u.s. department of the interior u.s. geological survey ...water resources division 10 bearfoot rd....

U.S. Department of the InteriorU.S. Geological Survey

Estimated Hydraulic Properties for theSurficial- and Bedrock-Aquifer System,Meddybemps, MaineBy FOREST P. LYFORD, STEPHEN P. GARABEDIAN, and BRUCE P. HANSEN

Open-File Report 99-199

Prepared in cooperation with theU.S. ENVIRONMENTAL PROTECTION AGENCY

Northborough, Massachusetts1999

U.S. DEPARTMENT OF THE INTERIOR

BRUCE BABBITT, Secretary

U.S. GEOLOGICAL SURVEYCharles G. Groat, Director

The use of trade or product names in this report is for identificationpurposes only and does not constitute endorsement by theU.S. Geological Survey.

For additional information write to:Chief, Massachusetts-Rhode Island DistrictU.S. Geological SurveyWater Resources Division10 Bearfoot Rd.Northborough, MA 01532

Copies of this report can be purchased from:U.S. Geological SurveyInformation ServicesBox 25286Federal CenterDenver, CO 80225-0286

CONTENTSAbstract 1Introduction 1Description of Study Area 2Estimation of Hydraulic Properties Using Analytical Methods 6

Specific-Capacity Measurements 6Aquifer Tests 10

Estimation of Hydraulic Properties using Numerical Methods 14Model Design and Hydraulic Properties 14Boundary Conditions 17Recharge and Wells 17Steady-State Calibration and Simulation 18Transient Calibration and Simulation of Well MW-1 IB Aquifer Test 21

Summary and Conclusions 25References 26

FIGURES

1,2. Maps showing:1. Location of the Eastern Surplus Superfund Site, study area, and numerical model area,

Meddybemps, Maine 32. Location of study area, extent of tetrachloroethylene (PCE) in ground water, and locations of wells 4

3. Geohydrologic sections A-A' and B-B' 54-7. Graphs showing:

4. Depth to water during recovery of water levels in selected bedrock wells after drilling and depths ofwater-bearing fractures or fracture zones 9

5. Water level for wells pumped during aquifer testing 116. Water level for observation wells that responded to pumping during aquifer testing 127. Depth to water in wells MW-1 IB and MW-22B during pumping and calculations of transmissivity 13

8. Map showing finite-difference grid, surficial geology of modeled area, and constant-head boundaries 159. Geologic section B-B' showing layers used in numerical model 16

10,11. Maps showing:10. Observed head contours for the surficial aquifer and simulated head contours for the uppermost

active layer 1911. Observed head contours for the bedrock aquifer and simulated head contours for model layer 4 20

12. Graph showing observed and simulated drawdown at well MW-10B while pumping well MW-1 IB 2213,14. Maps showing:

13. Simulated drawdown in model layer 4 after pumping MW-1 IB for 1 day 2314. Simulated drawdown in model layer 2 after pumping MW-1 IB for 1 day 24

TABLES

1. Well construction data, water levels and stage on September 9-10,1997, and wells equipped withwater-level recorders during aquifer testing, Meddybemps, Maine 7

2. Estimates of transmissivity for bedrock using specific-capacity data for wells 103. Estimates of transmissivity and hydraulic conductivity for surficial materials using specific-capacity

data for wells 104. Observed and simulated steady-state heads for wells 215. Observed and simulated drawdown in the pumped well and observation wells after pumping

well MW-1 IB for 24 hours 25

Contents III

CONVERSION FACTORS AND VERTICAL DATUM

CONVERSION FACTORS

Multiply

acrecubic foot per day (ft3/d)

foot (ft)foot per day (ft/d)

foot per minute (ft/mm)foot squared per day (ft2/d)

gallons per day (gal/d)gallon per minute (gal/mm)

gallon per minute per foot [(gal/mm)/ft)Jinch (in )

inch per year (in/yr)mile (mi)

By

04047283170304803048030480092900 00378500630902070

254254

1609

To obtain

hectareliters per day

metermeter per day

meter per minutemeter squared per day

cubic meter per dayliter per second

liter per second per metermillimeter

millimeter per yearkilometer

VERTICAL DATUM

Sea level: In this report, "sea level" refers to the National Geodetic Vertical Datum of 1929 (NGVD of!929)-a geodetic datum denved from a general adjustment of the first-order level nets of the UnitedStates and Canada, formerly called Sea Level Datum of 1929

IV Contents

Estimated Hydraulic Properties for the Surficial- andBedrock-Aquifer System, Meddybemps, Maine

By Forest P. Lyford, Stephen P. Garabedian, and Bruce P. Hansen

Abstract

Analytical and numerical-modelingmethods were used to estimate hydraulicproperties of the aquifer system underlying theEastern Surplus Company Superfund Site inMeddybemps, Maine. Estimates of hydraulicproperties are needed to evaluate pathways forcontaminants in ground water and to supportevaluation and selection of remediation measuresfor contaminated ground water at this site.

The hydraulic conductivity of surficialmaterials, determined from specific-capacity tests,ranges from 17 to 78 feet per day for wellscompleted in coarse-grained glaciomarinesediments, and from about 0.1 to 1.0 foot per dayfor wells completed in till. The transmissivity offractured bedrock determined from specific-capacity tests and aquifer tests in wells completedin less than 200 feet of bedrock ranges from about0.09 to 130 feet squared per day. Relatively highvalues of transmissivity at the south end of thestudy area appear to be associated with a high-angle fracture or fracture zone that hydraulicallyconnects two wells completed in bedrock.Transmissivities at six low-yielding (less than0.5 gallon per minute) wells, which appear to liewithin a poorly transmissive block of the bedrock,are consistently in a range of about 0.09 to 0.5 footsquared per day.

The estimates of hydraulic conductivity andtransmissivity in the southern half of the studyarea are supported by results of steady-statecalibration of a numerical model and simulation ofa 24-hour pumping test at a well completed inbedrock. Hydraulic conductivity values for thesurficial aquifer used in the model were 30 feet per

day for coarse-grained glaciomarine sediments,0.001 to 0.01 foot per day for fine-grainedglaciomarine sediments, and 0.1 to 0.5 foot perday for till. As part of model calibration, arelatively transmissive zone in the surficial aquiferwas extended beyond the hypothesized extent ofcoarse-grained sediments eastward to the DennysRiver.

Hydraulic conductivity values used forbedrock in the model ranged from 3x10~4 to1.5 feet per day. The highest values were in thefracture zone that hydraulically connects two wellsand apparently extends to the Dennys River. Thetransmissivity of bedrock used in the modelranged from 0.03 to 150 feet squared per day, withthe majority of the bedrock transmissivities set at0.3 foot squared per day. Numerical modelingresults indicated that a very low ratio of verticalhydraulic conductivity to thickness (lxlO~9days'1)was required to simulate a persistent cone ofdepression near a residential well that lies in thepreviously identified poorly transmissive block ofbedrock.

INTRODUCTION

Volatile organic compounds (VOCs) have beendetected in ground water in surficial materials andbedrock in two areas near the Eastern SurplusSuperfund Site in Meddybemps, Maine (Lyford andothers, 1998). The U.S. Environmental ProtectionAgency (USEPA) and local residents are concernedthat contaminants, principally tetrachloroethylene(PCE), can move to existing residential wells and limitfuture development of ground-water resources in thearea. Information on the hydraulic properties of theaquifer system is needed to assess the potential forcontaminants in ground water to affect residential wells

Introduction 1

and to help evaluate remediation approaches. During1997-98, the U.S. Geological Survey (USGS), incooperation with USEPA, studied characteristics ofthe fractured crystalline bedrock and the hydraulicproperties of the aquifer system near the EasternSurplus Superfund Site.

The purpose of this report is to provide estimatesof hydraulic properties for the surficial- and bedrock-aquifer system near the Eastern Surplus SuperfundSite. Estimates of transmissivity and hydraulicconductivity determined from specific-capacity dataand aquifer tests were refined by calibration of anumerical model for steady-state and transientconditions. Calibration of the numerical model alsoprovided estimates of vertical hydraulic conductivityand reinforced the conceptual ground-water-flowmodel for the study area. The study focused on the areanear the southernmost of two contaminant plumes thatis closest to existing residential wells. Thecharacteristics of fractures near the site are described ina companion report (Hansen and others, 1999).

Thanks are extended to Edward Hathaway,USEPA Project Manager, for logistical support duringthe study and to Mona Van Wart and Charlotte Smithfor access to their wells during aquifer testing. Also,thanks are extended to Madge Orchard, Terry Lord,Greg Smith, and Harry Smith for access to theirproperty.

DESCRIPTION OF STUDY AREA

The Eastern Surplus Superfund Site is on thewestern bank of the Dennys River at the outlet fromMeddybemps Lake (figs. 1 and 2). A study area ofapproximately 30 acres encompasses the 4-acreSuperfund site. The primary focus of this investigationwas on the southern half of the study area. Thefollowing description of the hydrogeology of the regionand study area is summarized from Lyford and others(1998).

