Gradient boosting-based soil wetness for forestry climate adaptation in HarvesterSeasons service -training a model to forecast soil water index SWI from a comprehensive set of IFS model predictors in Destination Earth

Gradient boosting-based soil wetness for forestry climate adaptation in HarvesterSeasons service -training a model to forecast soil water index SWI from a comprehensive set of IFS model predictors in Destination Earth was an exercise that clearly improved a crucial part of the forestry climate service HarvesterSeasons.com. Forestry in nordic countries has to adapt to good and sustainable winter conditions being less and less available. Dry summer conditions are being looked for to compensate for weak winter times.

We present our service, the machine learning method for the new product and the validation of the new product. For machine learning Xtreme Gradient was used to train the Earth Observation product Soil Water Index from ERA5-Land, soilgrids.org and other features. Predicting is then enabled from Destination Earth Extremes and Climate Adaptation Digital Twins.