Impacts of grazing exclusion on productivity partitioning along regional plant diversity and climatic gradients in Tibetan alpine grasslands

The biodiversity-productivity relationship is critical for better predicting ecosystem responses to climate change and human disturbance. However, it remains unclear about the effects of climate change, land use shifts, plant diversity, and their interactions on productivity partitioning above- and below-ground components in alpine grasslands on the Tibetan Plateau. To answer this question, we conducted field surveys at 33 grazed vs. fenced paired sites that are distributed across the alpine meadow, steppe, and desert-steppe zones on the northern Tibetan Plateau in early August of 2010-2013. Generalized additive models (GAMs) showed that aboveground net primary productivity (ANPP) linearly increased with growing season precipitation (GSP) while belowground net primary productivity (BNPP) decreased with growing season temperature (GST). Compared to grazed sites, short-term fencing did not alter the patterns of ANPP along climatic gradients but tended to decrease BNPP at moderate precipitation levels of 200 mm < GSP < 450 mm. We also found that ANPP and BNPP linearly increased with species richness, ANPP decreased with Shannon diversity index, and BNPP did not correlate with the Shannon diversity index. Fencing did not alter the relationships between productivity components and plant diversity indices. Generalized additive mixed models furtherly confirmed that the interaction of localized plant diversity and climatic condition nonlinearly regulated productivity partitioning of alpine grasslands in this area. Finally, structural equation models (SEMs) revealed the direction and strength of causal links between biotic and abiotic variables within alpine grassland ecosystems. ANPP was controlled directly by GSP (0.53) and indirectly via species richness (0.41) and Shannon index (- 0.12). In contrast, BNPP was influenced directly by GST (- 0.43) and indirectly by GSP via species richness (0.05) and Shannon index (- 0.02). Therefore, we recommend using a joint approach of GAMs and SEMs for better understanding mechanisms behind the relationship between biodiversity and ecosystem function under climate change and human disturbance.