Professional Documents
Culture Documents
Evans, Davies - 1998 - Causes of Concentrationdischarge Hysteresis and Its Potential As A Tool For Analysis of Episode Hydrochemistry
Evans, Davies - 1998 - Causes of Concentrationdischarge Hysteresis and Its Potential As A Tool For Analysis of Episode Hydrochemistry
RESEARCH,
VOL.34,NO.1,PAGES129-137,
JANUARY1998
-•- TotalDischarge
m 15
• lO
0 1 2 3 4 5 6 7 8 9 10
For a 3CM of groundwater,soil water, and surface event Elsenbeeret al. [1995a],timingsfor these components were
water components, the concentration of a conservative solute actuallyreversed.However,the assumedprecedenceof surface
in streamwaterat a given time is describedby the following eventwater is consistent with the variablesourceareaconcept,
massbalanceequation: in that flow from direct precipitation onto saturated areas
peaksbefore subsurface flow [Dunneand Black, 1970;Hewlett
C•Q• = CoQo + CsoQso+ Cs•Qs• (1) and Nutter, 1970]. This systemis likely to operatewidely in
where C is concentration,Q is discharge,and the subscripts T, humid forestedbasins.It alsoseemsreasonableto expectsoil
G, SO, and SE representtotal streamflow,groundwater,soil water responseto be lagged,giventhat a large proportionis
water, and surfaceeventwater, respectively.The methodsde- likely to derive from hillslope areas, away from the stream
veloped here require only that C•: and Q•: are measured. network, and that conditionsof saturationmust usuallyde-
Provided an event follows a period of low flow, Ca can be velop (becauseof either a risingwater table or the develop-
approximatedby the pre-eventC•:. Three-componenthydro- ment of perchedsaturation)before significantflow can occur
graphseparationsrequiremeasurementOf the two "stormflow [Lynchand Corbett,1985;Swistocket al., 1989;Hooperet al.,
components,"C so and C SE, but these are treated here as 1990].Groundwaterdominanceis consistentwith ridgingthe-
unknowns.Sincecomponentdischargesare alsounmeasured, ories[SklashandFarvolden,1979;Gillham,1984]or a simpler
precisevaluesfor C so and CSEcannotbe calculated.However, mechanismof old water displacement.The sensitivityof the
by makingcertainassumptions regardingthe natureof hydro- model to changesin assumptions is consideredbelow.
graph response,it is possibleto obtain estimatesfor C so and A simpleconceptualthree-component hydrograph,basedon
C SE relative to Ca and to each other so as to determine this flow model, is shownin Figure 1. The hydrographis the
"componentrankings." hypotheticalresponseto a single-peaked rain input. This was
An empiricallybased runoff sequenceis assumedfor this used to generate a set of C/Q plots based on a range of
study.It is essentiallythat describedfor MahantangoCreek, different componentconcentrations.Componentswere as-
Pennsylvania,
Wherea typicalprogression during signedconcentrationsof 50, 100, and 150 arbitraryconcentra-
of dominance
stormflowis believedto be (1) baseflow, (2) channelprecipi- tion units in a total of sixpossiblecombinations.
