An EOF-based technique to compute merged high resolution sea surface temperature fields

High quality sea surface temperature (SST) data sets are needed for various applications, including numerical weather prediction, ocean forecasting and climate research. The coverage, resolution and precision of individual SST satellite observations is not sufficient for these applications, therefore the merging of these complementary data sets is needed to reduce the final data set error. This is usually performed by optimal interpolation
(OI).We present an extension of the capabilities of DINEOF (Data INterpolating Empirical Orthogonal Functions) to merge data from different platforms. The analysis is based on the formalism of OI, but the crucial difference is that the error covariance is not parametrized a priori using an analytical expression, but expressed using a spatial EOF basis calculated by DINEOF. This EOF basis represents more realistically the complex variability
of SST data sets than the parametric covariance used in most OI applications. An example will be presented using data from a polar-orbiting satellite (AVHRR on MetOp) and a geostationary satellite (SEVIRI on MSG). The high spatial resolution of the polar-orbiting satellite and the high temporal resolution of the geostationary satellite are retained to create a very high spatial and temporal resolution field of the western Mediterranean
SST. The results are validated with independent data.