We used 1463 observations of early leaf discoloration presence or absence in Switzerland's European beech (Fagus sylvatica L.) forests between 2018 and 2023 and interpolated Sentinel-2 vegetation index time series from Koch et al. (2024) to train a support vector classifier of early leaf discoloration. We applied this classifier to the European beech dominated forest area of Switzerland over the same time period to estimate a probability of early leaf discoloration between 0 and 1 (high probability of early leaf discoloration). The maps intentionally exclude legacy effects of drought like crown dieback or mortality which occurred in prior years.

Early leaf discoloration maps are available at the end of every summer month (June, July, August, and September) 2017 to 2023. Additional maps are available every 5-days during the summer months of 2018. Probability of early leaf discoloration at each date is provided as a GeoTIFF file with a scaled value range of 0 to 10,000 such that a probability of 0.9 is represented by a value of 9000 (a high probability of early leaf discoloration). An optimized probability threshold of 0.4 (4000) is recommended for classification into discolored and not discolored pixels. More details on the methods and results can be found in the publication of Bloom et al. (2026).

These maps represent an interpreted remote sensing product that approximates ground observations of early leaf discoloration across Switzerland's European beech dominated forests.

Given ongoing efforts to operationalize an early discoloration pipeline, all data is currently available upon reasonable request. An example of the 2018 5-day time series in the Hardwald (Canton Basel-Land) is available for immediate download.