Twenty year (1984-2004) temporal and spatial burn severity patterns inferred from satellite imagery in the Gila National Forest, New Mexico


Holden, Zachary Alan.. (2008). Twenty year (1984-2004) temporal and spatial burn severity patterns inferred from satellite imagery in the Gila National Forest, New Mexico. Theses and Dissertations Collection, University of Idaho Library Digital Collections.

Twenty year (1984-2004) temporal and spatial burn severity patterns inferred from satellite imagery in the Gila National Forest, New Mexico
Holden, Zachary Alan.
Forest fires--New Mexico--Gila National Forest Remote sensing--New Mexico--Gila National Forest
Natural Resources
Recent increasing trends in fire extent have been documented, yet little is known about how climate, vegetation and topography influence the patterns of burn severity (defined here as the magnitude of vegetation change one year post-fire relative to pre-fire conditions) of those fires. Here, I use satellite-derived burn severity data to infer 20-year patterns of burn severity relative to topography and climate. A time series of Landsat Thematic Mapper (TM) satellite images were used to map 114 fires (195,600 hectares burned) on the Gila National Forest from 1984-2004. Burn severity of each fire was inferred from the Relative Differenced Normalized Burn Ratio (RdNBR), a derivative of the differenced Normalized Burn Ratio. Data from nearby weather and Snowpack Telemetry (SNOTEL) stations were used to evaluate the influence of Snow Water Equivalent (SWE) and precipitation patterns on severe fire occurrence. Vegetation and Digital Elevation Model-derived Geographic Information System (GIS) layers were used to analyze the spatial patterns of severe fire occurrence on the 1.4 million-hectare Gila National Forest.;Severe fire occurred more frequently at high elevations, in mesic spruce-fir and mixed-conifer vegetation types, on north-facing slopes and where solar radiation and heat load index values were low. Within drier Potential Vegetation Types, severe fire occurred more frequently where moisture was more available. However, this pattern shifts at higher elevations, where areas with high heat load indexes and exposed south-facing slopes increased the probability of severe fire occurrence during this twenty-year period. Random Forest predictions of severe fire occurrence using topographic variables as predictors yielded classification accuracies of 82% and 63% for two (high severity vs. other) and three (low, moderate, high severity) class burn severity grids.;Spring precipitation, SWE and precipitation-free periods during the fire season (April-July) were significantly related to area burned and area burned severely, with the length of dry periods explaining most of the variation in fire extent and severity. These precipitation metrics were strongly correlated with 17-year patterns of spring and early summer vegetation green-up inferred from the Advanced Very High Resolution Radiometer (AVHRR).;Spectral indices used in this study were derived from the Landsat TM sensor. However the life of this sensor may be limited and other remotely sensed data on burn severity patterns will likely be sought in the future. Using pre and post-fire images from 4 different satellite sensors with varying spatial and spectral resolutions (Quickbird, the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)), Landsat TM and the Moderate Resolution Imaging Spectroradiometer (MODIS) correlations between ground-based Composite Burn Index (CBI) plots and satellite-derived indices were compared. ASTER and Quickbird-derived indices performed as well or better than the Landsat-derived dNBR.
Thesis (Ph. D., Natural Resources)--University of Idaho, January 2008.
Major Professor:
Penelope Morgan.
Defense Date:
January 2008.
Format Original:
x, 149 leaves :ill., maps ;29 cm.

Contact us about this record

In Copyright - Educational Use Permitted. For more information, please contact University of Idaho Library Special Collections and Archives Department at
Standardized Rights: