Mapping The Relative Abundance of Soil Microbiome Biodiversity From eDNA and Remote Sensing - ScienceDirect

10/21/22, 11:34 AM Mapping the relative abundance of soil microbiome biodiversity from eDNA and remote sensing - ScienceDirect
Science of Remote Sensing

Volume 6, December 2022, 100065
Mapping the relative abundance of soil microbiome

biodiversity from eDNA and remote sensing
Andrew K. Skidmore a, b , Andjin Siegenthaler a, Tiejun Wang a, Roshanak Darvishzadeh a, Xi Zhu a, Anthony
Chariton b, G. Arjen de Groot c
Show more
Outline Share Cite
https://doi.org/10.1016/j.srs.2022.100065 Get rights and content

Under a Creative Commons license Open access
Highlights
• Combining image spectroscopy and eDNA for mapping soil microbiome
biodiversity.
• Mapping relative abundance of microbiome families across savanna, boreal and

tundra.
• Linear Discriminate Analysis (LDA) identifies indicator eDNA microbial families.
• PLSR predicts the relative abundance of soil microbial families from remote
sensing.
Abstract
Although an enormous number of plant and animal species have been directly observed and
recorded in online databases, the spatial variation in the composition of the microbiome
remains relatively largely unknown. In this study, for the first time, we demonstrate mapping
of the relative abundance of the soil microbiome for three terrestrial ecosystems across North
https://www.sciencedirect.com/science/article/pii/S266601722200027X 1/30
America (savanna, boreal and tundra) using airborne image spectroscopy and environmental
DNA (eDNA) data. We identified field plots of publicly available eDNA data co-occurring with
AVIRIS-NG hyperspectral imagery. An eDNA processing pipeline was developed to generate a
consistent profile of the relative abundance for thousands of microbiome operational
taxonomic units (OTU) and 225 microbiome families. Using Linear Discriminate Analysis
(LDA) scores for the eDNA data, we identified 81 families with the greatest explanatory power
based on the community composition between the three ecosystems. A case study example
demonstrates our conceptual approach by selecting a dominant and functionally important
bacterial family for each ecosystem, with each family representing a specific biomarker. A
partial least squares regression (PLSR) was applied to spatially predict the relative abundance
of each bacteria family (in the boreal, tundra, and savanna ecosystems) from the hyperspectral
imagery. For the boreal, Pseudomonadaceae is shown to be a dominant family taxon, as it is a
saprophytic family specialized in decomposing the dense organic matter of boreal forest soils.
Members of an understudied family of Acidobacteria, so far indicated as AKIW659, are
abundant in acidic Arctic soils and peat bogs. Finally, the Micromonosporaceae are dominant
and functionally important in drier regions with grass-tree dominated woodlands, being a
member of Actinobacteria with a high relative abundance in soils with high carbon content
and nitrate leaching. We demonstrate for the first time how the spatial prediction of relative
abundance of these bacteria taxa based on remote sensing, showing patterns of the soil
microbiome biodiversity and ecosystem function within and across the three ecosystems.
Previous Next
Keywords
Image spectroscopy; Environmental DNA; Biodiversity; Species abundance
1. Introduction
Traditional biodiversity monitoring involves the field observation of species by specialized
biologists, aided by skilled volunteers, whose expertise is restricted to specific biotic
groupings. Botanists may identify and record the presence of plant species and ornithologists
the bird biota, but ‘unpopular’ biotic groups such as fungi, bacteria and insects are under-
observed or escape identification altogether. Traditional methods for identifying soil
microbial communities from in situ samples have been based on direct human observation of
organisms cultured under laboratory conditions. Though traditional techniques have been
steadily refined they remain expensive, time consuming, inconsistent across time and space,
as well as only capturing a fraction of true biodiversity (Chariton et al., 2010). In recent years,
environmental DNA (eDNA) metabarcoding has enabled the rapid collection of
comprehensive species occurrence of a sample's microbial community composition and
functional potential (Taberlet et al., 2018).
Though eDNA remains expensive, the costs per amount of data are rapidly reducing, while
protocols (for field sampling of eDNA as well as laboratory analysis) are standardizing, and
from massive online reference databases it is possible to identify taxa from DNA-sequences
(Chariton et al., 2010; Taberlet et al., 2018). Another key innovation is the rapid growth in the
number of meta-databases of sequenced DNA samples, though samples are thin on the
ground especially when upscaled to landscape or global levels. For the microbiome, the spatial
distribution of taxa remains essentially unknown, as more intensive sampling to ‘fill the gaps’
at landscape to global levels would be prohibitively expensive in terms of fieldwork and
laboratory costs.
Concurrent with these advances in microbiology, imaging spectroscopy with a finer spatial
resolution pixel (2–5 m) can remotely sense the chemical, compositional, and functional
characteristics of above-ground plant communities using spectral reflectance and vegetation
indices (Lausch et al., 2016; Feilhauer et al., 2016; Rahman et al., 2003). Interestingly, even
ecosystem functions such as phenology and photosynthetic activity have been linked to
vegetation indices using coarser resolution hyperspectral satellites (250–500 m) such as
MODIS and Hyperion, (Zhang et al., 2003; Guerschman et al., 2009). The combination of very
high-resolution image spectroscopy and eDNA approaches link disparate ecosystem
components by capturing dominant ecosystem functions across multiple domains of life. The
use of very high-resolution image spectroscopy allows contiguous mapping, by ‘filling the
gaps’ between in situ eDNA field samples.
As biodiversity change has many dimensions and occurs at a range of spatial and temporal
scales, relying on observational in situ samples is proving inadequate for the 196 parties to the
UN Convention on Biodiversity (CBD) to meet their monitoring commitments (Marques et al.,
2014). The Essential Biodiversity Variables developed by the Group of Earth Observation
Biodiversity Observation Network (GEO BON) provide a framework for measuring
biodiversity at multiple levels (Pereira et al., 2013). Although some essential biodiversity
variables can already be monitored using broad-band optical satellite imagery such as from
Landsat and Sentinel-2, these are primarily measures of gross biodiversity metrics, e.g. land
cover and phenology (Potapov et al., 2017; Skidmore et al., 2015), and lack the high spectral
resolution needed to model ecosystem functions from eDNA in situ samples. To address this
shortfall requires fundamentally new approaches for rapidly quantifying and understanding
the complexity and changes in global biodiversity, including the relative abundance of species
and community composition. Exciting actual and planned innovations in next-generation
hyperspectral satellites (viz., the Environmental Mapper and Analysis Programme EnMAP
(Storch, 2022) from DLR-Germany, PRecursore IperSpettrale della Missione Applicativa
PRISMA (PRISMA, 2022) from ASI-Italy , Copernicus Hyperspectral Imaging Mission for the
Environment CHIME (ESA, 2022) from ESA-Europe, and Surface Biology and Geology SBG
(Nastal and Pavlick, 2022) from NASA-USA) further our abilities to upscale biodiversity
products to a global level and further understanding of how biodiversity responds to stress
and environmental change (Lausch et al., 2016).
Here we demonstrate for the first time the potential to spatially predict microbial community
composition from remote sensing. We further highlight examples of how ecosystem
functions, such as soil respiration and plant growth reduction in waterlogged anaerobic
conditions, can be synchronously measured by remote sensing as well as deduced from the
high relative abundance of specific soil microbiome families. This builds on the work of
spatial ecology and image spectroscopy, highlighting how functional ecology has been linked
to image spectroscopy (Asner and Martin, 2016) as well as Plant Functional Types (Ustin and
Gamon, 2010) as indicators of soil properties and the soil microbiome (Aponte et al., 2013).
Our research, using the image spectroscopy response to inform on environmental
niche/habitat, reveals for the first time how biodiversity such as the soil properties as well the
soil microbiome can be derived from image spectroscopy. In other words, through analysis of
co-occurring eDNA and image spectroscopy data, we aim to show how ecosystem function can
be interpreted from the literature, thereby filling the Linnaean gap (i.e., the gap caused by
most species not being described by traditional sampling) and the Wallacean gap (i.e., the
unknown geographical distribution and extent of most species) (Bini et al., 2006).
2. Methods
2.1. Remote sensing data

Using the NASA Jet Propulsion Laboratory (JPL) portal we located high resolution
hyperspectral (image spectroscopy) imagery obtained with the Airborne Visible InfraRed
Imaging Spectrometer - Next Generation (AVIRIS-NG) between 2014 and 2017. AVIRIS-NG
measures the radiance of wavelengths between 380 nm and 2510 nm with a 5 nm sampling
interval. Data were processed by the AVIRIS-NG instrument ground data system (IGDS). All
publicly available images were free-of-cloud. As there was noise in the radiance values, the
useable wavelength range was truncated to 431–2470 nm.
The hyperspectral data are delivered as standardized ortho-corrected as well as

radiometrically-corrected reflectance (https://avirisng.jpl.nasa.gov/ang_data_pipeline.html).
Depending on the height of the aircraft, the spatial resolution of the imagery ranged from 2 m
to 5.2 m. To standardize the image pixel size to the eDNA plot data (with some field locations
having more than 2 replicate samples) we assumed replicate points were located with a RMSE
≤10 m. The image pixels intersecting with each eDNA field plot (of 10 m × 10 m) were selected
and the average reflectance values of all the pixels in the 10 m × 10 m plot retrieved.
2.2. eDNA data

