International Journal of Electrical and Computer Engineering (IJECE) Vol.
No.
October 2025, pp.
ISSN: 2088-8708.
DOI: 10.
11591/ijece.
Remote sensing applied to cocoa crop identification, a thematic Luisa Fernanda Cuellar-Escobar1,2.
Vladimir Henao-Cyspedes3 Master Program in Remote Sensing.
Universidad Catylica de Manizales.
Manizales.
Colombia Sustainable Cacao Sourcing Team.
Casa Luker.
Manizales.
Colombia Faculty of Engineering and Architecture.
Universidad Catylica de Manizales.
Manizales.
Colombia
Article Info
ABSTRACT
Article history:
This article presents a thematic review of 25 publications related to the use of remote sensing techniques for the identification of cocoa crops from 2000 Although the use of remote sensing techniques is widely used for mapping different covers because it is very useful in discriminating them, the generation of maps of cocoa crops presents challenges due to their spectral behavior similar to that of forests.
This is because cocoa cultivation, being an agroforestry system that is developed in association with timber trees, causes the classification algorithms used to fail to differentiate between forest cover and cocoa crops.
For this reason, this study seeks to investigate the different remote sensing techniques used in the mapping of cocoa crops, as well as an analysis of the structure of the publications highlighting the connections between countries and the factors that motivated the authors to research this crop.
Received Oct 29, 2024 Revised Jun 11, 2025 Accepted Jun 30, 2025 Keywords:
Bibliometric Biblioshiny Cocoa Identification crops PRISMA Remote sensing This is an open access article under the CC BY-SA license.
Corresponding Author:
Vladimir Henao-Cyspedes Faculty of Engineering and Architecture.
Universidad Catylica de Manizales Cra 23 No 60-63.
Manizales.
Colombia Email: vhenao@ucm.
INTRODUCTION
Currently, cocoa in Colombia is positioned as a crop with an economic impact and that encompasses a social component associated with its impact in conflict zones since it is implemented as an alternative to illicit crops in the country.
In Colombia, cocoa is produced mostly by small and medium-sized farmers amounting to 38,000 families, achieving an annual production of 60,000 tons, which makes it a source of income .
Cocoa cultivation in some countries has a high proportion of forest loss, such as Ghana .
%).
Ivory Coast .
%), the two largest cocoa producers in the world, and Liberia .
%) .
, .
Due to the deforestation problem that occurs in the main cocoa-producing countries, regulation 2023/1115 .
was approved, known as the zero deforestation regulation, which is part of the green deal initiatives to promote actions that address climate change and protect the environment .
According to the regulation, no cocoa product can enter or leave the European Union (EU) market if deforestation or forest degradation occurs after December 31, 2020 .
To identify deforested areas, remote sensing and its different techniques have been useful, from the different studies on land use and cover .
Regarding cocoa cultivation, .
, it is carried out in association with service trees, whether fruit trees, timber trees, or a combination of these, which provide shade during their phenological development, which is why cocoa plantations have a great similarity with natural forests, making the process of classifying cocoa cover difficult due to the similarity between the spectral signatures of the covers .
Given the above, the objective of this article is to present a thematic review of the scientific production of remote sensing techniques that have allowed for the classification Journal homepage: http://ijece.
Int J Elec & Comp Eng
ISSN: 2088-8708
processes of cocoa crops.
It is expected that the results obtained will be an input in the search for remote sensing techniques that support compliance with regulation 1115 EUDR.
METHOD
Review structure The study of bibliographic data using statistical and numerical tools is called bibliometry and can be used to analyze patterns of scientific publications on a specific topic .
The bibliometric analysis was carried out with the bibliometrix package .
in R using the biblioshiny graphical interface to facilitate interaction with the user and the visualization of the analyzed data.
The PRISMA methodology in Figure 1 was applied to have a logic for analyzing the documents, which would allow applying inclusion and exclusion criteria .
, .
Figure 1.
PRISMA applied methodology Data collection The reference base consulted was Scopus, in the time window defined for the analysis was from the year 2000 to 2023.
The search equation applied was ( "Land cover types" OR "Land change" OR "Land cover dynamics" OR "Land cover" OR "Land use" ) AND ( "remotely sensing" OR "maps" OR "mapping" OR "classification" OR "remote sensing" OR "GIS" OR "Multiespectral" OR "imagery" OR "satellite images" ) AND ( "cacao" OR "cocoa" OR "theobroma cacao" OR "cocoa plantations" OR "cocoa agroforetrsy" OR "cocoas" ).
