Effects of slope vegetation removal on the ... - Wiley Online Library

0 downloads 0 Views 1MB Size Report
The effect on removal of lower, mid, and upper slope vegetation on the diurnal variation in ..... does not explain the changes in diurnal variation caused by ... the lag goes to zero. .... tallying up the number of positive and negative residuals be-.
WATER RESOURCES RESEARCH, VOL. 33, NO. 2, PAGES 321-331, FEBRUARY

1997

Effectsof slopevegetationremovalon the diurnal variations of a small Leon

mountain

stream

J. Bren

CooperativeResearchCentre for CatchmentHydrology,ForestryDepartment,Universityof Melbourne Creswick, Australia

Abstract. The effecton removalof lower, mid, and upper slopevegetationon the diurnal variation

in streamflow

from a 46-ha catchment was observed. The diurnal variation

in

streamflowof the small streamwas measurableduring the late-spring-to-late-autumn period. The amplitudein streamflowvariation reacheda maximumin early summerand declinedduring autumn.Observationof diurnal variationsduring the periodsof higher flow in winter and springshowedthat they may occurbut were maskedby much larger variations

associated with storm runoff. Simulation

of the characteristics

of the flow

measurementsystemshowedthat diurnal variationscan only be studiedusingV-notch weirsand float recordersduringperiodsof low flow. No effectof the clearingof slope vegetationon the phaseof the outflowcouldbe found. However,there was evidenceof a

significant increase in amplitude, probably dueto increased groundwater outflowfromthe slopes.It was concludedthat the diurnalvariationis due to transpirationby the riparian and near-riparianvegetationonly,and that the lowerto mid slopevegetationplayslittle

rolein thisvariation.Simulations suggested thatincreased amplitudeis associated with increasedflow rates, and that the amplitude is not directly affectedby water use of vegetationon the catchmentslopes.It was concludedthat the amplitudeof the variationis insensitiveto changesin slopehydrologyand cannotbe usedto provideinsightinto deep slopeprocesses. Introduction

Diurnal variationsin streamflowof large and small streams emanatingfrom forestedareashavebeen noted frequentlyin hydrologic literature from the days of Godwin [1931] at Wicken Fen. Analysesof the effectsof land use changeon the characteristicsof this variation are less common, although changeis sometimesnoted in passing[e.g.,Lawrence,1990]. This paper usesdata from 8 yearsof the CropperCreek multiple-catchmentstudyto characterizethe amplitudeand phase of the diurnal variation

in streamflow

before and after a treat-

mentwhichremovedthe slopevegetationbut left the riparian vegetationunchanged. The work wascarriedout aspart of the completeevaluationof the resultsof the Cropper Creek HydrologicProject [seeBrenand Papworth,1991].At the time of this analysisit was consideredthat changesin the diurnal variation as a result of the catchmenttreatment could give insightinto deep slopehydrologicprocesses. The aim of this work wasto investigatethe changesin the diurnal variation of streamflowas a result of slopevegetationdestructionand to examinewhat inferenceson deep slope processescould be made.

The Cropper Creek Project An accountof thismultiple-catchment projectand the early treatment effectson water yield and streamflowis given by Bren and Papworth[1991].The projectis locatedin a humid zone of Victoria approximately200 km northeastof MelCopyright1997by the AmericanGeophysicalUnion.

bourne,Australia(Figure 1). The neighboringClem and Ella Creekswere gaugedfrom 1975to 1987usingconventional120ø sharp-edgedV-notch weirsfitted with stream-heightrecorders and a network of recordingand nonrecordingrain gauges.In December1979the slopevegetationof Clem Creekwatershed (46 ha) was convertedfrom native eucalyptforest to radiata pine plantation.This conversioninvolvedcompletedestruction of the slopeforest,burningof the debris,and plantingof the slopeswith radiatapine.The streamsideswere protectedby a retained strip of native forest 30 m either side of the stream. Figure 2 givesan obliqueview of the catchmentafter treatment andshowsthe retainedriparianstrip.Figure3 givesa plan and crosssectionof the watershedshowingthe extent of the vegetation destruction.The adjacentwatershed(Ella Creek, 113 ha) retainingthe native eucalyptforestwas usedas a control for this study.Ella Creek often ceasedflowingin hot summer weather,in contrastto Clem Creek which alwaysflowed.This differencemakesdirect comparisonsof the diurnal flow variationsimpossibleover the full year.Brenand Papworth[1991] showed that Ella Creek formed

an excellent control for Clem

Creek for all but low flows.

Bren [1979] andBrenand Turner[1986]give an accountof the hydrologyof the Clem Creek watershed.Among other work, three gaugingstationswere installedupstreamof the main measurementweir, and measurementsof groundwater pressure,phreatic surfaceslope, groundwaterinflow to the stream,overlandflow, and streamchannelpropagationproperties were made. The results of this work showed the follow-

ing: 1.

Groundwater

inflow to the stream from the watershed

slopesvia a water table aquifer rechargedby soil infiltration was alwaysthe dominanthydrologicprocess.The aquifermaterial wasthe weatheredsedimentaryrock underlyingthe po-

Paper number 96WR02648. 0043-1397/97/96WR-02648509.00 321

322

BREN:

EFFECTS

OF SLOPE

VEGETATION

REMOVAL

A 65 7•

250

110

0

110

Kilometers

0

50• '•

/!