From the thousands of records of field measured eDNA that are publicly available, only 23
eDNA plots were found that co-occur with 5 hyperspectral (AVIRIS-NG) images across North
America (Supplementary Table S1). The hyperspectral images, coincident with the eDNA plots,
were acquired during summer months (June–August) between 2014 and 2017. The eDNA plot
samples were derived from two prominent international databases: the European Molecular
Biology Laboratory-European Bioinformatics Institute (EMBL-EBI) (Mitchell et al., 2018) and

the Earth Microbiome Project (EMP) (Thompson et al., 2017).
The publicly available eDNA data have bacterial 16S gene profiles, allowing inter-comparison
of bacteria family taxonomic data across the three ecosystems (savanna, boreal and tundra).
These profiles have been constructed using the well-established 515F-806R primer set
targeting the V4 region of the 16S SSU rRNA (Caporaso et al., 2011). The DNA profiles were
downloaded from the public databases as operational taxonomic unit (OTU) tables annotated
using the Greengenes reference database (v13.8). Data in these databases has been collected
and processed according to standardized protocols
(dx.doi.org/10.17504/protocols.io.nuudeww) and comply with the high data standards of these
databases (https://earthmicrobiome.org/protocols-and-standards/16s/,
https://www.ebi.ac.uk/ena/browser/about/data-standards), with the aim to reduce variation
between individual studies (Yilmaz et al., 2011). Meta-analyses of microbial community
samples collated in these databases have resulted in over 100 high-quality publications
exploring microbial diversity at local to global scales by utilizing the information stored about
the DNA extraction, amplification, sequencing and bioinformatic (QIIME 1) protocols
(Caporaso et al., 2010; Thompson et al., 2017; Nuccio et al., 2016; Gilbert et al., 2014). See
Supplementary Table S2 for source information of each dataset used in this study, including
the sampling and sequencing procedures. We proceeded with only those taxa assigned to
Eubacteria (i.e., true bacteria), removing any taxa assigned to Eukaryotic organelles (i.e.,
mitochondria and chloroplast) and Archaea (i.e., organisms with molecular characteristics
separating them from bacteria). Replicate samples from the same plot were averaged and
taxonomic units with a maximum of 9 or less counts per sample were removed on a sample-
by-sample basis to avoid false positives and low-frequency noise (De Barba et al., 2014;
Siegenthaler et al., 2019). The resulting table was rarefied to 10,000 reads per sample to
normalize for variation in sequencing depth (mean ± SD read count per sample:
90,693 ± 49,086) (Sanders, 1968; Weiss et al., 2017) and taxa were grouped into families and
then the relative read abundance per family was calculated by dividing the raw reads of each
family by the total number of reads in a sample. Relative abundances show the relative
contribution of each taxon to the overall composition of the amplified product and are less
influenced compared to raw read counts by stochastic processes introduced by the
amplification and sequencing process (Philippot et al., 2009; Taberlet et al., 2012). Relative
abundances is a well established approach for studying microbial community composition
(Orwin et al., 2018; Lozupone and Knight, 2007; Roesch et al., 2007; Jesmok et al., 2016) as well
as analysing diversity patterns in eDNA datasets (Philippot et al., 2009; Albaina et al., 2016;
Evans et al., 2016). We conducted Linear Discriminant Analysis (LDA) to select bacterial
families that form representative biomarkers for the 3 ecosystems. Each ecosystem (i.e.,
savanna, boreal and tundra) represented a class in a Linear Discriminant Analysis (LDA). The
variables in this LDA were the eDNA families and values were the relative abundance of each
family. Differentially abundant families between the three ecosystems (boreal, savanna and
tundra) were identified using LDA Effect Size (LEfSe) analysis (Segata et al., 2011) with α' ≤ 0.05
(factorial Kruskal-Wallis test) and a minimum logarithmic LDA score ≥3 thresholds (Segata et
al., 2011).
2.3. Modeling the bacterial families from remote sensing using PLSR
Based on the Linear Discriminant Analysis (LDA) thresholds, a total of 81 discriminant family
taxonomic units were extracted for each sample. Then, using a Partial Least Squares
regression (PLSR) the relative abundance of each family was predicted using the hyperspectral
imagery (with the AVIRIS image band being the predictor variable in the PLSR analysis). PLSR
reduces predictor variables (i.e., image bands) to a small set of uncorrelated components, and
regression is then started on this small set of components. As spectral measurements have
predictors (wavebands) which are highly collinear and heavily outnumber observations, PLSR
has become the preferred method of analysis for image spectroscopy and is frequently used
for modeling multiple response variables (Axelsson et al., 2013; Ramoelo et al., 2013; Bi et al.,
2019). The optimal number of PLSR components (with 'components' being a term often used
in remote sensing, versus ‘latent variables' which is used more in biology) was chosen based
on the lowest corrected Akaike Information Criterion (AICc) threshold values and lowest
RMSE, where AICc is the AIC corrected for small sample sizes (Segata et al., 2011). PLSR
models were run in R (v. 4.1.1) using 'pls' (v. 2.8-0) with NIPALS (Nonlinear Iterative Partial
Least Squares) (Mevik and Wehrens, 2007) as well as the Python (3.8.2) environment. Models
were validated using leave-one-out (LOO) cross-validated datasets (Philippot et al., 2009;
Albaina et al., 2016; Evans et al., 2016) (see also Supplementary Fig. S3). Examples of
discriminant bacterial families (Log-LDA score >3) which were highly correlated (R2 > 0.5) to
the image spectra in the PLSR models) are summarized in Table 1.
Table 1. Examples of bacterial families selected from the 81 family taxonomic units as extracted
for each sample. Families were selected that had the following thresholds: they showed a
significant (P < 0.05; Kruskal-Wallis test) and discriminant association with a ecosystem (Log-
LDA score >3) and they were highly correlated (R2 > 0.5) to the image spectra in the PLSR
models. The number of PLSR components was selected based the lowest AICc value. Mean
relative read abundance (MRA% in Table 1) are shown per family for the full dataset and for
individual ecosystems.
Family Ecosystem LDA PLSR MRA% All MRA% MRA% MRA%

score R2 data Savanna Boreal Tundra
Pseudomonadaceae boreal 3.64 0.54 0.26 0.02 0.67 0.05
AKIW659 tundra 3.74 0.85 0.21 0.00 0.00 0.61
Anaerolinaceae tundra 3.88 0.54 0.33 0.00 0.12 0.82
Syntrophaceae tundra 4.66 0.72 1.77 0.00 0.04 5.04
Micromonosporaceae savanna 4.40 0.85 1.04 3.36 0.03 0.03
Family Ecosystem LDA PLSR MRA% All MRA% MRA% MRA%

score R2 data Savanna Boreal Tundra
Solirubrobacteraceae savanna 4.35 0.65 1.03 3.22 0.10 0.04
Geodermatophilaceae savanna 4.02 0.72 0.51 1.50 0.03 0.11
To demonstrate the potential to map and further understand the distribution of microbial
families, we selected three example microbial families (from the 81 pre-selected families) as a
representative biomarker for each of the three ecosystems (Table 2). These three families were
selected based on a high R-squared value from the PLSR and known ecological information
explaining the presence of each family taxon across a landscape based on its ecological
function. We considered when selecting the three family taxa whether the typical habitat of
the family taxon could be deduced from known hyperspectral remote sensing wavelengths.
For example, it is known that 940–960 nm are wavelengths important for discriminating
tundra vegetation (Bratsch et al., 2017), while the 1305–1355 nm range is associated with the
waterlogged conditions in the Arctic tundra which exposes a greater component of the
soil/ground/water reflectance signal (Knipling, 1970). From the literature, we also considered
the habitat typical of the family taxon and links to ecological function (e.g., amount of green
biomass and non-photosynthetic components such as lignin and cellulose linked to
photosynthesis, the role of canopy nitrogen and phosphorus in nutrient cycling, and role of
biochemical compounds such as leaf proteins and polyphenols etc.) The first selected
microbial family AKIW659 is abundant in waterlogged, anaerobic, and highly acidic soils, and
proportionally more abundant in tundra wetlands. The second family is Pseudomonadaceae,
which is saprophytic and specialized in decomposing the thick organic matter of boreal forest
floors. Finally, the Micromonosporaceae family is dominant and functionally important in
savannah soils, and forms in herbage with a higher carbon content and leaching of nitrates
(see Table 2 for more details).
Table 2. Known ecological habitat of the three families (AKIW659, Pseudomonadaceae,