With this equation, it was possible to detect 105 records in SCOPUS, to which the following exclusion criteria were applied: .
records in a language other than English, i.
records outside the time window, and .
records that when reviewed were not related to the main topic.
Once the exclusion criteria were applied, 25 documents were obtained to carry out the thematic review.
RESULTS AND DISCUSSION
Geography distribution of authors Figure 2 shows that countries such as Ghana.
Germany, and Brazil have the largest number of The number of authors per country ranges between 8 and 26, with Ghana .
having the largest number of authors, followed by Germany .
and Brazil .
The countries mentioned have a strong connection with cocoa cultivation.
Ghana, located on the African continent, is one of the countries with the largest cocoa production worldwide .
while Germany is one of the main importing and exporting countries of cocoa .
Although Germany is not a producing country, it is familiar with all cocoa manufacturing, adding value through its processing for the production of chocolates and other cocoa-derived products that are exported worldwide.
For its part.
Brazil is considered the main producer of cocoa in Latin America, being one of the main ones worldwide .
However, it should not be overlooked in this scenario that countries such as the Netherlands and Cameroon, followed by China and Mexico, play a crucial role with 7 and 6 authors respectively, which stand out worldwide for their participation in the cocoa market.
China is one of the Asian countries with the highest cocoa production and manufacturing, and the Netherlands and Mexico are some of the countries with the highest participation in cocoa manufacturing issues worldwide.
Remote sensing applied to cocoa crop identification, a thematic review (Luisa Fernanda Cuellar-Escoba.
A ISSN: 2088-8708 Figure 1.
Authors by country Publication by year Analyzing the annual variation of publications, see Figure 3, it was possible to show that up to 2017, 20% of total productivity was generated, and in 2018 alone, 16% of productivity was recorded, that is, only 4% less than what was achieved in 12 years.
Since 2017, an upward trend has been observed in scientific research developed on the subject.
By delving a little deeper into the subject, some facts were identified that could motivate such behavior.
In 2017, important initiatives began to be generated that motivated the production of sustainable cocoa during the negative effects of climate change and premium chocolate trends driven by the market.
By 2018, the commitment of cocoa-producing countries to combat deforestation and the creation of international agreements such as the Cocoa & Forests Initiative gave way to the adoption of practices that encouraged more sustainable cocoa production.
And at the commercial level, the tension
between the United States and China impacted the markets for basic products, including cocoa, affecting its Cumulative Articles by year
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023
Years of publication Articles Cumulative Figure 3.
Temporal distribution of articles between 2006 and 2023 Most relevant articles According to the results obtained, the 10 most cited articles can be seen in Table 1.
These showed that the main topics addressed in the publications were mainly related to the problem of deforestation, land use and land cover, as well as the countries associated with this problem.
Ghana and Cameroon.
Int J Elec & Comp Eng.
Vol.
No.
October 2025: 4848-4855
Int J Elec & Comp Eng
ISSN: 2088-8708
Table 1.
The most relevant articles Title Cocoa crops are destroying the forest reserves of the classified forest of Haut-Sassandra (Ivory Coas.
Regional land cover mapping in the humid tropics using combined optical and SAR satellite dataAia case study from Central Sulawesi.
Indonesia Delineation of Cocoa Agroforests using Multiseason Sentinel-1 SAR images: a low grey level range reduces uncertainties in GLCM texture-based mapping Effects of tree-crop farming on land-cover transitions in a mosaic landscape in the Eastern Region of Ghana Detecting cocoa plantations in Cyte dAoIvoire and Ghana and their implications on protected areas Change vector analysis to categorize land cover change processes using the tasseled cap as biophysical indicator An assessment of land-use changes in the Cocoa Belt of south-west Nigeria Investigating relationships between Landsat-7 ETM data and spatial segregation of LULC types under shifting agriculture in southern Cameroon Spatial and temporal dynamic of land-cover/land-use and carbon stocks in Eastern Cameroon: a case study of the teaching and research forest of the University of Dschang Pixel-based and object-oriented approaches in segregating cocoa from forest in the Juabeso-Bia landscape of Ghana Year Journal Global Ecology and Conservation Total Citations International Journal of Remote Sensing International Journal of Geo - Information Environmental Management Ecological Indicators Environmental Monitoring and Assessment International Journal of Remote Sensing International Journal of Applied Earth Observation and Geoinformation Forest Science and Technology Remote Sensing Applications: Society and Environment Thematic review Table 2 .
n Appendi.
presents the main findings of the thematic review associated with each of the 10 most relevant articles.
