The rapid expansion of open-pit mining in Northwest China, driven by growing demands for energy and strategic minerals, poses significant challenges to environmental sustainability in arid and semiarid ecosystems. Leveraging high-resolution remote sensing data (e.g., Gaofen-2/6, ZY-3), this study systematically analyzed the spatiotemporal patterns of 12,680 mining polygons across 3,253 approved open-pit mines in four northwestern provinces (Xinjiang, Qinghai, Gansu, and Ningxia) from 2021 to 2022. Results revealed a pronounced spatial clustering of mining activities, with 92% of the total mining area concentrated in Xinjiang and Qinghai. Non-metallic mineral extraction (potash, lithium, coal) dominated, accounting for 71.8% of the total footprint, while metallic mining showed localized impacts. Global Moran’s I (0.156, p<0.01) and Getis-Ord Gi* analyses identified two high-intensity clusters in potassium-rich basins, demonstrating significant spatial autocorrelation. Desertification sensitivity assessments in key coal mining zones revealed that 55% of mining areas fell within high-sensitivity regions, correlating with post-2016 NDVI declines. Despite progress in avoiding protected areas (only 0.45% overlap), ecological trade-offs persisted-mining contributed 31% to Qinghai’s GDP but degraded 12% of grasslands annually. Methodologically, manual interpretation achieved 95% accuracy but faced scalability limitations compared to automated machine learning approaches. The findings underscore the necessity of environmental, landscape-scale governance frameworks to address spillover effects between protected and high-density mining zones. We advocate integrating AI-enhanced monitoring and blockchain traceability to strengthen green mining practices, emphasizing that sustainable resource development in fragile ecosystems requires balancing economic priorities with spatially explicit ecological safeguards. This work provides a replicable model for reconciling mineral extraction with environmental conservation in global arid regions.