Two-stream remote sensing model for water quality mapping : 2SeaColor

A fundamental measure to characterize water quality is the attenuation coefficient for downward planar irradiance in the water column. In this paper we present an analytical forward model and develop an inversion scheme (2SeaColor) to retrieve the downwelling attenuation coefficients from remote sensing data. The forward formulation of 2SeaColor provides an analytical and stable solution of the direct and diffuse components of the radiative transfer equation. The common high turbidity of inland waters is accounted for in 2SeaColor by projecting its effect on the inherent optical properties (IOPs) using the similarity transform. The inversion scheme of 2SeaColor uses a novel approach to parameterize the spectral dependency of IOPs. In this parameterization 2SeaColor uses the curvature information of observed remote sensing spectra to deduce the spectral-shape of the total absorption coefficient.

The forward model is verified using simulated data from the International Ocean Colour Coordinating Group (IOCCG data) and is compared with two existing models.

Statistical analysis results of model verification show that the forward model is able to derive Kd values with 3% accuracy (relative mean absolute error). Our solution provides an improvement on the accuracy of the estimated Kd when compared to other models, and especially at larger depths. The inversion scheme is validated against three sets of data: IOCCG, field measured and satellite observed matchups. Field measurements were collected in the turbid Naivasha Lake, Kenya. The mean absolute relative error in the derived Kd in all data sets did not exceed 15%, with the satellite matchup set resulting in the largest discrepancy.

The developed model, 2SeaColor, is stable to noise up to 15% and consistently produces accurate results which make it an ideal candidate model for the future Ocean and Land Colour Instrument (OLCI) on board the Sentinel-3 mission.