Data used in the study was collected from records of historical climate data available electronically as well as data generated by interpolation. Detailed description of the data is provided in the climate data section.
Evaluation of interpolated Climate data
To evaluate how well the new interpolation technique estimates climate data in western Canada, correlation analysis was performed between observed and interpolatioed climate data for 240 weather stations in the five provinces, and r-square was computed. R-square is an indicator of how well the interpolated data fits the observed data. Observed climate data was generated from past records of each weather station. Spatial information (latitude, longitude and elevation) at each of the weather staitons was used for the generation of interpolated climate data for the same period as that where past weather records were available. Observed and interpolated climate data was compared at selected time slices as well as over a 102 year period.
Past and Future climate change Trends
Interpolated climate data for description of general trends was generated for each grid (approx. 15 x 15 km) covering the entire study area. The purpose of using a coarse resolution grid was to ensure that data generated would easily be handled, since higher resolution data would be too large to be handled by most programmes used for the analysis. Pivot charts were used to generate average values for each ecozone for each year. Deviations of all the 1901-2002 values from the 1961-1990 normals were computed for each ecozone. In addition, the deviation of future climate from the 1961-1990 reference period for the entire study area were also computed.
Climate Characeterization of Ecozones
Interpolated climate data used for characterization of ecozones was generated for 1x1 km grids that had been subsampled to take every fouth grid resulting in a coverage roughtly 4x4km.
In order to describe general groupings within ecozone climate data principal component analysis was used. Principal component analysis-PCA is a multivariate technique for examining relationships among several quantitaive variables. The principal components are the eigenvectors of the covariance matrix. PCA is a useful tool in reducing the multidimentional data to a few dimensions to be able to examine them visually. Given ten ecoregions in western Canada, PCA served to show how climate data reveals differences and similarities within then ten ecoregions.
Discriminant analysis was used for the classification of data into ecozones. Discriminant analysis is a multivariate technique that develops discriminant criterion to classify each observation, from a set of observations with several quantitative variables and a classification variable defining groups of observation, into one of the groups. The discriminant criterion is determined by a measure of generalised squared distance and can be used for testing another set of data. In this case discriminant analysis was used as both a descriptive tool for the 1961-1990 climate data as well as predictive tool with regard to future climate data.
|
