Analysis of the spatiotemporal dynamics of grassland carbon sinks in Xinjiang via the improved CASA model

• Optimize the CASA model improve the accuracy of simulating net primary productivity. • Interannual and seasonal changes of grassland net ecological productivity increased. • Most areas of the grassland have changed from carbon sources to carbon sinks. • Humidity, soil pH, and livestock grazing had main effect on grassland carbon storage. Accurate assessment of grassland carbon sinks provides scientific data for quantifying the global carbon balance. Currently, methods for precisely calculating NPP (net primary productivity) and NEP (net ecosystem productivity) in arid grasslands are limited, and the changes in NEP and their driving mechanisms remain unclear. This study optimized LUE max (maximum light use efficiency) and W ε (water stress coefficient) in the CASA (Carnegie–Ames–Stanford Approach) model and developed eight improved CASA model configurations. NEP was calculated as the difference between the NPP estimated by the optimal CASA model and soil heterotrophic respiration (R h ). The changes in grassland NPP and NEP from 2000 to 2022 and the effects of different factors on grassland NEP were explored. The analysis revealed that (1) the R 2 value of the optimal CASA model was 0.70, and the RMSE was 22.30 gC·m −2 . (2) The 23-year average NEP in Xinjiang grassland was 113.00 gC·m −2 ·a −1 , with 82.20 % carbon sink areas and 17.80 % carbon source areas. (3) NEPs of meadows, steppes, and desert grasslands in the Xinjiang region tended to increase during spring, summer, and autumn. (4) During the 23-year period, 15.77 % of the Xinjiang grassland area transitioned from carbon sources to carbon sinks. (5) Relative humidity and soil pH strongly influenced the NEP of Xinjiang grassland. This study provides a more accurate model for calculating NPP and estimating the variation in carbon source and carbon sink areas in arid grasslands, providing theoretical guidance for the development of grassland management policies.