Micromonosporaceae) with respectively high relative abundance in the tundra, boreal and
savanna ecosystems.
Family Description
Family Description
AKIW659 AKIW659 is in the phylum Acidobacteria being characterized as an aerobic

heterotroph which thrives in acidic conditions and have a specialized role in
carbon cycling by oxidizing simple and complex carbohydrates such as starch,
cellulose and lignin, in wetlands as well as drier tundra areas which typically have
a buildup of carbon rich organic matter associated with a scarcity of nutrients and
low quality soils (de Chaves et al., 2019; Stark et al., 2014; Koyama et al., 2014). The
location of the tundra sample plots is in an ice-wedge polygonal tundra landscape
with flat-center polygons, typically with sedges, grasses and various mosses
(Davidson et al., 2016).
Pseudomonadaceae Pseudomonadaceae are in the phylum Proteobacteria, and a metabolically diverse

family that degrade a variety of low-molecular-weight organic compounds and
hydrolytic products generated by actinomycete and fungal growth. They are
aerobic and rarely grow below pH 5.0–6.0 (Cousins et al., 2004). As free-living
saprophytic family, Pseudomonadaceae are specialized in decomposing the
organic matter (Palleroni Starr et al., 1981) typical of boreal forest ecosystem.
Micromonosporaceae Micromonosporaceae of the phylum Actinobacteria, are aerobic and chemo-

organotrophic occurring on less acidic soils typical of the savannah. They can
survive high temperature exposure but are considered to be mesophilic (optimum
temperature between 20 °C and 40 °C). Micromonosporaceae are stress tolerant of
drought. They are known to fix nitrogen as well as decompose carbon compounds.
We ran the PLSR (NIPALS option) with leave-one-out (LOO) cross-validation of the
predictions (Mevik and Wehrens, 2007). For the three selected microbial families, the optimal
number of components (i.e., the lowest AICc and RMSEP) were 5 for AKIW659 and
Micromonosporaceae, and 3 components for Pseudomonadaceae (Table 3). To validate the
modeling of the bacterial families from the remote sensing using PLSR, the measured and
predicted values derived by the leave-one-out cross validation demonstrate that the PLSR
models are reliable with an accuracy comparable with similar studies (Ramoelo et al., 2013;
Askari et al., 2015; Fernandez-Habas et al., 2021) (Table 3 and Fig. S3).
Table 3. Predictions of microbiome family relative abundances using PLSR fitted with AVIRIS-
NG hyperspectral imagery. RMSE and R2values are based on bias-corrected cross-validation
estimates (NIPALS option). Total model variance explained using an optimal number of
components (i.e., the lowest AICc and RMSEP). See also Supplementary Fig. S1.
Family R2 (Taberlet et al., 2018) RMSE (adjCV) Components Variance explained
AKIW659 0.85 0.013 5 95.34
Family R2 (Taberlet et al., 2018) RMSE (adjCV) Components Variance explained
Pseudomonadaceae 0.54 0.028 3 69.42
Micromonosporaceae 0.85 0.034 5 93.99
We utilized the AVIRIS-NG hyperspectral imagery to predict the relative abundance of the 3
microbial families (AKIW659, Pseudomonadaceae, Micromonosporaceae) across the 3
ecosystems (viz. tundra, boreal and savanna). The PLSR model directly predicted the relative
abundance of each family derived from the eDNA profile data across each AVIRIS-NG image
(i.e., Python 3.8.2 PLSR regression ‘predict’ function), with the relative abundance values of the
three selected microbial families being written into three separate image matrices. We finally
projected RGB colour composite maps, by combining the relative abundance image of each
family into a colour composite image as the red (AKIW659), green (Micromonosporaceae) and
blue (Pseudomonadaceae) respectively (Schmidtlein et al., 2012) (Fig. 2). Using the visible bands
of the AVIRIS-NG imagery, aerial imagery was simulated as a true (also known as ‘natural’)
colour composition (Siok and Ewiak, 2020) (Fig. 2).
3. Results
DNA profiles contained on average (±SD) 90,693 ± 49,086 reads per sample after removal of
reads from Eukaryotic organelles, Archaea, and low-frequency noise. Rarefaction to 10,000
reads per sample resulted in a total number of 779 OTUs (267 ± 59 Mean ± SD OTUs per
sample). Seven OTUs were removed as result of the rarefaction. A total of 81 bacterial families
were detected in the dataset, with a mean (±SD) 52 ± 10 families per profile.
The PLSR provides insight into the microbial community, as well as indicating habitat
requirements by highlighting significant wavelength bands associated with specific microbial
families. From Supplementary Fig. S1, the plotted PLSR scores for the first 2 components
(latent factors 1 and 2) show good separation of the family taxon across the three ecosystems.
For example, AKIW659 has a high relative abundance with PLS1 in the tundra (green dots),
and a low relative abundance in the boreal and savanna, highlighting the importance of the
microbiome family as a biomaker for certain ecosystems.
Examining the PLSR coefficients in combination with the Variable Importance of Projection
(VIP) scores, indicate wavelengths having a significant impact on predicting the relative
abundance of each family. The wavelengths having an impact on the relative abundance of the
family AKIW659 (Fig. 1a) are at 760 nm, 940–960 nm (double peak), 1120 nm, 1305–1355 nm
(double peak), and 2015 nm). The wavelengths having an impact on the relative abundance of
the Pseudomonadaceae family associate with the wavelength range 750 nm, 900 nm, 1120 nm
with a negative association at 1970–2020 nm for the relative abundance of Pseudomonadaceae
(Fig. 1b). A common family present in savanna/Mediterranean landscapes is the
Micromonosporaceae, which occurred with high relative abundance in our eDNA profile data.
There is a high relative abundance of Micromonosporaceae associated with wavelengths at
680 nm, 1480–1490 nm, 1690–1710 nm, with a low relative abundance at 1425 nm 2100 nm
(Fig. 1c).
Download : Download high-res image (687KB) Download : Download full-size image
Fig. 1. The VIP scores and regression coefficients for families a) AKIW659 b)
Pseudomonadaceae c) Micromonosporaceae generated by the PLSR analysis for the three
ecosystems (boreal, savanna and tundra). VIP scores below the red line (VIP<1) indicate parts
of the spectrum that are non-informative for predicting the relative abundance of the
respective family. The vertical blue lines indicate the wavelengths that are informative with
respect to relative abundance – for example the Micromonosporaceae family is relatively

