Fig. 1: Trends and variability in TWS and rainfall concentration.
From: More concentrated precipitation decreases terrestrial water storage

a, Standard deviation of detrended water-year mean TWS from GRACE satellite observations25, over 2002–2022. b, Linear trend in water-year mean GRACE TWS over 2002–2022, shown as change per 20 years. c, Illustrative daily precipitation time series from Phnom Penh, Cambodia, for water-years 2003 (light blue) and 2016 (dark blue), the years with maximum and minimum daily precipitation concentrations, respectively, in the study period as quantified using the Gini coefficient. Data are from the Climate Prediction Center (CPC)22. Annotations indicate contrasting features of concentration in 2016 compared with 2003. d, Lorenz curves (fractional cumulative sum of ranked daily precipitation) for the two time series in a (light and dark blue). The grey dotted line shows the hypothetical Lorenz curve for equal daily precipitation. The daily precipitation Gini coefficient (GP) is calculated from the shaded integrals A, B and C, as annotated. e, Global dependence over 2002–2022 among interannual variations in GP and key hydroclimate variables, including extreme precipitation days (exceeding the 99th percentile of climatological nonzero daily precipitation), dry days (days with less than 0.1 mm precipitation), annual mean 2 m air temperature and annual total precipitation. Error bars show the range of estimates across three precipitation data products. f, Mean GP over 1980–2022, with the location of Phnom Penh indicated with a red cross. g, GP trends over 1980–2022, expressed as changes per 20 years. Data shown in f and g averages across the CPC, Global Precipitation Climatology Center (GPCC)23 and Global Precipitation Climatology Project (GPCP)21 daily precipitation datasets; results for individual data products are shown in Extended Data Fig. 3. Base map adapted from Natural Earth (https://www.naturalearthdata.com).