Investigating coupling relationships between ecological spatial network topological structures and desertification processes across the Mongolian Plateau

• Evaluating desertification across the Mongolian Plateau through feature space analysis. • Integrating pixel-level patches into ecological spatial networks. • Identifying spatiotemporal patterns in the topological structure of ecological spatial networks. • Investigating topological structural responses to desertification progression and regression processes. Global climate change and human activities have intensified desertification, fragmenting habitats and degrading ecosystem functions. Although the Mongolian Plateau is a global hotspot for desertification, the response mechanisms between vegetation spatial patterns and desertification processes remain unclear. This study assessed desertification trends from 2001 to 2023 using the MSAVI-Albedo feature space and Sen’s trend analysis. We constructed ecological spatial networks using the Minimum Cumulative Resistance (MCR) model and complex network theory to identify the spatiotemporal evolution of topological indicators. Finally, Pearson correlation coefficients were used to quantify the relationship between desertification and network topology. Results revealed that between 2001 and 2023, desertification on the Mongolian Plateau demonstrated an overall “reversal” phenomenon, with improvement occurring in 70 % of the region, predominantly manifested as non-significant improvement (54.13 %), and the desertification process exhibited evident nonlinear dynamic characteristics. The numbers of patches and corridors in the ecological spatial network experienced a process of initial sharp increase followed by gradual decline, primarily reflected in small-area patches and high-weight corridors. The ecological spatial network structure spatially presented a “core-periphery” pattern configuration, with relatively balanced hotspot and coldspot areas for degree centrality, eigenvector centrality, PageRank, and betweenness centrality; however, numerically dominated by small “coldspot” patches, whereas the hotspot and coldspot areas for closeness centrality and clustering coefficient exhibited hotspot dominance (79.79 %, with numerical coldspots) and coldspot dominance (70.72 %), respectively. Degree centrality, PageRank, and eigenvector centrality showed strong positive correlations with desertification (r > 0.7, p < 0.01), while closeness centrality and clustering coefficient showed no significant correlations. Unlike traditional studies, we integrated pixel-level patches to reveal how topological indicators respond differentially to desertification, providing scientific evidence for early warning and targeted conservation in arid regions.