'

•g

• •

'

250 Meten

x



•t•ned eucal•t

..•

Forest Removal

'•

Cross-Section

AB

Figure 3. Plan and crosssectionof Clem Creek catchment, to scale,showingforest before and after clearing.

Figure 1. Location of the Croppers Creek multiplecatchmentproject and placesreferred to in the text.

stream.This supportedthe use of Dupuit-Forcheimertheory in groundwatermodelingas an approximationof the groundwater behavior.The water table closeto the streamhad grarous clay loam soil. Overlandflow from the slopeswas negli- dientsrangingup to 25ø (or more), but generallythey were about 15ø during winter, reflectingthe nature of this steep gible [Bren,1979]. 2. Direct measurementof groundwaterlevelscloseto the mountainterrain(slopesof up to 30ø).In prolongeddry periods stream showeda positivecorrelationbetween the phreatic the phreaticsurfaceslopenear the streamwasa closeto zero. 3. Clem Creek commencesas the outflowfrom a springin surfacegradientand the volumeof groundwaterenteringthe

Figure 2. Oblique view of Clem Creek catchmentafter clearingand conversionto radiata pine The streamflowmeasurementweir is just abovethe lower limit of clearing.The retained riparian strip can be clearlyseen.

BREN:

EFFECTS

OF SLOPE

VEGETATION

REMOVAL

323

16-

'7,

12

o

'-•

8

E

O

4

,



,

,

,

I

I

i

24

o

E E

.__

Jan1

i

Jan•

,t. Jan 1

I Jan 16

, L Jan 21

Jan 26

Jan 31

1981

Figure 4. Example of the diurnal variation for January1981.

a watershedconcavity.Diurnal variationwasnonexistentin the springoutflow, but was detectableand displayedincreasing amplitudewith increasingdistancedownstream[Bren,1979]. Using the Dupuit-Forcheimerformulation of groundwater flow, castin a coordinatesystemthat approximatedthe geometry of the watershedcomponents,explainedthe generalform of the observedhydrographs,includingthe differencesin the behavior

of the stream head and the watershed

storm the differential

between

the cleared

catchment

and the

controlhad disappeared.It washypothesizedthat at this time the slopestoragesof bothwatersheds were similarand that the propertiesof the watershedmaterial were limiting the rate of water transmissionto the streams.The analysisshowed no changein summerflows as a result of the clearing.

flanks and the

sensitivityof stormflow (as definedby Hewlettet al. [1984]) to antecedentflow. The simulationssuggested that the stormflow responsewas functionally dependent on the slope of the groundwatersurfaceat the start of the storm.This was borne out by field studiesthat showeda statisticaldependenceof the storm flow responseon the antecedentflow and that the antecedentflow was in turn positivelycorrelatedwith phreatic surfacegradientsnear the stream.A full accountis given by Bren and Turner[1986]. Observationof the native eucalyptforestson the catchment suggeststhat the riparian environmentis both well-watered and shelteredcomparedto the adjacentslopes.The mostcommon eucalyptis drought-tolerant narrow-leafpeppermint(Eucalyptusradiata Sieber), with an understorey of bracken (Pteridiumesculentum Nakai). Predominantheightapproaches 40 m in the lower slopesand dropsto lessthan 20 m on upper ridges.Along the streamsthe lessxerophyticmanna gum (E. viminalisLabill.) is found, while the common fern is false bracken(CulcitadubiaMaxon). Both thesespeciesare associated with wetter

rechargethe watershedslopes.By the time of aboutthe fourth

conditions.

A full analysisof the effects of slope vegetation removal ("treatment") on water yield is given by Bren and Papworth [1991].The treatmentled to greaterstormflowsin earlywinter than expectedon the basisof comparisonwith the Ella Creek control. This responselasted for the first few major storms after the dry autumn (March-May) periods.The hypothesis developedwas that the enhancedearly winter storm flow responsereflected water stored deep within the watershed slopes,thereby reducingthe amount of water necessaryto

Diurnal

Variations

A diurnalvariationis a small-amplitudecyclicchange(of a 24-hourperiod) in flow, superimposed on underlyinglongerterm

trends.

In its most common

form

the flow

reaches

a

maximum in the early morning hours and a minimum in the late afternoon.Figure 4 showsan exampleof this variation over one calendarmonth (January1981). Fourier analysisof examplesshowa pronouncedspectralpeak with a period of 24 hoursand no other clear harmonics.Associatedaquifersalso commonlyshowa similarchangein water pressures.For small, groundwater-fed streams the variation appears linked to evapotranspirationof the watershedvegetation.Most examples cited in literature, includingthe examplesof this study, showpronounceddailyminima flow and less-pronounced daily maxima.

The variation can be viewed as small perturbationson an otherwisesmoothlycontinuousflow. Given this, the sensitivity of the measuringsystemis of importance. All V-notch or rectangular weirs show decreasedsensitivitywith increased flow. Additionally,float-typewater heighttransducersrequire a minimumwater level change("float lag") to overcomefriction and systeminertia before movingto the new level. For the 300-mm float diameter usedin this studythe lag was 1-2 mm [Leupoldand Stevens,Inc., 1972]. Thus at higher flows the ability of the systemto measure diurnal variationsbecomes doubtfulbecausethe changein weir level inducedby a flow changeis smaller than the changenecessaryto move the float up or down.

