Ecological Indicators 96 (2019) 628–634

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

Implications of some major human-induced activities on forest cover using T

extended change matrix quantity and intensity analysis based on historical
Landsat data from the Kloto District, Togo
⁎
Yawovi S. Kogloa,b, , Thomas Gaiserc, Wilson A. Agyared, Jean M. Sogbedjie, Kokou Kouamif
a
Institut Togolais de Recherche Agronomique (ITRA), BP 1163 Lomé, Togo
b
WASCAL Climate Change and Land Use, Department of Civil Engineering, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana
c
Institute of Crop Science and Resource Conservation, Crop Science Group, University of Bonn, 53115 Bonn, Germany
d
Kwame Nkrumah University of Science and Technology, Department of Agricultural Engineering, West African Science Service Centre on Climate Change and Adapted
Land Use, Kumasi, Ghana
e
Université de Lomé, Département des Sciences du Sol, BP 1515, Togo
f
Université de Lomé, Faculté des Sciences, Laboratoire de Recherche Forestière (LRF), West African Science Service Centre on Climate Change and Human Security, BP
1515 Togo

A R T I C LE I N FO A B S T R A C T

Keywords: This study analyses forest reference level in terms of loss, gain and transitions among forest, cocoa agroforestry,
Forest cassava, maize, settlement and others in the Kloto district (Togo) for REDD+ and sustainable forest and agri-
REDD+ culture. The pixel-based classification was adopted and combined with the extended change matrix quantity and
Extended matrix intensity analysis using 32-year (1985–2017) Landsat data and land use information from land owners and
Intensity analysis
farmers. Results indicate an active forest loss (19.5%) with dormant gain (0.8%). Forest is involved in most
Soil degradation
transitions as the most targeted category with the largest transition being a forest to cocoa agroforestry while the
Agroecology
avoiding transition was from forest, cocoa agroforestry, maize, cassava and settlement to unclassified classes
(e.g. road, water body) and vice versa. Other targeting categories were from forest to settlement, cassava and
maize Thus, both cash and food crops are major contributors of forest loss. The study concludes that cropland
land degradation is the main reason that explains the significant conversion of forest lands to stable agricultural
lands. Therefore, review of the existing cropping and farming systems by promoting agroecology systems (e.g.
agroforestry, rotational cropping, mixing cropping with pulses) to sustain and restore soil degradation while
mitigating climate change, forest degradation and provide food security for the rural communities is re-
commended. Economic measures such as: trade-off compensations for agroecology practices and afforestation
and reforestation through farmer’s association initiatives could be encouraged to limit forest extensions.

1. Introduction (2005) investigation on tropical deforestation revealed that 1–2 billion

tonnes (roughly 15–25% annual global greenhouse) of carbon are being
The deliberate conversion of tropical forests to agricultural lands, lost per year since 1990’s. Lasco (2002) also posited that land use
intensive deforestation, and forest degradation and inappropriate land change and forest conversion are a significant source of CO2 con-
management practices are the major impediment to forests and soil tributing to around 1.7 ± 0.6 Pg C per year (e.g., Parks and Hardie,
carbon sequestration potential (Denman et al., 2007; Lal, 2008a,b; 1995; Plantinga and Birdsey, 1995; Callaway and Mccarl, 1996;
Stockmann et al., 2013; Koglo et al., 2016, 2017) in changing climate. Stavins, 1999; Kanime et al., 2013). Besides, land use change (e.g. forest
According to Valentine et al. (2000), forests have a remarkable ability to agricultural and non-agricultural lands) is one of the primary drivers
to store carbon in both plants and soils and they are valuable natural of climate change and global warming (e.g., changes in albedo and
break and sinkers of atmospheric carbon dioxide. The largest source of radiative forces) with serious environmental and socioeconomic con-
GHG emissions in most tropical countries is from deforestation and cerns. Following this background, in Togo, close analysis of available
forest degradation. Fearnside and Laurance (2003) and Houghton statistics on greenhouse gas emissions revealed significant released

⁎
Corresponding author at: WASCAL CC&LU, College of Engineering, KNUST, Kumasi, Ghana.
E-mail addresses: koglo.y@edu.wascal.org (Y.S. Koglo), tgaiser@uni-bonn.de (T. Gaiser), wagyare@yahoo.co.uk (W.A. Agyare),
mianikpo@yahoo.com (J.M. Sogbedji), kokoukouami@hotmail.com (K. Kouami).