The region that encompasses the study area isunderlain mainly by the Meddybemps Granite. A smallarea centered on the study area is underlain by agabbro-diorite, which is most likely a detached body of

mafic rock within the Meddybemps Granite. Surficialmaterials include till, generally less than 10 to 20 ftthick, and extensive glaciomarine deposits, includingcoarse-grained and fine-grained (PresumpscotFormation) sediments deposited during deglaciation ofthe region. The coarse gravel and sand was deposited inan ancestral sea, probably as a subaqueous fan at theice margin during retreat of the glacier. Glaciomarinesilty clay of the Presumpscot Formation underliesmuch of the lowland area in the region. A sandy faciesin the upper section of the Presumpscot Formation wasdeposited as the land rose relative to sea level and theshoreline regressed southeastward through the area.

Hydrogeologic units in the study area includetill, coarse-grained glaciomarine deposits, fine-grainedglaciomarine deposits, and bedrock. The vertical andlateral distribution of hydrogeologic units is shownin sections on figure 3. Till thickness ranges fromless than 5 ft on the western side of the Dennys Riverto about 40 ft on the eastern side. The coarse-grained glaciomarine deposits are present at or nearthe surface in the western part of the study area;thickness ranges from 0 to more than 30 ft. The thick(more than 10 ft) coarse-grained sections are largelyabove the water table. Fine-grained glaciomarinedeposits (Presumpscot Formation) are present in thecentral and southern parts of the study area wherethickness ranges from 0 to about 20 ft. The silt-clayfacies of this unit is poorly permeable and serves as aconfining layer for ground water in underlying till andcoarse-grained glaciomarine deposits. Ground water inbedrock occurs principally in fractures. The occurrenceof water-yielding fractures ranges widely; in somewells only one or two fractures supply measurablequantities of water (more than 0.02 gal/min).

Ground-water levels in bedrock wells on thenorth side of the study area respond rapidly to rainfall.Responses to precipitation in surficial materials andbedrock are subdued or are not apparent where silts andclays of the Presumpscot Formation are present. Theannual recharge may approach a potential rate of 24 to26 in. where coarse-grained materials are present at thesurface, but is probably less where till, silts, and claysare at the surface.

2 Estimated Hydraulic Properties for the Surficial- and Bedrock-Aquifer System, Meddybemps, Maine

67'2230" 67°20'

V

Meddybemps Lake-., ':&,:<:, Lakeburfacel

77V;|p :̂r-tJ$>. r---' //" />'" V- ' ' "^ I /'" '" ,•%

W x ;i /£'./ /^ ! . -. ~x .-:X- ' ^i

Base from U.S. Geological SurveyMeddybemps Lake East and Meddybemps Lake WestQuadrangles, 1:24,000, provisional edition 1987 &*•-{ \ 0 .5 1 KILOMETER

MAINE ,TOPOGRAPHIC CONTOUR INTERVAL IS 10 FEET

Map Area NATIONAL GEODETIC VERTICAL» DATUM OF 1929

Figure 1. Location of the Eastern Surplus Superfund Site, study area, and numerical model area, Meddybemps,Maine. (Modified from Lyford and others, 1998, fig. 1.)

Description of Study Area 3

EXPLANATION

ESTIMATED EXTENT OF PCE INWATER-Shows the extent of concentrationsgreater than 5 micrograms per liter {fromLyford and others, 1998, figure 12)

— — ™ ~ SUPERFUND SITE BOUNDARY

A A' LINE OF GEOHYDROLOGICSECTION-See figure 3

MW-11B MONITORING WELL AND NAME* OR NUMBER

Van Wan WELL PUMPED DURING PERIO• OF AQUIFER TESTING

67°2T35

67*21 •25-

w u*.cstoffe—Dereference

.Point

River stagereferencepoint

Van Wart

45 02'15"-

40 60 METERS

EXPLANATIONFOUNDATION OR ROC K WALL

CD BUILDINGJ~~^> TREtllNE--~ tDOt OF WATER

1 70 — TOPOGRAPHIC CONTOUR Inlerval ^ f«tNational GeutklK. Venical Dalum of 1929

Base map modified from OEST Awx Int IW7

Figure 2. Location of study area, extent of tetrachloroethylene (PCE) in ground water and locations of wells, Meddybemps,Maine.

4 Estimated Hydaulic Properties for the Surficial- and Bedrock-Aquifer System, Meddybemps, Maine

Southeast

OJ

a

<uC/5UJ

D3

120-

100 -*-*•0 50 100 150 FEET

-f-VERTICAL EXAGGERATION X 3 75

0 10 20 30 40 METERS

Northwest SoutheastB'

-1 220UJP

200

Bottom of wenatilliwJeofWft

Bottom of weDat altitude of 45 ft

0 50 100 150 FEETI 1—"-I 1 1—1

0 10 20 30 40 METERS

GLACIAL DEPOSITS

Glaciomanne deposits of thePresumpscot Formation, fine grained

— Fine sand and silt— Silt and clay

Glaciomanne deposits, coarse grainedSand and gravelSand

Glacial till

VERTICAL EXAGGERATION X 3 75

EXPLANATION

BEDROCK

Gabbro-dionte intrusive complex

Meddybemps Granite

HYDROLOGY

PotentiometncSurface InSurfical Aquifer-Measuredon April 30,1997

f WATER LEVEL INBEDROCK (B) WELL-MeasuredonApnI30, 1997

1 SCREENED INTERVALIN SHALLOW (S) WELL

VOC LIMIT OF VOLATILE ORGANICCOMPOUNDS PLUME

Figure 3. Geohydrologic sections A-A' and B-B', Meddybemps, Maine.

Description of Study Area 5

Ground water in surficial materials generallyflows towards the Dennys River. The saturatedthickness of surficial materials under the study area isgenerally less than 10 ft, and several monitoring wellsscreened in surficial materials "go dry" duringextended periods of little or no recharge. South ofHighway 191, ground water in coarse-grainedsediments is confined below silts and clays of thePresumpscot Formation, but flow also is toward theDennys River.

Ground water in bedrock flows towards theDennys River from the eastern and western sides of thestudy area. In addition, hydraulic gradients aregenerally downward from the surficial aquifer tobedrock. Water-level data also indicate a potential forflow under the river from the western side to a cone ofdepression near a residential well on the eastern side.The cone of depression may extend laterally severalhundred feet and affect water levels within a block ofthe aquifer characterized by low-yielding wells (lessthan 0.5 gal/min). In contrast, it is likely that thehigher-yielding bedrock wells outside the low-yieldingzone intersect high-angle fractures that extend to theoverlying surficial aquifer.

Plumes of VOCs, including PCE andtrichloroethylene (TCE), have been detected in groundwater in two areas. VOCs in both plumes move throughsurficial materials and shallow bedrock towards theDennys River. Contaminants in the southern plumecould potentially move through fractures in bedrock tothe local cone of depression east of the Dennys River.

ESTIMATION OF HYDRAULICPROPERTIES USING ANALYTICALMETHODS

Specific-capacity data and drawdown data fromaquifer tests conducted in wells in the study area wereused to estimate aquifer transmissivity and hydraulicconductivity by applying conventional analyticalmethods. For tests in bedrock, an equivalent porousmedium and radial flow are assumed. This sectiondescribes results of these analyses.

Specific-Capacity Measurements

The specific capacity of a well is the ratio ofpumping rate of the well to drawdown at the well. Thefollowing formula presented by Cooper and Jacob(1946) is used here to estimate aquifer transmissivityfrom specific capacity data. Application of the methodhas been described by Fetter (1994).

T -2.25Tt

'' r2S(1)

where:T = transmissivity,s = drawdown,

Q = pumping rate,5 = storage coefficient,t = time since pumping began, andr = well radius.

Application of equation 1 requires an estimate ofthe storage coefficient and an initial estimate oftransmissivity. The final transmissivity is thendetermined through a series of iterations that uses thepreviously estimated or calculated value oftransmissivity. An initial estimate of transmissivity, infeet squared per day, can be derived by multiplying thespecific capacity of the well, in gallons per minute perfoot of drawdown, times 200. The transmissivitydetermined by the specific-capacity method is,however, subject to considerable uncertainty. Inaddition to using an estimate of storage coefficient,which could be in error by a factor of 10 or more, themethod requires the following assumptions: (1) well-entry losses are negligible, (2) the pumping period issufficiently long to satisfy the requirements forapplying the Cooper-Jacob formula (r^S/^Tt < 0.01),and (3) wellbore storage effects are negligible. At aminimum, transmissivity values determined fromspecific-capacity tests provide relative values that canbe used as indices for some types of hydrologicanalyses. Specific-capacity data were collected duringwater-quality sampling (Roy F. Weston, Inc., 1997;1998a), during borehole flowmeter tests, duringrecovery measurements in bedrock wells after drilling,and during aquifer tests. Well-construction data for thewells tested are summarized in table 1.


