An Evaluation of ARIMA and Holt Winters Time Series Models for Forecasting Monthly Precipitation and Monthly Temperature (Case Study: Latian Station)

Climatic parameters including temperature and precipitation have an important role in water resources management of river basin as well as agricultural planning. Time series models are a kind of short-term prediction for these parameters. Precipitation is one of the most important climate parameters that should be addressed in water resources management. This is especially important in Iran, with an average annual rainfall of about 250 mm. Another climate parameter is temperature, which changes the climate structure of each location. For this reason, the study on temperature at various time and space scales has been addressed in a large part of the climatological researches. Time series analysis is widely used as a tool for temperature and rainfall predictions. So far, various studies have been done to predict climate and hydrologic parameters using time series analysis models. Kaushik and Singh (2008) predicted monthly temperature and precipitation in India using the seasonal Arima Model. There are also other researchers focused on application of ARMIA model such as Naill and Momani (2009), Tularam and Ilahee (2010) and Mondal et al. (2014). Holt Winters is also one of the time series models used for prediction. For example, Costa et al. (2015) predicted water quality parameters using the Holt Winters model and presented its effectiveness in the prediction.
In this research, the ability of time series models for forecasting monthly temperature and precipitation of Latian station in Iran has been examined. Trend analysis was conducted using the Seasonal Mann- Kendall test and then, various Autoregressive Integrated Moving Average Models (ARIMA) as well as Holt Winters model were fitted to the data and the best time series model was finally selected.