Document VGqXbkKyBnBdGD9q3p7b9Gexp

AR226-2598 I ^ I I REVISED GROUNDWATER FLOW I MODEL DUPONT WASHINGTON WORKS I WASHINGTON, WV I I I "') * I Date: January 2003 I Project No,: 18983635 7423 I I I CORPORATE REMEDIATION GROUP An Alliance between DuPont and URS Diamond I Bariey Mill Plaza, Building 27 Wilmington, Delaware 19805 I w ASH020280 EID630708 I I I I I I tKSKjjl 'tW ",'' '|I I I I I I I I Revised Qroundwater Flow Modal _______________________Introduction TABLE OF CONTENTS 1.0 Introduction,-................................"..--.---..--...............................,,,.,...--..................,3 2,0 Site Setting, Geology, and Hydrogeology.................................................................. 4 2.1 Site Setting..........,,..................................,,....,,...,,,,,,.----................,,,,.,,.,,---- 4 2.2 Geology................................,.....................................--.-....----.------.--.---4 2.3 5 Hydrogeology.............................................................................................. 3.0 Primary Data Sources....................,...,...........,..,.......,,.............................,,........----. 7 4.0 Model Set-up................................................................................................. 9 4.1 Model Domain and iMscredzation................................................,,,,...........-- 9 4.2 Boundary Conditions ....,,,,...,,.....,,,,...,,.----..............,,,,.,,,........--.--............ 10 4.2-1 No Flow Boundaries ,,.,,,,...,,..,,.,,.....,,,,...........,,.........,,,,,,,,................ 10 4.2.2 Ohio River..................................................................................... 10 4.2.3 Minor Surface Water Features.................................................... 12 4.2.4 Pumping Wells................................................................................. 13 4.3 Main Input Values......--.........,........................,,..............-...........--.............. 16 4.3.1 Hydraulic Conductivity ............................................................ 16 4.3.2 18 Recharge.............................................................................. 5.0 Groiffldwater Model Calibration ..,,.,,................,,...,......,,...,,..,--.............................. 19 5.1 Calibration Strategy....................................................................................... 19 5.2 Calibration Results................................................................................. 20 6,0 Model Results.....,.,.........,.............,.,,.,.,,.-.-.-...................,..........----................2l 6.1 MassBalaflce...,............................,........,,,.,.--................,.,..,...--"..--..............21 6.2 Current Pumping Conditions........................................................................... 21 6.3 No DuPont Pumping.......................................................................... 22 6.4 No GE Pumping.............................................................................................. 23 6.5 No Pumping by DuPont or GE................................................................. 23 6.6 No Little HocldngWA Pumping......................................................... 23 7.0 Sensitivity Analysis....................,.,.,.,,.,,.,,..................................--....--.................... 25 8.0 Model Conclusions.......,..,.,,,,.--.........................................--....--..----...--...,,,.".26 9.0 27 References..............,.......,,.,.....-- .........,,...,,........,,.,,,,..,,..............,,..,,,,--...-..--.. 74Z3WWGWM-R01^oo Jan. 7.03 Wllmlnglon, DE ASH0202B1 EID630709 Revised Oroundwaier Plow Model Introduction TABLES Table 1 Primary Data Sources for the Gtoundwater Flow Model ,..,,., ...7 Table 2 Finite-Difference Grid Vertical Discretization .................. ...9 Table 3 Table 3. Riverbed Hydraulic Conductivity Values .............. ..11 Table 4 Model Pumping Rates .............................................. ,.14 Table 5 Aquifer Hydraulic Conductivity Values .......................... ..17 Table 6 Model Recharge Rates ............................................... .18 Table 7 Statistical Summary of Model Calibration Errors (Residuals) .20 Tableg Mass Balance Summary............................................ ..21 Table 9 Sensitivity Analysis Results...................................... ..25 FIGURES Figure I Figure 2 Figure3 Figure 4 FigureS Figure6 Figure 7 FigureS Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 14a Figure 15 Figure 16 Figure 17 Figure 18 Figure 19 Figure 20 Site Location Map Idealized Ohio River Valley Cross-Section and Block Diagram Terrace Surfaces Generalized Geologic Cross-Section at River Mile 190 Groundwater Elevation Map Model Finite Difference Orid Boundary Conditions Model Pumping Well Locations Hydraulic Conductivity Zonatkm Recharge Zonation Synoptic Water Level Measurement Locations Model Calibration Results Model Computed Head-Layer 1 Model Computed Head-Layer 2 Model Computed Heads-Layer 2-Site Area Model Computed Head-Layer 3 Predicted Groundwater Table with No DuPont Pumping Predicted Groundwater Table with No GE Pumping Predicted Gioundwater Table with No DuPont or GE Pumping Predicted Groundwater Table with No Little Hocking Pumping Sensitivity Analysis Results 7423WWeWM-R01,(toc Jan. 7,03 Wllmlngton, DE ASH020282 EID630710 RwlsodOrtiuntiwatef Plow Mods! Introduction 1.0 INTRODUCTION A steady-state groundwater flow model was developed for the DuPont Washington Works facility (the site) in 1999. This model, which was completed as part of the RFI, was documented in the RCRA Facility Investigation Report (DuPont, 1999). The original groundwater flow model has been revised and expanded. The primary objectives of the modeling woric documented in this report included: 0 Addressing EPA/ACOE comments on the previous groundwater flow model, Q Evaluating groundwater migration pathways at the site under current site and regional pumping conditions, and 0 Predicting future grouodwater migration pathways under various pumping scenarios. Feedback was sought from experts with the U.S. Environmental Protection Agency (EPA), U.S. Army Corps of Engineers (ACOE), and U.S. Geological Survey during the redevelopment of the model. Meetings were held with these experts at key stages in model development including completion of the conceptual geologic model, following preliminary model calibration, and then a-fter fitia] model calibration m order to give the experts the opportunity to review the work completed at each stage and to solicit feedback. The purpose of this report is to document the revised groundwater flow model for the DuPont Washington Works facility. The following sections describe the site setting, geology, and hydrogeology, and document the model set-up, calibration, and results. 