Hydrogeology Journal

, Volume 18, Issue 8, pp 1855–1866

Simulated impacts of artificial groundwater recharge and discharge on the source area and source volume of an Atlantic Coastal Plain stream, Delaware, USA

Authors

    • Department of Natural Resources and Environmental Control
  • Judith M. Denver
    • US Geological Survey
  • Thomas E. McKenna
    • Delaware Geological SurveyUniversity of Delaware
  • William J. Ullman
    • School of Marine Science and PolicyUniversity of Delaware
Report

DOI: 10.1007/s10040-010-0641-x

Cite this article as:
Kasper, J.W., Denver, J.M., McKenna, T.E. et al. Hydrogeol J (2010) 18: 1855. doi:10.1007/s10040-010-0641-x
  • 261 Views

Abstract

A numerical groundwater-flow model was used to characterize the source area and volume of Phillips Branch, a baseflow-dominated stream incising a highly permeable unconfined aquifer on the low relief Delmarva Peninsula, USA. Particle-tracking analyses indicate that the source area (5.51 km2) is ∼20% smaller than the topographically defined watershed (6.85 km2), and recharge entering ∼37% of the surface watershed does not discharge to Phillips Branch. Groundwater residence time within the source volume ranges from a few days to almost 100 years, with 95% of the volume “flushing” within 50 years. Artificial discharge from groundwater pumping alters the shape of the source area and reduces baseflow due to the interception of stream flow paths, but has limited impacts on the residence time of groundwater discharged as baseflow. In contrast, artificial recharge from land-based wastewater disposal substantially reduces the source area, lowers the range in residence time due to the elimination of older flow paths to the stream, and leads to increased discharge to adjacent surface-water bodies. This research suggests that, in this and similar hydrogeologic settings, the “watershed” approach to water-resource management may be limited, particularly where anthropogenic stresses alter the transport of soluble contaminants through highly permeable unconfined aquifers.

Keywords

Numerical modelingGroundwater/surface-water relationsSource areaSource volumeUSA

Impacts simulés de prélèvement et de réalimentation artificielle de l’aire d’alimentation d’un ruisseau de plaine côtière atlantique, Delaware, Etats-Unis

Résumé

Un modèle numérique d’écoulement souterrain était utilisé pour caractériser le bassin versant d’alimentation du Phillips Branch, un ruisseau généralement peu influencé par lesécoulement superficiel qui recoupe un aquifère libre fortement perméable dans le relief faible de la Péninsule de Delmarva, Etats-Unis. Le suivi des particules indique que la surface d’alimentation (5.51 km²) est inférieure d’environ 20 % à l’aire du bassin versant topographique (6.85 km²), et que environ 37 % du volume de réalimentation du bassin versant superficiel ne se retrouve pas à l’exutoire du Philips Branch. Le temps de séjour souterrain est compris entre quelques jours et 100 ans, avec 95 % du volume total étant « étant évacué » en moins de 50 ans. La décharge par pompage altère la forme du domaine d’écoulement et réduit le débit de base lié à l’interception des chenaux d’écoulement, mais a un impact limité sur les temps de séjour des eaux souterraines alimentant le débit de base. Par contre, la réalimentation artificielle par les installations d’assainissement diminue l’aire d’alimentation, réduit le temps de séjour en évitant les cheminements longs vers le ruisseau, et augmente l’alimentation des masses d’eau superficielles adjacentes. Cette recherche suggère que, dans un contexte hydrogéologique similaire, « l’approche par bassin versant » de la gestion des eaux peut présenter certaines limites, notamment lorsque des contraintes anthropiques modifient le transport de polluants solubles à travers des aquifères libres fortement perméables.

Impactos simulados de recarga y descarga artificial de agua subterránea en área y volúmenes de la fuente en una corriente de la Atlantic Coastal Plain, Delaware, EEUU

Resumen

Un modelo de flujo numérico de agua subterránea fue usado para caracterizar el área y volúmenes fuente de Phillips Branch, una corriente dominada por flujo base que incisiona un acuífero no confinado altamente permeable en el relieve bajo de la península Delmarva, EEUU. El análisis de seguimiento de partículas indican que el área fuente (5.51 km2) es ∼20% más pequeña que la cuenca definida topográficamente (6.85 km2), y que la recarga entrante que es ∼37% de la superficie de la cuenca no descarga al Phillips Branch. El tiempo de residencia del agua subterránea dentro del volumen de la fuente varía entre pocos días a a casi 100 años, con 95% del volumen “escurrido” en 50 años. La descarga artificial a partir del bombeo del agua subterránea altera la forma del área fuente y reduce el flujo de base debido a la intercepción de las trayectorias del flujo de la corriente, pero tiene impactos limitados en el tiempo de residencia del agua subterránea que descarga como flujo de base. En contraste, la recarga artificial a partir de la disposición de aguas residuales en la tierra reduce sustancialmente el área fuente, disminuye el intervalo del tiempo de residencia debido a la eliminación de viejas trayectorias del flujo hacia la corriente, y conduce una descarga incrementada a cuerpos adyacentes de aguas superficiales. Esta investigación sugiere que, en esta y similares configuraciones hidrogeológicas, el enfoque de la “cuenca de drenaje” para el manejo del recurso agua puede estar limitado, particularmente donde las tensiones antropogénica alteran el transporte de contaminantes solubles a través de acuíferos no confinados altamente permeables.

