This dissertation is founded on the importance of phytoplankton community composition to marine biogeochemistry and ecosystem processes and motivated by the need to understand their distributions on regional to global scales. The ultimate goal was to predict surface phytoplankton communities using satellite remote sensing by relating marine habitats--defined through a statistical description of environmental properties--to different phytoplankton communities. While phytoplankton community composition is governed by the interplay of abiotic and biotic interactions, the strategy adopted here was to focus on the physical abiotic factors. This allowed for the detection of habitats from ocean satellites based on abiotic factors that were linked to associated phytoplankton communities.; The research entailed three studies that addressed different aspects of the main goal using a dataset collected in the western Gulf of Maine over a 3-year period. The first study evaluated a chemotaxonomic method that quantified phytoplankton composition from pigment data. This enabled the characterization of three phytoplankton communities, which were defined by the relative abundance of diatoms and flagellates. The second study examined the cycles of these communities along with environmental variables, and the results revealed that the three phytoplankton communities exhibited an affinity to different hydrographic regimes. The third study focused on the implementation of a classifier that predicted phytoplankton communities from environmental variables. Its ability to differentiate communities dominated by diatoms versus flagellates was shown to be high. However, the increase in data imprecision when using satellite data led to lowered performance and favored an approach that incorporated fuzzy logic. The fuzzy method is well suited to characterize the uncertainties in phytoplankton community prediction, and provides a measure of confidence on predicted communities. The final product of the overall dissertation was a time series of maps generated from satellite observations depicting the likelihood of three phytoplankton communities.; This dissertation reached the main goal and, moreover, demonstrated that improvements in the predictive power of the method can be achieved with increased precision and more advanced satellite-derived products. The results of this research can benefit present bio-optical and primary productivity models, and ecosystem-based models of the marine environment.
展开▼
