LAND COVER AND CROP IDENTIFICATION USING MACHINE LEARNING ALGORITHMS
Abstract
Agriculture in Gabon is essentially composed of smallholder farms scattered across vast rural territories. It plays a key role in the country’s food security. However, the isolation of these areas makes their monitoring and characterization difficult. This study proposes a methodology for mapping agricultural areas and main crops in the commune of Lébamba and its surroundings. The approach integrates GPS surveys conducted from July to November 2024 across 51 plots (246 ground points), combined with drone images and Sentinel-2 satellite imagery. These data were tested through three experimental scenarios: Sentinel-2 and GPS data, Sentinel-2 and drone data, and Sentinel-2 multi-source fusion. Classifications were performed using machine learning algorithms (Random Forests, XGBoost, and Artificial Neural Networks). The results demonstrate that the use of drone imagery, coupled with the Random Forest algorithm, offers the best performance with an overall accuracy of 99.88% (kappa of 0.9964) for the identification of agricultural areas, and 94.74% for fine crop discrimination (kappa of 0.8941). This work provides a reproducible methodology, the scope of which for public agricultural policy planning in Gabon remains to be tested.
Related
- A geospatial and community perception analysis of the impacts of land use and land cover change on ecosystem service values
- Quantification of soil loss by using RUSLE model under different land use and land cover conditions in AKANYARU catchment in Rwanda
- Evaluating the performance and reliability of CA-Markov and CA-ANN models for land use/land cover change simulation in Southwestern Nigeria
- Spatiotemporal dynamics of ecosystem service values under land use and land cover change in selected districts Hadiya zone, Central Ethiopian Region
- Machine-learning projections of groundwater storage under climate, land-use, and human-driven changes in Madhesh Province, Nepal
BibTeX
@article{mbindzoukou2026land,
title = {{LAND COVER AND CROP IDENTIFICATION USING MACHINE LEARNING ALGORITHMS}},
author = {Marius Massala Mbindzoukou and Bruno Nkoumakali and Luckin Obiang Obiang},
journal = {Perspectives Plurielles},
year = {2026},
month = sep,
url = {https://hal.science/hal-05755207v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - LAND COVER AND CROP IDENTIFICATION USING MACHINE LEARNING ALGORITHMS
AU - Marius Massala Mbindzoukou
AU - Bruno Nkoumakali
AU - Luckin Obiang Obiang
PY - 2026
DA - 2026/09/18
JO - Perspectives Plurielles
UR - https://hal.science/hal-05755207v1
AB - Agriculture in Gabon is essentially composed of smallholder farms scattered across vast rural territories. It plays a key role in the country’s food security. However, the isolation of these areas makes their monitoring and characterization difficult. This study proposes a methodology for mapping agricultural areas and main crops in the commune of Lébamba and its surroundings. The approach integrates GPS surveys conducted from July to November 2024 across 51 plots (246 ground points), combined with drone images and Sentinel-2 satellite imagery. These data were tested through three experimental scenarios: Sentinel-2 and GPS data, Sentinel-2 and drone data, and Sentinel-2 multi-source fusion. Classifications were performed using machine learning algorithms (Random Forests, XGBoost, and Artificial Neural Networks). The results demonstrate that the use of drone imagery, coupled with the Random Forest algorithm, offers the best performance with an overall accuracy of 99.88% (kappa of 0.9964) for the identification of agricultural areas, and 94.74% for fine crop discrimination (kappa of 0.8941). This work provides a reproducible methodology, the scope of which for public agricultural policy planning in Gabon remains to be tested.
SN - 3105-8485
SN - 3105-8493
ER -