人工地下水补给和排泄对美国Delaware大西洋滨海平原某河流源区面积和体积的影响的模拟

摘要

应用地下水流动数值模型描述Phillips Branch河源区的面积和体积。该河以基流为主, 切割了美国地势平缓的Delmarva半岛上一个高渗透潜水含水层。质点径迹分析表明, 补给区 (5.51 km2) 较之按地形圈定的流域面积 (6.85 km2) 小 ∼ 20%, 地表流域 ∼37%的补给量未泄入Phillips Branch河。源区地下水滞留时间自几天至接近100年, 95%在50年内。由于阻截了径流路径, 抽取地下水造成的人工排泄改变了补给区的形状, 并使基流量减少, 但是对以基流形式排泄的地下水的驻留时间的影响有限。相反, 由于来自陆地污水处理场的人工补给消除了以前到河流的径流路径, 明显缩小了源区面积, 并减小了驻留时间的变化范围, 导致向邻近地表水体的排泄量增加。该研究表明, 在这种和类似的水文地质背景中, 水资源管理中的流域方法可能是有局限的, 特别是在人类活动改变了强透水潜水含水层中可溶污染物迁移的地区。

Simulação dos impactes da recarga artificial de águas subterrâneas e da descarga na área de origem na origem do volume de água de um ribeiro da Planície Costeira Atlântica, Delaware, EUA

Resumo

Foi utilizado um modelo numérico de fluxo de águas subterrâneas para caracterizar a área de origem e o volume de água de Phillips Branch, um ribeiro dominado por fluxo de base incidindo num aquífero livre de elevada permeabilidade nas terras baixas da Península Delmarva, EUA. Análises de rastreio de partículas indicam que a área de origem (5.51 km2) é aproximadamente 20% menor do que a bacia hidrográfica definida topograficamente (6.85 km2), e que a recarga que ocorre em cerca de 37% da superfície da bacia hidrográfica não descarrega para o ribeiro Phillips Branch. O tempo de residência das águas subterrâneas correspondentes aos volumes de descarga no ribeiro varia de poucos dias a quase 100 anos, com 95% do volume de “saídas” a ocorrem no prazo de 50 anos. A descarga artificial efectuada pelo bombeamento das águas subterrâneas modifica a forma da área de origem e reduz o fluxo de base, devido à intercepção das linhas de fluxo do ribeiro, mas tem impacte limitado sobre o tempo de residência das águas subterrâneas descarregadas como fluxo de base. Ao contrário, a recarga artificial por águas residuais reduz substancialmente a área de origem, diminui o intervalo no tempo de residência, devido à eliminação das linhas de fluxo antigas que corriam para o ribeiro, e leva ao aumento da descarga para massas de água superficiais adjacentes. Esta pesquisa sugere que, nas condições hidrogeológicas presentes e similares, a abordagem da “bacia hidrográfica” para a gestão dos recursos hídricos pode ser limitada, especialmente quando as pressões antropogénicas alteram o transporte de contaminantes solúveis em aquíferos não confinados de elevada permeabilidade.

Introduction

Watershed areas delineated on the basis of land-surface topography often serve as regions of interest for the protection, management, and restoration of surface-water quality (USEPA 2008). This approach is generally reasonable where regional topographic relief is high and overland flow is the primary mechanism of contaminant transport to surface water (exceptions are cited in Winter et al. 2003). This approach is less reliable, however, in hydrogeologic settings where regional topographic relief is low, unconfined aquifers are highly permeable and incised by baseflow-dominated streams, and contaminant transport to streams is dominated by sub-surface pathways. In such settings, effective water-resource management requires an understanding of a stream’s source area and volume, which collectively comprise the three-dimensional aquifer space that receives recharge and contributes baseflow to a stream (Modica et al. 1997). Information on the direction and rates of groundwater movement within the source volume is also needed to quantify both the spatial and temporal aspects of contaminant fluxes.

Source areas and volumes have distinct topologies within the regional groundwater-flow system that are determined by aquifer recharge, thickness, anisotropy, and heterogeneity in hydraulic properties (Modica et al. 1997). Characterizing the source areas and volumes of streams incising high-permeability, unconfined sand aquifers in regions of low topographic relief is difficult because: (1) the water table does not necessarily mimic land-surface topography, (2) multiple groundwater-flow systems of varying spatial scales may exist, and (3) groundwater divides may move due to natural and anthropogenic variations in recharge and discharge (Winter et al. 2003). Few studies have examined the latter issue and the associated impacts on groundwater residence time and stream baseflow (Winter et al. 2003) and this was one impetus for the study described in this report. Furthermore, dynamic source-area boundaries in some hydrogeologic terrains may limit the applicability of the common assumptions for characterizing and conceptualizing hydrologic systems (e.g., Winter 2001).

