Marine Science Faculty Publications

Influence of a Red Band-based Water Classification Approach on Chlorophyll Algorithms for Optically Complex Estuaries

Document Type


Publication Date



Chlorophyll algorithms, NTD675, Band indices, MODIS, Optically complex estuaries

Digital Object Identifier (DOI)


Empirical chlorophyll algorithms depend on the inter-relationships of the various in-water optical constituents which may change in both space and time in optically complex estuaries. Thus, algorithms developed for one case may not be applicable for others due to their different requirements on the model forms, wavebands, or model parameters. Here, using both in situ and satellite data collected from two estuaries in the United States, namely Tampa Bay and Chesapeake Bay, we attempted to address this technical challenge through water type classification and type-specific algorithm tuning. Based on an existing NTD675 (Normalized Trough Depth at 675 nm) water classification method that was originally devised for inland turbid waters, the estuarine waters were divided into three water types, and then used to develop type-specific empirical algorithms using different bands (2-band, 3-band, and 4-band algorithms). The 2-band model outperformed the other models, whose band selection varied with water types. Application of the type-specific models to the Moderate Resolution Imaging Spectroradiometer (MODIS) data showed uncertainties of 32.5% in mean absolute percentage error (MAPE) and 4.7 mg m− 3 in root mean square error (RMSE) for chlorophyll a concentration (Chla) ranging between 2 and 20 mg m− 3. As a comparison, a unified model tuned for all water types without classification showed inferior performance with a MAPE of 50.5% and a RMSE of 5.9 mg m− 3. By means of the in situ data, the type-specific models for the classified waters yielded a general improvement in MAPE by 8.4%, in comparison with the model for the non-classified water. Application to satellite imagery also showed reasonable spatial distribution patterns of Chla in Tampa Bay and Chesapeake Bay. These results suggest that the type-specific models coupled with the NTD675 water classification have the potential for improving Chla retrievals over optically complex waters, while general applicability of the approach to other estuaries still requires further research.

Was this content written or created while at USF?


Citation / Publisher Attribution

Remote Sensing of Environment, v. 155, p. 289-302