Extreme Sea Level Prediction Method Research Based on Hybrid NARX Model

Abstract

Abstract The ocean occupies 70% of the world's surface area, and extreme changes in sea level can have an even greater impact on humans. Therefore, it is necessary to make an accurate advance prediction of extreme sea level. In this paper, a model with robustness, high accuracy and universal applicability has been constructed based on Nonlinear Auto-Regressive Exogenous (NARX) Neural Network. The data-set collected from several observation stations are selected as the input factors. Moreover, a neuron pruning strategy based on sensitivity analysis is also introduced. Because of this strategy, the model structure can be adjusted accordingly. Meanwhile, a modular prediction method is introduced based on the tide harmonic analysis data so as to make the model prediction results more accurate. At last, a hybrid extreme sea level prediction model, Pruned Modular (PM)-NARX, is constructed. In this paper, the model is trained by using historical data and used for extreme sea level prediction along the southern America in 2020. The simulations on MATLAB show that the correlation between the predicted data and the observed data is stable above 0.99 at 12h in advance. The prediction speed, accuracy, and stability are higher than those of conventional models. In addition, two sets of follow-up tests show that the prediction accuracy of the model can still maintain a high level. It can even be applied to other time-series prediction problems beyond extreme sea level prediction as well.