Using machine learning to produce a cost-effective national building height map of Ireland to categorise local climate zones

Abstract. ECOCLIMAP-Second Generation (ECO-SG) is the land-cover map used in HARMONIE-AROME configuration of shared ALADIN-HIRLAM Numerical Weather Prediction system for short-range operational weather forecasting Ireland. The ECO-SG urban classification implicitly includes building heights. work presented this paper involved production first open-access height island Ireland which complements Ulmas-Walsh land cover map, a has improved horizontal extent areas over resulting will potentially enable upgrades to information future implementation HARMONIE-AROME. This study not only produced at 10 m × resolution, but assessed various types regression models trained using pre-existing Dublin City and selected 64 important spatio-temporal features, engineered from both Sentinel-1A/B Sentinel-2A/B satellites. performance metrics revealed that Convolutional Neural Network superior all aspects except computational time required create map. Despite accuracy Network, final created results ridge model provided best blend realistic output low complexity. method relies solely on freely available satellite imagery, cost-effective, can be updated regularly, applied other regions depending availability representative regional sample data.