ГИДРОЛОГИЯЛЫҚ ҮДЕРІСТЕР МЕН СПЕКТРАЛДЫҚ ДЕРЕКТЕРДІ ТАЛДАУ НЕГІЗІНДЕ ТОПЫРАҚТЫҢ ТҰЗДАНУ ДЕҢГЕЙІН БОЛЖАУДЫҢ ГИБРИДТІ МОДЕЛІН ӘЗІРЛЕУ

Abstract

Soil salinity is one of the most critical environmental and agricultural issues, especially in the arid and semi-arid regions of Kazakhstan. It leads to reduced soil fertility, land degradation, and poses a threat to food security. The aim of this study is to develop a hybrid model for predicting soil salinity levels by integrating spectral remote sensing data (Sentinel-2) and hydrological parameters from ERA5-Land. Principal Component Analysis (PCA), K-Means clustering, and the XGBoost algorithm were applied to identify informative features and improve prediction accuracy. Spatial data were processed in the Google Earth Engine environment, where spectral and hydrological parameters for two contrasting regions of Kazakhstan — Akmola region and the dried Aral Sea area — were combined. The results showed that the hybrid model reduced the mean squared error (MSE) by 26% compared to the baseline XGBoost model and successfully classified up to 98% of saline soils. The study demonstrates the high potential of an integrated approach combining remote sensing and hydrological modeling for land degradation monitoring and assessment.