Professional Documents
Culture Documents
Figure 1. Photo of BissenQuarry site and map of regional location in Door County, Wisconsin,United
Statesof America. Pumpingwell 13 shownwith PVC casingand PVC piping to surfacepump locatedaway
from ground-penetrating
radar (GPR) surveylocation.PVC wasusedto preventelectromagnetic interference
duringsurveying.
The water table is very shallowas indicatedby the floodedquarrypits at the top of Figure
1.
7
ß Tracenumber•)
19 9
.............
ill
%
121
•41
•91
8
........ •4........ 17
•...:101
0 Meters 5 6
1997 3-D datawere aimed at optimumimagingof shallowand installationof the deepwell. Water levelwasnot monitoredin
deepsectionsof the aquifer.The 19982-D datawere aimedat the pumpingwell during the surveybecauseno spacewas
shallowimagingof water table changesproducedby the pump- availablebetweenthe pump piping and the well casing.The
ing test at well 13. pumping
ratefor the testwas3.2 L s-• (50 gpm),measured
We appliedminimalprocessing to the GPR data in order to
usinga 20-gal.barrel and a stopwatch.
preservethe relative amplitude,frequency,and phaseof the The pumping test had been planned to coincidewith dry
signal.Processing of both 2-D and3-D surveysconsisted of (1)
weather and the rental of commercialGPR equipment.How-
debias,high-pass-frequency filtering designedby Sensorsand ever, intenseovernightthunderstormsinterrupted the pump-
Software to remove low-frequencynoise induced from one ing testand raisedwater levelsto the highesteverobservedfor
antennato another;(2) traceleveling(flattening)of the cor- the site. The GPR surveywas conductedthe following day
respondingconstantantenna offset direct airwave arrival to duringa rain-free period,when the drawdownwas at approx-
correct for instrumentationtime drifts; (3) trace bulk time imate steadystate conditionsand the surveyarea (exposed
shiftingto the corresponding dolomitequarryfloor) wasfree of pondedwater.During GPR
time zero, definedasthe time the
transmitterbegantransmitting;and (4) amplitudegaining,us- data acquisitionwe measuredlessthan 0.01-m changein head
inga [time]2 factorto compensate for the exponential
signal levels,with an averagechangeof 0.005 m, some of which is
amplitudedecaywhile preservingrelative amplitudeinforma- attributed to water level measurement errors.
tion (i.e., amplitudevariations) along horizontal reflectors
[Tsofiias,1999]. 3.2. Data Analysis and Modeling
The GPR surveysimagedcoherentreflectorsto 240-nstwo- In low-electrical-conductivity
media, suchas carbonatesand
way travel time (approximately12 m depth at an averageve- freshwater,the amplitudeof a radar reflection at a single
locityof 0.1 m ns-•). For comparison, twoGPR profilesare planar interface is controlledby the electromagneticvelocity
displayedover the same location near wells 12, 13, and 14 contrast(a functionof dielectricconstantcontrast)acrossthe
(Figure 3). The 2-D line 1 was acquiredin 1998 duringthe interface,the angle of incidence,and the polarizationof the
pumpingtest at well 13 with the water table near the survey wave field [Straton,1941].For a thin layer betweentwo inter-
floor (lessthat 0.5-m depth below surface,approximately10 faces,suchas a fracture or a dissolution-enhanced flow path,
ns).The 3-D inline53 wasacquiredin 1997underundisturbed the reflectionstrengthis alsoa functionof layer thicknessand
hydrologicconditionsat whichtime the water tablewasat 2-m signalfrequency.Figure5 showsa plot of reflectioncoefficient
depth (approximately40 ns). Depth to the water table was versuslayer thicknessfor an analyticalmodel with the dielec-
monitored at observation wells. tric propertiesof the Byron Dolomite and a 200-MHz radar
The pumpingtest for the studywas conductedas an open signal[Tsofiias,1999].
