Details

Time series modeling in water loss

by Chuang, Wen-Cheng

Abstract (Summary)
The water loss including infiltration and evaporation is an essential factor for designing irrigation system, water supply facility, and pollution control system. Due to the extreme uncertainties of infiltration and evaporation, no deterministic model satisfactorily interprets the behavior of the phenomenon of water loss. By formulating the annual water-loss series and investigating their statistical properties, this study attempts to simulate the behavior of annual water losses in the Ohio River Basin and to forecast annual water losses for the region. Initially, the law of water balance is applied to an isolated watershed to formulate the annual water- loss series. Data nonhomogeneity and inconsistency are examined by the simple moving average method, and tested by the t-statistic, respectively. It is found that the four annual water-loss series studied have significant mean-level changes which contribute to the data nonhomogeneity. A proposed method is used to remove the nonhomogeneity in the data. It is found that the removal of nonhomogeneity results in the removal of inconsistency for the studied data from the Ohio River Basin. Next, the Autoregressive Integrated Moving Average (ARIMA) process is introduced to model the annual water-loss time series. The procedure of the model building, including Identification, Estimation, Diagnostic Checking, and Best Model Selection, is applied to find the best model for forecasting the annual water losses for the region. The performance of parameter estimation is examined by differently inputting initial values. The estimation by utilizing the initial values obtained from the method of moment increase the computation efficiency to provide the better estimation procedure. The best model selection based on the Akaike Information criterion, tends to select the model with least parameters. By the comparison of forecasted with observed values and their confidence bounds, the proposed real-time forecasting method shows somewhat satisfactory results.
Bibliographical Information:

Advisor:

School:Ohio University

School Location:USA - Ohio

Source Type:Master's Thesis

Keywords:time series modeling water loss pollution control system

ISBN:

Date of Publication:01/01/1987

© 2009 OpenThesis.org. All Rights Reserved.