A negative mass-balance year did not make the next summer balance more negative beyond climate on 143-227 WGMS glaciers in 1950-2024; the forecast gain from the previous year was regional

Question: after a very negative glacier mass-balance year, does less firn and a darker surface make the next summer balance more negative than the weather of that summer predicts (melt memory)? Approach: we used glaciological balances from the WGMS Fluctuations of Glaciers database (version 2026-02-10) and 0.5 deg GHCN_CAMS air temperature. We used fixed-effects models with glacier and calendar-year effects and three tests: (1) a direct test of the previous-year annual balance with a winter-balance control outcome; (2) an exogenous test that compares the effect of the previous summer temperature with that of the next summer temperature, which is a placebo; (3) an out-of-sample forecast test (fit 1950-2005, test 2006-2024). Results: the previous-year balance coefficient on summer balance was 0.009 m w.e. per m w.e. (95% CI -0.039 to 0.058; 4,277 glacier-years, 154 glaciers). The previous-minus-next summer temperature contrast was +0.005 m w.e. per K (-0.048 to 0.057; 143 glaciers), which gives an implied memory coefficient of -0.012 (-0.072 to 0.051). For annual balances of 227 glaciers, it was 0.041 (-0.036 to 0.117). Results were similar across time systems, flags, periods and AAR groups. Out of sample, the own previous-year balance lowered test error by about 3%, but the mean previous-year balance of other glaciers in the region lowered it by 4-5%, and the own value then added at most 0.6%. Thus the useful persistence is regional, not glacier-specific. Limitations: year effects remove global shocks; temperature is a coarse grid proxy; many series are reanalysed; smaller regional subsets do not exclude a memory coefficient of about 0.3; elevation bands were not analysed.

glacier mass balance · WGMS · melt-albedo feedback · firn · persistence · placebo test · out-of-sample forecast · GHCN_CAMS · glacier runoff · fixed effects