Projecting precipitation-driven vegetation changes on the Tibetan Plateau using in-situ station data

Vegetation on the Tibetan Plateau (TP) plays a crucial role in continental energy, water and carbon budgets. Throughout the rest of the 21st century, TP vegetation faces two key pressures: First, the impact of climate change-induced atmospheric changes; and second, potential shifts imposed by direct land use. In this study we characterize rainfall-driven local vegetation changes through combining in situ station data and satellite observations. We then project future vegetation changes using simulated precipitation characteristics from the latest state-of-the-art Earth System Models (ESMs).

We find that 61 out of 90 stations have a significant precipitation-vegetation relationship compared to only 29 stations with a significant temperature-vegetation relationship. Therefore, to project future vegetation changes, we focus on local partial correlations between interannual variations of precipitation and satellite retrievals of the Normalized Difference Vegetation Index (NDVI). Most sites with significant partial correlations are characterized by short vegetation conditions (94 %). Significant stations are also associated with low climatological precipitation and dry surface conditions. We study connections between simultaneous interannual variations of NDVI and precipitation, as temporal trends of these variables are insufficient given the short time span of observations (20 years). At most short vegetation sites, precipitation increases NDVI. For example, at Madou, a well-established grassland site, NDVI increases by 0.05 (approximately 18 %) with an annual accumulation rainfall increase of 100 mm, which is within the limits of interannual precipitation variations at this site.

We take our site-specific partial correlations, where statistically significant, and merge them with projections of future rainfall to estimate NDVI throughout the rest of the century. Across the TP, we estimate typical NDVI increases of 10 %, when using the “business-as-usual” scenario to force ESMs. However, uncertainties in estimated NDVI occur due to substantial variations in projected precipitation by ESMs and site-specific attributes. Our analysis encourages constraining future precipitation projections by ESMs, developing land models specific to the TP, and increasing the duration and number of in situ observational sites.