Prescribed error estimation and diagnostic analysis in reconstruction of 3-D ocean temperature field from multi-source data

To reconstruct 3-D ocean temperature fields using both satellite measurements and in situ observations, the prescribed statistical information estimation in the optimal interpolation (OI) was optimized and the impact mechanisms of these two types of data were analysed. In the OI, two comparative schemes were implemented, with the background field set to a climatology (static scheme) and a synthetic field from satellite surface data (dynamic scheme), respectively. Before the OI, one a posterior diagnosis iterative method was performed to optimize the background error and observation error covariances. After the OI, the two schemes were compared using some diagnostic errors in the observational space and indices on the model grids. The main conclusions include:(1) The analysis error is smaller for the dynamic scheme than for the static one, in which the reduced absolute quantity is determined by the differences in their observation errors (i.e., noises), and the reduced relative extent is determined by the differences in their background errors (i.e., signals). (2) The background term is more important in the high and middle latitudes, and equivalent to the observation term along the equator, which is determined by the qusi-zonally-orientated distributed covariance scales. (3) The dynamic fields have higher spectral energy related to the temperature mesoscale signal features than the static fields by 1-3 orders of magnitude overall, but roughly the same in the tropics. (4) Satellite measurements reduces the total error and increase the effective resolution of the analysis field by resolving the mesoscale features that in situ observations cannot resolve well.