abundant at 680 nm, 1480–1490 nm,1690–1710 nm, and has a low relative abundance at
1425 nm and 2100 nm (Fig. 1c). (For interpretation of the references to colour in this figure
legend, the reader is referred to the Web version of this article.)
Download : Download high-res image (2MB) Download : Download full-size image
Fig. 2. a–f: True colour composite image segments of tundra (2a), boreal (2b), savanna (2c).
Relative abundance of the three families (viz. AKIW659, Pseudomonadaceae,
Micromonosporaceae) in the tundra (2c), boreal (2d) and savanna (2e). Red dots indicate
sample locations that are overlapping with the shown image segments – note that some
samples are located on adjacent imagery (which are not shown). (For interpretation of the
references to colour in this figure legend, the reader is referred to the Web version of this
article.)
The relative abundance of the 3 families in the tundra, boreal and savannah hyperspectral
images were modelled from reflectance values for each image and compared with a true
colour image (Fig. 2). To interpret the relative abundance imagery (Schmidtlein et al., 2012) the
colour of each pixel indicates the relative abundance of each family projected as a mix of the 3
colour guns, with the dominant colour indicating the respective higher relative abundance of
the microbial families in the landscape. A cursory inspection of the relative abundance images
for the tundra, boreal and savanna imagery (Fig. 2) highlights an overall reddish colour for the
tundra (where AKIW659 (red) dominates the tundra), boreal is dominated by blue
(Pseudomonadaceae), and savanna by green (Micromonosporaceae) – with the relative
abundances of the families being confirmed by summary statistics (Table 1). In Fig. 2, larger
areas of similar colour are hotspots of higher relative abundance for that specific mix of
microbial families. Black pixels are water pixels, filtered using a negative NDVI values.
4. Discussion
The family AKIW659 occurs with high relative abundance in the tundra, the
Psedomonadaceae family is abundant in the boreal, and for Mediterranean climates the
Micromonosporaceae family has a high relative abundance (Table 1). There is a relative
abundance of AKIW659 associated with an important (VIP >1.0) negative regression coefficient
at 760 nm, highlighting the contrast between the lack of leaf structure of lichens/mosses
(Kuusinen et al., 2020) (i.e., poorly developed NIR shoulder) in the tundra with the well-
developed NIR shoulder structure typical of grasses, shrubs and trees in tundra, boreal and
savanna vegetation (Schmidt and Skidmore, 2001; Hope et al., 1993) (Fig. 1a). The wavelength
range 660–680 nm has a high regression coefficient (Fig. 1a) and has been shown to be
statistically significant for differentiating vegetation communities in this region west of
Barrow Alaska (Davidson et al., 2016). This part of the visible spectrum, coinciding with a peak
contribution of the regression coefficient and VIP (0.7), differentiates grasses (including
hummock grass) and non-ericaceous shrubs from mosses (Tieszen and Johnson, 1968), with a
higher chlorophyll a pigment concentration being associated with green vegetation (Liu et al.,
2017). We do note that even though there is a high correlation between relative abundance of
AKIW659 and this wavelength, the VIP of 0.7 indicates the wavelength does not contribute a
large amount of information to the model. The double peak regression coefficient at 940–
960 nm has been noted as a wavelength range important for discriminating tundra vegetation
(Bratsch et al., 2017). This concurs with findings that structural carbon elements (including
lignin, starch, and cellulose) are strongly associated with the SWIR at 950–1000 nm (starch)
and at 1120 nm (lignin and cellulose) (Curran, 1989), and as noted in Table 2 the AKIW659
family has a specialized role in carbon cycling by oxidizing carbohydrates. The strong negative
coefficient feature at 1305–1355 nm is associated with the waterlogged conditions of the Arctic
tundra, which also has a lower leaf area index (LAI) contrasting with the boreal forests to the
south, exposing a greater component of the soil/ground/water reflectance signal, interspersed
with sparser shrub and grass vegetation (Knipling, 1970). The atmospheric absorption bands
for atmospheric cloud and water vapour around 1250–1400 nm are generally not of interest to
terrestrial remote sensing, being included in bands on satellites for automatic cirrus cloud
detection at 1375 nm (Richter et al., 2011).
The Pseudomonadaceae family occurred with high relative abundance in all boreal samples
(Table 1). In the boreal forest, tree spectra dominate an increasing part of hyperspectral image
reflectance. The 1120 nm wavelength is known to be associated with lignin (Curran, 1989)
which is an important structural carbohydrate component of the fibre content of leaves and
litter in boreal forests (Richardson, 2004), and the PLSR analysis indicates this wavelength
range is important for explaining the variability of the relative abundance of the
Pseudomonadaceae family (Fig. 1b). The importance of the red edge wavelengths 680 nm
(Gitelson et al., 1996) association with the relative abundance of Pseudomonadaceae (Fig. 1b)
can be explained by the chlorophyll content of the tree species of the Canadian and Alaskan
Shield reflecting relatively more near-infrared energy compared to the wetland areas of the
tundra. The 750–925 nm range is part of the spectrum associated with light scattering due to
leaf and canopy structure (Ali et al., 2016) coupled with low absorption by water, lignin and
cellulose, resulting in high plant reflectance and transmittance (Rautiainen et al., 2018). The
wavelength range 1970–2020 nm associates with higher protein/nitrogen as well as
cellulose/starch content leading to lower reflectance (Curran, 1989). In boreal forests, there is a
relative increase in the lignin and cellulose content resulting from slower litter
decomposition at the cooler temperatures of more northerly latitudes (Richardson, 2004). Low
soil macronutrient levels, especially nitrogen and phosphorus further reduce the rate of litter
decomposition (Archibold, 1995). The boreal forests are generally podzolic, acid soils, with a
well-developed leaf litter layer containing a relatively high content of cellulose and lignin
(Handley, 1954) providing a suitable habitat (Table 3) for the observed high relative abundance
of Pseudomonadaceae.
For Mediterranean climates, the Micromonosporaceae have been isolated from soil and plant
materials (Genilloud Trujillo et al., 2015), and in our data have a high relative abundance
(Table 1), as this family is able to withstand and colonize drier habitats evidenced by a number
of stress related genes being detected (Carro et al., 2019). Micromonosporaceae can withstand
high temperatures (Khan et al., 2020). This family has been isolated in semiarid grasslands and
are thought to have an important role in the degradation of cellulose (Yeager et al., 2017). The
640–660 nm wavelength associated with Micromonosporaceae (Fig. 1c) relates to increased
canopy nitrogen content and chlorophyll a+b absorption (Curran, 1989) resulting from
Micromonosporaceae taxa having a function in plant growth promotion (including nitrogen
fixation) (Hirsch and Valdes, 2010) through plant/rhizosphere nitrogen fixing nodules (Trujillo
et al., 2014). The ability of Micromonosporaceae to synthesize pigments, likely for UV

protection (Carro et al., 2018), is indicated by association with the 640–660 nm range.
Micromonosporaceae are soil-associated plant symbionts – including symbionts of grasses
(Yeager et al., 2017). The 750–800 nm wavelength is associated with internal leaf scattering of
the vegetation (Liu et al., 2014). A 1480–1490 nm feature is associated with leaf cellulose and
starch (Curran, 1989) of drier broadleaf plants, the feature at 1425 nm indicates a relatively
lower lignin content (Curran, 1989) and the feature at 2100 nm a relatively lower non-
structural carbohydrate concentration (Ramirez et al., 2015); these features have been noted for
broadleaf vegetation when compared to coniferous (needle) trees of the boreal ecosystem
(Prescott et al., 2004; Li et al., 2016). Lignin, starch, protein and nitrogen are associated with
the 1690–1710 nm feature indicative of nitrogen fixation, and the breakdown of cell wall
material (Hirsch and Valdes, 2010).
The promising results demonstrate the potential to interpret the relative abundance and
ecological function of the three selected families for the savanna, boreal and tundra habitats,
from the colour composite images (Fig. 2). For example, of the three selected species, in the
boreal (Fig. 2e) the Pseudomonadaceae (bright blue pixels) dominate. In Fig. 2e, the abundance
of Pseudomonadaceae is relatively higher (bluer colour) on the north facing slopes associated
with the mesic black spruce (coniferous forest) (Fig. 2b) compared with the south facing slopes
dominated by drier hardwood forests (LTER, 2021) – see the NEON field site network Caribou-
Poker Creeks Research Watershed data for reference (https://www.neonscience.org/field-
sites/bona). The southern facing slopes has Micromonosporaceae mixing with the
Pseudomonadaceae (i.e., greener tinge in Fig. 2e), which associates with increasing mixed
coniferous-broadleaf forest, broadleaf forest and open woodland/tussock vegetation (Haugen
et al., 1982). The warmer microclimate on the southern facing slopes also makes this habitat
more suitable for Micromonosporaceae compared to the colder northern facing slopes (Table
2). The broadleaf (birch-aspen dominated) forest is a persistent seral hardwood community
maintained by repeated disturbance approximately every 75 years, with closed canopy and a
shrub/herbaceous understory (Boucher and Reid, 2010), and is associated with a higher
relative abundance of Micromonosporaceae. Regular disturbance results in spruce exclusion
as this species is fire intolerant. Because conifer litter decays more slowly than hardwood litter
(Fassnacht and Gower, 1999; McClaugherty et al., 1985) the relative abundance of
Pseudomonadaceae will be higher. Further we note the mixing of the Micromonosporaceae
(green-blue areas) in the boreal-taiga vegetation as well as a trace of AKWI659 (red) in the
wetter bog areas (Fig. 2b).
Forest vegetation is the habitat driving the relative abundance of the microbiome. Land cover
mapping, such as for the Cariboh-Poker Catchment (Haugen et al., 1982), has distinct
vegetation types across which the microbial families subtly merge and these microbial
families associate with specific habitats across vegetation types (e.g., AKWI659 is in the
phylum Acidobacteria and characterized as an aerobic heterotroph typically occurring with
sedges, grasses and various mosses). This subtle grading of microbial families occur within
specific habitat types allows the relative abundance of the microbiome taxa to be visualized
across a vegetation or habitat type - our novel approach provides detailed spatial information
about the microbiome biodiversity across vegetation habitats, albeit requiring eDNA and
hyperspectral imagery and relatively complicated processing. As discussed above for the three
example microbial families that are representative biomarkers for each of the three
ecosystems, we deduce that the functional ecology of each microbiome family mirrors the
canopy functional traits through the soil resources, as proposed by the Spectranomics
approach (Asner and Martin, 2016). Canopy plant species have evolved plant functional types
and traits that echo soil structure and nutrient concentration (Massmann et al., 2022), which
in turn are further adapted by the abiotic conditions of a particular location including climate,
geology, and topography (Higgins et al., 2014). Image spectroscopy can associate ecosystem
processes with plants, for example through biogeochemical cycles, thereby mirroring soil
conditions (Asner and Martin, 2016) and specifically the soil microbiome (Aponte et al., 2013).
In our approach presented, spectral properties mirror the biochemical, structural, and
functional properties of vegetation. That is, the vegetation or Plant Functional Types (PFT) are
indicators of soil properties as well as the soil mirobiome. Plant Functional Types (PFTs) is an
important concept linking species with broad vegetation types, based on ecosystem structure,
ecosystem function (phenology and physiology), as well as mirroring plant responses to soil
and ecosystem conditions (Ustin and Gamon, 2010). Climate as well as the soil biophysical
environment, control ecological niche and a biome's vegetation structure and function (Ustin
and Gamon, 2010).
The relationship between above ground vegetation and soil processes are known to be both
species and ecosystem specific (Ma et al., 2020; Hess and Austin, 2014), with patterns and
processes even demonstrated at global scales (Ordonez et al., 2009). Understanding soil-plant
relationships requires knowledge of species interactions and soil processes at the level of
ecosystems. While spectral indicators of the biota may be context-dependent, many general
relations have been developed (Roy et al., 2006), which allows deductive interpretation of
rhizosphere functions, as demonstrated in this study for three ecosystems across North
America, and further increases understanding of soil microbial processes at a global scale
(Niemann et al., 2015). Soil microbiome biodiversity as measured by eDNA reveals a complex
community composition. Microbiome families respond to habitat, with the habitat
comprising the physical and biotic factors that support the species’ survival and reproduction
(Kiesewetter and Afkhami, 2021; Wang et al., 2022; Keet et al., 2019). By considering the co-
occurring eDNA and image spectroscopy data, our results demonstrate that ecosystem
functions and hence related processes such as the presence and relative abundance of
microbiome families, can be deductively interpreted, also in ecosystems with vegetation cover.
Given the eDNA profile data contain thousands of taxonomic groups, which in this study were
reduced to a list of 81 discriminative families (see Table 1 for some examples of these 81
families), it is straightforward to repeat our novel method for different families as well as at
different taxonomic ranks. For example, the three families (viz. AKIW659, Pseudomonadaceae,
Micromonosporaceae) chosen to demonstrate our method (Fig. 1) could be replaced in the
true colour composite images with any combination of three families from the 81
discriminative families, or indeed other taxonomic ranks such as species, order, or phylum. A
key insight of this paper is that it is possible, with the highly multivariate (“deep”) eDNA
profile data, to predict the relative distribution of any combination of the taxonomic groups
present in the study area from remote sensing. We show that it is also methodologically
straightforward to discover the (multivariate) relations between the microbiome taxa (in an
eDNA profile database), the image spectroscopy, and environmental data, as demonstrated
above for the Psedomonadaceae family.
The relatively limited number of samples may result in statistical models being potentially
overfitted (Shen et al., 2020). We acknowledge that ideally more samples would be available in
this study, though the scarcity of publicly available image spectroscopy imagery that coincide
with eDNA profile data across natural areas in North America constrained the study. Though
research indicates that as few as 20–30 samples may be adequate for significant and
meaningful PLSR analyses (Bastien et al., 2005; Tenenhaus et al., 2005; Kock and Hadaya,
2018), other researchers propose more samples (>100) (Burnett et al., 2021). Several image
spectroscopy studies using hard-to-collect biodiversity variables have successfully interpolated
environmental variables using n < 25 in addition to the range 26 < n < 50 (Axelsson et al., 2013;
Ramoelo et al., 2013; Bi et al., 2019; Askari et al., 2015; Cho et al., 2007). Additionally, by
reducing the number of PLSR predictor variables to between 3 and 5 regression components,
model overfitting is further diminished (Shen et al., 2020). We checked overfitting of the PLSR
model using standard leave-one-out (LOO) cross-validation for the predicted versus measured
scores, demonstrating that results are repeatable and robust (Ramoelo et al., 2013; Askari et al.,
2015; Fernandez-Habas et al., 2021) (Supplementary Fig. S1) for the three family taxa, with a
RMSEP of 0.01 and 0.04 as well as a R2 of 0.54 and 0.85 between the measured and predicted
values for the three families (Table 3). We anticipate other scientists will further test and
develop the concept described in this paper for robustness and repeatability when mapping
the microbiome from remote sensing, as more eDNA data become available that are
coincident with hyperspectral images.
As well as complementing in situ data collection, the method using remote sensing
demonstrated here may be further upscaled with respect to the number of taxa for whom the
relative abundance is estimated, potentially providing further insight into the ecosystem
health of ecosystems. The so-called ‘Linnean gap’ (i.e., gaps caused by most species not being
described with traditional sampling) may easily be filled by increasing the number of marker
genes for additional species, as part of the eDNA analysis process (Bini et al., 2006). If multiple
images obtained at different dates are available, then it will be possible to interpolate
microbial species relative abundance using time-series imagery (Zhang et al., 2003).
5. Conclusion
The requirements for monitoring anthropogenic biodiversity change, as detailed by the UN
Sustainable Development Goals as well as the UN Convention on Biodiversity Aichi Targets,
necessitates the application of novel technologies to provide spatially contiguous information
rapidly and accurately to update biodiversity information for companies, government, and
NGOs. The post-2020 Biodiversity Targets of the UN Convention on Biodiversity are not
merely aspirational; the implementation of policy is currently constrained by the lack of data
and indicators (Geijzendorffer et al., 2016). Several imaging spectroscopy satellites with spatial
sampling of 20–30 m have been and will launch between 2019 and 2026, ushering in real
opportunities to combine this newly emerging space technology with eDNA in situ taxonomic
data. Here we demonstrate that ecosystem function can be the bridge between these two
rapidly developing technological approaches, combining the massive information content of
in situ eDNA data with the high spatial accuracy and global overview provided by imaging
spectroscopy. As reference collections of eDNA barcode taxonomic databases are coupled with
next generation satellite imaging spectrometers, a more comprehensive and complete picture
of biodiversity delineating species across all kingdoms of life will become possible by applying
and building upon the techniques presented in this paper. Combining eDNA and imaging
spectroscopy holds the promise of accurately, rapidly, and cost-effectively monitoring
biodiversity over the terrestrial surface of the Earth.
Declaration of competing interest

