Modeling Remote-Sensing Reflectance and Retrieving Chlorophyll-a Concentration in Extremely Turbid Case-2 Waters (Lake Taihu, China)

Accurate assessment of concentration of chlorophyll a (Chla) and correct identification of algal blooms by remote sensing have previously been a great challenge in the optically complex Case-2 waters. In this paper, we used a large biooptical data set to model the remote-sensing reflectance in an extremely turbid and biologically productive Lake Taihu in China. The conceptual three-band model [R_rs ^-1 (lambda₁) - R_rs ^-1 (lambda₂₎] times R_rs ( lambda₃) (where R_rs represents remote-sensing reflectance just above the water surface) to retrieve Chla concentration was calibrated and validated, and a detailed assessment of its accuracy was obtained. Water samples were collected for four seasons from 2006 to 2007 at 50 sites, covering different ecosystem types, and contained three very variable optically active substances (tripton 7.9-281.7 mg middot L^-1, Chla 4.0-448.9 mug middot L^-1, and chromophoric dissolved organic matter [a_CDOM(440)] 0.27-2.36 m^-1). Secchi disk transparency ranged from 8 to 85 cm. The retrieval accuracies (r²) of the optimal three-band model and the related band-ratio method were 0.94 and 0.92, while the root mean-square errors (RMSE) and relative errors (RE) were 15.1 mug middot L^-1 (37.3% accounting for the mean value) and 18.0 mug middot L^-1, and 44.4% and 60.2%, respectively. Applications of the three-band model using MERIS central bands [R_rs ^-1(681) - R_rs ^-1(709)] times R_rs(754) also allowed accurate estimation of Chla, with r², RMSE, and RE of 0.92, 17.0 mug middot L^-1, and 48.1%, respectively. The establishment of a simple and robust biooptical model with high retrieval accuracy and known error budgets will help the rapid, accurate, and real-time assessment of algal blooms using in situ- - and satellite remote-sensing techniques.