We thank Criss and Winston (2007) for their interest in and analysis of our manuscript on the interaction between discharge and drainage area. We feel that their points help strengthen the conclusions of our original article (Galster et al., 2006).
‘k’ Values: Criss and Winston begin by discussing the k values from Equation 1, which we characterized as a “measure of river base flow.” In hindsight this was an oversimplification, as the units of k will vary, and we agree with their assessment that k is not a useful measure of discharge behavior. Watershed variables such as extrabasinal sources of groundwater, antecedent moisture conditions, and precipitation characteristics will change from one measurement of discharge to the next, and can result in different k values. Our goal was to characterize ‘c’ from Equation 1 (Q = kAc), not k, and we do not discuss or make any conclusions regarding k after our initial characterization. Criss and Winston accurately describe the k values in Table 3 as being log k values, but that description does not affect the c values, which are the focus of our paper and discussion. We urge caution when applying physical interpretations to empirically fitted equations.
Poor Fit of Equation 1 to Data: Criss and Winston comment that the regression of peak flow #6 (published at 0.19 ± 0.67) should be 0.07 (no published error or confidence interval). However, not only are these two values (0.19 ± 0.67 and 0.07) statistically the same given the large 95% confidence interval (0.67), but the difference can be explained by rounding issues—for publishing purposes, the values in Table 2 were shortened to two decimal places. For example, the listed discharge of event #6 for the Bowers station was 6.52, but a value of 6.516 was used in the linear regression. Other discrepancies noted by Criss and Winston are even smaller than the above example, and all are within the stated 95% confidence interval.
Figure 2: We agree that the caption and the x-axis of our Figure 2 are contradictory, with the x-axis being correct and the caption being wrong. The figure caption should read that these data are from June 2005 to mid-July 2005. In a drafting error, the hydrograph of Virginville was repeated and mislabeled “Kuztown” instead of “Game land.” However, the data listed in Table 3 remain correct, as well as the statistical analyses derived from the data.
Comparison of Discharges from Similarly Sized Drainage Areas: Criss and Winston also note the disparity in our discharge data for similar drainage areas in the Sacony and Little Lehigh watersheds. Theses differences can be explained by the seasonality of the data. As we noted, most of the discharge data from Sacony Creek were collected in fall/winter, while most of the Little Lehigh discharges were collected in spring/summer. The fall/winter setting has lower evapotranspiration, higher soil moisture, and the possibility of frozen soil, all of which result in higher peak discharges given similar drainage areas.
Comparison of Other Data Sets to our Conclusions: As Criss and Winston show in their Figure 2, large data sets comparing river discharge and drainage area show linear, or close to linear, results. There are several important differences to note from their Figure 2 and the conclusions of our study. First, our study examines the increase in discharges within a single small watershed, using multiple gauging stations. Previously published research, as cited in both our article and in Criss and Winston's Comment, compiles data from multiple watersheds and concludes that c values are not greater than one. In our paired watershed study, we tried to carefully control for most of the watershed and hydrologic variables that plague large data sets. Our c value of 0.83 ± 0.25 for Sacony Creek watershed, in which the land cover is consistent throughout the watershed, not only agrees with Criss and Winston's Figure 2 but also validates our research methodology for measuring discharges.
Our study specifically set out to test how urbanization has affected the increase of discharge moving downstream in a single small watershed. The hypothesis was that changes in downstream urbanization levels in the Little Lehigh watershed would increase the flood peaks moving downstream. Our data show that the peak discharges in this watershed covary nearly with the square of drainage area (c = 1.81 ± 0.28). We note that this particular pattern of downstream urbanization is not unique but is found in other similarly sized watersheds in eastern Pennsylvania. The distribution of impervious surfaces (Carlson, 2003) in the following watersheds (ranging in area from 128 to 906 km2) was determined by dividing each watershed into upstream and downstream halves and calculating the average percent of impervious land cover for each half. Fourteen have increasing levels of impervious surfaces moving downstream (Aquashicola, Brandywine, Bushkill, Chickies, Darby, French, Jordan, Manatawny, Monocacy, Neshaminy, Pequea, Perkiomen, Saucon, and Tulpehocken), while only three (Little Schuykill, Mahantango, and Tohickon) have less impervious surfaces downstream. We suggest that while our findings of c > 1 may only apply to similarly sized watersheds with urbanization concentrated downstream, this land use pattern is not unique to the Little Lehigh watershed (e.g., the 14 watersheds listed above) and that our conclusions are broadly portable.
In summary, we thank Criss and Winston for their interest and scrutiny of our original article. We appreciate the opportunity to further explore the covariance between discharge and drainage area in small watersheds undergoing acute urbanization pressure.