The influence of isolated observations on short-range numerical weather forecasts

Abstract

The influence of isolated observations on short-range numerical weather forecasts has been studied by comparing ECMWF forecasts over limited areas.We first identify situations where a single observation has a large impacton the analysis and then study the subsequent error growth by comparing 2 numerical forecasts with and without the observation. The study is carried out in data-sparse regions. We have analysed 152 cases and the results show a large spread in the perturbation growth rates which are calculated as the RMS differences between forecasted (up to 48 h) 500 mb geopotential heights. The average growth rate is close to earlier estimates of error growths, but this growth rate is not very representative for the ensemble of forecasts which have been investigated. Instead, there is a tendency towards a bimodal distribution, with either very rapid or slow perturbation growth rates. To identify the physical cause of this large variation in growth rates, we have evaluated local instability indexes for each case and compared these with the actual growth rates. The instability index is calculated from both the baroclinic and barotropic properties of the unperturbed, initial flow fields. We find that the occurrence of rapid perturbation growth rates seems to be coupled to the baroclinicity of the flow in the vicinity of the localized perturbation. Barotropic effects in terms of strong, local vorticity gradients do not seem to be associated with rapid perturbation growths.