We've Moved!
Visit SDSU’s new digital collections website at https://digitalcollections.sdsu.edu
Description
In recent decades, climate change and other human-induced factors accelerated and increased wildfire seasons, frequency, and durations, especially in the semi-arid southwestern United States. This has significant implications on the change in the hydrological cycle, which consequently affects the ecosystem, biological processes, and downstream communities. The first part of this thesis investigates the effects and changes of evapotranspiration (ET) in forested regions that undergo wildfires. ET is an important process in the hydrological cycle as it transfers moisture from the surface to the atmosphere, which eventually becomes precipitation. Ground-based measurements can provide frequent observations, but often have smaller spatial extents. Additionally, ground-based equipment can be impacted by large wildfires, therefore this thesis aims to examine both pre- and post-fire evapotranspiration with an emphasis on utilizing remote sensing. Specifically, a spatial ET model is coupled with soil burn severity to investigate the change in ET temporally (annual and season) and spatially (watershed comparisons). Significant declines in ET resulted across various burn severities due to the transformation of vegetation types after the fire. The second part of this thesis proposes the novel utilization of machine learning to conduct supervised learning on spatial actual ET with remotely sensed indices and variables. Traditional spatial ET methods are time and resource intensive; machine learning can provide an alternative solution and minimize user errors and computational problems. A water balance application of a burned watershed is included to validate and demonstrate the potential benefit of this approach for land managers to assess the effect of wildfires on the hydrological cycle. This thesis work is the first to fully utilize remote sensing to examine the potential effects of ET due to large wildfires and to couple a machine learning method to develop SVM-ET6 to predict ET in a post-fire environment.
