Assessing the impact of plateau pikas on carbon sources and Sinks: a case study of the Qinghai section of the Yellow River Basin

• 2001-2020 net carbon uptake trends mapped for Qinghai section of Yellow River Basin. • Plateau pika activity mapped and habitat shifts tracked over twenty years. • Moving-center analysis reveals long-term changes in suitable pika areas. • Joint impacts of pikas, weather, and land traits on carbon uptake quantified. • Local drivers of carbon uptake vary, showing region-specific pika effects. The role of the plateau pika ( Ochotona curzoniae ) as a keystone species in the ecosystems of the Tibetan Plateau and the Upper Yellow River remains highly debated. In particular, its influence on regional carbon sources and sinks is still unclear. Balancing the environmental impacts of pika activity with its effect on carbon dynamics requires further investigation. This study employs the Maxent model, the Carnegie-Ames-Stanford Approach (CASA) model, and empirical formulas to analyze the spatiotemporal patterns of pika activity intensity and Net Ecosystem Production (NEP) from 2001 to 2020. Additionally, the Partial Least Squares Structural Equation Modeling (PLS-SEM) model is used to evaluate the effects of environmental variables and pikas activity on NEP. To further explore the contributions of individual variables and their interactions, we introduce the Geographically Weighted Random Forest (GWRF) model coupled with SHapley Additive exPlanations (SHAP) analysis. The results indicate that in recent years, the suitable habitat for pika has expanded, with a discernible shift in distribution and centroid from the north to the south. Meanwhile, the annual average NEP exhibits a fluctuating upward trend, increasing at a rate of approximately 1.4 g·C/m 2 ·a. Between 2006 and 2020, the yearly NEP consistently exceeded the long-term average of 268.936 g·C/m 2 ·a. Among the environmental variables, grazing intensity and climatic factors—particularly precipitation—exert the most substantial effects on NEP. Although the direct impact of pikas on NEP is relatively minor (path coefficient = 0.07, P < 0.001), they indirectly influence NEP by affecting surface soil moisture and annual vegetation fractional cover (FVC). Across much of the study region, NEP dynamics are primarily driven by grazing intensity and climatic variables, land surface temperature (LST) and precipitation dominate the changes in NEP in most of the study area. The spatial distribution of the SHAP values reveals notable heterogeneity in the effect of pika activity, with the most pronounced impacts observed from Xinghai County to Dari County. Pikas activity shows a positive correlation with key variables such as grazing intensity, LST, and precipitation. Interaction analysis indicates that combined effects of overgrazing, high-intensity pika activity, and extreme climatic conditions lead to NEP decline. However, pika activity can also alleviate the adverse impacts of extreme rainfall on NEP.