7423WWGWM-RD1.doG Jan. 7, 03 WilmIiiBton, DE ASH020283 EID630711 Revised Groundwater Flow Model__________Sjite Setting, Gaology, and Hydrogeology 2.0 SITE SETTING, GEOLOGY, AND HYDR.OGEOLOGY I 2.1 Site Setting The site is located on Washington Bottom in Washington, West Virginia. The location of I the site and several surrounding properties are shown on Figure 1. The site, which is bounded on the north by the Ohio River, lies at the western (down-stream) end of Blermerhassett Island approximately between river miles 189 and 191. The DuPont site I shares its southwestern boundary with another manufacturing facility operated by General Electric (GE). I 2.2 Geology The site lies within the Ohio River Valley and rests on Quaternary alluvial sediments that I overly the Permian-aged Duokard Group. The two predominant facies of die Ohio River alluvium that have been identified include coarse-grained Ohio River Alluvium (Pleistocene-aged glacial outwash deposits) and fine-grained Ohio River Alluvium (Holocene overbank deposits(Simard, 1989). The Pleistocene deposits consist primarily of coarse-grained sand and gravel while the Holocene deposits consist primary of interbedded and laminated silt, clay and fine-grained sand. Simard also identified three minor facies of the Ohio River Alluvium including tributary deposits (silt, .sand and gravel), colluvial deposits (coarse sand, gravel, cobbles and boulders) and eolian sand and silt deposits. The Dunkard Group (bedrock) consists primarily of red and varicolored sandy shale; gray, green and brown sandstone; gray and light-gray siltstone; and minor beds of coal, claystone, black carbonaceous shale and limestone. The facies of the'Ohio River Alluvium formed in response to the glacial advances and I retreats of the pre-, early- and late-Wisconsinan and were deposited as successive phases ofaggradatlon and degradation of river valley alluvial materials. The coarse-grained I Pleistocene alluvium was deposited as glacial outwash during the primary valley aggradation event following (be glacial scouring of fee valley into the bedrock floor. During the subsequent degradation and aggradation cycles of the Pleistocene, the glacial I outwasb sediments were partially removed, re-worked and then redeposited to a lower elevation than the previous cycle, thus forming a terrace. This process formed a series of Pleistocene-aged terrace surfaces within the Ohio River Valley. These surfaces were I designated (youngest to oldest) as S4, S5, and S6 by Sfanard. Witfl each subsequent degredation/aggradation cycle, additional fines were incorporated into the Pleistocene deposits due to continual influx of finer-grained fluvial sediments from tributaries of the I Ohio River. As a result, the Pleistocene deposits become more highly re-worked and progressively finer-grained toward the center of the river valley, particularly in locations down-stream of significant tributaries (Siniard, 1989). The total thickness of Pleistocene sediments at Washington BWaWttnges from about 80 feet beneath the highest Pleistocene terrace surfaces to about 15 feet beneath the cuarent channel of the Ohio River. 7423WWGWM-ROt.ClOc Jan. T, 03 Wilmlngton, DE ASH020284 EID630712 Revised Grounqwaier Flow Model__________Site Setting, Geology, and Hydrogeology The Pleistocene alluvial deposits are overlain by the finer-grained Holocene sediments. The silts, clays, and fine sands were deposited on the surface of the Pleistocene terraces as well as on a series of more recent floodplains, which formed in the center of the Ohio River Valley during the Holocene. The thickness of the Holocene sediments typically ranges from 5 to 15 feet over the Pleistocene terrace surfaces and 25 to 35 feet over the Holocene floodplains. Simard designated the Holocene floodplain surfaces as Sl through S3 and the modern floodplain of the Ohio River as SO. Figure 2, modified from Simard I (1989), is a block diagram and idealized cross-section through the Ohio River valley depicting the complex set of Pleistocene terraces and Holocene floodplains which have formed in the valley, I Holocene floodplain surfaces and Pleistocene terrace surfaces were mapped using surface age and elevation data from Simard, 1989. A map of floodplain and terrace surfaces is I shown in Figure 3, Individual floodplain and ten-ace surfaces were not differentiated. Due to damming of (he river, the SO, Sl, and S2 Holocene floodplain surfaces are now flooded along the reach of the Ohio River at Washington Bottom. The S3 surface I remains above the current nonnal pool elevation of582-ft. MSL. A thin strip of this surface remains on the south side of the river along Washington Bottom. More extensive expanses of this surface are present along the norfb side of the river at Little Hocking I Water Association (WA), Blennerhassett Island, Shell Kraton, and Belpre. Additional expanses of this surface are present down-stream of Washington Bottom at Lubeck Public Service District (PSD). The remainder of the alluvial valley is occupied by the I Pleistocene terrace surfaces. A generalized north-south cross-section from the Little Hocking WAwell field in Ohio, through the Ohio River and across the DuPont facility, is presented to Figure 4, This cross-section shows (be floodplain and terrace surfaces, the Holocene silt and clay ; ^ overbank deposits overlying the Pleistocene sand and gravel outwash deposits and the re- worked Pleistocene alluvium in the center of me river valley. The alluvial terrace I deposits are underlain by a flat, river-scoured bedrock surface of the Dunkaid Group that rises steeply and forms the valley walls to the North of Little Hocking Water Association and to the south of the DuPont facility (Figure 3). I 2.3 Hydrogeology I Groundwater supplies in me region are obtained from the Dunkard Group bedrock and Ohio River alluvial terrace deposits. The saturated portion of the Ohio River alluvial terrace deposits comprise the principal regional aquifer used for water supply purposes. I Production wells completed in this aquifer have been known to yield up to 500 gallons per minute (gpm) (Schultz, 19B4). Based on these high yields^numerous industrial and commercial water supply companiesobtain water from the alluvial aquifer. I The Holocene silts and clays support perched groundwater zones. Along the southern riverbank at the DuPont plant site on Washington Bottom, are seven monitoring wells I mat are completed in this perched groundwater zone. Qroundwater elevations for these monitoring wells are typically 6 to 18 feet higher than elevations measured ia monitoring wells completed in the underlying primary site water-table aquifer, which is significantly I depressed as a results of pumping by DuPont and GE. During the February 2002 f 7423WWBWM-R01.dOC Jan, 7.03 Wilmington. DE ASH020285 EID630713 Revised aroundwater Flow Model___________Site Setting, Geology, and Hydrogeotogy synoptic groundwatcr level event, groundwater elevations in the perched water table at the site ranged ftom 571.91 feet above mean sea level (MSL) to 582.09 feet MSL (Figure 5). Groundwater elevations in the underlying primary site water-table aquifer ranged from 558.24 to 566.61 feet MSL. The Ohio River Alluvial Aquifer, which is the primary water-table aquifer in the area, occurs at a depth of about 60 to 70 feet below ground surface at the DuPont plant site. The saturated zone is approximately 30 to 40 feet thick, extending approximately to the surface of the underlying Dinakard Group bedrock. Numerous pumping tests have been completed in the alluvial aquifer in the Washington Bottom area as part of water supply investigations. The hydraulic conductivity of the alluvial aquifer in the area typically ranges from 100 to 300 ft/d. In contrast, the hydraulic conductivity of the underlying Dunkard Group bedrock aquifer is typically between 0.05 and 5 ftVd (see Section 4,3.1). Natural recharge to the alluvial aquifer comes from various sources, including: Q Infiltration of precipitation falling directly on the alluvium Q Lateral movement of the river water through the alluvium Q Seepage from stream tributaries that discharge to the Ohio River 0 Surface rim-off from the outcrop areas of the Dunkard Group, which forms steep slopes adjacent to the uppermost Pleistocene terrace. I I 1 I I I 7423WWGWM-R01,()OC Jan. 7.03 Wilmington, DE ASH0202B6 EID630714  Revised Sioundwater Plow Model r Primary Data Sources 3.0 PRIMARY DATA SOURCES Data was obtained from numerous sources for use in revising the Washington Works 1 grorodwater flow model. These sources and the nature of the data obtained from each source are summarized in Table 1 below. 