borehole test with observation wells. To minimize the intro- We employed2-D finite differencetime domainelectromag-
ductionof electromagneticnoiseto the GPR data, there was netic modeling [Poot, 1998] to simulatethe effectsof water
no metal piping in the borehole and the pump was placed saturationchangesin a horizontalthin layer (subresolution
approximately20 m away from the pumpingwell by laying thickness)to GPR signalresponse.Model parameterssimu-
PVC pipe on the quarryfloor (Figure 1). The well casingson lated a horizontallystratifieddolomite environmentwith ap-
the sitewere primarilyPVC, but a few wellsoutsidethe survey propriateelectromagnetic wavevelocityof propagation(v) and
area had metal casings.The head levelswere monitoredat 13 conductivity(or) for the field site and the same field data
piezometersand two openboreholesbefore and after the GPR acquisitiongeometry(antennaoffsetand fracturedepth).We
survey(Figure 4). Water level was continuouslymonitored determined unsaturatedand saturated dolomite electromag-
using a pressuretransducerand data logger (Hermit 2000, netic(EM) wavevelocities of 0.11m ns-• and0.08m ns-•,
In-Situ Corp.,Laramie,Wyoming)in the uppersectionof well respectively,usingvelocityanalysisof four multioffsetcommon
19, the only deepwell at the site (Figure 2). Well 19 had a midpoint(CMP) surveys[Tanerand Koehler,1969].Electrical
packerinstalledat a depth of 11 m to allow the pumpingtest conductivities were estimated to be 0.7 mS m -• for the unsat-
to be correlatedwith previousteststhat occurredprior to the uratedand1.0mSm- • for theshallow saturatedportionof the
1224 TSOFLIAS ET AL.: MONITORING PUMPING TEST RESPONSE IN FRACTURED AQUIFER
s 2D LineI N S 3D Inline53 N
TRC1 10 20 30 40 50 60 70 80 90 TRC 90 80 70 60 5.0 40 30 20 t0
•41 I, I, tI1•3, I, I, f, I112 •.14t, I, t, [ .13, !, !, [, !. 12
• ....:.:......:.:.:
......
.:.:
...............
:...........................
:..:.::':.:?'::.:":"...;:..•
(.......
.........
:.:..:..:::..•,.
...........
.................
•..........:....;.
.........
':::.:.:.::•;'.-;
......... ' •.•;•
...........
•' -•:'•:'•'•'• :.:•:':.'e":
:.;:.
'. •'•':'":"•:'"'< ".:.:
....
•-
.. .........................
.... •:•---:...:•
........ •.::.'
.............
,.•.-:..::::::
................
•:.•:::.:---'•"-
.........
:-...''.:':.•.:.......•....::½.:,.
:.•.;..:
.............
,•..
fl ".............
::..
:.......................................
..............
:......................
-•:•.•:. •:<•:•...•..•.•?•...•:.:.,•,•,:.
..........
?•........
-
...:...
.....
•,,.
. .......................
...... -..• ...................
•..-
........................
"--.-•"-•' -- . .....:.
........................................
................... .....................
:".•:
so-'.,:':•.•,:
....
':' :-:-."-':'":-'-:'--:-'-:
....::'-::"'.
..................................
:.....'....::'...'-
"'-'":•:':--:
...........
:::::•::-':::':
..... ,.:-•':::..:.'
..::'::•.:•:"•:.-.::'::':'.:l-s0
.:::::..:•..-: ............
.............
s0]•--': :':•..'
..........
..... ...........
:"":"":';,.':':::.--..""':-'.:":
::'"':::':-:-..:'"::'"::'
'?:•.-::--•:•:'::
.......
'•"'.'
...:.... . ........ ::.• ...........
'.".•s0
•:.:•"%,•5;:::;:'•,
................
::--:•/•;
..................
":'•
..............
".?•7%:•;"'"'%;"":-'•:•'"'"•'?
......................
•............
:.1 1:•;•:;:'""•"??:':-•":•'•'"
..............................................
•"?:::;:.
:.':'":•%'
.................................
.:•'.....................................
•:•:':•:.•:'•:
•
......
:.......:':.:....::.::..:::::?:::..::•?:-.:.:.:.:::...::.::.::.•:..:.:.:.:.:
--•%.:.:..;:.:.:..::::::::::T.
•?.::.::...:..:.:::..::.::.....?..:...?:•
....f2......
[::.:
.................
::
..........
::,.::....":,..:..".'":':
...........................
':..:
:.:.::?'........:.:::.:.....•,...:.:
;•
..............
.......................
:.:.•:
•
...................
'.•'
:...:.:;:3
.........
-........
,•.•.,••..
...................................
ß-.:...: ............
:.:.:-:• :.............
...........
.:::•:.:'::•;
...... •..:.
'".....
"" •:•
:•,_•
::• .................................
•:,.:•--<:'"'....'
......
