of multivariate data sets and to extract features (Hotelling, 1933;
Pearson, 1901). The unit variance and mean centring producer were used to
pre-process the data. To build the PCA model, spectral vegetation indices were
treated as independent variables, and the SVIs data of all 60 sample plots (21
infested and 40 healthy plots) were analysed. PCA models were built
independently for each SVIs group as follows: (a) including all SVIs (pigment,
water, and raw bands); (b) including only pigment-related indices; and (c)
including only water-related indices. This step is important because it allows

87
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

us to understand and identify the most effective spectral indices group for
separating the healthy from the infested sample groups (Fig. 5.2).

In order to identify the key SVIs influencing the spectral separability between
the healthy and infested plots, and to map bark beetle green attack, a novel
method known as a Random Frog (RF) was used. The RF is a state-of-the-art
variable selection algorithm, and it is a computationally efficient method using
the context of the reversible jump Markov chain Monte Carlo (MCMC) technique
(Green, 1995). It performs a search in the model space via both fixed-
dimensional and trans-dimensional moves between different models. After a
pseudo-MCMC chain is calculated, this can be used to calculate a selection
probability (SP) value for each variable included in the model. The key variables
can be identified regarding the ranking of all variables based on the SP value.
A detailed description of the calculation of RF can be found in Yun et al. (2013).
To achieve this, partial least squares-discriminant analysis (PLS-DA) was
employed as a modelling method in RF (Li et al., 2012). PLS-DA is a
classification technique allowing for the identification of variables that improve
the separation or classification between different groups (Wold et al., 2001).

In our study, SVIs with an SP value > 0.50 were selected as an important
variable to map bark beetle green attack. Following this, the selected SVIs
from both satellites (Sentinel-2 and Landsat-8) were used to map bark beetle
green attack. The box plot technique was used to display the distribution of
SVI values for healthy and infested plots. The threshold values for green-
attacked pixels were identified for each selected SVI. The criteria for selecting
the threshold values were based on the area (value) of each SVI, which
essentially characterised the difference (no overlap) between healthy areas
and those infested by bark beetle green attack (Fig. 5.9). Consequently, a
conditional decision was made using the identified threshold values from each
SVI to identify those pixels in the images falling within the identified threshold
range (assigned as 1) and those pixels falling outside of the threshold range
(assigned as 0). For example, the following conditional decision was utilised to
extract the threshold value from NDRE-2:
1, 𝑖𝑓 0.45 𝑁𝐷𝑅𝐸2 0.80
𝐺𝑟𝑒𝑒𝑛 𝑎𝑡𝑡𝑎𝑐𝑘𝑒𝑑 𝑝𝑖𝑥𝑒𝑙𝑠 5.1
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
From the conditional statement the following threshold values are extracted:
if the value of NDRE-2 is higher than 0.45 and less than 0.80, a 1 (green
attack) will be assigned to that cell location on the output raster; otherwise,
a 0 (false) will be assigned on the output raster. Similar conditions were applied
to the other selected SVIs (Figure 8)
Finally, to generate the final infestation map, the following equation was
applied to sum the infestation maps generated from each SVIs:
1, 𝑖𝑓 𝑖𝑛𝑑𝑖𝑐𝑒𝑠 𝐴 𝑎𝑛𝑑 𝐵 𝑎𝑛𝑑 𝐶 … 1
𝐹𝑖𝑛𝑎𝑙 𝑖𝑛𝑓𝑒𝑠𝑡𝑎𝑡𝑖𝑜𝑛 𝑚𝑎𝑝 5.2
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

88
 Chapter 5

All of the above statistical analyses were carried out using MATLAB 2016b
(MathWorks Inc, Natick, USA) and ArcMap (v.10.3).

5.2.5 Ancillary data and accuracy assessment

To assess how successfully the infestation maps produced in this study
matched with the existing infestation areas, reference data corresponding to
the locations of infested trees obtained from the Bavarian Forest National Park
administration were used for validation of the output infestation maps. The
flight on 11 Jun 2017 documented new dead wood (grey attack stage) from
the previous year 2016, with reference data collected from airborne colour-
infrared (VIS and NIR) aerial images (CIR) with 0.1 m spatial resolution. Full
details about the processing and interpretation of the aerial photography in the
Bavarian Forest National Park can be found (Heurich et al., 2009; Lausch et
al., 2013a).

The infestation data were in the form of polygons and were rasterised twice;
firstly into 20 m × 20 m grid cells to match the generated map of infestation
from Sentinel-2 SVIs, and secondly into 30 m × 30 m grid cells to match the
infestation map generated from Landsat-8 SVIs (Fig. 5.2). Since bark beetles
only infest old and mature Norway spruce (Picea abies (L.) Karst), the non-
spruce stands and young stands were masked using land cover data obtained
from the national park administration (Dupke et al., 2017). The masked land
cover was overlaid on the infestation maps generated in this study and
compared with the rasterised reference data. Finally, the total number of pixels
that correctly matched with the reference pixels (ground truth) were extracted
and calculated (Fig. 5.2).

89
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

Table 5. 2: Spectral vegetation indices applied to leaf reflectance measurements,

Sentinel-2 and Landsat-8 in the study area.
Sensor
Index Landsat- Sentinel- Equation Reference
8 2
(chlorophyll, pigments, greenness)
𝑁𝐼𝑅 𝑅𝑒𝑑𝑒𝑑𝑔𝑒1
Canopy Chlorophyll 𝑁𝐼𝑅 𝑅𝑒𝑑𝑒𝑑𝑔𝑒1
(Barnes et
Content Index × √ ------------------------
al., 2000)
(CCCI) 𝑁𝐼𝑅 𝑅𝑒𝑑
𝑁𝐼𝑅 𝑅𝑒𝑑

Chlorophyll Green 𝑅𝑒𝑑𝑒𝑑𝑔𝑒 3 (Gitelson et

× √
(Chlgreen) 𝐺𝑟𝑒𝑒𝑛 al., 2006)

Leaf Chlorophyll 𝑁𝐼𝑅 𝑅𝑒𝑑𝑒𝑑𝑔𝑒1 (Pu et al.,

× √
Index (LCI) 𝑁𝐼𝑅 𝑅𝑒𝑑 2008)

Normalized 𝑁𝐼𝑅 𝑅𝑒𝑑𝑒𝑑𝑔𝑒1 (Haboudane

Difference Red-Edge × √
𝑁𝐼𝑅 𝑅𝑒𝑑𝑒𝑑𝑔𝑒1 et al., 2004)
2 (NDRE2)

Normalized 𝑁𝐼𝑅 𝑅𝑒𝑑𝑒𝑑𝑔𝑒2

Difference Red-Edge × √ This study
𝑁𝐼𝑅 𝑅𝑒𝑑𝑒𝑑𝑔𝑒2
3 (NDRE 3)

Chlorophyll
𝑅𝑒𝑑 (Vincini et
vegetation index √ √ 𝑁𝐼𝑅
𝐺𝑟𝑒𝑒𝑛^2 al., 2008)
(CVI)

Green Difference
(Tucker et
Vegetation Index √ √ NIR Green
al., 1979)
(GDVI)

Green leaf index 2 𝐺𝑟𝑒𝑒𝑛 𝑅𝑒𝑑 𝐵𝑙𝑢𝑒 (Gobron et

√ √
(GLI) 2 𝐺𝑟𝑒𝑒𝑛 𝑅𝑒𝑑 𝐵𝑙𝑢𝑒 al., 2000)
Green Normalized
Difference 𝑁𝐼𝑅 𝐺𝑟𝑒𝑒𝑛 (Fu et al.,
√ √
Vegetation Index 𝑁𝐼𝑅 𝐺𝑟𝑒𝑒𝑛 2008)
(GNDVI)
Normalized
Difference 𝑁𝐼𝑅 𝑅𝑒𝑑 (Tucker,
√ √
Vegetation Index 𝑁𝐼𝑅 𝑅𝑒𝑑 1979)
(NDVI)
Normalize Green 𝐺𝑟𝑒𝑒𝑛
√ √ This study
(NG) 𝑁𝐼𝑅 𝑅 𝐺𝑟𝑒𝑒𝑛
Normalized
Difference
𝐺𝑟𝑒𝑒𝑛 𝑅𝑒𝑑
Green/Red (Hunt et al.,
√ √ 𝐺𝑟𝑒𝑒𝑛 𝑅𝑒𝑑
Normalized green 2011)
red difference index
(NGRDI)

90
 Chapter 5

Normalized
Difference NIR/Blue 𝑁𝐼𝑅 𝐵𝐿𝑈𝐸 (Hancock and
Blue-normalized
√ √ 𝑁𝐼𝑅 𝐵𝐿𝑈𝐸 Dougherty,
difference
2007)
vegetation index
(BNDVI)
Simple Ratio Blue / 𝐵𝑙𝑢𝑒
Red √ √ 𝑅𝑒𝑑 This study
SR-Blue/Red)

Chlorophyll index 𝑁𝐼𝑅 (Gitelson et

√ √ 1
green (CIG) 𝐺𝑟𝑒𝑒𝑛 al., 2003)

(Gamon and
𝐺𝑟𝑒𝑒𝑛
Green Index (GI) √ √ Surfus,
𝑅𝑒𝑑 1999)

Plant biochemical 𝑁𝐼𝑅

√ √ This study
index (PBI) 𝐺𝑟𝑒𝑒𝑛

Coloration Index 𝑅𝐸𝐷 𝐵𝐿𝑈𝐸 (Escadafal et

√ √
(CI) 𝑅𝐸𝐷 al., 1994a)

Plant pigment ratio 𝐺𝑟𝑒𝑒𝑛 𝐵𝐿𝑈𝐸 (Metternicht,

√ √
(PGR) 𝐺𝑟𝑒𝑒𝑛 𝐵𝐿𝑈𝐸 2003)
Water indices

Disease Stress 𝑁𝐼𝑅 𝐺𝑟𝑒𝑒𝑛 (Galvao et

√ √
water index (DSWI) 𝑆𝑊𝐼𝑅 𝑅𝑒𝑑 al., 2005)

𝑁𝐼𝑅 0.1 𝑆𝑊𝐼𝑅 0.02

Vegetation Moisture 𝑁𝐼𝑅 0.1 𝑆𝑊𝐼𝑅 0.02 (Ceccato et
√ √
Index (VMI) al., 2002)

Leaf Water Content 𝑙𝑜𝑔 1 𝑁𝐼𝑅 𝑆𝑊𝐼𝑅 (Cohen,

√ √
Index (LWCI) 𝑙𝑜𝑔 1 𝑁𝐼𝑅 𝑆𝑊𝐼𝑅 1991)

Normalized 𝑁𝐼𝑅 𝑆𝑊𝐼𝑅 (Hardinsky

Difference Infrared √ √ and Lemas,
𝑁𝐼𝑅 𝑆𝑊𝐼𝑅
Index (NDWI) 1983)
Simple Ratio SWIR 𝑆𝑊𝐼𝑅1
√ √ This Study
(SR-SWIR) 𝑆𝑊𝐼𝑅2
(Pinder and
Ratio Drought Index 𝑆𝑊𝐼𝑅1
√ √ McLeod,
(RDI) 𝑁𝐼𝑅 1999)

5.3 Results
5.3.1 Impact of bark beetle green attack on measured leaf
traits
The results of the ANOVA test showed significant differences (p < 0.05)
between healthy and infested samples for all measured leaf traits in this study

91
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

except for nitrogen concentration at plot level (Fig. 5.3). At the leaf level,
chlorophyll and leaf water content were significantly higher for healthy than for
infested foliar. In contrast, smaller differences were observed in foliar nitrogen
concentration (p < 0.04) between these two groups. Leaf dry matter content
(LDMC) was significantly higher for the infested leaves. Similarly, at plot level,
higher chlorophyll and water content was observed for the healthy plots (Fig.
5.3).

Figure 5. 3: Mean and standard deviation for measured leaf traits at both leaf (A) and
plot level (B). Leaf level data is obtained from the average measurements per tree and
plot level data is obtained from the average measurements of representative trees within
each plot.

5.3.2 Leaf and canopy spectral variations

Figure (5.4) shows the difference between the mean reflectance of healthy and
infested foliage. The difference is largest in the VIS region between 520-685
nm, the NIR (740-1130 nm) and the shortwave infrared region (1420-1850
nm and 2000-2200 nm). Similar results have been observed at canopy level
for the Sentinel-2 data, particularly in the NIR and SWIR infrared regions (Fig.
5.5 A). Infested trees tend to have a higher reflectance in the VIS and SWIR
regions than healthy trees. Moreover, the results of one-way ANOVA testing
shows that the mean reflectance spectra of healthy and green-attacked sample
plots were significantly different (p < 0.05) for all Sentinel-2 spectral bands at
both leaf (simulated spectra) and canopy level. In contrast, for Landsat-8, no
significant difference was observed at the canopy level. However, from the
simulated leaf spectra of Landsat-8 spectral bands, low significance differences
were observed for the red, NIR, SWIR1 and SWIR 2 bands (Table 5.3).

92
 Chapter 5

Figure 5. 4: Mean reflectance spectra of healthy and infested foliar at the green attack
stage. Grey areas depict the location of wavebands displaying is a significant difference
between healthy and infested spectra.

Figure 5. 5: (A) foliar and canopy reflectance using Sentinel-2 data (B) foliar and canopy
reflectance using Landsat-8 data.

Table 5. 3: One-way ANOVA test between healthy and infested reflectance data at both
leaf (simulated) and canopy level.
Significance level (p < 0.05)
Spectral bands Landsat -8 Sentinel- 2
Leaf level Canopy level Leaf level Canopy level
Blue • • ** *
Green • • ** ***
Red * • ** ***
Red-edge1 Not available Not available *** ***
Red-edge2 Not available Not available *** ***
Red-edge3 Not available Not available ** *
NIR * • ** **
NIR(a) Not available Not available ** *
SWIR-1 * • *** ***
SWIR-2 * • *** ***
(●) Not significant, (*) Hardly significant, (**) significant, (***) Strongly significant.

93
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

5.3.3 Principal component analysis (PCA) and ANOVA test

Significant differences (p < 0.05), were found for all SVIs calculated from
Sentinel-2 (except SR-Blue/Red, GDVI and CI) between healthy and infested
trees (Fig. 5.6). Whereas, the SVIs derived from Landsat-8 were only sensitive
to changes in water content (water indices) (DSWI, NDII and LWCI), and no
significant differences were found for the pigment indices (Fig. 5.6). Moreover,
the results of PCA revealed that the first two components (PC 1 and 2)
explained more than 70% of the variance in the samples of SVIs investigated
in this study. Figure 5.7 (A, B, and C) shows the clustering of two sample plots
(healthy and infested) within the space of the first two principal components
(PC1 and 2). As can be seen from Figure 5.7 (A), there was a slight crossover
between healthy and infested sample plots when all three groups of SVIs from
Sentinel-2 data were applied in the model. In contrast, the two sample plots
(healthy and infested) exhibited an apparent overlap and mixed scattering for
the Landsat-8 SVI groups. Furthermore, for Sentinel-2 SVIs, the majority of
healthy plots tended to be on the negative side of PC1, while the infested plots
were relatively scattered on the positive side of PC1 (Fig. 5.7 (B)). Notably, no
obvious improvement was observed when the PCA model was built
independently for each SVI group.

Figure 5. 6: Comparison of p-values from one-way ANOVA analysis between SVIs for
healthy and infested samples (red dash line represents the significant level (p < 0.05)).

94
 Chapter 5

Figure 5. 7: Cluster plots based on the first two PCs: (A) Cluster plots based on all spectral
indices including raw spectral bands (B) Cluster plots based on pigment indices and (C)
cluster plots based on water-related indices.

5.3.4 Mapping bark beetle green attack and validation

To determine the importance of each of the SVIs when mapping bark beetle
green attack, selection probability (SP) values were obtained from the RF
algorithm. A total of 8 out of 24 indices shown in Figure 5.8 for Landsat-8
indicated a higher SP (> 0.5), whereas 17 of 35 SVIs were recorded with an
SP value of > 0.5 for Sentinel-2 (Fig. 5.8).

95
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

Figure 5. 8: Selection probability value of SVIs obtained from PLS-DA Random frog
algorithm.

Moreover, higher variation and wider threshold values were observed between
these two sample groups (healthy and infested) using the SVIs considered in
this study for the Sentinel-2 data than for Landsat-8 (see Fig. 5.9). Likewise,
using raw spectral bands from Sentinel-2 indicates that red, red-edge1-3, NIR
and SWIR bands exhibited SP values of > 0.5. In contrast, the Landsat-8 red
band had an SP value > 0.5. From Sentinel-2 data, the indices selected to
generate a final infestation map included NDRE- 2, NDRE-3, GLI, GNDVI, NDVI,
NGRDI, CIG, PGR, DSWI, LWCI, NDWI, SR-SWIR, RDI, red, red-edge1&2, NIR
and SWIR1. However, from Landsat-8 data only CIG, GLI, NGRDI, PGR, DSWI,
NDWI, DRI, and red were selected.

