Land desertification threatens ecosystem sustainability and socioeconomic development globally. We assessed global desertification sensitivity from 2005 to 2020 using the Google Earth Engine-Mediterranean Desertification and Land Use (GEE-MEDALUS) model, integrating Climate (CQI), Soil (SQI), Vegetation (VQI), and Management (MQI) Quality Indices. Drivers were quantified with the Optimal Parameters-based Geographical Detector (OPGD) at the global scale, and causality was evaluated in China using the Geographical Convergent Cross Mapping (GCCM) model. Results showed persistent spatial heterogeneity, with high to extreme sensitivity concentrated in arid, semi-arid, and sub-humid margins, particularly across North Africa, the Middle East and Central Asia. Temporally, extreme sensitivity areas remained stable (11.5%-12.0%), while high sensitivity expanded slightly (20.8% to 21.8%). OPGD identified CQI and SQI as dominant drivers (mean q = 0.74 and 0.65), surpassing VQI and MQI (mean q = 0.56 and 0.53). Specifically, the Aridity Index and Soil Organic Matter were the most influential indicators (q = 0.80 and 0.79), and their interaction amplified explanatory power (q = 0.77-0.80). GCCM confirmed causal linkages for all indicators except Depth to Bedrock. These findings highlight a distinct climate-soil co-limitation, and help prioritize interventions to enhance resilience under environmental change.