https://doi.org/10.1016/j.ecolind.2018.09.042
Received 29 April 2018; Received in revised form 16 September 2018; Accepted 22 September 2018
Available online 03 October 2018
1470-160X/ © 2018 Elsevier Ltd. All rights reserved.
Y.S. Koglo et al. Ecological Indicators 96 (2019) 628–634

Fig. 1. Land use land cover map of the study area.

from land use change (forest to agriculture) compared to energy, in- from deforestation and forest abasement; protection of timberland
dustry, and waste management sector (MERF, 2001). This crucial step is carbon stocks; sustainable management of forests and improvement of
paramount for the implementation of precautionary measures to miti- forest carbon stocks (REDD+) have turned into the major theme of
gate the emissions resulting from forest conversion to agriculture lands. environmental change discussion. For instance, the 2030 Agenda for
Thus, it urges to answer following questions related to the changes, viz: Sustainable Development and the Development Goals and the United
when? (period and identification of the changes via remote sensing and Nations Strategic Plan for Forest (UNSPF) 2017 – 3030 made and bold
GIS technology), how much? (referring to the significance of the an ambitious commitment to halting deforestation in 2020 by reversing
changes), where? (locations) moreover, what to do? (what necessary the loss of forest cover and increasing forest area by 3% (www.cpfweb.
and sustainable decision or measure to make). Knowing the initial time org). REDD+ does not only rolls back climate change but also proffers
(time 1) and the transition time (time 2), change detection (pixel, area additional co-benefit in response to food security, conservation of
or percent of the map) analysis or transition matrix analysis can be nature and improvement of socio-economic livelihoods of rural and
performed to account for the net changes for each land use type from poorest communities that are likely to experience the adverse effects of
time1 to time 2. However, this method fails to account for the like- climate changes. Accordingly, the main question to ask is how much
lihood gains and losses of a category from one location to the other. forest changes and at which rates? and what are the impacts of an-
Moreover, it does not inform on the intensity of the changes (targeted/ thropogenic activities on forests carbon loss or gain in Kloto district
avoided and active/dormant categories) as well as the swap change (the (Togo)? Several studies have been carried out on land use, land cover
difference between the total change and net change or spatial reloca- changes from one period to another (e.g., Han et al., 2009;Munsi et al.,
tion), swap distance and persistent land use category (Pontius et al., 2010; Badjana et al., 2014; Folega et al., 2014; Sambou et al., 2015;
2004). Thus extended cross-tabulation matrix analysis combined with Diwediga et al., 2017), intensity analysis and its implications on carbon
the intensity of gains and losses and transition intensity accounts for cycle (e.g., Pontius et al., 2004; Manandhar et al., 2010; Aldwaik and
allocation changes based on land cover change signals. It is against this Pontius, 2012; Runfola and Pontius, 2013; Villamor et al.,2014).
background that, scientific communities and policy makers have given However, little is known about the differentiation of food and cash
credit to driven forces of deforestation and forest degradations projects crops extent (e.g. maize, cassava and cocoa agro-forestry) and the ex-
and researchers to limit forest depletion and degradation. The role of tent of urbanization (population growth). Moreover, these studies fail
forests in a changing climate is of utmost importance to meet the re- to integrate historical land occupation information and the extended
quirements of sustainable development goals. Diminishing emission quantity and intensity analysis. It is, therefore, a prerequisite to

629
Y.S. Koglo et al. Ecological Indicators 96 (2019) 628–634

investigating long-term transitional dynamics of annual and perennial statistic data) with the assistance of google earth historical data records,
farming system expansions and population growth on the gain and loss and classification was performed using a Maximum Likelihood
of native forest extent at a local and national level in developing Classifier. And the post-classification technique was initiated to derive
countries, notably Togo to facilitate forest reference levels (FRL) and the extended cross-tabulation matrix for land use change and intensity
forest emission reference level (FREL) assessment. It is a way to for- analysis. Images were classified with 100% accuracy with Kappa coef-
mulate solid arguments for the REDD+ establishment and accurate ficient equal to 1 (Appendices A and B).
measuring, reporting, and verification of REDD+ to United Nations
organs and Climate change, donors. From the foregoing introduction, 3.1. Quantity of land use change
this paper aims to assess 32 years transitional and intensity changes on
the native forest extent due to maize, cassava, cocoa agroforestry Change analysis of the thematic classes from time 1 (1985) to time 2
cropping, and urbanization extent for establishing a logical framework (2017) (Fig. 1) as percent of map was assessed in terms of persistence
model for sustainable forest management while increasing crop pro- (diagonal entries), total loss (total row of category i minus percentage
ductivity and establishing recommended landscape programmes. map of the same category of final year), total gain (total column of
Simply put its presents: (i) the quantity of changes in terms of gross category j minus category j of the initial year). The extended change
gains, losses, persistence, net and swap changes; (ii) the intensity of matrix was also used to derive the total, net and swap changes of each
gains and losses and (iii) the intensity of transition from forest to other land use category. The total change (TC) of a category is the sum of its
land use type (maize, cassava, cocoa agro-forestry farming and settle- gross gain and loss. At the main time, the net change (NC) of a category
ments). is the difference between its gross gain and loss while the swap change
(SC) of a category is the difference of a total change and net change for
2. Study area the category.

