Qilian Mountain National Park, serving as a vital ecological barrier in northwest China, urgently requires achieving a balance between conservation and management. This study aims to elucidate the spatiotemporal dynamics of landscape patterns, landscape ecological risks (LERs), and ecosystem services (ESs), and reveal their response relationships. Based on risk-service response optimization, ecological zoning is established, and adaptive strategies are proposed. The research integrates multi-source data within a "pattern-process-service-sustainability" framework, employing methods including landscape indices, the multiscale geographically weighted regression (MGWR) model, the LER-ES coupling model, and the Geodetector model. Key findings are as follows: (a) Landscape fragmentation decreased by 12.3% from 2000 to 2020, with an average annual reduction of 2.1% in LER and weakened spatial clustering intensity. An east-west gradient in LER persisted. (b) The comprehensive capacity of ESs improved, exhibiting a southeast-northwest gradient distribution. (c) The MGWR model confirmed that LER negatively impacts ecological services, with spatial nonstationarity present. (d) Ecological zones were classified into protected areas, stable areas, sensitive areas, and vulnerable areas, with obvious differences in driving factors-environmental vulnerability was primarily dominated by natural factors, while relatively stable areas were more significantly influenced by human activities. This represents the first zoning based on the risk-service relationship, breaking through traditional limitations.