Fire Risk, Burn Severity, and Vegetation Recovery for the Berry Fire, Grand Teton National Park
<p>Wildfires considerably impact environments and communities around the world. Wildfires can alter ecosystems, damage property and infrastructure, and harm the wellbeing of at-risk populations. In the summer of 2016, several wildfires occurred in the Greater Yellowstone Ecosystem, one of which was named the Berry Fire. The Berry Fire consumed over 8,000 ha making it the largest fire on record for Grand Teton National Park. This study explores the fire ecology of the Berry Fire by modeling pre-fire risk, estimating field measured fire effects using Sentinel-2 imagery and assessing vegetation recovery via regression fitted spectral indices and multiple-endmember spectral mixture analysis (MESMA). For pre-fire risk assessment, multicriteria evaluation based on fuel type, canopy cover, relative moisture content, slope, elevation, aspect and distance to roads, trails and structures was implemented. The resulting risks were then compared to burn severity levels using logistic regression. The relationship between risk and burn severity was found to be generally weak, with only two burn severity categories (unburned and moderate to high) possessing moderately strong relationships to the pre-fire risk.</p> <p>The next analysis examined the ability for spectral indices to estimate field measured fire effects related to burn severity. Currently, most burn severity research attempts to associate spectral indices with the field measured composite burn index (CBI), however this approach is limited due to CBI being optically assessed and therefore subjective. Studies which have attempted to measure fire effects using Landsat imagery have not found strong correlations between tested spectral indices and fire effects. The recent availability of red-edge bands at a moderate spatial resolution, thanks to the launch of the Sentienl-2 constellation, allows for the calculation of spectral indices not available to the Landsat satellites. Using all-possible-models multivariate regression, a total of thirty different spectral indices were calculated and compared to field measured fire effects collected by Turner et al. 2019. Six of the fire effects possessed models that possessed coefficients of determination and variance inflation factors that passed the criteria for a suitable model. The best models for each of these fire effects were then further explored. All six of these models included red-edge indices based on Sentinel-2 band five, which strengthens other research findings indicating the usefulness of this band for burn severity assessments.</p> <p>Finally, vegetation recovery was assessed using fractional vegetation cover (FVC) derived from a combination of field plots, regression fitted spectral indices and MESMA. A total of sixty field plots were collected in the summer of 2019 in each of which eight downward and eight upward hemispherical photographs were taken. The FVC was then calculated for each photograph belonging to a plot within CAN-EYE and the average FVC was calculated. Thirty-one of these plots were then used to derive the regression fits for the spectral indices, which were implemented using raster algebra. The resulting regression fit values were then compared to the remaining plots via linear regression to determine how accurately FVC was mapped. The MESMA, derived using three forest and three herbaceous endmembers, was compared to all sixty plots using linear regression. The results indicate that a red-edge index (NDVIre1n) outperformed all other methods for estimating FVC.</p> <p>The objective of this research was to examine three temporal stages of fire ecology for the Berry Fire: pre-fire risk, immediate post-fire fire effects (burn severity) and vegetation recovery three years post-fire. This was accomplished using geographic information systems and remote sensing analyses based on current research trends and newly available datasets. As each of these dynamics influence and impact the following dynamic, they combine to create a holistic view of the conditions and effects of the Berry Fire.</p>
Remote sensing, Fire ecology, GIScience
Szpakowski, D. M. (2020). <i>Fire risk, burn severity, and vegetation recovery for The Berry Fire, Grand Teton National Park</i> (Unpublished dissertation). Texas State University, San Marcos, Texas.