----?'---,•-•,-
lOO
:.•.:•.
..................
.•:..:•;:.:..•.:...•:
•..•... .................
::;;:•.'
;.-;•:•.
.......
ß...........
• ..........
•:<
............... ..................................................................................
'.'•........ -.:.-... ."•..................... : ..
..}•.•:• ..............
,::.---.-.•-•--.•..•-.•••••••••
:•:• • '"-"•"•-½•:•.•:•:• ,:..... ; ..................
••.•<•:.:•;•..::.•.......
;-• ......
....• .::....-..
..........
:..:.•,•::•..:,..•...•
•....:
...................
•:.•--..:.•.::.:.•:
:•¾
.......
.....-..•......_•?'-•:
..........
.,-.....• :•.:•:.::?:.½•
'•-•:•- .•..•--.•.•.•..-...•..•.•.•.•.....•
..<•••..:.;•:::::..:.,•¾S.-•{•:<•½•.:•%:...;•..,.:
..... ... .
..... Pumping..
'.........
, !!i::[•!._ i+.:.iiiii•i-•.:
i:' .....
:•.-...::..
Figure 3. GPR profilesdisplayingthe samesectionthroughwells12, 13, and 14. Trace numbers(TRC) and
well surfacelocations(circles)are identifiedon eachprofile. Caliper and resistivitylog curvesare displayed
alongthe well bores.On the left is the 19982-D line 1 collectedduringa pumpingtest at well 13. On the right
is the 19973-D inline 53 collectedunder undisturbedhydrologicconditions.Continuousstrongradar reflec-
tors correspondto knownflow paths(horizontaldiscontinuities fl, f2, and f3 and dissolutionzone zl).
aquifer basedon publishedrangesfor carbonates[Davisand rated with an upper air-filled layer and a lower water-filled
Annan, 1989] and the low permeability of the rock matrix. layer.The EM parametersof air andfreshwaterare (1) air rr =
Electrical conductivityvariation of 1 order of magnitudedid 0 mS m- • and •, - 0.3 m ns- •' and freshwater rr = 1.0 mS m- •
not yield significantchangesin the modelingof shallowinter- and • = 0.033m ns-• [DavisandAnnan,1989].The model
faces; thus the values selected for the model are considered employeda 1-cm grid size in two dimensions,a 0.01-ns time
adequate.The horizontalfracturewasunder three conditions: step,a 200-MHz Gaussiansourcepulse(wavelet),and a thin
(1) air filled; (2) freshwatersaturated;and (3) partially satu- horizontal layer thicknessof 2 cm.
31:
13 WeftNo. 7 5 a 6 ....... 19'" 8 ......2....
-0
Fracture f 1
dl
zl ............
o 25'
f2 . •, ...... -o
• f3 6o
d2 ......................
8 e•
,, • •.•.Monitodng
Intervals
19 10
0 10 20 30
4. Results
,
GPR profilesacquiredover the samelocationprior to and
0 0.01 0.02 0.03 0.04 0.05 duringthe pumpingtest are shownin Figure 3. Both radar
Layer
Thickness
(m) profiles
display
thesame
subsurface
features
consisting
ofhor-
izontallylayeredcarbonatefaciesand horizontaldiscontinui-
ties (fracturesand a diageneticzone). Caliper log maximum
and resistivitylog minimum deflectionsdisplayedalong the
wells identify the locationsof known horizontal fracturesfl
(1.35-m depth), f2 (4.25 m), and f3 (6.4 m) [Muldoonand
Bradbury,1996].Well coresidentifieddiageneticzonezl (3.35
Figure 5. Plot of reflectioncoefficientversuslayer thickness m) [Muldoon and Bradbury, 1996]. These high-hydraulic-
for a 200-MHz normal incidentplane wave to a saturatedand conductivityflow paths correlateto increasedamplitudecon-
unsaturatedthin layer of varyingthicknesscontainedwithin a tinuous radar reflectorsat 24-, 75-, 92-, and 137-ns two-way
rock matrix with the dielectricpropertiesof the Byron Dolo- traveltime (Figure3). Below35 ns,the two profilesare nearly
mite [Tsofiias,1999]. identicaland exhibit small amplitudedifferencesthat are at-
tributed to different antennaoffset, different electromagnetic
wavefield orientation,differentwater table depth,and the use
The pumping test analysisused the Neuman [1975] type- of differentradar systems (althoughof the samemodel). Small
curvemethodfor an unconfinedaquifer to obtain an estimate vertical time shifts of horizontal reflectors between the two
