Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2004

Weather elements are important inputs to many models in hydrology and agriculture. With the rise in agricultural and similar developmental activities in the Davao provinces, there is now a greater demand for weather data and studies. This study focused on analyzing daily weather data (maximum and minimum temperature, relative humidity, rainfall, wind speed and sunlight duration) collected from 1996-2002 in Davao City. Since each weather variable of this particular data set has not been studied before, these were first subjected to univariate analysis using the Box-Jenkins methodology (also called ARIMA for autoregressive integrated moving averages) to describe how each series moved with time and itself. This was followed by multivariate analysis where a multivariate time series model was developed using the Vector ARMA process. The methodology for multivariate analysis suggested that the given time series data be first translated into principal components based on the covariance matrix. These components, in contrast to the original variables, are independent making it possible to fit univariate ARIMA models for each of the principal components. These univariate models were converted back into a multivariate form for the original data, which resulted in a Vector ARMA (1,3). Five-step ahead forecasts from the univariate and multivariate analysis were the means for comparison of the acceptability of the model. It was found out, however, through the Principal Component Analysis and comparison of forecasts, that for this specific set of data, each weather variable can be treated independently without having to lose much information.