Optimal temperature of vegetation productivity and its linkage with climate and elevation on the Tibetan Plateau

Vegetation productivity first increases and then decreases with temperature; and temperature corresponding to the maximum productivity is called optimal temperature (T-opt). In this study, we used satellite derived near-infrared reflectance of vegetation (NIRv) data to map T-opt of vegetation productivity at the spatial resolution of 0.1 degrees on the Tibetan Plateau (TP), one of most sensitive regions in the climate system. The average T-opt of non-forest vegetation on the TP is about 14.7 degrees C, significantly lower than the T-opt value used in current ecosystem models. A remarkable geographical heterogeneity in T-opt is observed over the TP. Higher T-opt values generally appear in the north-eastern TP, while the south-western TP has relatively lower T-opt (<10 degrees C), in line with the difference of climate conditions and topography across different regions. Spatially, T-opt tends to decrease by 0.41 degrees C per 100 m increase in elevation, faster than the elevational elapse rate of growing season temperature, implying a potential CO2 regulation of T-opt in addition to temperature acclimation. T-opt increases by 0.66 degrees C for each 1 degrees C of rising mean annual temperature as a result of vegetation acclimation to climate change. However, at least at the decadal scale, there is no significant change in T-opt between 2000s and 2010s, suggesting that the T-opt climate acclimation may not keep up with the warming rate. Finally, future (2091-2100) warming could be close to and even surpass T-opt on the TP under different RCP scenarios without considering potential climate acclimation. Our analyses imply that the temperature tipping point when the impact of future warming shifts from positive to negative on the TP is greatly overestimated by current vegetation models. Future research needs to include varying thermal and CO2 acclimation effects on T-opt across different time scales in vegetation models.