Fig. 1: Impacts of forest restoration on water fluxes and water availability.

a, The tree-restoration potential²: the percentage area of each pixel that is suitable for tree restoration. b, Model ensemble mean versus observed streamflow (Q) measurements³⁰ for 19 validated river basins. The error bars for Q Budyko indicate the standard deviation over the six Budyko models. The error bars for Q validation indicate a 20% error. The river basins are Amazon (AM), Brahmaputra (BR), Colorado (COL), Congo (CON), Danube (DA), Ganges (GA), La Plata (LA), Mackenzie (MA), Mekong (ME), Mississippi (MI), Murray–Darling (MU), Niger (NIG), Nile (NIL), Orinoco (OR), Rhine (RH), Volga (VO), Yangtze (YA), Yenisei (YE) and Zambezi (ZA). c–f, The absolute annual change in water fluxes after tree restoration: change in evaporation (c), precipitation (d), water availability without evaporation recycling (e) and water availability with evaporation recycling (f). Note that e is the inverse of c: without the feedback of evaporation recycling, the local increase in evaporation equals the local decrease in water availability. g, The histogram shows the distribution of the global changes in water availability without and with evaporation recycling; 89% (without recycling) and 91% (with recycling) of the data fall within the displayed range of –20 mm yr^–1 to +10 mm yr^–1. All maps display the 0.1° mean values, except for c, which displays the 0.5° mean value.