324

BREN:

EFFECTS

OF SLOPE

Diurnal variationshave often been noted in passingin the literature, but lesscommonlyhave been studied.Diurnal variationshave been used for tentative computationsconcerning water use of the vegetation [e.g., Meyboom,1967]. Godwin [1931] at Wicken (England) showedthat the magnitudeof diurnalvariationsin groundwaterlevel in a marshwasconsistent with evapotranspirationreadings in a lysimeter. Hoyt [1936] reported that the variation occurredin a Maryland stream only when the skywas clear. Troxell[1936] examined the diurnal fluctuationin the groundwaterand flow of the SantaAna River in California,andwaspuzzledthat the phase of the variation did not changewith distancedownstream. From this he inferred that the diurnal variation in the larger stream representeda bank discharge-recharge phenomenon associated with the drawdownof bank groundwaterby riparian vegetation. Wicht [1941] provided clear examplesof strong diurnal variations

in South African

streams from forested wa-

tersheds with areas between 20 and 4000 ha. He noted that the

amplitude varied from stream to stream,with season,and apparentlywith recent weather conditions.The variations laggedthe maximumevaporativedemandvariations,andthere was a correlationbetweendaily amplitudeand maximumdaily temperature.Kobayashiet al. [1990, 1995]examinedthe occurrenceof diurnalvariationson two smallstreamsin Japan.They concludedthat the variation was due to evapotranspiration from

the forest

associated

with

the stream

and not due to

VEGETATION

REMOVAL

97.5-ha foothill eucalyptwatershednear Canberra,in southeastAustralia.The experimentprovideddata on how streamflow varied when transpirationwas suddenlystoppedby severely burningthe watershedvegetation.In 1980, on a day of high flammability,the eucalyptforestwasignitedby hand and aerial ignition.The watershedvegetationwascompletelyburnt out, with crown scorchoccurringover 95% of the watershed, and all litter and understoreyburnt back to the bare mineral soil. Before burning, the stream outflow exhibiteda strong diurnal variation. This immediately ceasedon burning, and streamflowsincreasedover the nextfew daysby a factor of 3.5, although there was no rain during this period. Barometric effects were ruled out because "their

effect would

have been

displayedequallybefore and after the fire" [O'Loughlinet al., 1982,p. 135]. Both the abovestudiesusedsmallsequences of data to demonstrateeffectsand did not considerthe longerterm recovery.Similarly,Lawrence[1990], in an analysisof historicflow recordsfrom the high plains areasof the Victorian mountains(50 km southeastof the studywatershed;see Figure 1) noted that the burningof the watershedsof major streamsby extremelysevere and extensivewildfires in 1939 causeda cessationof the diurnalvariationon the hydrographs collectedfrom routine gaugingstations. Towriley[1995a] presented analytical dimensionlesssolutionsfor the responseof a phreaticaquifer to periodicforcing. One caseconsideredwas the outflow to an adjacentstream from a water table aquifer subjectto a sinusoidalperiodic rechargeover the entire aquifer. He derived a dimensionless

direct evaporationfrom the streamwater surface.They also showedthat there wasa slightlyout-of-phasediurnalvariation in the specificconductanceof the streamwater and hypothe- parameter L 2S/TP,whereL is thelengthof theaquifer,S is sized that this was due to exclusion of ions from the water the aquiferstoragecoefficient(dimensionless), T is the aquifer transmissivity,and P is the period of the fluctuation.His sotranspired. In an interestingstudy,Constantzet al. [1994]demonstrated lutionshowed thatif L 2S/TPwaslarge,thenpeakgroundwathat for losing reacheswith significantdiurnal temperature ter levelsfar from the outflowboundaryof the aquiferlag the variations, the effect of stream temperature on streambed peaksin rechargeor abstractionby one quarter of a period. seepageis a major factor contributingto reduced afternoon Towriley [1995a, Figures 8 and 9] also showed that when by streamflows.The hypothesizedexplanationwas basedon the L2S/TP is large,outflowto a streamlagspeakrecharge effectof streamtemperatureon the hydraulicconductivityof 0.125P,whereP istheperiodof flow.WhenL 2S/TPissmall, the streambed,thereby leading to increasedbed infiltration. the lag goesto zero. This factor has not appearedpreviouslyin the literature. ApIn the Clem Creek catchmentthe pretreatmentslopesalong plicationof this hypothesis to Clem Creek is restrictedby the the streamflanksapproximatedthe caseanalyzedby Towriley fact that the stream was influent and shaded.Temperature [1995a].During the catchmenttreatmentthe slopevegetation measurementstaken in the studysuggesteda diurnal temper- was destroyed,and hence the periodic componentof water ature rangeof streamflowsof lessthan 5øC,whichis consistent abstractionfor this portion of the slopewas substantially rewith influent channels[Constantzet al., 1994]. The hypothesis duced, leaving only the lowest portion unchanged. This does not explain the changesin diurnal variation causedby changesthe slopefrom one that approximatesthe model andestructionof vegetationnoted in the work of some authors alyzedby Townley to a more complexcase.Thus it might be below. expectedthat the treatmentwouldinducea phasechangeand Diurnal variation has been lessstudiedin activemanipula- that the magnitudeand directionof this couldgiveinsightinto tion of vegetation.Of particularinterestis the studyof Dunford the effectsof the treatment on the slopehydrology. and Fletcher[1947].This work was "preliminary"in scopeand was never fully reported (W. Swank,CoweetaForest Experi- Methods and Results mental Station, personalcommunication,1993). Vegetation General Comments wascut alongthe streambank(15-50 m totalwidth) of a 10-ha watershedin the Coweeta Experimental Forest. Preliminary A studyof diurnalvariationsinvolvesextractinga relatively examinationof the data immediatelyfollowingcuttingshowed weak signalfrom a noisy environmentreceivingmany other "a virtual elimination of the diurnal fluctuation during the inputs.Data analysiswasrestrictedto the period from 1975 to growingseason"[Dunfordand Fletcher,1947, p. 105]. It was 1982 to reduce any effectsassociatedwith subsequentradiata ß inferred that the vegetation immediately adjacent to the pine plantationgrowthon the slopes.Figure5 showstwo 6-day streamwas makingappreciabledemandson the groundwater sequencesof the diurnal variation for correspondingdates suppliescontributingto streamflowand that riparian vegeta- before (i.e., 1979) and after (i.e., 1981) the slopevegetation tion removal would give significantgainsin watershedwater removal. Field observation at the time the data was collected yield. showedthat in summerthe groundwatergradientwascloseto O'Loughlin et al. [1982] conducted an experiment on a zero (occasionallythe stream became effluent over short