of transmissivity.Becauseof possible well borestorageeffects, profiles(up to 3-ns delayobservedin the shallowsectionon
variationsin pump speed,and thunderstormsin the early por- 3-D inline 53) are causedby the differentantennaoffsetsand
tion of the pumpingtest,the recoverysegment(1375-2660min associatedreflectiontravel pathsnot compensatedfor in pro-
on Figure 6) of the testwasusedfor analysisof the drawdown cessing.These effectsdecreasewith depth as travel pathsap-
data. Using thesedata, two steadystate modelsfor the ob- proachnormalincidenceto reflectinginterfaces.Additionally,
servedwater levelelevationsin the system(at about1300min small velocity differencesproducedby different water table
on Figure 6) were utilizedto determinewhichhorizontalfrac- depthscontributeto minor reflectortime shiftingat this low-
turesdrainedduringthe GPR survey.A confinedinfinite aqui- porositysite.Note that the top of the saturatedzone doesnot
fer model and an unconfined model, which assumed that the yield a radar reflection, although velocity analysisof CMP
drainageditch at the site acted as a hydraulicboundary22 m surveys indicated
velocities of 0.11m ns-• and0.08m ns-• for
the shallow(unsaturated)and deeper(saturated)sectionsof
the aquifer,respectively.The lack of a reflectoris due to the
29.00 lack of a sharpdielectricconstantcontrastat the unsaturated-
saturatedinterface in the low-porositydolomite matrix. It is
GPR Survey
the secondaryporosityfeatures(fracturesand the diagenetic
zone) that give rise to strongreflectors.At the level of the
p.•?
On • •Recharge
Event
28.75
28.00
Fracture
f1•.f••'-•-•
Pump
Off ,
tudesat the pumpingwell. Althoughreflectoramplitudevalues
cannotbe compareddirectlybetweenthe 2-D and 3-D profiles,
0 500 1000 1500 2000 2500 the reflectorvariabilityin eachprofile can be usedto identify
Time (minutes) differencesbetweenthe two profiles.The differencebetween
Figure 6. Water levelsin open borehole19 duringpumping the shallowsectionsof the two profiles(pumpingtest versus
test. The water level early in the pumpingtest drainedbelow undisturbed)showsthat the pumpingtest introducedchanges
fracture 1, but subsequentprecipitationcausedthe water level in the electromagneticpropertiesof the aquifer at the 24-ns
to rise above fracture 1 in well 19. region.We interpretthesechangesas the resultof changesin
1226 TSOFLIAS ET AL.: MONITORING PUMPING TEST RESPONSE IN FRACTURED AQUIFER
2D Line 1
S 13 N
TRC 1 10 20 30 40 50 60 70 80
0
lO o
20
0.)30 ß
E '5'
Figure 7. The 2-D line 1 duringpumpingtest at well 13. Singletrace insertsof the modeledwaveformsfor
saturated,unsaturated(alsoreferredaspartiallysaturated),anddrainedfracturemodelsare shownalongwith
the correspondingrecordedwaveformsoutlinedby open boxes.Refer to text for detailedanalysis.
the water saturation of horizontal fracture fl. Furthermore, pumpingwell 13 (traces42-48), wherethe fractureis expected
the observedwaveformchangesare not symmetricabout the to be drained, the observed radar waveform has the lowest
pumpingwell 13. This suggests that the drainagearea of this amplitudeand is stretchedand delayedin time, similarto the
shallowhorizontalflowpath is not concentricaboutthe pump- waveformpredictedby modeling.North of the pumpingwell,
ing well. Similar amplitudepatterns(i.e., high amplitudesto signalamplitudesremain at low levels,and waveformsappear
the south,lowestamplitudesat the pumpingwell, and inter- delayedin time and stretchedcomparedto saturatedand par-
mediateamplitudesto the north)wereobservedin all four 2-D tially saturatedwaveforms.These are also interpreted as re-
radial linescollectedduringthe pumpingtest. sponsesfrom nearly drained sectionsof the fracture. Similar
Analytical computationsof the magnitudeof reflectionco- amplitudeandwaveformvariationsare presentin all four 2-D
efficients,for saturatedand dry layers of varyingthickness radial lines.
