Dual-Model Assessment of Ecosystem Respiration Using Random Forest and Lloyd–Taylor Models in Two High-Altitude Agricultural River Basins: Spatiotemporal Dynamics

The alpine valley agricultural region is the most intensively human-impacted ecosystem type on the Qinghai–Tibet Plateau, and its ecosystem respiration (RE) plays an important role in regional carbon cycling and climate change responses. However, it remains unclear whether systematic differences exist in the spatiotemporal patterns of RE and its environmental controls across high-altitude agricultural watersheds. Using multi-source remote sensing and reanalysis data from 2000 to 2024, this study focuses on two representative valley agricultural basins on the Qinghai–Tibet Plateau—the Huangshui River Basin (HSH) and the Yijianglianghe River (YJLH). A dual-model framework combining a Random Forest model and a Lloyd–Taylor mechanistic model was developed to examine the spatiotemporal dynamics, driving mechanisms, and temperature sensitivity of RE. The results show that RE increased in both basins over the study period, while their dominant controlling mechanisms differed markedly. Gross primary productivity (GPP) was the most important driver in both basins (23.6% in HSH and 24.5% in YJLH). However, temperature and precipitation contributed more in YJLH (14.5% vs. 9.4%), suggesting that RE in the higher-altitude basin is more strongly constrained by hydrothermal conditions under colder and harsher climates. Spatially, high RE values in the HSH were mainly concentrated in mid- and low-elevation valley farmlands, whereas the YJLH exhibited a clear decrease in RE with increasing elevation, indicating a stronger topographic control. Both basins showed significant spatial clustering of RE (Moran’s I ≈ 0.93), with stronger spatial aggregation in the HSH. Temperature sensitivity (Q10) generally increased with elevation, and the YJLH exhibited markedly higher Q10 values in high-altitude regions, indicating stronger temperature responsiveness under extreme cold conditions. Empirical Q10 values were consistently higher than theoretical estimates, implying that ecosystem respiration is not only directly driven by temperature, but may also be amplified indirectly through enhanced vegetation productivity and increased substrate availability. Overall, this study reveals a clear divergence in RE control mechanisms across valley agricultural systems on the Qinghai–Tibet Plateau: productivity-driven regulation dominates in low-elevation regions, whereas hydrothermal constraints become increasingly important in high-altitude environments, leading to a transition from “productivity control” to “hydrothermal constraint control.” This shift highlights the nonlinear response of alpine agroecosystems to elevation gradients and climate change, providing mechanistic evidence for understanding carbon cycling in high-altitude anthropogenic ecosystems.