Quantitative Precipitation Prediction for Multi-Step-Ahead Flood Forecasting Using Artificial Neural Networks
Resource
台灣水利 55 (2): 25-33
Journal
台灣水利
Journal Issue
2
Pages
25-33
Date Issued
2007-06
Date
2007-06
Author(s)
Chiang, Yen-Ming
Abstract
The study built the real-time quantitative precipitation estimation/forecasting (QPE/F) form the meteorological radar data using recurrent neural network (RNN) and constructed the multi-step-ahead flood forecasting by training the back-propagation neural network (BPNN) utilizing the QPF information. First, a three-dimensional radar data structure which take into account the terminal velocity and the horizontal advection are used for training the RNN in QPE/F. The results of real-time rainfall estimation show that the RNN can produce much more accurate and stable performance than the Z-R power-law function. This work shows that the dynamic RNN can be applied successfully in real-time QPE/F using remote sensing data. Second,an exhilarating performance was found through the comparison of two recursive BPNN structures with different input patterns. This study demonstrates that the recursive structure with QPF outputs not only has the ability to improve the model accuracy but has the capability of reducing the time-delay problem that occurred in flood forecasting. Therefore, it is suggested that the recursive structure with whe output of QPF is an effective method for multi-step ahead flood forecasting.
Subjects
Meteorological radar
Artificial neural network
Quantitative precipitation forecasting
Recursive structure
Multi-step-ahead flood forecasting
Type
journal article
File(s)![Thumbnail Image]()
Loading...
Name
以類神經網路建構定量降雨及多階段.pdf
Size
1.81 MB
Format
Adobe PDF
Checksum
(MD5):af1b949724fc3ec4e79fb6accc7715e5