The authors declare that they have no known competing financial interests or personal
relationships that could have appeared to influence the work reported in this paper.
Acknowledgment
This project has received funding and support from the European Research Council (ERC)
under the European Union's Horizon 2020 research and innovation programme (grant
agreement n° 834709).
Appendix A. Supplementary data

The following is the Supplementary data to this article:
Download : Download Word document (120KB)
Multimedia component 1.
Recommended articles
References
Albaina et al., 2016 A. Albaina, M. Aguirre, D. Abad, M. Santos, A. Estonba
18S rRNA V9 metabarcoding for diet characterization: a critical evaluation with two
sympatric zooplanktivorous fish species
Ecol. Evol., 6 (2016), pp. 1809-1824, 10.1002/ece3.1986
View Record in Scopus Google Scholar
Ali et al., 2016 A.M. Ali, R. Darvishzadeh, A.K. Skidmore, I. van Duren
Effects of canopy structural variables on retrieval of leaf dry matter content and
specific leaf area from remotely sensed data
IEEE J. Sel. Top. Appl. Earth Obs. Rem. Sens., 9 (2016), pp. 898-909,
10.1109/jstars.2015.2450762
Aponte et al., 2013 C. Aponte, L.V. Garcia, T. Maranon

Tree species effects on nutrient cycling and soil biota: a feedback mechanism
favouring species coexistence
For. Ecol. Manage., 309 (2013), pp. 36-46, 10.1016/j.foreco.2013.05.035
Article Download PDF View Record in Scopus Google Scholar
Archibold, 1995 O.W. Archibold

Ecology of World Vegetation
(first ed.), Chapman & Hall Ltd (1995)
9,152,190,256
Google Scholar
Askari et al., 2015 M.S. Askari, S.M. O'Rourke, N.M. Holden

Evaluation of soil quality for agricultural production using visible-near-infrared
spectroscopy
Geoderma, 243 (2015), pp. 80-91, 10.1016/j.geoderma.2014.12.012
Asner and Martin, 2016 G.P. Asner, R.E. Martin

Spectranomics: emerging science and conservation opportunities at the interface of
biodiversity and remote sensing
Glob. Ecol. Conserv., 8 (2016), pp. 212-219, 10.1016/j.gecco.2016.09.010
Axelsson et al., 2013 C. Axelsson, A.K. Skidmore, M. Schlerf, A. Fauzi, W. Verhoef

Hyperspectral analysis of mangrove foliar chemistry using PLSR and support vector
regression
Int. J. Rem. Sens., 34 (2013), pp. 1724-1743, 10.1080/01431161.2012.725958
Bastien et al., 2005 P. Bastien, V.E. Vinzi, M. Tenenhaus

PLS generalised linear regression
Comput. Stat. Data Anal., 48 (2005), pp. 17-46, 10.1016/j.csda.2004.02.005
Bi et al., 2019 K.Y. Bi, S. Gao, Z. Niu, C.S. Zhang, N. Huang

Estimating leaf chlorophyll and nitrogen contents using active hyperspectral LiDAR
and partial least square regression method
J. Appl. Remote Sens., 13 (2019), 10.1117/1.Jrs.13.034513

Google Scholar
Bini et al., 2006 L.M. Bini, J.A.F. Diniz, T. Rangel, R.P. Bastos, M.P. Pinto
Challenging Wallacean and Linnean shortfalls: knowledge gradients and conservation
planning in a biodiversity hotspot
Divers. Distrib., 12 (2006), pp. 475-482, 10.1111/j.1366-9516.2006.00286.x
Boucher and Reid, 2010 T. Boucher, M.S. Reid

Terrestrial ecological system: western north American boreal mesic birch-aspen forest
https://explorer.natureserve.org/Taxon/ELEMENT_GLOBAL.2.817381/Western_North_
American_Boreal_Mesic_Birch-Aspen_Forest (2010)
Google Scholar
Bratsch et al., 2017 S. Bratsch, H. Epstein, M. Buchhorn, D. Walker, H. Landes

Relationships between hyperspectral data and components of vegetation biomass in
Low Arctic tundra communities at Ivotuk, Alaska
Environ. Res. Lett., 12 (2017), p. 14, 10.1088/1748-9326/aa572e
Google Scholar
Burnett, 2021 A.C. Burnett, et al.

A best-practice guide to predicting plant traits from leaf-level hyperspectral data using
partial least squares regression
J. Exp. Bot., 72 (2021), pp. 6175-6189, 10.1093/jxb/erab295
Caporaso et al., 2010 J.G. Caporaso, et al.

