Author(s): K. Vidyalashmi; Nandana J. S; L. Megha Chandana; Aparna Devi V. M; K. Sreya; B. Parvathy; Abhirami A. S; Priya K. L; Gubash Azhikodan; Katsuhide Yokoyama
Linked Author(s):
Keywords: Shtamudi; Hydrodynamics; Estuary; Artificial neural network; NARX; Water level forecasting
Abstract: Estuaries are highly dynamic ecosystems due to the mixing of fresh water and sea water at the river mouth. Due to the tidal influence, variation in river discharge and geomorphology of the estuary, the water level is fluctuating throughout and forecasting them accurately is of paramount importance due to their impact on the ecological and navigational sectors. This study presents Non-linear Autoregressive with Exogenous input (NARX) algorithm for single step ahead water level forecasting of Ashtamudi estuary in India, which has microtidal mixed semi-diurnal pattern. Two models were developed using the one-year four-hour interval data of river discharge and water level. Model I which was trained using the previous water level alone gave an R value of 0.9826 during testing, while Model II which was trained using the previous water level and river discharge gave an R value of 0.9801. From the performance indicator values, it can be inferred that both models were equally accurate in forecasting, though the second model is preferred for real time forecasting since it considers river discharge as an external parameter.
Year: 2024