Rocky Mountain Research Station Logo USDA Forest Service
Rocky Mountain Research Station
Forestry Sciences Laboratory - Moscow, Idaho
Moscow Personnel  |  Site Index  |  Site Map  |  Moscow Home
Project Information  |  Modeling Software  |  Library  |  Project Photos  |  Offsite Links  |  Eng. Home

Soil & Water
Engineering Publications

Project Leader:
William J. Elliot
email Bill

Contact Webmaster
email webmaster

Database updated
51 days ago

Using hyperspectral imagery to predict post-wildfire soil water repellency

Lewis, S.A.; Robichaud, P.R.; Frazier, B.E.; Wu, J.Q.; Laes, D.Y.M. 2008. Using hyperspectral imagery to predict post-wildfire soil water repellency. Geomorphology 95:192-205.

Keywords: Burn severity; Ash; Water repellent soils; Hayman Fire; Remote sensing; water repellency

Links: pdf PDF [1.09 MB]

Abstract: A principal task of evaluating large wildfires is to assess fire's effect on the soil in order to predict the potential watershed response. Two types of soil water repellency tests, the water drop penetration time (WDPT) test and the mini-disk infiltrometer (MDI) test, were performed after the Hayman Fire in Colorado, in the summer of 2002 to assess the infiltration potential of the soil. Remotely sensed hyperspectral imagery was also collected to map post-wildfire ground cover and soil condition. Detailed ground cover measurements were collected to validate the remotely sensed imagery and to examine the relationship between ground cover and soil water repellency. Percent ash cover measured on the ground was significantly correlated to WDPT (r=0.42; p-value<0.0001), and the MDI test (r=-0.37; p-value<0.0001). A Mixture Tuned Matched Filter (MTMF) spectral unmixing algorithm was applied to the hyperspectral imagery, which produced fractional cover maps of ash, soil, and scorched and green vegetation. The remotely sensed ash image had significant correlations to the water repellency tests, WDPT (r=0.24; p-value=0.001), and the MDI test (r=.0.21; p-value=0.005). An iterative threshold analysis was also applied to the ash and water repellency data to evaluate the relationship at increasingly higher levels of ash cover. Regression analysis between the means of grouped data: MDI time vs. ash cover data (R2=0.75) and vs. Ash MTMF scores (R2 =0.63) yielded significantly stronger relationships. From these results we found on-the-ground ash cover greater than 49% and remotely sensed ash cover greater than 33% to be indicative of strongly water repellent soils. Combining these results with geostatistical analyses indicated a spatial autocorrelation range of 15 to 40 m. Image pixels with high ash cover (>33%), including pixels within 15 m of these pixel patches, were used to create a likelihood map of soil water repellency. This map is a good indicator of areas where soil experienced severe fire effects--areas that likely have strong water repellent soil conditions and higher potential for post-fire erosion.

Moscow FSL publication no. 2008d