ORIGINAL RESEARCH
Application of Holt-Winters Time Series Models
for Predicting Climatic Parameters (Case Study:
Robat Garah-Bil Station, Iran)
More details
Hide details
1
Department of Civil Engineering, University of Malaya, Kuala Lumpur, Malaysia
2
Department of Water Resources and Harbor Engineering, Faculty of Civil Engineering, Fuzhou University,
Fuzhou-Minhou, China
3
Department of Civil Engineering, Tabriz University, Tabriz, Iran
4
Department of Water Engineering, Gorgan University of Agricultural Sciences and Natural Resources
Submission date: 2018-09-21
Final revision date: 2018-11-30
Acceptance date: 2018-12-04
Online publication date: 2019-09-10
Publication date: 2019-12-09
Pol. J. Environ. Stud. 2020;29(1):617-627
KEYWORDS
TOPICS
ABSTRACT
Predicting hydrological variables is a very useful tool in water resource management.
The importance of the forecast in environmental issues causes us to use more accurate statistical
methods for studying the weather and climate change. The main objective of this study is to investigate
the use of additive and multiplicative forms of the Holt-Winters time series model to predict
environmental variables such as temperature, precipitation, and sunshine hours for one year in advance.
As the Holt-Winters model uses a weighted average of current and past values to provide predictions, in
this study higher emphasis is placed on the recent observations by using larger weights for these data
compared to the older ones. As a case study, monthly environmental data (i.e., precipitation, maximum
temperature, minimum temperature and sunshine hours) collected for a span of 30 years (from 1981
to 2010) from Robat Gharah-BilStation located in Golestan, Iran was used. After modeling the data
through additive and multiplicative procedures, the main three smoothing parameters of the model are
optimized using a nonlinear optimization method. Based on this study, using the multiplicative form
of Holt-Winters time series results in an overall of 4% less mean absolute percentage error (MAPE)
compared to the additive one. The result showed that this model is more efficient in predicting and
modeling climate parameters, which show stable patterns of cycle and seasonality.