The HIRLAM Project

This site report describes the HIRLAM project, which deals with the development and maintenance of a state-of-the-art operational numerical short-range weather forecasting system. The report provides an insight into the diierent components of these systems, presents examples what can be achieved, and describes brieey what kind of CSE research is carried out. The international project: Its aim and organisation HIRLAM (HIgh Resolution Limited Area Modelling) is a project to develop and maintain a numerical short-range weather forecasting system for operational use by the participating institutes, which are the national meteorological and Sweden. The project was started in 1985. The project is in its third phase, HIRLAM-3. The second phase, HIRLAM-2, was successfully completed with the production of the HIRLAM system version 2 1]. This version is now used in routine weather forecasting at almost all participating institutes. The project is controlled by the HIRLAM council, which consists of the directors of the participating institutes. The project is managed by a management group consisting of the project leader, a system manager, and their deputies. The management group is advised by a Scientiic and Technical Advisory Committee. Modern numerical weather forecasting systems consist of three basic components: Analysis, forecast model and postprocessing. The analysis provides the initial conditions for the forecast model, based on recent observations and other sources of information. The forecast model then integrates the classical equations in time. These equations are, see e.g. 2], Newton's second law, mass conservation (for air, and for atmospheric constituents like water), and thermodynamics. For the integration, the equations are spatially and temporally discretised. Several discretisation techniques are in use, most frequently one encounters nite diierence methods and spectral methods for the spatial discretisation, and a leapfrog scheme for the discretisation in time. For forecasts with relatively long lead times the forecast model should cover the whole globe. But for weather forecasts with a lead time of up to two days it is not necessary to do the calculations globally. So, for shorter-range forecasts, the calculations are done on a domain of limited horizontal dimensions, in order to save computer resources. The disadvantage is that lateral boundary conditions have to be provided. Usually these are obtained from an earlier forecast with a global model. The forecast variables are usually temperature, wind, and humidity on a three-dimensional grid, and several two-dimensional elds. The latter include surface pressure and surface temperature. Postprocessing is the step to …