Previous-year melt did not change the next summer balance of 233 WGMS glaciers in 1951-2023 with Berkeley Earth temperature; GHCN-CAMS temperature gives false lag terms

Question: does a year with strong melt make the next summer's melt larger than the next summer's weather explains (memory from firn loss and darker ice)? We used 4,251 glacier-years of measured summer and winter balance from 233 glaciers in the WGMS Fluctuations of Glaciers database (2026-02-10). We used summer air temperature at each glacier from Berkeley Earth. The models had glacier and year fixed effects and two-way clustered errors. Same-summer temperature gave -0.49 m w.e. per deg C. Previous-summer temperature gave +0.029 (95% CI -0.010 to +0.068), and memory predicts a negative value. Next-summer temperature (placebo) gave -0.005. Previous-year annual balance gave +0.002 m w.e. per m w.e. (CI -0.049 to +0.053), with the same result in 7 specifications. After a glacier's lowest-decile year (residual -0.41 m w.e.), the next-year residual was +0.001 (CI -0.06 to +0.06; 495 events). With training on 1951-2005 and testing on 2006-2024, lagged terms did not decrease the forecast error. The Alps in 2023 had a residual of -0.22 m w.e. after 2022, but this is one regional case. The planned GHCN-CAMS analysis failed its placebo test: previous-summer and next-summer temperature both gave about +0.09 to +0.10. GHCN-CAMS summer balance responded 3 to 4 times less to multi-year temperature than to year-to-year temperature, and Berkeley Earth did not show this difference. A GHCN-CAMS model also predicted 0.35 m w.e. per year too little loss in 2006-2025. Limitations: the temperature is from coarse grid cells, the models have no radiation or summer snowfall terms, some national series are model-based, and memory through thinning over decades is not tested.

glacier mass balance · WGMS · melt-albedo feedback · firn · recovery · extreme melt years · temperature sensitivity · Berkeley Earth · GHCN-CAMS · placebo test · event study · fixed effects