Table 1 Well construction data, water levels and stage on September 9-10, 1997, and wells equipped with water-levelrecorders during aquifer testing, Meddybemps, Maine

[All depths in feet below land surface except water level and stage, which are in feet below the measuring point ft, feet. No , number, - no data,na, not applicable]

Well or refer-ence point

No. or name

MW-1BMW-3BMW-4SMW-4BMW-5S

MW-6SMW-7SMW-7BMW8SMW-8B

MW-9SMW-10SMW-10BMW-11SMW-11B

MW-12SMW-12BMW-13SMW 14BMW-15S

MW-15BMW-16SMW-I6BMW-17SMW-18S

MW-19SMW-20SMW-22BVanWartSmith

MeddybempsLake

Denny s Riverat RP-7

'Water level

Datedrilled

4/17/884/17/884/15/884/14/88

10/23/96

10/23/9610/28/9610/28/9610/25/9611/04/96

10/25/9611/06/96

1 1/4/9610/26/961 1/04/96

10/26/9611/4/96

10/29/9611/05/9611/06/96

11/05/9611/06/961 1/05/964/22/974/23/97

4/23/974/24/97

1950s----

na

na

Altitudeof landsurface

(ft)

2016017737174841747517986

1823417779177811673016904

1740317442174241693416969

199 1120013171 361857017846

1789718288182 181724217290

1770817857172351717817335

na

na

Altitude ofmeasuring

point(ft)

2041817989177601774318206

1847118009178751691416935

1755217613175641707017063

2002120134174 141873317932

1801118348183911743417482

1786018033174281741317455

17409

15641

in the Smith well was recovering slowly

Totaldepthof well

(ft)

5782331839713

70172

1178165

124

16523

12026

129

22138

1412038

24038

13823195

13580

49142420

na

na

Depth tobedrock orrefusal (r)

(ft)

3469

195195

r!3

r701718

H65205

r!652220

r2629

r22225

r!435

36

393638180180

11 860

1829--

na

na

Screened (s)or open-hole (0)interval

(ft)\11/

s 38-53s 13 3-23 3s 130-180s 24 7-39 7

s 10-13

045-70s 12-17

021-1178s 14-165

o 25 7-124

s 14-165s 18-23

o264-120s 21-26

o35 1-129

s 19-21 50277-138

s i 1-13 5094-120

s 26-36

o 46 9-240s 28-38

o423-140s 15-175s 16-18 5

s 9 3-1 1 8s 3 5-60

o 25 5-49039-142

--

na

na

Waterlevel or

stage belowmeasuring

point(ft)

3662896

138215481378

dry1822221212661345

dry1858189716551583

dry2672

dry14 171708

257612601339159817 10

drydry

1747915

110 10

233

385

Water-level

altitude(ft)

1675617093

163781619516828

-16157156631564815590

—1575515667154 1515480

--17462

--173 1616224

1543517088170521584715785

---

1568116498'6445

17176

15256

Equippedwith

recorderduringaquifertesting

nonononono

nonoyesyesyes

noyesyesyesyes

noyesnonono

yesyesyesyesyes

nonoyesyesyes

no

no

at the time of measurement

Estimation of Hydraulic Properties Using Analytical Methods 7

During sampling for water quality, wells werepumped at nearly constant rates that resulted inminimal and nearly constant drawdown. Pumpingcontinued until water-quality parameters measured inthe field had stabilized. Data are available for samplingperiods in December 1996, June 1997, and October1997. The pumping rate differed with well yield andwas typically less than 0.4 gal/min. Yields for wellsMW- US, MW-4B, MW-8B, MW-12B, MW-15B, andMW-16B (fig. 1) were too low to sustain a constantpumping rate. These wells were sampled after purgingand allowing the water level to recover (Roy F. Weston,Inc., 1998a). Several wells were dry or nearly dry inOctober 1997 after an extended period of lowprecipitation and could not be sampled. Specific-capacity data also were collected during borehole-flowmeter logging at the Van Wart well, during a brief(30 min) aquifer test at well MW-3B, and duringaquifer tests at wells MW-1 IB and MW-22B.

A storage coefficient of 0.1 was assumed formost wells completed in the surficial aquifer. Thisvalue is within the range commonly assumed forunconfined aquifers (Lohman, 1972, p. 8) and isconsidered to be a reasonable estimate for the short-term tests that were typically 45 to 90 min in duration.Exceptions were wells MW-8S and MW-10S, whereground water is confined by clays of the PresumpscotFormation, and wells MW-15S and MW-16S, wherethe well screen is considerably below the water table. Astorage coefficient of IxlO"4, which is within the rangecommonly assumed for confined aquifers (Lohman,1972, p. 8), was assumed for these wells. A storagecoefficient of IxlO"4 was also assumed for all wellscompleted in bedrock. This value is at the high end of arange from 5x10~7 to IxlO"4 reported for fractured-rockaquifers (Earl Greene, U.S. Geological Survey, writtencommun., 1997). The importance of the storage-coefficient estimates will be discussed later in thissection of the report.

An alternative approach to pumping was used todetermine the specific capacity of bedrock wellsMW-8B, MW-12B, MW-15B, and MW-16B. Waterlevels in these wells recovered slowly and at a nearlyconstant rate, indicating a constant inflow rate, forseveral hours to several days after they were drilled(fig. 4). Conceptually, when the water level in a welldeclines below the level of a bedrock fracture,drawdown at the well bore is effectively constant andflow to the well would be expected to gradually decline(Jacob and Lohman, 1952). A steady inflow rate while

water-bearing fractures are above the water level in thewell indicates either that a constant head source, suchas leakage from a surficial aquifer, was controlling theinflow rate, or that sufficient time had elapsed sochanges in discharge were very slow and notdiscernible in the recovery record. Computation ofchanges in flow with time using an equation presentedby Jacob and Lohman (1952), and estimates ofhydraulic properties at these low-yield wells, indicatedthat changes in flow should have been discernibleduring the recovery period. For this reason, leakagefrom a nearby source, causing steady flow to a wellafter a relatively short time, seems to be the more likelycause of the steady inflow rate. For this analysis, a timeof 100 min for flow to stabilize was assumed.

A transmissivity value was computed for majorfractures or fracture zones in these four bedrock wellsusing equation 1. Equation 1 was derived for constant-flow conditions, but it also applies to constant-drawdown conditions after very small values of time(Lohman, 1972, p. 23). The well yield, in cubic feet perday (ft3/d), was computed by multiplying the rate ofwater-level rise by the cross-sectional area of the well.The yield for each fracture was assumed to be thepercentage of total yield determined from borehole-flowmeter tests (Hansen and others, 1999). Thedrawdown for each fracture or fracture zone was thedepth to the fracture below a static water level that wasmeasured about 2 weeks after the well was drilled. Thedepths to water-bearing fractures and percentage offlow reported by Hansen and others (1999) are shownin figure 4. The transmissivity values reported intable 2 are the sum of transmissivity values computedfor all water-bearing fractures or fracture zones in awell.

The estimates of transmissivity for all thebedrock wells that were made on the basis of specific-capacity data range widely—from 0.09 ft2/d inwell MW-15B to 130 ft2/d in well MW-3B. Thetransmissivity range of 280 to 550 ft2/d for wellMW-22B shown in table 2 probably reflects thetransmissivity of the bedrock and surficial aquiferscombined, as discussed later in this report. For thisreason, the estimates of transmissivity for wellMW-22B are not included in this range. Thetransmissivity values for the low water-yielding wellsMW-7B, MW-8B, MW-12B, MW-15B, MW-16B, andpossibly MW-4B, are 0.5 ft2/d or less. For comparison,Lyford and others (1998) report a transmissivity of0.6 ft2/d for the Smith Well, another low-yielding well.


EXPLANATION

WATER-LEVEL MEASUREMENTS INWELLS AND TREND LINES BETWEENMEASUREMENTS

MW-8BMW-10BMW-11BMW-12BMW-14BMW-15BMW-16B

DEPTH OF MAJOR WATER-BEARING FRACTURE OR FRACTUREZONE—Number indicates percentage ofyield from fractures or fracture zonesdetermined from borehole-flowmeter data(Hansen and others, 1999) Fractures notshown for wells MW-10B and MW-11Bbecause of space limitations on graph

WATER-LEVEL RECOVERY TRENDINTERPOLATED TO DEPTHS OFWATER-BEARING FRACTURES

2600 1,000 2,000 3,000 4,000 5,000

TIME SINCE WELL WAS DRILLED AND PURGED OF WATER, IN MINUTES

Figure 4. Depth to water during recovery of water levels in selected bedrock wells after drilling and depths of water-bearing fractures or fracture zones.

The relatively high transmissivity values determinedfor wells MW-10B, MW-1 IB, MW-14B, and the VanWart well, can be attributed to the presence of one ormore water-yielding fractures or fracture zones(Hansen and others, 1999). The high transmissivitymeasured in well MW-3B may be attributable tofractures in shallow bedrock that provide a hydraulicconnection to the surficial aquifer and to MeddybempsLake, but fracture data were not available for this well.

