Satellite based quantitative rainfall estimation for flash flood forecasting / Wardah Tahir
The nature of flash flood which can be dreadfully torrential in a very short time requires an improvement in its flood forecasting-and-warning system (FFWS) to further mitigate flood damages and loss of life. The use of geostationary meteorological satellite images in estimating rainfall has become...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 語言: | English |
| 出版: |
2008
|
| 主題: | |
| 在線閱讀: | https://ir.uitm.edu.my/id/eprint/65317/1/65317.pdf |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | The nature of flash flood which can be dreadfully torrential in a very short time requires an improvement in its flood forecasting-and-warning system (FFWS) to further mitigate flood damages and loss of life. The use of geostationary meteorological satellite images in estimating rainfall has become an attractive option in improving the performance of FFWS. Although the estimates are indirect, meteorological satellites with fine temporal and spatial resolution cover broader areas that may be inaccessible or that may cause difficulties with the traditional rainfall measurement such as the oceans or rigid mountains, therefore should be taken as complementary to radar and rain gage measurements. In this study, a rainfall estimation algorithm using the information from the geostationary meteorological satellite infrared (IR) images is developed for potential input to a flood forecasting system. Data from the records of Geostationary Meteorological Satellite-5 (GMS-5) IR images have been retrieved for selected convective cells to be trained with the radar rain rate in a back-propagation Artificial Neural Network (ANN). The selected data as inputs to the neural network, are five parameters having a significant correlation with the radar rain rate: namely, the cloud top brightness-temperature of the pixel of interest, the mean and the standard deviation of the temperatures of the surrounding five by five pixels, the rate of temperature change, and the sobel operator that indicates the temperature gradient. In addition, three Numerical Weather Prediction (NWP) products namely the precipitable water content, relative humidity and vertical wind are also included as inputs; to provide some considerations on the meteorological factors that physically contribute to rain formation. The algorithm is applied for areal-averaged rainfall estimation in the upper Klang River Basin and compared with another technique, which uses power law regression between the IR cloud top brightness temperature and radar rain-rate. Results from both techniques are validated against recorded Thiessan areal-averaged rainfall with coefficient correlation values of 0.77 and 0.91 for the power-law regression and the ANN-based technique, respectively This satellite-based quantitative rainfall estimation using the ANN technique is then transformed to a satellite-based quantitative precipitation forecast (QPF) by using the cross correlation technique to track the direction of movement of the convective cloud cells. In addition, an example rainfall-runoff model is developed, namely the unit hydrograph technique to be linked with the satellite-based QPF to become a coupled hydro-meteorological FFWS. An extra lead time of around two hours is gained when the coupled hydro-meteorological model is applied to forecast several flash-flood events in the upper Klang River Basin. |
|---|