Abstract
Chlorophyll-a (Chl-a) concentration, a crucial indicator of phytoplankton biomass, is sensitive to seasonality. The variations in trophic states regarding seasonality and the changes of spectral properties of water bodies pose uncertainties to the accuracy of remote sensing semiempirical models. In particular, lakes in subtropical regions generally experience different trophic states in dry and wet seasons. In this study, a season-insensitive Chl-a retrieval model using multitask convolution neural network with multiple output layers (MCNN) is proposed. A layer-sharing network combined with data augmentation is adopted to alleviate the issue of insufficient quantity of in situ samples. In addition, a hyperparameter optimization is performed to automatically refine the MCNN architecture. To evaluate the accuracy of proposed method, Laguna Lake, one of the largest lakes in Southeast Asia, is selected as the validation target. The lake is characterized by oligotrophic and mesotrophic states in wet season, whereas the states change to mesotrophic and low-level eutrophic states in dry season. A collection of Sentinel-3 Ocean and Land Colour Instrument Level-2 images and 409 in situ samples with the Chl-a concentration range 1.24–22.30 mg $\cdot$ m $^{-3}$ were used for model calibration and evaluation. Experimental results showed that MCNN with the performance of average $\boldsymbol{R^{2}}$ = 0.74, RMSE = 2.06 mg $\cdot$ m $^{-3}$ , Pearson's $\boldsymbol {r}$ = 0.86 outperforms related semiempirical models, including normalized difference chlorophyll index, two-band and three-band models, and WaterNet. The Chl-a prediction accuracy was improved by 7.19–14.6%, in terms of RMSE, compared with WaterNet.
Highlights
E UTROPHICATION, or nutrient overenrichment, becomes a major concern for global inland waters [1], [2]
There is a positive correlation between eutrophication and abundant phytoplankton, and abundant phytoplankton is associated with elevated concentrations of chlorophyll-a (Chl-a) [4]
The developments of Chl-a-retrieval algorithms for turbid water are based on the ratios and combination of red-to-near infrared (NIRR)
Summary
E UTROPHICATION, or nutrient overenrichment, becomes a major concern for global inland waters [1], [2]. The optical property is dominated by phytoplankton, and Chl-a concentrations can be estimated by using blue–green ratios of remote sensing reflectance Rrs(λ). The blue–green ratios are unreliable for the Chl-a retrieval of turbid water, such as inland and parts of coastal water bodies. The developments of Chl-a-retrieval algorithms for turbid water are based on the ratios and combination of red-to-near infrared (NIRR). Gitelson et al [6] evaluated the applicability of two-band model (λ1 = 662–672 nm and λ2 = 743–753 nm) using moderate resolution imaging spectroradiometer (MODIS) spectral bands and three-band model (λ1 = 660–670 nm, λ2 = 703.75–713.75 nm, and λ3 = 750–757.5 nm) using medium resolution imaging spectrometer (MERIS) spectral bands for the water bodies with Chl-a concentrations ranging from 1.2–236 mg·m−3. The method was evaluated on water bodies with Chl-a concentrations ranging from 0.9–28.1 mg·m−3, and the coefficient of determination (R2) and root-mean-square
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