Estimates of transmissivity and hydraulicconductivity for surficial materials that were made onthe basis of specific-capacity data are summarized intable 3. The hydraulic conductivity values reported intable 3 were determined by dividing the transmissivityof each well by the saturated thickness of the aquifer at

the time of the tests. This approach for estimatinghydraulic conductivity requires the assumption that thefull saturated thickness of the aquifer contributed waterto the well during pumping. This assumption isreasonable if the saturated thickness is approximatelythe same as the length of the well screen, and for wellscompleted in coarse-grained materials. The assumptionmay yield values of hydraulic conductivity that aresomewhat lower than actual values in fine-grainedmaterials if vertical hydraulic conductivities arerelatively low, and if the screen length is considerablyless than the saturated thickness. The hydraulicconductivity estimates in table 3 for wells MW-8S,MW-15S, and MW-16S may be lower than actualvalues for this reason.


Table 2. Estimates of transmissivity for bedrock usingspecific-capacity data for wells, Meddybemps, Maine

[o length of open borehole s length of screen ff/d feet squared per dayft feet)

Well name

MW-1BMW-3BMW-4BMW-7BMW-8B

MW-10BMW-I1BMW-12BMW-14BMW-15B

MW-16BMW22BVan Wart

Transmissivityrange(ft2/d)

1 3-2360-130

< I 401-05

02

80-1570-110

0320-32

009

02280-550

12

Numberof tests

24121

34131

121

Length ofscreen or

open borehole(ft)

s 15s 10o!5o97o98

o94o94

o 1100 111

o!93

o98o23

o 103

Table 3. Estimates of transmissivity and hydraulicconductivity for surficial materials using specific-capacitydata for wells, Meddybemps, Maine

[ft foot ft/d feet per day ft~/d feet squared per day]

Wellname

MW4SMW5SMW7SMW8SMW IOS

MW 13SMW 15SMW 16SMW 17SMW L8S

Trans-mlf- Number fc"T8ivit* of tests le"£h

range (ft)(tf/d)

1 5-52150-200230-290

3 9

210-220

50

76-1929-80

11-18120

32

2

1

2

1

3

3

2

1

50

30

50

25

5

25

10

10

25

25

Saturatedthickness

range(ft)

55-92

62-86

37-37

74

50-72

2 7

184-237

23 1-306

36-67

5 7

Hydraulicconduc-

tivityrange(ft/d)

03-06

17-32

63-78

05

29-45

19

03-1 0

01-03

27-32

20

Relatively high values of transmissivity andhydraulic conductivity for surficial materials weremeasured in wells MW-5S, MW-7S, MW-10S, andMW-18S Estimates of hydraulic conductivity thatrange from 17 to 78 ft/d in these wells reflect the likelyrange of hydraulic properties for coarse-grainedglaciomanne sediments and is characteristic of siltysand and clean sand (Freeze and Cherry, 1979,table 2 2) Hydraulic conductivity values of 0 1 to

1 ft/d in wells MW-4S, MW-8S, MW-15S, andMW-16S reflect the likely range of hydraulicconductivity for till This range is within the rangeof values for tills derived from crystalline rocks insouthern New England and northern New Hampshire(Torak, 1979 Pietras, 1981, Melvm and others, 1992,table 3, Tiedeman and others, 1997, p 8)

Specific-capacity data are not available for wellsMW-9S, MW-12S, MW-19S, and MW-20S becausethey were dry or nearly dry and were not sampledSpecific-capacity data are not available for wellMW-11S because the yield was too low to sustainpumping during sampling

The major uncertainty associated with thespecific-capacity approach for estimating aquifertransmissivity is the storage coefficient The computedvalues of transmissivity, however, are relativelyinsensitive to the storage coefficient because thecoefficient appears in the log term of equation 1 Forexample, for well MW-12B completed in bedrock, areduction of the storage coefficient from 0 0001 to0 00001 increases the transmissivity from 0 16 to0 21 ft2/d For well MW-7S in the surficial aquifer,increasing the storage coefficient from 0 1 to 0 2decreases the transmissivity from 230 ft2/d to 207 ft2/d,and decreasing the storage coefficient from 0 1 to 0 05increases the transmissivity to 250 ft2/d For the low-yielding wells, the time required for the inflow rate tostabilize also is uncertain, but, as with the storagecoefficient, the transmissivity estimates are notparticularly sensitive to time because time also appearsin the log term of equation 1 For example, for wellMW-12B, a reduction of the time from 100 mm to10 mm in equation 1 reduces the computedtransmissivity of the highest-yielding fracture from0 16ft2 /dtoO 11 ft2/d

Aquifer Tests

Aquifer tests were conducted at wells MW-22Band MW-1 IB during September 10-14, 1997, to refineestimates of hydraulic properties for the ground-watersystem Water-level responses to pumping wereobserved in these two wells and the Smith and VanWart wells, which were pumped intermittently fordomestic purposes during the aquifer-test period(fig 5). Water levels in wells that responded topumping are shown in figure 6 Water levels in theDenny s River, Meddybemps Lake, and wells pnor toaquifer testing are summarized in table 1


170

140I , L 1 , L , L , 1

9-10-97 9-11-97 9-12-97 9-13-97 9-14-979-10-97 9-11-97 9-12-97 9-13-97

EXPLANATION

RECORDER MEASUREMENT

x MANUAL MEASUREMENT

MW-11B WELL NAME

Figure 5. Water-level hydrographs for wells pumped duringaquifer testing, Meddybemps, Maine.

The original plan for aquifer testing was to pumpwell MW-22B at a constant rate for at least 24 hours.After about 6 hours of pumping at a rate of 9 gal/min,however, drawdown suddenly increased and the waterlevel quickly fell to the pump intake. During part of thepumping period, turbidity in the water caused by finesand and silt indicated that the well may not be fullycased through surficial materials or that fractures inbedrock provided a short connection to the surficialaquifer. The test was terminated after 325 minutes, andwater levels were allowed to recover for about 18 hoursbefore a second test was started in well MW-1 IB.

Water-level data for the first 150 minutes of thetest in well MW-22B were used to estimate aquifertransmissivity by applying the straight-line method ofCooper and Jacob (1946). Drawdown during this part ofthe test followed an approximate straight-line trend on asemi-logarithmic plot, as shown in figure 7, before thewell started producing sand and silt and the water levelstarted declining at a greater rate. A transmissivity valueof 450 ft2/d estimated from the time and drawdown datais consistent with estimates using specific-capacity data(table 2). This estimate of transmissivity may be highrelative to other wells because it may represent acombination of the surficial and bedrock aquifers.

For the second test, well MW-1 IB was pumped ata rate of 4.5 gal/min for 24 hours. A uniform pumpingrate was difficult to maintain, and the small variations inpumping rate were apparent in the water-level record.Water levels in all wells monitored south of Route 191,except well MW-8B, responded to pumping from wellMW-1 IB (fig. 5). A rise of water level in well MW-8Bshortly after pumping started resulted from an estimated0.1 to 0.2 in. of precipitation that entered the uncoveredwell during a rain shower.

Water-level data for the first 200 min of the testin well MW-1 IB (fig. 7) were used to estimatetransmissivity by applying the straight-line method ofCooper and Jacob (1946). During this period, the effectsof leakage and well-bore storage were assumed to benegligible; however, both factors could have affected theanalysis. A transmissivity of 38 ft2/d calculated by thismethod is lower than estimates using specific-capacitydata (70-110 ft2/d in table 2), possibly because well-borestorage affected the rate of drawdown during the firsthour of pumping. The estimates of transmissivity for thiswell were refined using numerical methods discussed inthe next section.


CO0

oo

8

i(0

a0)

en

I

crcao>c

o>c"a.

Q.

I I I I I I

enco

UJ UJ

2 u

rr Ul UlUJ 5 5CE < Z

§ 1 a!DC 5 5

mo

T3<D

TJ

IO>0)

«5

(D

O

I

Q.

SO)o

2 8

H3A3T V3S 3AO9V 133d Nl '3aniinV 13A3n-d31VM

(0

CO

£


18

20

22

QC111

Ri26

28

30

7= 35 OAs

O = 4.5 gal/minAs = 4.2 ft

?•= 38 ft2/d

WELLMW-11B

10 20 50 100 200 500 1,000 2,000

LULL

ASQ = 9 gal/min

As = 0.7 ftr=450ft2/d

5 10 20 50 100 200 500 1,000TIME SINCE PUMPING BEGAN, IN MINUTES

EXPLANATION

+ MANUAL WATER-LEVEL MEASUREMENT

LINE USED FOR COMPUTATION OF TRANSMISSIVITY

O PUMPING RATE, IN GALLONS PER MINUTE (gal/min)

AS CHANGE IN DRAWDOWN, IN FEET (ft) OVER ONE LOGCYCLE

T TRANSMISSIVITY, IN FEET SQUARED PER DAY (frVd)

Figure 7. Depth to water in wells MW-11B and MW-22B during pumping and calculations oftransmissivity, Meddybemps, Maine.