The study was conducted in Kloto district, Togo and covers an ap- 3.2. Intensity analysis
proximate land area of 528.23 km2. It encompasses 13 sub-districts
located in the northwest of the capital, Lomé (Fig. 1). Kloto is located The gain and loss intensity (GI, LI) were duly derived from the
between 0.50° and 0.77° East and 6.75° and 7.0° North. The district change matrix knowing that the uniform intensity (UI) of the transition
covers a total area of 528.23 km2. The major economic activity is is the total gain or loss of the four thematic classes. The gain and loss
farming of food crops (e.g. maize, cassava) and cash crop (cocoa and intensity help to identify dormant (GI or LI less than UI) or active (GI or
coffee agroforestry). The average production of maize and cassava from LI greater than UI) categories. The LI and GI of a category were cal-
1990 to 2016 revealed an annual production of 10,928 and 19,498 culated by dividing the loss of a category in 1985 and a gain of a ca-
tonnes for maize and cassava over an area of 8291 and 3080 ha, re- tegory in 2017 with the size of the categories in 1985 and 2017, re-
spectively. From 2014 to 2016 cocoa agroforestry of 2762 ha produced spectively. Similarly, the loss and gain transition intensity of forest,
1041 tonnes annually (DSID, 2017: analyzed statistic data). cocoa agroforestry, maize and cassava farms were calculated from the
extended matrix. In this case, the transition intensity (TI) denotes tar-
3. Data collection and analysis geted (TI greater than UI) or avoided (TI smaller than UI) categories
when a particular category change into another category. Thus, the loss
Two Landsat data (5 and OLI) of March 1985 and April 2017 were transition intensity (LTI) from forest to cocoa agroforestry, for instance,
downloaded from the USGS with cloud cover less than 10% using path was obtained by dividing the loss of forest to cocoa agroforestry in 1985
(193) and row (055). We assumed same phonological conditions be- with the gain of cocoa agro-forestry in 2017. Therefore, the LTI is
cause of the bimodal climate season. Acquired images were pre-pro- compared to the hypothesized Uniform Loss/Gain Intensity (ULI/UGI)
cessed for atmospheric (cloud and noise removal) and radiometric of forest loss to cocoa agroforestry, maize, and cassava. The ULI/UGI is
(brightness) corrections as well as layer stacking and sub-setting in the union of forest loss to cocoa agroforestry, maize and cassava divided
ArcGIS. Thereafter, thematic analysis was performed in ENVI software by the sum of cocoa agroforestry, maize and cassava and vice versa.
based on six (06) Land Use, Land Cover types (Fig. 1), viz: forest, cocoa Moreover, the gain transition intensity (GTI) of forest from cocoa
agroforestry, maize, cassava farms, settlements and unclassified. agroforestry farms, for example, is computed by dividing the size of
A field campaign was organized from May to October 2017 for cocoa agroforestry in 1985 with the gain of forest from grassland.
historical land occupational information via an interview with land-
owners and farmers and 40 random points of each land use type were 4. Results and discussion
collected to train and validate the classification. Image calibration of
the two years was done using ground through and archive land occu- 4.1. Quantity of land use transition
pational geo-referenced points (40 in total) of each land use type of the
subsequent years of available statistics (DSID, 2017: Agriculture Census Results (Table 1 and 2) revealed significant persistence (49.50%) of

