WATER RESOURCES RESEARCH, VOL. 22, NO. 10, PAGES 1444-1454, SEPTEMBER 1986 A Comparison of Chemicaland Isotopic Hydrograph Separation RICHARD P. HOOPER 1 AND CHRISTINE A. SHOEMAKER Department of Environmental Engineering, Cornell University, Ithaca, New York As part of the construction of a simulationmodel to test an acid precipitationneutralizationmecha- nism,the stream hydrograph was separated into its baseflow and eventwater components using stable environmental isotopes of water,naturally occurring conservative tracer. Three snowmelt events and one storm event during the winter and spring of 1984 were studied at an instrumentedwatershed in the Hubbard Brook Experimental Forest, New Hampshire. Conditions for use of the isotopic tracerwerenot alwaysmet, however. During the latter part of the snowmelt and the storm the isotopiccontentof the groundwater and eventwater werenot distinguishable. Furthermore, the isotopic content of the meltwa- ter variedconsiderably over time, thereby reducing the precision of the hydrograph separation. Frequent sampling of the meltwater is mandatory to assess this variability. Because the concentration of major cationsand anions was measured as well, chemicaltracerscould be comparedto the isotopic tracer, when the isotopichydrograph separation was reliable,to test whetherthe chemical tracer was conserva- tive. Dissolved silica was found to act as a conservative tracer for this watershed. INTRODUCTION Traditional methods for separating the hydrographinto its event water (rainwater or meltwater) and base flow compo- nents have relied upon arbitrary graphical techniques. Con- sistent with Hortonian concepts of streamflow generation, vir- tually all of the hydrographic peak was assumed to be derived from overlandrunoff of the eventwater [Linsley et al., 1982,p. 210]. When conservative tracers have been used to separate the hydrograph, however, many studies have found that water that was presentin the watershed before the onset of the storm forms more than half of the peak flow [e.g., Martinec, 1974; SMashand Farvolden,1979; Rodhe,1981]. The physical mechanisms that permit sucha quick response are not under- stood and are still under discussion [Martinec, 1975; Sklash and Farvolden, 1979; Andersonand Butt, 1982; Ward, 1984]. Regardless of the manner in which this water makesits way to the streamthe implications are quite profoundfor water qual- ity modeling. To avoid any inferenceof the runoff process involved, we will use the terms "old" and "new" water to refer to water present in the watershed before the onset of the event and event water, respectively. This terminology was employed by Pilgrim et al. [1979] and is consistent with the "time aspect" classification of runoff generation terminology devel- oped by Sklash and Farvolden [1979]. In this study we undertook a coordinatedmodeling and field effort to test an acid precipitation neutralization hypoth- esis proposed by Johnson et al. [1981]. They suggest that acid rain is neutralized in a two-step process: The incident hy- drogenion acidity is quickly exchanged for aluminum acidity followed by a slowerneutralizationby basiccations(sodium, potassium, calcium,and magnesium). Accordingto their hy- pothesis, the chemistry of the wateris determined primarily by its residence time within the soil: Water of short residence time will be more acidic and have a higher aluminum con- centration, while water of long residencetime will be less acidic and have a lower aluminum'concentration. Because •Currently at Center for Industrial Research, Oslo, Norway. Copyright 1986by the AmericanGeophysical Union. Paper number 5W4289. 0043-1397/86/005W-4289505.00 acidic inputs (either storms or snowmelt) to a watershedare event phenomena, we could translate this residence time hy- pothesisinto a hydrograph separation problem where new water correspondsto water of short residencetime and old water corresponds to water of long residence time. This paper presents the results of the hydrograph separation portion of the study for three snowmeltevents and one storm event that occurred during the winter and spring of 1984 at an instrumented watershed in the Hubbard Brook Experimental Forest, New Hampshire. Samples were analyzed for a nat- urally occurringstable isotopic species of water (HDO, deu- terated water) as well as for major cations and anions. Deu- terated water was chosen as a tracer since it is unquestionably conservative. There are disadvantages in the useof stableisotopic tracers, however.Conditionsfor their use are not met in every event, and analysis of samples is expensive. In this studythe isotopic content of the old and new water were not distinguishable in the latter stages of the melt, and variability in the new water isotopiccomponent decreased the precision of the separation. An alternate chemical tracer, dissolved silica, is considered in this paper. For the first melt event when the isotopic separa- tion was very good, the separation obtainedfrom the isotopes is compared with that obtained by assumingthat dissolved silica is a conservative tracer. Similar results are obtained in this instance. A key consideration for testing the acid precipitation neu- tralization hypothesis is the precision of the hydrograph sepa- ration. Sensitivity analyses are presented to estimate this pre- cision. BACKGROUND For a conservative tracer, a simple mixing equation CoQo + CnQn = CtQt (1) where C is the concentration of each solution, Q is the flow, and the subscripts o, n, and t referto the old component of the flow, new component of the flow, and total flow, respectively, togetherwith a mass constraint Qo+ Qn= Qt (2) may be solved for Qo and Qn.The proportion of flow that 1444