Preparations for the assimilation of ERS-1 surface wind observations into numerical weather prediction models

This paper presents an investigation of how to use spatially dense and correlated observations in numerical weather prediction (NWP) models. As an example, we study the error characteristics of ERS-1 scatterometer winds, by comparison to collocated 10 m winds from the NWP model of The Norwegian Meteorological Institute. We attempt to determine standard deviations and horizontal correlation of the scatterometer wind errors. The potential of ERS-1 observations to improve the fields of geopotential height and wind in NWP models, is discussed within the context of statistical interpolation. The sensitivity of a three-dimensional, multivariate analysis scheme to the specification of the error structure of surface wind observations is studied within the same context. The results indicate that the analysis result is very sensitive to misspecification of the horizontal correlation when the observations are highly correlated and dense.