Table 1
Change matrix from 1985 to 2017 in the percentage of the map.
2017

Unclassified Settlement Cassava Maize Cocoa agroforestry Forest Total 1985 Loss

1985 Unclassified 49.49 0.00 0.02 0.00 0.01 0.00 49.53 0.04
Settlement 0.00 0.54 0.01 0.02 0.00 0.00 0.57 0.03
Cassava 0.00 0.05 0.70 0.08 0.43 0.05 1.32 0.62
Maize 0.00 0.71 3.14 0.51 1.04 0.17 5.58 5.06
Cocoa agroforestry 0.00 1.29 10.51 0.75 1.56 0.52 14.63 13.07
Forest 0.00 2.21 12.12 0.65 4.49 8.90 28.37 19.47

Total 2017 49.50 4.80 26.49 2.03 7.54 9.65 1.00 38.291
Gain 0.00 4.26 25.79 1.51 5.97 0.75 38.29

630
Y.S. Koglo et al. Ecological Indicators 96 (2019) 628–634

Table 2
Gain, loss, persistence, total change, net change and swap change per category.
Gain Loss Persistence Total Net change Swap
change change

Unclassified 0.00 0.04 49.49 0.04 −0.04 0.00

Settlement 4.26 0.03 0.54 4.29 4.23 0.06
Cassava 25.79 0.62 0.70 26.41 25.18 1.23
Maize 1.51 5.06 0.51 6.57 −3.55 3.02
Cocoa agroforestry 5.97 13.07 1.56 19.05 −7.10 11.95
Forest 0.75 19.47 8.90 20.22 −18.72 1.50

Total 38.29 38.29 61.71

Fig. 3. Gain and Loss intensity per category in percentage.

Fig. 2. Quantity of land use transition.

unclassified pixels (e.g. road, water bodies) with the marginal loss
(0.04) and gain (0.00), respectively. The transition of other classes
(settlement, Cassava, Maize, Cocoa Agroforestry, and forest) to such
pixels is quite negligible (0%) over 32 years (1985–2017) land use and
land cover transition analysis.
In the meantime, Table 2 shows a total gain/loss of 38.29% with
61.71% persistence. Very dynamic categories were distinguished in
terms of total changes (Fig. 2) namely cassava (26.41%), forest
(20.22%) and cocoa agroforestry (19.05%); dynamic classes encom-
passing maize (6.57%) and settlement (4.29%) and unclassified pixels
as a stable class (0.04%) in terms of loss-gain and vice versa over
32 years period. The net change (Fig. 2) reveals high forest land re-
duction (18.72%) follow by cocoa agroforestry (7.10%), maize (3.55%)
and others (0.04%) to the detriment of cassava and settlement which
are gaining 25.18 and 4.23%, respectively. The spatial relocation re-
garding loss or gain pixels shows a maximum distance change of
11.95 km for coca agroforestry and 1.50 km of forest lands. Cassava
Fig. 4. Land use land cover intensity of transitions expressed as (a) Forest, (b)
lands gain from others land use categories at a minimum distance of
Cocoa agroforestry, (c) Maize, (d) Cassava and (e) Settlement.
1.23 km less than maize (3.02 km) and settlement (0.06 km). In sum,
the quantitative analysis indicates that there are a change from forest to
agricultural (cassava, maize, cocoa agro-forestry) and non-agricultural reduced significantly. In the meantime, some gain more sites compared
(settlement and unclassified) land use systems. However, it did not give to others (Fig. 3). This is the case of forest which gains less than its
the information about the speed of the change as well as the intensity at losses. Forest gain is dormant (7.76 < 38.29) while the conversion of
which the changes are occurring. In that line, intensity analysis was forest to other land use, land cover types (e.g. settlement, cassava,
performed to determine dormant and active categories (Fig. 3) of the cocoa agroforestry and maize) is active. This expansion of cultivated
transition as well as the targeted and avoided categories (Fig. 4) of the and non-cultivated areas is intense for cassava farms (97.36) followed
32 years transition analysis.Fig. 1. by settlement (88.80), cocoa agroforestry (79.28) and maize (74.63)
farming systems. For the targeted and avoided categories of loss and
gain transitions, Fig. 4 (a–e) were used. The loss analysis (Fig. 4a) re-
4.2. Intensity analysis vealed that deforestation and forest degradation is mainly targeted by
cocoa agroforestry farms (59.60%) followed by housing expansion
Results (Fig. 3) depict the loss and gain intensity per category in (46.08%), cassava (45.73%) and maize (32.26%) farms in 2017. In the
percentage of map. Some categories, namely: unclassified and settle- meantime, the cocoa agroforestry loss (Fig. 4b) is targeted by cassava
ment is losing slowly while, cassava, maize, cocoa agroforestry, and (39.69%) and maize (36.96%) farming. In contrast, maize lost (Fig. 4c)
forest are losing faster. Simply put, unclassified and settlement is dor- 14.75, 13.79 and 11.86% to the detriment of the settlement, cocoa
mant regarding area expansion, while, the remaining categories