It has long been recognized that source areas do not necessarily coincide with drainage divides (Tóth 1962, 1963). The reader is referred to Winter et al. (2003) and the references therein for a historical perspective on the understanding of source areas/volumes in small watersheds, as well as examples from relatively recent studies conducted in various hydrogeologic terrains. Recent modeling studies involving the delineation of surface-water source areas in unconfined sand aquifers have included work in the Coastal Plain of New Jersey (Modica et al. 1997; Modica et al. 1998), Cape Cod, Massachusetts (Walter et al. 2004), and the Allequash Creek Basin of Wisconsin (Pint et al. 2003). Cambareri and Eichner (1998) delineated a source area on Cape Cod, Massachusetts, on the basis of observed water-table topography.

In this report, a three-dimensional, steady-state, groundwater-flow model with particle tracking was used to delineate the source area of Phillips Branch, a baseflow-dominated stream in southeastern Delaware, USA. Modeling results show that, due to the low regional topographic relief and high aquifer permeability in this Coastal-Plain setting, the Phillips Branch source area is markedly different from its surface watershed in terms of both area and shape. Impacts of artificial discharge (well pumping) and recharge (wastewater disposal) on the source area and volume are also simulated as these aquifer stresses are common in the region in response to suburban developments replacing forested and agricultural land uses. The impacts of these practices on the water resources and the ecology of Delaware’s Inland Bays, an estuary of national significance (NRC 2000), are largely unknown. Simulations demonstrate that these anthropogenic stresses have profound impacts on the shape and extent of the source area and volume, groundwater-flow paths and residence times, and baseflow to Phillips Branch. This work clearly indicates the necessity of groundwater-flow modeling as a component of water-resource management in Coastal-Plain and similar hydrogeologic settings.

Background

Phillips Branch is located in eastern Sussex County, Delaware, within the Mid-Atlantic portion of the Coastal Plain Physiographic Province (Fig. 1). In this region, the Coastal Plain consists of a seaward-thickening and -dipping wedge of unconsolidated to semi-consolidated sediments overlying crystalline basement rocks that crop out in the Piedmont, north of the study area (Fig. 1b). The Phillips Branch watershed, delineated based on 1:24,000-scale US Geological Survey (USGS) topographic maps with 1.5-m contour intervals, has an area of 6.85 km2 and is part of the larger Inland Bays watershed (McKenna et al. 2007; Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-010-0641-x/MediaObjects/10040_2010_641_Fig1_HTML.gif
Fig. 1

Maps showing a the conterminous United States of America (USA) and the location of Delaware (DE), b the physiographic provinces of DE and the extent of the groundwater-flow model within the Inland Bays watershed, and c key features of the modeled area. Some low-elevation islands and marshes in the Inland Bays are not shown in c. m meters, msl mean sea level

The study site and most of the Inland Bays watershed are within the “well-drained upland” hydrogeomorphic region of the Delmarva Peninsula (Denver 1993; Shedlock et al. 1999). Streams in this region are well connected to the unconfined groundwater system and their flows are dominated by baseflow, which accounts for up to 80% of total stream discharge (Johnston 1976). Much of the time, these Coastal-Plain streams act as drains for the unconfined groundwater-flow system (Johnston 1976).

Topography in the Inland Bays watershed is generally flat (Fig. 1b). In the Phillips Branch watershed, land-surface elevations range from ∼1.5 m at the confluence with Unity Branch to ∼11 m at the westernmost watershed divide (Fig. 2). Phillips Branch is a first-order stream with the exception of one agricultural ditch tributary (Fig. 2). This ditch is generally dry and disconnected from the groundwater system (Denver 1989). The stream gradient in Phillips Branch averaged over the entire reach (∼4.5 km) is ∼0.001 based on 1:24,000-scale USGS topographic maps. Downstream from its confluence with Phillips Branch, the non-tidal Unity Branch traverses a short distance (∼0.6 km) before it discharges to Hopkins Prong, a brackish and tidal tributary of the Inland Bays. Collectively, these surface-water features are the main discharge boundaries for groundwater flow at the study site.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-010-0641-x/MediaObjects/10040_2010_641_Fig2_HTML.gif
Fig. 2

Map of the interior part of model area (with a higher-resolution grid; Fig. 1c) showing wells and stream-gaging station established for previous investigations. Phillips Branch surface watershed after McKenna et al. (2007)

Numerous hydrogeologic studies have been conducted in and around the Phillips Branch watershed, and the observations and results of some of these studies serve as the basis for the modeling efforts presented in this report (Andres 1991a, 1991b, 1995; Denver 1989, 1993; Denver and Sandstrom 1991; Dunkle et al. 1993; Kasper 2006; Parkhurst and Plummer 1993; Plummer et al. 1993; Shedlock et al. 1999, 1993). Unpublished data, including geophysical logs and hydraulic conductivity estimates, are on file at the Delaware Geological Survey (DGS; Newark, DE).

