Runoff Coefficients of Floods in New England

Flooding and its consequences, including the loss of lives and infrastructure, have long been a major concern in New England (NE), especially because of spring rains and snow melt (Benson, 1963; Jahns, 1947; Patton, 1988; Wolman and Eiler, 1958). Moreover, the intensity and frequency of extreme precipitation events have been increasing in the northeastern US since the beginning of the 20^th century, with the most significant increasing trends within the more recent decades (Wuebbles et al., 2017). The complexity of watershed hydrological systems limits the ability to directly attribute trends in precipitation to trends in surface runoff, measured as the event-induced streamflow at the basin outlet (Ivancic and Shaw, 2015). However, recent studies have recognized positive trends in the frequency and magnitude of high-flow events in NE (Arnell and Gosling, 2016; Collins, 2009; Demaria et al., 2016; Ivancic and Shaw, 2015; Prosdocimi et al., 2015; Slater and Villarini, 2016). For high-flow events, the runoff coefficient (RC), which is the fraction of precipitation converted into surface runoff during an event, indicates basin vulnerability to flooding. RCs in different basins depend on geographic characteristics including land use which can impact surface runoff production and soil infiltration rates (Soil Conservation Services, 2004). Within a given basin, RC variability mainly results from differences in antecedent conditions such as snowpack and soil moisture (Woldemeskel and Sharma, 2016). Therefore, it is no surprise that historical floods in NE have mainly occurred under wet conditions when surface runoff is maximized (Paulson et al., 1991). This study explored the seasonal, temporal, and spatial variability of RCs for high-flow events in NE in basins with minimal anthropogenic manipulation, with the goal of understanding recent trends in climatic drivers of flood magnitudes.

This study explored recent spatial, seasonal, and temporal trends in RCs in NE for a period of 36 years from 1981 to 2016. We used daily stream-flow data from 28 long-term United States Geological Survey (USGS) Hydro-Climatic Data Network (HCDN) gauge stations in NE from basins with minimum human disturbance. This set of gauges had been previously used for streamflow trend analysis in basins free from human manipulation (Collins, 2009), which allowed us to focus on climatic rather than local anthropogenic factors. Daily precipitation information was obtained from PRISM (Parameter-elevation Regression on Independent Slopes Model) 2.5 arcmin (4 km) resolution gridded data, summed over the watershed for each gauge.

This study focused on high-flow events, which were determined to be those that yielded a daily maximum streamflow of equal to or greater than the 2-year return period for each basin. Surface runoff for each event was obtained by subtracting baseflow from measured streamflow. The corresponding precipitation volume was determined by integrating the hyetograph for a time frame of similar length, shifted back in time by a basin-specific lag time. RCs were obtained by dividing surface runoff volume by total precipitation volume. Analysis of RC trends were conducted using the non-parametric Mann-Kendall method. Data preparation, calculations, and statistical analyses were performed in R.

Analysis of 688 high-flow events indicated that RCs increase in magnitude and variability with distance from the Atlantic coast towards north and west in NE. Also, even though precipitation and streamflow are both independently increasing over time, RCs for high-flow events in NE were found to be generally stationary. The average RC of high-flow events in NE was 0.86, and RC magnitudes greater than one occurred in about 25% of events. In fact, 5 out of 28 basins have annual average RCs of greater than one during high-flow events. Overall, RCs were highest in spring (average RC of 1.04), and 99% of events with RCs greater than one occurred during March, April, or May. The observed seasonality of RCs highlights the role of snowmelt and high soil moisture content in producing high-flow events in NE.

Results from this study show that flood events in undisturbed basins have remained proportional to precipitation inputs, despite increases in extreme precipitation, possibly due to shifts in evapotranspiration, snowpack, and soil moisture. Flood management efforts should continue to focus on large springtime precipitation events, which generate the highest RCs. Finally, this study can serve as a reference point for future explorations of the flood susceptibility of basins with anthropogenic alterations such as dam construction and land use change.