Cloud-free satellite data for operational applications using DINEOF

DINEOF (Data Interpolating Empirical Orthogonal Functions) is an EOF-based technique to reconstruct missing data in satellite data sets, such as gaps created by the presence of clouds. It is parameter-free, meaning that no a priori information is needed (such as signal to noise ratio, or correlation length) to calculate the missing data: this information is extracted from the data through the EOF decomposition. In addition, computational time is lower than for other frequently used techniques to reconstruct missing data in satellites, such as optimal interpolation. Multivariate reconstructions can be also done, using extended EOFs. These characteristics make DINEOF very suitable for operational reconstruction of satellite data. Recently added to DINEOF is a technique to filter the temporal covariance matrix which allows to reduce spurious variability in the temporal EOFs, and therefore leads to improved reconstructions. We will present a general description of the technology, with examples of applications to different variables. We will also give an example of a near real time reconstruction of sea surface temperature in the western Mediterranean Sea. Conceived as a demonstration product for DINEOF, it is hosted at http://gher-diva.phys.ulg.ac.be/DINEOF/dineof.html and it is automatically updated daily, presenting the cloud-free sea surface temperature for the last ten days, as well as the original data, outliers and error fields.