| Summary: | 碩士 === 國立中興大學 === 土木工程學系所 === 102 === The average annual rainfall in Taiwan is about 2,500mm but most concentrated in typhoon season and the tremendous climate change intensifies loss caused by the disasters. If better rainfall prediction can be made, the data can be better used to lower the damage.
This study is based on grey prediction model along with radar rainfall as reference index for prediction and modification. First, the error from Radar rainfall prediction is discussed and revised by quadratic curve regression and residual grey model. Then, with radar rainfall prediction, we amended rainfall data which received from grey prediction model. The process is accomplished through Kalman filter、weight processing (one)、weight processing (two).
Chenyaolan Stream is the study case. Through comparison from prediction models,observed value and root mean square error as judgment index to certify the accuracy, we conclude that the accuracy from radar rainfall prediction is not ideal, but it predicts changes of rainfall well. Combining grey prediction model and radar rainfall along with the modification and the observed value is proved to amend the shortcomings of grey prediction model effectively. Being able to predict the blind spots of lag time and extreme value is an example. Kalman filter has best effect on prediction result.
|
|---|