tation and SOF (surfaceeventwater), (3) shallowsubsurface
flow (soilwater), and (4) baseflow [Pionkeet al., 1988;DeWalle 2.2. Interpretation of C/Q Plots
andPionke,1994].Other studiesin Pennsylvania by Swistock
et C/Q plots for each combinationof componentconcentra-
al. [1989]andFulcar [1990]haveobservedthe samesequence, tionsare shownin Figure 2. It is apparentthat eachcombina-
as have Hinton et al. [1994] for a June event at Harp 4-21, tion producesa distincthysteresis loop. Clockwiseloopsare
Ontario.SklashandFarvolden[1979]foundtwo components to classedas typesC1, C2, and C3, and anticlockwise loopsare
be sufficientfor streamsin Quebec and Ontario but predict classedastypesA1, A2, and A3. From (1), CT at a giventime
that in a three-componentsystem,eventwater would be most will tend towardsthe flow componentwhichdominatesat that
important on the rising limb and vadose soil water on the time. At base flow, CT is by definition equal to Ca; on the
falling limb. Consistentwith theseand other studies,ground- risinglimb it tendstowardsC SE,andon the fallinglimb it tends
water responseis assumedto be large and (as a result of its towardsCso. If, for instance,CSE> Ca > Cso, CT will peak
overalldominance)closelycorrelatedto total discharge. on the risinglimb and reach a minimumon the falling limb,
It is acknowledged that this model may not be universal:at generatinga type C1 loop.The oppositesituation(Cso > Co
Allt A'Mharcaidh, Scotland,Q sE and Q so peakstend to co- > CSE) will produce the equivalentA1 type anticlockwise
incide[Ogunkoya andJenkins,1993;Jenkinset al., 1994],andin loop. If C SEand C so are both either higheror lowerthan Ca,
one of two eventsmonitoredat South Creek, Queensland,by that is, Ca is "extreme"rather than "intermediate,"one limb
EVANS AND DAVIES: CONCENTRATION DISCHARGE HYSTERESIS 131
(a) C 1 (CSE> CG > Cso) Co)c2 (CsE> Cso > Co) (c) C3 (CG > CSE> Cso)
110
• 70
• 140
._•130
• '•2o
• 70 I • 40 0 20
i
40
c•11o 0 20 i
40
0 20 40
Discharge(arbitraryunits) Discharge (arbitraryunits)
Discharge(arbitraryunits)
(d) A1 (Cso > Ca > CSE) (e) A2 (Cso > CSE> Ca) (f) A3 (Co > Cso > CSE)
•' 80
110 • 70
100 130
120
80
0
I
2o
' I
4o
• 40 0 20
i
40
Discharge(arbitraryunits)
110
0 2040 Discharge (arbitraryunits)
Discharge(arbitraryunits)
130
plot will be intermediatebetweenconvexand concaveforms,
so that part of the loop becomeslinear (e.g., Figure 3b).
An increasein the magnitude of either Q SE or Q so will
110
causeCT to tend more towardsthe concentrationof that com-
ponent,especiallyon the limb where it dominates.An elevated
80%
proportionof flow from Q o will reducethe amountby which
CT deviatesfrom its pre-eventvalue, resultingin the "stretch-
ing" effect shownin Figure 4. Significantly,no changein com-
ponent dischargemagnitudewill affect basicshapecharacter-
istics,or subsequentinterpretation,in any way. Although a
"large"groundwaterresponsewas assumedin the model used,
0 10 20 30 40 then, the exactmagnitudeof Q o is unimportant.
Discharge(arbitraryunits) The hydrographmodel usedassumesthat QSEpeakson the
risinglimb, Qo peaksat or closeto maximumQT, and Qso
Figure 3. Set of modelled type C1 C/Q plots with varying peakson the fallinglimb. Again, any alterationin component
Q o and constantQ SE and Q so. Percentageson plot indicate
dischargetimingswhichdoesnot changethis sequencewill not
proportionof peak dischargegeneratedby Qo. QsE and Q so
are the same as in Figure 1. affectthe shapecharacteristics of modelledloops.However,if
the sequenceis altered, hysteresisforms may changesignifi-
cantly. In particular, a reversal of peaks for Q SE and Q so
would changethe rotational directionof all loops.This would
can often alsobe distinguishedusingC/Q plots. Most impor- invalidateinterpretation,althoughas discussedearlier sucha
tantly, a simple 2CM cannot generate the convexhysteresis scenariois expectedto be fairly unusual. A more realistic
types C1 and A1; this would require that CT tends towards possibility,observedby Bazemoreet al. [1994], is that ground-
valuesboth higher and lower than the pre-eventvalue. If any water maybe laggedbehindsoilwater. Given that groundwater
one solute exhibitsconvexhysteresis,this therefore suggests cannotby itself causeCT to deviatefrom pre-episodevalues,
that a 3CM is necessary.Similarly, a 2CM cannot generate the impact of this changemay be limited. However, a large
concaveloops in combinationsother than thosegiven above. delayed groundwaterinput is likely to distort C/Q relation-
If, for example, type C2 and C3 loops occur in the same ships,and the methodsdescribedshouldnot be appliedto a
episodefor two different solutes,a 2CM can againbe rejected. systemin which this situationis thoughtlikely. In practice,it
may be possibleto identifysucha systemwhere inferredcom-
2.4. Stability of Modelled C/Q Loops ponentrankingsare clearlyunrealistic,althoughobviouslythis
So far, C/Q plots have been generatedfor a single three- requiressomeprior knowledge.
