The model
Three global ensemble models, one local truth. Every forecast cycle the system extracts ECMWF IFS ENS, ECMWF AIFS ENS and DWD ICON-EU at the orchard, then:
calibrates each model's mean and spread against the on-site station, per lead time — fitted by maximum likelihood;
blends them by verified likelihood on identical samples;
decides: we calculate a probability distribution → determine the frost probability in the future.
extra in your orchard an additional data based model is used to complement the global forecast, improving the existing metrics.
This example shows that DWD-ICON and IFS ENS are short term better while the AIFS ENS model wins at longer dates.
Static snapshot of the Phase-1 back-analysis; full methods and math live in the repository's back-analysis report.