Skip to main content
Author(s):
Eva K. Strand, Jessie M. Dodge
Year Published:

Cataloging Information

Topic(s):
Mapping
Fuels
Fuel Treatments & Effects
Suppression treatments
Risk
Risk assessment

NRFSN number: 21596
Record updated:

With the past century of fire suppression in ponderosa pine (Pinus ponderosa) forests, there has been an accumulation of surface fuels, causing decreases in understory vegetation and increasing high severity fire risk. However, fire size and location can make it costly and unsafe to obtain ground measurements of understory vegetation cover and fuels. Remotely assessing heterogeneity and ground cover components within a fire perimeter can contribute in monitoring ecological trends post-fire. Landsat TM images are free, have a spatial resolution of 30 m, and have been used to assess burn severity since 1984 whereas the QuickBird sensor has a high spatial resolution of 0.6 to 2.4 m, though it has fewer bands, images take up more space, and take longer to process. This study sought to compare remotely sensed attributes related to burn severity, derived from QuickBird and Landsat TM images, to determine which images correlated more strongly to ground measurements as well as determine if effects of pre-fire fuel reduction treatments could be observed better at finer scales. The 2007 Egley Fire Complex of eastern Oregon prompted the Malheur National Forest to obtain QuickBird coverage of the fire complex at 0.6 m resolution July 26th, August 8th and August 13th, 2007. In 2008, overstory tree canopy cover and understory cover of green vegetation, non-photosynthetic vegetation (NPV), rock, soil and char were measured in 70 field sites. Sites were broadly distributed across elevation, aspect, and the burn severity gradient, determined using a rapid response, Burn Area Remote Classification (BARC) map based on immediate post-fire dNBR values. Field data were collected at five plots distributed within each site. Spearman’s rank correlations were used to correlate the Normalized Differenced Vegetation Index (NDVI) derived from QuickBird fine-scale (0.6 m) and Landsat TM 2007 (August 22nd, 2007) and 2008 (July 7th, 2008) images to surface variable covers. To determine the effects of spatial scale, we used focal statistic means in ArcMap at varying window sizes (1 x 1, 3 x 3, 5 x 5, 8 x 8, 17 x 17, 25 x 25, 50 x 50, 83 x 83, 167 x 167, and 417 x 417 pixels) of QuickBird imagery. We found considerable variation of NDVI correlation to surface covers among plots, though NDVI correlations became stronger with tree canopy cover, understory green, soil, and NPV as the window size increased. Site level correlations between NDVI and surface variable covers were much stronger than plot level correlations but were fairly consistent in strength of correlations among scales. QuickBird and Landsat TM NDVI correlations to tree canopy cover, NPV, and soil at the plot and site level were much stronger in untreated sites than treated sites and were generally stronger at fine QuickBird scales (1x1 to 5x5 pixel windows) than broad QuickBird scales (50 x 50 through 417 x 417 windows), though were still similar to Landsat 2007 and 2008 correlation coefficients. This study demonstrates the importance of matching pixel size of remote sensing images to field data scale. We also found that NDVI extracted from Landsat images are remarkably similar to QuickBird data, both at fine and resampled to broad scale, in the Egley Fire Complex (2007).

Citation

Strand EK, and Dodge JM. 2020. Effects of scale for assessing fuel treatment effectiveness and recovery post-fire in ponderosa pine. Final Report for JFSP PROJECT ID: 18-1-01-37. Moscow: University of Idaho, 25p.

Access this Document