Modeling of Hilsa (Tenualosa ilisha) landings in the lower stretch of Brahmaputra River (Assam, India) under time-series framework.

For effective fisheries management of Brahmaputra River in Assam, India as well as for sustainable fishery development in the river stretch, it is significant to know the change in pattern of fish landings in previous years. Statistical modelling helps in describing the dynamics of fish landings and their short-term predictions. Thus, we attempted modeling the quarterly data series (1987-2019) on Hilsa (Tenualosa ilisha) landings in the Indian part of Brahmaputra River at Uzanbazar (Guwahati) using univariate forecasting techniques, viz., ARIMA (Auto-Regressive Integrated Moving Average) and NNAR (Neural Network Auto Regression). A comparative performance of fitted models was assessed based on the forecast accuracy measures. Based on comparison of forecast accuracy of models, the ARIMA model outperformed the NNAR model. Based on Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values, an ARIMA(1,0,0)(0,1,1)4 was found to be the appropriate model for describing Hilsa landings of Brahmaputra River. Under the present conditions, the forecasts by fitted ARIMA model predicted stagnation in abundance of prized Hilsa fish in the River at around 3000 kg/year for the upcoming years, which call for formulating and implementing an effective fishery management plan for conservation of its stocks. This is the first attempt on developing time-series models to forecast the Hilsa landings of Brahmaputra River in the region. [ABSTRACT FROM AUTHOR]