No document available.
Abstract :
[en] DINEOF (Data Interpolating Empirical Orthogonal Functions) is an EOF-based method to reconstruct missing data in geophysical data sets. DINEOF can be used to reconstruct monovariate data sets (as sea surface temperature (SST), chlorophyll, etc), and multivariate data sets with little increase in complexity. For multivariate reconstructions, extended EOFs are used, which take into account the interrelationships between related variables to infer data at missing locations. Spatial maps of the standard deviation of the reconstruction error can be also calculated.
In the past, DINEOF has been compared to Optimal Interpolation (OI) techniques for the Adriatic Sea SST. The results showed that DINEOF was faster than OI, making it very suitable for operational applications.
DINEOF was also more accurate when compared to in situ data. Another advantage of DINEOF is that there is no need for a priori knowledge of the statistics of the reconstructed data set (such as covariance or correlation length), thus reducing the subjectivity of the analysis.
DINEOF has been successfully used to reconstruct a large variety of domains over the world ocean, mostly at the regional scale. In addition to an overview of the technique's capabilities, limitations and future developments, recent work aimed to improve the quality of the reconstructions at the global and local scales will be presented.