Based on the defined threshold value, infestation maps were generated using
the Landsat-8 and Sentinel-2 SVIs data. These maps were then overlaid with
ground truth reference data to calculate matched and mismatched pixels.
Figure 5.10 depicts those areas (pixels) correctly matched with ground truth
data (aerial photography), as well as pixels that have been falsely identified
from SVIs as green attacked areas. As can be seen from Table 5.4, the number
of correctly matched pixels with reference infestation data (visual
interpretation of aerial photography) was higher for Sentinel-2 (67 %) than

96
 Chapter 5

for Landsat-8 (36 %) SVI data. Similarly, the number of falsely identified pixels
(mismatched pixels) that indicated green attack using SVIs was lower for
Sentinel-2 (177 pixels) than for Landsat-8 (391 pixels).

Table 5. 4: Assessment of the generated maps from Landsat-8 and Sentinel-2 SVIs using
the reference data obtained from Aerial photography.

Identified pixels as Reference Pixels

Mismatched
green attack from pixels (Aerial correctly Error
pixels
(SVIs) photography) matched

Landsat-8 (612 417 (30 m) 221 391 64%

pixels) (36%)
Sentinel-2 (539) 687 (20 m) 362 177 33%
(67%)
* The error was calculated by dividing the total number of correctly matched pixels by
the total number of ground truth pixels.

Landsat-8

97
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

Sentinel-2

Figure 5. 9: The box plot shows the variation in SVI values calculated from Landsat-8
and Sentinel-2, between healthy and infested plots. The black and hollow boxes
represent healthy and infested plots, respectively. The red box shows the selected
threshold value of each SVI where there is no overlap between the two sample groups
(healthy and infested)

98
 Chapter 5

Figure 5. 10: Map showing the spruce cover under green attack stress in the Bavarian
Forest National Park in July 2016 based on spectral vegetation indices from Landsat-8
and Sentinel-2 selected through the Random frog algorithm.

5.4 Discussion
Spectral vegetation indices (SVIs) calculated from Sentinel-2 have a high
potential for mapping and detecting changes induced by bark beetle green
attack, particularly the red-edge and water-related indices. These changes
were only partly detectable by Landsat-8 due to the lower spectral and spatial

99
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

resolution of the OLI sensor. A greater number of pixels identifying green

attack from Sentinel-2 SVIs matched with ground truth data (362 out of 539
pixels), whereas Landsat-8 only matched 221 pixels (out of the 612) with
ground truth infestation data. Furthermore, leaf-level comparisons between
healthy and green-attacked foliar samples revealed that all leaf traits
considered in this study were significantly different (p < 0.05), particularly for
chlorophyll and water content. The reduction of chlorophyll and water content
in the infested trees caused changes in the foliar spectral measurements at
NIR and SWIR wavelengths.

In our study, the stand characteristics between the healthy and infested plots
were not significantly different (not shown). This is in line with earlier studies
that found stand characteristics do not play a major role in infestation during
an outbreak condition. In an epidemic level of bark beetle infestation all trees
with different conditions (healthy and stressed) and different stand
characteristics (large or small diameter) were under threat of this insect
(Wermelinge, 2004; Lausch et al., 2011; Allain et al., 2011). However, it is
important to not that in our study, DBH with the p-value of < 0.06 was narrowly
exceeding the threshold of 0.05, which probably indicate that this variable
may play a role in attracting the beetles. Previous findings by Hart et al. (2015)
and Six and Skov (2009) indicate that DBH played a major role in the endemic
level of bark beetle infestation.

A common observation in this study was that of the 19 pigment-related SVIs

(group-1) calculated from Sentinel-2 imagery, eight (i.e., GLI, NDVI, GNDVI,
CIG, PGR, NGRDI, and both red-edge-related indices (NDRE-2 and 3)) were
most important having an SP value of < 0.5 (Fig. 5.8). When only using VNIR
bands from Landsat-8, CIG, GLI, NGRDI, and PGR have the potential to
differentiate between healthy and infested sample plots. The underlying
reasons may relate to the existing dissimilarities in spectral and spatial
resolution between these two sensors (Mandanici and Bitelli, 2016). For
example, Landsat-8 has only one spectral band (30 m) in the near-infrared
region, while Sentinel-2 has a series of spectral bands (20 m) in the near-
infrared region (B5, B6, B7, and B8a). Furthermore, the availability of the three
red-edge bands is a unique feature that distinguishes Sentinel-2 from Landsat-
8. In our study, the indices developed from the red-edge bands (705–783 nm)
of Sentinel-2 (NRED2 and 3) showed the highest sensitivity to bark beetle
green attack, and a larger threshold was identified between healthy and
infested samples for these two indices (Fig. 5.8). This result is consistent with
results from previous studies (Eitel et al., 2011; Krofcheck et al., 2014;
Lottering et al., 2016; Modzelewska et al., 2017), indicating that NDRE
calculated from RapidEye and WorldView-2 has the capacity to detect forest
stress induced by drought and insect infestation in the early phase.

100
 Chapter 5

Interestingly, the spectral indices using the blue band (CI, SR-BLUE/RED and
BNDVI) were unable to discriminate trees stressed by bark beetle green attack
for both Landsat-8 and Sentinel-2 (Fig. 5.8). This is due to the reflectance of
these sensors in the blue region of the spectrum being insufficient to detect
spectral variation caused by bark beetle green attack. This result confirmed
earlier findings by Arellano et al. (2015), who revealed that the blue range
indices calculated from Hyperion images were unable to detect forest areas
polluted by hydrocarbon in the Amazon rainforest.

On the other hand, the indices calculated from a combination of SWIR and NIR
or VIS bands (water-related indices) performed well for both Sentinel-2 and
Landsat-8 imagery. For example, Figure 5.9 suggests that the indices DSWI,
NDWI, RDI, SR-SWIR, and LWCI successfully differentiated healthy from green
attacked plots using Sentinel-2. Due to the significantly lower water content
and higher leaf dry matter content (p < 0.05) of infested samples, their
spectral reflectance responded more profoundly in the SWIR region at both leaf
and canopy level than the reflectance of healthy samples did (Figure 3) (Ali et
al., 2016; Wang et al., 2011). This is because a reduction in leaf water content
is responsible for changes in SWIR reflectance (Bowman, 1989). This is also
demonstrated in Figures 5.4 and 5.5 (A and B), where distinctively higher
reflectance in the SWIR region was observed for the infested samples at both
leaf and canopy level. This is consistent with the study by Immitzer et al.
(2016), which highlighted the importance of the SWIR in Sentinel-2 data for
mapping different forest classes. Similar results have been reported for
wavelengths in the NIR and SWIR regions, which are sensitive to forest
disturbance caused by insect attacks. Foster et al. (2017) found that the SWIR
region from hyperspectral measurements collected from ASD FieldSpec Pro
was key to the detection of bark beetle (Dendroctonus rufipennis) green attack
in Engelmann spruce (Picea engelmannii).

The sensitivity of SWIR bands to variation in leaf water content is due to its
reflective nature, allowing it to bounce off objects while remaining invisible to
the human eye. This feature of the SWIR region makes variation in leaf water
content easily recognisable. Furthermore, it should be noted that not only the
SWIR region is sensitive to variation in leaf water content, but that the thermal
infrared region may also provide sufficient information in this regard. Landsat-
8 has two spectral bands in the thermal infrared region (10.60 – 12.51 nm).
Further research would, therefore, be useful to assess whether information
from the thermal region can provide sufficient information for mapping bark
beetle green attack.

Furthermore, the results of PCA analysis revealed that the method using
different SVI groups (pigment or water-related indices) did not improve
discrimination between healthy and infested sample plots when using Sentinel-

101
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

2 data (Fig. 5.7). However, with Landsat-8 data, the performance of the PCA
model decreased when SVIs groups were used separately. This reduction in
PCA performance was expected due to the spectral variation of infested plots
not being efficiently detected by Landsat-8 images (see Fig. 5.5 (A and B)).

It should also be noted that 177 and 391 pixels indicating green attack based
on Sentinel-2 and Landsat-8 data, respectively, were mismatched with the
ground truth data (Table 5.4). To better understand this discrepancy, we
studied the infestation data (obtained from visual interpretation of aerial
photography) of the previous years, 2014 and 2015, acquired from the
Bavarian Forest National Park administration. This allowed us to check whether
mismatched pixels are indicating green attack corresponded with previous
infestations in 2014 and 2015. To accomplish this, a similar process (as
explained in Section 2.5) was applied to 2014 and 2015 infestation data and
then overlaid on the mismatched pixels. As can be seen from Table 4, the
results revealed that, for Sentinel-2, 12 and 57 pixels (out of the 177
mismatched pixels) matched with the infestation data of the years 2014 and
2015, respectively. While for Landsat-8 data, only 11 and 31pixels matched
with infestation data for the years 2014 and 2015, respectively (Table 5.5).
This highlights the sensitivity of SVIs calculated from Sentinel-2 for detecting
canopies that are stressed by bark beetle green attack. It is important to note
that all the pixels that were identified as a green attack, using both sensors
considered in this study, were within 500 meters of the previous year’s
infestation when analysed using a 500-metre buffer zone around these data
(not shown). Hence, we could identify the previous years’ infestations, which
were now at the grey attack stage.

Table 5. 5: Assessment of the mismatched pixels from Landsat-8 and Sentinel-2 SVIs
using previous years’ (2014 and 2015) infestation data (obtained from Aerial
photography).
Mismatched pixels as Pixels correctly Pixels correctly
green attack from matched with matched with
(SVIs) infestation data infestation data
(2014) (2015)
Landsat-8 (391 pixels) 11 31
Sentinel-2 (177 pixels) 12 57

5.5 Conclusion
The spectral indices derived from Sentinel-2 data performed well at detecting
changes in the leaf biochemical properties (reduction in chlorophyll, increase
in leaf dry matter content and decrease in water leaf water content) and their
relation to canopy reflectance. The simulated Sentinel's data also showed good
accordance with measured leaf reflectance using an ASD spectrometer as well
as a superior response to changes in leaf biochemical properties over the whole
wavelength region, as almost all utilised SVIs performed well for detecting bark

102
 Chapter 5

beetle green attack. Although the total number of pixels matched with ground
truth data (362 pixels, 67%) obtained from the Sentinel-2 may not be high
enough for operational forestry practice and management purposes, it is a
promising technique for alerting to bark beetle green attacks. It can aid the
detection of bark beetle infestations in a timely manner over large areas and
thus form the basis for accurate and efficient bark beetle monitoring. It is
possible that another type of remote sensing data with a higher resolution
(such as from an Unmanned Airborne Vehicle) may provide a better result
detecting bark beetle green attack. As early detection of infestations is
essential for the successful control of an outbreak (Fahse and Heurich, 2011),
further research applying different approaches using Sentinel-2 imagery should
be undertaken with principal component analysis (PCA) and partial least square
discriminate analysis (PLS-DA) from this study to check the stability and
accuracy of the threshold values identified over a period of time.

103
Detection and mapping (Ips typographus, L) green attack using Landsat-8…

104
Chapter 6 Synthesis: Remote sensing of
European spruce (Ips typographus, L.) bark
beetle green attack

105
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

6.1 Introduction
During the last three decades, there have been massive and increasing
numbers of severe European spruce Ips typographus, L. bark beetle outbreaks
in the coniferous forests in central Europe. These have resulted in significant
ecological changes in terms of forest structure, the composition of wildlife
habitats and degradation of large areas within existing forests (Christiansen
and Bakke, 1988b; Fahse and Heurich, 2011; Raffa et al., 2008; Seidl et al.,
2011; Seidl et al., 2014). In other words, the bark beetle count is an agent of
stress and change in forest structure and function and has an effect on the
ecosystem both directly and indirectly. The direct impact of the bark beetle on
the ecosystem includes increasing tree mortality rates, changing forest stand
density and altering the forest microclimate by reducing the forest canopy
layer, which can increase the amount of sunlight to the forest floor (Beudert et
al., 2015; Kurz et al., 2008; Lehnert et al., 2013; Lindenmayer and Franklin,
2002; Mikkelson et al., 2013).To control or preclude a mass outbreak and
minimise the economic loss in the forest industry, early detection of a bark
beetle infestation is an essential step. Detection of a bark beetle infestation at
the green attack stage means locating infested trees when the beetle larvae
are still confined within the tree. In this respect, remote sensing is a cost-
effective and repetitive technique which is an optimal approach to monitor and
assess forest stress in comparison to the more traditional (field survey)
approach, which is not practical and is actually inefficient in large areas
because it is significantly laborious and costly. Therefore, the current thesis
has studied the utility of different types of remote sensing data obtained from
a variety of sensors, including ASD FieldSpec3 and multispectral satellites
(RapidEye, SPOT-5, Sentinel-2 and Landsat-8) for early detection of bark
beetle infestation at both the leaf and canopy levels. Furthermore, this study
provided an important opportunity to advance the understanding the impact of
early bark beetle infestation on the biochemical properties of the infested trees
(chlorophyll, nitrogen, chlorophyll fluoresces, leaf water content, dry matter
content and stomatal conductance) (Fig. 6.1).

The results of the study have been presented and published in four scientific
articles which form the main four chapters of this thesis (Chapter 2-5). While
the general introduction to the bark beetle infestation, aim and objectives of
the study are presented in Chapter one. In Chapter 2, foliar biochemical
(chlorophyll and nitrogen concentration) and spectral reflectance properties
(400–2000 nm) obtained from an ASD FieldSpec3 equipped with an integrating
sphere were used to study the impact of bark beetle green attack on the
needles of the infested spruce trees. Subsequently, in Chapter 3, temporal
high-resolution satellite data from RapidEye and SPOT-5, parallel with field
measurements of leaf properties , were used to understand and explore the
dynamics of leaf traits and canopy reflectance of Norway spruce during the

106
 Chapter 6

early to advanced stage of bark beetle attack (green to red attack) (Ips
typographus, L.). Leaf spectral measurements together with leaf traits
(nitrogen, chlorophyll, chlorophyll fluorescence, water content and stomatal
conductance) were studied from three repeated sequential measurements. In
Chapter 4, the utility of thermal (TIR) and visible-short wave infrared (VIS-
SWIR) data was evaluated to detect the temporal variations of the canopy due
to bark beetle infestation and to study thermal sensitive foliar properties (i.e.
stomatal conductance, chlorophyll fluorescence and water content). In Chapter
5, the potential of data from two freely available satellites—Sentinel-2 and
Landsat-8—were compared to detect and map bark beetle infestation at the
early stage with the use of field measured data (foliar reflectance and
biochemical properties).

Finally, this chapter (Chapter six) synthesises the main results and provides a
summary of the main findings. The chapter further discusses the practical
relevance of using remote sensing data for detection of bark beetle green
attack.

Figure 6. 1: Research framework and logical established links between

chapters and utilised data.

6.2 Effects of bark beetle green attack on foliar

reflectance and biochemical properties
Very few spectroscopic data are available for Norway spruce trees attacked by
the European bark beetle at the leaf level, despite such data being highly
beneficial for detecting pre-visual stress. Hence, we examined the potential of
hyperspectral measurements (400–2,200 nm) from an ASD FieldSpec3
equipped with an integrating sphere to detect possible changes in foliar
reflectance induced by bark beetle infestation.

107
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

The results of this study revealed that the mean reflectance spectra of attacked
green foliage were statistically different (p < 0.05) compared to healthy foliage
for 917 wavebands out of the 1600 wavebands considered in this study from
400–2200 nm. The findings demonstrated that the infestation of bark beetle
at the green attack stage reduced the concentration of chlorophyll and nitrogen
within the infested foliage. Previously, it was shown by Ahern (1988b) that the
most promising spectral bands for the detection of mountain pine beetle green
attack at the leaf level were located near the red edge of the spectrum. In
addition, the results of variable importance in projection (VIP) obtained from
partial least squares regression (PLSR) revealed that the wavelength range
between 730 and 790 nm (red-edge) is the most informative spectral area
affected by the European spruce bark beetle, and more pronounced in regards
to chlorophyll variation between healthy and infested foliage (Fig 6.2).

Our results further demonstrated that the reflectance at NIR and SWIR regions
(730–1,370 nm) was distinctively lower for the green attacked foliage than the
healthy ones. This phenomenon is mostly due to the deficit in water content
observed in infested trees, leading to a degeneration of the internal leaf
structure at the cellular level (Slaton et al., 2001b; Zhang et al., 2012a).

The research also found that the infestation of the bark beetle at the green
attack stage affected the estimation accuracy of foliar chlorophyll and nitrogen
concentration using hyperspectral measurements and the accuracy decreased
when the tree was infested by a bark beetle green attack. This was confirmed
from the results of PLSR (Fig. 6.3). Accordingly, the study concluded that the
retrieval of biochemical properties from hyperspectral measurements may be
used as an indicator for efficient landscape-wide detection of a bark beetle
green attack. It is, however, important to note that retrieval accuracies for the
total chlorophyll and foliar nitrogen concentrations at the leaf level would
probably be different from those at the canopy level because of structural and
external factors, such as illumination and atmospheric conditions, that may
affect the reflectance spectra at the canopy level (Ollinger, 2011; Wang, 2016).