QIIME allows analysis of high-throughput community sequencing data
Nat. Methods, 7 (2010), pp. 335-336, 10.1038/nmeth.f.303
Google Scholar
Caporaso et al., 2011 J.G. Caporaso, et al.

Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample
Proc. Natl. Acad. Sci. U. S. A, 108 (2011), pp. 4516-4522, 10.1073/pnas.1000080107
Carro et al., 2018 L. Carro, et al.

Genome-based classification of micromonosporae with a focus on their
biotechnological and ecological potential
Sci. Rep., 8 (2018), 10.1038/s41598-017-17392-0
Google Scholar
Carro et al., 2019 L. Carro, et al.

Uncovering the potential of novel micromonosporae isolated from an extreme hyper-
arid Atacama Desert soil
Sci. Rep., 9 (2019), 10.1038/s41598-019-38789-z

Google Scholar
Chariton et al., 2010 A.A. Chariton, L.N. Court, D.M. Hartley, C.M. J, C.M. Hardy
Ecological assessment of estuarine sediments by pyrosequencing eukaryotic ribosomal
DNA
Front. Ecol. Environ., 8 (2010), pp. 233-238
CrossRef View Record in Scopus Google Scholar
Cho et al., 2007 M.A. Cho, A. Skidmore, F. Corsi, S.E. van Wieren, I. Sobhan
Estimation of green grass/herb biomass from airborne hyperspectral imagery using
spectral indices and partial least squares regression
Int. J. Appl. Earth Obs. Geoinf., 9 (2007), pp. 414-424, 10.1016/j.jag.2007.02.001
Cousins et al., 2004 M.A. Cousins

R.K. Robinson, C.A. Batt, B.D. Patel (Eds.), Encyclopedia of Food Microbiology (2004)
997-893 (978-0-12-227070-3
Google Scholar
Curran, 1989 P.J. Curran

Remote-sensing of foliar chemistry
Remote Sens. Environ., 30 (1989), pp. 271-278, 10.1016/0034-4257(89)90069-2
Davidson et al., 2016 S. Davidson, et al.

Mapping arctic tundra vegetation communities using field spectroscopy and
multispectral satellite data in North Alaska, USA
Rem. Sens., 8 (2016), p. 978, 10.3390/rs8120978
De Barba et al., 2014 M. De Barba, et al.

DNA metabarcoding multiplexing and validation of data accuracy for diet assessment:
application to omnivorous diet
Mol. Ecol. Resour., 14 (2014), pp. 306-323, 10.1111/1755-0998.12188
de Chaves et al., 2019 M.G. de Chaves, et al.

Acidobacteria subgroups and their metabolic potential for carbon degradation in
sugarcane soil amended with vinasse and nitrogen fertilizers
Front. Microbiol., 10 (2019), p. 16, 10.3389/fmicb.2019.01680
Google Scholar
ESA, 2022 ESA

CHIME (Copernicus)
https://directory.eoportal.org/web/eoportal/satellite-missions/c-missions/chime-
copernicus (2022)
Google Scholar
Evans et al., 2016 N.T. Evans, et al.

Quantification of mesocosm fish and amphibian species diversity via environmental
DNA metabarcoding
Mol. Ecol. Resour., 16 (2016), pp. 29-41, 10.1111/1755-0998.12433
Fassnacht and Gower, 1999 K.S. Fassnacht, S.T. Gower

Comparison of the litterfall and forest floor organic matter and nitrogen dynamics of
upland forest ecosystems in north central Wisconsin
Biogeochemistry, 45 (1999), pp. 265-284, 10.1007/bf00993003
Feilhauer et al., 2016 H. Feilhauer, D. Doktor, S. Schmidtlein, A.K. Skidmore

Mapping pollination types with remote sensing
J. Veg. Sci., 27 (2016), pp. 999-1011, 10.1111/jvs.12421
Fernandez-Habas et al., 2021 J. Fernandez-Habas, et al.

Investigating the potential of Sentinel-2 configuration to predict the quality of
Mediterranean permanent grasslands in open woodlands
Sci. Total Environ., 791 (2021), 10.1016/j.scitotenv.2021.148101
Google Scholar
Geijzendorffer et al., 2016 I.R. Geijzendorffer, et al.

Bridging the gap between biodiversity data and policy reporting needs: an Essential
Biodiversity Variables perspective
J. Appl. Ecol., 53 (2016), pp. 1341-1350, 10.1111/1365-2664.12417
Genilloud Trujillo et al., 2015 O. Genilloud

M.E. Trujillo, et al. (Eds.), Bergey's Manual of Systematics of Archaea and Bacteria, 1-28,
Wiley & Sons, Inc. (2015)
Google Scholar
Gilbert et al., 2014 J.A. Gilbert, J.K. Jansson, R. Knight

The Earth Microbiome project: successes and aspirations
BMC Biol., 12 (2014), 10.1186/s12915-014-0069-1
Google Scholar
Gitelson et al., 1996 A.A. Gitelson, M.N. Merzlyak, H.K. Lichtenthaler

Detection of red edge position and chlorophyll content by reflectance measurements
near 700 nm
J. Plant Physiol., 148 (1996), pp. 501-508, 10.1016/s0176-1617(96)80285-9

Guerschman et al., 2009 J.P. Guerschman, et al.

Estimating fractional cover of photosynthetic vegetation, non-photosynthetic
vegetation and bare soil in the Australian tropical savanna region upscaling the EO-1
Hyperion and MODIS sensors
Remote Sens. Environ., 113 (2009), pp. 928-945, 10.1016/j.rse.2009.01.006
Handley, 1954 W.R.C. Handley

Mull and More Formation in Realtion to Forest Soils
Imperial Forestry Institute, London (1954)
1-114
Google Scholar
Haugen et al., 1982 R.K. Haugen, C.W. Slaughter, K.E. Howe, S.L. Dingman
Hydrology and Climatology of the Caribou-Poker Creeks Research Watershed
(1982), p. 41
Washington, D.C.
Google Scholar
Hess and Austin, 2014 L.J.T. Hess, A.T. Austin

Pinus ponderosa alters nitrogen dynamics anddiminishes the climate footprint in
natural ecosystems of Patagonia
J. Ecol., 102 (2014), pp. 610-621, 10.1111/1365-2745.12228
Higgins et al., 2014 M.A. Higgins, et al.

Linking imaging spectroscopy and LiDAR with floristic composition and forest
structure in Panama
Hirsch and Valdes, 2010 A.M. Hirsch, M. Valdes

Micromonospora: an important microbe for biomedicine and potentially for
biocontrol and biofuels
Soil Biol. Biochem., 42 (2010), pp. 536-542, 10.1016/j.soilbio.2009.11.023
Hope et al., 1993 A.S. Hope, J.S. Kimball, D.A. Stow

The relationship between tussock tundra spectral reflectance properties and biomass
and vegetation composition
Int. J. Rem. Sens., 14 (1993), pp. 1861-1874
Jesmok et al., 2016 E.M. Jesmok, J.M. Hopkins, D.R. Foran

Next-generation sequencing of the bacterial 16S rRNA gene for forensic soil
comparison: a feasibility study
J. Forensic Sci., 61 (2016), pp. 607-617, 10.1111/1556-4029.13049
Keet et al., 2019 J.H. Keet, A.G. Ellis, C. Hui, J.J. Le Roux
Strong spatial and temporal turnover of soil bacterial communities in South Africa's
hyperdiverse fynbos biome
Soil Biol. Biochem., 136 (2019), 10.1016/j.soilbio.2019.107541
Google Scholar
Khan et al., 2020 M.J. Khan, S.D. Jurburg, J.Z. He, G. Brodie, D. Gupta
Impact of microwave disinfestation treatments on the bacterial communities of no-till
agricultural soils
Eur. J. Soil Sci., 71 (2020), pp. 1006-1017, 10.1111/ejss.12867
Kiesewetter and Afkhami, 2021 K.N. Kiesewetter, M.E. Afkhami

Microbiome-mediated effects of habitat fragmentation on native plant performance
New Phytol., 232 (2021), pp. 1823-1838, 10.1111/nph.17595
Knipling, 1970 E.B. Knipling

Physical and physiological basis for the reflectance of visible and near-infrared
radiation from vegetation
Remote Sens. Environ., 1 (1970), pp. 155-159
Article Download PDF Google Scholar
Kock and Hadaya, 2018 N. Kock, P. Hadaya

Minimum sample size estimation in PLS-SEM: the inverse square root and gamma-
exponential methods
Inf. Syst. J., 28 (2018), pp. 227-261, 10.1111/isj.12131
Koyama et al., 2014 A. Koyama, M.D. Wallenstein, R.T. Simpson, J.C. Moore
Soil bacterial community composition altered by increased nutrient availability in
Arctic tundra soils
Front. Microbiol., 5 (2014), p. 16, 10.3389/fmicb.2014.00516
Kuusinen et al., 2020 N. Kuusinen, J. Juola, B. Karki, S. Stenroos, M. Rautiainen

A spectral analysis of common boreal ground lichen species
Remote Sens. Environ., 247 (2020), p. 13, 10.1016/j.rse.2020.111955
Google Scholar
Lausch et al., 2016 A. Lausch, et al.