From Table 2, it can be concluded that four of the 10 most cited articles carried out their study using optical and radar satellite images, of which the study carried out by .
obtained an accuracy of 89%, being the highest of the group of articles analyzed.
Regarding the precision algorithms used, the majority was the maximum likelihood algorithm with quite good accuracies, all above 80%.
Three of the four studies that used optical and radar images as inputs used the random forest (RF) algorithm with which the highest precision mentioned above was obtained, however, the remaining study used the k-nearest neighbor .
NN) algorithm with which it obtained an accuracy of 69%, this being the lowest of the 10 articles On the other hand, the main motivation for the development of the studies was to identify the dynamics of land use in forest areas since cocoa cultivation is associated with deforestation processes.
CONCLUSION
Although there are different methods to identify cocoa crops, the studies that used the combined use of optical and radar images with the Random Forest algorithm were the ones that obtained the highest precision .
%).
On the other hand, although the cocoa publications in the range of years analyzed are relatively low .
, a growing trend is evident from 2017 onwards, which is likely to be driven by the new regulations and guidelines on environmental conservation that will require the implementation of advanced technologies for the efficient monitoring of these dynamics.
In terms of publications, although a fairly diverse geographical distribution was observed, it is observed how each of these countries plays a crucial role in the cocoa production chain, either as producing countries or in cocoa manufacturing processes, a behavior that is directly related to the degree of participation of each country in the said chain.
ACKNOWLEDGEMENTS
The authors would like to thank the Catholic University of Manizales and the master's program in remote sensing.
REFERENCES
-B.
Mejya Castro Diana Carolina - Almansa Varela.
AuEquity in the cocoa chain in the department of cundinamarca (In Spai.
,Ay Universidad de los Andes, 2020.
Kalischek et al.
AuCocoa plantations are associated with deforestation in Cyte dAoIvoire and Ghana,Ay Nature Food, vol.
4, no.
384Ae393.
May 2023, doi: 10.
1038/s43016-023-00751-8.
Masolele et al.
AuMapping the diversity of land uses following deforestation across Africa,Ay Scientific Reports, vol.
14, no.
1, p.
Jan.
2024, doi: 10.
1038/s41598-024-52138-9.
Parlamentos Europeo y del Consejo.
AuEuropean Parliament and Council.
EU 2023/1115.
European Union, 2023 (In Spanis.
,Ay 2023.
Remote sensing applied to cocoa crop identification, a thematic review (Luisa Fernanda Cuellar-Escoba.
ISSN: 2088-8708
Europeo.
AuEuropean green deal: key to a climate-neutral and sustainable EU | News | European Parliament (In Spanis.
,Ay Fernandes et al.
AuHidden costs of EuropeAos deforestation policy,Ay Science, vol.
379, no.
6630, pp.
341Ae342, 2023.
Siwe and B.
Koch.
AuChange vector analysis to categorise land cover change processes using the tasselled cap as biophysical indicator,Ay Environmental Monitoring and Assessment, vol.
145, no.
1Ae3, pp.
227Ae235.
Oct.
2008, doi: 10.
1007/s10661-007-0031-6.
Somarriba.
AuCocoa in Agroforestry Systems (In Spanis.
,Ay Agroforesterya en las Amyricas, vol.
6, no.
22, 1999.
Dash.
Sanders.
Parajuli, and Y.
Ouyang.
AuImproving the accuracy of land use and land cover classification of landsat data in an agricultural watershed,Ay Remote Sensing, vol.
15, no.
16, p.
Aug.
2023, doi: 10.
3390/rs15164020.
Garcys-Gymez and V.
Henao-Cyspedes.
AuInternational Journal of Electrical and Computer Engineering: a bibliometric analysis,Ay International Journal of Electrical and Computer Engineering (IJECE), vol.
12, no.
6, p.
Dec.
doi: 10.