Figure 6. 2: Importance of wavelengths corresponding to the highest value of variable

importance in the projection scores of partial least squares regression in healthy and
infested samples.

108
 Chapter 6

Figure 6. 3: Measured versus predicted foliar chlorophyll and nitrogen concentration for
healthy (a) and infested (b) samples derived from the PLSR analysis.

6.3 Understanding dynamic changes of Norway

spruce trees at both leaf and canopy level under
bark beetle infestation during green to red attack
stage
Continuous monitoring is essential for investigating and understanding the
dynamic characteristics of leaf traits and canopy reflectance experiencing a
bark beetle infestation, in order to provide possible information about changes
in the infested forests at the early stage of the attack and instigate possible
management interventions. Therefore, the dynamics of leaf and canopy
reflectance of Norway spruce trees under bark beetle Ips typographus, L
infestation was investigated for the first time.

Chapter 3 in this thesis discussed the potential use of two high-resolution

satellites from RapidEye and SPOT-5 for the detection of a bark beetle
infestation and monitoring possible changes in infested trees. In situ
measurements of chlorophyll, chlorophyll fluoresces, nitrogen content, leaf
water content and stomatal conductance were obtained for three consecutive

109
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

repeated time measurements during a field campaign in 2015, concomitant

with the timing of image acquisition.

The findings showed that the variation between healthy and infested trees was
increased with the progression of infestation for all studied leaf traits
(particularly for stomatal conductance and chlorophyll fluorescence). Stomatal
closure is the first stress-induced physiological response of infested trees in
order to preserve water. These findings confirm results in section 6.4 which
showed that canopy surface temperature (CST) affected by a bark beetle
infestation is higher compared to the healthy situation. One observed result is
an increase in leaf and canopy temperature due to a decline in transpiration
cooling.

The spectral regions selected for differentiating and monitoring spectral

changes induced by bark beetle infestation were the red edge and SWIR bands
from RapidEye and SPOT-5, respectively. These two spectral regions
outperformed other spectral regions in detecting early bark beetle infestation
from June to July 2015 (Fig. 6.4). Similar results were observed at the leaf
level (section 6.2), for these same two spectral regions. Furthermore, we found
that the spectral vegetation indices calculated from the red edge and SWIR
spectral bands, such as NDRE, DSWI, NDWI and LWCI, were the most
sensitive indices that differentiated healthy from infested sites (Fig. 6.4).

This study concluded that red-edge and SWIR spectral bands of the
multispectral satellites can significantly improve the monitoring and detection
of forest stress under red attack, with subsequent important implications for
European bark beetle management and future studies.

RapidEye

A A

110
 Chapter 6

SPOT-5

A A A

Figure 6. 4: (A) Temporal variation of studied spectral vegetation indices for healthy and
infested plots in the Bavarian Forest National Park using RapidEye and SPOT-5 satellite
data. Green and black lines represent healthy and infested plots, respectively; (B)
Unpaired t-test of canopy reflectance between healthy and infested samples. Dark
squares indicate spectral wavebands that were significantly different (P ≤ 0.05). The red
box shows the spectral region that was significantly different overall temporal data
considered in this study.

6.4 Sensitivity of Landsat-8 OLI and TIRS Data to

Foliar Properties of Early Stage Bark Beetle (Ips
typographus, L.) Infestation
Several studies have shown that TIR data have significant potential for
detecting plant diseases and pathogens at a stage before the plants exhibit
visual signs (Aldea et al., 2005; Moller et al., 2007; Ni et al., 2015; Oerke et

111
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

al., 2006; Xu et al., 2006). In this respect, Chapter 3 sought to explore and
compare the potential of both optical and thermal infrared (TIR) data for the
detection of pre-visual symptoms induced by a bark beetle green attack. Three
Landsat-8 images from May, July and August 2016 were studied, representing
early, advanced, and post-infestation stages, respectively.

These results showed that CST calculated from TIR data was superior to the
spectral vegetation indices obtained from optical bands for detecting subtle
canopy changes due to a bark beetle infestation for the three months studied
(Fig. 6.5). In addition, a strong correlation between CST with leaf water
content, chlorophyll fluoresces and stomatal conductance was observed in the
healthy leaves (r = –0.46, –0.55 and –0.48, respectively), whereas for the
green attacked leaves, they were r = 0.05, –0.15 and –0.25, respectively (Fig.
6.6). CST was found to be an important variable for estimating measured leaf
traits (VIP > 1) and improving the differentiation between healthy and green
attacked sites when used with other SVIs. Accordingly, TIR may prove to be a
considerably useful indicator in this context. This study concludes that the early
stress induced by a bark beetle infestation is more effectively represented by
data obtained from TIR compared to optical sources. This could have positive
implications for future forest practice.

May July August

Figure 6. 5: Temporal variation of CST for healthy and green attacked sample plots in
the Bavarian Forest National Park. Green and red boxes represent healthy and infested
plots, respectively. A * indicates the significant difference between healthy and infested
plots, respectively. A * indicates the significant difference between healthy and infested
plots obtained using Student t-tests, and a blue o indicates an outlier.

112
 Chapter 6

Figure 6. 6: Correlation between studied leaf traits (chlorophyll fluorescence, leaf water
content, and stomatal conductance) and SVIs, in the Bavarian Forest National Park, July
2016.

6.5 Detection and mapping bark beetle green attack

using spectral vegetation indices from Landsat-
8 and Sentinel-2 data
Detecting and mapping Ips typographus, L green attack with medium
resolution satellite data was addressed in Chapter 5. The potential of spectral
vegetation indices from Sentinel-2 and Landsat-8 alongside field data
measurements to detect a bark beetle green attack was investigated.
Furthermore, the variation in reflectance spectra between healthy and infested
plots was determined at the leaf and canopy level.

The research revealed that the spectral vegetation indices from Sentinel-2
satellite data have a strong potential to map and detect a bark beetle
infestation at its earliest stage. Moreover, the results showed that the
difference between the mean reflectance of healthy and infested foliage is most
substantial in the VIS region between 520 and 685 nm, the NIR (740–1,130
nm) and the shortwave infrared region (1,420–1,850 nm and 2,000–2,200
nm). Similar results have been observed at the canopy level for the Sentinel-
2 data when compared with interpolated field ASD spectrometer at the leaf
level, particularly in the NIR and SWIR infrared regions, while for the Landsat-
8, this variation was not detected at the canopy level (Fig. 6.7). Moreover, a
good agreement exists between canopy reflectance and leaf spectra

113
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

(interpolated to Sentinel-2 spectral bands); in particular, for the red-edge and

SWIR wavebands, the correlation was r = 0.50 to 0.61 (p < 0.05), while
weaker correlations were observed between the canopy and interpolated leaf
spectra for Landsat-8 image data for both healthy and infested plots (Fig. 6.7).

The spectral bands selected as the best combination of the vegetation indices
for detecting and mapping bark beetle green attack were found in the red-edge
and SWIR bands in Sentinel-2, including NDRE2-3, SR-SWIR, NDWI, DSWI and
LWCI. This confirmed our earlier research, described in section 6.3, which
studied the temporal dynamics of canopy reflectance under a bark beetle green
attack and revealed that both red-edge and SIWR bands were the most
informative in terms of maintaining their sensitivity for monitoring and
detecting a bark beetle infestation from early to advanced stages of infestation
(Fig. 6.4). An increased numbers of pixels, identified as a green attack from
Sentinel-2 SVIs, were matched with ground truth data (67%), while for
Landsat-8 (36%), they were matched with ground truth infestation data. This
result confirmed that the spectral vegetation indices calculated from Sentinel-
2 spectral bands performed better in differentiating between healthy and
infested plots compared to Landsat-8 SVIs.

The study concluded that both red-edge and SIWR regions have the potential
to provide valuable information about infested areas and can produce a reliable
map of which areas in a forest are affected by a bark beetle green attack.

Figure 6. 7: foliar and canopy reflectance using (A) Sentinel-2 data (B) Landsat-8 data
(C & D) correlation between canopy reflectance and foliar spectra (interpolated from
Field ASD spectrometer to Sentinel-2 and Landsat-8 spectral bands) for both healthy and
infested plots.

114
 Chapter 6

6.6 Practical relevance

This study contributes to applied research on the field of remote sensing of
bark beetle green attack. The findings and methods applied in this work have
the potential to produce useful information about bark beetle infestation at an
early stage of the attack.

The results confirmed that it is possible to detect the stress induced by the
bark beetle at an early stage, and prior information derived from field
measurements can significantly improve this process through remote sensing
data and approaches. This will facilitate regular monitoring of bark beetle
infestation in conifer forests. However, the key issue is how the findings
obtained by this research and reported in this thesis can be used for detecting
bark beetle infestation. In other words, how, when and which spectral regions
are capable of using driven information from space to detect a bark beetle
green attack. We can conclude that in the remotely sensed survey of green
attacked Norway spruce trees affected by the bark beetle, that high-quality
observation in space, time and spectral characteristics are required, to make
this survey useful in a management context, as outlined in the following points:

Spectral information: Norway spruce trees impacted by the European spruce

bark beetle exhibits stress symptoms in a gradual process. Therefore, it is
necessary to use remote sensors which can record and collect information over
all portions of the electromagnetic spectrum (visible, near infrared-shortwave
infrared and thermal infrared). At the beginning of a bark beetle infestation,
the most common initial response is the closure of plant stomata to preserve
water content. Therefore, it may not be possible to detect such changes and
physiological stress using the visible portion (400–700 nm) of the
electromagnetic spectrum. In contrast, satellites such as Landsat-8 and
Sentinel-2 with the capabilities to further capture information in near,
shortwave and thermal infrared regions of the electromagnetic spectrum would
be more sensitive to this kind of physiological stress.

Moreover, in remote-sensing studies of bark beetle infestation, spectral

characteristics is another important issue and should be considered—results
from our study suggest that hyperspectral data are very important in
accurately detecting forest stress induced by bark beetle green attack.
Hyperspectral data can provide significant improvements in spectral
information content and a closer correlation with the biochemical and
physiological properties of targeted leaves or canopies when compared with
data from broad-bands sensors for detecting plant stress. For example, such
data are able to capitalise on both biochemical and structural properties of the
target forest area and offer enhancement in detecting subtle changes in tree
canopy reflectance due to physiological stress from insects or pathogens. The

115
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

hyperspectral data capturing can be performed at the early stages of bark

beetle infestation, and hence improve an ability to detect the stress induced
by bark beetle at its early stage. This can significantly benefit the control of
massive bark beetle outbreaks and minimises economic losses compared to
multispectral remote sensing, which can only detect canopy stress when
infestations are at their advanced stages and are widespread.

In the optical spectral region, we identified two spectral regions—the red edge
and SWIR—as being the most important spectral channels at both the leaf and
canopy levels for detecting subtle changes in Norway spruce trees under a bark
beetle green attack. Hence, remote sensors such as Sentinel-2, RapidEye,
SPOT-5 and the HYpex camera mounted on a UVA with the capacity to collect
information over these two spectral regions are providing better opportunities
to detect insect-stressed forests.

Temporal information: Because Norway spruce trees infested with European

spruce bark beetles do not die immediately, continuous monitoring of forests
from remote sensing platforms is critically important for characterising and
understanding the temporal response changes of infested trees. Multidate
images, within the same months, are preferably collected when the beetles
start to swarm during the European springtime. We reported in our studies
that employing multi-temporal imagery from RapidEye, SPOT-5, Sentinel-2
and Landsat-8 images are providing valuable information related to the stress
induced by bark beetles. We were able to identify for the first time in the
European situation (and in contrast to earlier works in the United States) that
mid-June to the beginning of July is an appropriate time frame for identifying
early stress induced by a bark beetle infestation. It is important to note that
our results at the leaf level show that the foliar properties change affectedly
within 3-4 weeks from the infestation time. Therefore, it is assumed it believes
that using high temporal resolution data (e.g. 3-5 day) may further improve
the time frame for early detection of European spruce bark beetle.

Radiometric Resolution: One of the main issue in the study of bark beetle
green attack when using remote-sensing data is the intra-class variability of
the reflectance values between healthy and green attacked trees, an issue
which is hampering the separability of infested and healthy classes. In this
case, the radiometric resolution of remotely sensed data plays an important
role. Radiometric resolution describes the actual information content in an
image and offers the ability to discriminate very slight differences in energy.
Therefore, the finer the radiometric resolution, the more sensitive it is to detect
small differences in reflected data over different objectives on the ground.
However, it is important to note that the radiometric resolution itself is
insufficient for detecting such differences. For example, Sentinel-2 and
Landsat-8 both have a radiometric resolution of 12-bits; however, in our study,

116
 Chapter 6

we showed that the potential of Sentinel-2 is higher than Landsat-8 imagery

for determining subtle canopy changes induced by bark beetle due to the
higher spectral and spatial resolution of Sentinel-2 data.

Spatial information: spatial resolution is a key issue in remote sensing

studies of a bark beetle green attack. Different remote sensing instruments
capture and collect data at different levels of scale and resolution: the higher
spatial resolution may outperform lower-resolution for bark beetle green attack
detection. This is especially true when they combine with high spectral and
radiometric resolution data. In other words, the combination of high spectral,
spatial, temporal and radiometric resolution can provide valuable information
regarding bark beetle green attack detection.

Moreover, It should be noted that the importance of high spatial resolution

data is increased when the bark beetle infestation occurs at an endemic level
(non-outbreak). In that case, the majority of the infested trees will be within
small isolated patches. Therefore, a higher spatial resolution is required
because it is more appropriate for small patch detection. However, when the
infestation is at an epidemic (outbreak) level, as was reported earlier in
previous chapters (5), larger clusters of trees are infested; in this case, there
is a possibility to detect the subtle spectral differences caused by a bark beetle
green attack using moderate-resolution imagery.

6.7 The relative importance of ground-based

measurements for bark beetle green attack
detection
During the past decades, fundamental research has been conducted on the
applicability of satellite remote sensing observations for forest parameter
estimation using extensive field data measurements. The results and findings
of these studies have been used in extensive calibration/validation experiments
and technological innovation, to propose and develop upcoming satellite
missions (i.e., Sentinels, EnMap, CHIME and HyspIRI).

In general, the purpose in ground-based measurements is to aid the calibration

and interpretation of remotely sensed data. However, in the field of bark beetle
infestation, ground-based measurements can provide even more contextual
information to obtain high accuracy results. An on-site gathering of location
and the associated characteristics of the infested tree that properly
characterise states, conditions, and parameters associated with the early stage
of bark beetle infestation can add value to physical imaging remote sensing
observations and possibilities to interlink the canopy stress assessment with
bark beetle infestation. In other words, when both in situ and remotely sensed
data are combined in a complimentary sense, they may provide additional

117
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

information in comparison to using a single source (e.g. field data or remote

sensing data alone) would be too imprecise.

As described in chapter one, the potential of remotely sensed data to provide

stress signal induced by bark beetle infestation depends on the physiological
response of trees and biology (lifecycle) of bark beetle. As such, in chapter
two, we identified spectral wavebands that are sensitive to variation in
chlorophyll and nitrogen concentration due to bark beetle green attack.
Moreover, we showed that the most common initial response of bark beetle
infestation is the closure of plant stomata to preserve water content. In chapter
three we selected the best time frame for bark beetle green attack detection
using temporal field measurements and remote sensing data. Moreover, in
chapter four and five we demonstrated the potential of multispectral satellite
data in detection bark beetle infestation. This information together can be used
to; a) produce predictive Models to detect or assess forest health status under
bark beetle infestation, and b) to propose or develop new remote sensing
sensors to fulfil all requirements for bark beetle green attack detection.

6.8 Implications for commercial forestry and

protected area management:
Prevention and suppression of bark beetle outbreaks comprise particularly
important applied strategies in commercially managed forests (Wermelinger,
2004). However, in protected areas, management of bark beetle infestation is
more complicated. Bark beetles can be viewed as natural renewal agents and
therefore an essential part of ecosystems, with the integrity of their protected
areas being mandated to guard against human interference. However, if left
uncontrolled, bark beetles may move outside park borders, cause damage on
commercial forests or private forested areas (McFarlane and Witson, 2008).
Similarly, the disturbance caused by bark beetle infestation negatively impacts
protected areas’ visual beauty and thus potentially on the visitor experience
(Müller and Job, 2009; Sheppard and Picard, 2006). Therefore, management
strategies for bark beetle infestation in protected areas are always required to
maintain a balance between ecological, social and economic demands.

Various bark-beetle-infestation strategies have been adopted. These

approaches include salvage logging, and large-scale commercial harvesting of
infested stands at the green attack stage before the new brood emerges and
migrates (Wermelinger 2004; Wulder et al. 2009). Forest managers need to
know the location of green attack trees in order to strategically and
operationally allocate resources for mitigation and control. In this regard, our
findings can provide useful information for forest management, particularly
with respect to the potential of Sentinel-2 data for detecting canopy stress
induced by bark beetle infestation at the green attack stage. This strategy will

118
 Chapter 6

reduce the various costs involved and facilitate bark beetle survey practice
since the data available free of charge.

Moreover, our findings can improve bark beetle management activities by

providing useful information regarding how, when and which remote-sensing
data could be applied to such survey practice. For example, our findings in
Chapter 3 showed that the most appropriate time for European field bark beetle
management is between Jun to July (European conditions). Furthermore, the
findings of our study in Chapter (2) revealed that biochemical properties (such
as chlorophyll, nitrogen and water content) in infested trees are most likely
affected during the green attack stage. Typically, chlorophyll, nitrogen and leaf
water content have been used by an ecologist to study the interaction between
the ecosystem and biological inhabitants; for example, to assess changes in
plant health associated with environmental stress and disease (Roumet et al.,
2006). Therefore, administrative bodies of forested areas may use such
information as an indicator for the efficient landscape-wide detection of bark
beetle green attack.

6.9 Future of remote-sensing platforms in

monitoring bark beetle infestation: UAV and
space-borne hyperspectral satellites
The traditional methods of remote-sensing data for bark beetle infestation still
cannot meet the needs of forest management practice due to logistical and
technological limitation factors as explained earlier in Chapter (1). As
mentioned in section 6.6, in bark beetle green attack analyses, the spectral,
spatial and temporal resolutions of remote-sensing platforms play a key role.
In the last few decades, airborne hyperspectral sensors have been among the
most important platforms providing high-spectral and -spatial resolution data.
The data captured by this platform are less affected by atmospheric
perturbation. However, airborne hyperspectral sensors are often relatively
expensive because of their limited spatial coverage, with multiple flight lines
typically being required to cover a study area. Further, a limitation such as;
cloud cover and data processing is often complex and can lead to significant
error. Similarly, due to technical and practical limitations such as challenging
signal-to-noise ratios (SNR) in particular, as well as bottom-of-atmosphere
reflectance, sensor cost, data volume and associated data processing costs and
time, until now, hyperspectral satellites have been poorly represented in
spaceborne missions compared to multispectral ones. Nonetheless, the future
of hyperspectral remote sensing is promising, considering forthcoming
launches of hyperspectral satellites. There are a number of hyperspectral
satellites planned to be launched in the near and distant future. These missions
will provide a better opportunity for bark beetle green attack detection (Figure
6.8).

119
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

Figure 6. 8: The main planned hyperspectral sensors to be launched in the next five
years.

EnMap, CHIME and HyspIRI commissioned hyperspectral sensors providing

more imagery alternatives, and their newly developed image processing
algorithms offering more analytical tools. Hyperspectral remote sensing is
positioned to become one of the core technologies for geospatial research,
exploration and monitoring of forest stress and insect infestation. For example,
EnMap, CHIME and HyspIRI can all collect data over visible to shortwave
infrared wavelengths (400–2,500 nm). This provision is important for the study
of bark beetle infestation as we demonstrated the potential of hyperspectral
measurement (400–2,500 nm) for detection of Ips typographus, L. green
attack Chapter (2). Furthermore, the HyspIRI satellite will be the first
hyperspectral sensor in space able to provide eight spectral bands in the
thermal infrared region from the 3–5- and 8–12-micron regions of the
spectrum. These satellites will offer a breakthrough regarding the control,
prevention and mitigation of large-scale bark beetle outbreaks. Our study
Chapter (4) revealed the importance of thermal bands in detecting canopy
stress induced by bark beetle infestation at a very early stage. Similarly, TIR
data were found to have a stronger correlation and able to better represent
the stress signals from leaf thermal properties (such as leaf water content,
stomatal conductance and chlorophyll fluorescence) under bark beetle
infestation.

Furthermore, other remote-sensing platforms such as unmanned aerial

vehicles (UAVs) as a new means of remote sensing offer significant potential
concerning future remote-sensing applications in forest management.
Currently, UAVs can collect data at low altitude and provide high-quality data
comparable to airborne sensors. The advantages of UAV-collected data include

120
 Chapter 6

low cost, high accuracy abundant data, real-time capability and the ability for
rapid data acquisition while transmitting image, video and orientation data in
real time to ground control stations (Table.6.1).

The UAV technology has developed at a significantly rapid pace, and today,
different sensors can be attached to UAVs onboard such as positioning sensors
(GPS), inertial navigation sensors (INS), micro-electro-mechanical systems
(MEMS), gyroscopes, accelerometers and altitude sensors (AS), all of which
allow for the realization of remote sensing missions. Moreover, various sensors
such as multispectral, hyperspectral, LIDAR and thermal cameras can be
installed onto UAV platforms. As a result, recent data collected by UAVs are
widely used for forest management practices.

The current limitation of UAVs is related to their flight time (i.e., limited battery
power) and processing time of the imagery. However, there is growing interest
in the forest management community in using UAVs as a promising and
decision-support tool for forest disturbance studies, including bark beetle
infestation management. Soon UAV can potentially fill the gap between
satellite/airborne platforms with ‘ground truth’ data collected using ASD field
spectrometers to study detection of bark beetle at the green attack stage.

Table 6. 1: Comparison between UAV and other remote-sensing platforms.

121
Synthesis: Remote sensing of bark beetle (Ips typographus, L) green attack

6.10 Future research Avenues

The results of this thesis outline the potential of using remotely sensed data
for the detection of early-stage European spruce bark beetle infestation. In line
with field data measurements and employed statistical approaches in this
dissertation, further studies could be undertaken along two lines: to examine
the retrieval of other foliar properties under bark beetle green attack, and the
use of different remotely sensed datasets.

In this study, we revealed that the retrieval of foliar biochemical properties

(chlorophyll and nitrogen concentration) from hyperspectral measurements
decreases when the trees are under bark beetle infestation, and that other leaf
traits (e.g. leaf water content, stomatal conductance and leaf dry matter
content) is important equally in the study of bark beetle green attack.
Therefore, further research could be undertaken on other vegetation traits in
the future.

In Chapter 2, foliar hyperspectral measurements from ASD FieldSpec3 have

been applied. However, due to the unavailability of airborne hyperspectral data
(at the canopy level ) when the field survey was undertaken, optical and TIR
data from Sentinel-2, Landsat-8, RapidEye and SPOT-5 were used in next
Chapters. Further research is critical for exploring and comparing the utility of
airborne hyperspectral data in both optical and TIR domains for early detection
of bark beetle green attack. Another possible extension of this study would be
with respect to the use of the newly available hyperspectral camera mounted
on UAV, and future planned hyperspectral satellite like EnMap, CHIME and
HyspIRI these sensors able to record information for red-edge, SWIR infrared
and thermal infrared region.

122
Bibliography
Abdullah, H., Darvishzadeh, R., Skidmore, A.K., Groen, T.A. and Heurich, M.,
2018a. European spruce bark beetle ( Ips typographus, L.) green attack
affects foliar reflectance and biochemical properties. International
Journal of Applied Earth Observation and Geoinformation, 64: 199-209.
Abdullah, H., Skidmore, A.K., Darvishzadeh, R., Heurich, M., Pettorelli, N. and
Disney, M., 2018b. Sentinel-2 accurately maps green-attack stage of
European spruce bark beetle (Ips typographus, L.) compared with
Landsat-8. Remote Sensing in Ecology and Conservation.
Adler-Golden, S.M., Matthew, M.W., Bernstein, L.S., Levine, R.Y., Berk, A.,
Richtsmeier, S.C., Acharya, P.K., Anderson, G.P., Felde, J.W. and
Gardner, J., 1999. Atmospheric correction for shortwave spectral
imagery based on MODTRAN4, SPIE's International Symposium on
Optical Science, Engineering, and Instrumentation. International Society
for Optics and Photonics, pp. 61-69.
Ahern, F., 1988a. The effects of bark beetle stress on the foliar spectral
reflectance of lodgepole pine. International Journal of Remote Sensing,
9(9): 1451-1468.
Ahern, F.J., 1988b. The effects of bark beetle stress on the foliar spectral
reflectance of lodgepole pine. International Journal of Remote Sensing,
9(9): 1451-1468.
Aldakheel, Y. and Danson, F., 1997. Spectral reflectance of dehydrating leaves:
measurements and modelling. International journal of remote sensing,
18(17): 3683-3690.
Aldea, M., Hamilton, J.G., Resti, J.P., Zangerl, A.R. and Berenbaum, M.R.,
2005. Indirect effects of insect herbivory on leaf gas exchange in
soybean. Plant, Cell & Environment, 28(3): 402-411.
Ali, A.M., Darvishzadeh, R., Skidmore, A.K., Duren, I.v., Heiden, U. and
Heurich, M., 2016. Estimating leaf functional traits by inversion of
PROSPECT: Assessing leaf dry matter content and specific leaf area in
mixed mountainous forest. International Journal of Applied Earth
Observation and Geoinformation, 45: 66-76.
Allain, M., Nguyen, A., Johnson, E., Williams, E., Tsai, S., Prichard, S., Freed,
T. and Skiles, J., 2011. A Geospatial Assessment of Mountain Pine Beetle
Infestations and Their Effect on Forest Health in Okanogan-Wenatchee
National Forest. Proceedings of the American Society of Photogrammetry
and Remote Sensing. Bethesda, Maryland: American Society for
Photogrammetry and Remote Sensing: 20-31.
Arellano, P., Tansey, K., Balzter, H. and Boyd, D.S., 2015. Detecting the effects
of hydrocarbon pollution in the Amazon forest using hyperspectral
satellite images. Environ Pollut, 205: 225-39.

123
Bibliography

Atzberger, C. and Werner, W., 1998. Needle reflectance of healthy and

diseased Spruce stands, 1st EARSeL workshop on imaging spectroscopy,
pp. 271-283.
Ayala-Silva, T. and Beyl, C.A., 2005. Changes in spectral reflectance of wheat
leaves in response to specific macronutrient deficiency. Advances in
Space Research, 35(2): 305-317.
Bannari, A., Asalhi, H. and Teillet, P., 2002. Transformed difference vegetation
index (TDVI) for vegetation cover mapping, Geoscience and Remote
Sensing Symposium, 2002. IGARSS'02. 2002 IEEE International. IEEE,
pp. 3053-3055.
Bannari, A., Morin, D., Bonn, F. and Huete, A.R., 1995. A review of vegetation
indices. Remote Sensing Reviews, 13(1-2): 95-120.
Barnes, E., Clarke, T., Richards, S., Colaizzi, P., Haberland, J., Kostrzewski,
M., Waller, P., Choi, C., Riley, E. and Thompson, T., 2000. Coincident
detection of crop water stress, nitrogen status and canopy density using
ground based multispectral data.
Bässler, C., Förster, B., Moning, C. and Müller, J., 2008a. The BIOKLIM-Project:
biodiversity research between climate change and wilding in a temperate
montane forest–the conceptual framework. Waldökologie,
Landschaftsforschung und Naturschutz, 7: 21-33.
Bässler, C., Förster, B., Moning, C. and Müller, J., 2008b. The BIOKLIM-Project:
biodiversity research between climate change and wilding in a temperate
montane forest—the conceptual framework.
Bentz, B. and Endreson, D., 2003. Evaluating satellite imagery for estimating
mountain pine beetle-caused lodgepole pine mortality: current status,
Mountain Pine Beetle Symposium: Challenges and Solutions. Natural
Resources Canada, Canadian Forest Service, Pacific Forestry Centre,
Kelowna, British Columbia, pp. 154-163.
Bentz, B.J., Régnière, J., Fettig, C.J., Hansen, E.M., Hayes, J.L., Hicke, J.A.,
Kelsey, R.G., Negrón, J.F. and Seybold, S.J., 2010. Climate Change and
Bark Beetles of the Western United States and Canada: Direct and
Indirect Effects. BioScience, 60(8): 602-613.
Berni, J.A., Zarco-Tejada, P.J., González-Dugo, V. and Fereres, E., 2009.
Remote sensing of thermal water stress indicators in peach, VII
International Peach Symposium 962, pp. 325-331.
Beudert, B., Bässler, C., Thorn, S., Noss, R., Schröder, B., Dieffenbach-Fries,
H., Foullois, N. and Müller, J., 2015. Bark Beetles Increase Biodiversity
While Maintaining Drinking Water Quality. Conservation Letters, 8(4):
272-281.
Bowman, W.D., 1989. The relationship between leaf water status, gas
exchange, and spectral reflectance in cotton leaves. Remote Sensing of
Environment, 30(3): 249-255.
Bright, B.C., Hicke, J.A. and Meddens, A.J.H., 2013. Effects of bark beetle-
caused tree mortality on biogeochemical and biogeophysical MODIS

124
 Bibliography

products. Journal of Geophysical Research: Biogeosciences, 118(3):

974-982.
Buitrago Acevedo, M.F., Groen, T.A., Hecker, C.A. and Skidmore, A.K., 2017.
Identifying leaf traits that signal stress in TIR spectra. ISPRS Journal of
Photogrammetry and Remote Sensing, 125: 132-145.
Buitrago, M.F., Groen, T.A., Hecker, C.A. and Skidmore, A.K., 2016. Changes
in thermal infrared spectra of plants caused by temperature and water
stress. ISPRS Journal of Photogrammetry and Remote Sensing, 111: 22-
31.
Cailleret, M., Heurich, M. and Bugmann, H., 2014a. Reduction in browsing
intensity may not compensate climate change effects on tree species
composition in the Bavarian Forest National Park. Forest Ecology and
Management, 328: 179-192.
Cailleret, M., Heurich, M. and Bugmann, H., 2014b. Reduction in browsing
intensity may not compensate climate change effects on tree species
composition in the Bavarian Forest National Park. Forest Ecology and
Management, 328(Supplement C): 179-192.
Campbell, P., Middleton, E., McMurtrey, J. and Chappelle, E., 2007.
Assessment of vegetation stress using reflectance or fluorescence
measurements. Journal of environmental quality, 36(3): 832-845.
Carlson, T.N. and Ripley, D.A., 1997. On the relation between NDVI, fractional
vegetation cover, and leaf area index. Remote sensing of Environment,
62(3): 241-252.
Carol, L.J., Paul, R.W., Niels, O.M., Marvin, L.S. and Roshani, J., 2004.
Estimating Water Stress in Plants Using Hyperspectral Sensing.
Carrascal, L.M., Galván, I. and Gordo, O., 2009. Partial least squares
regression as an alternative to current regression methods used in
ecology. Oikos, 118(5): 681-690.
Carter, G.A., 1993. Responses of leaf spectral reflectance to plant stress.
American Journal of Botany: 239-243.
Carter, G.A. and Knapp, A.K., 2001. Leaf optical properties in higher plants:
linking spectral characteristics to stress and chlorophyll concentration.
American journal of botany, 88(4): 677-684.
Carter, G.A. and Miller, R.L., 1994. Early detection of plant stress by digital
imaging within narrow stress-sensitive wavebands. Remote Sensing of
Environment, 50(3): 295-302.
Carter, G.A., Seal, M.R. and Haley, T., 1998. Airborne detection of southern
pine beetle damage using key spectral bands. Canadian Journal of Forest
Research, 28(7): 1040-1045.
Ceccato, P., Gobron, N., Flasse, S., Pinty, B. and Tarantola, S., 2002. Designing
a spectral index to estimate vegetation water content from remote
sensing data: Part 1: Theoretical approach. Remote Sensing of
Environment, 82(2–3): 188-197.

