DOI: http://dx.doi.org/10.26483/ijarcs.v8i8.4590 Volume 8, No. 8, September-October 2017 International Journal of Advanced Research in Computer Science RESEARCH PAPER Available Online at www.ijarcs.info © 2015-19, IJARCS All Rights Reserved 292 ISSN No. 0976-5697 MONTHLY TEMPERATURE PREDICTION BASED ON ARIMA MODEL: A CASE STUDY IN DIBRUGARH STATION OF ASSAM, INDIA K. Goswami Department of Statistics, Dibrugarh University, Assam, India-786004 J. Hazarika Department of Statistics, Dibrugarh University, Assam, India-786004 * A. N. Patowary College of Fisheries, Assam Agricultural University Raha, India-782103 *Corresponding Author Abstract: The forecasting of temperature on a seasonal time scales has been attempted by many researchers by different techniques at different time across the globe. It is a challenging task to forecast temperature on monthly and seasonal time scale. In this paper, an attempt has been made to develop a Seasonal Autoregressive Integrated Moving Average (SARIMA) model to long term temperature data of Dibrugarh, Assam, for the period of fifty (50) years (1966-2015). The analysis revels that the best seasonal models which are satisfactory to describe the data are SARIMA(2,1,1)(0,1,1)12 for monthly maximum and SARIMA(2,1,1)(0,1,1)12 for monthly minimum temperature data respectively. Keywords: Hydroclimatology, Komogorov-Smirnov test, KPSS test, SARIMA 1. INTRODUCTION Climate change, now a days is one of the biggest environmental threat to all over the world. It is one of the growing problems to water resources, livelihoods and forest diversity. The analysis of long-term data sets on hydroclimatic plays an important role in climatic studies. In recent years, it has been becoming a field of interest to the scientific community of the whole world. But it is a difficult task to analyze the changing pattern of climate as it is happens due to various reasons, some of which are local and some are global factors. Analysis on hydroclimatic variables can provide information on how the climate has evolved over time. Since, the events are evolving with respect to time and they have some successive relation, hence it is relevant to apply time series analysis on the time dependent data. The main focus of time series analysis is to give some future prediction by analyzing the past data through modelling. To assess the nature of the climate change in different regions of the world, a number of time series studies have been conducted in recent years. Out of the existing approaches, auto regressive integrated moving Average (ARIMA) model is the most widely used methods in recent times in the field of hydroclimatology. It was firstly proposed by Box and Jenkins in 1970 [18]. It can not only grasp more original time series information but also its flexibility nature, it is widely used in meteorology [13], [5]. Another advantage of ARIMA model is that, it can also apply in non-stationary time series data with some clearly identifiable trends [18]. The model is generally written as ARIMA(p, d, q), where p and q are non-negative integers that correspond to the order of autoregressive and moving average process respectively whereas d stands for order of difference the model. Since, periodicity of periodical time series is usually due to seasonal changes or any other natural reasons and so to account seasonal parameter, we can build a seasonal model viz., ARIMA(P,D,Q) model [11], where the parameters P, D and Q are seasonal autoregressive parameter, seasonal integrated parameter and seasonal moving average parameter respectively. Depending upon the data, one can build a multiplicative seasonal autoregressive moving average model (SARIMA), in short SARIMA(p, d, q)(P, D, Q)s model [19] where s represents period of time series data. In practical, the order of SARIMA model is generally not too large [9]. [17] applied SARIMA models that counted for 92% of the total variability in the monthly means of air temperature and found good agreement with the actual observed values of temperature. According to them for highly variable time series, SARIMA models gives better forecasts report than the simple models which are only based on means of previous observations. [12] analyzed weather variability and the incidence of cryptosporidiosis with the comparison of time series Poisson regression and SARIMA models. They used time series Poisson regression and SARIMA models in testing the potential impact of weather variability on the transmission of cryptosporidiosis and found SARIMA model having better predictive ability than the Poisson regression model. [7] applied SARIMA on hourly bicycle count and temperature data and modelled Vancouver Bicycle Traffic using weather variables. By using SARIMA model on average monthly temperature, [1] found that the average temperatures are rising over time in Ahwaz station, which was an indication the fact that the globe is warming. For forecasting of monthly minimum and maximum temperatures in the Moulvibazar and Sylhet districts of Bangladesh, [15] used SARIMA model and found SARIMA (1,1,1) (1,1,1)12 , SARIMA (1,1,1) (0,1,1)12 and SARIMA (1, 1, 0) (1, 1, 1)12 , SARIMA (0, 1, 1) (1, 1, 1)12 for the maximum and minimum temperatures at Sylhet and Moulvibazar district, respectively. According to them these results would help the researchers to estimate the missing