Modelled Historical Data
Visit also the modelled historical data website, where you can download in csv format multi-year time series of meteorological fields from 2008 to 2010, with 2-km horizontal grid spacing and 10 min time resolution.
The Atlas is based on a set of quadrilateral meso-scale domains called "tiles", and the juxtaposition of 65 partially overlapping tiles constitutes a representation of Canada's wind potential. On each tile, climate is characterized by a large set of weather conditions, providing over 200 different possible atmospheric states. The climatic modelling then consists in finding a spatial solution for the wind flow in each of those states. Results are then post-processed with a statistical model representing the dominant winds in order to obtain weighted average wind velocities.
The statistical-dynamical downscaling method (Frey-Buness et al., 1995) is applied to produce the wind atlas. The method consists of using large scale long term atmospheric data and their statistical properties to run a mesoscale model and post-process its output in order to get a small scale picture of atmospheric motion. This method involves the following steps:
- Wind climate classification
- Mesoscale simulations
- Statistical post-processing
- Microscale modelling
Wind climate classification
The 3D representation of atmospheric state available every six hours at 2.5 degree resolution over the globe, known as NCAR/NCEP reanalysis (Kalnay et al., 1996), is chosen as data base. The elements of the data base are classified (Frank et al., 2001) using three parameters: the geostrophic wind direction at 0 m (16 equal sectors), the geostrophic wind speed at 0 m (9 classes 2 m/s wide from 0 to 18 m/s and 5 classes 4 m/s wide from 18 to 38 m/s which gives all together 14 classes) and the sign of 0-1500 m geostrophic wind shear (positive or negative). This way 432 bins (16x13x2+16 as no shear distinction is done for wind speeds below 2 m/s) called climate states are defined.
Each element of the long time series (every 6 hours for 43 years from 1958 until 2000) of the geostrophic wind vector at 0 m from the data base is attributed to a particular climate state. For each data base grid point, the classification procedure allows determining the climate states that occurred during the analysed period and the number of their realizations which defines their frequency of occurrence. This information will be necessary to initialize the mesoscale model and to do the post-processing.
The simulations are performed with the Mesoscale Compressible Community (MC2) model (Benoit et al., 1997), a state-of-art atmospheric model widely used by Environment and Climate Change Canada, Canadian Universities, and others worldwide. The Polar Stereographic grid with 5 km resolution at latitude 60 N is chosen. This grid is split into 65 partially overlapping (40%) domains of 175 by 175 points each to cover the whole Canadian territory. There are 28 unevenly distributed vertical levels with the two lowest model levels for wind calculations at 50 and 150 m. The orography and land use data are interpolated to the model resolution from the U.S. Geological Survey data base at 900 m resolution. The roughness field is determined entirely from the land use data. The centre of each domain is associated with the nearest grid point of the NCAR/NCEP global data base. Due to the classification procedure, this point is characterized by a specific set of climate states and their frequency of occurrence. For each climate state, a simulation with the MC2 model is performed. This includes the initialization with the climate states data which are this way downscaled to the model resolution. The simplified physics scheme without radiation, condensation or diurnal cycle is used in order to accelerate model convergence to the final state. The time step is 120 seconds and there is a nine-hour adaptation period for the initial flow to the surface geophysical properties.
For each domain, the entire set of model outputs is combined using the frequencies of the climate states simulated as their weight. Seasonal atlas is created using seasonal frequencies of the climate states as weights to the same model outputs used for the annual atlas.
This gives a set of 2D data at model resolution characterizing wind potential of the domain. The mean wind speed (EU) and power (E1) are available as images at three levels: 30, 50 and 80 m in the maps section of the atlas. For the 50 m level the information about the frequency of occurrence of 12 wind sectors and 27 wind classes is also available in the form of respectively a wind rose and a wind speed histogram at each model point. All variables describing the wind potential are available in RPN standard files. Two of them, EU and E1, are available in MID/MIF files, the format compatible with a wide range of professional GIS and Image Processing softwares. Those two types of files are available in the download part of this site.
Statistical post-processing prepares data, available in RPN standard files, to be used by a microscale model such as MS-Micro of Environment and Climate Change Canada (Walmsley et al., 1986) or WAsP of the Riso Laboratory of Denmark (Troen and Petersen, 1989) to refine the wind flow near complex surfaces. The most important field for microscale modelling is the frequency distribution of mean wind speed by sector and class (variable UHR) which allows producing a bivariate frequency table of mean wind versus wind force and wind direction. The other data from the RPN standard file can be used as well if needed. But this is only one piece of information necessary to run the microscale model the other being the high resolution description of surface properties which is not provided here.
See the complete bibliography.
- Date modified: