Baseflow separation with digital filters /
Baseflow separation with digital filters has often been characterized in the literature as the isolation of the low frequency (baseflow) and high frequency (runoff) components of a streamflow record. To test this classification several representative time series were decomposed into their constitue...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
2000.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=727853391&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | Baseflow separation with digital filters has often been characterized in the literature as the isolation of the low frequency (baseflow) and high frequency (runoff) components of a streamflow record. To test this classification several representative time series were decomposed into their constituent sinusoids by Fourier transformation. The runoff models typically comprise more low frequencies than high frequencies. With respect to digital filtering the most significant finding is that the frequency content of baseflow and runoff overlap in the low frequency part of the spectrum. The effect of digital filters is frequency dependent. Because part of the baseflow and runoff "signal" occupy the same portion of the frequency spectrum perfect baseflow separation with digital filters is not possible. However, the procedure can be optimized. The Fourier transform of a time domain filter is a frequency domain representation called the frequency transfer function (FTF). The magnitude and phase of the FTF quantifies the effect that filtering will have on the magnitude and phase of each sinusoid of the input time series. A commonly used recursive baseflow filter was evaluated in the frequency domain to suggest an optimal filtering strategy. The result can be generalized to any recursive baseflow filter. All of the commonly used baseflow filters are implemented so that they will cause some phase distortion of the constituent sinusoids of the input time series. A simple algorithmic change produced a filter that can not cause phase distortion. This new class of filters is also able to remove more runoff signal without attenuating the baseflow signal. Non-recursive filters are not commonly used for baseflow attenuation. Several time and frequency domain baseflow filters were designed and tested. Most performed poorly. It was shown that a frequency domain filtering scheme may be useful as a pre-processing step prior to recursive filtering. All common recursive baseflow filters are of order one. Second and higher order recursive filters of several types were designed and tested. All performed poorly. However, two second order filters used in a zero-phase algorithm appear to be the best performing baseflow filters yet designed. |
|---|---|
| Item Description: | Vita. "Major Subject: Civil Engineering". |
| Physical Description: | xv, 243 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilm Inc. |
| Bibliography: | Includes bibliographical references (leaves 207-216). |