Weather models

Three different weather prediction models (IFS, HIRLAM, AROME) and one localized wind model (WAsP) were used to produce the Wind Atlas.

Weather prediction models in general

A weather prediction model computes the evolution of pressure, wind, temperature, humidity, clouds, rain, sunshine and other elements of the weather step by step for a large number of grid points, thereby forming a complete, three-dimensional picture of the atmospheric state that evolves with time according to the laws of physics. When a model is used to predict the actual weather, it is repeatedly restarted from initial conditions corresponding to observations of the real atmosphere, e.g. every six hours. In this way, the simulated evolution is forced to follow the real evolution as closely as possible.

The output of a weather prediction model can be treated in the same way as observations. In particular, averages and other statistical characteristics can be computed for any model variable or combination of variables anywhere within the domain of the model. Similar completeness and coverage could not be provided by any existing observation system for even one single moment in time.

In both nature and a model, the weather evolves following the same laws of physics, expressed in the model as a set of mathematical equations, discretized on a certain computational grid. The denser the grid, the smaller is the domain represented by each grid point and the finer, in general, is the scale of motion systems and structures that can be simulated. Still, the model output is not strictly comparable to observations because the latter will always be influenced by structures and processes beyond the reach of the model.



The spatial resolution of a weather prediction model depends on the grid spacing. The figure shows how topographic height and aerodynamic roughness are resolved in models having, from left to right, a horizontal grid spacing of 2.5 km, 7.5 km and 80 km, respectively.

The AROME weather prediction model

The non-hydrostatic AROME meso-scale weather prediction model is designed to resolve phenomena on horizontal scales down to several kilometres, such as large thunderstorms or sea-breeze circulations. It is under continuous development by a large European consortium, and is in daily use for operational forecasting at the FMI and elsewhere. The variables predicted by AROME are horizontal wind, vertical motion, pressure, temperature, humidity, two species of cloud condensate, cloud cover, three species of precipitating particles, the kinetic energy of small- scale turbulence, soil temperature and humidity, snow cover, and the exchange of heat, moisture and momentum between the surface and the atmosphere.

The surface of a given grid cell can consist of water or land. The water may be frozen or open, and the land may be covered by various types of forests or fields, or by built-up areas, each of which can be covered by snow. The exchange of heat, moisture and momentum in a grid cell is computed separately for each of the surface types present in the cell, and averaged taking into account the fractional coverage of each surface type. Each surface type is characterized by its own aerodynamic roughness length, albedo and thermal emissivity.

The effective roughness in AROME differs from that in WAsP. In AROME, the aerodynamic roughness length of open water depends on the wind speed according to Charnock's formula. Over ice, a constant roughness length of 1 mm is assumed. Over land, the roughness length depends on the vegetation and buildings, and on the snow cover. Vegetation type and buildings are described according to the ECOCLIMAP database. In AROME, the roughness length for heat and humidity exchange differs from the roughness length for momentum exchange.

The presence of sea ice influences the lower atmospheric flow via two mechanisms:

1. The roughness of sea ice differs from the roughness of open water. The ice may be level, in which case it is smoother than wavy open water, or it may pile up into a field of ridges, in which case it is rough compared to open water.

2. The ice cover acts as an insulator between the water and the air, blocking the flow of heat into the air. Therefore the air near the surface may cool considerably and form a very stable layer in which vertical mixing is suppressed, with weak winds prevailing near the surface.

In AROME, only the latter effect is taken into account, as it is considered the most important for conditions in Finland.

The computational domain covers Finland and nearby areas. There are 300 grid points in the east-west direction and 600 grid points in the north-south direction. The grid spacing is 2.5 km. In the vertical, there are 40 terrain-following levels, nine of which are located in the lowest kilometre of the atmosphere. The lowest computational level is located 30 m above the ground. The computation progresses in steps of one minute.

The HIRLAM model

The HIRLAM model differs from the AROME model primarily by its coarser resolution, allowing the description of motion systems of scales down to several tens of kilometres assumed to be in hydrostatic balance. HIRLAM has been maintained by a European consortium since 1985, and is used for short-range weather prediction at the FMI and elsewhere. The grid spacing used for the Wind Atlas is 7.5 km. In the vertical there are 60 levels following the terrain. Because of the different horizontal resolution, the effective roughness of HIRLAM differs from that of AROME.

The IFS model

The IFS (Integrated Forecasting System) model is a global hydrostatic model in use at the European Centre for Medium Range Weather Forecasts. In the ERA interim project, the model was applied with a horizontal grid spacing of 80 km at 60 vertical levels.