|
 |
It was decided that the application of a simple forecasting function to the interpolated precipitation data set derived from historical values should be used for forecasting. The potential forecasting will be limited to a simple ratio model. A prediction for precipitation at point Y will be given by the equation,
Yc = Yh / Xh * Xc
where Yc is the predicted value based on current conditions and Yh is the historically derived values for the point of interest. Xc and Xh are the current and historically derived values for the closest known station. Xh will be an aggregate monthly value derived from the station records and Yh will be an aggregate monthly value of the same interval derived from our interpolated precipitation field. For example, if it is found that the interpolated historical value at GM Place (no measured value) for January is 300 mm and the historical value for January at YVR is 150 mm, then the ratio for GM Place will be 2. In other words, it can be estimated that for the month of January, it rains twice as much at GM Place than at YVR. If it is then known that it is currently (say January 6th) going to rain 5 mm at YVR, it can be estimated that GM Place will receive 10 mm.
This method encounters many problems and the values will not be particularly accurate. The problem of relating this ratio based on monthly data to daily values is encountered. The model also takes on the inaccuracy of the original interpolated model that is used to create the ratio values. This method is also dependent on the number of stations that provide current data. If only one station (YVR) is available then estimated points near that station will be more accurate than points further away. The final problem is that this model takes on the same temporal assumptions as the interpolation method in that it is also assumes the differences to be equal over an entire month.
The reason that this model is being selected is that it is simple to run and
achievable by the project. The estimates that it creates should be in the general
area of the real value and will be useful for those requiring such general values
for areas where accurate and up-to-date forecast data are unavailable
 |
back to
top |
 |