Date of Award

3-2014

Degree Name

MS in Forestry Sciences

Department/Program

Natural Resources Management

College

College of Agriculture, Food, and Environmental Sciences

Advisor

Brian Dietterick

Advisor Department

Natural Resources Management

Advisor College

College of Agriculture, Food, and Environmental Sciences

Abstract

Swanton Pacific Ranch is an approximately 1,300 ha working ranch and forest in northern Santa Cruz County, California, managed by California Polytechnic State University, San Luis Obispo (Cal Poly). On August 12, 2009, the Lockheed Fire burned 300 ha of forestland, 51% of the forested area on the property, with variable fire intensity and mortality. This study used existing inventory data from 47 permanent 0.08 ha (1/5 ac) plots to compare the accuracy of classifying mortality resulting from the fire using digital multispectral imagery and LiDAR. The percent mortality of trees at least 25.4 cm (10”) DBH was aggregated to three classes (0-25, 25-50, and 50-100%). Three separate Classification Analysis and Regression Tree (CART) models were created to classify plot mortality. The first used the best imagery predictor variable of those considered, the Normalized Difference Vegetation Index (NDVI) calculated from 2010 National Agricultural Imagery Program (NAIP) aerial imagery, with shadowed pixel values adjusted, and non-canopy pixels removed. The second used the same NDVI in combination with selected variables from post-fire LiDAR data collected in 2010. The third used the same NDVI in combination with selected variables from differenced LiDAR data calculated using post-fire LiDAR and pre-fire LiDAR collected in 2008. The imagery alone was 74% accurate; the imagery and post-fire LiDAR model was 85% accurate, while the imagery and differenced LiDAR model was 83% accurate. These findings indicate that remote sensing data can accurately estimate post-fire mortality, and that the addition of LiDAR data to imagery may yield only modest improvement.

Share

COinS