Convective-Scale Data Assimilation
Providing initial conditions for limited-area forecasting models with small grid-lengths (typically 1-4km).
In high resolution limited-area models, the focus is on local and surface conditions. Data assimilation development for the 1.5km grid-length UK model is aimed at improving forecasts of key weather elements such as rain, cloud, visibility, 2-metre temperature and 10-metre wind, out to two days ahead but with a main focus on improving forecasts of severe weather in the 0-12 hour period.
The operational UK model currently uses 3D-Var with a 3-hourly assimilation cycle, but a project is in progress to develop 4DVAR with an hourly cycle to support 'NWP Nowcasting' for the 0-12 hour forecast period.
The UK model assimilates some observations not used by the global model. Most of the extra data assimilated provide information on the moisture distribution. Cloud top temperature data from infrared satellite imagery and surface cloud reports are assimilated separately to improve the three-dimensional cloud analysis. These cloud data have been found beneficial in the analysis and prediction of stratocumulus and in thunderstorm situations. During a stratocumulus episode, for example, assimilation of cloud data can lead to significant improvements in the accuracy of 2-metre temperature.
Since 1996, rainfall data from the UK Weather Radar Network and nearby European countries have been assimilated by a technique known as latent heat nudging, giving improvements in precipitation forecasts. In future we hope to replace LHN with a 4DVAR scheme for radar reflectivity. Doppler radial wind data are also assimilated from the radar network.
Visibility is related to the relative humidity and aerosol content of the atmosphere in a very nonlinear way, which poses a challenge for variational assimilation systems. Nevertheless, a scheme to assimilate visibility observations into regional models has been developed and run operationally since 1999. Observations of near-surface temperature and humidity from the UK's roadside sensor network now supplement the standard surface observing stations and help to improve short-period forecasts of low visibility.
To develop data assimilation schemes for the 1.5km UK model and convective-scale models in other parts of the globe, aiming to improve forecasts of key weather elements such as rain, cloud, visibility, screen temperature and 10-metre wind.
Develop an hourly data assimilation cycle with 4DVAR for the 1.5km UK model.
Measure the marginal benefit of the convective-scale data assimilation system relative to a 'forecast-only' downscaler model.