11591/ijece.
Aria and C.
Cuccurullo.
AuBibliometrix : an R-tool for comprehensive science mapping analysis,Ay Journal of Informetrics, 11, no.
4, pp.
959Ae975.
Nov.
2017, doi: 10.
1016/j.
Page et al.
AuThe PRISMA 2020 statement: an updated guideline for reporting systematic reviews,Ay bmj, vol.
Page et al.
AuPRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews,Ay BMJ, p.
Mar.
2021, doi: 10.
1136/bmj.
Suh and E.
Molua.
AuCocoa production under climate variability and farm management challenges: some farmersAo perspective,Ay Journal of Agriculture and Food Research, vol.
8, p.
Jun.
2022, doi: 10.
1016/j.
y T.
and Procolombia.
AuCocoa and cocoa derivatives sector in the world,Ay 2018.
Corponor.
AuCurrent state of the pyramo in the northeastern region of the Santurbyn biogeographic unit - Norte de Santander Department, municipalities of Villacaro.
Chitaga.
Cachira.
Salazar.
Arboledas.
Cucutilla.
Pamplona.
Mutiscua.
Cycota, and Silos.
Chapter I.
Pre,Ay 2009.
Ashiagbor et al.
AuPixel-based and object-oriented approaches in segregating cocoa from forest in the Juabeso-Bia landscape of Ghana,Ay Remote Sensing Applications: Society and Environment, vol.
19, p.
Aug.
2020, doi:
1016/j.
Akinyemi.
AuAn assessment of land-use change in the Cocoa Belt of south-west Nigeria,Ay International Journal of Remote Sensing, vol.
34, no.
8, pp.
2858Ae2875.
Apr.
2013, doi: 10.
1080/01431161.
Asubonteng.
Pfeffer.
Ros-Tonen.
Verbesselt, and I.
Baud.
AuEffects of tree-crop farming on land-cover transitions in a mosaic landscape in the Eastern Region of Ghana,Ay Environmental Management, vol.
62, no.
3, pp.
529Ae547.
Sep.
2018, doi:
1007/s00267-018-1060-3.
-O.
Abu.
Szantoi.
Brink.
Robuchon, and M.
Thiel.
AuDetecting cocoa plantations in Cyte dAoIvoire and Ghana and their implications on protected areas,Ay Ecological Indicators, vol.
129, p.
Oct.
2021, doi: 10.
1016/j.
Yemefack.
Bijker, and S.
De Jong.
AuInvestigating relationships between Landsat-7 ETM data and spatial segregation of LULC types under shifting agriculture in southern Cameroon,Ay International Journal of Applied Earth Observation and Geoinformation, vol.
8, no.
2, pp.
96Ae112.
Jun.
2006, doi: 10.
1016/j.
Barima et al.
AuCocoa crops are destroying the forest reserves of the classified forest of Haut-Sassandra (Ivory Coas.
,Ay Global Ecology and Conservation, vol.
8, pp.
85Ae98.
Oct.
2016, doi: 10.
1016/j.
Temgoua.
Momo Solefack.
Nguimdo Voufo.
Tagne Belibi, and A.
Tanougong.
AuSpatial and temporal dynamic of land-cover/land-use and carbon stocks in Eastern Cameroon: a case study of the teaching and research forest of the University of Dschang,Ay Forest Science and Technology, vol.
14, no.
4, pp.
181Ae191.
Oct.
2018, doi: 10.
1080/21580103.
Numbisi.
Van Coillie, and R.
De Wulf.
AuDelineation of cocoa agroforests using multiseason sentinel-1 SAR images: a low grey level range reduces uncertainties in GLCM texture-based mapping,Ay ISPRS International Journal of GeoInformation, vol.
8, no.
4, p.
Apr.
2019, doi: 10.
3390/ijgi8040179.
Erasmi and A.
Twele.
AuRegional land cover mapping in the humid tropics using combined optical and SAR satellite dataAia case study from Central Sulawesi.
Indonesia,Ay International Journal of Remote Sensing, vol.
30, no.
10, pp.
2465Ae2478.
May 2009, doi: 10.
1080/01431160802552728.
APPENDIX
Table 2 presents the main findings of the thematic review associated with each of the 10 most relevant articles .
