Building Height Estimation in Halle Using Geoclimate and Machine Learning

This GeoTIFF dataset contains estimated building heights for Halle, Germany, produced using the Geoclimate model—an open-source geospatial processing toolbox implementing a random forest regression algorithm. The model was trained on a dataset covering 14 French communes, using ground truth building heights provided by BDTopo (IGN France).

For inference, the model automatically retrieves building footprints from OpenStreetMap (OSM) and computes a set of 62 geospatial indicators describing each building's local environment (e.g. density, proximity to other structures, land use context). These indicators serve as input features to the trained random forest, which predicts the building height.

This dataset is an output of FAIRiCUBE Use Case 4 (UC4): Spatial and temporal assessment of neighbourhood building stock, which aims to evaluate energy consumption and material stocks in urban environments using harmonized methods across European cities.