Surrogate-based global sensitivity analysis for turbulence and fire-spotting effects in regional-scale wildland fire modeling

TL;DR

This paper develops and compares surrogate modeling techniques to efficiently perform global sensitivity analysis on wildfire spread models, focusing on turbulence and fire-spotting effects influenced by wind.

Contribution

It introduces a surrogate-based approach using gPC and Gaussian Process models to identify key parameters affecting wildfire behavior, optimizing computational efficiency.

Findings

gPC with sparse LAR performs best with small training sets

Surrogates effectively identify wind as the dominant factor

Sparse surrogates enable analysis of complex nonlinear models

Abstract

In presence of strong winds, wildfires feature nonlinear behavior, possibly inducing fire-spotting. We present a global sensitivity analysis of a new sub-model for turbulence and fire-spotting included in a wildfire spread model based on a stochastic representation of the fireline. To limit the number of model evaluations, fast surrogate models based on generalized Polynomial Chaos (gPC) and Gaussian Process are used to identify the key parameters affecting topology and size of burnt area. This study investigates the application of these surrogates to compute Sobol' sensitivity indices in an idealized test case. The wind is known to drive the fire propagation. The results show that it is a more general leading factor that governs the generation of secondary fires. This study also compares the performance of the surrogates for varying size and type of training sets as well as for varying parameterization and choice of algorithms. The best performance was achieved using a gPC strategy based on a sparse least-angle regression (LAR) and a low-discrepancy Halton's sequence. Still, the LAR-based gPC surrogate tends to filter out the information coming from parameters with large length-scale, which is not the case of the cleaning-based gPC surrogate. For both algorithms, sparsity ensures a surrogate can be built using an affordable number of forward model evaluations, while the model response is highly multi-scale and nonlinear. Using a sparse surrogate is thus a promising strategy to analyze new models and its dependency on input parameters in wildfire applications.