HYDROLOGICAL PROCESSES Hydrol. Process. 15, 1903���1924 (2001) DOI: 10.1002/hyp.246 Quantifying contributions to storm runoff through end-member mixing analysis and hydrologic measurements at the Panola Mountain Research Watershed (Georgia, USA) Douglas A. Burns,1* Jeffrey J. McDonnell,1��� Richard P. Hooper,2# Norman E. Peters,2 James E. Freer,1�� Carol Kendall3 and Keith Beven4 1 State University of New York, College of Environmental Science and Forestry, Syracuse, NY 13210, USA 2 US Geological Survey, Atlanta, GA, USA 3 US Geological Survey, Menlo Park, CA, USA 4 Lancaster University, Lancaster, UK Abstract: The geographic sources and hydrologic flow paths of stormflow in small catchments are not well understood because of limitations in sampling methods and insufficient resolution of potential end members. To address these limitations, an extensive hydrologic dataset was collected at a 10 ha catchment at Panola Mountain Research Watershed near Atlanta, GA, to quantify the contribution of three geographic sources of stormflow. Samples of stream water, runoff from an outcrop, and hillslope subsurface stormflow were collected during two rainstorms in the winter of 1996, and an end-member mixing analysis model that included five solutes was developed. Runoff from the outcrop, which occupies about one-third of the catchment area, contributed 50���55% of the peak streamflow during the 2 February rainstorm, and 80���85% of the peak streamflow during the 6���7 March rainstorm it also contributed about 50% to total streamflow during the dry winter conditions that preceded the 6���7 March storm. Riparian groundwater runoff was the largest component of stream runoff (80���100%) early during rising streamflow and throughout stream recession, and contributed about 50% to total stream runoff during the 2 February storm, which was preceded by wet winter conditions. Hillslope runoff contributed 25���30% to peak stream runoff and 15���18% to total stream runoff during both storms. The temporal response of the three runoff components showed general agreement with hydrologic measurements from the catchment during each storm. Estimates of recharge from the outcrop to the riparian aquifer that were independent of model calculations indicated that storage in the riparian aquifer could account for the volume of rain that fell on the outcrop but did not contribute to stream runoff. The results of this study generally indicate that improvements in the ability of mixing models to describe the hydrologic response accurately in forested catchments may depend on better identification, and detailed spatial and temporal characterization of the mobile waters from the principal hydrologic source areas that contribute to stream runoff. Copyright ��� 2001 John Wiley & Sons, Ltd. KEY WORDS stormflow end-member mixing analysis mixing model riparian groundwater Georgia, bedrock outcrop hillslope runoff runoff model INTRODUCTION Many studies of small catchments have used hydrograph separation techniques to identify the principal source components of stormflow (see reviews in Bonell (1993), Buttle (1994), and Genereux and Hooper (1998)). Separation techniques generally use isotope (Sklash et al., 1976 Pearce et al., 1986) or chemical tracers * Correspondence to: D. A. Burns, US Geological Survey, 425 Jordan Road, Troy, NY 12180, USA. E-mail: daburns@usgs.gov ��� Now at Oregon State University, Corvallis, OR, USA. # Now at US Geological Survey, Northborough, MA, USA. �� Now at Lancaster University, Lancaster, UK. Received 1 March 2000 Copyright ��� 2001 John Wiley & Sons, Ltd. Accepted 10 September 2000

1904 D. A. BURNS ET AL. (Caine, 1989 Eshelman et al., 1993) to solve a mass balance expression for the relative proportions of stream runoff derived from: (1) precipitation (sometimes referred to as new water or event water), and (2) water stored in the catchment prior to the onset of a hydrologic event (sometimes referred to as old water or pre-event water). The tracers in these two components of runoff are assumed to mix conservatively in the catchment during the event to provide an accurate representation of stream runoff, an assumption that is sometimes violated (Pilgrim et al., 1979). Applications of a two-component modelling approach to small, forested catchments generally have found that the pre-event component dominates stream runoff during high flow (Buttle, 1994 Genereux and Hooper, 1998). The pre-event and event water are considered time-source components because they provide a temporal separation of stream runoff (Sklash et al., 1976 Genereux and Hooper, 1998). DeWalle et al. (1988) found that inclusion of a third time-source component���soil water���was necessary to account for variations in 18O values in a Pennsylvania stream during rainstorms. Subsequent stormflow models have been developed using mixing diagrams, resulting in a geographic source separation of streamflow into three runoff components (Christophersen et al., 1990 Hooper et al., 1990 Mulholland, 1993 Ogunkoya and Jenkins, 1993). Despite the wide application of tracer-based, multi-component mixing models to describe the runoff response to rainfall and snowmelt, little work has been done to systematize