A physics-informed deep learning urban canopy model for Mediterranean cities: Development, validation, and application to Thessaloniki, Greece

2026-07-09

Ioannis Stergiou *

Serafim Kontos

Georgios Spyrou

Dimitrios Melas

DOI：10.1016/j.uclim.2026.103039

Original

Original request for help

Save

Abstract

En 中文

• Physics-informed DNN urban canopy model trained on 16 global sites. • Independent Heraklion validation confirms Mediterranean transferability. • DNN-UCM-PI is ∼225× faster than offline TEB. • Thessaloniki Bowen ratio of 4.9 indicates dry urban energy partitioning. • Local UHI causes 37% Qh underestimation when ERA5 forcing is used.

Keywords:

Urban canopy model

Deep learning

Physics informed

Surface energy balance

Urban-PLUMBER

Mediterranean climate

Urban heat island

A physics-informed deep learning urban canopy model for Mediterranean cities: Development, validation, and application to Thessaloniki, Greece

Abstract

Journal

Organization

Cited Papers

Citing Papers