Water-surface chlorophyll detection by remote sensing and vertical structure of chlorophyll analysis in Lake Superior: Water-surface chlorophyll as an estimate of water-column-integrated chlorophyll

This research, consisting of three parts, investigates the application of remote sensing to both the detection of water-surface chlorophyll a concentration ([Chl-a]) and the estimation of water-column-integrated [Chl-a] in Lake Superior. In Chapter 1, we describe four methods for determining remote sensing reflectance ( Rrs), using a combination of above-surface, below-surface, and polarization measurements of radiance, in western Lake Superior during the summer of 2003. The work allows us to examine in detail the effects of the unwanted surface reflectance from a rough water surface. The results of Chapter 1 provide consistent values of Rrs and indicate that the four methods are practical and helpful in determining Rrs reliably. In Chapter 2, we evaluate two standard empirical [Chl-a] retrieval algorithms: one for Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and one for Moderate Resolution Imaging Spectroradiometer (MODIS). We compare the results from satellite data with the corresponding in-situ data collected in western Lake Superior in the summer of 2004. The linear correlations between the empirically-derived [Chl-a] and the in-situ measurements are poor (R2 around 0.1) for both algorithms. We find that more detailed information on particulate backscattering is required to test and derive a regionally optimized semi-analytical algorithm. To overcome the problems associated with retrieving water-surface [Chl-a] from the empirical and the semi-analytical algorithms, we apply artificial neural networks (ANNs). The results appear to provide better estimates of [Chl-a] than the empirical algorithms. In Chapter 3, we discuss the difference between the [Chl- a] profiles collected in offshore and inshore areas of western and southeastern Lake Superior and explore the possibility of applying a Gaussian distribution to the modeling for the offshore [Chl-a] profiles. We establish statistical relationships between the water-surface [Chl- a] and its vertical distribution characteristics: the deep chlorophyll maximum (DCM), the DCM depth (zmax), and the depth-integrated [Chl-a]. ANNs are used to relate the [Chl-a], the temperature, and the vertical attenuation coefficient (K d) near water surface to the depth-integrated [Chl- a]. Using these two kinds of inverse models, we determine the depth-integrated [Chl-a] distribution for western Lake Superior.