Title Pixel-based and object-oriented approaches in segregating cocoa from forest in the Juabeso-Bia landscape of Ghana Reference .
Methodology Data from Sentinel-1 and Sentinel-2 satellites were used to generate three data sets: spectral bands, vegetation indices, and image objects derived using a multi-resolution segmentation These data sets were then classified using the Random Forest algorithm.
The accuracy of the classifications was assessed, and the McNemar test was used to compare statistical differences between the data sets.
An assessment of land-use change in the Cocoa Belt of south-west Nigeria The study uses multispectral satellite imagery to analyze land use and land cover changes in the Cocoa Belt in southwestern Nigeria over 25 Landsat TM and ETM images from 1986, 2002, and 2011 were used.
Image classification was performed using a maximum likelihood (MLC) algorithm, which identified Int J Elec & Comp Eng.
Vol.
No.
October 2025: 4848-4855 Findings The classification performed with the third dataset, which integrated spectral bands, backscatter data, and image objects, had an accuracy of 89.
76% and a kappa coefficient of This set was the most effective in differentiating agroforestry cocoa from open canopy forest, outperforming other sets that presented accuracies of 79.
02% and 80.
Image objects were identified as the most relevant predictors to increase classification accuracy.
The study reveals significant land use changes in the Cocoa Belt of Nigeria between 1986 and During this time, built-up areas increased 32% and gallery vegetation by 43.
while cultivated land and forest cover decreased 91% and 0.
The conversion of agricultural and forest land to built-up areas Int J Elec & Comp Eng
ISSN: 2088-8708
Table 2 presents the main findings of the thematic review associated with each of the 10 most relevant articles .
Title Reference Effects of Treecrop Farming on Land-cover Transitions in a Mosaic Landscape in the Eastern Region of Ghana Detecting cocoa plantations in Cyte dAoIvoire and Ghana and their implications on protected areas .
Investigating between Landsat-7 ETM data and spatial segregation of LULC types under shifting agriculture in Cameroon Cocoa crops are destroying the forest reserves of the classified forest of HautSassandra (Ivory Coas.
Methodology categories such as built-up areas, forest cover, water bodies, and cultivated land.
Field visits were conducted to validate the data and ensure the accuracy of the classifications, in addition to analyzing the spatial and temporal dynamics of land use and its main drivers.
The study analyses land cover changes in a semi-deciduous forest in Ghana using Landsat imagery from 1986 to 2015.
One hundred and fifty-eight images were downloaded, from which those from 1986 and 2015 were selected after filtering for quality.
Since there were no historical maps, information on landscape composition was obtained through Fifty-nine reference polygons were collected for 1986 and 908 for 2015, and various classification techniques were used, assessing accuracy with a change detection The study used 2019 Sentinel-1 and Sentinel-2 imagery in Google Earth Engine (GEE) to analyze land cover in Ghana.
A total of 158 filtered images were downloaded to ensure quality and reduce cloud interference.
19,196
reference points were collected to validate the results, including 3,842 cocoa points collected in the field.
The analysis included the application of pre-processing techniques to improve image quality, such as the calculation of the Normalized Difference Vegetation Index (NDVI) and Tasseled Cap Analysis, which considers brightness, greenness, and moisture.
A Random Forest model was used for land use classification, and its accuracy was assessed through metrics such as overall accuracy and producer accuracy.
The study was conducted in the tropical region of southern Cameroon, where shifting agriculture .
lash-and-bur.
causes forest conversion to land use diversity (LULC).
Its objective was to analyze the relationships between LULC.
Landsat-7 ETM imagery, and landscape structure.
171 crop plots representing eight LULC types were identified, and an ETM image obtained two months after field data collection was used.
A maximum likelihood classifier (MLC) was used to classify LULC types, allowing the creation of a map with an accuracy of 81% after consolidating certain land use types, such as fallows and cocoa plantations.
The landscape metrics obtained indicated a high level of patch diversity and connectivity, facilitating the simulation of predictive maps in areas with satellite imagery often covered by clouds.
The study was conducted in an area of 102,400 ha of forest surrounded by commercial crops such as coffee, cocoa, rubber, and cashews, as well as some areas of subsistence agriculture.
Five satellite images from the years 1997, 2002, 2006, 2013, and 2015 were used, which were orthorectified and classified using a maximum likelihood algorithm.
