ASSIMILATION FOR FORECASTING OF THE WEATHER AND THE NATURAL ENVIRONMENT

Operational Numerical Weather Prediction (NWP) uses large complex models, physically based on the equations of fluid dynamics, and the representation of atmospheric processes. Given complete and accurate initial conditions, they are capable of forecasting details of the atmospheric flow, temperature and humidity for several days ahead. In practice most errors in forecasts are due to inaccuracies in the initial conditions; the atmosphere is a chaotic system in which some small errors grow rapidly. The process of data assimilation has been developed to obtain the best possible representation at any time, from incomplete observations distributed over time. Global NWP models, and an operational global observing system, have existed for over 30 years. Recently, with increasing computer power, we have been able to apply data assimilation theory with fewer approximations, leading to significant improvements in forecasts.

Many other aspects of the natural environment are strongly dependent on the weather. So a global assimilation and NWP system is an important component in systems for forecasting these. The Met Office runs models for forecasting regional high-resolution weather, site-specific weather, the state of the land surface, the stratosphere, the oceans, waves, storm surges, seasonal climate, and the dispersion of nuclear and other pollution. These models are driven by the global assimilation and forecasts, which are run every 6 hours. Most assimilate the observations that are available for their particular environmental variables (e.g. the ocean), but this is not essential - often knowledge of the topography and atmospheric forcing is sufficient for a good representation and forecast (e.g. of soil moisture). Some can feed back boundary conditions into the atmospheric model. The whole is developing into a complex modelling system for representing and forecasting most of the natural environment, run operationally on one of the largest computer systems in the world.