Widespread Drying over Land Since the 1950s

Global warming is expected to increase surface evaporative demand of water vapor, which could lead to more droughts. Analyses of global climate model simulations suggest that wide-spread drying could occur over many land areas in the 21st century, and some studies suggest that some of this drying may have already been occurring since the 1980s. However, in a sharp contrast to several previous studies, a recent paper in Nature claims that there has been little drying trend since the 1950s based entirely on their calculated Palmer Drought Severity Index (PDSI). This has caused considerable confusion in the climate community. Here we update our PDSI analysis and present an evaluation of the various land precipitation products that show large disparities since around 1995 due to a sharp decline in analyzed rain-gauge data. We show that the CRU precipitation product (CRU TS 3.10.01), which was used in the Nature study and contained much fewer rain-gauge data than the GPCC v6 product for the recent years, exhibits a wet bias since 1995 compared with three other products and it results in much weaker drying trend for the last 60 years than the other products suggest. This wet bias leads to a spurious drop in global-mean runoff ratio around 1995 and generally lower correlation with the GRACE satellite observations of land water storage during 2003-2009 than the GPCC product. A new version of the CRU precipitation product (CRU TS 3.21) includes more gauge data and substantially reduces this wet bias for the period since 1950. Thus, it is likely that the wet bias since 1995 in the CRU product is spurious and it has contributed to the insignificant drying trend found in the Nature study. An updated analysis of the historical records of annual streamflow, precipitation, and PDSI all confirm our previous findings of significant global drying since the 1950s, with considerable contributions from surface warming since the 1980s.