Date of Award
Program or Major
Master of Science
The timing and magnitude of spring snowmelt events impact riverine flooding and inform reservoir operations. This study evaluates the ability of the Diurnal Amplitude Variation (DAV), Frequency Difference (FD) and Polarization Ratio (PR) melt onset detection algorithms to determine melt onset dates (MOD) in the mid-latitudes of the United States. The methods are evaluated using satellite remotely sensed passive microwave observations from the Advanced Microwave Scanning Radiometer – EOS (AMSR-E) sensor and compare against in situ snow measurements from 763 Snow Telemetry (SNOTEL) and Soil Climate Analysis Network (SCAN) stations. The DAV method performs best in Alaska, predicting the MOD with a mean absolute error of 9.4 days, while the Frequency Difference and Polarization Ratio methods produce mean absolute errors of 12.5 and 11.9 days, respectively. The DAV method also clearly produced the best results in Vermont, the FD method worked best in South Dakota and the PR method performed best in Arizona. None of the study’s methods are recommended for California, Minnesota, Oregon and Washington. The remaining states did not have an algorithm that worked notably better than the others and it was discovered that the methods do not work for a shallow snowpack. Tree cover was also found to have little effect on the performance of the melt onset detection methods for pixels having less than 50% tree cover.
Osborne, Douglas, "Evaluating Melt Onset Date in the United States using Remotely Sensed Passive Microwave Derived Brightness Temperature" (2016). Master's Theses and Capstones. 865.