Supplementary material to "Advances in Land Surface Model-based Forecasting: A Comparison of LSTM, Gradient Boosting, and Feedforward Neural Networks as Prognostic State Emulators in a Case Study with ECLand"

S2.1 European subregions for horizons computationA northern European subset was selected on Southern Scandinavia with a grid box on minimum and maximum latitudes of 55 and 71 degree respectively and a minimum and maximum longitude of 5 and 20 degree respectively.This resulted in a subset of 755 grid cells.For the southern European region, a grid box was created over France with minimum and maximum latitudes of 41.5 and 51.1 degree respectively and a minimum and maximum longitude of -5.1 and 6 degree respectively.Summary statics for the prognostic state variables in these regions are listed in table S1.Table S1: Exemplary summary statistics of the seven prognostic target variables over two European training data subsets, northern and southern.Mean, standard deviation and their ratio (Signal-to-noise ratio, SNR) are aggregated over times and grid cells.Northern Europe Southern Europe Mean Standard dev.SNR Mean Standard dev.SNR