ESTIMATION OF HYDRAULICPROPERTIES USING NUMERICALMETHODS

Numerical methods of analysis were used tosimulate conditions in the surficial and bedrockaquifers underlying the study area. Numerical methodshave advantages over analytical methods becauseboundary and initial conditions, along with aquiferproperties, can be varied across the area of interest,increasing the degree to which the models represent theconditions found in the aquifer systems. The numericalmethod used in these analyses is the MODFLOWmodel, a three-dimensional finite-difference simulatororiginally documented by McDonald and Harbaugh(1988) and updated by Harbaugh and McDonald(1996).

The use of MODFLOW assumes that ground-water flow in the fractured bedrock underlying thestudy area can be analyzed (modeled) using acontinuum approach. (See Freeze and Cherry, 1979,p. 73, for a discussion of this approach.) This approachinvolves a representation of the fracture systemsuch that values of hydraulic head and hydraulicconductivity are continuous within the aquifer systemand that ground-water flow follows Darcy's Law (thatis, flow is proportional to the hydraulic gradient). Thisapproach has been found to be valid and useful in thosecases in which a number of open, hydraulicallyconnected fractures are available for ground-water flowat the scale of the model simulation (Domenico andSchwartz, 1990, p. 83-88).

Although it is difficult to assess the degree towhich the continuum assumption is valid in the studyarea, it is clear from borehole-geophysical logs(Hansen and others, 1999) that many water-yieldingfractures are present with a spacing between 20 and50 ft, depending on the orientation of the fracture sets.With this spacing of fractures, it is likely that groundwater flows in interconnected sets of fractures acrossthe study area and thus can be simulated as acontinuous ground-water-flow system. It should benoted, however, that the hydraulic properties estimatedfor the bedrock by the continuum approach are averagevalues for blocks of the bedrock, and not of individualfractures or of the non-fractured block. In addition,because of the discrete nature of fractures and theirdistribution, the hydraulic characteristics of thebedrock system will vary widely, and the localmovement of water could deviate from the averagemovement simulated by the continuum model.

The numerical analysis involved a two-stepprocess, the first of which was to create a simulation offlow in the aquifers at the Meddybemps site underlong-term conditions (that is, a steady-state condition).The result from the steady-state simulation thenprovided a stable initial condition for the second step,which involved transient simulation of the MW-1 IBaquifer test. These steps were repeated as calibration ofthe aquifer properties proceeded, with the goal ofimproving the degree of agreement of simulated headswith measured aquifer heads after each set of changesin aquifer properties made during model calibration.Aquifer properties were changed during modelcalibration to produce simulated head conditionssimilar to the heads measured in the aquifer bothbefore the aquifer test (steady-state simulations) andduring the aquifer test at well MW-1 IB (transientsimulations).

Model Design and HydraulicProperties

The design of the Meddybemps numerical modelclosely follows the description of the ground-watersystem (conceptual model) presented by Lyford andothers (1998) and in the description of the study area inthis report. The principal area of interest is the southend of the study area, which includes the southerncontaminant plume (fig. 2). The extent of the model(figs. 1 and 8) is much larger than the area of interest soas to minimize the effect of relatively unknownboundary conditions on simulations in the area ofinterest. To simulate ground-water flow, it wasnecessary to assign initial conditions, boundaryconditions, aquifer properties, and internal sources andsinks of water across the modeled area. A rectangulargrid of aquifer blocks was laid out horizontally acrossthe modeled area, and the ground-water flow systemwas subdivided vertically into layers. The horizontalextent of the grid across the site is shown in figure 8.The modeled area is 40 blocks long along the columndirection (northwest to southeast) and 35 blocks widealong the row direction (northeast to southwest). Aseach block is 50 ft square, the extent of the modeledarea is 2,000 ft (column direction) by 1,750 ft (rowdirection). By convention, the origin in all MODFLOWmodels is the upper left-hand corner block (column 1,row 1), which in the Meddybemps model is thenorthernmost corner (fig. 8).


C(01

30

3510

0 100 200 300 400 FEETI I I I II n r^ r

0 40 80 120 METERS

15 20COLUMNS

25 35 40

EXPLANATION

PRESUMPSCOT FORMATION (SILT AND CLAY)

GLACIOMARINE (SAND)

J TILL

| CONSTANT HEAD BLOCKS

"1 INACTIVE AREAS

-100 TOPOGRAPHIC CONTOUR-Contour interval is 10 feet.Datum is sea level. Topography from Meddybemps LakeUSGS quadrangle, 1:24,000, provisional edition, 1987.

^ HOUSE OR BUILDING

Figure 8. Finite-difference grid, surficial geology of modeled area, and constant-head boundaries, Meddybemps, Maine.

Estimation of Hydraulic Properties Using Numerical Methods 15

The vertical distribution along section B-B' (lineof section shown on fig. 2) of the four layers used in theMeddybemps model is shown in figure 9. The uppertwo layers represent the surficial (glacial) materials,with the top layer (model layer 1) representing theuppermost saturated materials. The bottom of layer 1was established as 5 ft above the bedrock surface.Because most of the surficial aquifer is unconfined, thethickness of layer 1 varied across the modeled area andthis layer was inactive in those areas where thesaturated thickness of the surficial materials was less

than 5 ft. Surficial materials vary across the modeledarea as shown in figure 8 and include coarse-grained(sand) glaciomarine deposits on the bedrock ridge tothe west of the Dennys River, till, and fine-grainedglaciomarine deposits of the Presumpscot Formationwest of the river. The hydraulic conductivity of layer 1varied with the type of surficial material and rangedfrom 30 ft/d for coarse-grained glaciomarine material,0.1-0.5 ft/d for till, to 0.001-0.01 ft/d for fine-grained(Presumpscot Formation) material.

NorthwestR

220

overlies bedrock Q

Bottom of wellat attitude of 45 ft\

Layer 1Layer 2Bedrock surface

VERTICAL EXAGGERATION X 3 75

10 20 30 40 METERS

GLACIAL DEPOSITS

Glaciomarine deposits of thePresumpscot Formation, fine grained

•— Fine sand and silt— Silt and clay

Glaciomanne deposits, coarse grainedSand and gravelSand

Glacial till

EXPLANATION

BEDROCK

Gabbro-dionte intrusive complex

HYDROLOGY

PotentiomelncSurface In Surficial Aquifer-Measured on April 30. 1997

BOUNDARY BETWEENMODEL LAYERS

WATER LEVEL INBEDROCK (B) WELL-Measuredon April 30,1997

SCREENED INTERVALIN SHALLOW (S) WELL

Figure 9. Geologic section B-B' showing layers used in numerical model, Meddybemps, Maine.


Model layer 2 represents the till that mantlesmost of the bedrock across the modeled area, althoughthere are significant absences in areas across andadjacent to the bedrock ridge (fig. 9). Layer 2 in thoseareas represents the lower 5 ft of coarse-graineddeposits and was the uppermost active layer in severalblocks where the saturated thickness of surficialmaterial was less than 5 ft. The bottom of layer 2 is thebedrock surface across the modeled area. Hydraulicconductivity in layer 2 ranged from 10-30 ft/d forsandy material to 1.0 ft/d for till.

The lower two model layers represent thefractured bedrock aquifer (fig. 9). Layer 3 representsthe upper part of the fractured bedrock aquifer, and thebottom of this layer was established at 20 ft below thebedrock surface. Hydraulic conductivity of layer 3 wasset almost uniformly to 0.003 ft/d but was increased to0.5-1.5 ft/d in the area between wells MW-10B andMW-1 IB because a highly transmissive, high-anglefracture (or fracture set) hydraulically connects thesetwo wells. Layer 4 represents the deeper bedrockaquifer. Because the depth of active ground-water flowin this part of the system in the study area is not wellknown, the hydraulic properties of this layer wererepresented using variable transmissivity values.Transmissivity values for layer 4 generally were0.24 ft2/d but increased to 40-120 ft2/d for that areabetween wells MW-10B and MW-1 IB with the highlytransmissive (high-angle) fracture or fracture set. Thecombined transmissivity of the two bedrock layersranged from 0.03 to 150 ft2/d across the model area,and the majority of the combined bedrocktransmissivities were set at 0.3 ft2/d. A value of0.3 ft2/d is consistent with values estimated usingspecific-capacity data for low-yielding wells (table 2).

Vertical leakage between model layers inMODFLOW is controlled by the VCONT parameter,which is defined as the vertical hydraulic conductivitydivided by the distance between vertically adjacentmodel blocks (thickness); thus VCONT has units ofinverse time (day"1). Generally, VCONT values werecalculated using equation 51 in the report byMcDonald and Harbaugh (1988), which uses thehydraulic conductivity and thickness of the two verti-cally adjacent blocks. Isotropic conditions wereassumed for these calculations (that is, vertical hydrau-lic conductivity values were held equal to horizontalvalues). VCONT values between layers 1 and 2 rangedfrom 4xlO~4 days"1 for fine-grained material to0.39 days"1 for coarse-grained material. VCONT

values between layers 2 and 3 were almost uniformly3xlO"4 days"1. VCONT values between layers 3 and 4were generally 6xlO"5 days"1. During calibration,VCONT values were increased to 0.005 days"1

between layers 2 and 3 and 1.0 days"1 between layers 3and 4 in the region of the high-angle fracture betweenwells MW-1 OB and MW-1 IB, creating a highly trans-missive vertical-flow condition in this area of theground-water flow system. It was also found duringmodel calibration that it was necessary to significantlydecrease VCONT values between layers 3 and 4 in thearea around the Smith well to match the observeddrawdowns in head around this well; VCONT values inthis area were 1 x 10~9 days"'.