631
Y.S. Koglo et al. Ecological Indicators 96 (2019) 628–634

and population growth on forest extinction. These results are consistent

with similar studies in Indonesia (Villamor et al., 2014; Gao et al.,
2016); Ghana (Aloô and Pontius, 2008) and in China (Huang et al.,
2012; Zhou et al., 2014) where, transition analysis revealed a sys-
tematic transition from either forest to croplands and from croplands to
build-up. However, they failed to account for the part of individual
systems for forest reference and emission level determination as a
benchmark for REDD+ projects. Kloto district is a forest zone with the
dominance of indigenous trees and cocoa agroforestry systems. These
results could be explained by population growth, the market price of
cash and food crops and farming and cropping systems in use. The
perception of farmers and landowners bring to the conclusion that,
population growth from 1985 to 2017 had implied a tremendous in-
crease in food demand and expansion of residences. Recent population
census (MPDAT, 2010) depicted a significant increase of rural popula-
tion at national level to the rate of 25.2% (1981) for 37.7% (2010). The
population density of Kloto in 1981 was between 50 and 100 in-
habitant/km2 (Anipah et al., 1989) against 263 inhabitants/km2 in
2010 for a total population of 139,043 inhabitants (MPDAT, 2010).
Moreover, economic incentives to cocoa and cassava prices in the
market had promoted the conversion of many maize farms to cassava
and cocoa agroforestry farms. Interview from farmers revealed that
most of the farmers are adopting cocoa agroforestry farming systems
(e.g. cocoa and coffee; cocoa, coffee, and cassava or cocoa, cassava and
plantain) to increase and diversify their level of income. For others, the
price of these cash and food crops are more stable at national and in-
ternational markets compared to maize price. In addition to cocoa-
cassava agroforestry systems, farmers also practice mono-cropping
cassava systems with either indigenous or improved varieties while
maize cultivation is mainly mono-crop. In some areas, maize is culti-
vated in association with cassava or as sequential cropping systems. The
intensity of mono-cropping and the decline of soil fertility oblige
farmers to clear more forests to the detriment of cocoa agroforestry,
cassava and maize farms as revealed by the 32 years land use transition
analysis (Koglo et al., 2018). These analyses are similar to recent studies
conducted in northern Togo by Diwediga et al. (2017). An evaluation of
multifunctional landscapes progression in the mountainous basin of the
Mo River (Togo, West Africa) from 1972 to 2014 uncovered an im-
perative anthropogenic land change prompting land degradation. This
also explained part of the results of this study whereby, conventional
agriculture intensifications and population growth are targeting native
forest stability. The systematic land use transition was from forest to
cocoa agroforestry followed by settlement expansion, cassava and
maize farming with the total dormancy of forest gain and very active
non-forest land cover types.
Fig. 4. (continued)
5. Conclusions
agroforestry, and cassava farm. The transition intensity of cassava,
settlement, and unclassified categories (Fig. 4d), Fig. 4 depicts the high Land use transition analysis at sub certain level using extended
proportion of cassava loss to cocoa agroforestry (5.75%) followed by matrix analysis, remote sensing and land occupational data information
maize (4.10%) which also targeted settlement and others land use cover is scarce in developing countries for accurate REDD+ implementation,
categories at 1.19 and 0.10%, respectively. Transition to and from accountability and establishment of cost-effective sustainable land and
settlement is reciprocal with maize farming (Fig. 4e), whereby, settle- forest sustainable management policies. This study uses this technique
ment losses and gains simultaneously from maize. Furthermore, the with pertinent results. Conclusions drawn revealed an intensive loss
marginal gain of the forest came from cassava, cocoa agroforestry, and (19.47%) of forest in the past 32 years (1985–2017) to the detriment of
maize at 3.69, 3.56 and 3.13%, respectively while cocoa agroforestry cocoa agroforestry, maize, cassava farming and settlement expansion.
took 32.89 and 31.37% from cassava and forest rather than maize In contrast, forest gain (0.75%) slowly while, settlement, cocoa agro-
(18.64%). In the meantime, cassava targeted cocoa agroforestry farms forestry, cassava, and maize are gaining faster. The systematic transi-
(71.84%) compared to maize (56.33%) and forest (42.70%). Results tion was from forest to cocoa agroforestry followed by cassava, settle-
also revealed that settlement expansion is 12.70, 8.81 and 7.80% due to ment, and maize. These results are more explicit in formulating
maize and cocoa agroforestry farms and forest lands reduction respec- sustainable policies for each land use land cover type in forest man-
tively whereas, unclassified gain did not target any land use classes. agement while promoting sustainable agriculture in forested zones
From the analysis, it is quite an evident that deforestation and forest (Kloto, Togo). This confirms the robustness of the land use transition
degradation transition intensity analysis in Kloto district is both agri- analysis when combining satellite data with strong post classification
cultural (cocoa, agroforestry, cassava, and maize) and non-agricultural analysis and historical land use information from landowners and/or
(settlement expansion) issues. The results present the role of agriculture farmers. In other words, it gives the exact effect of each land use type