the varied observations among catchments. A review of data from more than 90 storms that were analysed through a two-component modelling approach indicated that peak flow in forested and agricultural catchments has a greater proportion of pre-event water than peak flow in urban catchments, and that peak flow resulting from rainfall has a greater proportion of pre-event water than peak flow resulting from snowmelt (Buttle, 1994). Within a catchment, the proportion of pre-event water in peak flow was found to decrease as rainfall intensity and amount increased (Hill and Waddington, 1993 Brown et al., 1999), and was found to increase as antecedent soil moisture increased (Burns and McDonnell, 1998). Additional studies are needed in a wide variety of catchment environments, however, to confirm these findings. In particular, an overrepresentation of mixing model studies from forested catchments in northern temperate climates and few studies in sub-tropical and tropical environments has been noted (Bonell, 1993). Many studies that have developed three-component models to describe catchment runoff have used throughfall, soil water, and groundwater as the three runoff components (DeWalle et al., 1988 Bazemore et al., 1994 Rice and Hornberger, 1998). Although this approach recognizes the importance of soil water as a runoff component, it has generally failed to provide insight into the geographic distribution of hydrologic source areas in a catchment. Of particular interest in many catchments, yet little explored to date, is the interaction between mobile waters on the hillslope and groundwater stored in the riparian area (Robson et al., 1992 Peters and Ratcliffe, 1998 McGlynn et al., 1999). The Panola Mountain Research Watershed (PMRW), a 41 ha forested catchment in the Piedmont region of Georgia, has been the site of several studies of storm runoff (McDonnell et al., 1996 Freer et al., 1997) and stream chemistry during rainstorms (Shanley and Peters, 1988 Hooper et al., 1990 Peters, 1994 Peters and Ratcliffe, 1998 Peters et al., 1998). The response of the stream to rainfall in the catchment is affected strongly by a 3��6 ha bedrock outcrop in the headwaters that provides rapid runoff during storms (Shanley and Peters, 1988). Some studies have concluded that runoff from this outcrop has little direct effect on stream chemistry (Hooper et al., 1990 Shanley and Peters, 1993), whereas others have discerned a large direct effect (Shanley and Peters, 1988 Peters and Ratcliffe, 1998). This apparent discrepancy probably results, in part, from temporal and seasonal variations in runoff, and from differences in the spatial scale at which these studies were conducted. Additionally, the effects of direct runoff from the outcrop on stream chemistry may depend on the combined effects of the size and intensity of the storm, and the antecedent soil moisture conditions. The recent addition of a hillslope trench at PMRW (Burns et al., 1998) has allowed the collection of mobile subsurface flow during rainstorms. The chemistry of this subsurface flow is distinct from groundwater in the stream riparian area (Hooper et al., 1998), thus providing an opportunity to examine the relative role of these two geographic source areas in stream runoff during rainstorms. In this study, we describe the results of a three-component model of storm runoff derived by end-member mixing analysis (EMMA) during two Copyright ��� 2001 John Wiley & Sons, Ltd. Hydrol. Process. 15, 1903���1924 (2001)

QUANTIFYING CONTRIBUTIONS TO STORM RUNOFF 1905 rainstorms in the winter of 1996. We address three questions about the response of winter runoff in this catchment. (1) What is the relative importance to stream runoff of direct runoff from the outcrop compared to hillslope and riparian groundwater? (2) What can the results from these two winter storms indicate about how runoff processes vary with storm size, rainfall intensity, and antecedent wetness conditions? (3) Are the EMMA modelling results consistent with physical hydrologic measurements in the catchment? STUDY SITE The PMRW is about 25 km southeast of Atlanta, GA, in the southern Piedmont province (Figure 1). The forested watershed is dominated by hickory, oak, tulip poplar, and loblolly pine (Carter, 1978). The 10 ha upper catchment at Panola is underlain by the Panola Granite, which is a biotite���oligoclase���quartz���microcline granite (Crawford et al., 1995). The 41 ha research catchment downstream from the 10 ha sub-catchment is underlain by the Clairmont Formation, a melange �� of amphibolite, granite, gneiss, and other rock types 84��10���35������ 30������ 25������ Piedmont Province Map Area GEORGIA Atlanta 33�� 37��� 40������ 50������ Base from U.S. Geological Survey Panola Mountain 1:1,000, 1985 260 260 250 240 230 Well 642 Well 611 Granite outcrop Riparian area Trench 270 EXPLANATION Elevation, in meters Stream Watershed boundary Stream gage Recording well Outcrop runoff Rain gage Sampling Piezometer 0 100 200 300 METERS Figure 1. Map of 10 ha catchment at PMRW, Georgia, showing locations of the sampling and monitoring sites, riparian area, and outcrop Copyright ��� 2001 John Wiley & Sons, Ltd. Hydrol. Process. 15, 1903���1924 (2001)