Objective This study aims to elucidate the response mechanisms of vegetation productivity to drought events across multiple time scales in arid and semi-arid regions, reveal their spatiotemporal patterns, and provide a scientific basis for regional carbon cycle regulation and climate adaptation management.

Method Taking the Helan Mountains of northwestern China as the study area, we applied the Denitrification-Decomposition (DNDC) and Forest-DNDC models to simulate the spatiotemporal dynamics of gross primary productivity (GPP) from 2001 to 2021. Drought conditions were characterized using the Standardized Precipitation Evapotranspiration Index (SPEI) at four time scales: SPEI-1 (monthly), SPEI-6 (semi-annual), SPEI-12 (annual), and SPEI-24 (multi-year). Linear correlation analysis and generalized additive models (GAMs) were used to quantitatively evaluate the effects of drought characteristics (duration, severity, intensity, and frequency) on GPP.

Result (1) During 2001−2021, GPP in the Helan Mountains showed a significant increasing trend, with an average annual growth rate of 0.20 g/(m²·year) (expressed as carbon), and showed a spatial gradient decreasing from the center to the surrounding areas; (2) GPP responses to drought were strongly time-scale-dependent and nonlinear. At the short-term timescale (SPEI-1), GPP was negatively correlated with drought duration, severity, and frequency, while weak drought intensity showed a slight positive effect, suggesting compensatory growth responses; (3) As the time scale extended (SPEI-6, SPEI-12, and SPEI-24), the inhibitory effect of drought on GPP progressively intensified, showing nonlinear threshold behavior; (4) Among ecosystems, grasslands were the most sensitive to medium- and long-term droughts, with GPP reductions significantly greater than those in forests and shrublands.

Conclusion GPP responses to drought stress in the Helan Mountains exhibit pronounced time-scale dependence and nonlinear behavior, with distinct sensitivity differences among ecosystem types. This study highlights the need to integrate drought characteristics and vegetation sensitivity into multi-scale drought risk assessment and ecosystem management strategies, providing a scientific basis for enhancing carbon sequestration capacity in arid regions.