125
Bibliography

Chaves, M.M., Maroco, J.P. and Pereira, J.S., 2003. Understanding plant
responses to drought—from genes to the whole plant. Functional plant
biology, 30(3): 239-264.
Chen, G. and Meentemeyer, R., 2016. Remote Sensing of Forest Damage by
Diseases and Insects, 145-162 pp.
Cheng, T., Rivard, B., Sánchez-Azofeifa, G.A., Feng, J. and Calvo-Polanco, M.,
2010. Continuous wavelet analysis for the detection of green attack
damage due to mountain pine beetle infestation. Remote Sensing of
Environment, 114(4): 899-910.
Chinellato, F., Faccoli, M., Marini, L. and Battisti, A., 2014. Distribution of
Norway spruce bark and wood-boring beetles along Alpine elevational
gradients. Agricultural and Forest Entomology, 16(2): 111-118.
Chong, I.-G. and Jun, C.-H., 2005. Performance of some variable selection
methods when multicollinearity is present. Chemometrics and Intelligent
Laboratory Systems, 78(1-2): 103-112.
Christiansen, E. and Bakke, A., 1988a. The spruce bark beetle of Eurasia,
Dynamics of forest insect populations. Springer, pp. 479-503.
Christiansen, E. and Bakke, A., 1988b. The Spruce Bark Beetle of Eurasia. In:
A.A. Berryman (Editor), Dynamics of Forest Insect Populations: Patterns,
Causes, Implications. Springer US, Boston, MA, pp. 479-503.
Cohen, W.B., 1991. Response of vegetation indices to changes in three
measures of leaf water stress.
Collins, J.B. and Woodcock, C.E., 1996. An assessment of several linear change
detection techniques for mapping forest mortality using multitemporal
landsat TM data. Remote Sensing of Environment, 56(1): 66-77.
Colombo, R., Meroni, M., Marchesi, A., Busetto, L., Rossini, M., Giardino, C.
and Panigada, C., 2008. Estimation of leaf and canopy water content in
poplar plantations by means of hyperspectral indices and inverse
modeling. Remote Sensing of Environment, 112(4): 1820-1834.
Coops, N.C., Johnson, M., Wulder, M.A. and White, J.C., 2006. Assessment of
QuickBird high spatial resolution imagery to detect red attack damage
due to mountain pine beetle infestation. Remote Sensing of
Environment, 103(1): 67-80.
Coulson, R.N., Amman, G.D., Dahlsten, D.L., DeMars Jr, C. and Stephen, F.,
1985. Forest-bark beetle interactions: bark beetle population dynamics.
Integrated pest management in pine-bark beetle ecosystems. John
Wiley & Sons, New York: 61-80.
Cudmore, T.J., Björklund, N., Carroll, A.L. and Lindgren, B.S., 2010. Climate
change and range expansion of an aggressive bark beetle: evidence of
higher beetle reproduction in naïve host tree populations. Journal of
Applied Ecology, 47(5): 1036-1043.
Curran, P.J., 1989. Remote sensing of foliar chemistry. Remote sensing of
Environment, 30(3): 271-278.

126
 Bibliography

Curran, P.J., Dungan, J.L., Macler, B.A., Plummer, S.E. and Peterson, D.L.,
1992. Reflectance spectroscopy of fresh whole leaves for the estimation
of chemical concentration. Remote Sensing of Environment, 39(2): 153-
166.
Darvishzadeh, R., Skidmore, A., Schlerf, M. and Atzberger, C., 2008. Inversion
of a radiative transfer model for estimating vegetation LAI and
chlorophyll in a heterogeneous grassland. Remote Sensing of
Environment, 112(5): 2592-2604.
Daszykowski, M., Serneels, S., Kaczmarek, K., Van Espen, P., Croux, C. and
Walczak, B., 2007. TOMCAT: A MATLAB toolbox for multivariate
calibration techniques. Chemometrics and intelligent laboratory
systems, 85(2): 269-277.
Daughtry, C., Biehl, L. and Ranson, K., 1989. A new technique to measure the
spectral properties of conifer needles. Remote Sensing of Environment,
27(1): 81-91.
Demetriades-Shah, T.H., Steven, M.D. and Clark, J.A., 1990. High resolution
derivative spectra in remote sensing. Remote Sensing of Environment,
33(1): 55-64.
Deshayes, M., Guyon, D., Jeanjean, H., Stach, N., Jolly, A. and Hagolle, O.,
2006. The contribution of remote sensing to the assessment of drought
effects in forest ecosystems. Annals of Forest Science, 63(6): 579-595.
Doughty, C.E., Field, C.B. and McMillan, A.M., 2011. Can crop albedo be
increased through the modification of leaf trichomes, and could this cool
regional climate? Climatic Change, 104(2): 379-387.
Dozier, J. and Warren, S.G., 1982. Effect of viewing angle on the infrared
brightness temperature of snow. Water Resources Research, 18(5):
1424-1434.
Dupke, C., Bonenfant, C., Reineking, B., Hable, R., Zeppenfeld, T., Ewald, M.
and Heurich, M., 2017. Habitat selection by a large herbivore at multiple
spatial and temporal scales is primarily governed by food resources.
Ecography, 40(8): 1014-1027.
Dye, M., Mutanga, O. and Ismail, R., 2008. Detecting the severity of
woodwasp, Sirex noctilio, infestation in a pine plantation in KwaZulu-
Natal, South Africa, using texture measures calculated from high spatial
resolution imagery. African Entomology, 16(2): 263-275.
Egan, J.M., Jacobi, W.R., Negron, J.F., Smith, S.L. and Cluck, D.R., 2010.
Forest thinning and subsequent bark beetle-caused mortality in
Northeastern California. Forest Ecology and Management, 260(10):
1832-1842.
Eidmann, H., 1992. Impact of bark beetles on forests and forestry in Sweden.
Journal of Applied Entomology, 114(1-5): 193-200.
Eitel, J.U., Gessler, P.E., Smith, A.M. and Robberecht, R., 2006. Suitability of
existing and novel spectral indices to remotely detect water stress in
Populus spp. Forest Ecology and Management, 229(1-3): 170-182.

127
Bibliography

Eitel, J.U.H., Vierling, L.A., Litvak, M.E., Long, D.S., Schulthess, U., Ager, A.A.,
Krofcheck, D.J. and Stoscheck, L., 2011. Broadband, red-edge
information from satellites improves early stress detection in a New
Mexico conifer woodland. Remote Sensing of Environment, 115(12):
3640-3646.
Escadafal, R., Belghith, A. and Ben Moussa, H., 1994a. Indices spectraux pour
la dégradation des milieux naturels en Tunisie aride, 6ème Symp. Int.
Mesures Physiques et Signatures en Télédétection, ISPRS-CNES, Val
d’lsère, France, pp. 253–259.
Escadafal, R., Belghith, A. and Ben Moussa, H., 1994b. Indices spectraux pour
la télédétection de la dégradation des milieux naturels en Tunisie aride.
Ewers, B., Mackay, D. and Samanta, S., 2007. Interannual consistency in
canopy stomatal conductance control of leaf water potential across seven
tree species. Tree Physiology, 27(1): 11-24.
Fabre, S., Lesaignoux, A., Olioso, A. and Briottet, X., 2011. Influence of Water
Content on Spectral Reflectance of Leaves in the 3–15-$\mu\hbox {m}
$ Domain. IEEE Geoscience and Remote Sensing Letters, 8(1): 143-147.
Faccoli, M. and Bernardinelli, I., 2014. Composition and Elevation of Spruce
Forests Affect Susceptibility to Bark Beetle Attacks: Implications for
Forest Management. Forests, 5(1): 88-102.
Fahse, L. and Heurich, M., 2011. Simulation and analysis of outbreaks of bark
beetle infestations and their management at the stand level. Ecological
Modelling, 222(11): 1833-1846.
Farrés, M., Platikanov, S., Tsakovski, S. and Tauler, R., 2015. Comparison of
the variable importance in projection (VIP) and of the selectivity ratio
(SR) methods for variable selection and interpretation. Journal of
Chemometrics, 29(10): 528-536.
Fassnacht, F.E., Latifi, H., Ghosh, A., Joshi, P.K. and Koch, B., 2014. Assessing
the potential of hyperspectral imagery to map bark beetle-induced tree
mortality. Remote Sensing of Environment, 140: 533-548.
Fensholt, R. and Sandholt, I., 2003. Derivation of a shortwave infrared water
stress index from MODIS near- and shortwave infrared data in a semiarid
environment. Remote Sensing of Environment, 87(1): 111-121.
Feret, J.-B., François, C., Asner, G.P., Gitelson, A.A., Martin, R.E., Bidel, L.P.R.,
Ustin, S.L., le Maire, G. and Jacquemoud, S., 2008. PROSPECT-4 and 5:
Advances in the leaf optical properties model separating photosynthetic
pigments. Remote Sensing of Environment, 112(6): 3030-3043.
Filchev, L., 2012a. An assessment of European spruce bark beetle infestation
using Worldview-2 satellite data, Proc. of European SCGIS Conf.:“Best
practices: Application of GIS technologies for conservation of natural and
cultural heritage sites, pp. 9-16.
Filchev, L., 2012b. An assessment of european spruce bark beetle infestation
using WorldView-2 Satellite data, Proceedings of 1st European SCGIS
Conference with International Participation―Best Practices: Application

128
 Bibliography

of GIS Technologies for Conservation of Natural and Cultural Heritage

Sites‖(SCGIS-Bulgaria, Sofia), Sofia, Bulgaria, pp. 21-23.
Foster, A.C., Walter, J.A., Shugart, H.H., Sibold, J. and Negron, J., 2017.
Spectral evidence of early-stage spruce beetle infestation in Engelmann
spruce. Forest Ecology and Management, 384: 347-357.
Franklin, S., Wulder, M., Skakun, R. and Carroll, A., 2003a. Mountain pine
beetle red-attack forest damage classification using stratified Landsat
TM data in British Columbia, Canada. Photogrammetric Engineering &
Remote Sensing, 69(3): 283-288.
Franklin, S.E., Wulder, M.A., Skakun, R.S. and Carroll, A.L., 2003b. Mountain
Pine Beetle Red-Attack Forest Damage Classification Using Stratified
Landsat TM Data in British Columbia, Canada. Photogrammetric
Engineering & Remote Sensing, 69(3): 283-288.
Fu, J., Zhou, Q., Liu, J., Liu, W., Wang, T., Zhang, Q. and Jiang, G., 2008. High
levels of heavy metals in rice (Oryzasativa L.) from a typical E-waste
recycling area in southeast China and its potential risk to human health.
Chemosphere, 71(7): 1269-1275.
Galvao, L.S., Formaggio, A.R. and Tisot, D.A., 2005. Discrimination of
sugarcane varieties in Southeastern Brazil with EO-1 Hyperion data.
Remote Sensing of Environment, 94(4): 523-534.
Gamon, J. and Surfus, J., 1999. Assessing leaf pigment content and activity
with a reflectometer. The New Phytologist, 143(1): 105-117.
Gartland, L.M., 2012. Heat islands: understanding and mitigating heat in urban
areas. Routledge.
Geladi, P. and Kowalski, B.R., 1986. Partial least-squares regression: a tutorial.
Analytica chimica acta, 185: 1-17.
Gersony, J.T., Prager, C.M., Boelman, N.T., Eitel, J.U.H., Gough, L., Greaves,
H.E., Griffin, K.L., Magney, T.S., Sweet, S.K., Vierling, L.A. and Naeem,
S., 2016. Scaling Thermal Properties from the Leaf to the Canopy in the
Alaskan Arctic Tundra. Arctic, Antarctic, and Alpine Research, 48(4):
739-754.
Gimbarzevsky, P., Dawson, A.F. and Van Sickle, G., 1992a. Assessment of
aerial photographs and multi-spectral scanner imagery for measuring
mountain pine beetle damage, 333.
Gimbarzevsky, P., Dawson, A.F. and Van Sickle, G.A., 1992b. Assessment of
aerial photographs and multi-spectral scanner imagery for measuring
mountain pine beetle damage, 333.
Gitelson, A.A., Buschmann, C. and Lichtenthaler, H.K., 1999. The chlorophyll
fluorescence ratio F 735/F 700 as an accurate measure of the chlorophyll
content in plants. Remote Sensing of Environment, 69(3): 296-302.
Gitelson, A.A., Kaufman, Y.J. and Merzlyak, M.N., 1996. Use of a green channel
in remote sensing of global vegetation from EOS-MODIS. Remote
Sensing of Environment, 58(3): 289-298.

129
Bibliography

Gitelson, A.A., Kaufman, Y.J., Stark, R. and Rundquist, D., 2002. Novel
algorithms for remote estimation of vegetation fraction. Remote sensing
of Environment, 80(1): 76-87.
Gitelson, A.A., Keydan, G.P. and Merzlyak, M.N., 2006. Three-band model for
noninvasive estimation of chlorophyll, carotenoids, and anthocyanin
contents in higher plant leaves. Geophysical research letters, 33(11).
Gitelson, A.A., Viña, A., Arkebauer, T.J., Rundquist, D.C., Keydan, G. and
Leavitt, B., 2003. Remote estimation of leaf area index and green leaf
biomass in maize canopies. Geophysical Research Letters, 30(5).
Giunta, A., Jenkins, M., Hebertson, E. and Munson, A., 2016. Disturbance
Agents and Their Associated Effects on the Health of Interior Douglas-
Fir Forests in the Central Rocky Mountains. Forests, 7(4): 80.
Gobron, N., Pinty, B., Verstraete, M.M. and Widlowski, J.-L., 2000. Advanced
vegetation indices optimized for up-coming sensors: Design,
performance, and applications. IEEE Transactions on Geoscience and
Remote Sensing, 38(6): 2489-2505.
Goel, N.S. and Qin, W., 1994. Influences of canopy architecture on
relationships between various vegetation indices and LAI and FPAR: A
computer simulation. Remote Sensing Reviews, 10(4): 309-347.
Goheen, D. and Hansen, E., 1993. Effects of pathogens and bark beetles on
forests. In: T.S.G. Filip (Editor), Beetle Pathogen Interactions in Conifer
Forests. Academic Press, London, pp. 175-196.
Göthlin, E., Schroeder, L.M. and Lindelöw, A., 2000. Attacks by Ips
typographus and Pityogenes chalcographus on windthrown spruces
(Picea abies) during the two years following a storm felling. Scandinavian
Journal of Forest Research, 15(5): 542-549.
Haboudane, D., Miller, J.R., Pattey, E., Zarco-Tejada, P.J. and Strachan, I.B.,
2004. Hyperspectral vegetation indices and novel algorithms for
predicting green LAI of crop canopies: Modeling and validation in the
context of precision agriculture. Remote sensing of environment, 90(3):
337-352.
Hais, M., Jonášová, M., Langhammer, J. and Kučera, T., 2009. Comparison of
two types of forest disturbance using multitemporal Landsat TM/ETM+
imagery and field vegetation data. Remote Sensing of Environment,
113(4): 835-845.
Hais, M. and Kučera, T., 2008. Surface temperature change of spruce forest as
a result of bark beetle attack: remote sensing and GIS approach.
European Journal of Forest Research, 127(4): 327-336.
Hancock, D.W. and Dougherty, C.T., 2007. Relationships between blue-and
red-based vegetation indices and leaf area and yield of alfalfa. Crop
science, 47(6): 2547-2556.
Hardinsky, M.A. and Lemas, V., 1983. The influence of soil salinity, growth
form, and leaf moisture on the spectral reflectance of Spartina

130
 Bibliography

alternifolia canopies. Photogrammetric Engineering and Remote

Sensing, 49: 77-83.
Hart, S.J., Veblen, T.T., Mietkiewicz, N. and Kulakowski, D., 2015. Negative
feedbacks on bark beetle outbreaks: widespread and severe spruce
beetle infestation restricts subsequent infestation. PLoS One, 10(5):
e0127975.
Hatala, J.A., Crabtree, R.L., Halligan, K.Q. and Moorcroft, P.R., 2010.
Landscape-scale patterns of forest pest and pathogen damage in the
Greater Yellowstone Ecosystem. Remote Sensing of Environment,
114(2): 375-384.
Havašová, M., Bucha, T., Ferenčík, J. and Jakuš, R., 2015. Applicability of a
vegetation indices-based method to map bark beetle outbreaks in the
High Tatra Mountains. Annals of Forest Research, 58(2).
Heath, J., 2001a. The detection of mountain pine beetle green attacked
lodgepole pine using Compact Airborne Spectrographic Imager (CASI)
data, University of British Columbia.
Heath, J., 2001b. The detection of mountain pine beetle green attacked
lodgepole pine using Compact Airborne Spectrographic Imager (CASI)
data, University of British Columbia, Vancouver, BC, Canada, 1-72 pp.
Heinze, M. and Fiedler, H.J., 1976. Beziehungen des Chlorophyllgehaltes zu
Standortsfaktoren, Ernährungszustand und Wachstum bei Koniferen.
Flora, 165(3): 269-293.
Hendry, G.A., Houghton, J.D. and Brown, S.B., 1987. The degradation of
chlorophyll—a biological enigma. New Phytologist, 107(2): 255-302.
Heurich, M., Beudert, B., Rall, H. and Křenová, Z., 2010a. National parks as
model regions for interdisciplinary long-term ecological research: the
Bavarian Forest and Šumavá National Parks underway to transboundary
ecosystem research, Long-Term Ecological Research. Springer, pp. 327-
344.
Heurich, M., Ochs, T., Andresen, T. and Schneider, T., 2009. Object-orientated
image analysis for the semi-automatic detection of dead trees following
a spruce bark beetle (Ips typographus) outbreak. European Journal of
Forest Research, 129(3): 313-324.
Heurich, M., Ochs, T., Andresen, T. and Schneider, T., 2010b. Object-
orientated image analysis for the semi-automatic detection of dead trees
following a spruce bark beetle (Ips typographus) outbreak. European
Journal of Forest Research, 129(3): 313-324.
Holm, S., 1979. A simple sequentially rejective multiple test procedure.
Scandinavian journal of statistics: 65-70.
Hotelling, H., 1933. Analysis of a complex of statistical variables into principal
components. Journal of educational psychology, 24(6): 417.
Hunt, E.R., Daughtry, C., Eitel, J.U. and Long, D.S., 2011. Remote sensing leaf
chlorophyll content using a visible band index. Agronomy Journal,
103(4): 1090-1099.

