Predicting the spread and environmental impacts of wildland fires remains an inexact science. Extreme temperature shifts due to climate change alongside continuous societal expansion have caused the wildland urban interface to grow. As this interface grows, these natural disasters will become deadlier and costlier each year. There has been a push for physics-based predictive fire spread models, yet simulations remain challenging due to the immense computational requirements introduced by the complex chemistry of biomass pyrolysis, ignition, and combustion. Furthermore, the large separation of spatial-temporal scales requires most simulations to resolve large scales and model small scales. The range of spatial scales in wildland fires spans from global weather patterns, on the order of thousands of kilometers, to chemical reactions on sub-millimeter scales. The range of temporal scales spans from days to reaction scales on the order of microseconds. The research presented in this thesis is focused on enabling physics-based computational modeling of chemical processes occurring at small spatial scales and on short time scales, with a view towards implementing the resulting models in much larger-scale and more computationally intensive simulations of wildland fire spread.

This dissertation outlines the steps taken to address the computational and time requirements required for simulations of reacting flows. These consist of a priori kinetic mechanism reduction, the integration of adaptive mesh refinement and dynamic load balancing with a multi-phase simulation, the use of tabulated chemistry with various skeletal models and the derivation of a lumped specie for use in an infinitely-fast chemistry simulation. Ultimately, the work encompassed here provides a computationally-efficient framework for simulations of biomass pyrolysis, ignition, and combustion.

The introductory chapter provides additional motivation for the study of wildland fire simulations, an expansive outline of the thesis and a listing of the contributions made to the research community. The second chapter provides background on mechanism reduction and its application to a comprehensive, biomass-specific mechanism to produce something more suitable for a large simulation. The third chapter discusses the development and verification of a computational framework that is compared to a cone-calorimeter experiment. The fourth chapter investigates the role of reduced models in the characteristics of reacting plumes. The final chapter discusses the overall conclusions of this work and contributions made in the dissertation.