componenthydrograph,using a singleset of componentcon-
centrations. It is therefore useful to consider the extent to 2.5. Effects of Temporally Varying Component
Concentrations
which plots retain a similar form if conditionsalter. Three
aspectswhich may vary are componentconcentrations,com- An inherent assumptionof the model used, and of mixing
ponent dischargemagnitudes,and componentdischargetim- modelsin general, is that componentconcentrationsremain
ings. constant.The violationof this assumptionwill only affect C/Q
The rankingof componentconcentrationshas alreadybeen formswhere componentrankingschangeduringthe episode.
identified as a critical determinantof hysteresisform. Beyond One situationin whichthismay occuriswhere solublematerial
this, however,absolutecomponentconcentrationshave no ef- hasaccumulatedprior to the event,for instancebecauseof dry
fect on the shapecharacteristics definedabove.The only situ- depositionor biologicprocesses. Flushingis likely to generate
ations where different forms will arise are special casesin high event water concentrationsat the start of the episode,
which two componentshave the same concentrations.If the followedby exhaustion[Wallingand Webb,1986].Significantly,
C SE= C so, plotswill approachthe curvilinearform shownin in a 2CM thiscantransforma typeC3 loopinto a typeC1 loop,
Figure 3a. If Co is equal to one of the other components,the which could otherwisebe produced only by a 3CM, if event
.a 70
õ•o
I '1 o 40 I I
0 20 40 60 0 20 40 6O
Discharge(arbitraryunits) Discharge(arbitraryunits)
Figure 4. Modelled C/Q loops for which two componentshave equal concentrations.All component
dischargesasin Figure 1. (a) Cso = 150,CSE= 150,andCo = 50. (b) Cso = 150,CSE-- 50, and Co = 50.
EVANS AND DAVIES: CONCENTRATION DISCHARGE HYSTERESIS 133
300
200
100
• •-125
=.E.100
O
• 75
0 .... 50 ,.
2 4 6 8 0 2 4 6
(c) Na (d) K
110 25
100 20
•- 15
80
70 =.E.
10
60 5
50- - - - ß : 0 , , ! , , i, , , , !
0 2 4 6 8 0 2 4 6
Figure 5. C/Q plots for June event, Fly Pond Outlet, Adirondacks.
30 .--. 140
20 ß 120
v
10 '•' 100
o
0 , ! , o• 80
-10 60 ,
5 10 0 5
Discharge(cfs) Discharge(cfs)
(c) Na (d) K
5O 25
2O
.--. 40
v
= 30
z 20
lO ! i 0 ! !
