Remote sensing reflectance reconstruction to obtain water optical properties from MERIS multispectral satellite images

Abstract

We use a multiple regression analysis and a data bank of about 400 reflectance spectra to reconstruct hyperspectral reflectances between 400 and 900 nm with a 5 nm step using only the values known at the wavelengths of the MERIS sensor level 2 data. For in situ remote sensing reflectances measured during different oceanographic missions, the reconstruction is within 2 per cent almost over the entire spectrum. The main difference (to a maximum of 4 per cent) usually occurs at the inflexion point of the reflectance curve between 580 and 600 nm. Observed in-situ remote sensing reflectances or reconstructed spectra are inverted using a Water Colour Simulator bio-optical model (WASI) to obtain the inherent optical properties (IOP) of the water. The values derived by the model are compared with the measurements when available. To validate the reconstruction, we compare the results of the model inversion using the initial spectrum or the reconstructed one as input. Preliminary results show that the derived values from the inversion of the reconstructed spectrum are very close to the values generated from the inversion of the initial spectrum, especially in case 1 waters. This reconstruction technique is used to generate hyperspectral remote sensing reflectances from reflectance data calculated by the MERIS sensor. We use the reconstructed spectra as input to be inverted in the WASI model in order to quantify the substances' concentrations; in particular, the inversion is working well for the suspended particulate matter concentrations.