BREN:

EFFECTS

OF SLOPE

Flow,L s -1

3.001

1979 1981

ß

2.50-

VEGETATION

REMOVAL

325

tion was taken as the sum of the maximumpositiveand negativedeviationsfrom the straightline linkingthe flowsat either end of each 24-hour period for both Clem and Ella Creeks. The set of paired values computedwas used in a residuals analysis.Thus, usingElla Creek as the control watershed,for the preclearingperiod a relation

C* = f(E)

(1)

was derived, where

2.002

C* (t)

1.50 Jan 21 Jan 22 Jan 23

Jan

Jan ;•.• I Jan 26

Time

estimateddaily amplitudeof variation in Clem Creek,L s- •' E(t) measureddaily amplitudeof variation in Ella Creek, L s-1; t days; f suitableform of the regressionequationderived usingpreclearingdata.

Figure 5. Examplesof the Clem Creek diurnalvariationfor 6 daysin preclearing(1979) and postclearing(1981). Stream Given the estimateC*, a residualfor each dailyvalue,R (in hydrographshavebeen sampledat 2-hour intervals. liters per second),can be computedas

R(t)=C*(t)-C(t) reaches)and that the "aquifer thickness"(vertical distance betweenthe appearanceof seepageon the ground and very solid rock) near the streamwas probablyof the order of a meter about 2 m laterally from the stream.The groundwater level near the stream showedan amplitude of variation of 10-20 mm. The formulationof Brenand Turner[1978]showed that the changeof streamlevelassociated with a typicaldiurnal flow variation was about 6 mm. Analysis of all flow data showedthat reliable tracesshowingthe diurnal variation can not be obtained before November of eachyear becauseof the high springflow levelsand many storms. Weir Sensitivity

The output distortionof the input given by a weir-floatrecorder systemwas modeled using the continuity equation basedon the numericalmethod of McCuen [1989] and using the float lag given by Leupold and Stevens,Inc. [1972]. The model incorporatedthe measuredweir rating and weir pool area as a function

of weir water

level. This showed that the

ability of the 120ø weirs to measurea small diurnal variation progressively decreasedto zero at flowsgreaterthan about8 L

(2)

where C is the observedamplitude.The valueR canbe viewed as being composedof a clearingeffect (posttreatment)and a regressionerror. The regressionequation derived using pretreatmentdata was

log•0C* = 0.4698 log•0E - 0.4311

R2= 0.4262 (intransformed units)

(3)

whereE is the daily amplitudeof Ella Creek flow, in litersper second.

Using the logarithmictransformationgavea better distribution of pointsfor the regressionand removedheteroscedasticity. Although the regressionwas significantat P - 0.01, there was considerablescatter. Figure 6 showsthe amplitude of variationfor Clem CreekC (t), the estimatedvalueC * (t), and the ratio of C (t)/C* (t) as a functionof time. It can be seen that there is an apparent increasein the ratio of C/C* after treatment, suggestingthat the removal of slope vegetation increasedamplitude overall. This can be further tested by tallying up the number of positiveand negativeresidualsbe-

(Table1) andapplying a X2testto the s-1 andwasverysensitive to thefloatlag.Visualexamination foreandaftertreatment of recorder charts showed that diurnal variations sometimes resultingcontingency table.The computed valueof • was could be detected

at flows far above this. In such cases the

changein weir stagewasvery small,but becauseof the sensitivity of the outflow to weir stagewith increasingstreamflow the computedamplitudeof diurnalvariationin streamflowwas quitelarge.A possibleexplanationis that float lag decreasedat higherflowsbecauseof better lubricationof the recordershaft pulleysdue to more movement.Given the unreliabilityof the measurementsystemsfor smallchangesin flowsof more than 8 L s-1, it is concludedthat the existenceof diurnalvariations at higherflowsrequiresfurther study.The insensitivityof the V-notch weirs at higher flows effectivelyobscuressuch data, and an adequateexaminationof the phenomenonat highflows is not possiblewith the weir-float-recordertechnologythen used.

Control Catchment Comparison of Effect of Clearing on Amplitude

This residualsanalysisused sequencesof 2 or more days duringwhich rain was not recorded.The amplitudeof devia-

significantat p - 0.001, leadingto the conclusionthat there wasa real, if variable,increasein the amplitudeof the variation after the treatment.

