Remote estimation of chlorophyll concentration in hyper-eutrophic aquatic systems: Model tuning and accuracy optimization

Abstract

Accurate assessment of phytoplankton chlorophyll a (chl a) concentration by remote sensing is challenging in turbid hyper-eutrophic waters. This paper assessed methods to resolve this problem. A hand-held spectroradiometer was used to measure subsurface spectral reflectance ( R) in the visible and near infrared range of the spectrum. Water samples were collected concurrently and contained variable chlorophyll a concentration (chl a from 107 to more than 3000 mg/m 3) and turbidity (from 11 to 423 NTU) levels. The conceptual three-band model [ R − 1 ( λ 1) − R − 1 ( λ 2)] × R( λ 3) and its special case, the two-band model R( λ 3)/ R( λ 1), were spectrally tuned in accord with optical properties of the media to optimize spectral bands ( λ 1, λ 2 and λ 3) for accurate chlorophyll a estimation. Strong linear relationships were established between analytically measured chl a and both the three-band [ R − 1 (650) − R − 1 (710)] × R(740) and the reflectance ratio model R(714)/ R(650). The three-band model accounted for 7% more variation of chl a concentration than the ratio model (78 vs. 71%). Assessment of the model accuracy in dense algal blooms is hampered by the spatial and temporal inhomogeneity of algal distributions—in these waters, non-random algal distributions accounted for more than 20% spatial and up to 8% temporal variation in chlorophyll a concentration. The findings underlined the rationale behind the conceptual model and demonstrated the robustness of the algorithm for chl a retrieval in very turbid, hyper-eutrophic waters.