Parameter Automatic Calibration Approach for Neural-Network-Based Cyclonic Precipitation Forecast Models

Der-Chang Lo,Chih-Chiang Wei,En-Ping Tsai

doi:10.3390/w7073963

Der-Chang Lo, Chih-Chiang Wei + Show 1 more

Open Access

https://doi.org/10.3390/w7073963

Copy DOI

Abstract

This paper presents artificial neural network (ANN)-based models for forecasting precipitation, in which the training parameters are adjusted using a parameter automatic calibration (PAC) approach. A classical ANN-based model, the multilayer perceptron (MLP) neural network, was used to verify the utility of the proposed ANN–PAC approach. The MLP-based ANN used the learning rate, momentum, and number of neurons in the hidden layer as its major parameters. The Dawu gauge station in Taitung, Taiwan, was the study site, and observed typhoon characteristics and ground weather data were the study data. The traditional multiple linear regression model was selected as the benchmark for comparing the accuracy of the ANN–PAC model. In addition, two MLP ANN models based on a trial-and-error calibration method, ANN–TRI1 and ANN–TRI2, were realized by manually tuning the parameters. We found the results yielded by the ANN–PAC model were more reliable than those yielded by the ANN–TRI1, ANN–TRI2, and traditional regression models. In addition, the computing efficiency of the ANN–PAC model decreased with an increase in the number of increments within the parameter ranges because of the considerably increased computational time, whereas the prediction errors decreased because of the model’s increased capability of identifying optimal solutions.

Highlights

IntroductionPacific and has an area of 35,981 km; the Central Mountain Range runs from north to south, and the Tropic of Cancer passes through the south
Taiwan is a long and narrow island located between Japan and the Philippines in the WesternPacific and has an area of 35,981 km2; the Central Mountain Range runs from north to south, and the Tropic of Cancer passes through the south
The traditional multiple linear regression model was selected as the benchmark for comparing the accuracy of the artificial neural network (ANN)–parameter automatic calibration (PAC) model

Summary

Introduction

Pacific and has an area of 35,981 km; the Central Mountain Range runs from north to south, and the Tropic of Cancer passes through the south. Approximately 80 tropical cyclones, known as typhoons, form annually worldwide, of which approximately 30 form in the western North. In the late 1980s, research on artificial neural network (ANN) applications advanced after the introduction of backpropagation training algorithms for feedforward ANNs [3]. ANNs, which simulate the biological nervous system and brain activity, have become the preferred forecasting approach in hydrology and hydrometeorology (e.g., [4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19]). ANNs are advantageous because feedforward networks are universal approximators capable of learning continuous functions with any desired degree of accuracy. Most ANN models have several parameters that users can adjust for realizing different scenarios and objectives, and the results produced by such models are typically distinct, which renders identifying the unique optimal solution difficult [20]

Methods

Results

Conclusion