Geovisual explainable AI for understanding frozen ground in Qinghai-Tibet Plateau urban region

• GeoVisXRD framework integrates XAI and causal discovery for MTFG analysis. • Recursive SHAP decomposition quantifies cascading climatic-geographic drivers. • 30.41 °N latitude, 935.70 mm precipiation and 2,662 m elevation define critical boundaries for MTFG. • Climate factors dominate over limited human impacts on MTFG on QTPUR. Urban areas are typically located on anthropogenic soils that have been heavily modified or created by human activities. However, many cities on the Qinghai-Tibet Plateau are exceptionally built directly on frozen ground. Therefore, geospatial analysis of frozen ground is indispensable for understanding the geographic environment of these cities. Existing research, however, is limited to mapping and predicting frozen ground distribution, lacking exploration of the direct and indirect driving factors that influence it. To address this gap, we propose a framework that integrates geovisualization, causal discovery, and explainable artificial intelligence (XAI). Within this framework, we perform recursive decomposition modeling of XAI values based on causal relationships to quantitatively characterize both direct and indirect drivers of frozen ground distribution. Our findings reveal that climatic-geographic factors are the primary drivers of frozen ground. The 30.41 °N latitude watershed demarcates distinct frozen ground conditions, while thresholds of 2,662 m in elevation and 935.70 mm of precipitation define the nonlinear responses of frozen ground systems to altitudinal and hydrological gradients. We also found that human-environment interactions have limited influence on frozen ground. Transportation infrastructures, such as railways and highways, induce an Euclidean-distance distribution pattern on frozen ground. Localized disturbances of frozen ground are observed within approximately 1 km of major watersheds. These findings suggest the importance of accounting for frozen ground degradation patterns in infrastructure planning of urbanization on the Qinghai-Tibet Plateau.