More than 50 monitoring wells were installed in the unconfined aquifer at the study site during previous investigations (Denver 1989, 1993; Fig. 2). These wells were constructed of 5-cm diameter polyvinyl chloride (PVC) with screen lengths ranging from 1 to 1.5 m. The well network included 12 clusters consisting of multiple adjacent wells of varying depths to assess vertical variations in groundwater chemistry and hydraulic head within the unconfined flow system. (Well details are tabulated in Kasper 2006.) Historical hydraulic head and testing data for these wells were used in this study to characterize and model the unconfined groundwater-flow system at the study site. These data are summarized in Denver (1989, 1993) and Kasper (2006).

Hydrogeology

The Pliocene-age Beaverdam Formation is the principle geologic unit containing the unconfined (Columbia) aquifer (Fig. 3), the primary focus of this study. The Beaverdam Formation is comprised predominantly of unconsolidated, fine-to-medium quartzose sand deposited in fluvial to estuarine environments (Ramsey 2003). Beaverdam sands are generally non-reactive and lack organic matter and other reduced components. At the study site, groundwater is predominantly oxic and contaminants normally subject to chemical reduction, such as nitrate, persist in the unconfined groundwater-flow system (Denver 1989, 1993; Shedlock et al. 1999).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-010-0641-x/MediaObjects/10040_2010_641_Fig3_HTML.gif
Fig. 3

Gamma-ray log for borehole Ph13-36 showing geologic and hydrogeologic units. Borehole location shown in Fig. 2. CB-1, -2, and -3 denote confining beds (see text). Geologic formations: Tbd Beaverdam, Tbt Bethany, Tcat Cat Hill (undivided), Tsm St. Marys. Dashed lines denote approximate formation boundaries. msl mean sea level

Over much of the model area, confining beds in the Bethany Formation serve as the base of the unconfined flow system (Andres and Klingbeil 2006). At Ph13-36 (Figs. 2 and 3), confining beds at the top and bottom of the Bethany Formation (CB-1 and -2, respectively, in Fig. 3) are separated by a sandy section (Pocomoke aquifer). In some areas, confining bed CB-1 is absent (due to erosion or facies change) and the Pocomoke aquifer likely functions as part of the unconfined flow system. There are also some places where both CB-1 and CB-2 are absent (i.e., the Bethany Formation is entirely sandy) and sands of the Manokin aquifer (Fig. 3) may function as part of the unconfined flow system. The St. Marys Formation (CB-3 in Fig. 3) is predominantly comprised of muds and functions as the base of the unconfined aquifer where shallower (younger) confining beds are absent (Andres and Klingbeil 2006).

Horizontal hydraulic conductivity (Kh) of the unconfined aquifer ranges from 0.34 to 150 m/day with a mean of 33.2 m/day based on unpublished slug-test data for 17 wells on file with the DGS (Kasper 2006). Analysis of slug-test data did not account for vertical anisotropy in hydraulic conductivity (A. Scott Andres, DGS, unpublished communication, 2005) and, therefore, Kh values may be underestimated (Zlotnik 1994). Johnston (1973) determined that the ratio between Kh and vertical hydraulic conductivity (Kv) for the unconfined aquifer is approximately 10:1. Analyses of an on-site aquifer test indicate a transmissivity (T) of 1,700 m2/day and a specific yield (Sy) of 0.16 (Denver 1989). This value of T translates to a Kh of 65 m/day assuming an aquifer thickness of 26 m. Aquifer-test wells used by Denver (1989) were screened near the base of the Beaverdam Formation where gravel beds are common (Ramsey 2003) and they are likely more transmissive than shallower sediments (Kasper 2006).

Groundwater flow is from west to east and generally parallels Phillips Branch (Denver 1989, 1993). Hydraulic head measurements taken on 19 July 1989 and 16 February 1989 are representative of average and extremely dry water-table conditions, respectively, based on the analysis of long-term hydrographs for study-site wells (Kasper 2006). Horizontal hydraulic gradients on 19 July 1989 ranged from ∼0.001 to 0.002, roughly equivalent to the Phillips Branch stream gradient. Vertical hydraulic gradients were 0.001 downward at cluster C1 near the western boundary of the watershed, and 0.003 upward at cluster C11 near the eastern boundary of the watershed (Fig. 2).