5 10 0 5 10
Discharge(cfs) Discharge(cfs)
ered likely that further K + is releasedfrom the watershed hysteresisis observedduring this period, but concentrations
surface.Where overlandflow hasbeen sampled,albeit at quite are uniformly low for the remainder of the year.
differentbasinsin Amazoniaand Queensland,it appearsthat Previousmeasurementsin Pennsylvaniahavebeen givenfor
K + increases
significantly
on contactwith the surface[Elsen- PeavineHill by DeWalleet al. [1988b]and for Fish Run (a
beeret al., 1995a, b]. ANC was not measuredin any of the tributaryof Linn Run) by Swistock et al. [1989].Measurements
previousstudies,but an indicationcan be gainedfrom values have also been made of shallowsoil water, throughfall, and
givenfor H +. As expected, groundwater is relativelyalkaline, baseflowby the firstauthorduringfall 1995at BennerRun (C.
and soilwater is acidic.However,the consistentlyacidicnature Evanset al., manuscriptin preparation,1997).Again, overland
of throughfallappearsto be at oddswith the high ANC pre- flow was not sampled.All three studies,like thosefor the
dictedfor surfaceeventwater from C/Q analyses.It is thought, Adirondacks, supportpredictions of highSO42-in soilwater
however, that this discrepancycan be explained by surface andhighK + in throughfall. ANC (on the basisof H + values)
interactions,particularlywithinwetlands.Hill [1993]hasshown is highestin groundwaterbut lower than expectedin through-
thatconcentrations
of Ca2+,Mg2+,andNa+ in throughfall
can fall relativeto soil water. Na+ is alsolowestin throughfall,
increaseby 35-50% within a minute of contactingwetland whichis inconsistentwith the C SE> C so rankingindicatedby
substrates.This would explainhigher than expectedANC in clockwisehysteresisat Benner Run. This againsuggests that at
surface event water. Na + results are also consistent with this BennerRun at least,surfaceeventwater is modifiedby contact
hypothesis;previousstudiesshowhigh levelsin groundwater, with the watershed surface.
but throughfall concentrationsare lower than those in soil
water. Again, therefore,somesurfaceinteractionis necessary
5. Conclusions
to explainobservedhysteresis (althoughnot presentedhere,
thesamefindings areobtained for Ca2+ andMg2+). It has been demonstratedthat C/Q hysteresistakes on one
Before comparingPennsylvania resultsto previousstudies,it of a rangeof characteristic
formsdependingon the flow system
is usefulto examinedifferences betweenthe two streams,par- in operationand takeson the rankingof soluteconcentrations
ticularlywith regardto Na+ andK +. At BennerRun, Na+ is in different components.While, as emphasizfidby Christo-
thoughtto derivefrom a groundwaterbrine source.As a result, phersenand Hooper [1992], absolutecomponentconcentra-
behavioris similar to that in the Adirondacks,where Na + is tions cannot be determined from stream data alone, the meth-
mainlyweathering-derived[Munsonet al., 1990;Evans et al., odspresentedhere make it possibleto predict relative values
1996].At Linn Run there appearto be neither major weath- basedon simplecriteria.
eringor brinesources, andNa+ is consistently low. Contrasts C/Q hysteresisobservedfor the Adirondack and Pennsylva-
in K+ hysteresis maybe explained by hydrologic differences. nia ERP streamshas been interpreted in terms of a 3CM of
At Benner Run, wetland areasgeneratesignificantSOF, and groundwater,soil water, and surfaceevent water. The hydro-
interactionwith surficialmaterialsproduceshigh surfaceevent logic model used is thought to be realistic in terms of flow
waterK + throughout theyear.At Linn Run,withoutwetlands, generationtheoriesand is consistentwith most previous3CM
surfaceeventwater inputsare thoughtto be limitedto direct hydrograph separations. In general, predicted component
channel interception [DeWalle et al., 1988a; Swistocket al., rankingsfor a range of solutesagreewell with measurements
1989].Thiswill retaina throughfallcomposition, with K + en- obtainedduringother studies.This would seemto supportthe
richmentconfinedto the growingseason.As a result,type C1 assumptions of the modelused,in that an invalidmodelwould
136 EVANS AND DAVIES: CONCENTRATION DISCHARGE HYSTERESIS
not have givenrealisticresults.Those discrepancies which do the chemicalcompositionof four streamsin the Adirondack Moun-
arise can be explainedby base cation releasefrom wetlands tains, New York, J. Hydrol., 185, 297-316, 1996.