Boundary Conditions

Model boundary conditions include constant-head blocks around the exterior of the modeled area,along the shore of Meddybemps Lake, and along theDennys River and its unnamed tributary to the south(fig. 8). Head conditions around the model exteriorwere estimated from surface topography, assuming a 5-to 10-foot depth to ground water. Head conditions forMeddybemps Lake and the Dennys River were basedon values reported by Lyford and others (1998) forApril 30, 1997. Head conditions along the unnamedtributary to the Dennys River were estimated fromsurface topography. The constant-head boundaryconditions on the exterior of the model, in addition tothose on the lake shore and the unnamed tributary onthe southern edge of the model, were used in modellayers 1 to 4. Constant-head conditions along theDennys River, transversing the central part of themodel, were used only in layers 1 and 2 of the modelbecause the surficial material is absent from most of thestreambed.

Recharge and Wells

The source of most ground water underlyingthe Meddybemps site is recharge from precipitation.Because the rate of infiltration of precipitation isgenerally controlled by soil permeability, the rechargerates across the study area were varied by the type ofmaterial at the surface. An estimated recharge rate of10 in/yr was used in areas with till at the surface, a rateof 20 in/yr was used in areas with sandy material, and0.5 in/yr was used in areas underlain by silt and clay


(Presumpscot Formation). The estimates for till andsandy material are consistent with rates reported forsimilar geologic settings in Connecticut (Mazzaferroand others, 1979). The recharge rate of 20 in/yr thatwas used in the model is somewhat lower than potentialrates of 24 to 26 in/yr (Lyford and Cohen, 1988) toallow for some surface runoff from fairly steep slopes.The value for silt and clay was assumed to be very low,but supporting information is not available.

Two domestic wells were included in thesimulation, the Smith well (row 17, column 25) and theVan Wart well (row 18, column 31) (figs. 2 and 8). Bothdomestic wells were estimated to pump 100 gal/d fromthe deeper bedrock layer (layer 4) on the basis ofsteady water-level recovery rates between pumpingperiods in the Smith well (fig. 5) and a single occupantin the Van Wart residence.

Steady-State Calibration and Simulation

Model simulations of the long-term, steady-statehead conditions in the ground-water-flow systemunderlying the Meddybemps site were conducted tomatch the geologic and hydrologic observations(calibration), and to provide a stable initial conditionfor the transient simulations of the MW-1 IB aquifertest. However, it was found that MODFLOW washighly unstable in the steady-state mode. Thisinstability was the result of the nonlinear nature ofthe unconfined flow conditions in the model, in whichthe saturated thickness of the aquifer depends on thehead in the model block. This inherently nonlinearsituation was made even more unstable in the regionof the bedrock ridge near Stone Road (fig. 2) by thelarge contrast in hydraulic conductivities betweenthe overlying sand and the fractured bedrock below. Itwas not uncommon during the iterative-solutionprocedure for the steady-state simulations for wideswings to occur in calculated heads in the region of thebedrock ridge. These swings created high heads in oneiteration immediately followed by low heads in asecond iteration—an oscillation of calculated head thatnever allowed the model to converge to a final solution.This model instability occurred in an area of theground-water system where field measurementsindicated that wells that penetrate the surficial

materials become dry during some periods of the year,indicating a thin, saturated thickness of the surficialaquifer as an average condition (fig. 9).

The model oscillation problem could not beresolved by adjusting model-solver parameters butwas overcome by using a transient simulation thatconverged to a steady-state condition over a 1,000-day period. The changes in head during the iterative-solution procedure were slow because of thedampening effect of the storage coefficients used inthe transient simulation to keep the model solutionstable at each time step (0.5 days). The result, after2,000 time steps, is a condition in which the change inhead, with respect to time, has slowed to the point atwhich the amount of water from storage representsonly 1 percent of the total flux through the system, andinputs minus outputs equal the change in storage. Theresulting steady-state head solution included severalblocks in layer 1 along the bedrock ridge that becameinactive (dry) during the simulation, reflecting thesmall saturated thickness of the surficial material in thispart of the ground-water system (fig. 9).

The results of the steady-state simulation areshown in figures 10 and 11. The distribution ofsimulated head in the uppermost active layer acrossthe modeled area is shown in figure 10; in some areasthe uppermost active layer is layer 2, because thecalculated head is below the bottom of layer 1. Thecontoured distribution of head observed in wells also isshown in figure 10; these head contours are modifiedfrom Lyford and others (1998, fig. 9). A comparisonof the two sets of contours, although limited in arealextent, shows good agreement in absolute magnitudeof head and in the direction of ground-water flow.Generally, the primary control on head in the surficialmaterial in the area of interest (that is, the central partof the model) is the movement of ground water fromrecharge areas in topographically high areas todischarge areas along the Dennys River. To achievea suitable calibration, the hydraulic conductivity forlayer 2 near wells MW-8S, MW-10S, and MW-1 IS(fig. 2) was increased to a value representative ofcoarse-grained materials. Low hydraulic conductivityvalues representative of till were initially assumedbecause of the low yields observed from wells MW-8Sand MW-1 IS. These wells may not fully penetratecoarse materials, or the coarse materials may not bepresent near these wells.


I

30

35

0 100 200 300 400 FEET

I ' l ' I ' l '0 40 80 120 METERS

20

COLUMNS

• • • 160 - • • HEAD CONTOUR BASED ON WATER-LEVELMEASUREMENT-Contour interval is 5 feetDatum is sea level (Data from Lyford and others, 1998)

100 HEAD CONTOUR BASED ON MODELSIMULATION—Contour interval is 5 feetDatum is sea level

1 FINITE-DIFFERENCE GRID

40

Figure 10. Observed head contours for the surficial aquifer and simulated head contours for the uppermost active layer,Meddybemps, Maine.


1

30

35

10

0 100 200 300 400 FEETI I I I I

15 20

COLUMNS

25 35 40

I40

\ I80 120 METERS

EXPLANATION

• • • 160 • • • HEAD CONTOUR BASED ON WATER-LEVELMEASUREMENT— Contour interval is 5 feetDatum is sea level. (Data from Lyford and others, 1998)

100 HEAD CONTOUR BASED ON MODELSIMULATION— Contour interval is 5 feetDatum is sea level

FINITE-DIFFERENCE GRID

VAN WART WELL

SMITH WELL

Figure 11. Observed head contours for the bedrock aquifer and simulated head contours for model layer 4, Meddybemps,Maine.


In contrast to conditions found in the surficialaquifer, the primary factor affecting the head in thedeeper bedrock (layer 4) in the area of interest appearsto be pumping from the Smith well (fig. 11). A well-developed cone of depression is simulated around theSmith well, with a smaller cone developed around theVan Wart well. These simulated head patternsreasonably match the observed head patterns (modifiedfrom fig. 11 in Lyford and others, 1998). It should benoted that the vertical leakage parameter, VCONT, wasdecreased significantly between layers 3 and 4 in thearea around the Smith well to simulate the magnitudeand extent of the observed head pattern as shown byLyford and others (1998, fig. 11).

The average difference between the measuredand simulated heads (table 4) was about 1 ft, and thestandard deviation of those differences was about 4 ft.The largest difference in head was for well MW-1B(fig. 2), a bedrock well on the edge of the area ofinterest. Most of the differences were small; 85 percentof the differences in head were between -5 and 5 ft. Acomparison of well pairs (S-shallow, B-bedrock)indicates that the proper direction of head change withdepth was simulated for five of the eight pairs (fig. 2and table 4). In fact, the largest head difference withdepth, measured at wells MW-15S and MW-15B, wasclosely matched by the simulated head difference withdepth at that location. Two well pairs (MW-4S and 4Band MW-7S and 7B) show that measured headsdecrease with depth near the Dennys River, whereasthe simulated heads increase with depth. This situationmay indicate that the effect of drawdown from theSmith well pumpage on heads on the opposite side ofthe river may be stronger than is simulated by themodel.

