IAHR Document Library

« Back to Library Homepage « Proceedings of the 34th IAHR World Congress (Brisbane, 2011)...

Filling Missing Data in Streamflow Series for Supporting Models

Author(s): Mijail Arias-Hidalgo; Gonzalo Villa-Cox; Ann Van Griensven; Arthur E. Mynett

Linked Author(s): Arthur Mynett

Keywords: Streamflow time series; Hodrick-Prescott filter; Maximum likelihood estimation; Homoscedastic; Trend; Noise

Abstract: Several previous publications have addressed the problem of missing data in time series, which is a typical problem in many developing regions. This paper proposes an alternative when to develop a rainfall-runoff model to predict streamflows given rainfall series is time demanding or when such a data are simply not available. The Hodrick-Prescott filter has been employed in two modalities: heteroscedastic and homoscedastic by blocks (high and low discharges). The signal was split in trend and noise where the former is simulated and extended using Fourier Series and the latter is a function of the trend and computed using the Maximum Likelihood Estimation criterion for the unknown values. Afterwards, summing up the simulated trend and noise, a new total signal was obtained. The applicability of this technique was tested in some streamflow stations in Guayas River Basin (34000 Km2), Ecuador. Further models can be benefited from the products of these interpolations, either in setup stage (e.g. boundary conditions) or for further application such as recognition of overbanking periods.


Year: 2011

Copyright © 2024 International Association for Hydro-Environment Engineering and Research. All rights reserved. | Terms and Conditions