Methods
In this study we use the Weather Research and Forecasting (WRF) NWP model for the forecasts. This model is a next-generation mesoscale numerical weather prediction model collaboratively developed by the National Center for Atmospheric Research (NCAR), the National Oceanic and Atmospheric Administration and other organizations. For further information about the WRF model please refer to http://www.wrf-model.org/.
In order to assess the outputs of the NWP model we need to have the actual weather observations. For this purpose we use a data set from the Research Data Archive (RDA) which is maintained by the Computational and Information Systems Laboratory (CISL) at the National Center for Atmospheric Research (NCAR). NCAR is sponsored by the National Science Foundation (NSF). The original data are available from the RDA (http://dss.ucar.edu) in dataset number “ds472.0”. The data we are focusing on here is the subset of these observations that fall within the province of British Columbia comprising of about 50 weather observation stations (approximate locations are shown in the map).
Data Pre-processing
For the data preparation process we developed a C++ application as in the left figure which first selects the observation records from the NCAR sources that are inside BC and are within a specific time range (here we only use the Summer of 2008). Next, we scan each WRF model output file which has its own format (about 50 files each about 300 MB for the summer forecast in BC) so that we will assign the prediction values to each BC observation record based on time and distance. Finally, the matched record holding the prediction and observation together will be written to an output file after some post-processing to change the units or to compute new derived values.
