Land Use and Land Cover Classification in Google Earth Engine Using Sentinel-2 Based Random Forest: A Case Study of Katsina State, Nigeria

DOI

10.2991/978-94-6239-713-2_31How to use a DOI?

Keywords

Land Use and Land Cover (LULC); Random Forest; Sentinel-2; Google Earth Engine (GEE); Remote Sensing Classification; Semi-Arid Environment

Abstract

Considering the ongoing rapid land transformation across semi-arid sub-Saharan Africa, accurate and up-to-date information on land use and land cover (LULC) is increasingly important for environmental monitoring, agricultural planning, and sustainable land management. No small part of this challenge is captured in Katsina State, northwestern Nigeria, where exacerbating anthropogenic pressures continue to induce significant landscape change but remain largely undocumented in the literature. The study utilized the random forest (RF) machine learning algorithm to map LULC on Katsina State using Sentinel-2 Level-2A surface reflectance imagery collected over the October 2025 late wet-to-dry seasonal transition, processed within the Google Earth Engine (GEE) cloud computing environment. Classification inputs included Sentinel-2 multispectral bands, covering the visible, red-edge, near-infrared, and shortwave-infrared spectral regions. The RF classifier, composed of 100 decision trees, was trained using 80% of the stratified sample points, with the remaining 20% set aside for independent validation. A total of six classes were identified in the LULC data: cropland, vegetation, water, barren, arroyos, and buildup. These led to a confusion matrix-based metric of Overall Accuracy (94.35%) and standard deviation (0.02851) for classification performance; Vegetation, Cropland, and Water were classified with comparatively higher producer accuracy (100%, 99.36% and 98.45%) while buildup, barren, and arroyos exhibited lower accuracies due to inter-class spectral overlapping under dry season surface conditions. The overall accuracy of 94.35% achieved here is well above that reported for semi-arid LULC classification by single-date images (78-92%), suggesting a better performance of the proposed approach. The results validate that Sentinel-2 imagery and RF classification are a strong and repeatable technique within cloud-based geospatial frameworks, where scalability is limited only by the computing power available in GEE.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Ismail Dauda Abubakar
AU  - Narayan Vyas
PY  - 2026
DA  - 2026/06/25
TI  - Land Use and Land Cover Classification in Google Earth Engine Using Sentinel-2 Based Random Forest: A Case Study of Katsina State, Nigeria
BT  - Proceedings of the International Conference on Advances in Computing Technology and Artificial Intelligence (COMPUTATIA 2026)
PB  - Atlantis Press
SP  - 406
EP  - 418
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6239-713-2_31
DO  - 10.2991/978-94-6239-713-2_31
ID  - Abubakar2026
ER  -