Interquartile Distance Analysis

"Interquartile distance"was usedto derive monthlyvalues representativeof the diurnal variation before and after clearing on Clem Creek. This methodwasselectedbecauseit tends to subdueerrors in "noisy" time sequences,with the major sourceof noisebeing smallrainfall events.The stepswere as follows:

1. The digitizedsequences of datawere resampledto give flow at 2-hour intervals.The resultant time sequencewas heavily smoothedby fitting a running median based on 12 valueseither side of a givenpoint and the point (i.e., 25 values). 2. A sequenceof points,c(t), was computedfrom

c(t) = r(t) - M(t)

(4)

BREN:

326

Observed

EFFECTS

OF SLOPE

VEGETATION

REMOVAL

2.00

Amplitude,

Ls-I

1.00

I

I

0.00

Estimated

2.00

AmRlitude,

Ls-I

1.00

0.00

Observed

6.0

Estimated 4.0

2.0

/'1"•i' ' 0.0

,

1975

,

1976

,

1977

,

1978

1979

1980



1981

1982

SlopeVegetation • Destroyed

Figure 6. The observedClemCreekamplitude,the estimatedamplitudegivenby equation(1), andthe ratio of observed/estimated as a function to time for the months of low flow.

wherec(t) is the deviationof the flow from the runningmedian,r(t) is the recordedflow,andM(t) is the runningmedian at time t, expressedin 2-hour increments. 3. The statisticaldistributionof c(t) for each month was computedand the value

supportsa conclusionthat there is no reductionin the amplitude of diurnal variation as a consequence of the reductionin lower, mid, and upper slope vegetation.Table 2 gives the monthly mean interquartile variation amplitude before and after treatment.

This also shows an increase in the mean am-

plitude. The pooled data for the pretreatmentand posttreatment period for January-Marchshoweda significantcorrelawas calculated,where c7s and c2s refer to the 75% and 25% tion (at p -- 0.01) between monthly mean flow and the quartiles, respectively,and each value of a(t) is for one amplitudeof variation, suggestingthat larger amplitudesof calendermonth.Use of theseinterquartiledifferencesexcludes variationwere associatedwith larger monthlyflows.The analthat therewassome largevariationsdue to rainfall.The variablea (t) is regardedas ysisof BrenandPapworth[1991]suggested increase in summer flows but was unable to show this statistia measureof the amplitudeof the diurnal variation over the cally.Thus the increasedamplitudeof posttreatmentvariation month. This method was used only for periods of relatively low may be viewed as associatedwith a small increasein summer streamflowattributableto the destructionof the slopevegetarainfall and small amounts of storm-induced recessions and tion. hencewas restrictedto the monthsfrom Novemberto May. Figure 7 showsthe amplitudeas givenby the interquartile Is There A Phase Change After SlopeVegetation Removal? method, togetherwith the monthly flow. No amplitude dimiTo determinewhetherthere wasa changein phaseinduced nutionat the time of slopevegetationremovalis evident.This by the slopeclearing,the times at which the daily minimum occurredwere usedas an indicatorof the phase.The timesat which it occurredduring the monthsof Januarywere assemTable 1. Tallying of the Sign of Residuals,C - C*, for bled for the preclearingand postclearingperiods.The null Daily Values in Pretreatmentand PosttreatmentPeriods hypothesiswas made that there was no differencein the staPretreatment Posttreatment Total tisticaldistributionsof thesedata setsand any differenceswere due to chance;this was testedusinga standardt test. 33 227 Negativeresiduals 194 136 381 Positive residuals 245 The analysisof phaseshiftin the diurnalvariationcausedby the slopeclearingshowedno evidenceof any effect.The preTotal 439 169 608 clearingand postclearingminima were achievedon averageat

a(t) = c7s- c2s

(5)

BREN:

EFFECTS

OF SLOPE

VEGETATION

REMOVAL

327

Interquartile 0.4Variation,

L/s

0.2'

0.G 20•

Monthly Mean Flow, ML

10'

1975

1976

1977

1978

1979

1980

1981

1982

Year

Figure 7. Monthly mean amplitudeof the diurnal variation in Clem Creek as shownby the interquartile variation,and monthlymean flow. Data for the low-flowperiod only is shown.

4:55 and 5:00 P.M., respectively,with standarddeviationsof about 2.4 and 1.4 hours,respectively.The examplesshownin Figure5 were typicalof the data. Given sucha smalldifference, it is concludedthat there is no evidenceto supportany hypothesisof slope clearingcausinga phase shift. Analysisattempting to use the time of Ella Creek minimum flow as a predictor of Clem Creek minimum flow also suggestedno changein phase but had a large error induced by the weir errorsat higher flowsand had a distinctlynonrandompattern of error residuals,andwashenceviewedasunreliable.Sapflow measurementsin the Melbourne Water experimentalwatersheds200 km southeastshowedthat the peak tree transpiration rate in summeris achievedbetweenmiddayand 3:00 P.M. (R. Benyon, Melbourne Water, personal communication, 1995).Thus the phaselag of the outflowappearsto be about3 hours.This comparesto the observationof maximumtemperature and lowesthumidityat 3:00 P.M. and minimum outflow at 5:00 P.M. (i.e., 2-hourlag) by Wicht[1941]in SouthAfrica. Townley's [1995a]dimensionless analysisof a slopeundergoing a periodicrechargesuggested that the maximumphaselag of

