Sea-water-level prediction via combined wavelet decomposition, neuro-fuzzy and neural networks using SLA and wind information

Abstract

Sea-water-level (SWL) prediction significantly impacts human lives and maritime activities in coastal regions, particularly at offshore locations with shallow water levels. Long-term SWL forecasts, which are conventionally obtained via harmonic analysis, become ineffective when nonperiodic meteorological events predominate. Artificial intelligence combined with other data-processing methods can effectively forecast highly nonlinear and nonstationary inflow patterns by recognizing historical relationships between input and output. These techniques are considerably useful in time-series data predictions. This paper reports the development of a hybrid model to realize accurate multihour SWL forecasting by combining an adaptive neuro-fuzzy inference system (ANFIS) with wavelet decomposition while using sea-level anomaly (SLA) and wind-shear-velocity components as inputs. Numerous wavelet-ANFIS (WANFIS) models have been tested using different inputs to assess their applicability as alternatives to the artificial neural network (ANN), wavelet ANN (WANN), and ANFIS models. Different error definitions have been used to evaluate results, which indicate that integrated wavelet-decomposition and ANFIS models improve the accuracy of SWL prediction and that the inputs of SLA and wind-shear velocity exhibit superior prediction capability compared to conventional SWL-only models.