TY - JOUR
T1 - What determines variation in remotely sensed fire severity? Consideration of remote sensing limitations and confounding factors
AU - Gale, Matthew G.
AU - Cary, Geoffrey J.
N1 - Publisher Copyright:
© 2022 The Author(s) (or their employer(s)). Published by CSIRO Publishing on behalf of IAWF.
PY - 2022/3
Y1 - 2022/3
N2 - Analyses of the effects of topography, weather, land management, and fuel on fire severity are increasingly common, and generally apply fire severity indices derived from satellite optical remote sensing. However, these indices are commonly interpreted with insufficient appreciation for their limitations and may be inappropriately invoked as representing physical fire effects and fire behaviour. For a large wildfire in southeast Australia, we investigated three considerations for inferring robust insights from fire severity analyses-the potentially confounding influences of pre-fire vegetation height and tall vegetation cover, and the choice of fire severity response variable. Using nonparametric regression, we found that different fire severity indices gave rise to substantially different modelled relationships with commonly invoked environmental predictors, as is consistent with dissimilarities in index design. Further, pre-fire vegetation height was a strong control of fire severity, with equivalent importance to weather. Importantly, strong covariation between vegetation height and environmental predictors suggests that modelled fire severity effects are strongly influenced by variation in vertical distance between flames and vegetation, and this confounds fire behaviour insights. To enable more robust and mechanistic insights into the determinants of fire severity, we recommend greater consideration of the meaning and limitations of optical remote sensing indices.
AB - Analyses of the effects of topography, weather, land management, and fuel on fire severity are increasingly common, and generally apply fire severity indices derived from satellite optical remote sensing. However, these indices are commonly interpreted with insufficient appreciation for their limitations and may be inappropriately invoked as representing physical fire effects and fire behaviour. For a large wildfire in southeast Australia, we investigated three considerations for inferring robust insights from fire severity analyses-the potentially confounding influences of pre-fire vegetation height and tall vegetation cover, and the choice of fire severity response variable. Using nonparametric regression, we found that different fire severity indices gave rise to substantially different modelled relationships with commonly invoked environmental predictors, as is consistent with dissimilarities in index design. Further, pre-fire vegetation height was a strong control of fire severity, with equivalent importance to weather. Importantly, strong covariation between vegetation height and environmental predictors suggests that modelled fire severity effects are strongly influenced by variation in vertical distance between flames and vegetation, and this confounds fire behaviour insights. To enable more robust and mechanistic insights into the determinants of fire severity, we recommend greater consideration of the meaning and limitations of optical remote sensing indices.
KW - airborne LiDAR
KW - ecosystems: Temperate
KW - fire behaviour
KW - fire intensity
KW - fire severity
KW - remote sensing
KW - spectral indices
KW - vegetation cover
KW - vegetation height
UR - http://www.scopus.com/inward/record.url?scp=85127465741&partnerID=8YFLogxK
U2 - 10.1071/WF21131
DO - 10.1071/WF21131
M3 - Article
SN - 1049-8001
VL - 31
SP - 291
EP - 305
JO - International Journal of Wildland Fire
JF - International Journal of Wildland Fire
IS - 3
ER -