theriparianstrip,theeffective valueofL 2S/TPislargeenough to mask any changein the dynamicsof the system.Thus the absenceof change in the period suggestsa slow travel of groundwateracrossthe retainedriparian strip. Simulations

of Diurnal

Variations

To aid in interpretingthe aboveresults,a numberof simulations of periodic abstractionfrom an aquifer water table systemwere made. These used the finite element package AQUIFEM-N [Townley,1995b] and viewed the stream as a sink of thickness1 m at the toe of a low gradientwater table aquifer(Figure8) of a widthof 250m, passing waterto a reach of stream of 250 m length. This approximatedthe slopesof Clem Creek. Simulationsuseda family of differentinitial conditionsgeneratedby the equationof Singhand Rai [1989] to yield the followingexpressionfor the height of the phreatic surface:

h(x) = (2x- x2)ø'Sh(250) + 1

(6)

about3 hoursoccurred whenL2S/TP waslarge.Thiswork wherex = d/250, whered (in meters)is the distancefrom the [Townley,1995a,Figure 6b] alsoshowedthat if the value was large (>1000), then the upslope90% (approximately)of the aquifer would exert little influence on the variation on the timing of the diurnal variation in stream outflow. Thus it is

stream (maximum of 250 m); and h(250) is the height of phreatic surfaceat d = 250 m (i.e., under the ridge). The value of h(0) = 1 at x - 0 reflectsthe stream thickness.

and that evenif the slopesystemafter cuttingis viewedasonly

from regional measurementsof the fractured rock characteristics.Transpirationwas set as a stepfunction(12 hourstran-

Table 2. Mean Monthly Amplitude of the Interquartile Variation for the Preclearingand PostclearingPeriods

Hydraulicconductivity wastakenas1 m d-•, andthestorage concluded thatfor theslopeL 2S/TPwaslargebeforecutting coefficientof the aquiferwas0.05, thesevaluesbeingadopted

spiration at the rateof 1 mmh-•, 12 hoursnontranspiring);

Overall,

this would correspondto severelydryingdaysin this environment. In presentingthese simulationsone is consciousof the substantialgeometricapproximationsof the real case,the sto-

Month

L s-•

L s-•

L s-•

chastic nature of the variables involved, and the nondimen-

Dec. Jan. Feb. March

0.10 0.11 0.08 0.06

0.20 0.24 0.19 0.15

0.13 0.15 0.12 0.09

April May

0.02 0.02

0.08 0.02

0.04 0.02

Preclearing,

Postclearing,

sionalpresentationof the results.The simulationsare justified as helping to give qualitativeinsightinto the nature of the diurnal variations

in this case.

The first set of simulationslookedat whetherthe presence of a transpiringbandof vegetationof a width 10% of the slope (25 m) at the foot of a slopecouldbe reasonablyexpectedto

328

BREN:

EFFECTS

OF SLOPE

VEGETATION

REMOVAL

Evaporationat 1 mm per hourfor 12 hours,12 hourswith no evaporation. Spatialextentvarieswith simulation.

ttttt Stream head

I

always 1 m

I

]50 m

250 m

I•' d=

250

Initial phreaticsurfacegivenby z = h (2x - x^2)^0.5 where x = d/250.

K = 1 m/day,StorageCoefficient= 0.05

Figure 8. Illustrationof the slopecrosssectionandstreamreachedusedin the Dupuit-Forcheimer modelused.

have a substantialimpact on streamflow.Figure 9 showsthe slopeoutflowof simulationsof the case(the outflowsbeing scaledby the outflowon day 1). Simulationsusingdifferent initial conditionsand lesserratesof evaporationgavequalitativelysimilarresultsshowinga strongreductionin streamflow. As might be expected,the water table near the streamin the transpiringcaseis loweredrelativeto the nontranspiring case (Figure 10). The simulationsshowedthat the band of trees leadsto a greaterlossof water (transpirationplusstreamflow)

from within the slope but a reduced componentof stream outflow. The interpretationof this is that the presenceof a diurnal variation

is associated with a reduction

in outflow from

the catchment slopes because of the interaction between groundwaterlevels,transpirationrate, and streamoutflow. A second set of simulations

looked

at the relation

Relative Outflow to Stream 1.00

0.75

No transpiration 0.50

Transpiration

0.25

0.0

I I l.l 0.0

l.l 4.0

between

diurnalvariationamplitude,outflow,and extentof forestfrom the slopetoe. Valuesof h(250) usedwere 5, 10, 15, 20, and 25 m. The amplitudewas taken as that given4 daysinto the

1 Transpiration Rate

!, !8.0 ! ! I ,12.0 -,-0 mmperhour

Time (Days)

Figure 9. Simulatedslope outflow (stream inflow) for a bare slope and a slopewith the lower 25 m

transpiring at 1 mmhour-•.Outflows havebeenscaled bytheinitialoutflow.

