Relationships between vegetation indices and surface reflectance: Implications for detecting and monitoring sandification in arid regions

• Combined vegetation indices and surface parameters to develop sandification monitoring model for arid wetlands. • The NDVI-albedo model achieved 88.55% accuracy in monitoring arid wetland sandification. • Regional sandification displayed a transformation-differentiation pattern with severe-grade areas declining markedly. • Soil factors served as key mediating variables by integrating climate and human influences in the sandification process. Terminal lakes in arid regions are increasingly vulnerable to sandification under water scarcity and climate stress. Taking the Qingtu Lake region in China’s Shiyang River Basin as a case study, we evaluated different vegetation indices and surface parameter combinations to identify optimal monitoring model, analyzing the spatiotemporal dynamics of sandification and primary driving factors using long-term remote sensing data (2000–2023). The NDVI–albedo combination outperformed other index–parameter combinations in the feature space models (FSMs), achieving an overall classification accuracy of 88.55 %. This superior combination’s temporal trends exhibited strong inverse relationships, with 70 % of pixels having significant negative correlations between NDVI and albedo. The model effectively captured fine-scale spatial details of sandification levels with high ground truth consistency compared to other tested models. The regional sandification patterning revealed a distinct “transformation-differentiation” dimension in 2000–2023. Temporally, sandification intensity has greatly declined, with the area of extremely severe sandification shrinking from 2282 to 377 km 2 ; spatially, sandification has occurred along a pronounced northeast–southwest gradient. Climate factors persistently imposed significant negative effects on sandification dynamics over past the two decades, whereas the direct influence of human activities showed a marked increase from 0.18 to 0.38. Soil factors functioned as key mediating variables by integrating climate and human influences, while geographical factors exhibited minimal contribution to the overall model (direct effects < 0.1). In conclusion, this study provided a reliable technical framework to better quantitatively assess wetlands’ sandification, thus bolstering essential information for developing targeted prevention and control strategies in arid regions.