Linking Earth Observation and taxonomic, structural and functional biodiversity: local
to ecosystem perspectives
Ecol. Indicat., 70 (2016), pp. 317-339, 10.1016/j.ecolind.2016.06.022
Li et al., 2016 N. Li, N.P. He, G.R. Yu, Q.F. Wang, J. Sun
Leaf non-structural carbohydrates regulated by plant functional groups and climate:
evidences from a tropical to cold-temperate forest transect
Ecol. Indicat., 62 (2016), pp. 22-31, 10.1016/j.ecolind.2015.11.017
Liu et al., 2014 L.Y. Liu, W.J. Huang, R.L. Pu, J.H. Wang
Detection of internal leaf structure deterioration using a new spectral ratio index in
the near-infrared shoulder region
J. Integr. Agric., 13 (2014), pp. 760-769, 10.1016/s2095-3119(13)60385-8
Liu et al., 2017 N.F. Liu, P. Budkewitsch, P. Treitz

Examining spectral reflectance features related to Arctic percent vegetation cover:
implications for hyperspectral remote sensing of Arctic tundra
Lozupone and Knight, 2007 C.A. Lozupone, R. Knight

Global patterns in bacterial diversity
Proc. Natl. Acad. Sci. U. S. A, 104 (2007), pp. 11436-11440, 10.1073/pnas.0611525104
Google Scholar
LTER, 2021 B.C. LTER

Study Sites & Design: Regional Site Network
University of Alaska, Fairbanks (2021)
http://www.lter.uaf.edu/research/study-sites-regional
Google Scholar
Ma et al., 2020 J. Ma, et al.

Herbaceous layer determines the relationship between soil respiration and
photosynthesis in a shrub-dominated desert plant community
Plant Soil, 449 (2020), pp. 193-207, 10.1007/s11104-020-04484-6
Marques et al., 2014 A. Marques, et al.

A framework to identify enabling and urgent actions for the 2020 Aichi Targets
Basic Appl. Ecol., 15 (2014), pp. 633-638, 10.1016/j.baae.2014.09.004
Massmann et al., 2022 A. Massmann, M.A. Cavaleri, S.F. Oberbauer, P.C. Olivas, S. Porder
Foliar stoichiometry is marginally sensitive to soil phosphorus across a lowland
tropical rainforest
Ecosystems, 25 (2022), pp. 61-74, 10.1007/s10021-021-00640-w
McClaugherty et al., 1985 C.A. McClaugherty, J. Pastor, J.D. Aber, J.M. Melillo
Forest litter decomposition in relation to soil-nitrogen dynamics and litter quality
Ecology, 66 (1985), pp. 266-275, 10.2307/1941327
Mevik and Wehrens, 2007 B.-H. Mevik, R. Wehrens

The pls package: principal component and partial least squares regression in R
J. Stat. Software, 18 (2007), pp. 1-23, 10.18637/jss.v018.i02
Mitchell et al., 2018 A.L. Mitchell, et al.

EBI Metagenomics in 2017: enriching the analysis of microbial communities, from
sequence reads to assemblies
Nucleic Acids Res., 46 (2018), pp. D726-D735, 10.1093/nar/gkx967
Nastal and Pavlick, 2022 J. Nastal, R.P. Pavlick

Surface Biology and Geology (SBG)
(2022)
https://science.nasa.gov/earth-science/decadal-sbg
Google Scholar
Niemann et al., 2015 K.O. Niemann, G. Quinn, R. Stephen, F. Visintini, D. Parton

Hyperspectral remote sensing of mountain pine beetle with an emphasis on previsual
assessment
Can. J. Rem. Sens., 41 (2015), pp. 191-202, 10.1080/07038992.2015.1065707
Nuccio et al., 2016 E.E. Nuccio, et al.

Climate and edaphic controllers influence rhizosphere community assembly for a wild
annual grass
Ecology, 97 (2016), pp. 1307-1318, 10.1890/15-0882.1
Ordonez et al., 2009 J.C. Ordonez, et al.

A global study of relationships between leaf traits, climate and soil measures of
nutrient fertility
Global Ecol. Biogeogr., 18 (2009), pp. 137-149, 10.1111/j.1466-8238.2008.00441.x
Google Scholar
Orwin et al., 2018 K.H. Orwin, I.A. Dickie, R. Holdaway, J.R. Wood
A comparison of the ability of PLFA and 16S rRNA gene metabarcoding to resolve soil
community change and predict ecosystem functions
Soil Biol. Biochem., 117 (2018), pp. 27-35, 10.1016/j.soilbio.2017.10.036
Palleroni Starr et al., 1981 N.J. Palleroni

Mortimer P. Starr, et al. (Eds.), The Prokaryotes: A Handbook on Habitats, Isolation,
and Identification of Bacteria, Springer Berlin Heidelberg (1981), pp. 655-665
CrossRef Google Scholar
Pereira et al., 2013 H.M. Pereira, et al.

Essential biodiversity variables
Science, 339 (2013), pp. 277-278, 10.1126/science.1229931
Philippot et al., 2009 L. Philippot, et al.

Spatial patterns of bacterial taxa in nature reflect ecological traits of deep branches of
the 16S rRNA bacterial tree
Environ. Microbiol., 11 (2009), pp. 3096-3104, 10.1111/j.1462-2920.2009.02014.x
Potapov et al., 2017 P. Potapov, et al.

The last frontiers of wilderness: tracking loss of intact forest landscapes from 2000 to
2013
Sci. Adv., 3 (2017), p. 13, 10.1126/sciadv.1600821
Google Scholar
Prescott et al., 2004 C.E. Prescott, L. Vesterdal, C.M. Preston, S.W. Simard
Influence of initial chemistry on decomposition of foliar litter in contrasting forest
types in British Columbia
Can. J. For. Res., 34 (2004), pp. 1714-1729, 10.1139/x04-040
PRISMA, 2022 A.S.I. PRISMA

https://www.asi.it/en/earth-science/prisma/ (2022)
Rahman et al., 2003 A.F. Rahman, J.A. Gamon, D.A. Sims, M. Schmidts
Optimum pixel size for hyperspectral studies of ecosystem function in southern
California chaparral and grassland
Remote Sens. Environ., 84 (2003), pp. 192-207, 10.1016/s0034-4257(02)00107-4
Ramirez et al., 2015 J.A. Ramirez, et al.

Near-infrared spectroscopy (NIRS) predicts non-structural carbohydrate
concentrations in different tissue types of a broad range of tree species
Methods Ecol. Evol., 6 (2015), pp. 1018-1025, 10.1111/2041-210x.12391

Ramoelo et al., 2013 A. Ramoelo, et al.

Non-linear partial least square regression increases the estimation accuracy of grass
nitrogen and phosphorus using in situ hyperspectral and environmental data
ISPRS J. Photogrammetry Remote Sens., 82 (2013), pp. 27-40,
10.1016/j.isprsjprs.2013.04.012
Rautiainen et al., 2018 M. Rautiainen, et al.

Spectral properties of coniferous forests: a review of in situ and laboratory
measurements
Rem. Sens., 10 (2018), p. 207
Richardson, 2004 A.D. Richardson

Foliar chemistry of balsam fir and red spruce in relation to elevation and the canopy
light gradient in the mountains of the northeastern United States
Plant Soil, 260 (2004), pp. 291-299, 10.1023/b:Plso.0000030179.02819.85
Richter et al., 2011 R. Richter, X.J. Wang, M. Bachmann, D. Schlapfer

Correction of cirrus effects in Sentinel-2 type of imagery
Int. J. Rem. Sens., 32 (2011), pp. 2931-2941, 10.1080/01431161.2010.520346
Roesch et al., 2007 L.F. Roesch, et al.

Pyrosequencing enumerates and contrasts soil microbial diversity
ISME J., 1 (2007), pp. 283-290, 10.1038/ismej.2007.53
Roy et al., 2006 D.R. Roy, L. Boschetti, S.N. Trigg

Remote sensing of fire severity: assessing the performance of the normalized Burn
ratio
Geosci. Rem. Sens. Lett. IEEE, 3 (2006), pp. 112-116, 10.1109/lgrs.2005.858485
Sanders, 1968 H.L. Sanders

Marine benthic diversity - a comparative study
Am. Nat., 102 (1968), pp. 243-&, 10.1086/282541
Google Scholar
Schmidt and Skidmore, 2001 K.S. Schmidt, A.K. Skidmore

Exploring spectral discrimination of grass species in African rangelands
Int. J. Rem. Sens., 22 (2001), pp. 3421-3434
Schmidtlein et al., 2012 S. Schmidtlein, H. Feilhauer, H. Bruelheide

Mapping plant strategy types using remote sensing
J. Veg. Sci., 23 (2012), pp. 395-405, 10.1111/j.1654-1103.2011.01370.x
Segata et al., 2011 N. Segata, et al.

Metagenomic biomarker discovery and explanation
Genome Biol., 12 (2011), 10.1186/gb-2011-12-6-r60
Google Scholar
Shen et al., 2020 L.Z. Shen, et al.