BREN:

EFFECTS

OF SLOPE

VEGETATION

REMOVAL

329

Head Difference, m 0.20

Extentof transpiringzone 0.00

-0.20

-0.40

-0.60

-0.80

0

50

100

150

200

250

Distance from Stream, m

Figure 10. Differencebetweensimulatedphreaticsurfacelevelsfor the tree and nontreecaseof Figure 9.

simulation(see Figure 9) to allow the simulationto "settle" into the diurnalvariations.Figure 11 showsthe resultsgenerated. There was a clear dependenceof the amplitudeon the initial conditions(or initial slopeoutflow),which agreedwith the finding of a positivecorrelationbetween monthly interquartilevariationand monthlymeanflow. It is a simpleexercise to use a finite different formulation of the DupuitForcheimerequationsnear the streamto showthat for a given changein groundwaterlevel near the stream,the changein flow inducedwill be positivelycorrelatedwith the flow. Thus the noted positivecorrelationbetweenoutflowand amplitude reflectsthe nature of the groundwatersystem.The simulated amplitudeof the variationwasnot affectedby the extentof the transpiringzone from the streamprovidedthe zone was not very narrow. As the width of the zone increasedfrom 0 to about20 m, the amplitudereacheda maximumand decreased slightlyand then becameindependentof the width. The exact

causeof this is unclear but is probablyassociatedwith the severeboundaryconditionof a constanthead in the stream.It may alsoreflectthat for smalldistancesapproximatingintegral multiplesof the hydraulicconductivitythere may be a phase effectpresentwhichleadsto enhancedamplitudefluctuations. The effect,if real, is not large. The simulationssuggest that unlessthe width of the riparian strip had been narrowed substantiallyless than the 30 m adopted,there wouldnot havebeen a reductionin the amplitude. This findingis similarto the interpretationdrawnfrom the lack of changeof phaseof the amplitude.The increasein amplitude observedis most likely attributable to enhanced summerstreamflowsdue to the reductionin slopetranspiration. However,the modelingalsoshowedan "optimalwidth" for maximumamplitudeat a givendaily flow rate. These findings lead to the conclusionthat the diurnal variation is an indicatorof riparian vegetationtranspirationbut that the am-

Amplitude of diurnal

0.2variation, Ls-1

25

0.1 20

Height ofPhreatic 1

80%.

Surface under Ridge,

m

10

o• • 40o/;•$\o• e•

Figure 11. Amplitudeof diurnalvariation(taken on day4 of the simulation;seeFigure 9) as a functionof initial height of the phreaticsurfaceunder the ridge and percentageof the slopeforested(startingfrom stream).

330

BREN:

EFFECTS

OF SLOPE

VEGETATION

REMOVAL

Table 3. Summaryof Treatment Effectson the Diurnal Variation Authors

Vegetation Killed or Removed

Dunfordand Fletcher[1947] riparianvegetationonly eliminated

O'Loughlinet al. [1982]

slopesand riparianvegetationburnt by experimentalfire in small catchment

Lawrence[1990]

slopesand riparianvegetationburnt by wildfire in 1939;usedroutine gauging lower, mid, and upper slope vegetationeliminated;30 m riparian strip retained

This study

Effectson Diurnal Variation

Effectson Streamflow

almosteliminated;shortdata sequences only eliminated;slopeoutflowtrebled after burningin dry summer; short data sequencesonly

substantialflow increases increasedthreefold immediately

eliminated

generallyincreased

increasedamplitude

low flowsprobablyincreased; higher early stormflows

after burn

plitude is unlikely to be useful as an indicator of upslope of a diurnal variation is probably an indicator of reduced groundwaterconditionsor slopewater use. stream inflow due to transpirationalong its length. The inIt wasconcludedthat the simulationscouldreproducemost creased amplitude noted after slope vegetation destruction of the observed features of the diurnal variation and that the probablyreflectsthe increasedslopeoutflowdueto diminished existenceof a diurnal variation is functionallydependenton slopevegetationwater use.The amplitudeof the diurnalvarithe presenceof a stronglytranspiringvegetationalong the ation doesnot appearto giveinsightinto deep slopeprocesses. stream-slope interface.Thisvariationis superimposed on long- A diurnalvariationin streamflowprobablyoccursall year but er-termvariationsassociated with sloperecharge.The simula- is only detectableat low flows becauseof characteristicsof tions suggestedthat water balancepartitioningof the fate of common measurementsystemsand larger-amplitudevariastoredslopewater into transpirationand streaminflowis com- tions associated with storm flows. plex and is certainlya functionof the placementof vegetation on the slope relative to the stream. This is a topic that is Acknowledgments. This work is a contribution of the riparian substantiallyneglectedin studies. buffer programof the CRC in CatchmentHydrology,and the author acknowledges the leadershipof Peter Hairsine and Cathy Wilson of that organization.Much of thiswork wasundertakenwhile on sabbatical at Oregon State UniversityDepartment of Forest Engineering, and stimulatingdiscussions with Bob Beschtaof that organizationare acknowledged.Modeling of the weirs was helped by the "Extend" packageof "ImagineThat," andIan O'Neill of the Departmentof Civil

Discussion

The resultsof smallwatershedprojectsknownto the author in whichinformationis givenaboutthe diurnalvariationand in which vegetationhas been altered have been summarizedin andAgricultural Engineering provided advice onnumerical solution of Table 3. Of particularinterestin the studiesof O'Loughlinet al. the continuityequation.Lloyd Townleyof CSIRO Divisionof Water [1982] and Dunfordand Fletcher[1947] is that the elimination Resourcesprovidedboth assistancein the use of AQUIFEM-N for of the diurnal

variation

was associated with immediate

and

simulationsof slopeoutflow and stimulatingdiscussions. Refereesof draftsprovidedmuchthoughtfulhelp.

substantialincreasesin flow. Unfortunately, the resultspresentedgive no informationon how long theseeffectspersisted or how suchincreasein outflowscomparewith the reduced References evapotranspiration.The observationsare all consistentwith the field observations and modelingundertakenin this paper, Bren, L. J., The numericpredictionof hydrologicprocesseson a small forested catchment,Ph.D. thesis, 180 pp., Univ. of Melbourne, suggesting that the riparian vegetationonly producesthe diParkville, Australia. urnal variation.