1 Table 1. Primary Data Sources for the Oroundwater Flow Model. 1 ^S^^^^. : - """ WijsM ^ata'TyptW WPwat WtshingtoB Worta RCBA Facility Investigation Report, WcB and sofl 'boringlogs I DuPont CRG, 1S99 Groundwater elevation data Aquifel lestlDg results I Hydrogeotogic Evaluation for Additional "Water Supply fltnn WeHlog* BlenneibaEsett Island, Leggette, Brashears & Graham, Inc., 1986 Aquifel testing results BicnnerlBXsett Island 'Water Supply Well Drilling and Test Well log! Pumping, Burgess & Niple, Ltd., 19S8 Aquifer testing xesults Grain size analysis results Washington Worits 'i*relimm2iy Hydrogeolopc Asscsantait, Wen Bud soil boring logs I DuPont, 1991 Aquifer hydraulic conductivity data RCRA fwaSty Investigation Report SheB Kraton PiBrt, Bctpre, Well Bad soil boring logs I OlriOt all Chemical Owapssy, 1999 Aquifer hydraulic conductivity dato 1 Geologic History of <he Lowa- Teliaces and Floodjilams ofttie Upper Ohio River Valley, West Virginia Geological Survey Open PBe Report 8903, Siraatd, 1989 Lithology, ihiclnicss, and depwitioaal history of die Ohio River alluvial sediments Historical low water levels oflhc Ohio River I Terrace ages and elevations I Aquifcr-CharacteMtics Data for West Virginia, USGS WterRcsonrccs IsvestigBtions Report 01-4036, 'KjatWt sad Maltel, 2001 Aquifer hydraulic conductivity data Estimated recharge rates WcB Head Protection Plan, Little Hocking Water Association, 1996 WdllogB Aquifer hydraulic conductivity data I 7423WW6WhHW-(loc Jan. 7,03 7 PI Wllniinslon. DE ASH020287 E:tD630715 Revtaed Groundwater Flow Model Primary Data Sources Well Head Protection Area Survey, Lubeck Public Service District, 1998 Well logs Aquifer testing results I Geotechmcal Boring Logs, Planned 892 Corridor, Burgess & Niple, 2001 Lithologic descriptions I Oeotechnical Boring Logs, Planned 3B,t. 50 Coirridor D over Ohio River and Blenncrhassctt Island, Michael Baker Jr., Inc., 2002 Lithologic descriptions I Ohio River Bathymetric Data, ACOE, 2002 Monthly Operation Reports for Lubeck PSD, Little Hocking Ohio River water depth and bed elevations Daily piunpisg rates I WA, and City ofBelpre DuPont and Little Hocking Puinpmg Rate Data Hourly or daily rates from days of synoptic water measwernents Well Logs from Shell and GE Sites Lithologit descriptions I Synoptic Oroundwater Levels fiom Februwy 2,2002 and August Groundwater elevations at Lnbeck, OE, DuPom 21,2002 plant, Blenn. Isle., Little Hocking, Shell ,1 I I I I I I f 7423WW6WM-R01,(toc Jan. 7. 03 WImlngton, DE ASH0202B8 EID630716 I I I I I I? !* I I I I I I Revised Bmundwater Flow Model Model Set-up 4.0 MODEL SET-UP Groioidwater flow in the Ohio River Alluvial Aquifer at the DuPont Washington Works facility and the sun-oundmg area was simulated using United States Geological Survey's MODular Groundwater FLOW modeling code (MODFLOW. McDonald and Harbaugh. 1984). MODFLOW was selected because of its wide acceptance within the industry as well as its versanlily in simulating various types of complex groundwater flow boundaries. The grouodwater model was constructed and calibrated usbg Groundwater Vistas (v. 3.36), a graphicalpre- and post-processor that interfaces with MODFLOW. published by EnviroiHncntal Simulations, Inc. (ESI). 4.1 Model Domain and Discrifftization The Washington Works Groundwater Model domain covers an area of 39.5 square miles, extending 41,56? feet east-west and 26,394 feet north-south. The origin of the finite difference grid is located at UTM Coordinates of 4,341,000 northing and 437,800 easting (UTM, NAD 27 meters). The finite difiference model grid is shown on Figure 6. The model domain was cliscretized as follows: Q 153 rows by 235 columns Q Slayers Q 107,865 cells (69.705 active cells) Q Grid spacing: 25m to 100m (82 ft, to 328 ft:) The Ohio River Alluvial Aquifer is represented in the model by layers 1 and 2, while the bedrock aquifer (minor aquifer) is simulated as layer 3. The vertical discretization and bydrostratigraphic units simulated in each layer is summarized in Table 2 below. Table 1. Finite-Difference Grid Vertical Discretization. -1 "":""""""--i-- ..Model liayere - fc:,'.'':';"'':asw''{f'^t.-abovet.MSIi.JI Layer 1 Variable-based 560 on topography Layer 2 560 535 Layer 3 535 500 %!S.i'%tli''&tlrt'ttgrapUhnleit,s: ,,' Pine-grained Holocene sediments Pleistocene Sands and Gravels Bedrock - Dunkard Group Pleistocene Sands and Gravels Re-worked Pleistocene Sands and Gravels - main river channel Re-worked Pleistocene Sands and Gravels - Blenoerhassett bland Bedrock - Dunkaid Gioup Bedrock Dunkard Group 7423WWGWM-R(M.(tae Jan. T, 03 Wilmingtan. DE ASH0202B9 EID630717 Revised Gfoundwater Flow Model___________________________________Model Set-Up Tlie base of layer 1 is set at an elevation of 560 ft. above MSL throughout the model layer. This elevation corresponds approximately to the elevation of the contact between the fine-grained Holocene sediments and the underlying Pleistocene sand and gravel beneath the Holocene floodplains. The top of layer 2 coincides with the base of layer 1. The base of layer 2, which is also constant throughout the model, is set at an elevation of 535 ft. above MSL. This elevation corresponds approximately to the elevation of the contact between the Pleistocene sand and gravel and me underlying shales of the Dunkard Group (bedrock) within the river valley. These contact elevations are based on well and soil boring logs from the DuPont Washington Works plant site, Blennerhassett Island, and Little Hocking WA (Figure 4). 4.2 Boundary Conditions Under normal stress conditions, the alluvial aquifer would be expected to discharge to the Ohio River with infiltration being the predominant source of recharge to the aquifer. Due to the high volume ofgroundwater pumping from the alluvial aquifer along this reach of the river, the Ohio River is the main source of recharge to the alluvium within the area of the model domain while infiltration is a secondary recharge source. Correspondingly, discharge ofgroundwater from the alluvial aquifer within the area of the model domain occurs primarily through the numerous industrial and pubic supply wells. Minor volumes ofgroundwater recharge and discharge to the alluvial aquifer along several minor surface water features. The treatment of these boundary conditions in the groundwater flow model is described in the following sections. The model boundary conditions are shown in Figure 7. 