Applying local model approach for tidal prediction in a deterministic model

Abstract

SUMMARYIn recent years, a practice of tidal prediction based on a deterministic model or by a time series forecasting model has been established. A deterministic model can predict tidal movement and capture the dynamics of the flow pattern over the entire domain. However, due to the simplification of model settings and near shore effects, the accuracy of the numerical model can diminish. Time series forecasting is capable of capturing the underlying mechanism that may not be revealed in the deterministic model simulation. However, such data-driven forecast fails to maintain accuracy with the progress of forecast horizon. In this paper, a scheme that combines the advantages of these two methods is introduced. The model errors are forecasted to different time horizons using a data-driven approach, and are then superimposed on the simulation results in order to correct the model output. Based on the proposed method, it is found that the accuracy is significantly improved with more than 50% of the errors removed on the average.

Enhancing water level prediction through model residual correction based on Chaos theory and Kriging

“…Previous studies of dynamic systems reveal that the random and irregular behavior in natural systems may arise from purely deterministic dynamics with unstable trajectories . Configuration of the observed process can be represented as a state through abstraction .…”

“…The local LM essentially reflects the underlying dynamics of a time series by using the most similar trajectories from the past data. The approach has been effectively applied to predict the time series even in cases of highly non‐linear systems . The approach simulates the evolution of a dynamical system to provide accurate short‐term predictions.…”

“…Although Sun et al . used outlined as an approach to forecast model residuals with satisfying results, it has been found that the LM approach is less competent in capturing the trajectories of the state vectors in the higher dimensional phase spaces for long forecast horizon. Other authors have extended the chaotic model from univariate model to multivariate model to predict the storm surges.…”

Efficient data assimilation method based on chaos theory and Kalman filter with an application in Singapore Regional Model

“…At the measurement stations, error correction can be carried out using error forecast models, which predicts errors using records of past residuals between model outputs and observations Babovic et al, 2001;Sun et al, 2008). However, error forecast models only correct the modelling system at the measurement stations and do not provide any information about updating other model points in the computational domain.…”

Artificial neural networks as routine for error correction with an application in Singapore regional model

“…As an efficient way to combat the presence of model errors, great efforts have been devoted to developing a variety of error prediction methods, such as chaos theory (Babovic et al 2005;Sun et al 2009) and artificial neural networks (Babovic et al 2001;Sun et al 2010). …”