The land cover classes considered were: forest, bare soil, habitat, crop, and fallow.
The most recent image was initially classified with 54 training zones validated by field visits.
The results obtained were used to adjust the classifications of the previous images, and transition matrices were created to identify changes in the landscape over time.
Findings largely explains this increase.
95 kmA of forest was lost, largely due to the expansion of cocoa production, although 180.
50 kmA of cultivated land was recovered as forest.
The accuracy of the classifications was high, with values of 87.
33% in 1986 and 91.
33% in 2011.
The overall accuracies of the land cover maps for 1986 and 2015 were 91.
2% and 78.
respectively, although the area estimates reduced these accuracies by about 6%.
The accuracy of the 1986 map was better than that of the 2015 map.
Forest and food crop land cover decreased by 9%, with an annual rate of change of 1.
84% for forest and 1.
75% for food crop land.
On the other hand, oil palm experienced the largest increase in land cover at 11.
2%, followed by cocoa at 8.
The annual changes for oil palm and cocoa were 93% and 1.
06%, respectively.
The study generated a thematic map of cocoa plantations using a Random Forest classifier, achieving an accuracy of 82.
9% for the producer 2% for the user.
The area under cocoa cultivation is estimated at 3.
69 Mha in Cyte d'Ivoire and 2.
15 Mha in Ghana.
Approximately 18 Mha of the 5.
8 Mha of plantations were found to be within protected areas (PA.
, affecting 362 PAs, which represents about 70% in the studied region.
Cocoa encroachment is more prevalent in PAs with a management category that allows human-nature interaction, compared to areas whose priority is biodiversity conservation.
However, 40% of the PAs hosting cocoa plantations do not have information on their management category, showing a high percentage of cocoa encroachment.
The study examined Landsat ETM spectral responses to map land use and land cover (LULC) types in the rainforest of southern Cameroon.
Although the spectral bands were able to discriminate between 50% and 60% of LULC types, crop fields, and short fallows showed more than 80% accuracy.
Due to the difficulty in differentiating some plots, several LULC types were consolidated, allowing a map with 81% accuracy to be generated.
The maximum likelihood classifier (MLC) was effective, and the landscape analysis showed high diversity and connectivity, useful for predictive mapping.
The study showed an overall accuracy and kappa coefficient of over 80%.
An increase in crop and fallow area coverage was observed due to the reduction of forest areas.
The largest disturbances identified were cocoa plantations, felling and burning of trees, and trees burned at the base.
Approximately 95% of the disturbances were of human origin and 5% of natural origin.
Classification accuracy decreased in more heterogeneous environments, due to the difficulty of capturing detail when units are smaller than a The highest accuracy was achieved in more homogeneous areas before the conflicts.
Remote sensing applied to cocoa crop identification, a thematic review (Luisa Fernanda Cuellar-Escoba.
A ISSN: 2088-8708
Table 2 presents the main findings of the thematic review associated with each of the 10 most relevant articles .
Title Spatial and temporal dynamic of land-cover/landuse and carbon stocks in Eastern Cameroon: a case study of the teaching and research forest of the University of Dschang Reference .
Delineation of Cocoa Agroforests Using Multiseason Sentinel-1 SAR Images: A Low Grey Level Range Reduces Uncertainties in GLCM TextureBased Mapping .
Change vector analysis to categorise land cover change processes using the tasselled cap as .
Regional land cover mapping in the humid tropics using combined optical and SAR satellite dataAia case study from Central Sulawesi.
Indonesia Methodology Findings The study was carried out in the teaching and The study mapped land cover into four classes:
research forest of Dschang University in bare soil and built-up area, dense forest, light Cameroon to analyze changes in land cover and forest, and savanna.
Bare soil and built-up areas use, as well as carbon stocks.
Landsat satellite were grouped together due to the similar images .
TM from 1984, 7ETM from 2000 reflectance of thatched roofs and soil.
The dense and 8OLI from 2.
were used, and processed forest includes primary, secondary, and swamp with ENVI and ArcGIS software.
Interviews, forests, while the light forest is degraded, group discussions, and participatory mapping sometimes coinciding with cocoa plantations or were used to identify human activities.
A forest farms.
Savannas and croplands were also grouped inventory was carried out using 12 km long together due to the same reflectance in satellite transects to measure dendrometric parameters Between 1984 and 2016, forest cover and classify land use types.
