STRUCTURAL TIME SERIES ANALYSIS AND FORECASTING FISH AND TIGER SHRIMP SEED COLLECTION FROM TWO TIDAL ESTUARIES IN SUNDARBANS K. K. GOSWAMI, UTPAL BHAUMIK, M. K. MUKHOPADHYAY, N. P. SHRIVASTAVA AND A. P. SHARMA Central Inland Fisheries Research Institute, Barrackpore, Kolkata 700 120 (Received : 16.02.2014; Accepted : 22.04.2014) Tidal estuaries in and around the Sundarbans in South 24 Parganas district of West Bengal are natural sources of fin and shell fish seeds. In this context, Structural Time Series Models (STSM) were developed to forecast daily fish seed collection from data generated under intensive study conducted in two tidal estuarine rivers of Sunderbans during the period 1994 to 1998. The trend plus cyclical models were used to estimate lunar effect, trend and seasonality from daily collection data and also for forecasting future values (R 2 value are 0.82 and 0.87) of fish seed and Tiger shrimp seed collection (R 2 value from 0.66 and 0.89). All the components of the models were found to be significant and both cycles were stationary and represented seasonality or lunar effect. Key words: Structural Time Series Model, trend, cycles, forecasting, Sundarbans J. Inland Fish. Soc. India, 46 (1) : 01-08, 2014 01 Introduction Resource management and forecasting are of great importance to fisheries management. Reliable forecasts, one year in advance, are helpful to resource managers, market planners/decision makers, and fishermen alike. A prime concern of the fishery science has been dynamic system modeling, specifically the construction of mathematical models that determine values of fish yield (harvest) through time. In most elemental form, this system is composed of an input, fishing effort, an output, catch and a mathematical model, which translates values of fishing effort to catch. Tidal estuaries in and around the Sundarbans are rich in commercially important prawn and fish seeds. Quantitative assessment of the natural seed resources of lower zone of the estuaries is of importance to determine the recruitment level and for the exploitation of important cultivable seeds for aquaculture. Information on the abundance of the seeds of shrimps and fishes for post- Farakka period is available in Mandal and Bhowmik (1984), Bhaumik et al. (1993), De and Sinha (1997) and Bhaumik and Mitra (2013). Harvey and Shephard (1993) described structural time series model as one, which is set up in terms of components, which have direct interpretation. Structural time series models are the state space form with the state of the system representing various unobserved components such as trends and seasonal. The estimate of the unobservable state can be updated by means of a filtering procedure as new observations become available. Predictions are made by extrapolating these estimated components for the future, while smoothing algorithms give the best estimate of the state at any point within the sample. A structural model can, therefore, not only provide forecasts, but can through estimates of the components, present a set of stylized facts. Freeman and Kirkwood (1995) have described methods for estimating stock biomass and recruitment from catch and effort data by using structural time series or state-space methods.