enclosedin dolomitematrix,indicatethat significantly stronger When the aboveobservations are correlatedalongall of the
amplitudereflectionsresultfrom a water-saturated layer(Fig- 2-D surveylines, an interpreteddrainagepattern for the hor-
ure 5) [Tsoflias,1999].Two-dimensional finite differencemod- izontalfracturefl canbe developed.The 2-D (pumpingtest)
eling of radar signalresponseshowsthat the strongestampli- data showsignificantamplitudevariation(0-430 •V) at the
tude corresponds to a fully saturated fracture and an 28-nspeak reflectorfor all 2-D linesaroundthe pumpingwell,
intermediate amplitude responsecorrespondsto a partially comparedto the clearlysmaller-amplitude variation(80-160
saturatedfracture(Figure7 inserts).A drainedfractureyields /xV) along 3-D inline 53 (undisturbed)(Plate 1). High and
the lowestamplituderesponseand an apparentdelayin time intermediateamplitudesare consistently southof the pumping
and stretchingof the waveform,as a result of the reversalof well, andlow amplitudesare centeredon well 13 and extendto
reflectivityand the decreasedsharpness of the reflectiveinter- the north. Waveform interpretation yields the contoured
face comparedto the saturatedmodel. Note that becauseof drainagepattern for the fracture(Plate 1). The low to inter-
uncertaintyin the modelparameters(rockmatrixand intersti- mediateamplitudetransitionat the northwesternend of line 2
tial water electrical conductivities,complex dielectric con- is used as the northernmostcontrol point of the drainedarea
stants,velocitiesof propagation,time-variantfrequencycon- contour.Note that drainageof the horizontalfracture is not
tent, roughnessof the interfaces,and fracture aperture), the symmetricabout the pumpingwell and indicatesa recharge
modeling is used to assistinterpretationrather than for a boundaryfrom the south.This drainagepattern agreeswith
precisesimulationof field conditions. the settingof the field site where a drainageditch, locatedto
Inspectionof 2-D line 1 (Figure 7) in the 24- to 28-nstime the southof the pumpingwell, providedrapid rechargeof the
interval(trough-peak)alongthe fracturereflectionshowsthe shallow horizontal fracture.
strongestamplitudesto the southof the section(traces1-20).
Intermediate amplitudesare observedsouth of the pumping 4.2. Hydraulic Data
well (traces21-33) decreasing to minimumamplitudesaround The monitoringdata at well 19 (Figure6) indicatedthat the
the pumpingwell (traces42-48). Small-amplitudelevelsand hydraulicheadat the sitewasbelowthe level of the uppermost
stretchedwaveformsare observednorth of well 13. According fracture during the early portion of the pumpingtest. Subse-
to modelingresultsthe strongamplitudesto the southof the quentrechargeraisedwater levelsat well 19 abovethe level of
radar line (traces1-20) indicatefully saturatedsectionsalong the fracture.Examiningall availablehead data indicatedthat
the fracture.The intermediateamplitudesshowinga decreas- the head distributionon the sitewasvery complex,asexpected
ing trend toward the pumpingwell (traces21-33) indicate in a heterogeneousfracturedaquifer.
partially saturatedareas along the fracture with decreasing The pumpingtesttransmissivity
of 2.5 x 10-3 m2 s-• is in
water saturationwhen approachingwell 13. Closestto the close agreement with tests conducted on the site at other
TSOFLIAS ET AL.: MONITORING PUMPING TEST RESPONSE IN FRACTURED AQUIFER 1227
2D
ßOverburden Amplitude
Quarry
Wall --- -50
100
-
150
•
3
__200•
250 •
SATURATED -- 300 c3
- 350•
_ 400 •
6
N
I I I I il•i[i
ll I
[
•,•
"ill)
0
II I
Meters
I
5
'½--
• • •mmmmlimmmmmlmm
• llll lli i i i m I Iiiiiiii i i lllli iii m I i ll!l• •i
lllll I III
'-o,,I //•/I,F•
III I I
ooded
I iii i i mI i Drainage
I lllllll illIll lllll
Ditch
i i i iii I illiii Iill!i Im
I
I I Illll
I
,,,,,... ,,,-- -IIII....