4.2.1 No Flow Boundaries Many of the cells within the model domain are set as inactive (no flow cells). These cells are in areas of the domain where the alluvium is absent and the low permeability shales of the bedrock are present exclusively. The area of bedrock immediately surrounding me alluvial aquifer, both adjacent and below the alluvium, is included as active cells within the model to discretely simulate recharge to the alluvial aquifer from the bedrock, The area of active bedrock cells laterally surrounding the alluvial aquifer was delineated based on the estimated locations ofgroundwater divides within the bedrock aquifer. The locations of me divides were estimated from surface topography. Additional no flow cells were assigned to the center of the Ohio River in model layer 1 where the river bed elevation was below the base of that layer (560 ft. MSL). River bed elevation in each model cell was interpolated from Ohio River bathymetric data (US ACOE, 2002) 4.2.2 Ohio PUver The Ohio River is Einrnlated in the model as a MODPLOW river boundary. River boundary cells were set within the model domain based on the current river channel width and riverbed elevation. The river boundary cells representing the Ohio River were assigned to either model layer 1 or 2, depending on the elevation of the riverbed at the T423WWGWM-RO,doe Jan. 7. 03 10 Wilmington, DE ASH020290 EID630718 Revised Onmndwatef Row Model Model Set-up center of each model cell. Where the river bed elevation was below the base of layer 1 (560 ft. MSL), (be river boundary was assigned to layer 2 and the overlying cell in layer 1 was set as a no flow cell (inactive) as previously described. River boundaries allow &e simulation ofgroundwater recharge or dischargeon the basis I of the gradientbetween the predictedground-water bead surrounding fee river cell and fhe specifiedriver stage elevation. The volumetric rate ofgroimdwater flow across the I boundary is governed by the hydraulic gradient between the groundwater head and the river stage and the hydraulic conductance of the riverbed material. The conductance of the riverbed is calculated for each river cell independently, according to the following I fotinula: I CwraKLW/U Where: Cssr =' hydraulic conductance of the river-aqiufer interconnection K =a hydraulic conductivity of the riverbed material I L = length of the riverbed within the model cell W s- width of (fae riverbed within the model cell I M = thickness of the riverbed within the model cell Because entire model cells are contained within the Ohio River, the length and width of 11 the riverbed within each river boundary cell was taken to be the length and width of the I model grid cell The thickness of the riverbed was assumed to be 5 feet. The hydraulic conductivity of the rivesrbed material, which was adjustedduringmodel calibration, varied depending upon the nature of the geologic materials that the river cell was founded I upon (riveibed K for river cells overtying Holocene silt and clay were lower than those overlying Pleistocene sand and gravel). The final riverbed hydraulic conductivity values for river cells are summarized in Table 3 below. I Tables. Riverbed Hydiaulic Conductivity Values. I '^^'^l^^'l^^^^ I Holoceae alluvion - silt, clay, fine sand Re-worked Pleistocene alluvium - sand, gravel. Bill iuUc Coaductivity ^ol'aiysrtred-tft^d)'- 0.1 OJ Re-worked Pleistocene alluvium at west end ofBlGnnerbassett Island 30 I 7423WWjaWM-RO'UtoG,lm.!.0.3 11 Wilmlngton. DE . .... ASH020291 EID630719  I I I I I iliSI '<< "^fc;!;^ I I I I Revised Oroundwaler Bow Model___________________________ Mpd&l Set-Up The normal pool elevation for this Teach oflhe Ohio River in 582 ft. above MSL. River elevations measured at the Little Hocking WA and SheU Kraton plant on August 21, 2002 were 582.24 and 582,05, respectively. The river elevation should be higher at the Shell Kraton plant than at Little Hocking WA, which is down-stream, indicating that there is some error in one or both of these measurements. Therefore, the normal pool elevation of 582 ft above MSL was used as the river stage elevation in the groundwater flow model. A constant river elevation was used throughout the model domain (e.g., no stream fall was simulated) due to river damming, which has led to relatively insignificant change in river stage elevation within the model domain. 4.2.3 Minor Surface Water Features Numerous minor surface water features occur within the bedrock outcrop areas as well as on the alluvial terrace surfaces within fee model domain. The majority of these features are ephemeral streams in which the flow is limited primarily to storm water run-off. The stream reaches within the bedrock outcrop area are included in the model as MODFLOW drain boundaries. This boundary type is appropriate for simulating the ephemeral streams in the bedrock areas, which are minor discharge points for the bedrock aquifer but do not contribute significant recharge. Drain boundaries allow the simulation of groundwater discharge on the basis of gradient between the predicted head surrounding the drain and the specified drain elevation (5,c,, the surface water elevation). Similar to river boundaries described previously, the rate of flow across the boundary is governed by me hydraulic gradient between the surrounding groundwater head and the drain stage elevation and the hydraulic conductance of the drain bed material (streambed). Contrary to river boundaries, drain boundaries will not contribute positive flow into the model domain under inward gradient conditions (e,g., if the sunounding heads are lower than the specified boundary head). The drain elevation assigned to each drain boundary cell was based on the topographic elevation along the stream at the center of the cell. The drain conductance in each boundary cell was calculated using estimated stream dimensions (length, width and bed thickness) and the streambed hydraulic conductivity. A nominal streambed hydraulic conductivity of 1 ft. / d was used so as not restrict groundwater discharge along these boundaries. Three perennial streams are present within the model domain. These streams include Sandy Creek near Lubeck PSD, an unnamed stream across the Ohio River from Lubeck, and an unnamed stream that flows through the Shell Kraton facility on the north side of the Ohio River near Blennerhassett Island, It is likely that each of these streams provides some recharge to (he alluvial aquifer and has some influence on the local groundwater elevations. Therefore, these streams were simulated as MODFLOW river boundaries to allow for the simulation of this recharge component. The river boundary elevation in each cell was based on the topographic elevation along the stream. The stream conductance in each river boundary cell was calculated using estimated stream dimensions (length, width and bed thickness) and a streambed hydraulic conductivity of 0.1 ft. / d, which is appropriate for the Hblocene streambed sediments. 