An Environmental Support System for Predicting the Normalized Difference Vegetation Index (NDVI) in the Green Mountain Region, Libya
DOI:
https://doi.org/10.5281/56fc3n87Keywords:
desertification, decision support, NDVI index, remote sensing, XGBoost, explainable artificial intelligenceAbstract
This study presents an intelligent environmental decision support system for monitoring vegetation cover and desertification indicators in Libya's Green Mountain region, by integrating remote sensing, GIS and machine learning techniques into a single platform. The study relied on seasonal data for the period 2000-2025 extracted from Landsat 5, Landsat 7, Landsat 8 and Sentinel-2 images, as well as climate data from ERA5-Land via the Google Earth Engine platform.
The XGBoost Regression algorithm was used to build a model that predicts the NDVI_mean value based on temperature, precipitation, humidity, dew point, and statistical NDVI characteristics, with performance assessed using MAE, RMSE, R2, and correlation coefficient indicators. SHAP technology was employed to interpret model decisions and determine the impact of each variable on results, enhancing system transparency and reliability.
The results demonstrated the model’s ability to perform well in predicting the mean values of the NDVI index, achieving a coefficient of determination (R² = 0.6629), a root mean square error (RMSE = 0.0791), a mean absolute error (MAE = 0.0621), and a correlation coefficient of 0.8259, confirming its potential for use in environmental decision-support systems, with the possibility of further developing the system in the future by updating the data and expanding its scope of application to include other regions within Libya.
References
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Alqalam Almoner

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.


