Water transparency represented by the Secchi disk depth (Zsd) plays an important role in understanding water ecology environment variations, especially for optically complex and shallow lake waters. In this study, using in situ measured remote sensing reflectance (Rrs), diffuse attenuation coefficient (Kd), and Zsd data collected in Lake Taihu (China), a regional algorithm for estimating Kd from Rrs was designed, and the semi-analytical model proposed by Lee et al. (2015) (hereafter called Lee_2015 model) was refined using a linear scaling correction for remote sensing of Zsd. The results showed that a good agreement between the derived Kd and in situ measured data (mean absolute percentage error (MAPE) = 26% for Kd(490); MAPE < 5% for Kd at 443, 555, and 660 nm). The in situ Rrs-derived Zsd results using the refined Lee_2015 model compared well with the in situ measured Zsd (R2 = 0.72 and MAPE = 36%), which was an obvious improvement over the Lee_2015 model in our study region. Subsequently, the refined Lee_2015 model was applied to the geostationary ocean color imager (GOCI) observations between 2012 and 2018 to yield the spatial and temporal variations of water transparency in the Lake Taihu waters. The long-term mean distribution of Zsd revealed that water transparency values in the northeastern Lake Taihu were generally higher than those in the southwest part. Monthly climatological Zsd patterns suggested that the Zsd distributions had large temporal variability, and distinct monthly patterns of Zsd existed in different subregions of Lake Taihu. The significant interannual variations of Zsd in Lake Taihu are probably affected by a combination of the water column stability mainly caused by wind, water temperature, human activity, and riverine discharge. The present study can provide a new approach for quantifying water visibility and serve for water-color remote sensing of optically complex and highly turbid waters.
Remote sensing of water transparency variability in the Ibitinga reservoir during COVID-19 lockdown