7423WWSWM-R01.!loc Jan, 7,03 12 Wilmington, DE ASH020232 EID630720 Revised Qroundwater Flow Model Model Set-up 4.2.4 Pumping Wells A total of 50 industrial and public water supply wells are located within the model domain. These wells, which are shown on Figure 8, are located at the following sites: Q GE (14 wells) Q DuPont Washington Works Plant (13 wells) Q Bleonerhassett Island (12 wells) Q Lubecic PSD (6 wells) 0 Little HocldngWA (4 wells) Q City ofBelpre (1 well) Forty-four of these -wells were actively pumping during the collection of synoptic groundwater elevation measurements used for model calibration. Therefore, pumping is simulated at these 44 wells in the grouodwater flow model. The pumping wells, which are all completed in the lower portion of the alluvial aquifer, are simulated in layer 2 of the model. In most cases, pumping rates (ft^/d)used in the model were calculated from total daily flows on February 02,2002 (day of synoptic groundwater level measurements), assuming continuous operation. The exception is pumping rates for Little Hocking WA wells, which were not pumping during the collection of synoptic groundwater levels on February 02- Additional groundwater level measurements were collected on August 21, 2002 while the Little Hocking pumping wells were active (See Section 5.1 Calibration Strategy). Tborefore^ pumping rates for these wells were calculated from hourly system flows on August 21,2002. The majority of the DuPont water supply wells are monitored individually, including the East Well Field wells (331 to 337), Bleimeriiassett Island wells (435 to 446) and me Ranney collector well. The pianpjng rates for these wells were calculated from the total daily flow of each well. Due to model convergence problems associated with model cells going dry around groups of closely-spaced pumping wells, the individual rates for the East Well Field wells were totaled and re-apportioned equally among the seven wells. The DuPont Ranney collector well, which includes the central, well and 6 lateral collector wells (horizontal), was simulated in the model as 5 individual wells with the pumping rate apportioned equally between them. The location of the 5 wells used to simulate the Raimey collector was based on the actual orientation and length of the 5 lateral collectors. Flow rates are not monitored at individual wells in the DuPont-Lubeck Well Field. Thus the total flow rate for the field was calculated from the total daily flow and apportioned equally among the 5 wells (DL-1 to DL-5). Total system abstraction rates for me Lubeck PSD and the City ofBelpre were obtained J&om Monthly Operation Reports for these supply systems, which were provided by the West Virginia Department ofEnvironmentel Protection or Ohio Environmental Protection Agency (OHEPA). The total system flow rate for each system was calculated from me total daily abstraction and apportioned equally between the pumping wells. T423WWGWM-R01.dOG Jan. T.OS 13 Wllmington, DE ASH020293 EID630721 Revised Oroundwatef Plow Model Model Set-up Hourly flow rates for Little Hocking WA were provided by OHBPA. Flow rates for the time of day during which groundwater levels were measured at Little Hocking were used in the modeL The total flow rate for the system was apportioned equally between wells PW-1, PW-2, and PW-3, all of which were pinnping when the groundwater levels were measured (August 21,2002). Pumping rate data for GE was unavailable. Pumping rates for toe GE wells were estimated from the "normal" operational pumping rates for individual wells, which were provided for fbe majority of the wells by GE, The actual pumping rates usually vaiy somewhat from the "normal" operation rates. Therefore, the pumping rates for several of the GB wells were adjustedduring model calibration to roughly match the observed draw-downs. The pumping rates used in the model are summarized m Table 4 below. Table 4. Model Pumping Rates. ''''^t^^iy^,^^ Belpre Bleimerhflssett Island Blemuxhassett Island Bletinfriiassctt Island Bleimeriiasaett bland Bleimerhassett Inland Blend^rhassBtt Island J^IamfitliABSOtt Islfixijd Dicsimfirliassett Island Bicimerhassfttt Island Blenncrhassctt bbod Blenncrhassctt Island BlenneAassctt Island DuPoat Wash. WodtS Plant DuP&at Wash. Wato Plant DnPont Wah- Works Plant DuPont Wash. Warfat Plant DuPont Wash. Wtttfa Plant DuPont Wash. Worte Plant DuPoat Wash. Worts Plant DuPont Warii, WoriB Plant DuPoiBt wash. Works Plant Disport WaA. Worts Plant DttPont Wash. Wwfa Plant WeHl 435 436 437 438 43? 440 441 442 443 444 445 446 331 332 333 334 335 336 337 DL-1 DL-2 DL.3 DL-4 Puniplitt6JR^iHireb2.M" odel Puroplue HstA (pm) f ^f-W^spa^'^-iS W 697 465 465 485 485 442 442 297 297 489 489 409 409 323 323 236 236 414 414 444 444 417 417 0 0 159 165 69 165 203 165 247 165 147 165 84 165 247 165 69 69 69 69 69 69 69 69 7423WWBWM-R01.doc Jan. 7,03 14 WIWngton, DE ASH020294 EID630722 I I I I I I I '^ I I I I Ravlsed Groundwirter Row Modal Model Set-up DuPont Wash. Works Plant DuPont Wash. 'Works Plant DuPont Wash. Worts Plant DuPont Wash. Works Plant EhiPoni Wash. Works Plant DuPont Wash. Works Plant DuPoat Wash- Works Plant Little Hocking WA1 Little Hocking WA1 Little Hocking WA1 Little HodnngWA' LubcckPSD LubcckPSD LubeckPSD LubeckPSD LubeckPSD LubeckPSD GE2 GE1 DL-S Gallery Well RanneyWell 1 RamieyWellZ RaimcyWell3 RjmneyWeU4 RanneyWell5 PW-1 PW-2 PW-3 PW-5 PW-A PW-B PW-C PW-D PW-E PW-F Well 3 Well 4 69 0 885 -- - .- -287 287 287 0 111 111 111 111 111 111 200 130 69 0 177 177 177 177 177 287 287 287 0 111 111 111 111 111 111 200 130 GE2 Well 5 0 0 GB2 Well 6 ISO 150 GE2 wen? NA 0 GE2 Wett8 150 150 GE1 Well 9 150 150 GE2 Well 10 NA 0 GE2 Well 11 NA 150 GE2 Well 12 60 60 OB2 Well 13 NA 100 GE2 Well 14 600 350 GE' Well 15 500 350 GE2 Well 16 NA 0 1. All pumpingrate me basefl on nicasurcd daily Hows fmm February 02,2002 exctpl for Litfle Hocking WA wells where Augua 21,2002 howly rates wire used. 2 - Pumpingnrte dil forQE wlls was unavailable. Pumping rates for GE write were estimmtd froin AssignOpacities and adjusted during model lalibtaflon. T423WWGWM-R01.dOC Jan. 7, 03 15 Wilmlnston.DE ASH020295 EID630723 I I I I I I I )< I I I I I I RevigadGroundwaler Flow Modal____________^_______________Model Sst-Up 4.3 Main Input Values 4.3.1 Hydraulic Conductivity Distribution of hydraulic conductivity values used in the model was based on predicted sub-surface distribution of Holocene and Pleistocene aged alluvial sediments and depositional history of those sediments as described in Section 2.2 Site Geology. Model layer 1 was discretized vertically such that all fine-grained Holocene sediments occur within this model layer (See Section 4.1). The vertical discretization of model layer 2 was such that only the coarser grained Pleistocene sediments, which underlie the Holocene sediments and overiie the bedrock, occur within this model layer. The top of model layer 3 coincides with the estimated elevation of (he top of the bedrock surface underlying the alluvium. Hydraulic.conductivity zonation used in the model is shown in Figure 9. The lateral distribution of hydraulic conductivity values in each model layer (zonation) reflects the predicted sub-surface distribution of alluvial sediments based on the terrace surface ages as previously described in Section 11. Areas of layer 1 that underlie Holocene aged floodplains in the central portion of the river valley were assigned hydraulic conductivity values appropriate for the silt and clay overbank deposits, which occur in the shallow sub-surface beneath the floodplains. Areas in both layer 1 and layer 2 underlying the higher Pliestocene-aged terraces were assigned conductivity values appropriate for the coarse sand and gravel glacial oulwash deposits, which occur in the sub-surface in these areas. The areas of layer 2 beneath the existing Ohio River channel were assigned a lower conductivity than the surrounding Pleistocene sediments to account for the significant degree of re-working of the Pliestocene sediments and contribution of additional fines which has occurred in the center of the river valley. Hydraulic conductivity values assigned to model layer 3 were appropriate for the shales and sandstones of the Dunkard Oroup. A range of appropriate hydraulic conductivity values was determined for each of the geologic units simulated in me flow model from available aquifer testing results and published literature values. Results from numerous aquifer pumping tests from water supply wells completed in the coarse-grained Pleistocene sand and gravels were considered in determining the appropriate range of values for these sediments. The appropriate range for the other geologic writs (Holocene sediments, re-worked Pleistocene sediments, and bedrock) were estimated from published literature values. The conductivity value for each of the geologic units was adjusted during model calibration, but maintained within the range of conductivity values deemed appropriate for the unit. The range of appropriate hydraulic conductivity values for each geologic unit, the source of me conductivity data used, and the conductivity values assigned to each geologic unit in the groundwater flow model are presented to Table 5 below. 