Assessment of chlorophyll-a concentration derived from Sentinel-3 satellite images using open source data

The concentration of chlorophyll-a is considered a very important water quality parameter due to the role it plays in the eutrophication. It can be estimated by remote sensing using empirical methods and/or semi-analytical methods. The main aim of this paper is to assess the chlorophyll-a concentration, derived from the Sentinel-3 Satellite, with in situ measurements covering the Mediterranean Sea. The Sentinel-3 Ocean and Land Colour Instrument (OLCI) was utilized. Two algorithms examined on their efficiency: i) the OC4Me Maximum Band Ratio algorithm, a polynomial algorithm based on the use of a semi-analytical model which uses a maximum band ratio approach of reflectances at 443, 490 and 510 nm, over the 560 nm and ii) a neural net (NN) algorithm, that uses an Inverse Radiative Transfer Model to estimate the water constitutes and estimate the chlorophyll-a concentration. A dedicated data set from the Copernicus Marine Environmental Service (CMES) with in situ chlorophyll-a concentrations was utilized. The parameters of interest were extracted and chlorophyll-a values at different depths were extracted. Also, to assure the accuracy of the in situ measurements the quality control parameters provided by the marine Copernicus were applied. The concentration of chlorophyll-a (CM) at a penetration depth (Zpd) was calculated from the in situ dataset. Then, a comparison was performed using the two algorithms against the in situ data. Statistical indexes were calculated to illustrate the correlation between the two algorithms and the in situ measurements. No significant correlation was observed for OC4Me algorithm. However, examining the time difference among the in situ data and the satellite acquisition the best correlation is observed between a time difference of two hours from the satellite and the in situ dataset. Last but not least, no correlation was observed between the chlorophyll-a calculated from the neural nets and the in situ dataset.