Regional objective analysis for merging MERIS, MODIS/Aqua and SeaWiFS chlorophyll-a data from 1998 to 2008 on the European Atlantic Shelf at a resolution of 1.1Km

In this paper we define the method used to merge high resolution multi-sensor chlorophyll-a data on the Ireland-Biscay-Iberia Regional Ocean Observing System (IBI-ROOS) area at a resolution of 0.015deg. The method is based on geostatistics and is known as kriging. The merged variable is the anomaly of chlorophyll-a, the anomaly being defined for a day as the difference between the daily image and the mean historical field for that day. The used chlorophyll-a dataset is derived from the daily level-2 water leaving radiances of the Sea-viewing Wide Field of View Sensor (SeaWiFS) of the Orbview platform, the Moderate Resolution Imaging Spectroradiometer (MODIS) of the Aqua platform and the Medium Resolution Imaging Spectrometer Instrument (MERIS) of the ENVISAR platform are obtained using specific algorithm developed by Ifremer, known as OC5 products. Before merging, each satellite-derived chlorophyll-a data have been compared to
in situ
data and validated using a matchup dataset. After this validation against
in situ
data, inter-comparisons between the satellite datasets have been performed. As the chlorophyll-a anomaly is not stationary on the whole area, local space-time semi-variograms have been calculated. These semi-variograms are defined by their nugget effect (noise), range (maximum distance for a non null covariance between the anomalies) and sill (maximum variance). The ranges of the semi-variograms have been approximated using local estimations on a regular grid. The nuggets and the sills have been deduced from the square of the mean of the chlorophyll-a concentration (the historical mean reference) as there is a classical proportionality effect between the square of the chlorophyll-a mean, the variance of the distribution and the parameters of the semi-variograms. Compared to each original product, the analysis shows a complete coverage and equivalent uncertainties in the retrieved variables and offers number of applications for environmental monitoring such as the application of the Water Framework Directive.