College of Science and Mathematics

Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2003

This study is concerned with stochastic-modeling of the Talomo River flow using the non-stationary class of models. It aimed to construct models based on the monthly flow of data, validate the best fitting model through comparison of statistical characteristics of the output of the model and the observed data, assess the predictive abilities of the model to forecast five years ahead and compare the best fitted model with other suggested models. The data went through a series of model testing. The first stage was identifying the general information and description of the watershed. The second was investigating the different statistical characteristics using the first eleven years of the data, which includes the histogram, monthly means and standard deviation and as well as the spectral density of the empirical data. The third stage was fitting the data with the available time series models. The fourth stage is assessment of the predictable quality of the SS model using the remaining two years and the last stage is, forecasting. Four tentative models were generated: Winters Method ? Additive, Seasonal Exponential Smoothing, and Log Winters Method ? Additive and Log Seasonal Exponential Smoothing. From the statistical and verification results, among the four models, Log Winters Method Additive appeared to be superior, which could model the Talomo River flow. It generated with the least standard error. Moreover, the model forecasted such flows for the next four years for the watershed.