7423WWGWM.R01.dOB Jan. 7, 03 16 Waifnington, DE ASH020296 EID630724 Revteea erountlwatM Flaw Motlcl Model Set-41^ Table 5. Aquifer Hydraulic Conductivity Values, GtologteTNt Holocene Alluvial Sediments ^/; I>taBrtpfli>n"'''; .; v-' ^'S^^'V^w^KM.) ^ ' '"',.- '' ',, '''','''' . ^S.-^'R^.,.',.'-^.^ ^rotBer' 'iaiir'? ^ 'tii-.i;-?....' ^SW^ 'i.;?' ;;.!!! ,'-i:i;- ;' , ^ t J ,1'SWwAls: Geonietriti .; .iMi-y"?' t ' \ '''' '.'' Interbedded iind laminated sflt, clay, 0.03 2,8 and&aesnd Model K Vahns tflrfd.) , Kx,Ky1.0 Kv=0.01 Data Sources) 4 Pleistocene Alluvial Sediments Coarse-grained sand and gravel 37 35 871 146 Ka!,Ky=300 1,2,3,6.7, KvlSO 8,9 Plicstocanc Sedun6St9 Underlying Bleanerittssett Island Re-worked Pleistocene Alluvial Sediments Coarae-grained sand nnd gravel latertieddcd aand and gavel wilfa silt Ounkard OtDUpBedrock' iBteibcddcd shale and sandstone with minor UlSifistDHjc aod coiil beds S 124 374 235 &t,Ky=200 1,6 Kv-100 2.8 284 Kit^-y-30 4 K.v-3 3 0.02 5.6 02 &tiKy0.1 5 Kv"lxl0'4 NoB t. VsmSu flom Wwx? Coanty onty OWB utMl IB Hie imlyri. Hyfliwllc cTMaucriity latellnted BOTOreponed ttannmBivity md Win dqilli. DirttScurcet: 1. Buiseai&Niplc.Ltl-.l&ga 2. DttPwit, 1%1 3. Dul'BttaW.1999 *, FMW.198B 5. KOKB-md MathK, 2001 6. LeHjcttE.Bnuhnre&GraIiani. inc., 1986 7. Unie BockiBg'WiltCT AiBodHicn, 1996 8. LubedcPublic Service Eftfel,lSi 9. SheB,19% 7423WWGWM-R01.stoc Jan. 7,03 Wllmlnolan, DE I? ASH020297 E1D630725 Revlfd Greundwater Flew Model Model Set-up 11 4.3.2 Recharge ^|| Distribution of recharge values used in the model was based on the predicted distribution igl of sitfficial soil types and thicknesses as described in Section 2-2 Site Geology. Pleistocene terraces, Holocene Boodpkdns, and bedrock outcrop areas were assigned d i f f e r inin th g r ese ar e e c a h s arge due ra to tes to acc differing ount soil for varying types and th infiltrati icknesse on rat s. An es ar t e h a at o would f increa be anticipate sed recharge d was assigned along the outer edges of the Pleistocene terraces adjacent to the bedrock o u t c r o p areas to account for increased infiltration of storm water run-off from the bedrock slopes in these areas. In addition, reduced recharge rates were assigned to large areas of impervious surface cover at fhe GE and DuPont plant sites. Recharge zonation used in the model is shown in Figure 10. The recharge rates for the alluvial surfaces were based on estimates made by USGS (M. 1 ^ D. Kozar, oral communication). The rates for bedrock areas were estimated on the basis of topography and soil types present A low infiltration rate was assigned to the bedrock areas to account for fhe steep slopes and low pcnaeabitity of the surficial soils in these a r e a s .groiin Recha dwater r f ge r low ates we model ns ar adjustedduring mode e presented in Table 6 l calibra below, t i o n . The rates used in the tt Table 6. Model Recharge Rates. r . - '..-'^ilS^sai&d^^^^ I t-1": . ,;,,.,,'^<^&.^:|^..;s^:^;^^^ Holocene floodplains pHiec^irgeRat(ein*/ yr) ' 'y,*,;;-"-!-''. ,.. .' 4 )< Pleistocene terraces 8 1 Pleistocene terraces along bedrock outcrop 20 I Bedrock and areas ofnxipervious surface cover 0.1 I I I 1 1 T423WWSWM'R01.doc Jan. 7, 03 18 WllmlnBton, DE ASH0202&8 EID630726 Revised Groundwater plow Model Groundwater Model Calibration 5.0 GROUNDWATER MODEL CALIBRATION 5.1 Calibration Strategy The steady-state groundwater flow model was calibrated against two sets of synoptic groundwater elevation measurements that were collected specifically for use in model calibration. The first set was collected on February 02,, 2002. During this event, groundwater levels were measured in a total of 84 observation wells. Levels were measured at the following sites: Q DuPont Plant (44 wells) 0 Blennerhassett Island (8 wells) Q Little Hocldng WA (8 wells) Q LubeckPSD(6wells) Q GE Plant (13 wells) 0 Shell Kraton Plant (5 wells) It was determined that the pumping wells at Little Hocking WA were inactive at the time of measuring groundwater levels on February 02,2002. Therefore, a second round of groundwater level measurements was made at the Little Hocldng WA site on August 21, 2002 to obtain groundwater elevations for this site under active pumping conditions. Groundwater levels were also measured in selected wells at the DuPont Plant (15 wells) and the Shell Kraton Plant (1 well) for comparison between the two monitoring events. The extent of seasonal variation in groundwater elevation between the two monitoring events was determined by comparing groundwater elevations from the two events from well MW-11 at the Shell Kraton Plant. This well was used for the evaluation because it was the only well monitored during both events that was unlikely to be influenced by pumping activities DuPont and Little Hocldng WA. The groundwater elevation decreased by 0.55 feet at MW-11 between the February and August monitoring events. This seasonal variation is likely to be insignificant relative the variation in groundwater elevation that occurs in response to changes in pumping rates wifhin the model domain. Given the small seasonal variation observed in groundwater elevation, the two sets of groundwater elevations were combined and used for model calibration targets (Figure 11). With the exception of those from Little Hocldng WA, groundwater elevations from February 02,2002 were used as the calibration targets in the tnodsL The August 21, 2002 groundwater elevations from the Little Hocking WA site were used as model canbration targets in place of those from February in order to calibrate the model against groundwater elevations at this site under the influence of active pumping. Similarly, pumping rates from February 02,2002 were used for all pumping wells other than those at Little Hocking WA where the August 21,2002 rates were used instead (see Section 4.2.4 Pumping Wells). 7423WW6WM.R01.doc Jan. 7,03 ^ Wllmington, D6 ASH020299 EID630727 I I I I I I I ')- I I I I I I I Revised Groundwater Plow Model Croundwater Model Calibration 5.2 Calibration Results The results of the model calibration indicate that the groundwater flow model is accurately predicting the groimdwater heads in areas where groundwater elevation data is available. The overall success ofmodel calibration was measured in terms of the error (residual) between the observed (measured) groundwater elevation or head at each calibration target (observation well) and the head computed by the model at (he target location. Calibration target residuals from the calibrated flow model along with a plot of observed versos calculated head are shown in Figure 12. The plot of observed versus computed heads shows a very good correlation between observed and computed head for a regional-scale groundwater flow modal. The success of model calibration can also be dcmcmstrated through statistical analysis of the srois between observed and cornfnrted heads (target residuals). One particular measure is the overall residual error expressed as the peieeotage of the raoge in observed head, which is calculated as standard deviation of the residuals divided by fne observed range in head. In this case, fhestandatd deviation is 9.9% of the observed range. A result of 10% or less is generally considered to represent a good calibration. Table 7 below srounarizes the results of the statistical analysis of the target residuals. Table 7. Statistical Swniaary of Model Calibration Errors (Residuals). -------------"iii':"- :.