Diagnosing atmospheric motion vector observation errors for an operational high resolution data assimilation system

摘要

Atmospheric motion vectors (AMVs) are wind observations derived by tracking cloud or water vapour features in consecutive satellite images. These observations are incorporated into Numerical Weather Prediction (NWP) through data assimilation. In the assimilation algorithm, the weighting given to an observation is determined by the uncertainty associated with its measurement and representation. Previous studies assessing AMV uncertainty have used direct comparisons between AMVs with co-located radiosonde data and AMVs derived from Observing System Simulation Experiments (OSSEs). These have shown that AMV error is horizontally correlated with characteristic length scale up to 200 km. In this work, we take an alternative approach and estimate AMV error variance and horizontal error correlation using background and analysis residuals obtained from the Met Office limited area, 3 km horizontal grid length data assimilation system. The results show that the observation error variance profile ranges from 5.2 to 14.1 s m2s− 2, with the highest values occurring at high and medium heights. This is indicative that the maximum error variance occurs where wind speed and shear, in combination, are largest. With the exception of AMVs derived from the High Resolution Visible channel, the results show horizontal observation error correlations at all heights in the atmosphere, with correlation lengthscales ranging between 140 and 200 km. These horizontal lengthscales are significantly larger than current AMV observation thinning distances used in the Met Office high resolution assimilation.