Nature 545, 169–174 (2017); doi:https://doi.org/10.1038/nature22062

In this Article, I estimated net glacial melt volumes on the river-basin scale from long-term precipitation and temperature records (1951–2007), taking into account the various mass contributions from avalanching, sublimation, snow drifting and so on. To this component (the seasonally delayed turnover of water in the glacial system) I added an estimate of the contribution due to sustained glacial mass losses, based on sparse observations of multi-decadal change. I then accounted for meltwater losses through evaporation, and compared this to net precipitation, distributed across river basins and across the catchments of a large number of dams. I estimated the second meltwater component (the additional contribution from glacier losses) as −0.35 to −0.40 metres water-equivalent per decade based on a global compilation of long-term mass-balance observations (from table 2 in ref. 32 of the Article). In this table, losses are described as “decadal averages (millimetres water equivalent)” but the units are actually intended to be decadally averaged annual values. Hence, the loss components of total meltwater that I used in my calculations are too small and the summed meltwater volumes reported here should be larger. Asia’s glaciers are thus regionally a more important buffer against drought than I first stated, strengthening some of the conclusions of this study but also altering others. I am therefore retracting this Article. I thank L. Zhao and J. Moore for bringing the error to my attention.