Carbon stocks were decreased, with losses of 937 ha between 1984estimated using allometric models based on 2000 and 931 ha between 2000-2016, while non-destructive methods.
Aboveground and savanna vegetation increased.
The Kappa index belowground biomass were calculated and the and overall accuracy were above 80%, indicating data were analyzed to assess carbon stock a good representation of reality.
dynamics and forest cover changes.
The study was conducted in the Bakoa Random Forest (RF) classification models landscape in the central region of Cameroon, achieved accuracies above 70%, although with 28 kmA of bush savannah, variations in classification error and sensitivity subsistence agriculture, and cocoa agroforests.
between different land cover classes.
The most RapidEye optical imagery with 5-meter reliable models were RE1.
GLI3 and GL3.
The resolution and five spectral bands were used, combination of polarization bands (VV and VH) creating a mosaic from four images obtained in exceeded the accuracy of their individual use.
the same period.
Fifty Sentinel-1A dualHowever, the inclusion of backscatter GLCM polarized (VV and VH) C-band radar (SAR) texture measures and optical vegetation indices images were also used, which were predid not improve classification.
The RE1 model processed using SNAP to remove noise and excelled in identifying areas without vegetation, calibrate radiometry.
while SAR imagery offered reliable results for Field data on land cover and land use were distinguishing cover types such as savannas and collected between 2015 and 2017.
Several cocoa agroforests.
Although multispectral optical vegetation indices such as NDVI and SAVI imaging showed better overall performance, it were calculated using Random Forest was less effective in discriminating between algorithms in R.
For SAR images, gray-level evergreen agroforests and transitional forests.
co-occurrence matrix (GLCM) textures were analyzed in 5y5 windows, measuring features such as entropy and contrast at different The study used bi-temporal Landsat images The results showed that the majority of the from 1987 and 2002, applying the Change studied area experienced some type of change.
Vector Analysis (CVA) technique to analyze 75% gaining biomass and 21.
7% losing it.
the differences in tasseled cap trajectories The overall accuracy of the analysis was 70%.
hich measure brightness and greennes.
Areas of biomass loss and deforestation were between the two years.
The images were detected mainly in areas close to villages or along geometrically and topographically corrected.
main roads, while areas of forest regeneration The magnitudes and directions of changes in were mainly located adjacent to these deforested forest cover were calculated from the The technique allowed the identification of differences in brightness and greenness.
areas of positive and negative vegetation change.
Thresholds were established to classify the although the lack of complete historical data changes into three intensity categories: low, limited the validation of these results.
medium, and high.
Finally, the directions of change were identified and classified into 13 different categories of forest cover modification.
The study was conducted in Central Sulawesi The study analyzed land cover classification in region.
Indonesia, using Landsat-ETM and Central Sulawesi.
Indonesia, using multitemporal Envisat-ASAR satellite data to classify land synthetic aperture radar (SAR) and Landsat use and land cover in a tropical environment.
It satellite data.
Co-polarized SAR data achieved an focused on identifying intensive agricultural accuracy of 69% with a Kappa index of 0.
systems, such as cocoa plantations and rice while cross-polarized data had a lower accuracy crops, and on differentiating cultivated from of 54% and a Kappa index of 0.
The forested areas.
A multitemporal approach was combination of multitemporal data and SAR used with synthetic aperture radar (SAR) data textures achieved an accuracy of 68% (Kappa and the effectiveness of different polarization index of 0.
Among the Landsat data, the 2004 modes .
o-polarization and cross-polarizatio.
data, with SLC-off, achieved an accuracy of 73% was evaluated to improve classification (Kappa index of 0.
, and the 2002 data showed SAR data were preprocessed to the best overall accuracy, with 85% and a Kappa correct geometric and radiometric errors, and index of 0.
The combination of Landsat and filters were applied to reduce noise.
Indices and co-polarized SAR data resulted in the highest statistics were calculated to facilitate the overall accuracy, reaching 89% (Kappa index of separation of land use classes.
These results highlight the effectiveness of Int J Elec & Comp Eng.
Vol.
No.
October 2025: 4848-4855
Int J Elec & Comp Eng
ISSN: 2088-8708
Table 2 presents the main findings of the thematic review associated with each of the 10 most relevant articles .
Title Reference