Fire rate-of-spread model inversion: what can be inferred from remote sensing observations of fire behaviour?

Abstract

Listen

Translate article

Understanding the drivers of wildfires is crucial to evaluating fire's impact on natural processes and their feedback on the environment and society. Due to the difficulty of directly observing fires in the wild, our current knowledge of wildfire behaviour heavily relies on lab experiments and open-controlled fires, which are the main source for our models and assumptions. Thus, there are still several gaps on how spot samples upscale to ecosystems, how different fuel types with varying moisture contents should be represented in the models, or how fire responds to the intrinsic structural complexity of ecosystems. Some of this information is currently encoded in fuel models, which are primarily based on vegetation types and used as prescribed inputs in our models. However, they are a documented source of uncertainties and inaccuracies when applied beyond the spatial domain from which they were originally formulated. Global fire behaviour observed through remote sensing could provide more general insights on the emergent response of fire to this complexity. In this work, by inverting Rothermel’s model for fire rate of spread (ROS), we show how field observations of fuel load, fuel moisture and environmental variables (temperature, wind speed) relate to the remotely sensed ROS and fire radiative power (FRP). We used fuel load measurements provided by the Public LANDFIRE Reference Database, fuel moisture data provided by the National Fuel Moisture Database, fire ROS from the Fire Atlas, FRP from the MCD14ML product, and vegetation backscattering signal from Sentinel -1 mission over the conterminous United States. Preliminary results show only partial agreement between the components of the ROS equations calculated from observations and estimated by model inversion. However, these findings suggest the possibility of bypassing the empirical estimations of fuel load and fuel moisture content, which use remote sensing, and directly use these signals to calculate a potential ROS as a fraction of wind speed.