Illll III ill I I I Illllll llll I I I [
Plate 1. Amplitude of the 28-ns peak alongthe 2-D GPR data during the pumpingtest at well 13 and the
3-D inline 53 prior to the pumping test. Contour lines indicate drainage pattern interpretation for the
shallowestfracture reflector. Well locationsare given to comparewith Figure 2. Amplitude legend corre-
spondsto 2-D profilescollectedunder pumpingconditions.The 3-D inline 53 is givento illustratethe lack of
amplitudevariabilityin the prepumpingtest data.
boreholes [Muldoon and Bradbury, 1998]. The steady state tures than in a low-porosityrock matrix. Porous rock matrix
models indicate that fracture fl is drained to a maximum would require prolongedpumpingin order to drain sufficient
radius of 16 m and a minimum radius of 6 m (Figure 8). moisturefrom the pore spaceto generatedielectricconstant
Without head data closer to the borehole in the individual
contrastsdetectableby surfaceGPR systems. At BissenQuarry
fracture the radius of fracture draining cannotbe determined the lack of a water table reflector under undisturbedhydro-
with greater accuracy. logic conditionssuggeststhat even prolongedpumping might
not generate detectable dielectric constant contrastsin the
5. Discussion dolomite matrix.
... 28.6
WaterLevelinFlooded
Ditch
0.4
0.8,-.,
signalfrequency.Finite differencemodelingof the response
a 100-MHz signalto a 1-cmlayerproducedsimilarbehaviorto
the 200-MHz and 2-cm layer model of Figure 7 but with re-
of
2.0
•e
Model
uratedversusdrainedwaveformsis what is expectedfrom the
dielectricconstantcontrasts betweendolomite(8), water(80),
10 20 and air (1). Also, the stretchedwaveformof the drainedfrac-
Radiusfrom Well 13 (m) ture is a resultof the low-reflectiveinterfacecomparedto the
Figure 8. Water elevationsduringGPR surveyingandsteady saturatedfractureinterface.Althoughdetailedreview,model-
state models for confined and unconfined conditions. The di- ing,andanalysisof thin layertheoryis beyondthe scopeof this
amondsrepresentwater levelsobtainedfrom the piezometers study,we havedemonstratedthat subresolution thicknesslayer
shownin Figure 4. The elevationof fracturefl is depictedby responsecan be correlatedto physicalpropertiesand inter-
a horizontalline. The water levelin the floodeddrainageditch preted with confidencein GPR data. Suchanalysesare well
shown in Plate 1 is provided to illustrate the interpreted suitedto hydrologicstudiesbecausethe large dielectriccon-
boundarycondition.The curvesdesignatesteadystate solu-
stantof water introducesdetectabledielectriccontrastsin geo-
tions expectedfor fully penetratingwells under confinedand
unconfinedconditions.Sincethe majorityof monitoringloca- logic formations.
tionsare not fully penetratingand this is not a homogeneous Hydrogeologicdata supportingthe GPR interpretationare
porousaquifer,it is not expectedthat the curvesshouldfit the not ideal,asexpectedfor a fracturedaquifer.In suchaquifers,
data. The water level data suggesta drainedradiusfor fracture piezometersshouldideallymonitorisolatedfracturesor high-
fl of between 6 and 16 m. permeabilityzonesto determine their response.However, in
general,water level data supportthe interpretationfrom the
GPR data. The pumpingtest data, not unexpectedly, do not
plitude strengthand reflectorarrival times.Furthermore,dif- yield a simplesteadystateresponsethat is preciselycorrelated
ferencesin weather(precipitationandtemperature),different to the changein GPR amplitude.An alternativeapproachto
GPR systems,and different wave field orientationresult in providesupportfor the GPR interpretationwouldbe to mon-
additionalsourcesof amplitudedifferencesbetweenour 2-D itor fracturefl by installingseveralpiezometersin fracturefl
and 3-D surveys.Thus we comparethe relativechanges(or aroundthe pumpingwell at varyingorientationsanddistances.