632
Y.S. Koglo et al. Ecological Indicators 96 (2019) 628–634

(e.g. food, cash crop or residential) on forest stability, loss or gain. This Acknowledgements
study recommends demographic policies (e.g. family planning; in-
centive measures for small households), capacity building of land This work is part of Yawovi S. Koglo PhD thesis on Climate Change
owners and farmers (e.g. sustainable forest management). As both cash and Land Use at Kwame Nkrumah University of Science and
and food crops are major impediments to forest loss, it urges to review Technology (KNUST), Kumasi. It is fully funded by the German Ministry
the existing cropping and farming systems in the area by promoting of Education and Science through the West African Science Service
agro-ecology systems (e.g. agroforestry, rotational cropping, mixing Centre on Climate Change and Adapted Land Use (WASCAL). Authors
cropping with pulses) as natural fertilizers to regain and restore soil also present their profound gratitude to Kloto district farmers for their
fertility while mitigating climate change, forest decline and providing collaboration and to Dr. Ayi K. Adden, Mr. Gle Kossivi and field assis-
more foods for the rural communities. tants for their tremendous supports in the fulfilment of this project. We
also appreciate the peer review works from anonymous reviewers in
strengthening the quality of the paper.

Appendix A. Analysis report land use land cover map 1985.

Overall Accuracy = (1162188/1162188) 100%

Kappa Coefficient = 1

Table a. Confusion matrix

Class Unclassified Settlement Cassava Maize Cocoa agroforestry Forest Total

Unclassified 100 0 0 0 0 0 100

Settlement 0 100 0 0 0 0 100
Cassava 0 0 100 0 0 0 100
Maize 0 0 0 100 0 0 100
Cocoa agroforestry 0 0 0 0 100 0 100
Forest 0 0 0 0 0 100 100
Total 100 100 100 100 100 100

Table b. Omission and Commission error

Class Producer accuracy (%) User accuracy (%) Producer accuracy (Pixels) User accuracy (Pixels)

Unclassified 100 100 575228/575228 575228/575228

Settlement 100 100 9216/9216 9216/9216
Cassava 100 100 83425/83425 83425/83425
Maize 100 100 93908/93908 93908/93908
Cocoa agroforestry 100 100 148596/148596 148596/148596
Forest 100 100 251815/251815 251815/251815

Appendix B. Analysis report land use land cover map 2017.

Overall Accuracy = (1162188/1162188) 100%.

Kappa Coefficient = 1

Table a. Confusion matrix

Class Unclassified Settlement Cassava Maize Cocoa agroforestry Forest Total

Unclassified 100 0 0 0 0 0 100

Settlement 0 100 0 0 0 0 100
Cassava 0 0 100 0 0 0 100
Maize 0 0 0 100 0 0 100
Cocoa agroforestry 0 0 0 0 100 0 100
Forest 0 0 0 0 0 100 100
Total 100 100 100 100 100 100

Table b. Omission and Commission error

Class Producer accuracy (%) User accuracy (%) Producer accuracy (Pixels) User accuracy (Pixels)

Unclassified 100 100 627/627 627/627

(continued on next page)

633
Y.S. Koglo et al. Ecological Indicators 96 (2019) 628–634

Table b. Omission and Commission error (continued)

Class Producer accuracy (%) User accuracy (%) Producer accuracy (Pixels) User accuracy (Pixels)

Settlement 100 100 658000/658000 658000/658000

Cassava 100 100 206470/206470 206470/206470
Maize 100 100 49902/49902 49902/49902
Cocoa agroforestry 100 100 101300/101300 101300/101300
Forest 100 100 145889/145889 145889/145889

Appendix C. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.ecolind.2018.09.042.

