An investigation of corrective approaches for uncertain winds and analysis of impacts on smoke model performance
| Authors: | Zongrun Li, Susan O'Neill, Rime El Asmar, Yongtao Hu, Adam K. Kochanski, Angel Farguell, Jan Mandel, David J. Tanner, L. Gregory Huey, Armistead G. Russell, Rodney J. Weber, M. Talat Odman |
| Year: | 2026 |
| Type: | Scientific Journal |
| Station: | Pacific Northwest Research Station |
| DOI: | https://doi.org/10.1016/j.agrformet.2025.110885 |
| Source: | Agricultural and Forest Meteorology |
Abstract
In wildland fires, wind affects fire propagation, emission intensity, and smoke transport; therefore, uncertainties in wind simulations can critically impact PM2.5 concentration produced by coupled fire-atmosphere models or chemical transport models. These uncertainties must be addressed prior to assessing other parameters that may affect model-predicted PM2.5 concentrations through comparisons with observations. This study simulated prescribed fire events with accurate ignition pattern design and high spatial-temporal resolutions using the coupled fire-atmosphere model WRF-SFIRE. We designed, implemented, and evaluated various wind bias reduction and smoke model evaluation methods to quantitatively capture the impacts of wind uncertainty on simulated smoke concentrations. For the wind bias reduction methods, incorporating wind observations into initial and boundary conditions proved effective in reducing bias, especially for wind speed. The suggested method resulted in RMSE values of 63 degrees for wind direction and 0.5 m/s for wind speed, lower than the standard WRF nudging benchmark’s 71 degrees and 1.5 m/s, respectively. This improvement in wind simulation accuracy enhanced the smoke simulation performance, successfully identifying 5 out of 8 detected PM2.5 peaks (≥ 35 μg/m3) missed in the benchmark WRF simulation using nudging. Among smoke model evaluation methods, which are post-analysis algorithms on the numerical modeling results, the equal time backward/forward trajectory method was the most effective approach. It successfully captured 7 PM2.5 peaks, which were not well simulated in the nudging benchmark. These smoke evaluation methods estimated uncertainty in smoke concentration by considering the simulated wind bias and demonstrated that the smoke concentration simulation is highly sensitive to the wind bias from the wind simulation. This study is unique in suggesting novel solutions to improving modeled wind field data and creative approaches to post-analysis of smoke transport simulations.