Chi-square spectral fitting for concentration retrieval, automatic local calibration, quality control, and water type detection

Abstract

In this study, the inverse bio-optical model HYDROPT was calibrated with regional specific inherent optical properties (SIOPs) and various local SIOPs to examine the effect of these calibrations on the retrievals. The study area, the Wadden Sea, is an estuary and tidal flat area with very high concentrations of chlorophyll a (Chl a), suspended particulate matter (SPM), and coloured dissolved organic matter (CDOM). HYDROPT could derive concentrations of Chl a, SPM, and CDOM with a reasonable degree of accuracy when in situ above-water reflectances were used as input (root mean squared error of 0.19–0.52 mg·m−3 for Chl a, 0.28–0.46 mg·m−3 for SPM, and 0.20–0.34 m−1 for aCDOM). However, quality control showed that 70% of the in situ input reflectance spectra were ambiguous; these spectra could be modelled with various sets of SIOPs. Therefore, automatic local calibration based on the spectral fit (χ2) value of the fitting procedure did not necessarily lead to the best results; this was expected to be an advantage of χ2 fitting. When MERIS data were used as input, the concentration maps showed distributions according to the expectations, although tidal flats and nearby land affected the results at the locations that matched with in situ stations. The water types (water with similar SIOPs) that could be detected based on MERIS data were new in this study.