Improved estimation of radar rainfall using multivariate Kalman filtering approach / Sharifah Nurul Huda Syed Yahya
Radar rainfall estimates has become an indispensable complementary to rain gauge data as input to flood prediction model. The research work had focused on methods to improve radar rainfall estimates. The techniques used in this study will help better rainfall estimation after reducing some errors co...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2015
|
| Subjects: | |
| Online Access: | http://ir.uitm.edu.my/id/eprint/15734/ http://ir.uitm.edu.my/id/eprint/15734/1/TM_SHARIFAH%20NURUL%20HUDA%20SYED%20YAHYA%20EC%2015_5.pdf |
| Summary: | Radar rainfall estimates has become an indispensable complementary to rain gauge data as input to flood prediction model. The research work had focused on methods to improve radar rainfall estimates. The techniques used in this study will help better rainfall estimation after reducing some errors correlated with weather radar. The radar rainfall data from Terminal Doppler Radar, KLIA Malaysia and rain gauge data from the Department of Irrigation and Drainage Malaysia were used. The hourly rainfall accumulation from the Doppler radar in mm/hr was estimated and compared with the hourly rainfall rate from the rain gauge network. In the first part of this thesis, Kriging interpolation for a network of rain gauges in Klang River Basin is presented. The kriged rain values were compared with radar rainfall estimates to assess the advantages of both methods. Three types of semivariogram functions namely Spherical, Exponential and Gaussian had been applied in the interpolation. The results show that the Spherical model performs the best in the interpolation with r2 = 0.78. It is also found that Kriging interpolation of rain gauge data can provide useful comparative values to the radar rainfall estimates. The second part of this thesis presents state space model, which applied Kalman Filter method. In this work, multivariate technique was integrated in Kalman Filter method to reduce noise in radar rainfall data. The novel approach of Kalman Filter combined with multivariate analysis help to improve radar rainfall estimate by correcting, updating and forecasting. The integration process needs to state the threshold point in the system of noise variance (Q) and observation noise variance (R). The implementation of this technique also involved the weather parameter elements that affect rainfall occurrence such as temperature, wind and humidity. Results showed that Kalman filter integrated with multivariate analysis using measured weather parameter reduce radar rainfall errors. For example, Kalman filtering with multivariate analysis for Petaling Jaya station has improved r2 values from 0.68 to 0.87. |
|---|