Hip — Historical data — machine learning and uncertainty

The model calculations provide a nationwide overview of the depth of the off-road groundwater in high spatial resolution (10x10 meters) for a typical winter situation and a typical summer situation on the basis of data for the period 1990-2019. For the two situations, an uncertainty band based on 10 and 90 % confidence limits is expressed. The depth is indicated in centimeters. The depth of ground water is calculated in 10 m spatial solution. The two models are trained against 75 % of data and 25 % of data are used for validation. A remedy absolute error is used in the training process. The parameters of the machine learning model were optimised with an automated trial-and-error method. The same parameter set is used for both models, winter and summer. The modelling calculations can be used as a starting point and basic data for assessments where the spatial variation of ground water is in focus, e.g. urban development planning, LAR, infrastructure planning, risk assessment of construction, design of drainage and drainage systems, etc. Risk assessment of soil and groundwater pollution, agriculture and other land use, etc. The insecurity band is useful to characterise the uncertainty of the ML model at grid level (10 m) for the whole country. The data set can usefully be used as zoom on similar and other data from 100 m model, as the two themes can be usefully interpreted.