Bren,L. J., andM. Papworth,Early wateryieldeffectsof conversionof

slopesof a eucalyptforest catchmentto radiata pine plantation, With the exceptionof that of Dunford and Fletcher[1947] Water Resour. Res., 27, 2421-2428, 1991. there has been little experimentalwork involvingdirect reBren, L. J., and A. K. Turner, Wave propagationin steep, rough moval of the riparian zone. The resultssuggestthat the presmountainstreams,J. Hydraul.Div. Am. Soc. Civ. Eng., 104(HY5), enceof a diurnalvariationis an indicatorthat manipulationof 745-754, 1978. the vegetationcouldgive substantialand surprisinginfluences Bren, L.J., and A. K. Turner, Hydrologicbehaviourof a smallforested catchment,J. Hydrol., 76, 333-350, 1986. on summerstreamflows.In further studiesof thisphenomenon a direct approachthat removesthe riparian vegetation and Constantz,J., C. L. Thomas, and G. Zellweger, Influence of diurnal variationsin streamtemperatureon streamflowlossand groundwameasuresgroundwaterconditionsand outflowswould be the ter recharge,WaterResour.Res.,30(12), 3253-3264,1994.

most efficient.

Dunford, E.G., and P. W. Fletcher, Effect of removal of stream-bank

vegetationuponwateryield,Eos Trans.AGU, 28(1), 105-110,1947. Godwin,H., Studiesin the ecologyof Wicken fen, 1, The groundwater Conclusions

It is concluded that the destruction of the lower, mid, and

level of the fen, J. Ecol., 19, 449-473, 1931.

Hewlett, J. D., J. C. Fortson,and G. B. Cunninghan,Additional tests on the effectsof rainfall intensityon stormflow and peak flow from

wild land basins, Water Resour.Res., 20, 985-989, 1984. upper slopevegetationhas left the diurnalvariationin slope outflow unchangedin phase and slightlyincreasedin ampli- Hoyt, J. C., Droughtsof 1934-36, U.S. Geol.Surv.WaterSupplyPap. 680, 106 pp., 1936. tude relative to its preclearingconditionand that the diurnal Kobayashi,D., K. Suzuki, and M. Nomura, Diurnal fluctuationsin variation in outflowis a consequence of riparian transpiration streamflowand in specificelectricconductanceduringdroughtperemovingwater from the phreaticaquiferin the vicinityof the riods,J. Hydrol.,115, 105-114, 1990. stream,therebyreducinginflowinto the stream.The presence Kobayashi,D., Y. Kodana,Y. Ishii, Y. Tanaka, and K. Suzuki,Diurnal

BREN:

EFFECTS

OF SLOPE

variationsin streamflowand water quality during the summerdry season,Hydrol.Proc.,9(7), 833-841, 1995. Lawrence, R. E., The interaction between the environment, land use,

and hydrologyof the BogongHigh Plainsarea from 1850 to 1985, Ph.D. thesis,798pp., Univ. of Melbourne,Parkville,Australia,1990. Leupold and Stevens,Inc., StevensWaterResources Data Book, 2nd Ed., 154 pp., Leupold & StevensInc., Beaverton,Oreg., 1972. McCuen,R. H., Hydrologic AnalysisandDesign,867 pp., Prentice-Hall, EnglewoodCliffs,N.J., 1989. Meyboom,P., Groundwaterstudiesin the AssiniboineRiver drainage basin,II, Hydrologiccharacteristics of phreatophyticvegetationin south-centralSaskatchewan, Geol.Surv.Can.Bull. 139, 64 pp., 1967. O'Loughlin, E. M., N. P. Cheney, and J. Burns, The Bushrangers Experiment:Hydrologicalresponseof a eucalyptcatchmentto fire, in The First National Symposiumon ForestHydrolog3,, 1982, Natl. Conf. Publ. 82/6, edited by E. M. O'Loughlinand L. J. Bren, pp. 132-139, Inst. of Eng. of Aust., Canberra,Australia, 1982. Singh,R. N., and S. N. Rai, A solutionof the non-linearBoussinesq equationfor phreaticflow usingan integral balanceapproach,J. Hydrol., 109, 319-323, 1989.

VEGETATION

REMOVAL

331

Townley,L. R., The responseof aquifersto periodic forcing,Adv. Water Resour., 18, 125-146, 1995a.

Townley,L. R., AQUIFEM-N: A Multi-LayeredFiniteElementAquifer Flow Model, User'smanualand description,Townley& Associates Pty. Ltd., Claremont,Australia, 1995b. Troxell, H. C., The diurnalfluctuationsin the groundwaterand flow in the SantaAna River and its meaning,Eos Trans.AGU, 17, 496-504, 1936.

Wicht, C. L., Diurnal fluctuations in Jonkershoek streams due to

evaporationand transpiration,J. S.Aft. For. Assoc.,7, 34-49, 1941.

L. J. Bren, CRC for CatchmentHydrology,ForestryDepartment, Universityof Melbourne,Creswick,Victoria 3363, Australia.(e-mail: leon_bren@muwayf. unimelb.edu.au)

(ReceivedMay 10, 1995;revisedAugust26, 1996; acceptedAugust28, 1996.)