Case Studies of Methane Plume Detection Using WRF-LES

Forecasting of urban dispersion of hazardous materials is crucial for emergency response planning. Uncertainty in atmospheric conditions and the chaotic nature of plume mixing makes a single deterministic solution insufficient. This study investigates the use of ensemble modeling in simulating methane gas leaks using the Weather Research and Forecast (WRF) model with an embedded large-eddy simulation (LES) in comparison to in-situ methane concentration measurements for two different case studies.

The WRF model simulates both single (Case 1) and double (Case 2) leaks at a test site in Central California with a 4.5 km x 4.5 km outer domain (300 m horizontal resolution) and a 625 m x 625 m nested domain (25 m horizontal resolution). Only the inner domain utilizes LES and WRF-Chem, while the outer domain uses a parameterized boundary layer. Five leak locations along a North/South line are simulated. For each location four ensemble members, each with different forecast lead times and physics parameterizations, are simulated. Due to the relatively coarse LES resolution, the WRF simulations under predict the methane concentration in both cases. However, the simulated leak locations closest to the observed leak location do show a qualitative agreement with observations and RMSE values are lowest for the actual leak location. Analysis of the ensemble members shows that the members do not help in plume detection in these test cases.

Simulated wind speed and direction also differed compared with observations in both cases. For Case 1, the magnitude of the wind speed was identical to observations while the direction was offset -47 degrees. To preserve the angle between the in-situ measurement path and the wind direction, the simulated plume is rotated by the difference in wind direction. The RMSE for the rotated plume at the actual location is slightly improved compared to the unrotated plume. In Case 2, the difference in wind direction was 58 degrees, while the simulated magnitude was 4.89 m/s compared with the observed value of 1.74 m/s. Rotation only produces a RMSE that is greater than the unrotated plume. Preliminary testing on how to simultaneously account for the differences in wind direction and speed will be discussed.