131
Bibliography

Hunt, E.R. and Rock, B.N., 1989. Detection of changes in leaf water content
using near-and middle-infrared reflectances. Remote sensing of
environment, 30(1): 43-54.
Immitzer, M. and Atzberger, C., 2014. Early Detection of Bark Beetle
Infestation in Norway Spruce (<I>Picea abies</I>, L.) using WorldView-
2 Data<BR> Frühzeitige Erkennung von Borkenkä ferbefall an Fichten
mittels WorldView-2 Satellitendaten. Photogrammetrie - Fernerkundung
- Geoinformation, 2014(5): 351-367.
Ismail, R. and Mutanga, O., 2010. A comparison of regression tree ensembles:
Predicting Sirex noctilio induced water stress in Pinus patula forests of
KwaZulu-Natal, South Africa. International Journal of Applied Earth
Observation and Geoinformation, 12: S45-S51.
Jackson, T., 2004. Vegetation water content mapping using Landsat data
derived normalized difference water index for corn and soybeans.
Remote Sensing of Environment, 92(4): 475-482.
Jacquemoud, S. and Baret, F., 1990. PROSPECT: A model of leaf optical
properties spectra. Remote sensing of environment, 34(2): 75-91.
Jang, J.D., Viau, A.A. and Anctil, F., 2006. Thermal-water stress index from
satellite images. International Journal of Remote Sensing, 27(8): 1619-
1639.
Jensen, J.R., 2009. Remote sensing of the environment: An earth resource
perspective 2/e. Pearson Education India.
Jiang, Z., Huete, A.R., Didan, K. and Miura, T., 2008. Development of a two-
band enhanced vegetation index without a blue band. Remote sensing
of Environment, 112(10): 3833-3845.
Jönsson, A.M., Appelberg, G., Harding, S. and Bärring, L., 2009. Spatio-
temporal impact of climate change on the activity and voltinism of the
spruce bark beetle, Ips typographus. Global Change Biology, 15(2): 486-
499.
Junttila, S., Vastaranta, M., Hämäläinen, J., Latva-käyrä, P., Holopainen, M.,
Hernández Clemente, R., Hyyppä, H. and Navarro-Cerrillo, R.M., 2016.
Effect of forest structure and health on the relative surface temperature
captured by airborne thermal imagery – Case study in Norway Spruce-
dominated stands in Southern Finland. Scandinavian Journal of Forest
Research, 32(2): 154-165.
Kim, Y., Still, C.J., Hanson, C.V., Kwon, H., Greer, B.T. and Law, B.E., 2016.
Canopy skin temperature variations in relation to climate, soil
temperature, and carbon flux at a ponderosa pine forest in central
Oregon. Agricultural and Forest Meteorology, 226-227: 161-173.
Klein, W.H., 1973. Beetle-killed pine estimates. Photogrammetric Engineering,
39(4).
Kováč, D., Navrátil, M., Malenovský, Z., Štroch, M., Špunda, V. and Urban, O.,
2012. Reflectance continuum removal spectral index tracking the

132
 Bibliography

xanthophyll cycle photoprotective reactions in Norway spruce needles.

Functional Plant Biology, 39(12): 987.
Krofcheck, D.J., Eitel, J.U.H., Vierling, L.A., Schulthess, U., Hilton, T.M.,
Dettweiler-Robinson, E., Pendleton, R. and Litvak, M.E., 2014. Detecting
mortality induced structural and functional changes in a piñon-juniper
woodland using Landsat and RapidEye time series. Remote Sensing of
Environment, 151: 102-113.
Krokene, P. and Solheim, H., 1998. Pathogenicity of four blue-stain fungi
associated with aggressive and nonaggressive bark beetles.
Phytopathology, 88(1): 39-44.
Kümmerlen, B., Dauwe, S., Schmundt, D. and Schurr, U., 1999. Thermography
to measure water relations of plant leaves. Handbook of computer vision
and applications, 3: 763-781.
Kurz, W.A., Dymond, C.C., Stinson, G., Rampley, G.J., Neilson, E.T., Carroll,
A.L., Ebata, T. and Safranyik, L., 2008. Mountain pine beetle and forest
carbon feedback to climate change. Nature, 452(7190): 987-90.
Larcher, W., 2003. Physiological plant ecology: ecophysiology and stress
physiology of functional groups. Springer Science & Business Media.
Latifi, H., Schumann, B., Kautz, M. and Dech, S., 2014. Spatial characterization
of bark beetle infestations by a multidate synergy of SPOT and Landsat
imagery. Environ Monit Assess, 186(1): 441-56.
Laurent, E.J., Shi, H., Gatziolis, D., LeBouton, J.P., Walters, M.B. and Liu, J.,
2005. Using the spatial and spectral precision of satellite imagery to
predict wildlife occurrence patterns. Remote Sensing of Environment,
97(2): 249-262.
Lausch, A., Fahse, L. and Heurich, M., 2011. Factors affecting the spatio-
temporal dispersion of Ips typographus (L.) in Bavarian Forest National
Park: A long-term quantitative landscape-level analysis. Forest Ecology
and Management, 261(2): 233-245.
Lausch, A., Heurich, M. and Fahse, L., 2013a. Spatio-temporal infestation
patterns of Ips typographus (L.) in the Bavarian Forest National Park,
Germany. Ecological Indicators, 31: 73-81.
Lausch, A., Heurich, M., Gordalla, D., Dobner, H.J., Gwillym-Margianto, S. and
Salbach, C., 2013b. Forecasting potential bark beetle outbreaks based
on spruce forest vitality using hyperspectral remote-sensing techniques
at different scales. Forest Ecology and Management, 308: 76-89.
Lawrence, R. and Labus, M., 2003. Early detection of Douglas-fir beetle
infestation with subcanopy resolution hyperspectral imagery. Western
Journal of Applied Forestry, 18(3): 202-206.
Lehnert, L.W., Bässler, C., Brandl, R., Burton, P.J. and Müller, J., 2013.
Conservation value of forests attacked by bark beetles: Highest number
of indicator species is found in early successional stages. Journal for
Nature Conservation, 21(2): 97-104.

133
Bibliography

Li, H., Xu, Q. and Liang, Y., 2014. libPLS: an integrated library for partial least
squares regression and discriminant analysis. PeerJ PrePrints, 2:
e190v1.
Lichtenthaler, H.K., 1987. [34] Chlorophylls and carotenoids: pigments of
photosynthetic biomembranes. Methods in enzymology, 148: 350-382.
Lieutier, F., 2002. Mechanisms of resistance in conifers and bark beetle attack
strategies, Mechanisms and deployment of resistance in trees to insects.
Springer, pp. 31-77.
Lindenmayer, D. and Franklin, J.F., 2002. Conserving forest biodiversity: a
comprehensive multiscaled approach. Island Press.
Liu, L. and Zhang, Y., 2011. Urban heat island analysis using the Landsat TM
data and ASTER data: A case study in Hong Kong. Remote Sensing, 3(7):
1535-1552.
Lobinger, G., 1994a. Air temperature as a limiting factor for flight activity of
two species of pine bark beetles, Ips typographus L. and Pityogenes
chalcographus L.(Col., Scolytidae). Anzeiger für Schädlingskunde,
Pflanzenschutz, Umweltschutz, 67(1): 14-17.
Lobinger, G., 1994b. Die Lufttemperatur als limitierender Faktor für die
Schwärmaktivität zweier rindenbrütender Fichtenborkenkäferarten, lps
typographus L. undPityogenes chalcographus L.(Col., Scolytidae).
Anzeiger für Schädlingskunde, 67(1): 14-17.
Lobinger, G. and Skatulla, U., 1996. Untersuchungen zum Einfluss von
Sonnenlicht auf das Schwärmverhalten von Borkenkäfern. Anzeiger für
Schädlingskunde, Pflanzenschutz, Umweltschutz, 69(8): 183-185.
Lottering, R., Mutanga, O. and Peerbhay, K., 2016. Detecting and mapping
levels of Gonipterus scutellatus-induced vegetation defoliation and leaf
area index using spatially optimized vegetation indices. Geocarto
International: 1-16.
Mahlein, A.-K., Rumpf, T., Welke, P., Dehne, H.-W., Plümer, L., Steiner, U. and
Oerke, E.-C., 2013. Development of spectral indices for detecting and
identifying plant diseases. Remote Sensing of Environment, 128: 21-30.
Maitra, S. and Yan, J., 2008. Principle component analysis and partial least
squares: Two dimension reduction techniques for regression. Applying
Multivariate Statistical Models, 79: 79-90.
Malenovský, Z., Albrechtová, J., Lhotáková, Z., Zurita-Milla, R., Clevers, J.,
Schaepman, M. and Cudlín, P., 2006. Applicability of the PROSPECT
model for Norway spruce needles. International Journal of Remote
Sensing, 27(24): 5315-5340.
Mandanici, E. and Bitelli, G., 2016. Preliminary Comparison of Sentinel-2 and
Landsat 8 Imagery for a Combined Use. Remote Sensing, 8(12): 1014.
Marini, L., Økland, B., Jönsson, A.M., Bentz, B., Carroll, A., Forster, B.,
Grégoire, J.-C., Hurling, R., Nageleisen, L.M., Netherer, S., Ravn, H.P.,
Weed, A. and Schroeder, M., 2016. Climate drivers of bark beetle
outbreak dynamics in Norway spruce forests. Ecography: n/a-n/a.

134
 Bibliography

Markham, B.L. and Barker, J.L., 1986. Landsat MSS and TM post-calibration
dynamic ranges, exoatmospheric reflectances and at-satellite
temperatures. EOSAT Landsat technical notes, 1(1): 3-8.
Marx, A., 2010. Detection and classification of bark beetle infestation in pure
norway spruce stands with multi-temporal RapidEye imagery and data
mining techniques. Photogrammetrie-Fernerkundung-Geoinformation,
2010(4): 243-252.
Marx, A. and an der Havel, B., 2010a. Erkennung von borkenkäferbefall in
fichtenreinbeständen mit multi-temporalen rapideye-satellitenbildern
und datamining-techniken.
Marx, A. and an der Havel, B., 2010b. Erkennung von borkenkäferbefall in
fichtenreinbeständen mit multi-temporalen rapideye-satellitenbildern
und datamining-techniken. Photogramm. Fernerk. Geoinf, 4: 243-252.
McFarlane, B.L. and Witson, D.O., 2008. Perceptions of ecological risk
associated with mountain pine beetle (Dendroctonus ponderosae)
infestations in Banff and Kootenay National Parks of Canada. Risk
Analysis: An International Journal, 28(1): 203-212.
McFarlane, J., Watson, R.D., Theisen, A.F., Jackson, R.D., Ehrler, W., Pinter,
P., Idso, S.B. and Reginato, R., 1980. Plant stress detection by remote
measurement of fluorescence. Applied Optics, 19(19): 3287-3289.
Meddens, A.J., 2012. Mapping and detecting bark beetle-caused tree mortality
in the western United States. AAT 3524572 Thesis, University of Idaho,
Dissertation Abstracts International.
Meddens, A.J.H., Hicke, J.A., Vierling, L.A. and Hudak, A.T., 2013. Evaluating
methods to detect bark beetle-caused tree mortality using single-date
and multi-date Landsat imagery. Remote Sensing of Environment, 132:
49-58.
Mesarch, M.A., Walter-Shea, E.A., Asner, G.P., Middleton, E.M. and Chan, S.S.,
1999. A revised measurement methodology for conifer needles spectral
optical properties: evaluating the influence of gaps between elements.
Remote Sensing of Environment, 68(2): 177-192.
Méthy, M., Olioso, A. and Trabaud, L., 1994. Chlorophyll fluorescence as a tool
for management of plant resources. Remote sensing of environment,
47(1): 2-9.
Metternicht, G., 2003. Vegetation indices derived from high-resolution airborne
videography for precision crop management. International Journal of
Remote Sensing, 24(14): 2855-2877.
Mikkelson, K.M., Bearup, L.A., Maxwell, R.M., Stednick, J.D., McCray, J.E. and
Sharp, J.O., 2013. Bark beetle infestation impacts on nutrient cycling,
water quality and interdependent hydrological effects. Biogeochemistry,
115(1-3): 1-21.
Miller, Wu, J.Y., Boyer, M.G., Belanger, M. and Hare, E.W., 1991. SEASONAL
PATTERNS IN LEAF REFLECTANCE RED-EDGE CHARACTERISTICS.
INTERNATIONAL JOURNAL OF REMOTE SENSING, 12(7): 1509-1523.

135
Bibliography

Mirzaie, M., Darvishzadeh, R., Shakiba, A., Matkan, A.A., Atzberger, C. and
Skidmore, A., 2014. Comparative analysis of different uni- and multi-
variate methods for estimation of vegetation water content using hyper-
spectral measurements. International Journal of Applied Earth
Observation and Geoinformation, 26: 1-11.
Module, F., Atmospheric correction module: Quac and flaash user’s guide.
Modzelewska, A., Stereńczak, K., Mierczyk, M., Maciuk, S., Bałazy, R. and
Zawiła-Niedźwiecki, T., 2017. Sensitivity of vegetation indices in relation
to parameters of Norway spruce stands. Folia Forestalia Polonica, 59(2).
Moller, M., Alchanatis, V., Cohen, Y., Meron, M., Tsipris, J., Naor, A., Ostrovsky,
V., Sprintsin, M. and Cohen, S., 2007. Use of thermal and visible imagery
for estimating crop water status of irrigated grapevine. J Exp Bot, 58(4):
827-38.
Morris, J.L., Cottrell, S., Fettig, C.J., Hansen, W.D., Sherriff, R.L., Carter, V.A.,
Clear, J.L., Clement, J., DeRose, R.J., Hicke, J.A., Higuera, P.E., Mattor,
K.M., Seddon, A.W.R., Seppä, H.T., Stednick, J.D., Seybold, S.J. and
Marini, L., 2017. Managing bark beetle impacts on ecosystems and
society: priority questions to motivate future research. Journal of Applied
Ecology, 54(3): 750-760.
Mullen, K.E., 2016. Early Detection of Mountain Pine Beetle Damage in
Ponderosa Pine Forests of the Black Hills Using Hyperspectral and
WorldView-2 Data.
Müller, J., Bußler, H., Goßner, M., Rettelbach, T. and Duelli, P., 2008. The
European spruce bark beetle Ips typographus in a national park: from
pest to keystone species. Biodiversity and Conservation, 17(12): 2979-
3001.
Müller, J. and Bütler, R., 2010. A review of habitat thresholds for dead wood:
a baseline for management recommendations in European forests.
European Journal of Forest Research, 129(6): 981-992.
Müller, M. and Job, H., 2009. Managing natural disturbance in protected areas:
Tourists’ attitude towards the bark beetle in a German national park.
Biological conservation, 142(2): 375-383.
Munoz-Huerta, R.F., Guevara-Gonzalez, R.G., Contreras-Medina, L.M., Torres-
Pacheco, I., Prado-Olivarez, J. and Ocampo-Velazquez, R.V., 2013. A
review of methods for sensing the nitrogen status in plants: advantages,
disadvantages and recent advances. Sensors (Basel), 13(8): 10823-43.
Murtha, P. and Wiart, R., 1989. PC-based digital analysis of mountain pine
beetle current-attacked and non-attacked lodgepole pine. Canadian
journal of remote sensing.
Murtha, P.A., 1972a. A guide to air photo interpretation of forest damage in
Canada. Publication (Canadian Forestry Service) ; no. 1292. Canadian
Forestry Service, Ottawa, vi, 63 p. pp.
Murtha, P.A., 1972b. guide to air photo interpretation of forest damage in
Canada.

