Civil and Environmental Research www.iiste.org ISSN 2224-5790 (Paper) ISSN 2225-0514 (Online) Vol.7, No.1, 2015 69 Forecasting Monthly Precipitation in Sylhet City Using ARIMA Model S H BARI 1 ,MT RAHMAN 2 , MM HUSSAIN 2 , SOURAV RAY 2 1 Department of Civil Engineering, Leading University, Sylhet, Bangladesh 2 Department of Civil & Environmental Engineering, Shahjalal University of Science and Technology, Sylhet- 3114, Bangladesh *Corresponding Author <sourav.ceesust@gmail.com> ABSTRACT In this study a seasonal ARIMA model was built using Box and Jenkins method to forecast long term rainfall in Sylhet. For this purpose rainfall data from 1980 to 2010 of Sylhet station were used to build and check the model. Rainfall data from 1980 to 2006 were used to develop the model while data from 2007 to 2010 were used to verify the prediction precision. Four basic chronological steps namely: identification, estimation, diagnostic checking, and forecasting were fitted out in developing the model. Validity of the model was tested using standard graphical explanation of residuals given by Box and Jenkins. As a second step of validation, forecasted values of monthly rainfall were checked using actual data series. After completion of necessary checking and forecast observation, the ARIMA(0, 0, 1) (1,1, 1) 12 was found to be the most effective to predict future precipitation with a 95% confidence interval. It is expected that this long term prediction will help decision makers in efficient scheduling of flood prediction, urban planning, rainwater harvesting and crop management. Keywords: Nonlinear time series analysis, ARIMA model, rainfall forecasting, Sylhet. INTRODUCTION Rainfall is very non-linear in nature and very complicated to predict. Due to adverse effects of climate change rainfall pattern has also been changing rapidly Short term and long term forecast of rainfall have significant relevance to agricultural, tourism, flood prevention and management strategy and water body management which influence the economy of a country. To predict such event, numerous techniques including numerical and machine learning processes have been adopted based on historical time series and radar data (Chander et al. 2002, Ingsrisawang et al. 2008). Still, currently most common methodology for rainfall prediction uses radar image data available from various organizations and analyzing them to predict rainfall. However various statistical methods are often useful to predict rainfall (Bisgaard and Kulahci, 2011). Among which the most effective approaches for analyzing time series data is the model introduced by Box and Jenkins (1976) and modified by Box, et al. (1994), also known as ARIMA (Autoregressive Integrated Moving Average).ARIMA has widely been exercised over the years to predict the rainfall trend (Mahsin et al. 2012,Kaushik and Singh 2008, Shamsnia et al. 2011, Thapaliyal 1981, Momani et al.2009), reservoir and river modeling (Dizon 2007, Cui 2011, Peng et al. 2000, Valipour et al. 2012), economics and production(Nochai et al. 2006),evapotranspiration(Valipour 2012).The method has some interesting features that made it more desirable for researchers. It eases the forecasting process allowing researchers to use only single variable time data series while also allow multiple for more complex cases. Rainfall forecast study for Dhaka division of Bangladesh has been done by Mahsin et al. (2012). However, forecasting of rainfall for Sylhet hasn’t been done yet. The area is one of the top tourist attracting locations in Bangladesh. Although the area is nearby the world’s wettest place (Cherapunji), urban Sylhet and some other region including Surma river basin are facing a rapid ground water depletion rate. Besides, excessive iron and arsenic contamination has an adverse impact in drinking water supply. To pursue a sustainable alternative source of supply, rainwater has become a suitable option in many parts of the region. Therefore the study objective of this study is to focus on development of a reliable forecasting of rainfall over Sylhet to manage water resources as well as handle flash flood effective and timely. METHODOLOGICAL APPROACH: (i) Study location and data collection Bangladesh meteorological department (BMD) collects rainfall data for Bangladesh through its 34 stations. Rainfall data for Sylhet station (Fig.:01) collected from BMD were used in the study. This station covers Sylhet District and it’s nearby areas.