References effects on tropical agriculture systems: the role of improved soil management prac-
tices in regulating soil moisture, temperature and carbon losses. J. Soil Sci. Environ.
Manage. 8 (6), 104–112. https://doi.org/10.5897/JSSEM2017.0619.
Aldwaik, S.Z., Pontius Jr., G., 2012. Intensity analysis to unify measurements of size and Koglo, Y.S., Agyare, W.A., Diwediga, B., Sogbedji, J.M., Adden, A.K., Gaiser, T., 2018.
stationarity of land changes by interval, category, and transition. Landscape Urban Remote sensing-based and participatory analysis of forests, agricultural land dy-
Plann. 106 (1), 103–114. https://doi.org/10.1016/j.landurbplan.2012.02.010. namics, and potential land conservation measures in Kloto District (Togo, West
Aloô, C.A., Pontius Jr., R.G., 2008. Identifying systematic land-cover transitions using Africa). Soil Syst. 2, 49. https://doi.org/10.3390/soilsystems2030049.
remote sensing and GIS: the fate of forests inside and outside protected areas of Lal, R., 2008b. Soil carbon stocks under present and future climate with specific reference
Southwestern Ghana. Environ. Plann. 35, 280–296. https://doi.org/10.1068/ to European ecoregions. Nutrient Cycling Agroecosyst. 81 (2), 113–127. https://doi.
b32091. org/10.1007/s10705-007-9147-x.
Anipah, K., Gozo, K., Nyassogbo, K., 1989. La population togolaise dans ses grandes Lal, R., 2008a. Sequestration of atmospheric CO2 in global carbon pools. Energy Environ.
lignes. Le Fonds des Nations-Unies pour la Population (FNUAP), Lome, Togo, 8p. Sci. 1 (1), 86–100.
(accessed 04.09.18). Lasco, R.D., 2002. Forest carbon budgets in Southeast Asia following harvesting and land
Badjana, H.M., Selsam, P., Wala, K., et al., 2014. Assessment of land-cover changes in a cover change. Science in China series c life sciences-English edition-, pp. 55–64.
sub-catchment of the Oti basin (West Africa): a case study of the Kara River basin. Manandhar, R., Odeh, I.O., Pontius, R.G., 2010. Analysis of twenty years of categorical
Zbl. Geol. Paläont Teil I, Jg 1, 151–170. land transitions in the Lower Hunter of New South Wales. Australia. Agric. Ecosyst.
Callaway, J.M., Mccarl, B., 1996. The economic consequences of substituting carbon Environ. 135 (4), 336–346. https://doi.org/10.1016/j.agee.2009.10.016.
payments for crop subsidies in US agriculture. Environ. Resour. Econ. 7 (1), 15–43. MERF, 2001. Première communication nationale initiale du Togo. Lomé. http://unfccc.
Denman, K.L., Brasseur, G., Chidthaisong, A., Ciais, P., Cox, P.M., Dickinson, R.E., int/resource/docs/natc/tognc1.pdf (accessed 20.03.16).
Hauglustaine, D., Heinze, C., Holland, E., Jacob, D., Lohmann, U., Ramachandran, S., Ministère de la Planification, du Développement et de l’Aménagement du Territoire
Da Silva Dias, P.L., Wofsy, S.C., Zhang, X., 2007. Couplings between changes in the (MPDAT), 2010. Recensement général de la population et de l’habitat (06 au 21
climate system and biogeochemistry. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Novembre 2010). République Togolaise, Lomé, Togo, 65p. (accessed 04.09.18).
Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Climate Change 2007: The Munsi, M., Malaviya, S., Oinam, G., Joshi, P.K., 2010. A landscape approach for quan-
Physical Science Basis. Contribution of Working Group I to the Fourth Assessment tifying land-use and land-cover change (1976–2006) in middle Himalaya. Regional
Report of the Intergovernmental Panel on Climate Change. Cambridge University Environ. Change 10, 145–155. https://doi.org/10.1007/s10113-009-0101-0.
Press, Cambridge, New York. Parks, P.J., Hardie, I.W., 1995. Least-cost forest carbon reserves: cost effective subsidies