^ " A^ i ^...., ^^fe^a.d Hea^ ^ ^"n" ,,"->? ir.,,fl"-" ''^ifcS, peratioi < % .i|Sy,;:;|J|,ai,i,,;^,,.;,, Residual Mean (ft.) Residual Standard Deviation (ft0 Sum of Squared Rwiduals (ft2) Absolute Residual Mean (ft,) Minimum Residual (ft.) Maadimun Residual (ft.) Range m Observed Head (A.) Residual Standard Deviation / Range in Observed Head (fL/ft.) Result 0.104 3.285 &07.491 2.308 -9.249 12.060 33.350 0.099 74Z3WWBWM--fl01 .doc Jan. T. 01 20 Wllinlnatoft, BE ASH020300 EID630728 Revised GreuiKhraler Row Model Model Results 6.0 MODEL RESULTS 6.1 Mass Balance The total water balance (mass balance) of the groundwater flow model indicates very little eiror in the simulation. The total error (total inflow minus total outflow) is less than 0.01 percent, A summary of the mass balance for the groundwater flow model is shown in Table 8 below. TableS. Mass Balance Summary. - '' $";l;^!<3olnpOIl^tl:'^a!"^ Storage ^il^Iltr '.^ 0 Constant Head 0 Wells 0 Drams 0 Recharge 4,343,758 River Leakage 17.581300 Total 21,925,058 Absolute Error (In - Out) Percent Discrepancy Out 0 0 20,818,613 3,947 0 1103300 21,925,860 .802 -0.0037 0.2 Current Pumping Conditions The groundwater flow model is calibrated to observed groundwater elevations under the existing pumping conditions at sites within the model domain (as of February 02,2002), The computed beads for the calibrated flow model for layers 1,2 and 3 are shown in Figures 13,14, and IS, respectively. With the exception of local areas surrounding major pumping centers, the computed heads within the Pleistocene alluvium in layers 1 and 2 are approximately the same. Computed beads in the Holocene alluvium in layer 1 are slightly higher (less than I foot) than in the underlying Pleistocene alluvium. A slight downward gradient is predicted in most areas of the alluvial aquifer in response to pumping from model layer 2. Increased downward gradient is predicted in layer 2 locally around pumping centers. A groimdwatcr divide is predicted beneath the Ohio River in the Pleistocene alluvion (layer 2). A divide is predicted beneath the main channel and on either side of Blenaerhasseitt Island. The divides separate me areas of significant draw-down centered ?423VlW<3VWH<01.(toc Jan. 7,03 21 wnmineton.DE ASH020301 EID630729 Revised Groundweter Flow Model Model Results at DuPont and GE, Blemierhassett Island, and Little Hocking WA from one another. Due to the presence of (bis divide, no groundwater migration pathway is predicted beneath river in the alluvial aquifer under the current pumping rates. The only exception is at the east end ofBleiaierhassett Island where the river has not incised as deeply into the I underlying Pleistocene alluvium (the entire width of the river boundary remains in layer 1). At this location, a groundwater divide is absent in layer 2 and the cone of depression from Belpre extends southward across the river. Figure 14a shows the approximate I extent of the capture zones associated with the DuPont and Little Hocking Pumping systems. Computed heads in bedrock areas generally decrease from layer 1 to layer 3. An upward I gradient from the bedrock to the overlying alluvial aquifer is predicted throughout most of the domain, with the highest gradient predicted near pumping centers. The highest predicted upward gradient is about 9 ft. at Blennerhassett Island. I A groundwater divide is not predicted in the bedrock aquifer beneath the Ohio River, Groundwater flow from Little Hocking WA to the DuPont and GE plant sites is predicted I in the bedrock aquifer. However, the bedrock aquifer is simulated discretely in the model in order to provide recharge to the overlying alluvial aquifer. Groundwater elevations in the bedrock aquifer were not calibrated and therefore should be considered an I approximation. Simplification of complex stratification of the bedrock aquifer may have resulted itt inaccuracies in predicted groundwater elevations in the bedrock aquifer. This simplification has little or no effect on predictions of groundwater flow in alluvium, The majority of the groundwater in the alluvial aquifer at the DuPont plant site i$ currently being captured by the on-site pumping activities. Some limited off-site groundwater migration may be occurring in the northwest comer of the DuPont plant site }t in response to pumping at GE wells 3 and 4. 6.3 No DuPont Pumping A simulation was nm to determine the influence that pumping activities at the DuPont plant and Blennerhassett Island are having on groundwater elevations at surrounding sites. To complete this simulation, pumping wells at the DuPont plant and at Blennerhassett Island were tuned-off while leaving all other pumping well rates as per the calibrated simulation. The predicted heads in model layer 2 for this scenario are shown in Figure 16. The model predicts an increase m groundwater heads in the central areas of the DuPont and GE plant sites of approximately 15 and 9 ft, respectively. An increase of approximately 2 ft. is predicted at Lubeck PSD. I The model predicts no change in grouadwater heads at Little Hocking WA. Shell Kraton, or Belpre would occur in response to cessation in pumping by DuPont. Conversely, this simulation suggests strongly that current pumping activities at the DuPont plant and I Blennerhassett Island are not influencing groundwater heads at these sites. f 7423VWVGWM-R01 .doc Jan. 7,03 22 Wiliriingtori, DE ASH020302 EID630730 I ^ Revised CfoUftjwaierFlow Model Results __________________________Modol 6.4 No GE Pumping Similar to the previously described simulation, a simulation was iun to investigate the influence that pumping activities at the GE plant are having on groundwater elevations at I surrounding sites. To complete this simulation, pumping wells at the GE plant were tuned-off while leaving all other pumping well rates as per the calibrated simulation. The predicted heads in model layer 2 for mis scenario are shown in Figure 17, I The model predicts an increase in groundwater heads in the central areas of the DuPont and GE plant sites of approximately 10 and 16 ft., respectively. An increase of approximately 3.5 ft. is predicted at Lubeck PSD. I Again, the model predicts no change in groundwater beads at Little Hocking WA would occur in response to cessation m pumping by GE. I 6.5 No Pumping by DuPont or GE I The previous 2 simulations were combined to investigate the influence that pumping at Little Hocking WA would have on groundwater elevations at Washington Bottom if both DuPont and GE ceased pumping activities. To complete this simulation, pumping wells I at the DuPont and GE plant sites were tuned-off while leaving all other pumping well rates as per the calibrated simulation. The predicted heads in model layer 2 for this scenario are shown in Figure 18. I The model predicts an increase in groundwater beads in the central areas of the DuPont and GE plant sites of approximately 22 and 24 ft,, respectively, which is equivalent to the