The Counterpropagation Fuzzy-Neural Network for Infilling Missing Daily Weather Data
Abstract
Hydrologists are often concerned with short duration data sets which also suffer from missing data values. Rainfall-runoff model calibration either for streamflow forecasting or simulation needs reliable, representative and complete time series of daily weather data (precipitation, maximum and minimum temperature). Our work deals with a particular artificial neural network namely the counterpropagation neural network with a fuzzy optimization method (CFNN). The CFNN has been used for real time streamflow forecasting. In this study, this network is used because it appears to be suitable to deal with data sets showing spatiotemporal patterns such as weather data. The performance of this network is assessed for infilling missing daily data such as precipitation, minimum temperature and maximum temperature. The CFNN is very effective at infilling the gaps when a principal component analysis is performed during the preprocessing stage of the input data.
- Publication:
-
AGU Spring Meeting Abstracts
- Pub Date:
- May 2004
- Bibcode:
- 2004AGUSM.H51F..04E
- Keywords:
-
- 1800 HYDROLOGY;
- 1854 Precipitation (3354);
- 1860 Runoff and streamflow;
- 3210 Modeling;
- 9820 Techniques applicable in three or more fields