Convective-Scale Data Assimilation and Nowcasting

Nowcasting is the prediction of weather with 0 to 7 hour lead times. This is vital for the prediction of severe weather, in particular flooding events.

Nowcasts of rainfall and associated weather, such as hail and lightning, are the most widespread and advanced applications, though temperature, wind and other weather elements are also produced in some systems.

Traditionally nowcasting has been done by human weather forecasters or by computer-based extrapolation techniques of the current or most recent observations, particularly radar and satellite observations. Increases in available computer power allowing faster, more complex models with higher resolution mean that it is now possible to start developing an NWP-based nowcasting system - this will allow for prediction of initiation, development and decay of storms, by use of fundamental fluid dynamics and physical processes, which pure extrapolation techniques cannot represent.

A major challenge is to exploit available observations, especially high resolution ground and satellite based remote sensing, to describe the initial conditions for the nowcasts using data assimilation.

At JCMM a 1.5km resolution NWP-based hourly nowcasting system producing forecasts from 0 to 7 hours is being developed to use surface observations, radar radial Doppler winds, reflectivity and refractivity data, Meteosat satellite imagery, wind profiler, GPS and aircraft data using 3D-Var and 4D-Var data assimilation methods. The forecasts must be available very quickly after the time of the observations which is challenging not only for the data assimilation and NWP systems but also for transmission and pre-processing of the observations.

There is still a need to produce analyses that closely match the observations and a new project will look at techniques to try to improve on the current system.

Key Aims:

  • Development of efficient NWP-based nowcasting system for forecasting high impact weather
  • Exploitation of novel data sources such as radar, geostationary imagery, high resolution polar orbiter sounding data and ground based remote sensing

Current Projects:

  • Development of hourly data assimilation and forecast cycle at 1.5km resolution
  • Comparison of use of 3D and 4D variational data assimilation at 1.5km resolution
  • Development of techniques to directly assimilate radar data such as radial Doppler winds, reflectivity, derived surface precipitation rates and refractivity and geostationary satellite imagery at 1.5km resolution
  • Development of techniques for improved specification of forecast background errors at 1.5km resolution
  • Collaboration on project to define future upper air network for nowcasting and high resolution NWP over UK
  • Development of improved analysis techniques for UK observations

Last updated: 15 September 2014