Grazing amplifies grassland productivity sensitivity to climate variability through altered water regulation in drylands

Dryland grasslands are increasingly exposed to the combined pressures of climate warming, drying and intensified grazing, yet how grazing alters ecosystem sensitivity to climate variability remains poorly understood. In particular, the ecohydrological mechanisms through which grazing influences vegetation productivity under climate change are still unclear. Here, we investigate how grazing modifies the climate sensitivity of net primary productivity (NPP) in a temperate dryland grassland ecosystem.

We focused on a grassland watershed in Inner Mongolia, China, and combined process-based modelling with interpretable machine learning to analyze long-term NPP responses to climate variability under contrasting grazing intensities. Using an HRU-based DNDC model, we reconstructed multi-decadal NPP dynamics under grazing and no-grazing scenarios. Random forest models and SHAP analysis were then applied to quantify changes in the relative importance of precipitation characteristics, soil moisture and temperature.

Our results show that total precipitation, precipitation frequency and early to mid-growing season soil moisture (June–July) are the dominant hydrometeorological controls on NPP in this dryland system. However, grazing substantially amplifies NPP sensitivity to climate variability by weakening ecosystem water regulation capacity. This amplification effect differs among vegetation types. Low-coverage grasslands exhibit strong sensitivity to water availability, while medium-coverage grasslands maintain higher regulatory capacity. In contrast, high-coverage grasslands are most vulnerable to the combined impacts of grazing and climate stress.

Threshold analysis reveals vegetation-dependent grazing intensities beyond which ecosystem stability declines. High-coverage grasslands show reduced stability at relatively low grazing intensity (~0.25 beef ha⁻¹), whereas medium- and low-coverage grasslands tolerate higher intensities (~0.3 beef ha⁻¹). Historical NPP trajectories further indicate nonlinear responses, including lag effects and metastable states, suggesting that shifts in grazing pressure can trigger abrupt changes in ecosystem functioning.

Our findings demonstrate that grazing acts as a key amplifier of climate sensitivity in dryland grasslands by altering ecohydrological controls on productivity. These results highlight the importance of incorporating vegetation type and water availability into adaptive grazing management strategies under increasing climate variability.