Diwediga, B., Agodzo, S., Wala, K., Le Quang, B., 2017. Assessment of multifunctional to convert marginal agricultural land to forests. Land Econ. 71 (22), 136.
landscapes dynamics in the mountainous basin of the Mo River (Togo, West Africa). Plantinga, A.J., Birdsey, R.A., 1995. Carbon fluxes resulting from U.S. Private Timberland
J. Geograp. Sci. 27 (5), 579–605. https://doi.org/10.1007/s11442-017-1394-4. Management. Clim. Change 23, 37–53.
Fearnside, P.M., Laurance, W.F., 2003. Comment on ‘determination of deforestation rates Pontius, R.G., Shusas, E., Mceachern, M., 2004. Detecting important categorical land
of the world’s humid tropical forests’. Science 299, 1015. changes while accounting for persistence. Agric. Ecosyst. Environ. 101 (2), 251–268.
Folega, F., Zhang, C., Zhao, X., et al., 2014. Satellite monitoring of land-use and land- https://doi.org/10.1016/j.agee.2003.09.008.
cover changes in northern Togo protected areas. J. For. Res. 25, 385–392. Runfola, D.S.M., Pontius Jr., R.G., 2013. Measuring the temporal instability of land
Gao, Y., Pontius Jr, R. G., Giner, N. M., Kohyama, T. S., Osaki, M., Hirose, K. 2016. Land change using the flow matrix. Int. J. Geograph. Inf. Sci. 27 (9), 1696–1716. https://
Change Analysis from 2000 to 2004 in Peatland of Central Kalimantan, Indonesia doi.org/10.1080/13658816.2013.792344.
Using GIS and an Extended Transition Matrix. https://doi.org/10.1007/978-4-431- Sambou, S., Lykke, A.M., Sambou, H., Guiro, I., Sambou, B., Mbow, C., 2015. Land use-
55681-7. land cover change and drivers of deforestation in the Patako protected area (Center-
Han, J., Hayashi, Y., Cao, X., Imura, H., 2009. Evaluating land-use change in rapidly West of Senegal). Am. J. Environ. Prot. 4 (6), 306–317.
urbanizing China: case study of Shanghai. J. Urban Plann. Dev. 135 (4), 166–171. Stavins, R.N., 1999. The costs of carbon sequestration: a revealed-preference approach.
Houghton, R.A., 2005. Tropical Deforestation as a Source of GHGS Emissions Tropical Am. Econ. Rev. 89 (4), 994–1009.
Deforestation and Climate Changed Mutinho and Schwartzman. IPAM, Belem. Stockmann, U., Adams, M.A., Crawford, J.W., Field, D.J., Henakaarchchi, N., Jenkins, M.,
Huang, J., Pontius Jr., R.G., Li, Q., Zhang, Y., 2012. Use of intensity analysis to link Wheeler, I., 2013. The knowns, known unknowns and unknowns of sequestration of
patterns with processes of land change from 1986 to 2007 in a coastal watershed of soil organic carbon. Agric. Ecosyst. Environ. 164, 80–99. https://doi.org/10.1016/j.
southeast China. Appl. Geogr. 34, 371–384. https://doi.org/10.1016/j.apgeog.2012. agee.2012.10.001.
01.001. Valentine, R., Matteucci, G., Dolman, A.J., 2000. Respiration as the main determinant and
Kanime, N., Kaushal, R., Tewari, S.K., Raverkar, K.P., Chaturvedi, S., Chaturvedi, O.P., carbon balances in European forests. Nature 404, 861–865.
2013. Biomass production and carbon sequestration in different tree-based systems of Villamor, G., Pontius Jr., R.G., Noordwijk, M.V., 2014. Agroforest’s growing role in re-
central Himalayan tarai region. For Trees Livelihoods 22, 38–50. ducing carbon losses from Jambi (Sumatra). Indonesia 14 (2), 825–834. https://doi.
Koglo, Y.S., Abdulkadir, A., Feliciano, D., Okhimamhe, A.A., 2016. Efficacy of integrated org/10.1007/s10113-013-0525-4.
straw formulations on lowland rice field organic carbon and greenhouse gas emis- Zhou, P., Huang, J., Pontius Jr., R.G., Hong, H., 2014. Land classification and change
sions using CCAFS-MOT model in Niger State, Nigeria. Am. J. Exp. Agric. https://doi. intensity analysis in a coastal watershed of Southeast China. Sensors 11640–11658.
org/10.9734/AJEA/2016/27088. https://doi.org/10.3390/s140711640.
Koglo, Y.S., Abdulkadiri, A., Bissadu, K.D., Akamah, A.K., 2017. Mitigating droughts

634