Transient Calibration and Simulation ofWell MW-11B Aquifer Test

Calibration of model parameters during thetransient phase of simulations primarily involved theestimation of aquifer storage properties and changesto aquifer hydraulic properties in the region aroundwells MW-1 IB and MW-10B. The type of layer inMODFLOW determines the storage coefficient appliedduring a simulation. Layer 1 (the uppermost surficialmaterial) was set as a type 1, and an unconfined storagecoefficient (specific yield) was used. Layer 2 (surficialmaterial, primarily till) was set as a type 3, or as a

Table 4. Observed and simulated steady-state heads forwells, Meddybemps, Maine

[Measured head values are water levels, in feet above sea level, on April 30,1997 (Lyford and others, 1998, table 3)]

Well nami

MW-1BMW-3BMW-4SMW-4BMW-5S

MW-6SMW-7SMW-7BMW-8SMW-8B

MW-9SMW-1 OSMW-1 OBMW-1 ISMW-1 IB

MW-12SMW-12BMW-13SMW-14BMW-15S

MW-15BMW-16SMW-16BMW-17SMW-18S

MW-20SMW-22B

AverageStandard

MeasuredB head

(hm)

169.57173.63165.93162.76174.41

175.40163.61158.79157.93157.01

158.43158.90157.90154.99155.49

177.60174.22159.71178.13166.07

155.62178.16176.54160.57160.02

175.44158.36

deviation

Simulatedhead(h.)

179.77170.94166.01169.27175.78

179.29163.79166.27157.37156.65

161.73161.97161.04155.18154.95

174.90169.49159.52183.02168.50

158.70172.61172.53160.76161.79

173.36159.11

Difference(ru-hm)

10.20-2.69

.086.511.37

3.89.18

7.48-.56-.36

3.303.073.14

.19-.54

-2.70-4.73

-.194.892.43

3.08-5.55-4.01

.191.77

-2.08.75

1.083.67

convertible layer. If a block in layer 1 is active, thestorage coefficient used for the underlying layer 2 is aconfined storage coefficient. When the head in a blockin layer 1 drops below the bottom of the block, thestorage coefficient in the underlying layer 2 blockconverts to an unconfined storage coefficient. Layers 3and 4 (bedrock) were set as confined units, andconfined storage coefficients were used. Specific yield(unconfined storage) values used in layers 1 and 2 wereset at 20 percent. The confined storage coefficient wasset at IxlO'5 for layer 2 and IxlO"4 for layers 3 and 4.The confined storage coefficient for layer 2 was set


lower than that of layers 3 and 4 because layer 2 isthinner (5 ft) than either layers 3 (20 ft) or 4 (unknown,but several tens of feet at a minimum).

The results of the simulation of the aquifertest are presented in figures 12-14, and table 5. Theobserved and simulated drawdown over time atobservation well MW-10B during the 24-hour testconducted at well MW-1 IB is shown in figure 12.Simulation time steps ranged logarithmically from 1to 200 minutes. The shape of the simulated drawdowngenerally matches the observed drawdown curve.Major features of the observed response in wellMW-1 OB are the rapid increase in drawdown andthe equally rapid stabilization of drawdown. Tosimulate the rapid increase in drawdown over timein the observation well, which is 205 ft from thepumped well, it was necessary to greatly increase thetransmissivity of the bedrock aquifer (layers 3 and 4)between these two wells. This change in aquiferproperties in the model is supported by geophysicalevidence (Hansen and others, 1999), which indicatesthat a high-angle fracture (or fracture set) hydraulicallyconnects these two wells. In addition to the increase intransmissivity between the two wells, it was alsonecessary to decrease the transmissivity of the bedrockimmediately around the high-transmissivity zone (to

2.5

HIKM.5

<ro0.5

01 '0.0001

Model

0.001 0.01 0.1 1TIME SINCE PUMPING BEGAN, IN DAYS

10

Figure 12. Observed and simulated drawdown at well MW-1 OBwhile pumping well MW-11B, Meddybemps, Maine.

0.03 ft2/d, corresponding to a hydraulic conductivity ofSxlO"4 ft/d) to reduce the drawdown response observedin bedrock wells outside this zone. (See, for example,well MW-22B in table 5.) Although the model wascalibrated to simulate the response in MW-22B as if itwere open solely to the bedrock system, the lowdrawdown in this well may be due to vertical flow fromthe surficial aquifer through a local vertical fracture ora broken casing seal. Therefore, the calibration resultof a low hydraulic conductivity zone around the highlytransmissive fracture may be an artifact of thequestionable observed drawdown at well MW-22B.

The stabilization of the drawdown response inMW-10B (fig. 12) required sources of water to sustainthe pumping at MW-1 IB; the model simulationsindicate that the likely sources of water were leakagefrom the overlying surficial materials and flow from theDennys River into the bedrock aquifer. A comparisonof drawdowns in layer 4 (fig. 13) and layer 2 (fig. 14)shows the effects of these two sources on thedrawdown patterns around the pumping well. Thesteep, elongated cone of depression after 1 day ofpumping shown in figure 13 is shortened in thedirection of the overlying constant head block in layer2, indicating the simulated downward leakage of waterfrom the Dennys River into the bedrock fracture thatwas assumed to extend to the river. The surficialmaterials are thin along this segment of the DennysRiver, providing a potential hydraulic connectionbetween the river and the bedrock fracture, if itunderlies the river. In addition to leakage from theriver, the model simulations also indicated that waterleaking downward from the overlying surficial materialinto the fracture was needed to stabilize drawdowninduced by pumping of MW-1 IB. Because the dipangle of the fracture is high (>45 degrees), the outcropof the fracture at the bedrock surface is expected to bepresent near the pumping and observation wells, andthe vertical leakage parameter VCONT was increasedto 0.005 days"1 between layers 2 and 3 and 1.0 days"1

between layers 3 and 4 to simulate this ability to movewater vertically.

The drawdown pattern in the surficial aquifer(layer 2) after 1 day of pumping is shown in figure 14.Although the drawdowns in layer 2 were significantlyless than those in the fracture (fig. 13), they are areallyextensive and indicate that water was moving throughthe overlying materials toward and into the fracture,providing another source of water for drawdownstabilization. As noted earlier, the hydraulic


CO

21

22

23

24

25

20 21 220 50 100 FEET1 . V , iI I I I0 10 20 30 METERS

23 24 25

COLUMNS

26 27 28 29

\©

EXPLANATION

CONSTANT-HEAD BLOCK IN OVERLYING LAYER (LAYER 2)

HIGH TRANSMISSIVITY ZONE

DRAWDOWN CONTOUR-In Feet

APPROXIMATE WELL LOCATION

Figure 13. Simulated drawdown in model layer 4 after pumping MW-11B for 1 day, Meddybemps, Maine.

conductivity of the surficial aquifer in this area (layer2) was increased from initial estimates to simulate thedrawdowns observed in these wells (table 5) and toimprove the steady-state head results with respect topre-test conditions.

The overall magnitude of simulated drawdownvalues was reasonable for the five observation wellsand the pumped well at the end of the aquifer test(table 5). Differences between observed and simulateddrawdowns at the observation wells can be explained,to some degree, by differences in the distances betweenobserved and simulated wells (table 5). Additionally, itshould also be recognized that the drawdown value inthe pumped well (MW-1 IB) is expected to be largerthan that of the simulated drawdown in the blockcontaining the pumped well, because the numericalmodel assumes a distributed pumping source (that is,the water is pumped from all of the 50-ft square block).

Head losses at the open well bore (also called entrylosses) are considered to be negligible at the pumpingrate of 4.5 gal/min. It is particularly interesting to notethe greatly reduced drawdown in well MW-22Brelative to MW-1 OB, although it is significantly closerto the pumped well than MW-1 OB (fig. 2). However,water levels in MW-22B may not be suitable forcomparison to modeling results for reasons statedearlier. In addition, although not shown in table 5, itshould be noted that no drawdown was observable inbedrock well MW-8B during the aquifer test, eventhough 0.04 ft of drawdown was recorded in theshallow well at that location. These observationssupport the view that the major control on the flow ofwater in this region of the bedrock system is the highlytransmissive fracture set between wells MW-1 IB andMW-1 OB.


20

21

29

30

20 21 22

0 50 100 FEET

I 1 r1—i '0 10 20 30 METERS

4.

\

23 24 25 26 27 28 29

COLUMNS

EXPLANATION

CONSTANT-HEAD BLOCK

HIGHTRANSMISSIVITY ZONE IN UNDERLYING LAYERSLAYERS (3 AND 4)

0.1 DRAWDOWN CONTOUR-In Feet

(J) APPROXIMATE WELL LOCATION

30

Figure 14. Simulated drawdown in model layer 2 after pumping MW-11B for 1 day, Meddybemps, Maine.


Table 5. Observed and simulated drawdown in the pumped well and observation wells after pumping well MW-11B for 24hours, Meddybemps, Maine

Wellname

MW-8SMW-10SMW-10B

MW-1 IS

MW-1 IB

MW-22B

Distance topumped well

(feet)

202

210

205

15

0

169

Observeddrawdown

(feet)

0.04

.16

1.90

1.12

11.78

.10

Simulationdistance

(feet)

212

200

200

12

12

158

Simulateddrawdown

(feet)

003

.22

2.12

.71

9.91

.45

Difference(simulated-observed)

(feet)

-001

06

22

-41

-1 87

35

SUMMARY AND CONCLUSIONS

The hydraulic conductivity of surficialmaterials near the Eastern Surplus Superfund Sitein Meddybemps, Maine, determined from specific-capacity tests, ranges from 17 to 78 ft/d for wellscompleted in coarse-grained glaciomarine sedimentsand from about 0.1 to 1.0 ft/d for wells completed intill. The transmissivity of fractured bedrock determinedfrom specific-capacity tests and an aquifer test inbedrock ranges from about 0.09 to 130 ft2/d. Relativelyhigh values of transmissivity at the southern end of thestudy area appear to be associated with a high-anglefracture or fracture zone that hydraulically connectstwo wells completed in bedrock. Transmissivities at sixlow-yielding (less than 0.5 gal/min) wells, whichappear to lie within a poorly transmissive block of thebedrock, were consistently in a range of about 0.09 to0.5 ft2/d.