Hyperspectral estimation of soil organic matter content using different spectral
preprocessing techniques and PLSR method
Rem. Sens., 12 (2020), 10.3390/rs12071206
Google Scholar
Siegenthaler et al., 2019 A. Siegenthaler, O.S. Wangensteen, C. Benvenuto, J. Campos, S.

Mariani
DNA metabarcoding unveils multiscale trophic variation in a widespread coastal
opportunist
Mol. Ecol., 28 (2019), pp. 232-249, 10.1111/mec.14886
Siok and Ewiak, 2020 K. Siok, I. Ewiak

The simulation approach to the interpretation of archival aerial photographs
Open Geosci., 12 (2020), pp. 1-10, 10.1515/geo-2020-0001
Skidmore et al., 2015 A.K. Skidmore, et al.

Agree on biodiversity metrics to track from space
Nature, 523 (2015), pp. 403-405, 10.1038/523403a
Stark et al., 2014 S. Stark, M.K. Mannisto, A. Eskelinen

Nutrient availability and pH jointly constrain microbial extracellular enzyme activities
in nutrient-poor tundra soils
Plant Soil, 383 (2014), pp. 373-385, 10.1007/s11104-014-2181-y
Storch, 2022 T. Storch

EnMAP
https://www.dlr.de/eoc/en/desktopdefault.aspx/tabid-5514/20470_read-47899/ (2022)
Google Scholar
Taberlet et al., 2012 P. Taberlet, E. Coissac, F. Pompanon, C. Brochmann, E. Willerslev

Towards next-generation biodiversity assessment using DNA metabarcoding
Mol. Ecol., 21 (2012), pp. 2045-2050, 10.1111/j.1365-294X.2012.05470.x
Taberlet et al., 2018 P. Taberlet, A. Bonin, L. Zinger, E. Coissac

Environmental DNA for Biodiversity REsearch and Monitoring
Oxford University Press (2018)
Google Scholar
Tenenhaus et al., 2005 M. Tenenhaus, V.E. Vinzi, Y.M. Chatelin, C. Lauro

PLS path modeling
Comput. Stat. Data Anal., 48 (2005), pp. 159-205, 10.1016/j.csda.2004.03.005
Thompson et al., 2017 L.R. Thompson, et al.

A communal catalogue reveals Earth's multiscale microbial diversity
Nature, 551 (2017), p. 457, 10.1038/nature24621
Tieszen and Johnson, 1968 L.L. Tieszen, P.L. Johnson

Pigment structure of some arctic tundra communities
Ecology, 49 (1968), p. 370, 10.2307/1934476
Trujillo et al., 2014 M.E. Trujillo, et al.

Genome features of the endophytic actinobacterium micromonospora lupini strain
lupac 08: on the process of adaptation to an endophytic life style?
PLoS One, 9 (2014), 10.1371/journal.pone.0108522
Google Scholar
Ustin and Gamon, 2010 S.L. Ustin, J.A. Gamon

Remote sensing of plant functional types
New Phytol., 186 (2010), pp. 795-816, 10.1111/j.1469-8137.2010.03284.x
Wang et al., 2022 J.J. Wang, et al.

Embracing mountain microbiome and ecosystem functions under global change
New Phytol., 234 (2022), pp. 1987-2002, 10.1111/nph.18051
Weiss et al., 2017 S. Weiss, et al.

Normalization and microbial differential abundance strategies depend upon data
characteristics
Microbiome, 5 (2017), 10.1186/s40168-017-0237-y
Google Scholar
C.M. Yeager, et al.

Yeager et al., 2017
Polysaccharide degradation capability of actinomycetales soil isolates from
a semiarid grassland of the Colorado plateau
Appl. Environ. Microbiol., 83 (2017), 10.1128/aem.03020-16
Google Scholar
Yilmaz et al., 2011 P. Yilmaz, et al.

Minimum information about a marker gene sequence (MIMARKS) and minimum
information about any (x) sequence (MIxS) specifications
Nat. Biotechnol., 29 (2011), pp. 415-420, 10.1038/nbt.1823
Zhang et al., 2003 X.Y. Zhang, et al.

Monitoring vegetation phenology using MODIS
Remote Sens. Environ., 84 (2003), pp. 471-475, 10.1016/s0034-4257(02)00135-9
Article Download PDF Google Scholar
Cited by (0)
Crown Copyright © 2022 Published by Elsevier B.V.
Copyright © 2022 Elsevier B.V. or its licensors or contributors.

ScienceDirect® is a registered trademark of Elsevier B.V.

Mapping The Relative Abundance of Soil Microbiome Biodiversity From eDNA and Remote Sensing - ScienceDirect

Uploaded by

Copyright:

Available Formats

Mapping The Relative Abundance of Soil Microbiome Biodiversity From eDNA and Remote Sensing - ScienceDirect

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mapping The Relative Abundance of Soil Microbiome Biodiversity From eDNA and Remote Sensing - ScienceDirect

Uploaded by

Copyright:

Available Formats

10/21/22, 11:34 AM Mapping the relative abundance of soil microbiome biodiversity from eDNA and remote sensing - ScienceDirect

Science of Remote Sensing

Mapping the relative abundance of soil microbiome

Outline Share Cite

https://doi.org/10.1016/j.srs.2022.100065 Get rights and content

• Mapping relative abundance of microbiome families across savanna, boreal and

• Linear Discriminate Analysis (LDA) identifies indicator eDNA microbial families.

2.1. Remote sensing data

The hyperspectral data are delivered as standardized ortho-corrected as well as

2.2. eDNA data

Biology Laboratory-European Bioinformatics Institute (EMBL-EBI) (Mitchell et al., 2018) and

Family Ecosystem LDA PLSR MRA% All MRA% MRA% MRA%

Pseudomonadaceae boreal 3.64 0.54 0.26 0.02 0.67 0.05

AKIW659 tundra 3.74 0.85 0.21 0.00 0.00 0.61

Anaerolinaceae tundra 3.88 0.54 0.33 0.00 0.12 0.82

Syntrophaceae tundra 4.66 0.72 1.77 0.00 0.04 5.04

Micromonosporaceae savanna 4.40 0.85 1.04 3.36 0.03 0.03

Family Ecosystem LDA PLSR MRA% All MRA% MRA% MRA%

Solirubrobacteraceae savanna 4.35 0.65 1.03 3.22 0.10 0.04

Geodermatophilaceae savanna 4.02 0.72 0.51 1.50 0.03 0.11

Table 2. Known ecological habitat of the three families (AKIW659, Pseudomonadaceae,

AKIW659 AKIW659 is in the phylum Acidobacteria being characterized as an aerobic

Pseudomonadaceae Pseudomonadaceae are in the phylum Proteobacteria, and a metabolically diverse

Micromonosporaceae Micromonosporaceae of the phylum Actinobacteria, are aerobic and chemo-

Family R2 (Taberlet et al., 2018) RMSE (adjCV) Components Variance explained

AKIW659 0.85 0.013 5 95.34

Family R2 (Taberlet et al., 2018) RMSE (adjCV) Components Variance explained

Pseudomonadaceae 0.54 0.028 3 69.42

Micromonosporaceae 0.85 0.034 5 93.99

Download : Download high-res image (687KB) Download : Download full-size image

respect to relative abundance – for example the Micromonosporaceae family is relatively

Download : Download high-res image (2MB) Download : Download full-size image

et al., 2014). The ability of Micromonosporaceae to synthesize pigments, likely for UV

Declaration of competing interest

Appendix A. Supplementary data

Download : Download Word document (120KB)

Aponte et al., 2013 C. Aponte, L.V. Garcia, T. Maranon

Archibold, 1995 O.W. Archibold

Askari et al., 2015 M.S. Askari, S.M. O'Rourke, N.M. Holden

Asner and Martin, 2016 G.P. Asner, R.E. Martin

Axelsson et al., 2013 C. Axelsson, A.K. Skidmore, M. Schlerf, A. Fauzi, W. Verhoef

Bastien et al., 2005 P. Bastien, V.E. Vinzi, M. Tenenhaus

Bi et al., 2019 K.Y. Bi, S. Gao, Z. Niu, C.S. Zhang, N. Huang

J. Appl. Remote Sens., 13 (2019), 10.1117/1.Jrs.13.034513

Boucher and Reid, 2010 T. Boucher, M.S. Reid

Bratsch et al., 2017 S. Bratsch, H. Epstein, M. Buchhorn, D. Walker, H. Landes

Burnett, 2021 A.C. Burnett, et al.

Caporaso et al., 2010 J.G. Caporaso, et al.

Caporaso et al., 2011 J.G. Caporaso, et al.

Carro et al., 2018 L. Carro, et al.

Carro et al., 2019 L. Carro, et al.

Sci. Rep., 9 (2019), 10.1038/s41598-019-38789-z

Cousins et al., 2004 M.A. Cousins

Curran, 1989 P.J. Curran

Davidson et al., 2016 S. Davidson, et al.

De Barba et al., 2014 M. De Barba, et al.

de Chaves et al., 2019 M.G. de Chaves, et al.

ESA, 2022 ESA