Drier summers were followed by later, not earlier, autumn leaf colouring at German phenology stations in 1951-2024; the Swiss estimate had the same sign but was uncertain

Question: do dry, hot, or hot-and-dry summers change the date when 50% of tree leaves have autumn colour? Data: 578,945 DWD records (6,402 German stations) and 17,988 MeteoSwiss records (173 Swiss stations), 1951-2024, for beech, birch, horse chestnut, larch, and oak in Germany only. Weather came from E-OBS v31.0e (0.25 degree): June-August Tmax, June-August precipitation z-score, and September temperature. For Germany in 1991-2024 we also used DWD 1 km soil moisture. Models had station and 5-year period fixed effects and two-way clustered errors. We wrote the plan before we fitted the models. Result: drier summers were followed by later colouring, not earlier colouring. In Germany the delay was +0.52 days per 1 SD drier summer (95% CI +0.14 to +0.90). It was +1.5 days after summers below -1.5 SD, and +1.49 days per 1 SD drier soil (CI +0.85 to +2.13). The sign was the same for all five species, in both periods and in four regions. Swiss estimates had the same sign for all four species, but the pooled CI included zero (+0.42, CI -0.13 to +0.97), and a placebo with next summer's rain had a similar size. No sample showed earlier colouring after dry summers with a CI that excluded zero. Summer heat and the heat x dryness interaction were not robust: they changed sign with the design and with the period, and the placebos were not zero. A 1 C warmer September delayed colouring by 1.2-1.4 days. Limitations: coarse weather grid; no observer identity; subjective 50% phase; within-year (station + year) models gave much smaller precipitation effects; colouring in the 2018 drought was earlier than predicted; the weather model did not predict 2018-2024 better than a trend.

autumn phenology · leaf senescence · leaf colouring · drought · summer heat · compound extremes · soil moisture · DWD phenology · MeteoSwiss phenology · E-OBS · Fagus sylvatica · Betula pendula · Larix decidua · Quercus robur · Aesculus hippocastanum · fixed effects · placebo test · observation bias · climate change · Germany Switzerland