Foster, N. W., M. J. Mitchell, I. K. Morrison, and J.P. Shepard,
during saturationoverland flow. It is also apparent that the
Cyclingof acidandbasecationsin deciduous standsof Huntington
exact nature of "surfaceevent" water may differ betweenba- Forest,NewYork, andTurkeyLakes,Ontario,Can.J. For.Res.,22,
sins;in the AdirondackstreamsandBennerRun, it is thought 167-174, 1992.
to consistmainly of saturationoverlandflow, whereasat Linn Fulcar, G. E., Determination of the componentsof stormflow in
Run it may be confinedlargelyto direct channelinterception. streamson an urbanizedand forestedbasinusingoxygen-18,M.S.
It is thoughtthat the methodspresentedhere may be appli- Thesis,98 pp., Schoolof ForestResources,Pennsylvania StateUrii-
versity,UniversityPark, 1990.
cable to a wider range of basins,althoughclearlyit is essential Gillham, R. W., The capillaryfringe and its effect on water table
that a particular basin shouldconform to the assumptionsof response,J. Hydrol.,67, 307-324, 1984.
the hydrologicmodel used.The approachshouldprimarilybe Hendrickson, G. E., andR. A. Kreiger,Geochemistryof naturalwaters
of use for those studies,such as the ERP, where component of the Blue Grassregion•Kentucky,U.S. Geol. Surv.WaterSupply
compositions
havenotbeenquantified.
Byobtaining
chemical Pap. 1700, 1964.
Hewlett, J. D., and W. L. Nutter, The varyingsourcearea of stream-
signaturesfor streamcomponents,it may be possibleto infer flow from uplandbasins,Proceedings of the Symposium on Interdis-
runoffmechanisms andpotential sources of chemicalmodifi- ciplinaryAspectsof Watershed Management, pp. 65-83, Am. Soc.of
cation during trhnsit. These resultscan provide a basisfor Civ. Eng., New York, 1970.
future sampling.The approachmay alsobe useful for studies Hill, A. R., Basecation chemistryof storm runoff in a forestedhead-
in which componentshave been identified and sampled,in water wetland,WaterResour.Res.,29, 2663-2673,1993.
Hinton,M. J., S. L. Schiff,andM. C. English,Examining
the contri-
order to verify that the model proposed is consistentwith butionsof glacial till water to storm runoff usingtwo- and three-
stream variations for a wide range of solutes,rather than a componenthydrographseparations,WaterResour.Res.,30, 983-993,
limited set of tracers. 1994.
Hooper, R. P., N. Christophersen,and N. E. Peters,Modellingstream-
water chemistryas a mixture of soilwaterend-members--anappli-
cation to the Panola Mountain catchment,Georgia, U.S.A., J. Hy-
Acknowledgments.Collectionof ERP data usedin this studywas drol., 116, 321-343, 1990.
fundedby the U.S. EnvironmentalProtectionAgency,However,the Jenkins,A., R. C. Ferrier, R. Harriman, and Y. O. Ogunkoya,A case
manuscripthas not been subjectedto the EPA's internal review and studyin catchmenthydrochemistry: Conflictinginterpretationsfrom
doesnot necessarilyreflect the viewsof the Agency;no officialen- hydrologicaland chemical observations,Hydrol. Processes, 8, 335-
dorsementshouldbe inferred,We hppreciatecommentsmadeon the 349, 1994.
manuscript
by DaveDeWalleandhisassistance,
alongwith thatfrom Johannessen, M., A. Skartveit,and R. F. Wright, Streamwaterchem-
othersat Pennsylvania
StateUniversity,duringfieldwork.We alsowish istry before, during and after snowmelt,paper presentedat Inter-
to acknowledgethe constructivecommentsof Helmut Elsenbeer, nationalConferenceon the EcologicalImpact of AcidicPrecipita-
G.I.U.B.-Hydro-Biogeochemistry,
Hallerstrasse
12, 3012Bern, Swit- tion, SNSF Proj., Norway, 1980.
zerland, and of two anonymousreviewers. Johnson,F. A., and J. W. East, Cyclicalrelationships betweenriver
dischargeand chemicalconcentrationsduring flood events,J. Hy-
drol., 57, 93-106, 1982.
