Developed Seasonal ARIMA Model to Forecast Streamflow for Savitri Basin in Konkan Region of Maharshtra on Daily Basis

Abstract

Forecasting of streamflow is very useful for designing of soil and water conservation structures and planning of watershed activities, design of hydraulic structures, etc. Stochastic based model are found to be very reliable in estimation and forecasting streamflow on short and long term basis. In present study, streamflow on daily basis of Savitri basin were forecasted using seasonal ARIMA model developed in SPSS software following initial analysis of time series, identification model, apply diagnostic checks, adoption of model and forecast the stream flow by adopted model with checking the statistical indices such as correlation coefficient ( R), root mean square errpr (RMSE), coefficient of efficiency (CE), volumetric error (EV), mean absolute deviation (MAD) and mean absolute percentage error (MAPE). The streamflow data for duration of 1992 to 2011 (20 years) measured at Mahad were used to calibrate, validate and forecast the streamflow by the ARIMA model. It is observed that, the data used in ARIMA model is stationary, no trends and no seasonality found in data sets on daily basis. The two models were identified as ARIMA (111,111)³¹ and ARIMA (202,000)³¹ as best model for forecasting the streamflow with R value more than 0.9 during calibration and forecasting period. The other statistical indices RMSE, CE, EV, MAD, MAPE were also found in appropriate range to forecast the runoff with goodness of fit. Hence, developed ARIMA model could be adopted to forecast the runoff for sub tropical coastal region of Maharashtra on short and long term.