Detecting Changes in River Flooding Patterns and Flood Frequencies in the Upper Midwest

Abstract

Several statistical techniques for detecting changes in river flooding patterns have been applied to the peak annual flow series of 101 USGS stations found in the Upper Midwest and the implications on flood frequency statistics based on splitting the systematic flood records are then examined. Mann-Kendall trend test is one of the techniques used to determine if an increasing trend exists overall in the entire annual peak flows series up to the latest year in the record. For each station that exhibits a trend, moving average and ratio of means methods are performed on the series in preparation for a two tailed t-test which pinpoints a trend change year. The flood frequency statistics of the stations are updated using the trend change year found. In order to study the flooding trends more closely, wavelet analysis is used to detect trend changes in the daily flow series. Wavelet coefficient plots are made at several cycle lengths allowing the determination of trend changes occurring at each scale. The research show varying effectiveness of the techniques used in detecting trend. Through wavelet analysis, the influence of ENSO may be shown as a possible cause of trend change although the splitting of a series may be necessary as indicated by the statistical techniques.