Assimilation of radar data in the Met Office mesoscale and convective scale forecast systems

The Met Office is developing a high resolution NWP forecasting capability with the aim of ultimately replacing existing nowcasting techniques. A 4km resolution model for the UK has been introduced operationally this year and the aim is for a 1km resolution system once sufficient computer power is available. This is based on the nonhydrostatic unified model.

The 4km model is currently being run once a day without data assimilation using interpolated 12km resolution analysis/forecast as the initial conditions. A data assimilation system is under development for use at 4km and 1km resolution.

Radar data and satellite imagery are important sources of observations for high resolution modelling. Currently surface precipitation rate analyses derived from operational radar data are exploited operationally in the 12km UK and 12km North Atlantic and European forecast systems via latent heat nudging. Other data is analysed using 3D-Var and then the analysis increments are nudged into the 12km resolution forecasts along with latent heat increments derived from forecast/analysis precipitation rate differences and humidity increments derived from forecast/analysed cloud cover differences. Doppler radar radial winds outside the UK are assimilated via VAD profiles in 3D-VAR.

The complete 3D-Var and nudging system has been trialled at 4km resolution and the nudging system using the 4km 3D-Var analysis increments and its own latent heat and moisture increments has been trialled at 1km resolution. These trials have shown the importance of the rain rate and cloud cover analyses in forecasts of precipitation, which is an important end-use of the system for flood prediction/ warning. Work is under way to test use of higher time frequency precipitation (every 15mins compared with hourly) and cloud cover analyses (every hour compared with 3 hourly) and to reduce overprediction of precipitation in the very short range forecasts.

The Met Office should have its first operational doppler radar radial wind products by the end of the year. VAD profiles from the first radar will be included in the operational analyses after monitoring to test the quality of the data.

Code to assimilate Doppler Radar radial winds directly has been incorporated into the Met Office Variational Data Assimilation system. The code has been tested using data derived from a PPI scan by the Chilbolton Advanced Meteorological Radar. The very high resolution raw data from Chilbolton is averaged (or "superobbed") to model resolution before assimilation. The project with the Chilbolton data is being carried out as a collaboration between the Met Office and a team at the Telford Institue of Environmental Science, Salford University (Dr. F. Rihan and Prof. C. Collier). Work is underway to ingest the operational radar data when it is available and to develop quality control and monitoring systems as well as operational systems to specify observation error and perform superobbing.

Projects in other groups at Reading University are also just starting to investigate the potential use of radar reflectivity and refractivity data in high resolution data assimilation systems. These are at an early stage concentrating on issues relating to the observations themselves. Other projects are looking at techniques for inclusion of precipitation data, and also reflectivity data, in the variational analysis system itself. Work is underway to enhance the linear physics in the perturbation (linear) forecast model 4D-Var system.

This paper will describe the work being undertaken at, or in collaboration with, the Met Office to exploit radar data within the mesoscale and convective scale data assimilation systems.