Precipitation averages 117 cm/year of which ∼33 cm/year is net groundwater recharge (Johnston 1976). Instantaneous discharge measurements at the Phillips Branch gaging station (Fig. 2) on 22 June 2004 (3,181 m3/day) and 30 August 1999 (563 m3/day) are taken as representative of average summer and extremely dry baseflow conditions, respectively. The use of non-synchronous baseflow and hydraulic head data for modeling purposes is justified on the basis of water-table observations at well Ng11-01, ∼22 km north/northwest of the study area (DGS 2009). Based on 488 observations over ∼45 years, the water table depth at Ng11-01 ranged from 2.11 to 4.46 m below land surface (bls) and averaged 3.46 m bls. During the periods when average hydraulic head (19 July 1989) and Phillips Branch summer baseflow (22 June 2004) were determined, the water table at Ng11-01 was 3.26 and 3.16 m bls, respectively, slightly shallower than the long-term average, but within or nearly within the interquartile range (3.17–3.77 m bls). During the extremely dry periods when hydraulic head (16 February 1989) and Phillips Branch baseflow (30 August 1999) were determined, the water table depth at Ng11-01 was 4.00 and 4.20 m bls, respectively, close to the 90th percentile water depth in this well (4.04 m bls).

Numerical methods

Steady-state groundwater flow was simulated using Visual MODFLOW (version 3.1.0; WHI 2003), which is based on the USGS modular three-dimensional finite-difference groundwater-flow model (MODFLOW-2000; Harbaugh et al. 2000). Visual MODFLOW implements MODPATH (Pollock 1989, 1994) and Zone Budget (Harbaugh 1990) for particle tracking and water-budget calculations, respectively.

The horizontal extent of the finite-difference grid is 15 × 18 km (225 km2) and consists of 181 rows and 326 columns (Fig. 1). The model extent is much larger than the primary area of interest near Phillips Branch (refined grid in Fig. 1c) so as to incorporate major regional hydrologic features and to minimize effects of prescribed boundary conditions. Cell sizes range from 300 × 300 m along the periphery of the modeled area to 30 × 30 m where the refined higher-resolution grid encompasses the Phillips Branch watershed (Figs. 1c and 2). Cell sizes grade horizontally from smallest to largest by a factor of no more than 1.5 (Anderson and Woessner 1992; WHI 2003). The grid is rotated 15° west of north to parallel Phillips Branch and regional groundwater flow at the study site.

The grid is discretized into five layers (Fig. 4) to resolve vertical flow in the simulations. The land surface (the top of layer 1 in Fig. 4) is based on a 30-m resolution digital elevation model (DEM) produced by the University of Delaware’s Spatial Analysis Lab (SpatLab 2005). The base of the model (the bottom of layer 5 in Fig. 4) is from a 30-m resolution grid of the base of the Columbia aquifer created using data from Andres and Klingbeil (2006). For the purpose of modeling, a continuous impermeable confining bed (no-flow boundary) at the base of the unconfined aquifer is assumed; however, the depth to the aquifer base is locally increased to account for the observed discontinuities in the confining beds of the Bethany Formation (see section Hydrogeology). The lateral extent of the finite-difference grid (Figs. 1 and 4) is also a no-flow boundary. The water table computed by Visual MODFLOW (Fig. 1c) is the upper variable-head boundary.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-010-0641-x/MediaObjects/10040_2010_641_Fig4_HTML.gif
Fig. 4

Model cross section A–A’ along row 100 of the finite-difference grid. Refer to Fig. 1c for cross-section location and orientation. msl mean sea level

Non-tidal streams are simulated using the Drain Package, an appropriate conceptualization where surface runoff is small relative to the groundwater contribution (Reilly 2001). This assumption is consistent with the observed regional hydrology (Johnston 1976). Drain reaches were digitized based on 1:24,000-scale digital line graph (DLG) data for the USGS Fairmount, Harbeson, Millsboro, and Frankford 7.5-min quadrangles (USGS 1998a, 1998b, 1998c, 1998d). Drain elevations were assigned where topographic contours cross streams and linearly interpolated between contour intervals to account for changes in stream gradient. Globally, a drain conductance per unit length (Cglobal) of 3 m/day was assigned initially; however, Phillips Branch drain conductance (Cphillips) was reduced during the calibration process to better simulate the observed discharge. Ponds were treated as constant-head boundaries (Figs. 1 and 2), the elevations of which were assigned based on topographic maps. Tidal surface-water bodies of the Inland Bays, which have a mean tidal range of less than 0.77 m (NOAA 2010), also were treated as constant-head boundaries with an elevation of 0 m. Recharge (R) to the uppermost model layer was 33 cm/year (Johnston 1976).

Horizontal isotropy of Kh was assumed as there are no data to suggest otherwise (Andres et al. 2003). The vertical anisotropy ratio (Kh:Kv) for all simulations was assumed to be 10:1 (Johnston 1973), a ratio used successfully in other local modeling studies (Andres et al. 2003; Johnston 1977). Effective porosity (ne), used in MODPATH to compute groundwater velocities, was assigned a value of 0.16 based on the Sy determined from the only available on-site aquifer test (Denver 1989). This value of Sy is within the range of values (0.14–0.17) reported by Johnston (1973) for the unconfined aquifer at three locations in Sussex County, DE.

