High-resolution WRF modeling in the Pyrenees: Sensitivity to static data for complex terrain

Complex terrain poses significant challenges for Numerical Weather Prediction (NWP) models, particularly in capturing localized boundary layer phenomena such as thermal circulations, katabatic flows, and temperature inversions. This study focuses on the Pyrenees mountain range, a region where accurate high-resolution forecasting is critical for understanding local weather extremes and variability, especially during synoptically quiescent conditions.

As part of a doctoral research project integrating Artificial Intelligence with high-resolution NWP, this work presents the foundational optimization of the Weather Research and Forecasting (WRF) model (v4.6.1). The modeling setup utilizes a one-way nested domain configuration bridging synoptic scales down to turbulence-resolving resolutions (333 m and 111 m LES), driven by ERA5 and GFS boundary conditions. We hypothesize that standard static input data provided by default in the WRF Preprocessing System (WPS) are insufficient to resolve the intricate surface heterogeneity of the Pyrenees. To address this, we conduct sensitivity experiments comparing the default USGS/MODIS configurations against enhanced high-resolution static datasets: 1-arc-second (~30 m) SRTM topography and the 100 m Copernicus Global Land Cover (CGLS-LC100). We evaluate the model’s performance in reproducing key local effects, focusing on surface wind fields, valley-floor cold pools, and thermal gradients under stable stratification.

Preliminary results quantify the bias reduction achieved by updating surface boundary conditions, establishing a robust baseline configuration. These findings are a prerequisite for subsequent full Large Eddy Simulations (LES) and the development of AI-driven bias correction schemes aimed at reducing computational costs while preserving accuracy in complex terrain.

This research has been funded by projects ARTEMIS (PID2021-124253OB-I00) and LIFE22-IPC-ES-LIFE PYRENEES4CLIMA.