Premium
Snow water equivalent time‐series forecasting in Ontario, Canada, in link to large atmospheric circulations
Author(s) -
Sarhadi Ali,
Kelly Richard,
Modarres Reza
Publication year - 2014
Publication title -
hydrological processes
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.222
H-Index - 161
eISSN - 1099-1085
pISSN - 0885-6087
DOI - 10.1002/hyp.10184
Subject(s) - autoregressive integrated moving average , series (stratigraphy) , autoregressive–moving average model , autoregressive model , time series , north atlantic oscillation , climatology , statistics , nonparametric statistics , environmental science , moving average , meteorology , snow , econometrics , mathematics , geography , geology , paleontology
The present study applies different time‐series models for forecasting daily and monthly snow water equivalent (SWE) data in Ontario, Canada, during 1987–2011. For daily time series, which showed a significant negative trend, four categories of the autoregressive moving‐average (ARMA) and ARMA model with exogenous variables (ARMAX) were applied. The North Atlantic Oscillation, Southern Oscillation Index and Pacific/North American Pattern, as large‐scale atmospheric anomalies, as well as temperature time series are considered as exogenous variables for ARMAX models. According to the multicriteria performance evaluation, a time‐trend ARMAX model demonstrated the best performance for modelling and forecasting daily SWE. Two models, seasonal autoregressive integrated moving average (SARIMA) and SARIMA with exogenous variables (SARIMAX), were also fitted to the monthly SWE time series. The results revealed that the SARIMAX model showed a better performance than the SARIMA model according to multicriteria evaluation. The three nonparametric tests, Wilcoxon, Levene and Kolmogorov–Smirnov for forecasting evaluation demonstrated that the selected time‐series models had enough reliability for short‐term SWE forecasting in Ontario. The results of this study also demonstrate the importance of incorporating both trend and appropriate exogenous variables for SWE time‐series modelling and forecasting. Copyright © 2014 John Wiley & Sons, Ltd.