Filling of missing rainfall data in Luvuvhu River Catchment using artificial neural networks
► We patched rainfall data which had gaps in the LRC. ► Artificial neural networks will help in estimating missing rainfall data. ► Reliable rainfall data has been produced for the study are for the selected stations. Incomplete data with gaps is always a challenge in hydrological modeling and water...
Gespeichert in:
Veröffentlicht in: | Physics and chemistry of the earth. Parts A/B/C 2011, Vol.36 (14-15), p.830-835 |
---|---|
Hauptverfasser: | , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | ► We patched rainfall data which had gaps in the LRC. ► Artificial neural networks will help in estimating missing rainfall data. ► Reliable rainfall data has been produced for the study are for the selected stations.
Incomplete data with gaps is always a challenge in hydrological modeling and water resources planning and management. Complete and reliable data is required for water resources planning and management. A study was done in the Luvuvhu River Catchment (LRC) with the aim of filling missing rainfall data. This was done with the aid of artificial neural networks (ANNs) using a radial basis function. The Root Mean Square Error (RMSE) was used as an objective function in the calibration phase. The Shuffled Complex Evolution (SCE) was used to find optimal parameters of the ANNs. Reliable rainfall data from surrounding stations was used as inputs to fill in missing rainfall data for an output station. A double mass curve was plotted to check the quality of rainfall data of the output station against the surrounding stations. Not all the stations in the LRC showed good correlation. However, data selected for training and testing from all the patched stations performed well. The measures of model performance fell within the acceptable ranges in hydrological modeling for all stations. During the calibration phase the Nash–Sutcliffe Efficiency (NSE) ranged from 0.55 to 0.85 and the percent bias ranged from −2% to 23%. In the validation process NSE range was between 0.49 and 0.75 and percent bias was between 2% and 19%. The values of NSE and percent bias were satisfactory to good, and acceptable respectively. This study has shown that ANNs are suitable for estimating missing rainfall data in the LRC. The study has produced reliable rainfall data that can be used in hydrological modeling and water resources planning and management. |
---|---|
ISSN: | 1474-7065 1873-5193 |
DOI: | 10.1016/j.pce.2011.07.041 |