The estimates of hydraulic conductivity andtransmissivity in the southern half of the study areawere supported by steady-state calibration of anumerical model and simulation of a 24-hour aquifertest at a well completed in bedrock. Model calibrationindicated an extension of a relatively transmissive zonein the surficial aquifer beyond the mapped extent ofcoarse-grained sediments eastward to the DennysRiver. It was necessary to use high values of verticalhydraulic conductivity along the fracture or fracturezone that connects the pumped well and an observationwell to match drawdowns during model calibration tothe aquifer tests. In addition, modeling results indicatedthat a very low vertical hydraulic conductivity was

needed to simulate a persistent cone of depression neara residential well that may lie within the poorlytransmissive block of bedrock.

Considerable uncertainty is attached to estimatesof transmissivity and hydraulic conductivity deter-mined by analytical methods because of the many sim-plifying assumptions. Nevertheless, calibration of anumerical model yielded values of the hydraulic prop-erties of the surficial- and bedrock-aquifer system thatwere similar to those determined by analytical methodsfor wells at the Meddybemps site. Numerical modelingalso yielded estimates of vertical hydraulic conductiv-ity that would be difficult to determine analyticallyfrom available data. Generally, the numerical modelingresults support the conceptual model of ground-waterflow. The following observations were based onnumerical modeling:

• Hydraulic properties within the surficial materialsand the fractured bedrock are quite variable.Hydraulic conductivities range between 0.001 to30 ft/d for surficial materials, and transmissivitiesrange from 0.03 to 150 ft2/d for fractured bedrock.The high contrast in values in hydraulic propertiesindicate the presence of preferential pathways forwater flow in the Meddybemps site that couldaffect the direction and rate of contaminanttransport at the site.

• Although low yields in wells MW-8S and MW-1 ISindicate that the transmissivity of surficialmaterials decreases eastward and southeastwardtoward the Dennys River, model calibrationindicated that the surficial materials in this area arerelatively transmissive.

Summary and Conclusions 25

• The rapid stabilization of the drawdown response topumping at well MW-1 IB indicates that likelysources of water are leakage from the overlyingsurficial materials and flow from the Dennys Riverinto the bedrock aquifer. Simulation of downwardmovement of water from surficial materials andlateral movement from the Dennys River to theprincipal fracture zone yielding water in wellMW-1 IB produced asymmetric and ellipticaldrawdown patterns in bedrock and surficialaquifers.

• Model simulations indicate that the vertical leakageto the deeper bedrock around the Smith residentialwell must be very low to maintain the depressedwater levels that are observed in nearby bedrockwells and in the Smith well itself. The low verticalleakage and the low transmissivity of the deeperbedrock in this part of the study area is supportedby the specific-capacity data and headobservations for wells open to this part of thebedrock. The inferred low vertical leakage, incombination with the low transmissivity of thebedrock in this part of the study area, couldconstrain contaminant transport. VOCs had notbeen detected in the Smith well prior to the springof 1998. The high degree of variability in aquiferhydraulic properties, however, must be taken intoconsideration to achieve an accurate assessment ofthe vulnerability of an individual well tocontamination.

The estimates of hydraulic properties presentedin this report were determined on the basis of availablespecific-capacity data, aquifer tests, and numericalmodeling done within the time and budgetaryconstraints of the project. Additional aquifer testingand further refinement of the model through additional

calibration might narrow the range of hydraulicproperty values and refine the conceptual flow modelfor the aquifer system; however, the degree ofrefinement needed will depend on the intended use ofthe information. The information presented in thisreport could be used to estimate ground-water-flowpatterns and flow rates of water through the aquifersystem and to assess remediation approaches. Designof a ground-water remediation system and simulationof solute transport might require additionalinformation, particularly the hydraulic properties offractures.

The numerical model described in this reportwas designed to analyze aquifer tests, evaluate theconceptual model of ground-water flow, and refineestimates of hydraulic properties for the aquifersystem. Other uses, such as a design of a ground-waterremediation system or as a basis for solute-transportmodeling, may not be appropriate because fracturedistribution and properties have been characterizedonly in some areas and other conceptual models ofground-water flow may be possible. Thepreponderance of evidence, however, indicates that theconceptual model is the most likely interpretation ofinformation collected thus far at the Meddybemps site.

REFERENCES

Cooper, H.H., Jr., and Jacob, C.E., 1946, A generalizedgraphical method for evaluating formation constantsand summarizing well-field history: AmericanGeophysics Union Transactions, v. 27, no. 4,p. 526-534.

Domenico, P.A., and Schwartz, F.W., 1990, Physical andchemical hydrogeology: New York, John Wiley & Sons,824 p.


Fetter, C.W., 1994, Applied hydrogeology, 3d edition:Englewood Cliffs, N.J., Prentice-Hall, Inc., 691 p.,1 diskette.

Freeze, R.A., and Cherry, J.A., 1979, Ground water:Englewood Cliffs, N J., Prentice-Hall, Inc., 604 p.

Hansen, B.P, Stone, J.R., and Lane, J.W., Jr., 1999,Characteristics of fractures in crystalline bedrockdetermined by surface and borehole geophysicalsurveys, Eastern Surplus Superfund Site, Meddybemps,Maine: U.S. Geological Survey Water-ResourcesInvestigations Report 99-4050,27 p.

Harbaugh, A.W., and McDonald, M.G., 1996, User'sdocumentation for MODFLOW-96, an update to theU.S. Geological Survey modular finite-differenceground-water flow model: U.S. Geological SurveyOpen-File Report 96-485,56 p.

Jacob, C.E., and Lohman, S.W., 1952, Nonsteady flow to awell of constant drawdown in an extensive aquifer:American Geophysics Union Transactions, v. 33,p. 559-569.

Lohman, S.W., 1972, Ground-water hydraulics: U.S.Geological Survey Professional Paper 708,70 p.

Lyford, P.P., and Cohen, A J., 1988, Estimation of wateravailable for recharge to sand and gravel aquifers in theGlaciated Northeastern United States, in Randall, A.D.,and Johnson, A.I., eds., Regional aquifer systems of theUnited States—the northeast glacial aquifers: AmericanWater Resources Association Monograph Series 11,p. 37-61.

Lyford, P.P., Stone, J.R., Nielsen, J.P., and Hansen, B.P.,1998, Geohydrology and ground-water quality, EasternSurplus Superfund Site, Meddybemps, Maine: U.S.Geological Survey Water-Resources InvestigationReport 98-4174,68 p.

Mazzaferro, D.L., Handman, E.H., and Thomas, M.P., 1979,Water resources inventory of Connecticut, part 8,Quinnipiac River Basin: Connecticut Water ResourcesBulletin 27,88 p.

McDonald, M.W., and Harbaugh, A.W., 1988, A modularthree-dimensional finite-difference ground-water flowmodel: U.S. Geological Survey Techniques of Water-Resources Investigations, book 6, chap. Al, 586 p.

Melvin, R.L., deLima, V.A., and Stone, B.D., 1992,Stratigraphy and hydraulic properties of tills in southernNew England: U.S. Geological Survey Open-FileReport 91-481,53 p.

Pietras, T.W., 1981, Leaching of nutrients in two watershedsunder different land uses in eastern Connecticut: Storrs,Conn., University of Connecticut, unpublished master'sthesis, 302 p.

Roy F. Weston, Inc., 1997, Groundwater monitoring wellsampling summary report, Eastern Surplus CompanySuperfund Site, Meddybemps, Maine: Burlington,Mass.,70 p.,2 app.

1998a, Remedial investigation groundwatermonitoring well sampling summary report, EasternSurplus Company Superfund Site, Meddybemps,Maine: Burlington, Mass., 104 p., 7 attachments.

1998b, Remedial investigation residential wellsampling summary report, Eastern Surplus CompanySuperfund Site, Meddybemps, Maine: Burlington,Mass., 56 p., 3 attachments.

Tiedeman, C.R., Goode, D.J., and Hsieh, P.A., 1997,Numerical simulation of ground-water flow throughglacial deposits and crystalline bedrock in the MirrorLake area, Grafton County, New Hampshire: U.S.Geological Survey Professional Paper 1572,50 p.

Torak, L.J., 1979, Determination of hydrologic parametersfor glacial tills in Connecticut: Storrs, Conn., Universityof Connecticut, unpublished master's thesis, 161 p.

References 27

u.s. department of the interior u.s. geological survey ...water resources division 10 bearfoot rd....

Documents