References Kenrledy,V. C., C. Kendall, G. W. Zellweger,T. A. Wyerman, and
R. J. Avanzino, Determination of the componentsof stormflow
Bazemore, D. E., K. N. Eshleman, and K. J. Hollenbeck, The role of usingwater chemistryand environmentalisotopes,Mattole River
soilwaterin stormflow generationin a forested
headwhter catch- basin,California,J. Hydrol.,84, 107-140, 1986.
ment:synthesis of naturaltracerand hydrometricevidence,J. Hy- Likens,G. E., C. T. Driscoll,D.C. Buso,D. F. Siccama,C. E. Johnson,
drol., 162, 47-75, i994. G. M. Lovett, D. F. Ryan, T. Fahey, and W. A. Reiners,The
Bond, H. W., Nutrient concentrationpatternsin a streamdraininga biogeochemistry of potassiumat Hubbard Brook,Biogeochemistry,
montaneecosystem
in U[ah,Ecology,
60, 1184-1196,1979. 25, 61-125, 1994.
Christophersen,
N., and R. P. Hooper, Multivariateanalysisof stream Lynch,J. A., and E. S. Corbett,Sourceareavariabilityduringpeak-
water chemicaldata: The use of principalcomponentsanalysisfor flow:A functionof antecedentmoisturecontent,paperpresentedat
the end-membermixingproblem,WaterResour.Res.,28, 99-107, SymposiumSponsoredby Committeeon WatershedManagement,
1992.
Irrig, and Drain. Div., Am. Soc. Civ. Eng., Denver, Colo., 1985.
Cronan,C. S., Biogeochemical influenceof vegetationandsoilsin the McDonnell, J. J., M. K. Stewart,and I. F. Owens,Effect of catchment-
ILWAS watersheds,WaterAir Soil Pollut.,26, 355-371, 1985.
scalesubsurface mixingon streamisotopicresponse,WaterResour.
David, M. B., and C. T. Driscoll,Aluminiumspeciationand equilibria Res., 27, 3065-3073, 1991.
in soilsolutionsof a haplorthodin the AdirondackMountains(New
Miller, W. R., andJ. I. Drever,Water chemistryof a streamfollowing
York, U.S.A.), Geoderma,33, 297-318, 1984.
a storm,AbsarokaMountains,Wyoming, Geol. Soc.Am. Bull., 88,
DeWalle, D. R., and H. B. Pionke,Streamflowgenerationon a small
286-290, 1977.
agricultural
catchment
duringautumnrecharge,
II, Stormflow
peri-
ods,J. Hydrol., 163, 23-42, 1994. Mollitor,A. V., andD. J. Raynal,Acid precipitationandionicmove-
ments in Adirondack forest soils,Soil Sci. Soc.Am. J., 46, 137-141,
DeWalle, D. R., B. R. Swistock,and W. E. Sharpe,Three-component
1982.
tracer model for stormflowon a smallAppalachianforestedcatch-
ment, J. Hydrol, 104, 301-310, 1988a. Munson, R. K., C. T. Drisc011,
and S. A. Gherini, Phenomenological
DeWalle, D. R., W. E. Sharpe,andP. J. Edwards,Biogeochemistry
of Analysisof ALSC chemistrydata, in AdirondackLakesSurvey:
An
two Appalachiandeciduousforest sites in relation to episodic Interpretive
Analysisof Fish Communitiesand WaterChemistry,1984-
stream acidification, WaterAir Soil Pollut., 40, 143-156, 1988b. 87, pp. 2-27-2-69, AdirondackLakesSurv.Corp., Ray Brook,N.Y.,
1990.