136
 Bibliography

Murtha, P.A., 1978. Remote sensing and vegetation damage: a theory for
detection and assessment. Photogrammetric Engineering and Remote
Sensing, 44(9).
Näsi, R., Honkavaara, E., Blomqvist, M., Lyytikäinen-Saarenmaa, P., Hakala,
T., Viljanen, N., Kantola, T. and Holopainen, M., 2018. Remote sensing
of bark beetle damage in urban forests at individual tree level using a
novel hyperspectral camera from UAV and aircraft. Urban Forestry &
Urban Greening, 30: 72-83.
Näsi, R., Honkavaara, E., Lyytikäinen-Saarenmaa, P., Blomqvist, M., Litkey,
P., Hakala, T., Viljanen, N., Kantola, T., Tanhuanpää, T. and Holopainen,
M., 2015. Using UAV-Based Photogrammetry and Hyperspectral Imaging
for Mapping Bark Beetle Damage at Tree-Level. Remote Sensing, 7(12):
15467-15493.
Negrón, J.F., Allen, K.K., Ambourn, A., Cook, B. and Marchand, K., 2017.
Large-Scale Thinnings, Ponderosa Pine, and Mountain Pine Beetle in the
Black Hills, USA. Forest Science, 63(5): 529-536.
Neinavaz, E., Darvishzadeh, R., Skidmore, A.K. and Groen, T.A., 2016.
Measuring the response of canopy emissivity spectra to leaf area index
variation using thermal hyperspectral data. International Journal of
Applied Earth Observation and Geoinformation, 53: 40-47.
Netherer, S., Matthews, B., Katzensteiner, K., Blackwell, E., Henschke, P.,
Hietz, P., Pennerstorfer, J., Rosner, S., Kikuta, S., Schume, H. and
Schopf, A., 2015. Do water-limiting conditions predispose Norway spruce
to bark beetle attack? New Phytol, 205(3): 1128-41.
Netherer, S. and Nopp-Mayr, U., 2005. Predisposition assessment systems
(PAS) as supportive tools in forest management - Rating of site and
stand-related hazards of bark beetle infestation in the High Tatra
Mountains as an example for system application and verification, 207,
99-107 pp.
Ni, Z., Liu, Z., Huo, H., Li, Z.-L., Nerry, F., Wang, Q. and Li, X., 2015. Early
Water Stress Detection Using Leaf-Level Measurements of Chlorophyll
Fluorescence and Temperature Data. Remote Sensing, 7(3): 3232-3249.
Nicolaï, B.M., Beullens, K., Bobelyn, E., Peirs, A., Saeys, W., Theron, K.I. and
Lammertyn, J., 2007. Nondestructive measurement of fruit and
vegetable quality by means of NIR spectroscopy: A review. Postharvest
Biology and Technology, 46(2): 99-118.
Niemann, K.O., Quinn, G., Stephen, R., Visintini, F. and Parton, D., 2015.
Hyperspectral Remote Sensing of Mountain Pine Beetle with an Emphasis
on Previsual Assessment. Canadian Journal of Remote Sensing, 41(3):
191-202.
Niemann, K.O. and Visintini, F., 2005a. Assessment of potential for remote
sensing detection of bark beetle-infested areas during green attack: a
literature review.

137
Bibliography

Niemann, K.O. and Visintini, F., 2005b. Assessment of potential for remote
sensing detection of bark beetle-infested areas during green attack: A
literature review, Natural Resources Canada, Canadian Forest Service,
pp. pp. 1–14.
Nikolov, C., Konôpka, B., Kajba, M., Galko, J., Kunca, A. and Janský, L., 2014.
Post-disaster Forest Management and Bark Beetle Outbreak in Tatra
National Park, Slovakia. Mountain Research and Development, 34(4):
326-335.
Oerke, E., Steiner, U., Dehne, H. and Lindenthal, M., 2006. Thermal imaging
of cucumber leaves affected by downy mildew and environmental
conditions. Journal of experimental botany, 57(9): 2121-2132.
Öhrn, P., 2012. The spruce bark beetle Ips typographus in a changing climate.
Ollinger, S.V., 2011. Sources of variability in canopy reflectance and the
convergent properties of plants. New Phytologist, 189(2): 375-394.
Ortiz, S., Breidenbach, J. and Kändler, G., 2013. Early Detection of Bark Beetle
Green Attack Using TerraSAR-X and RapidEye Data. Remote Sensing,
5(4): 1912-1931.
Paine, T., Raffa, K. and Harrington, T., 1997a. Interactions among scolytid bark
beetles, their associated fungi, and live host conifers. Annual review of
entomology, 42(1): 179-206.
Paine, T.D., Raffa, K.F. and Harrington, T.C., 1997b. INTERACTIONS AMONG
SCOLYTID BARK BEETLES, THEIR ASSOCIATED FUNGI, AND LIVE HOST
CONIFERS. Annual Review of Entomology, 42(1): 179-206.
Pasztor, F., Matulla, C., Rammer, W. and Lexer, M.J., 2014. Drivers of the bark
beetle disturbance regime in Alpine forests in Austria. Forest Ecology and
Management, 318: 349-358.
Pearson, K., 1901. LIII. On lines and planes of closest fit to systems of points
in space. The London, Edinburgh, and Dublin Philosophical Magazine and
Journal of Science, 2(11): 559-572.
Peñuelas, J. and Filella, I., 1998. Visible and near-infrared reflectance
techniques for diagnosing plant physiological status. Trends in plant
science, 3(4): 151-156.
Peterson, D.L., Aber, J.D., Matson, P.A., Card, D.H., Swanberg, N., Wessman,
C. and Spanner, M., 1988. Imaging SpectrometryRemote sensing of
forest canopy and leaf biochemical contents. Remote Sensing of
Environment, 24(1): 85-108.
Peterson, D.L., Westman, W.E., Stephenson, N.J., Ambrosia, V.G., Brass, J.A.
and Spanner, M.A., 1986. Analysis of forest structure using Thematic
Mapper simulator data. IEEE transactions on geoscience and remote
sensing(1): 113-121.
Pfeifer, E.M., Hicke, J.A. and Meddens, A.J., 2011. Observations and modeling
of aboveground tree carbon stocks and fluxes following a bark beetle
outbreak in the western United States. Global Change Biology, 17(1):
339-350.

138
 Bibliography

Pierce, L.L. and Congalton, R.G., 1988. A methodology for mapping forest
latent heat flux densities using remote sensing. Remote Sensing of
Environment, 24(3): 405-418.
Pierce, L.L., Running, S.W. and Riggs, G.A., 1990. Remote detection of canopy
water stress in coniferous forests using the NS001 thematic mapper
simulator and the thermal infrared multispectral scanner. PE&RS,
Photogrammetric Engineering & Remote Sensing, 56(5): 579-586.
Pinder, J.E. and McLeod, K.W., 1999. Indications of relative drought stress in
longleaf pine from Thematic Mapper data. Photogrammetric Engineering
and Remote Sensing, 65: 495–501.
Porcar-Castell, A., Tyystjärvi, E., Atherton, J., van der Tol, C., Flexas, J.,
Pfündel, E.E., Moreno, J., Frankenberg, C. and Berry, J.A., 2014. Linking
chlorophyll a fluorescence to photosynthesis for remote sensing
applications: mechanisms and challenges. Journal of experimental
botany, 65(15): 4065-4095.
Pu, R., Gong, P. and Yu, Q., 2008. Comparative analysis of EO-1 ALI and
Hyperion, and Landsat ETM+ data for mapping forest crown closure and
leaf area index. Sensors, 8(6): 3744-3766.
Qin, Z., Karnieli, A. and Berliner, P., 2001. A mono-window algorithm for
retrieving land surface temperature from Landsat TM data and its
application to the Israel-Egypt border region. International Journal of
Remote Sensing, 22(18): 3719-3746.
Raffa, K.F., Aukema, B.H., Bentz, B.J., Carroll, A.L., Hicke, J.A., Turner, M.G.
and Romme, W.H., 2008. Cross-scale drivers of natural disturbances
prone to anthropogenic amplification: the dynamics of bark beetle
eruptions. Bioscience, 58(6): 501-517.
Raffa, K.F., Grégoire, J.-C. and Staffan Lindgren, B., 2015. Natural History and
Ecology of Bark Beetles, Bark Beetles, pp. 1-40.
Reid, R., 1961. Moisture changes in lodgepole pine before and after attack by
the mountain pine beetle. The Forestry Chronicle, 37(4): 368-375.
Roberts, A., Northrup, J. and Reich, R., 2005. Mountain pine beetle detection
and monitoring: replication trials for early detection, Analysis of Multi-
Temporal Remote Sensing Images, 2005 International Workshop on the.
IEEE, pp. 20-24.
Rohde, M., Waldmann, R. and Lunderstädt, J., 1996. Induced defence reaction
in the phloem of spruce (Picea abies) and larch (Larix decidua) after
attack by Ips typographus and Ips cembrae. Forest Ecology and
Management, 86(1): 51-59.
Roumet, C., Fortunel, C., Kazakou, E., Garnier, E., Vile, D., Papadimitriou, M.,
Zarovali, M.P., Papanastasis, V.P., Skarpe, C., Rusch, G., Sternberg, M.,
Thébault, A., Quétier, F., Grigulis, K., Robson, M., Lavorel, S., Jouany,
C., Theau, J.-P., Cruz, P., Ansquer, P., Castro, H., Freitas, H., Lepš, J.,
Dolezal, J., Quested, H., Eriksson, O., Golodets, C., Kigel, J., Kleyer, M.,
Meier, T., Lehsten, V. and Pakeman, R., 2006. Assessing the Effects of

139
Bibliography

Land-use Change on Plant Traits, Communities and Ecosystem

Functioning in Grasslands: A Standardized Methodology and Lessons
from an Application to 11 European SitesGarnier et al. — Methodology
to Assess Effects of Land-use ChangeGarnier et al. — Methodology to
Assess Effects of Land-use Change. Annals of Botany, 99(5): 967-985.
Rullan-Silva, C.D., Olthoff, A.E., Delgado de la Mata, J.A. and Pajares-Alonso,
J.A., 2013. Remote Monitoring of Forest Insect Defoliation -A Review.
Forest Systems, 22(3): 377.
Runesson, U.T., 1991. Considerations for early remote detection of mountain
pine beetle in green-foliaged lodgepole pine, University of British
Columbia.
Safranyik, L., 1974. Management of lodgepole pine to reduce losses from the
mountain pine beetle.
Safranyik, L. and Carroll, A.L., 2006. The biology and epidemiology of the
mountain pine beetle in lodgepole pine forests.
Safranyik, L., Carroll, A.L. and Wilson, B., 2007. The biology and epidemiology
of the mountain pine beetle in lodgepole pine forests. The mountain pine
beetle: a synthesis of biology, management and impacts on lodgepole
pine: 3-66.
Sambaraju, K.R., Carroll, A.L., Zhu, J., Stahl, K., Moore, R.D. and Aukema,
B.H., 2012. Climate change could alter the distribution of mountain pine
beetle outbreaks in western Canada. Ecography, 35(3): 211-223.
Savitzky, A. and Golay, M.J., 1964. Smoothing and differentiation of data by
simplified least squares procedures. Analytical chemistry, 36(8): 1627-
1639.
Schelhaas, M.J., Nabuurs, G.J. and Schuck, A., 2003. Natural disturbances in
the European forests in the 19th and 20th centuries. Global Change
Biology, 9(11): 1620-1633.
Schmidt, K.S. and Skidmore, A.K., 2003. Spectral discrimination of vegetation
types in a coastal wetland. Remote Sensing of Environment, 85(1): 92-
108.
Schneider, C.A., Rasband, W.S. and Eliceiri, K.W., 2012. NIH Image to ImageJ:
25 years of image analysis. Nat Meth, 9(7): 671-675.
Schowalter, T.D., 2012. Ecology and Management of Bark Beetles (Coleoptera:
Curculionidae: Scolytinae) in Southern Pine Forests. Journal of
Integrated Pest Management, 3(2): 1-7.
Schulze, E.-D. and Hall, A., 1982. Stomatal responses, water loss and CO2
assimilation rates of plants in contrasting environments, Physiological
plant ecology II. Springer, pp. 181-230.
Seidl, R., Rammer, W., Jäger, D. and Lexer, M.J., 2008. Impact of bark beetle
(Ips typographus L.) disturbance on timber production and carbon
sequestration in different management strategies under climate change.
Forest Ecology and Management, 256(3): 209-220.

140
 Bibliography

Seidl, R., Schelhaas, M.-J. and Lexer, M.J., 2011. Unraveling the drivers of
intensifying forest disturbance regimes in Europe. Global Change
Biology, 17(9): 2842-2852.
Seidl, R., Schelhaas, M.-J., Rammer, W. and Verkerk, P.J., 2014. Increasing
forest disturbances in Europe and their impact on carbon storage. Nature
Clim. Change, 4(9): 806-810.
Seidl, R., Thom, D., Kautz, M., Martin-Benito, D., Peltoniemi, M., Vacchiano,
G., Wild, J., Ascoli, D., Petr, M., Honkaniemi, J., Lexer, M.J., Trotsiuk,
V., Mairota, P., Svoboda, M., Fabrika, M., Nagel, T.A. and Reyer, C.P.O.,
2017a. Forest disturbances under climate change. Nature Climate
Change, 7(6): 395-402.
Seidl, R., Thom, D., Kautz, M., Martin-Benito, D., Peltoniemi, M., Vacchiano,
G., Wild, J., Ascoli, D., Petr, M., Honkaniemi, J., Lexer, M.J., Trotsiuk,
V., Mairota, P., Svoboda, M., Fabrika, M., Nagel, T.A. and Reyer, C.P.O.,
2017b. Forest disturbances under climate change. Nat Clim Chang, 7:
395-402.
Sepulcre-Cantó, G., Zarco-Tejada, P.J., Jiménez-Muñoz, J., Sobrino, J., De
Miguel, E. and Villalobos, F.J., 2006. Detection of water stress in an olive
orchard with thermal remote sensing imagery. Agricultural and Forest
Meteorology, 136(1-2): 31-44.
Sheppard, S. and Picard, P., 2006. Visual-quality impacts of forest pest activity
at the landscape level: a synthesis of published knowledge and research
needs. Landscape and Urban Planning, 77(4): 321-342.
Simard, M., Powell, E.N., Raffa, K.F. and Turner, M.G., 2012. What explains
landscape patterns of tree mortality caused by bark beetle outbreaks in
Greater Yellowstone? Global Ecology and Biogeography, 21(5): 556-567.
Six, D.L. and Skov, K., 2009. Response of bark beetles and their natural
enemies to fire and fire surrogate treatments in mixed-conifer forests in
western Montana. Forest Ecology and Management, 258(5): 761-772.
Skakun, R.S., Wulder, M.A. and Franklin, S.E., 2003. Sensitivity of the thematic
mapper enhanced wetness difference index to detect mountain pine
beetle red-attack damage. Remote Sensing of Environment, 86(4): 433-
443.
Slaton, M.R., Hunt, E.R. and Smith, W.K., 2001a. Estimating near-infrared leaf
reflectance from leaf structural characteristics. American Journal of
Botany, 88(2): 278-284.
Slaton, M.R., Raymond Hunt Jr, E. and Smith, W.K., 2001b. Estimating near-
infrared leaf reflectance from leaf structural characteristics. American
Journal of Botany, 88(2): 278-284.
Smith, K., Steven, M. and Colls, J., 2004. Use of hyperspectral derivative ratios
in the red-edge region to identify plant stress responses to gas leaks.
Remote sensing of environment, 92(2): 207-217.

141
Bibliography

Snyder, W.C., Wan, Z., Zhang, Y. and Feng, Y.-Z., 1998. Classification-based
emissivity for land surface temperature measurement from space.
International Journal of Remote Sensing, 19(14): 2753-2774.
Sobrino, J.A., Jiménez-Muñoz, J.C., Sòria, G., Romaguera, M., Guanter, L.,
Moreno, J., Plaza, A. and Martínez, P., 2008. Land surface emissivity
retrieval from different VNIR and TIR sensors. IEEE Transactions on
Geoscience and Remote Sensing, 46(2): 316-327.
Sprintsin, M., 2011. Combining land surface temperature and shortwave
infrared reflectance for early detection of mountain pine beetle
infestations in western Canada. Journal of Applied Remote Sensing, 5(1).
Sprintsin, M., Chen, J.M. and Czurylowicz, P., 2011. Combining land surface
temperature and shortwave infrared reflectance for early detection of
mountain pine beetle infestations in western Canada. Journal of Applied
Remote Sensing, 5(1): 053566-053566-13.
Stadelmann, G., Bugmann, H., Meier, F., Wermelinger, B. and Bigler, C., 2013.
Effects of salvage logging and sanitation felling on bark beetle (Ips
typographus L.) infestations. Forest Ecology and Management, 305:
273-281.
Stoner, W.A. and Miller, P.C., 1975. Water relations of plant species in the wet
coastal tundra at Barrow, Alaska. Arctic and Alpine Research: 109-124.
Sun, Q., Tan, J. and Xu, Y., 2010. An ERDAS image processing method for
retrieving LST and describing urban heat evolution: a case study in the
Pearl River Delta Region in South China. Environmental Earth Sciences,
59(5): 1047-1055.
Tchakerian, M.D. and Coulson, R.N., 2011. Ecological impacts of southern pine
beetle. In: Coulson, RN; Klepzig, KD 2011. Southern Pine Beetle II. Gen.
Tech. Rep. SRS-140. Asheville, NC: US Department of Agriculture Forest
Service, Southern Research Station. 223-234., 140: 223-234.
Tenenhaus, M., 1998. La régression PLS, edition TECHNIP. Paris.
Thom, D. and Seidl, R., 2015. Natural disturbance impacts on ecosystem
services and biodiversity in temperate and boreal forests. Biological
Reviews.
Thom, D. and Seidl, R., 2016. Natural disturbance impacts on ecosystem
services and biodiversity in temperate and boreal forests. Biol Rev Camb
Philos Soc, 91(3): 760-81.
Tucker, C.J., 1979. Red and photographic infrared linear combinations for
monitoring vegetation. Remote sensing of Environment, 8(2): 127-150.
Tucker, C.J., Elgin Jr, J.H., McMurtrey Iii, J. and Fan, C., 1979. Monitoring corn
and soybean crop development with hand-held radiometer spectral data.
Remote Sensing of Environment, 8(3): 237-248.
Ullah, S., Skidmore, A.K., Ramoelo, A., Groen, T.A., Naeem, M. and Ali, A.,
2014. Retrieval of leaf water content spanning the visible to thermal
infrared spectra. ISPRS Journal of Photogrammetry and Remote
Sensing, 93: 56-64.

