A comparison of fire severity measures: An Australian example and implications for predicting major areas of soil erosion

Modelled results of two popular fire severity measures NBRdiff and NDVIdiff were compared with field data for two small wildfires that occurred in eucalypt-dominated forest near Sydney, Australia during the spring of 2006. NBRdiff and NDVIdiff are computed by differencing specific spectral reflectance properties of satellite images captured before and after the fire event. In this case indices were derived from the Landsat 5 Thematic Mapper platform. It is demonstrated that the spatial distribution of severity classes computed using NBRdiff were not statistically different to severity classes computed for NDVIdiff in the Tallowa fire, which was a ‘hot’ burn. However, in the ‘cooler’ Thirlmere Lakes fire, their respective classifications differed significantly. NBRdiff was clearly superior to NDVIdiff in discriminating lower fire severities, which predominantly burn understorey vegetation whilst leaving the canopy un-scorched. It is also demonstrated using post-fire soil profiles from 50 sample sites that charring depth, which is related to the hydrophobic properties of the soil landscape, is significantly correlated to fire severity. This has important implications for the probability of developing a soil erosion risk model based on remotely sensed data, because in this environment, charred soil landscapes will have lost some of their naturally highly water repellent behaviour during burning, which in turn alters their susceptibility to overland flow development and surface wash erosion.