The steady-state model was calibrated by varying Cphillips and global Kh to simulate average hydraulic heads (19 July 1989) and average Phillips Branch summer baseflow conditions (22 June 2004). Fixed and calibrated parameters for the groundwater-flow model are summarized in Table 1; simulated water-table contours based on model calibration are shown in Fig. 1c. The root-mean-squared error (RMSE) and normalized RMSE (NRMSE) between simulated and observed head were 0.13 m and 1.8%, respectively, below the a priori calibration criteria of 0.3 m and 10%, respectively. The absolute difference (AD) between simulated and observed baseflow (0.8%) also was below the a priori calibration criteria (10%). Potential error in ne was evaluated based on a comparison of flow- and chlorofluorocarbon-modeled groundwater ages for six well clusters (Dunkle et al. 1993). Over the range in ne for Sussex County, DE (0.14–0.17; Johnston 1973), the value of 0.16 resulted in the least error in groundwater age (RMSE 5.2 years, NRMSE 20.2%; Kasper 2006). The model was verified by reducing R to simulate the extremely dry period when hydraulic head (16 February 1989) and Phillips Branch baseflow (30 August 1999) were determined. An R of 16 cm/year resulted in a corresponding best fit for both hydraulic head (RMSE 0.11 m; NRMSE 2.1%) and baseflow (AD 8%).
Table 1

Fixed and calibrated parameters for the groundwater-flow model

Parameter

Units

Value

Kha

m/day

40

Kva

m/day

4

R

cm/year

33

Cphillipsa

m/day

2

Cglobal

m/day

3

ne

Unitless

0.16

aCalibrated value

Kh horizontal hydraulic conductivity, Kv vertical hydraulic conductivity, R recharge, Cphillips drain conductance for simulated Phillips Branch (see Fig. 1), Cglobal drain conductance for simulated streams (not including Phillips Branch), ne effective porosity

Sensitivity analyses involved independently increasing and decreasing R, Kh, and Kv relative to final model values (Table 1) and evaluating the resulting impacts to calibration statistics for average head and summer baseflow. Results of these analyses indicate the model is more sensitive to R than Kh (maintaining a 10:1 vertical anisotropy ratio). Specifically, variations of ±10% in Kh (36 and 44 m/day) maintain baseflow within the a priori calibration criteria; in contrast, corresponding ±10% variations in R (29.7 and 36.3 cm/year) result in baseflow errors exceeding the calibration criteria. The model is relatively insensitive to the vertical anisotropy ratio, suggesting the groundwater system is dominated by horizontal flow. The vertical anisotropy ratio of 10:1, however, provides the best overall calibration with respect to both head and baseflow, further supporting the use of this ratio for this effort and perhaps other local groundwater-modeling studies. Additional details regarding model development and testing may be found in Kasper (2006).

Both forward and backward particle tracking were used to delineate the Phillips Branch source area for ambient conditions (no artificial aquifer stresses), with particle pathlines determined from the calibrated steady-state flow model. Particles were initially placed in selected model cells adjacent to and below drain cells representing Phillips Branch; those particles were traced backward in time. Particles were then released from the uppermost model layer along the periphery of the initial source-area delineation and traced forward in time to refine the delineation. Groundwater residence time, the elapsed time between recharge to and discharge from the source volume (Modica et al. 1997), was evaluated by introducing 523 particles to the uppermost model layer on a 100-m grid spacing. MODPATH-generated endpoint data for these particles are the basis for residence-time estimates for ambient conditions. These data also were used as a proxy to indicate the nature of groundwater seepage along the stream channel for ambient conditions.

Artificial discharge from well pumping and artificial recharge from land-based wastewater disposal was simulated separately, but at the same central location, within the Phillips Branch source area for the direct comparison of results. The purpose of these simulations was to investigate the impacts artificial aquifer stresses have on the shape of the source area, stream baseflow, and the residence time of groundwater that results in baseflow. Artificial discharge was simulated using a single, fully-penetrating pumping well. Artificial recharge was simulated using a 300 × 300-m recharge boundary condition in layer 1 of the model. Evaluated artificial discharge and recharge scenarios ranged from 473 m3/day (125K gal/day) to 3,780 m3/day (1M gal/day; Tables 2 and 3). For artificial-discharge simulations, particles were released adjacent to the pumping well near the aquifer base and tracked backwards in time to delineate the origin of pumped water and the extent of well influence (i.e., “capture zone”). For artificial-recharge simulations, a total of 100 particles (one per model cell representing land-based wastewater disposal) was released in the uppermost model layer and traced forward in time to delineate the extent of impacted groundwater. For all simulations, particles were released within and below the drain representing Phillips Branch to delineate the stream’s source area and determine the residence time of groundwater that results in stream baseflow. Zone Budget was used in each simulation to track baseflow gain along the entire stream reach, both up and downstream from the gaging station (Fig. 2).
Table 2

Artificial-discharge scenarios and their impacts on groundwater residence time in the Phillips Branch source volume and stream baseflow. P percentile, ∆ change

Artificial-discharge scenario (m3/day)

Fig. 6a–e

Residence time statistics for the Phillips Branch source volume (years)

