The accuracy of weather prediction is important in many levels of society. For instance, airlines depend on the prediction of cloud ceiling, visibility, and wind for the safety of passengers (Klein et al. 2009). Weather forecasts also provide information for agricultural workers about planting date, hybrid, nitrogen fertilizer amount, and plant density to manage the agricultural land (Jones et al. 2000). Moreover, students may rely on weather forecasts to decide how to dress accordingly. For the safety of students, schools will have some course cancellations in winter, according to the short term weather forecasting. The ubiquity of mobile devices means that more and more people are relying on weather apps to make decisions in their lives and professions. Therefore, the accuracy of the weather forecast can be said to play an important role in everyone’s life.

An extensive literature is available on weather forecasting techniques (Ripley and Archibold, 2002). According to Ripley and Archibold’s study done in Canada, the coastal stations have lower forecast errors than inland sites because of lower temperature variability. Another research done by Lupo and Market (2002) also shows that seasonal variations of forecast accuracy exist. As the temperature prediction error shown is relatively low, we have decided to find out whether the weather prediction of coastal area UBC has as high of accuracy as the coastal temperature predictions. Besides this, Gordon (1974) has suggested that weather forecasts have not yet achieved a satisfactory level of usefulness as there are limitations when it comes to the accuracy of the computer technology and statistical power. Existing techniques are used based on an analogue, and matching techniques where the overall weather circulation of a particular month in the past is used to predict the weather condition in the preceding month. Therefore, it is clearly seen that weather forecast may not be a reliable tool to predict the future weather conditions as there is a technological limitation.

On the other hand, nowadays, as technology has improved, the weather forecasting results have become more accurate, especially when predictions are made using the nowcasting methodology. A nowcast will give a weather forecast for a short term period, such as 12-hours period. This method has been used at least since 1992, when it was used for automated forecasting of thunderstorms (Dixon). Therefore, we have chosen to analyze weather forecasts which are 12 hrs in advance, and use this data to compare to the actual weather conditions at the time.

Gordon (1974) has introduced a classification system to evaluate the accuracy of the prediction where a table that consists of nine boxes is constructed, three for each class of prediction: good, bad and very bad. The class of prediction will be determined depending on the matching results of the observed weather condition and the predicted weather condition. Therefore, a similar idea is used in our research where we match the predicted weather based on reliable sources and the observed weather to further prove our hypothesis.