Based on the proportion of runoff that reached the outlet on each of the 5 LDK378 purchase days following a rain or snow melt event, we were able to determine a best-fit Tp which minimized the root-mean-square error between the predicted and observed runoff shape
(see Appendix A for further details). We used the take-one-out approach to evaluate the degree to which any one watershed influenced the relationships between the best-fit Tp and Tc. We performed three independent tests on our model: (1) we used a leave-one-out approach to see how well our model would predict the hydrograph of a watershed that was not used to determine the regional model parameters, (2) we compared our predicted storm runoff locations to shallow water table measurements, and (3) we compared our predicted storm runoff locations to measured soil moisture. To understand how the model would perform in ungauged watersheds, we considered the recalculated relationships between S and SWDd and between Tc and Tp, determined by systematically excluding one watershed in a leave-one-out approach ( Arlot and Celisse, 2010). We then used these relationships to model the excluded watershed and compare the predicted and observed discharge hydrographs; note, in the earlier part of this paper we were only investigating how sensitive the parameters
were to any one watershed and here we are evaluating learn more model performance. The values of the coefficients for the relationships between measured and model parameters when excluding each watershed are reported in Table 2. Modeled results were compared to USGS daily streamflow measurements at each location. In addition to the Nash-Sutcliffe efficiencies (NSE),
we determined the ratio of the root mean square error to the standard deviation of observed streamflow (RSR) and the percent bias (PBIAS) for each watershed (Nash and Sutcliffe, 1970). Moriasi et al. (2007) proposed that a model is satisfactory if NSE > 0.50, RSR < 0.70, and has an absolute PBIAS < 25%. We also calculated NSE on an event basis, where runoff events were initiated by a 1 day rise in the observed USGS hydrograph after at least 2 days of decreasing flows. We created a LIDAR-derived STI (Fig. 3) for comparison to water 3-mercaptopyruvate sulfurtransferase table height measurements from Town Brook Watershed, using: (i) a 3 m LIDAR-derived DEM from the NY Department of Environmental Protection (DEP), (ii) maximum triangular slope (Tarboton, 1997), (iii) the Multiple Triangular Flow Direction method (Seibert and McGlynn, 2007) as per Buchanan et al. (2013). We then binned STI values into equal-area wetness classes, such that low-numbered wetness classes are wetter areas (large STI values) and high wetness classes signify dry areas of the watershed (low STI values). This allowed us to assign a location as “wet” or “dry” during a storm event based on the saturated extent predicted by the model. Lyon et al.