Date of Award
Level of Access Assigned by Author
Master of Science (MS)
Second Committee Member
Third Committee Member
The derivation of oceanographic and biological parameters from remote sensing is well documented across decades of research. Careful evaluation of satellite products provides insight into the optimal algorithms for image processing, research, and various biogeographical applications. Archived multi-satellite data from the United States Geological Survey offers users decades of continuously updated global data, and the agency has recently updated the Landsat portion of its catalog with Collection 2 files, which offers both Level 1 and Level 2 processed data products. Here, we evaluate the Collection 2 improvements using several published algorithms currently used to derive sea surface temperature, chlorophyll, and suspended particulate matter (SPM) across the coastal Gulf of Maine. Additionally, we used two publicly available level-2 processing softwares, Acolite and Polymer, and compared their resultant remotely sensed ocean color spectra, then used these product outputs as inputs for algorithms to estimate chlorophyll-ɑ and SPM. Algorithms tested include both standard empirical explicit algorithms as well as implicit algorithms produced by a neural network trained on United States coastal data. For sea surface temperature, the USGS processed temperature estimated with the Collection 2, Level 2 data was found to be significantly better than other temperature retrieval algorithms. When assessing the color bands for quality of spectra, Polymer significantly outperformed Acolite for all datasets in both Collections. C2 files provided slightly higher quality spectra; however, when comparing algorithms for chlorophyll-ɑ and SPM using these Level-2 processed spectra as input, the original Landsat Collection 1 did not significantly differ from Collection 2 when compared against validation data from buoys along the GOM. Chlorophyll-ɑ algorithms demonstrated significantly more range when using Acolite input data, especially when using machine learning-based implicit algorithms. For SPM, Polymer data generated the better overall product, but the empirical single-band algorithm(Nechad et al 2010) outperformed the neural network algorithm. Overall, Collection 2 offers similar quality or even improved input data for water-quality algorithms, and the Collection 2, Level-2 surface temperature product especially reduces researcher workload with a single-satellite source for high-quality thermal data. These results improve Landsat’s utility for long-term coastal mapping of water-quality parameters, and with the launch of Landsat-9 in 2021, it is recommended to use Collection-2 for quality consistency when considering the data archive as a whole.
Hesketh, Gabriel, "High Resolution Remote Sensing As a Tool To Improve Coastal Habitat Mapping in The Gulf of Maine" (2021). Electronic Theses and Dissertations. 3506.
Files over 10MB may be slow to open. For best results, right-click and select "save as..."