142
 Bibliography

Valor, E. and Caselles, V., 1996. Mapping land surface emissivity from NDVI:
Application to European, African, and South American areas. Remote
sensing of Environment, 57(3): 167-184.
Van de Griend, A. and Owe, M., 1993. On the relationship between thermal
emissivity and the normalized difference vegetation index for natural
surfaces. International Journal of remote sensing, 14(6): 1119-1131.
van Maarschalkerweerd, M. and Husted, S., 2015. Recent developments in fast
spectroscopy for plant mineral analysis. Frontiers in plant science, 6.
Vanderhoof, M., Williams, C.A., Ghimire, B. and Rogan, J., 2013. Impact of
mountain pine beetle outbreaks on forest albedo and radiative forcing,
as derived from Moderate Resolution Imaging Spectroradiometer, Rocky
Mountains, USA. Journal of Geophysical Research: Biogeosciences,
118(4): 1461-1471.
Verbesselt, J., Robinson, A., Stone, C. and Culvenor, D., 2009. Forecasting
tree mortality using change metrics derived from MODIS satellite data.
Forest Ecology and Management, 258(7): 1166-1173.
Viña, A., Gitelson, A.A., Nguy-Robertson, A.L. and Peng, Y., 2011. Comparison
of different vegetation indices for the remote assessment of green leaf
area index of crops. Remote Sensing of Environment, 115(12): 3468-
3478.
Vincini, M., Frazzi, E. and D’Alessio, P., 2008. A broad-band leaf chlorophyll
vegetation index at the canopy scale. Precision Agriculture, 9(5): 303-
319.
Vlassova, L., Pérez-Cabello, F., Mimbrero, M., Llovería, R. and García-Martín,
A., 2014. Analysis of the Relationship between Land Surface
Temperature and Wildfire Severity in a Series of Landsat Images.
Remote Sensing, 6(7): 6136-6162.
Wang, L., Qu, J.J., Hao, X. and Hunt, E.R., 2011. Estimating dry matter content
from spectral reflectance for green leaves of different species.
International Journal of Remote Sensing, 32(22): 7097-7109.
Wang, Z., 2016. Mapping spatial variation of Foliar Nitrogen using
hyperspectral remote sensing.
Wang, Z., Skidmore, A.K., Darvishzadeh, R., Heiden, U., Heurich, M. and
Wang, T., 2015a. Leaf Nitrogen Content Indirectly Estimated by Leaf
Traits Derived From the PROSPECT Model. IEEE Journal of Selected
Topics in Applied Earth Observations and Remote Sensing, 8(6): 3172-
3182.
Wang, Z., Skidmore, A.K., Wang, T., Darvishzadeh, R. and Hearne, J., 2015b.
Applicability of the PROSPECT model for estimating protein and
cellulose+lignin in fresh leaves. Remote Sensing of Environment, 168:
205-218.
Waring, K.M., Reboletti, D.M., Mork, L.A., Huang, C.-H., Hofstetter, R.W.,
Garcia, A.M., Fulé, P.Z. and Davis, T.S., 2009. Modeling the impacts of
two bark beetle species under a warming climate in the southwestern

143
Bibliography

USA: ecological and economic consequences. Environmental

management, 44(4): 824-835.
Waring, R.H., 1983. Estimating forest growth and efficiency in relation to
canopy leaf area. Advances in Ecological Research, 13: 327-354.
Wermelinger, B., 2004. Ecology and management of the spruce bark beetle
Ips typographus—a review of recent research. Forest Ecology and
Management, 202(1-3): 67-82.
Wold, S., Johansson, E. and Cocchi, M., 1993. PLS-partial least squares
projections to latent structures. 3D QSAR in drug design, 1: 523-550.
Wold, S., Sjöström, M. and Eriksson, L., 2001. PLS-regression: a basic tool of
chemometrics. Chemometrics and Intelligent Laboratory Systems,
58(2): 109-130.
Wood, D.L., 1982. The role of pheromones, kairomones, and allomones in the
host selection and colonization behavior of bark beetles. Annual review
of entomology, 27(1): 411-446.
Wulder, M.A., Dymond, C.C., White, J.C., Leckie, D.G. and Carroll, A.L., 2006a.
Surveying mountain pine beetle damage of forests: A review of remote
sensing opportunities. Forest Ecology and Management, 221(1): 27-41.
Wulder, M.A., White, J., Bentz, B., Alvarez, M. and Coops, N., 2006b.
Estimating the probability of mountain pine beetle red-attack damage.
Remote Sensing of Environment, 101(2): 150-166.
Wulder, M.A., White, J.C., Bentz, B., Alvarez, M.F. and Coops, N.C., 2006c.
Estimating the probability of mountain pine beetle red-attack damage.
Remote Sensing of Environment, 101(2): 150-166.
Wulder, M.A., White, J.C., Carroll, A.L. and Coops, N.C., 2009. Challenges for
the operational detection of mountain pine beetle green attack with
remote sensing. The Forestry Chronicle, 85(1): 32-38.
Xu, H., Zhu, S., Ying, Y. and Jiang, H., 2006. Early detection of plant disease
using infrared thermal imaging, Optics East 2006. SPIE, pp. 7.
Xu, Q.-S. and Liang, Y.-Z., 2001. Monte Carlo cross validation. Chemometrics
and Intelligent Laboratory Systems, 56(1): 1-11.
Yamaoka, Y., Swanson, R. and Hiratsuka, Y., 1990. Inoculation of lodgepole
pine with four blue-stain fungi associated with mountain pine beetle,
monitored by a heat pulse velocity (HPV) instrument. Canadian Journal
of Forest Research, 20(1): 31-36.
Yoder, B.J. and Pettigrew-Crosby, R.E., 1995. Predicting nitrogen and
chlorophyll content and concentrations from reflectance spectra (400–
2500 nm) at leaf and canopy scales. Remote sensing of environment,
53(3): 199-211.
Zarco-Tejada, P. and Sepulcre-Cantó, G., 2007. Remote sensing of vegetation
biophysical parameters for detecting stress condition and land cover
changes. Estudios de la Zona no Saturada del Suelo Vol. VIII: 37-44.
Zeppenfeld, T., Svoboda, M., DeRose, R.J., Heurich, M., Müller, J., Čížková, P.,
Starý, M., Bače, R., Donato, D.C. and Bugmann, H., 2015. Response of

144
 Bibliography

mountainPicea abiesforests to stand-replacing bark beetle outbreaks:

neighbourhood effects lead to self-replacement. Journal of Applied
Ecology, 52(5): 1402-1411.
Zhang, C., Chen, K., Liu, Y., Kovacs, J.M., Flores-Verdugo, F. and de Santiago,
F.J.F., 2012a. Spectral response to varying levels of leaf pigments
collected from a degraded mangrove forest. Journal of Applied Remote
Sensing, 6(1): 063501.
Zhang, C., Liu, Y., Kovacs, J.M., Flores-Verdugo, F., de Santiago, F.F. and
Chen, K., 2012b. Spectral response to varying levels of leaf pigments
collected from a degraded mangrove forest. Journal of Applied Remote
Sensing, 6(1): 063501-1-063501-14.
Zhang, J., Wang, Y. and Li, Y., 2006. A C++ program for retrieving land surface
temperature from the data of Landsat TM/ETM+ band6. Computers &
Geosciences, 32(10): 1796-1805.
Zhang, Y., Chen, J.M., Miller, J.R. and Noland, T.L., 2008. Leaf chlorophyll
content retrieval from airborne hyperspectral remote sensing imagery.
Remote Sensing of Environment, 112(7): 3234-3247.
Zuber, M. and Benz, G., 1992. Untersuchungen über das Schwärmverhalten
von Ips typographus (L.) und Pityogenes chalcographus (L.)(Col.,
Scolytidae) mit den Pheromonpräparaten Pheroprax und Chalcoprax.
Journal of Applied Entomology, 113(1-5): 430-436.

145
Bibliography

146
Summary
Forest disturbance in Europe, induced by the European spruce bark beetle Ips
typographus, L., results in regional-scale dieback and causes major economic
loss to the forest industry, particularly in Norway Spruce (Picea abies) forest
stands. Early stress detection induced by bark beetle infestation (the so-called
‘green-attack’ stage — when trees are yet to show distinct symptoms
observable by the human eye) is crucial and can lead to effective forest
management and reduce economic losses. In this respect, remote sensing is a
cost-effective and repetitive technique which offers an optimal approach for
monitoring and assessing forest stress in comparison to the more traditional
(field survey) approach, which is not practical and is inefficient in large areas
because it is significantly laborious and costly.

The current thesis presents research regarding the potential of different types
of remote sensing data obtained from a variety of sensors, including ASD
FieldSpec3 and multispectral satellites (RapidEye, SPOT-5, Sentinel-2 and
Landsat-8) for early detection of bark beetle infestation at both leaf and canopy
levels. Moreover, the research provided an important opportunity to advance
understanding regarding the impact of early bark beetle infestation on the
biochemical properties of infested trees (chlorophyll, nitrogen, chlorophyll
fluoresces, leaf water content, dry matter content and stomatal conductance).
The leaf-level spectrometry demonstrated that a significant difference (p <
0.05) exists in the mean reflectance spectra between healthy and infested
needle samples at the green-attack stage, with the most pronounced
differences being observed in the NIR and SWIR regions. The results of our
analysis demonstrated that infestation at the green-attack stage reduced the
foliar biochemical concentrations (chlorophyll and nitrogen) and weakened
their correlations with reflectance. The findings confirmed the importance of
hyperspectral measurement as well as foliar biochemical properties (i.e.
chlorophyll and nitrogen) for the detection of Ips typographus, L. at the green-
attack stage.

Furthermore, research has moved on to the spaceborne level to evaluate the

potential of different types of multispectral satellite data (RapidEye, SPOT-5,
Sentinel-2 and Landsat-8) in parallel with the collection of field data (leaf traits)
for early bark beetle detection. The results from analysis of different spectral
vegetation indices at both leaf and canopy levels under bark beetle infestation
during the green-attack stage revealed that red-edge and SWIR in the optical
domain were the most important spectral regions at both leaf and canopy levels
for detecting subtle changes in Norway spruce trees due to bark beetle
infestation. Furthermore, from the results of temporal analysis using a time
series of seven RapidEye scenes and six SPOT-5 scenes in parallel with field
measurements, we identified for the first time in a European situation (in

147
Summary

contrast to earlier works in the US) that mid-June to the beginning of July is
an appropriate time frame for detection of early stress induced by bark beetle
infestation.

Our study furtherhighlighted the importance and potential of thermal infrared

data from Landsat-8 for the early detection of bark beetle infestations and the
production of reliable infestation maps at the green-attack stage.

In general, this study contributes to applied research in the field of remote

sensing of bark beetle green attack. The findings and methods applied in this
work can potentially be used to produce valuable information regarding bark
beetle infestation at an early stage of the attack. Our findings can improve
bark beetle management activities by providing useful information regarding
how, when and which remote-sensing data could be applied to such survey
practice, and therefore administrative bodies of forested areas may benefit
from this information as an guidance for landscape-wide detection of bark
beetle green attack.

148
Samenvatting
Bosverstoring in Europa, veroorzaakt door de Europese sparschorskever Ips
typographus, L., resulteert in een terugval op regionale schaal en veroorzaakt
grote economische verliezen voor de bosbouwindustrie, met name de
bosgebieden met sparren (Picea abies). Vroege stressdetectie veroorzaakt
door schorskever infestatie (de zogenaamde 'groene-aanval'-fase - wanneer
bomen nog verschillende symptomen moeten vertonen die door het menselijk
oog kunnen worden waargenomen) is cruciaal en kan leiden tot meer effectief
bosbeheer alsmede het beperken van economische verliezen. In dit opzicht is
teledetectie een kosteneffectieve en repetitieve techniek die een optimale
aanpak biedt voor het bewaken en beoordelen van bosstress in vergelijking
met de meer traditionele (veldonderzoek) benadering, wat niet praktisch is en
in grote gebieden inefficiënt is, omdat het aanzienlijk arbeidsintensief en
kostbaar is.

Het huidige proefschrift presenteert onderzoek naar het potentieel van

verschillende soorten teledetectiegegevens die zijn verkregen door een
verscheidenheid aan sensoren, waaronder ASD FieldSpec3 en multispectrale
satellieten (RapidEye, SPOT-5, Sentinel-2 en Landsat-8) voor vroege detectie
van schorskevers. op zowel blad- als kruinniveau. Bovendien bood het
onderzoek een belangrijke mogelijkheid om begrip te kweken voor de impact
van een vroege schorskevers infestatie op de biochemische eigenschappen van
aangetaste bomen (chlorofyl, stikstof, chlorofylfluorescentie,
bladwatergehalte, droge stofgehalte en stomatale geleiding).

Het onderzoek op bladniveau toonde aan dat er een significant verschil (p

<0,05) bestaat in de gemiddelde reflectantiespectra tussen gezonde en
geïnfecteerde naaldmonsters in de fase van de groene aanval, waarbij de
meest uitgesproken verschillen worden waargenomen in de NIR- en de SWIR-
regio's. De resultaten van de studie toonden aan dat de plaag in de groene-
aanvalfase de blad biochemische concentraties (chlorofyl en stikstof)
verminderde en hun correlaties met de reflectie verzwakte. De bevindingen
bevestigden het belang van hyperspectrale meting evenals biochemische
eigenschappen van bladeren (d.w.z. chlorofyl en stikstof) voor de detectie van
Ips typographus, L. in de fase van de groene aanval.

Bovendien is het onderzoek verplaatst op ruimtebereik om het potentieel van

verschillende soorten multispectrale satellietgegevens (RapidEye, SPOT-5,
Sentinel-2 en Landsat-8) te evalueren, parallel met de verzameling
veldgegevens (bladkenmerken) voor vroege schors kever detectie. De
resultaten van de analyse van verschillende spectrale vegetatie-indices op
zowel blad- als kruinniveau onder schorskever infestatie tijdens het groene-
aanvalstadium onthulden dat Red-Edge en SWIR in het optische domein de

149
Samenvatting

belangrijkste spectrale gebieden waren op zowel blad- als kruinniveau voor

detectie. subtiele veranderingen in Noorse sparren bomen als gevolg van
schors keverplaag. Verder hebben we uit de resultaten van tijdanalyse met een
tijdreeks van zeven RapidEye-zintuigen en zes SPOT-5-zintuigen parallel aan
de verzameling in het veld, voor het eerst in een Europese situatie
geïdentificeerd (in tegenstelling tot eerdere werken in de VS) ) dat medio juni
tot begin juli een geschikt tijdsbestek is voor detectie van vroege stress
veroorzaakt door schorskevers.

Ook wezen de bevindingen van het onderzoek in hoofdstuk 4 uit naar het
belang en het potentieel van thermische infraroodgegevens van Landsat-8 voor
de vroege detectie van schimmelkeverbesmettingen en de productie van
betrouwbare infestatiekaarten in het stadium van de groene aanval.

Over het algemeen draagt dit onderzoek bij aan toegepast onderzoek op het
gebied van teledetectie van schorskevers ongevallen. De bevindingen en
methoden die in dit werk worden toegepast, kunnen in een vroeg stadium van
de aanval nuttige informatie opleveren over de aantasting van schorskevers.
Onze bevindingen kunnen ook de beheersing van schorskevers verbeteren
door nuttige informatie te verstrekken over hoe, wanneer en welke teledetectie
gegevens kunnen worden toegepast op dergelijke enquêtepraktijken, en
daarom kunnen bestuursorganen van beboste gebieden dergelijke informatie
gebruiken als een indicator voor een efficiënte landschaps-brede detectie van
de groene aanval van de schorskever.

150
 Author’s Biography
Haidi Jamal Abdullah was born on 5th January
1986 in Sulaymaniyah, Iraqi Kurdistan. He
studied geography at Salahaddin University in
Erbil, Iraq, and received a B.Sc. in July 2007.
After graduation from 2007-2009, he was
working as an assistant lecturer in the
department of geography, College of Arts. In
2010 he was awarded KRG Human Capacity
Development Program scholarship to pursue his
master research at the department of
geography, at the University of Leicester, UK. He
obtained his M.Sc. in 2011 with distinction. In 2014, he was awarded another
scholarship from European Commission Erasmus Mundus scholarships to
pursue his doctoral research at the Faculty of Geo-information Science and
Earth Observation (ITC) at the University of Twente, which resulted in this
thesis.

151

View publication stats