The vertical distribution of biomass burning aerosol (BBA) is important in regulating their impacts on weather and climate. The plume-rise process affects the injection height of wildfire aerosols and interacts with air parcel lifting and the cloud processes, but these processes are not represented in most global climate model. In this study, we designed two approaches to address this issue: a process-based modeling technique and machine learning (ML) modeling technique.

In the former approach we replaced the fixed vertical profiles of monthly BBA emissions in E3SM with an interactive fire plume-rise model. The vertical distribution of BBA emissions for each model column was calculated as a function of ambient thermodynamic conditions from the host E3SM, and distributions of fire sizes and sensible heat fluxes from observations. The maximum fire radiative power (FRP) technique was used to determine the active fire size and scaled MODIS FRP observations for the wildfire sensible heat release. Daily BBA emission, superimposed with a fire diurnal cycle retrieved from satellite observation, was included in model simulations. The model showed improved agreement with satellite retrievals and in-situ measurement during the NOAA WE-CAN campaign. The results demonstrate the importance of the plume-rise model and fire diurnal cycle assumption in determining the BBA emissions. We also find that the new model produces a larger carbonaceous aerosol burden, leading to 0.13 Wm^-2 warming at the top of atmosphere compared to the default E3SM. 

Even though the plume-rise model can explicitly and accurately simulate the plume injection heights, it is computational expensive. Therefore, in the second approach, we utilized artificial intelligence and machine learning (AI/ML) techniques to develop an emulator for simulating the plume-rise process. Specifically, we trained a multilayer perceptron (MLP) artificial neural network using fire properties and atmospheric thermodynamic fields as predictors and the plume injection heights observed by NASA MISR as the target variable. The training and evaluation results demonstrate that MLP can reasonably predict the plume injection height with an error of approximately 200 meters.