Central Asian Research Index

3i intellect idea innovation - интеллект идея инновация, 2026, т. 3, № 1, с. 193-204

MAPPING GROUNDWATER POTENTIAL USING SATELLITE DATA AND A MACHINE LEARNING ALGORITHM: A CASE STUDY OF THE ILI DISTRICT OF ALMATY REGION, SOUTH KAZAKHSTAN

КАРТИРОВАНИЕ ПОТЕНЦИАЛА ПОДЗЕМНЫХ ВОД С ПРИМЕНЕНИЕМ СПУТНИКОВЫХ ДАННЫХ И АЛГОРИТМА МАШИННОГО ОБУЧЕНИЯ НА ПРИМЕРЕ ИЛИЙСКОГО РАЙОНА АЛМАТИНСКОЙ ОБЛАСТИ ЮЖНОГО КАЗАХСТАНА

Zhuldyzbek Onlassynov, Malis Absametov, Yermek Murtazin, Mira Muratova

Akhmedsafin Institute of Hydrogeology and Environmental Geosciense, Satbayev University

Аннотация · Abstract

This study utilizes an integrated approach based on the combined use of remote sensing data and machine learning methods to predict the spatial distribution of potential groundwater zones. Random Forest, a supervised learning method that demonstrates high robustness to input data noise, was selected as the primary classification algorithm. The overall study design includes four main stages: generating input spatial variables; preparing a training set; training and tuning the Random Forest model; and validating and assessing the accuracy of the results. The input variables included spectral indices (NDVI, NDWI); morphometric parameters (Slope, DEM); and hydrological factor (Distance to Rivers). The training data included water intake facility locations in the Ili district of the Almaty region. To model groundwater potential, the Random Forest algorithm was implemented in Python using the Scikit-learn library. Satellite and geomorphometric data were used as predictors and processed in ArcGIS software. All variables were converted to a common spatial resolution and coordinate system. To ensure data comparability, values ​​were normalized, reducing the impact of differences in feature ranges on the model training process. The trained model was applied to the entire study area using raster overlay, resulting in a spatially continuous groundwater potential map classified into low, medium, and high potential zones.

Сведения · Details