Trend assessment in rainfall-runoff behavior in urbanizing watersheds

A typical flood frequency analysis is based on gauged annual maximum discharges. One assumption behind the analysis is that the measured discharge signal is stationary. The validity of this assumption can be difficult to establish, particularly where urbanization has occurred within the gauged watershed, altering the response of the affected watershed to precipitation. This alteration can produce a nonstationary streamflow signal that can be significant, depending on the percentage of the watershed altered. As urbanization increases, peak discharges are shown to increase, producing a positive trend in the annual maximum series. Urbanization occurring during and/or after the gauging period is quantified using spatially and temporally distributed land use data. Three statistical tests (a parametric t-test on the slope of the linear relationship between the flood series and time and two nonparametric tests: the Kendall's Tau and the Spearman Rank Correlation) are performed on both the annual maximum discharge and annual maximum discharge-precipitation ratios series to test for trends or nonstationary signals corresponding to periods of urbanization. A case study suggests that the ratios are more effective than the discharges alone for identifying nonstationarity resulting from urbanization. In addition, relationships between measures of urbanization and the presence or absence of significant trends in the discharge series are presented.