Phillips Branch baseflow

Min

P 5

P 25

P 50

P 75

P 95

Max

(m3/day)

(% ∆)

0

a

0.1

0.1

3.4

10.4

23.5

48.7

99.1

4,725

0

473

b

0.1

0.1

3.4

10.3

23.4

49

97.9

4,623

–2.2

945

c

0.1

0.1

3.7

11.2

23.6

48.4

101.2

4,793

+1.5

1,890

d

0.1

0.1

3.2

10.0

22.9

49.5

100.1

4,328

–8.4

3,780

e

0.1

0.1

2.8

9.2

21.6

49.5

122.1

3,956

–16.3

Table 3

Artificial-recharge scenarios and their impacts on groundwater residence time in the Phillips Branch source volume and stream baseflow. P percentile, ∆ change

Artificial-recharge scenario (m3/day)

Fig. 6f–j

Residence time statistics for the Phillips Branch source volume (years)

Phillips Branch baseflow

Min

P 5

P 25

P 50

P 75

P 95

Max

(m3/day)

(% ∆)

0

f

0.1

0.1

3.4

10.4

23.5

48.7

99.1

4,725

0

473

g

0.1j

0.1

3.5

9.8

20.9

45.7

100.5

4,810

+1.8

945

h

0.1

0.1

3.5

9.3

19.4

40.5

102.8

4,929

+4.3

1,890

i

0.1

0.1

3.8

8.8

19.1

32.1

82.7

5,165

+9.3

3,780

j

0.1

0.1

3.5

8.0

18.2

28.2

37.4

5,635

+19.2

Results and discussion

Ambient source area

The modeled source area (Fig. 5) is markedly different in appearance and size than the topographically defined surface watershed delineated by McKenna et al. (2007). The source area is shaped like an elongated ellipse tapered in both upgradient and downgradient directions. The source area (5.51 km2) is ∼20% smaller than the surface watershed (6.85 km2) and ∼22% of the source area (1.20 km2) is outside of the surface watershed. Moreover, a substantial portion of the surface watershed (2.54 km2 or 37%) is outside of the source area and, therefore, a large volume of recharge to the surface watershed (2,296 m3/day) discharges to receiving waters other than Phillips Branch.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-010-0641-x/MediaObjects/10040_2010_641_Fig5_HTML.gif
Fig. 5

Contrast between the Phillips Branch surface watershed and its modeled source area. Groundwater residence time contours based on particle-tracking analyses also shown

Most of the surface watershed extending beyond the Phillips Branch source area (1.92 km2 of 2.54 km2) is south of the source area (Fig. 5). Particle-tracking results suggest that recharge (∼1,555 m3/day ) entering most of this area (1.72 km2) discharges to Hopkins Prong or to the section of Unity Branch immediately upstream from Hopkins Prong. The remaining recharge to this area (∼181 m3/day) discharges south to Guinea Creek and tributaries of Indian River Bay (Fig. 1). The recharge entering the surface watershed north of the source area (∼560 m3/day from 0.62 km2 of watershed) discharges to Unity Branch.

Groundwater residence time in the source volume ranges from ∼4 day to almost 100 years. Residence time, however, is highly skewed with most of the particles (∼95%) discharging to Phillips Branch within 50 years (Table 2). The arithmetic mean residence time was 15 years. These results agree in general with geochemical age-dating work of Dunkle et al. (1993) and Andres (1995) for wells at this site and with Böhlke and Krantz (2003), who report that groundwater beneath Indian River Bay (Fig. 1) was recharged within the last 50 years. The results also are consistent with the findings of Robinson and Reay (2002), who used a flow model to evaluate residence time in a comparable-sized watershed (6.2 km2) in the Atlantic Coastal Plain of Virginia. Their results indicate an average residence time of 21 years with 90% of the residence times being less than 50 years.

The oldest groundwater enters the flow system near the upgradient terminus of the source area and travels along deep and long flow paths; younger groundwater is recharged proximal the Phillips Branch stream channel and travels along shallow and short flow paths (Fig. 5). The stream begins to receive baseflow near well S2 (Fig. 2) even though the drain boundary condition extends farther west. Groundwater discharging to this portion of the stream is young (≤10 years old) and relatively homogeneous in age, but generally becomes both older and more diverse in age with distance downstream. This trend indicates the diversity of groundwater provenance and transport pathways contributing to stream baseflow (Modica et al. 1997, 1998). Mixing of diverse groundwaters along the stream channel may help to explain the sharp contrast between nitrate in baseflow (<5 mg/L) and nitrate in some portions of the unconfined aquifer (up to ∼40 mg/L) reported by Denver (1989). Particle-tracking data indicate that the nature of groundwater seepage along the stream channel is “parabolic” (Modica et al. 1997) with groundwater discharge focused to the mid-reaches of Phillips Branch.

