Selecting
temporal resolution for the final forecasting model became important because
of the findings in the exploratory data analysis. As mentioned before, it has
been found that wide ranging values exist for a single day precipitation value.
The high variation of daily precipitation is essentially useless in making any
meaningful inferences. The entire process can be thought of as a statistical
sampling process at the daily scale. The sampling mean of the sampling distribution
for a given day is, for example, 9.79 mm at SFU Burnaby Mountain. However, due
to large variance in the sampling distribution, one can only infer that the
population mean daily value lies between 18.11 mm to 1.47 mm (Standard Error
= 4.16 mm) 95% of the time if sampling on this day was repeatedly done many
times. It was concluded that any meaningful trend is only attainable from more
aggregated data such as monthly or seasonally. However, seasonal aggregated
data does not quite fit into the overall goal of this project. The goal of the
project is to model the orographic effect on a scale that will be practical
for everyday life rather than making seasonal trend inferences. All considered,
monthly aggregated data is used to accommodate both the requirement of higher
confidence on mean values and also the applicability aspect of this project.
