Deriving inherent optical property for highly turbid productive inland water from MERIS data by semi-analytical model: A case study in Taihu Lake, China

Abstract

Listen

Translate article

Remote estimation of inherent optical properties was greatly challenged by significant spatial-temporal variation and the extreme complexity of bio-optical properties in inland turbid water. The multiband quasi-analytical algorithm has advantages over traditional band ratio and semi-analytical algorithm in that it is based on the remote sensing reflectance model derived from radiative transfer equation and does not need the parameterization of absorption coefficients. An improved model, which used, was developed to retrieve inherent optical properties in high turbid inland water. As a first step, the backscattering coefficient at reference wavelength [bbp(λ0)] was retrieved directly by support vector machine optimization algorithm instead of step 2 in the quasi analytical algorithm for the high correlation between bbp(λ0) and remote sensing reflectance at the near-infrared wavelength. The second step, a semi-analytical support vector machine algorithm, was used to retrieve spectral shape of bbp(λ) instead the step 4 in the quasi analytical algorithm. Part of field-measured dataset collected on November 2006, November 2007, November 2008 and April 2009 in Taihu Lake was used to train the support vector machine model, and the other part was used to test this algorithm. Results indicated that the mean square root of percentage between the derived and measured value of bbp(532 nm) was less than 3.73% and root mean square percentage of ap(442 nm) and ap(532 nm) were 15.29% and 30.45%, respectively. Furthermore, the potential application of this algorithm to MERIS data was investigated by the reduced resolution MERIS satellite image. The result shows that satellite-derived data using the support vector machine model is consistent with in situ measured data. This study advances the semi-analytical model and broadens the application of MERIS data in highly turbid inland waters.