Estimating pre-fire biomass for the 2013 California Rim Fire using airborne LiDAR and Landsat data

Wednesday, 17 December 2014
Mariano Garcia-Alonso1, Ángeles Casas Planes1, Alexander Koltunov1,2, Susan Ustin1, Matthias Falk1 and Carlos Ramirez2, (1)University of California Davis, Davis, CA, United States, (2)USDA Forest Service, Mcclellan Afb, CA, United States
Accurate knowledge of the amount and distribution of fuels is critical for appropriate fire planning and management, but also to improve carbon emissions estimates resulting from both wildland and prescribed fires. Airborne LiDAR (ALS) data has shown great capability to determine the amount of biomass in different ecosystems. Nevertheless, for most incidents a pre-fire LiDAR dataset that would enable the characterization of fuels before the incident is not available. Addressing this problem, we investigated the potential of combining a post-fire ALS dataset and a pre-fire Landsat image to model the pre-fire biomass distribution for the third-largest wildfire in California history, the Rim fire. Very high density (≈ 20 points/m2) ALS data was acquired covering the burned area plus a 2 km buffer. 500+ ALS-plots were located throughout the buffer area using a stratified random sampling scheme, with the strata defined by species group (coniferous, hardwood, and mixed forests) and diametric classes (5-9.9”; 10-19.9”; 20-29.9” and >30”). In these plots, individual tree crowns were delineated by the Watershed algorithm. Crown delineation was visually refined to avoid over- and under-segmentation errors, and the tree biomass was determined based on species-specific allometric equations. The biomass estimates for correctly delineated trees were subsequently aggregated to the plot-level. The next step is to derive a model relating the plot-level biomass to plot-level ALS-derived height and intensity metrics as explanatory variables. This model will be used to map pre-fire biomass in the buffer area outside the burn. To determine pre-fire biomass inside the fire perimeter, where ALS data are not available, we will use a statistical approach based on spectral information provided by a pre-fire Landsat image and its relationships with the 2 km buffer LiDAR-derived biomass estimates. We will validate our results with field measurements collected independently, before the fire.