Remote sensing-based hybrid modeling of sediment sampling suitability for sustainable geochronological reconstruction in the Karakum and Kyzylkum sandy deserts, Central Asia

Identifying optimal sediment sampling sites is crucial for accurate geochronological analysis and paleoclimatic reconstruction of arid regions. However, systematic site-suitability mapping in extensive desert environments remains limited. This study introduces a novel large-scale hybrid modeling framework that integrates multi-criteria decision-making (AHP and Entropy-TOPSIS) and data-driven ensemble machine learning models (Random Forest and Extreme Gradient Boosting) to optimize the suitability of sediment sampling sites for optically stimulated luminescence (OSL) dating in the Karakum and Kyzylkum Deserts. Nine key indicator parameters, including environmental, climatic, and geomorphic contexts, were employed to develop suitability layers. Accessibility factors were also considered for practical field sampling. The findings revealed that the landscape was categorized into five suitability levels: very low (27.4%), low (18.9%), moderate (23.0%), high (8.7%), and very high (21.9%). The uncertainty analysis identified 7.84% of the study area as very high confidence suitability zones, mainly in the northern Karakum, while 92.16% remained low confidence due to incomplete model agreement. The SHAP results revealed that mineral quartz and thermal exposure significantly influenced suitability predictions. This study is the first to provide a high-resolution sediment sampling suitability model for Central Asia. The employed integrated approach exhibits high predictive performance and spatial credibility, providing a transferable framework for prioritizing sites within comparable geomorphic settings. The findings serve as a foundation for targeted field campaigns, thereby enhancing the efficiency and accuracy of sediment-based paleoenvironmental research in desert environments, while also contributing to broader regional efforts aimed at climate resilience and Sustainability.