Forecasting model for extreme rainfall using artificial neural network
Successive days of rainfall are known to cause flood. The forecasting of daily rainfall helps to estimate the occurrences of rainfall and number of wet days, while with a maximum of five consecutive days of rainfall, the magnitude of rainfall within a specified period can predict what may signify r...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | eng eng |
| Published: |
2015
|
| Subjects: | |
| Online Access: | https://etd.uum.edu.my/5282/1/s815184.pdf https://etd.uum.edu.my/5282/2/s815184_abstract.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Successive days of rainfall are known to cause flood. The forecasting of daily rainfall helps to estimate the occurrences of rainfall and number of wet days, while with a maximum of five consecutive days of rainfall, the magnitude of rainfall within a specified
period can predict what may signify rainfall extremes. In this study, data mining and back
propagation neural network (BPNN) have been established in developing the extreme rainfall forecasting models. Four forecasting models were developed to forecast the maximum five consecutive days of rainfall amount (PX5D) of the next month. The models only use the extreme rainfall indices outlined by STARDEX as predictors in forecasting. The first developed model uses six extreme rainfall indices in forecasting, the second model uses the values of the PX5D index of a three-month delay, the third model uses the previous six-month PX5D values, while the fourth model was developed
to forecast the PX5D using the values of the same index of a twelve-month delay. It was
found that when using the six extreme rainfall core indices, the forecasting error was the
lowest. A regression model has been developed using the six extreme rainfall indices to
compare the performance measurements with the BPNN model that uses the same indices |
|---|