Effects of Precipitation Regimes on Ecosystem Respiration in Agricultural Regions of the Southern Tibetan Plateau

Abstract

Understanding how the spatiotemporal variability of precipitation affects ecosystem respiration (RE) is central to carbon–climate feedback in climate-smart agriculture, yet remains unresolved for the alpine agricultural region of the southern Qinghai–Tibet Plateau, where flux observations are sparse. Using 25 years (2000–2024) of monthly gridded climate and remote sensing data for the Yarlung Zangbo River Basin and Its Two Tributaries Basin, we developed a flux tower-constrained reference–respiration (Rref) environment-matching model in which Rref varies with the enhanced vegetation index (EVI) and land surface temperature (LST) to correct the Lloyd–Taylor parameterization. The correction reduced the RE root mean square error by 54.8% (1.04 → 0.47 gC·m−2·month−1) and eliminated systematic bias (+0.80 → −0.001) relative to an independent gridded RECO product. We then constructed a multidimensional index of precipitation variability (intra-annual concentration, interannual variability, long-term trend, spatial clustering) and combined random forest, spatial regression, lag analysis, and structural equation modeling (SEM) to disentangle direct and indirect pathways from precipitation variability to RE. The central finding is an indirect-conduction mechanism: precipitation concentration (PCI) affects RE almost entirely through vegetation productivity (PCI → GPP → RE, indirect effect −0.676) rather than directly (direct effect +0.076, opposite in sign), because low temperature and high soil water holding capacity buffer the immediate soil moisture response. The basin functions as a net carbon source (mean NEP = −0.550 gC·m−2·month−1) with a significant warming-driven interannual RE increase (Sen’s slope = 0.0025 yr−1, p = 0.022) that is independent of the stable precipitation total. The framework offers a transferable paradigm for carbon flux attribution in alpine regions under sparse observation.