资源环境科技发展态势分析平台

With the goal of popularizing the use of readily available data sets from numerical weather prediction models in crop yield forecasting, we present a comparative analysis of end of season forecasts of wheat and canola from daily values of maximum and minimum temperature and total daily precipitation across the Canadian Prairies from (1) the Global Ensemble Prediction System (GEPS) and (2) statistically generated values from climate stations. The analysis was done using the Canadian Crop Yield Forecaster (CCYF), a tool for conducting crop yield outlooks within the growing season. We found that the GEPS data sets provided skillful forecasts of spring wheat and canola from selected Census of Agricultural Regions (CARs) in Alberta, Manitoba and Saskatchewan. Aggregated results for the Prairie region showed that the GEPS data had a similar predictive skill as the statistically generated values of temperature and precipitation for spring wheat and showed improved prediction skill overall for canola from the provinces of Alberta and Saskatchewan. In the Canadian Prairie environment, where climatic records are short and spatially insufficient, the GEPS data set, which is produced every Thursday and gridded at 45 km resolution, can be used as a substitute for, or supplement to, station-generated climate variables. Because of the continuous improvement in numerical prediction models such as GEPS in terms of skill score and resolution, the testing of crop forecasting models should be done at regular intervals to take advantage of these data sets as they become available.