Reconstructing synthetic profiles from surface data

A method for the extrapolation of vertical profiles of temperature (and/or steric heights) from measurements of sea surface elevation and sea surface temperature has been developed and described here. The technique, called Coupled Pattern Reconstruction (CPR), bases on a multivariate analysis of the coupled variability of vertical profiles from historical hydrographic data and on the hypothesis that only few modes are needed to explain most of the covariance of the fields. Through a linear regression between the amplitudes of the coupled modes it is possible to reconstruct the first two modes by solving a simple linear system written for the surface values, that are supposed known. The CPR method has been applied and tested on nine years of Conductivity-Temperature-Depth (CTD) measurements collected in the northern Mediterranean sea during the DYFAMED (DYnamique des Flux de mAtière en MEDiterranée) program (1994-2002). The first six years were used as a training dataset, while the last three were set aside as independent test measurements. Results have demonstrated a substantial improvement in terms of absolute error respect to an ad hoc climatology, and slightly worse performances comparing to the most advanced technique found in literature, that consists in the single Empirical Orthogonal Function Reconstruction (sEOF-R) through a multivariate regression of the amplitudes, except for the very first meters. However, CPR demonstrated to be much more robust and performing respect to sEOF-R when considering the errors associated to real measurements of sea surface elevation.