Streamflow prediction using non-parametric regression methods at Pahang river catchment / Dewi Anneka Halid
Flood disaster cannot be stopped or avoided, but all stakeholders should learn how to face this problem with an aggressive approach and maximum efforts to minimize the losses in terms of fatalities and economic. One of the remaining ways to minimize the flood problem is to constantly seek new knowle...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2016
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/37822/1/37822.pdf https://ir.uitm.edu.my/id/eprint/37822/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Flood disaster cannot be stopped or avoided, but all stakeholders should learn how to face this problem with an aggressive approach and maximum efforts to minimize the losses in terms of fatalities and economic. One of the remaining ways to minimize the flood problem is to constantly seek new knowledge and alternatives to improve the flood forecasting system. An ongoing need to achieve the best accuracy of flood forecasting obviously has been stimulating many recent studies to give more attention in many advances technique of flood prediction. Therefore, the aim of this study is to investigate the potential of two non-parametric regression methods as a flood predicting tool, where the application is relatively new in the hydrologic problems. The approaches studied are K-Nearest Neighbours (KNN) and Multivariate Adaptive Regression Splines (MARS). Pahang River situated in Pahang, Malaysia has been selected as an area of interest of this study. 30 years of historical data set of daily rainfall and streamflow at upstream tributaries of Pahang River were used as input data to develop and evaluate the effectiveness of both approaches in one-year-ahead prediction of streamflow. The effect of different length of data sets to the performance of models was also examined. Simulation results showed that longer period data can provide significant improvement to the performance of both approaches. However, based on the comparison of performance between KNN model and MARS model, all the error values and efficiency percentage for KNN model (CC=0.7241 to 0.7601; MRSE=0.591 to 0.2503; MRAE=0.464 to 0.3361; CE=11.62 % to 31.72 %) at all predictive points are not as good as the error values and efficiency percentage for MARS model (CC=0.9898 to 0.9921; MRSE=0.0278 to 0.0184; MRAE=0.1098 to 0.0962; CE=97.97 % to 98.45 %). This clearly indicated that satisfactory result of streamflow prediction only appeared superior for MARS model. Overall, all findings that emerge from this study have gone some way towards enhance the understanding of the capability and limitation of KNN and MARS model in long term flood prediction. |
|---|