lack of changes)observedalong each horizontal reflector This would require installationof three or more additional
within a survey.GPR surveysacquiredunder identicalcondi- piezometersif a priori knowledgeaboutthe drainingradiusof
the fracture were available. We do not consider this method
tions (acquisitionparameters,weather,shorttime lapse,and
equipment)shouldallowmorequantitativecomparison of sur- practical.The degreeof hydrologicmonitoringrequiredto
veysconductedbefore and duringpumpingtests. obtainthe equivalentinformationthat wasobtainedby nonin-
In analyzingshallowGPR reflectors,cautionmustbe usedin vasive GPR techniquesdemonstratesthe strengthof this
distinguishing the direct groundwave energy(transmitterto method.GPR is not proposedas a stand-alonetool for char-
receiver)and possiblesystemringingthat may be superim- acterizingthe hydraulicpropertiesof fractures,but GPR data,
posedon the reflectedenergy.In the 0.5-m offset2-D data the combinedwith hydrologicaldata, can yield insightsnot avail-
directgroundwaveis receivedat 4.5 ns,and in the 1.5-moffset able from other methods.The ability to observehydraulic
3-D data the directgroundwaveis receivedat 13.5ns,whereas boundariesand partiallysaturatedconditionswith a noninva-
the reflector of interest is in the range of 24-30 ns. System sivetechniqueare the major strengthsof this method.
ringingcanbe identifiedby comparingdatacollectedat varying
antennaoffsetsand/or correlatingreflectiondata to CMP re- 6. Conclusions
flectorswith hyperbolicmoveout.At BissenQuarryboth tech-
niqueswere employed,and no ringingwas observedin the Ground-penetratingradar was successfully employedfor
GPR data even at the small 0.5-m antenna offset. three-dimensional,
real-timemonitoringof a pumpingtestin a
The vertical resolution of radar data must be examined when fracturedmedium. Forward modelingof radar responseto
determiningthe GPR signalresponsefor thin layers.Radar varying fracture saturationidentified distinct differencesin
surveyscollectedin geologicformationswith depth-imaging signalwaveformbetweensaturatedand unsaturatedconditions
objectivesof the order of meterscommonlyemployfrequen- in fractures.GPR surveys takenbeforeand duringa pumping
ciesof 50-200 MHz with a corresponding verticalresolutionof testrecordedthe signalexpectedfor theseconditions,and the
approximately1 m to 25 cm. Fracture aperturesare highly GPR surveysare confirmedwith head measurements in pi-
variable,and for mostgeologicenvironments, exceptperhaps ezometers.In addition, GPR profiles providedinformation
karst, the aperturesare ordersof magnitudelessthan radar abouthydrologicboundaries,despiteadverseweathercondi-
resolution.EM theorypredictsthat thin layers(subresolution tions that limited the usefulness of pumpingtest data. This
thickness)are capableof introducingdetectablechangesin study demonstratesthat GPR surveyingcombinedwith hy-
radar data [Straton,1941; Tsofiias,1999]. For the simplified draulicdata can provideimprovedunderstanding of fractured
caseof a planewavenormalincidentto a layer,signalchanges formationfluid flow propertieson the well scale.
TSOFLIAS ET AL.: MONITORING PUMPING TEST RESPONSE IN FRACTURED AQUIFER 1229
Acknowledgments. We would like to thank Maureen Muldoon, variationsand water contentestimatedfrom multi-offset,ground
KennethBradbury,the staff at WGNHS, the PeninsularAgricultural penetratingradar, Geophysics, 61, 683-695, 1996.
Research Station, and Ivan Bissen.We would also like to thank Wil- Mellett, J. S., Groundpenetratingradar applications
in engineering,
liam Clement and anonymousreviewersfor their commentswhich environmentalmanagement,and geology,J. Appl. Geophys.,33,
improvedthe manuscript.Computingsupportwas providedby The 157-166, 1995.
Universityof TexasInstitutefor Geophysics. This work wasfundedby Muldoon,M. A., and K. R. Bradbury,Hydrogeology of the fractured
a GeologicalSocietyof America StudentResearchAward, the W.R. Siluriandolomiteaquifer,Door County,Wisconsinin The Silurian
MuehlbergerField GeologyScholarship Fund, the Gulf CoastAsso- DolomiteAquiferof the Door Peninsula: Facies,Sequence Stratigra-
ciationof GeologicalSocieties,the Owen-CoatesFundof the Geology phy, Porosityand Hydrogeology: Fieldtrip Guidebookfor 1996 Fall
Foundation of the University of Texas at Austin, and a National FieldConference of theGreatLakesSectionof theSEPM,pp. 26-124,
ScienceFoundationTraineeshipin Hydrology (NSF grant GER- Soc. for Sediment Geol., Tulsa, Okla., 1996.
9454098). Muldoon,M. A., andK. R. Bradbury,Tracerstudyfor characterization
of groundwatermovementand contaminanttransportin fractured
dolomite,Final Rep., WOFR 98-2, Wis. Geol. Nat. Hist. Surv.,Mad-
References ison, 1998.
