Dissertation Information

Title: Smoke3d–Wildfire Smoke Transport Modeling with Adaptive Mesh Refinement

Program: Computing Ph.D. Computational Math Science and Engineering emphasis

Advisor: Dr. Donna Calhoun

Committee Members: Dr. Jodi Mead, Dr. Michal Kopera, Dr. Hans F. Schwaiger

Abstract
Models for tracking airborne pollutants, such as wildfire smoke, often struggle with uncertainties stemming from coarse parameterizations and limited physics. These challenges lead to inaccuracies in predicting plume behavior and emission concentrations, hindering effective decision-making.

We propose to develop Smoke3d, a model designed to simulate the long-range transport of wildfire smoke and estimate PM2.5 concentrations. To address uncertainties, Smoke3d will incorporate observational data through weak-constraint variational data assimilation. Built on the adaptive mesh refinement capabilities of ForestClaw,  Smoke3d dynamically refines the mesh in regions where smoke is present, enhancing resolution without excessive computational cost. 

The development of Smoke3d will be organized through three related projects: (1) Data assimilation, where we incorporate errors into the model to reduce uncertainties. The solution of the forward and adjoint transport problems using realistic wind field data, and validate the results using both artificial and realistic velocity fields; (2) Adjoint solver and representers, where we delve into the implementation of the adjoint-forward modeling framework and the use of representer functions to decouple these equations, enabling full data assimilation for reducing uncertainties; and (3) Testing the efficacy of the adjoint formulation with synthetic data and the full Smoke3d with real-time PM2.5 observations, enabling validation in controlled and realistic settings. This data-enabled adaptive modeling framework offers a path toward more accurate simulations of wildfire smoke transport and improved PM2.5 concentration estimates.