Development and Assessment of an Ensemble Forecast System for Fire Weather and Public Safety Power Shutoffs in the Idaho Region

In the context of increasing fire weather concerns and the necessity of Public Safety Power Shutoffs (PSPS) for mitigating wildfire risks, we designed an ensemble forecast system. This presentation details the development and assessment of a convective-resolving ensemble prediction system tailored for the Idaho region, in collaboration with the Idaho Power Company (IPC).

We used the Weather Research and Forecast (WRF) model with a 5.4-km outer domain grid spacing and a 1.8-km inner nest to explore various ensemble generation strategies that effectively capture both model and initial/boundary condition uncertainties using the Global Ensemble Forecast System (GEFS) for initialization. Our investigation to design an operational ensemble configuration included exploring a multiphysics ensemble approach, as well as stochastic perturbation schemes—the Stochastic-Kinetic Energy Backscatter (SKEB) and the Stochastically Perturbed Physics Tendency (SPPT) schemes—along with ensemble time-lagging. The initial ensemble design and development focused on a week-long forecast featuring a period of extreme wind and thunderstorm activity in early August 2022. The design also took into consideration IPC’s real-time requirements and computational resource constraints. The prototype ensemble is being run in real-time for ongoing monitoring and evaluation.

Based on the week-long forecast tests, we propose a GEFS-based time-lagged ensemble augmented by the SKEB and SPPT schemes for optimized and reliable computational performance, as well as a balanced model spread in relation to observation errors, particularly in the 10-m wind speed. The skill of the ensemble forecast system is assessed using surface observations from the ASOS network available within our Idaho-centered WRF domain. Our assessment includes both deterministic (e.g., root mean-squared error, mean error) and probabilistic (e.g., reliability diagram, rank histogram) evaluation metrics to assess the initial performance of the system. Through this evaluation, we determined the initial performance of our ensemble prediction system, laying the foundation for ongoing tuning and calibration efforts to enhance its reliability and accuracy.