Temperature reconstruction errors caused by representation of a priori information with use empirical ortogonal functions

The restoration of mean temperature of water layer is considered when the expansion of inhomogeneities of temperature field on empirical orthogonal functions (EOF) is used for description of a priori information. In practice only several first EOF giving the basic contribution into average square of a sound speed variation are taken into account. Even small modification of statistical model of inhomogeneities can noticeably change for the worse the quality of restoration of parameters of inhomogeneities. It is shown that the neglect by higher EOF, despite of their rather small contribution, can noticeably increase an error of the restoration of mean temperature of water layer.