This work evaluated the improvement to the accuracy of chlorophyll-a (chl-a) estimating algorithms derived from Sea-Viewing Wide Field-of-View Sensor (SeaWIFS) archives of an optically-shallow, subtropical bay. Preliminary investigation into the in situ chl-a measurements showed that the fine spatial and temporal resolution currently only available through satellite remote sensing are required to adequately understand the dynamics of coastal chl-a. The in situ datasets, however, were found to be useful for developing chl-a algorithms by allowing for 1) identification of appropriate times of year for classifying benthic habitats and 2) the assumption of annually invariable bottom reflectance. Benthic type-specific algorithms were developed where benthic class was established through image-based supervised classification of Landsat images of the study area. The overall accuracy of the classifier, using available field data, was 67% and 76% for the two validation years. Although improvement to the accuracy of satellite-retrieved chl-a was demonstrated, the accuracy of the improved chl-a estimates remained low. Algorithms tuned to the sparse-low seagrass bottom (r2 = 0.234, mean absolute percent difference (APD) = 71%) performed better than those associated with medium-dense seagrass (r2 = 0.332, mean APD = 66%). The positive bias produced by the operational SeaWiFS chl-a algorithm was removed through the regionally-tuned algorithms but the residuals of the medium-dense seagrass chl-a did suggest a seasonality in the bias of the improved estimates. The accessibility of the studied methodology, in terms of equipment, software and expertise required, and the lack of research into the SeaWiFS archive for multi-temporal analyses of coastal dynamics support continued development of the novel methodology. Atmospheric correction procedures derived specifically for normalizing surface reflectances across images are likely to improve the transferability of image-based classifiers as well as the performance of empirical chl-a algorithms. Testing the transferability of image-based optical signatures in space to other study areas is an important next step for this methodology. A well-defined spectral library of image-based classes would improve assessment of global chl-a dynamics, which is especially important given global climate change.