Neuman,S. P., Analysisof pumpingtestdata from anisotropicuncon-
Beres,M., Jr., and F. P. Haeni, Applicationof ground penetrating fined aquifersconsideringdelayedgravityresponse,WaterResour.
radar methodsin hydrogeologic studies,GroundWater,29, 375-386, Res., 11, 329-342, 1975.
1991.
Poot, H., 2D finite differencetime domainmodelingfor GPR, Rep.
Birken, R., and R. Versteeg,Use of four dimensionalGPR and ad- TA/TG 98.06 CTG/TG 98.05, Delft, Univ. of Technol., Delft, Neth-
vanced visualization methods to determine subsurfacefluid flow, in erlands, 1998.
SeventhInternationalConference on GroundPenetrating
Radar, vol. Straton,J., A., Electromagnetic
Theory,McGraw-Hill, New York, 1941.
1, pp. 377-381, Univ. of Kans.,Lawrence,1998. Taner, M. T., and F. Koehler,Velocityspectra-digital
computerderi-
Bradbury,K. R., and M. A. Muldoon,Hydrogeologyand groundwater vation and applicationsof velocityfunctions,Geophysics,
39, 859-
monitoringof fractureddolomitein the Upper Door PriorityWa- 881, 1969.
tershed,Door County,Wisconsin,OpenFile Rep., WOFR 92-2, Wis. Thiem, G., Hydrologische Methoden,Gebhardt, Leipzig. Germany,
Geol. Nat. Hist. Surv., Madison, 1992. 1906.
Brewster,M. L., and A. P. Annan, Ground-penetratingradar moni- Tsoflias,G. P., and J. M. Sharp, Three-dimensionalhydrogeologic
toringof a controlledDNAPL release:200 MHz radar, Geophysics, characterizationof fractured carbonate aquifers using ground-
59, 1211-1221, 1994. penetratingradar, Trans.Gulf CoastAssoc.Geol.Soc.,XL(VIII),
Davis, J. L., and A. P. Annan, Ground penetratingradar for high- 439-447, 1998.
resolutionmappingof soilandrockstratigraphy, Geophys.Prospect., Tsoflias,G. P., Hydrogeologiccharacterization of fracturedcarbonate
37, 531-551, 1989. aquifersemployingground-penetrating radar, Ph.D. thesis,Univ. of
Dubois,J. C., Boreholeradar experimentin limestone:Analysisand Tex. at Austin, 1999.
data processing,FirstBreak,13, 57-67, 1995. van Overmeeren, R. A., GPR and wetlands in the Netherlands, in
Endres,A. L., D. L. Rudolph, and W. P. Clement, Monitoring of a SeventhInternationalConference on GroundPenetrating Radar, vol.
pumpingtest in an unconfinedaquifer with ground penetrating 1, 251-256, Univ. of Kans., Lawrence, 1998.
radar,in Symposium on theApplicatiohof Geophysics to Engineering
andEnvironmental Problems,vol. 1, pp. 483-492, Environ.and Eng. T. Halihan, Schoolof Geology,OklahomaStateUniversity,Stillwa-
Geophys.Soc.,Reno, Nev., 1997. ter, OK 74078.
Endres,A. L., W. P. Clement,andD. L. Rudolph,Groundpenetrating J. M. Sharp Jr., Departmentof GeologicalSciences,Universityof
radarimagingof an aquiferduringa pumpingtest,GroundWater,in Texas at Austin, Mail Code Cl101, Austin, TX 78712.
press,2000. (jmsharp@mail'utexas'edu)
Fetter, C. W., AppliedHydrogeology, 3rd ed., Macmillan,Indianapolis, G. P. Tsoflias,MineralsManagementService,U.S. Departmentof
Indiana, 1994. the Interior, 1201 Elmwood Park Blvd., New Orleans, LA 70123.
Geier, J., W. S. Dershowitz, P. C. Wallmann, and T. W. Doe, Discrete (George.Tsoflias@mms.gov)
fracturemodellingof in-situ hydraulicand tracer experiments,in
Fracturedand JointedRock Masses,pp. 511-518, A. A. Balkema,
Brookfield, Vt., 1995. (ReceivedOctober18, 1999;revisedAugust31, 2000;
Greaves,R. J., D. P. Lesmes,J. M. Lee, and M. N. Toksoz,Velocity acceptedSeptember21, 2000.)