GAN-Generated Elevation Models In Computational Fluid Dynamics: A Feasibility Study For Complex Urban Terrain

Kurzfassung

Recently developed methods to simulatevery high-resolution(VHR)wind fields over complex urban terrain rely on high-quality three-dimensional vector representations of buildinginformation. Unfortunately data of that kind is sparsely avail-able on a worldwide scale. In this work, we investigate theapplicability ofcomputational fluid dynamics (CFD)on 2.5Ddigital surface models (DSMs)automatically generated bygenerative adversarial network (GAN)from globally avail-able satellite data which includes photogrammetric DSMs andpan-chromatic (PAN)images. The obtained results demon-strate that the GAN-based DSMs are reasonable alternativesto rarely availablelevel of detail 2 (LoD2)vector data, pro-moting large coverage, continuous wind field derivation overcomplex terrain.