A crucial challenge in ecosystem management is the identification of trade-off effects among multiple ecosystem services (ESs), particularly in complex and ecologically vulnerable mountainous protected areas. Nevertheless, investigations into the spatio-temporal dynamics of trade-offs and synergies among ES within such protected regions remain limited. This study concentrated on the Qilian Mountain, a critical ecologically fragile protected area, utilized the Invest model to comprehensively analyze six ESs over the period 1985–2020. Employed Pearson correlation coefficient, we scrutinized trade-off and synergy relationships of these ES across varying grid scales (30 m, 1 km, and 10 km). Furthermore, GeoDetector, in tandem with multi-source data, is employed to unveil the potential drivers influencing the patterns of ES. Our findings revealed that: (1) Increasing data spatial resolution refines grid scales, yielding smoother spatial distributions of ES, and accentuates spatial disparities. (2) Synergistic relationships prevailed among protected area ES from 1985 to 2020, with correlations between carbon storage - water yield and habitat quality - water yield shifted from synergy to trade-off in 2015. (3) Distinct grid scales exert non-linear impacts on ES correlations, affecting both the magnitude and direction of these relationships. (4) Interactions among natural factors, climate, and landscape patterns are principal drivers of spatial heterogeneity in ES, with the contributions of dominant drivers exhibiting temporal variability. Remarkably, factors with stronger driving forces demonstrate greater temporal instability in their contribution rates. Our research underscores the necessity for comparative assessments of ES and their interrelations at diverse data spatial resolutions, advocating for enhanced ES management strategies.