amount of draw-down. An increase of approximately 4 ft. is predicted ai Lubeck PSD. X Again, the model predicts no change in groundwater heads at Little Hocking WA would occur in response to cessation in pumping by DuPont and GE, However, with no I pumping on the south side of the Ohio River, fhe model predicts that (he groundwater divide below the river would be overcome by pumping at Little Hocking WA and the capture zone for that site would extend across the river to the south. I 6.6 No Little Hocking WA Pumping I A final simulation was run to determine the influence that pumping activities at me Little Hocking WA are having on groundwater elevations at surrotthdmg sites. To complete mis simulation, pumping wells at Little Hocldng were tuned-offwhile leaving all other I pumping well rates as per the calibrated simulation. The predicted heads in model layer 2 for fliis scenario are shown in Figure 19. The model predicts an increase in groundwater heads in the central area of the Little I Hocking well field of approximately 7-ft. An increase in groundwater elevation of about 0.5 ft. is predicted at the Shell Kraton site. No change in groundwater elevation is predicted at Belpre, Blennerhassett Island, the DuPont and GE plant sites, or at Lubeek I PSD in response to cessation ofpumping at Little Hocldng. 1 7W4i2lm3lWngWloGnW, DME-R01.tloc Jan. 7.03 23 ASH020303 EID630731 Similar to die previous simulation, the model predicts that with no pumping on the north side of the Ohio River, the groundwater divide below the river would be overcome and the DuPont plant capture zone would extend across the river to the north. 7423WWGWM-R01.(loc Jan. r, 03 Wllmlrigton. DE 24 ASH020304 EID630732 Revised Sroundwater Flow Model Sensitivity Analysis 7.0 SENSITIVITY ANALYSIS A sensitivity analysis was completed to determine the sensitivity of the model to uncertainties in the primaly input parameters including hydraulic conductivity, recharge, and river boundary conductance. This analysis involved pertiubing tile input value for a single input parameter by fixed increments and re-running the model. The overall calibration error from each of the sensitivity runs were then evaluated and compared to detenuine the relative sensitivity of the model to errors in each of the main input parameters. For each input parameter, multiplication factors of 0.25,0.75,1.0,1.25, and 1.5 were used in the sensitivityanalysis, The input values perturbed during the sensitivity analysis and the relative sensitivityof toe model to each value are summarized in Table 9 below. Graphs of the absolute residual mean and mean change in computed head for each input parameter tested during the sensitivity analysis are shown in Figure 20. The sensitivity analysis indicates that the model is most sensitive to uncertainties in the hydraulic conductivity value assigned to the re-worked Pleistocene alluvium beneath the Ohio River, The reduced conductivity of this unit serves to restrict the flow of groundwater beneath the river and thus strongly influences the response to pumping predicted by the model- Table 9. Sensitivity Analysis Results. '.; .Input PararoeterA'^'iil few^'h :..,,, :';,^ ^ - ^ ^ ' m ^ ^ ".i,;te';;r^, :,,... ',;;. ^'''.'"Si-'l':;;?,,,,,!,,;!?5-;! .;.,,,,,.,,.,,. Hydraulic Conductivity Holoceoe alluvium ^Relative Model 'SeM8Uytty(ranlted jBroinl.lMgheBtlta ^ lOpowegfl) 7 Pliestocene alluvium 2 PHestocene alluvium - Blennerhassett Island 6 Re-worked Pliestocene alluvium 1 Recharge Holoeene floodplams 10 Pleistocene terraces 4 Pleistocene ten-aces along bedrock outcrop 5 River Boundary Holocene alluvium 8 Conductance Re-worked Pleistocene aUuvimn 3 Re-worked Pleistocene alluvium at west end of 9 BlennerhasseU Island 7423WWGWlifrR01Ae Jan.7,fl3 25 Wllmington, D6 ASH02030S EID630733 Revised Broundwater Flow Model HMI"oMdWel vConclusions 8.0 MODEL CONCLUSIONS The primary conclusions of the revised Washington Works Groundwater Flow Model are summarized as follows: Q The Ohio River is creating a groundwater divide within ihe Pleistocene sediments beneath the river. Due to the presence of this divide, no groundwater migration pathway is predicted beneath river in the alluvial aquifer under the current pumping rates at the various pumping centers simulated in the model. Q A groundwater divide is not predicted in the bedrock aquifer beneath the Ohio River. Groundwater flow from Little Hocking WA to the DuPont and GE plant sites is predicted in the bedrock aquifer. However, groundwater elevations in the bedrock aquifer were not calibrated and should be considered on approximation. Q The maj ority of the groundwater in the alluvial aquifer at the DuPont plant site is currently being captured by the on-site pumping activities. Some limited off-site groundwater migration may be occurring in the northwest comer of the DuPont plant site in response to pumping at GE wells 3 and 4, 74Z3WWGWM-R01 .floe Jan, 7, 03 Wllmington, DE 26 ASH020306 EID630734 Revised Oroundwater Flow Model Ref9 Fences 9.0 REFERENCES Michael Baker Jr., Inc. 2002. Geotecnnical Boring Logs, Planned Rt. 50 Corridor D over Ohio River and Btennerhassett Island, unpublished. Burgess & Niple, Ltd, 1988. Blennerhassett Island Water Supply Well Drilling and Test Pumping^ unpublished report. Burgess & Niple. 2001, Geotechmcal Boring Logs, Planned 892 Corridor, unpublished. DuPont, 1991. Washington Works Preliminary Hydrogeologic Assessment, unpublished report. DuPont Corporate Remedialion Groups 1999. RCRA Facility Investigation Report, DuPont Washington Works, Washington. West Virginia, unpublished report. Fetter, C.W-1988. AppliedHydrogeology, Second Edition. MacmWm Publishing Company, New Yo, New York, 592 p. Leggette, Brashears & Graham, Inc. 1986. Hydrogeologic Evaluation for Additional Water Supplyfrom Blennerhassett Island, unpublished report. Kozar, M.D, aadM.V. Mathes, 2001. Aquifer-Characteristics Data for West Virginia. Water-Resources Investigations Rjeport 01-4036, United States Geological Survey, 74 p. Little Hocking Water Association. 1996. Well Head Protection Plan, unpublished report. Lubeck Public Service District 1998. Well Head Protection Area Survey, unpublished report. Shell Chemical Company, 1999, RCRA Facility Investigation Report, Shell Kraton Plant, Belpre, Ohio, unpublished report. Simard, C.M, 1989. Geologic History of the Lower Terraces and Floodplains of the Upper Ohio River Valley, Open File Report 8903, West Virginia Geological Survey, 160 p. U.S. Army Corps of Engineers. 2002. Ohio River Bathymetric Data, survey date: July 2002, unpublished. 7423WWOWM.R01.doC Jan. 7,03 27 Wllmington, OS- A3H020307 EID630735 3 0 C 3) m en ASH020308 EID630736 FIGURES ASH020309 EID630737 ASH020310 EID630738 ASH020311 EID630739 ASB020312 EID630740 PLEISTOCENE TERRACE HOLOCENE Fl-OODFLAIN HOlfiCENE n-oonpLAiN r^t r OUPO Tuna LEtSeNO; 'A MOtOCENEq<10!BAM<l<ISnSSILT WO CLAY '.6Q... 6o:|| POtOBUSR'TMOCSAEN(D6SALNAD(OaMRA.OVUBT. OASHOEP08tIS- REWO'RKEO PlfiSTOCEIE ALLUVIUM SWOANOGRWa. BEDROCK HORtZOHTA- SSVl-e. h=- Ss Coiporot* Bint Sirir/UI} rew KllilJW. C. ASH020314 BID630742 ASH020315 EID630743 ASH020316 EID630744 ASH020317 EID630745 AS8020318 EID630746 ASB02031& EID630747 ASH020320 EID630748 ASH020321 EID630749 ASH020322 EID630750 ASH020323 EID630751 ASH020324 EID630752 Asao2032S EID630753 ASH020326 EID630754 ss^oesaia /.2E030HSV ASH020328 BID630756 ASH020329 EID630757 ASH020330 EID63075S Bariav Mill Plaza. BuMIng 27_ Lancaster Pite & Poms 1^1 jyyilmlnqton. PS 19a35 j ASH020331 EID630759