Horizontal correlations in assimilated satellite-derived downwelling radiative fluxes

Horizontal correlations in satellite-derived downwelling longwave (LW) and shortwave (SW) fluxes are investigated within a Bayesian merging framework. The framework utilizes a 2D filter such that more than one measurement is assimilated for any given location. The number of measurements assimilated increases as a function of the user-defined influence length. Filter performance is based on comparisons of both prior (unconditioned) and posterior (conditioned) estimates against an independent, ground-based observational network in the Southern Great Plains of the United States. Preliminary results suggest assimilation of multiple measurements (i.e., when the influence length is greater than the spatial resolution of the measurements) yields a superior LW flux estimate during cloudy-sky conditions relative to that of the 1D filter (i.e., when the influence length equals the spatial resolution of the measurements). However, the 2D filter actually degrades clear-sky LW and all-sky SW flux estimates. These results suggest horizontal correlations exist, but only within a relatively small (~1 degree) region of space, particularly during the summer months when cloud systems are more dynamic and variable. Beyond this region of space, a poorly defined horizontal covariance structure often leads to the incorporation of inferior information relative to the ground-based observations.