Remote sensing of chlorophyll a concentration in turbid coastal waters based on a global optical water classification system

Accurate chlorophyll a concentration (Chla) retrieval in coastal waters from ocean color remote sensing faces challenges due to the significant optical complexity compared to clear oceanic waters. In this paper, a novel technique for Chla retrieval in turbid coastal waters was proposed and tested in the Bohai Sea based on a global optical water classification system. Firstly, the in situ measured spectral remote sensing reflectance spectra (R-rs (lambda)) (n = 559) were classified into different optical water types. Secondly, optimal algorithms were identified with newly tuned model parameters for each water type to achieve accurate Chla retrieval. We found that (1) among all the 23 optical water types (1-23), the water types of 9-22 are present in the Bohai Sea, indicative of moderate or high turbidity, while other water types corresponding to clear oceanic waters are generally absent from our data. (2) Clear spatio-temporal patterns of the water types are revealed. Three basins of the Bohai Sea (the Bohai Bay, Laizhou Bay and Liaodong Bay) receiving tremendous terrestrial inputs and with high turbidity are dominated by the optical water types of 15-22. Water types 9-14 that are less turbid, are mainly distributed in the Bohai Strait and the central Bohai Sea, which are far from the coast (and thus with less terrestrial influences) and relatively deep (and thus with less bottom suspension influences), compared to the 3 basins. (3) Through the identification of the optimal retrieval algorithm and parameter tuning for each water type, the uncertainty of chlorophyll a retrievals has been reduced from 54% (root mean square error of 2.76 mg/m(3)) to about 36% (1.84 mg/m(3)). Independent validation with the in situ-satellite match-ups further demonstrates the algorithm's validity (uncertainty of about 35%).The global optical classification system together with the optimal retrieval algorithm for each optical class, is proved to be a feasible way for ocean color retrieval in high accuracy over optically complex waters.