Dunne, T., and R. D. Black,An experimentalinvestigationof runoff
productionin permeablesoils,WaterResour.Res.,6, 478-490, 1970. Ogunkoya,O. O., and A. Jenkins,Analysisof stormhydrographand
Elsenbeer,H., D. Lorieri,andM. Bonell,Mixingmodelapproaches to flow pathwaysusing a •three-componenthydrographseparation
estimatestormflow sourcesin an overlandflow-dominatedtropical model, J. Hydrol., 142, 71-88, 1993.
rain forest catchment,WaterResour.Res.,31, 2267-2278, 1995a. Oxley,N. C., Suspended sedimentdeliveryratesand soluteconcen-
Elsenbeer,H., A. Lack, and K. Cassel,Chemicalfingerprintsof hy- trations of stream dischargein two Welsh catchments,in Fluvial
drological compartmentsand flow paths at La Cuenca, western Processes in Instrumented Watersheds,
Spec.Publ. 6, editedby K. J.
Amazonia, WaterResour.Res.,31, 3051-3058, 1995b. GregoryandD. E. Walling,pp. 141-154,Inst.of Br. Geogr.,Lon-
Evans,C. D., T. D. Davies,P. J. WigingtonJr., M. Tranter, andW. A. don, 1974.
Kretser, Use of factor analysisto investigateprocessescontrolling Pionke, H. B., J. R. Hoover, R. R. Schnabel, W. J. Gburek, J. B.
EVANS AND DAVIES: CONCENTRATION DISCHARGE HYSTERESIS 137
Urban,andA. S. Rogowski,Chemical-hydrologic
interactions
in the Walling,D. E., and B. W. Webb,Solutesin river systems,
in Solute
near-stream zone, Water Resour.Res., 24, 1101-1110, 1988. Processes,editedby S. T. Trudgill,pp. 251-327,JohnWiley, New
Shanley,J. B., andN. E. Peters,Variationsin aqueous
sulfateconcen- York, 1986.
•trations
at PanolaMountain,Georgia,J. Hydrot.,146, 361-382, 1993. Wigington,P. J., Jr., J.P. Baker, D. R. DeWalle, W. A. Kretser, P.S.
Shepard,J.P., M. J. Mitchell, T. J. Scott, and C. T. Driscoll, Soil Murdoch, H. A. Simonin,J. Van Sickle, D. V. Peck, and W. R.
solutionchemistryof an AdirondackSpodosol:
Lysimetryand N Barchet, Episodic acidificationof small streamsin the Northeast
dynamics,Can. J. For. Res.,20, 818-824, 1990. United States:EpisodicResponseProject,Ecol.Appl., 6, 374-388,
Sklash,M. G., andR. N. Farvolden,The roleof groundwater
in event 1996.
runoff, J. Hydrol., 43, 45-65, 1979.
Swistock,B. R., D. R. DeWalle, andW, E. Sharpe,Sourcesof acidic T. D. Davies and C. Evans,Schoolof EnvironmentalSciences,
stormflowin an Appalachianheadwaterstream,WaterResour.Res., Universityof East Anglia,NorwichNR4 7TJ, UK. (e-mail:e273@
25, 2139-2147, 1989. uea.ac.uk)
Toler,L. G., Relationbetweenchemical
qualityandwaterdischarge in
SpringCreek,Southwestern Georgia,U.S. Geol.Surv.Prof.Pap.
525-C, C209-C213, 1965.
Walling, D. E., and I. D. L. Foster, Variations in natural chemical
concentration
of river water duringfloodflows,and the lag effect: (ReceivedOctober28, 1996;revisedMay 28, 1997;
Somefurther comments,J. Hydrol.,26, 237-244, 1975. acceptedJune25, 1997.)