Source area with artificial discharge

In Delaware, thousands of water-supply wells tap the unconfined aquifer to satisfy potable, irrigation, and industrial water needs (Andres and Klingbeil 2006). More than half (∼57%) of the State’s groundwater withdrawals are from the unconfined aquifer (Wheeler 2003). In Sussex County (Fig. 1), groundwater accounts for ∼60% of freshwater use, most of which is used for public-supply and irrigation purposes (Wheeler 2003). Rapid development in Sussex and other Delaware counties will likely result in increasing use of groundwater resources.

Well pumping has profound impacts on the shape of the source area (Figs. 6a–e). The most noticeable impacts occur in the western, upgradient portion of the source area, with the eastern, downgradient portion relatively unaffected by pumping for each scenario. Even at the lowest pumping rate (473 m3/day; Fig. 6b), flow from a substantial portion of the ambient source area is captured by the well, forcing the stream’s source area to expand outward in the vicinity of the pumping center. As pumping increases, expansion of the upgradient portion of the source area continues. Under maximum pumping conditions (3,780 m3/day; Fig. 6e), approximately one third of the ambient stream source area no longer contributes water to Phillips Branch. Maximum pumping also causes some particles that ultimately discharge to the stream to move along longer, more tortuous flow paths (Fig. 6e).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-010-0641-x/MediaObjects/10040_2010_641_Fig6_HTML.gif
Fig. 6

Effects of ae artificial discharge and fj artificial recharge on the Phillips Branch source area. Recharge/discharge scenarios (from top to bottom): 0 m3/day (a and f), 473 m3/day (b and g), 945 m3/day (c and h), 1,890 m3/day (d and i), and 3,780 m3/day (e and j)

Most, but not all, well-pumping scenarios cause reductions in Phillips Branch baseflow (Table 2). For example, pumping at 473 m3/day causes a small (–2.2%) reduction in baseflow, while pumping at 945 m3/day actually causes a small (+1.5%) increase in baseflow due to the areal expansion of the source area. The increase in median residence time (11.2 years) relative to the other artificial discharge scenarios supports this observation (Table 2). Additional pumping leads to additional decreases in baseflow.

Pumping has little effect on the residence time of groundwater that discharges to Phillips Branch (Table 2). Under the maximum pumping scenario (3,780 m3/day), however, the maximum residence time increases by 20+ years (Table 2), suggesting that some particles are forced to travel along deeper, longer, and (or) more tortuous flow paths. These results suggest that, when dealing with soluble contaminants such as nitrate, which is largely conservative in the oxic groundwaters in the region (Shedlock et al. 1999), the discharge of contaminated groundwater to the stream may be delayed due to well pumping.

Source area with artificial recharge

Land-based wastewater disposal via spray irrigation or infiltration basins is common in Delaware, and increasing due to the elimination of surface-water discharge permits to meet total maximum daily load (TMDL) requirements for nutrients in the Inland Bays watershed (DNREC 1998). Although municipal wastewater systems exist, community-based systems are becoming common in non-sewered areas in the watershed.

As with artificial discharge, the shape of the Phillips Branch source area is highly sensitive to artificial recharge (Figs. 6f–j). For the reference simulation (Fig. 6f), water naturally recharged in the disposal area discharges to the upper one third of the stream and some particles travel below the stream channel to discharge farther downstream. As artificial recharge increases (473 and 945 m3/day; Figs. 6g and h, respectively; Table 3), water from the artificial-recharge area moves farther downstream and discharges along larger fractions of the stream channel. In addition, the fraction of this water that follows pathlines beneath the stream channel increases. Groundwater mounding from artificial recharge also causes the Phillips Branch source area to become more tapered upgradient from the recharge area, while the ambient shape of the source area is retained downgradient from the recharge area (Figs. 6g and h). At the highest artificial recharge rates (1,890 and 3,780 m3/day; Figs. 6i and j, respectively; Table 3), mounding beneath the recharge area increases and forces discharge along most or all of the Phillips Branch stream channel. Discharge also is forced to Unity Branch, both above and below the confluence with Phillips Branch, and directly to the tidal waters of Hopkins Prong. The upgradient portion of the source area tapers further under the 1,890 m3/day scenario (Fig. 6i) and is altogether eliminated under the 3,780 m3/day scenario where the source area takes on a tear-drop appearance (Fig. 6j).

Relative to the artificial-discharge simulations, artificial recharge has contrasting impacts on Phillips Branch baseflow (Table 3). Artificial recharge for the reference simulation represents only a small fraction (∼2%) of Phillips Branch baseflow (Table 4). As artificial recharge increases (473 and 945 m3/day; Figs. 6g and h, respectively; Tables 3 and 4), Phillips Branch baseflow increases along with the fractional contribution of artificial recharge to Phillips Branch baseflow. Although all of the artificial recharge discharges to Phillips Branch under these intermediate recharge scenarios, the increases in total baseflow are substantially less than the respective artificial-recharge rates (Table 3). This result is due to the tapering of the Phillips Branch source area upgradient from the artificial recharge area due to local mounding and the alteration of the source-area boundaries of adjacent streams (Figs. 6g and h). In other words, artificial recharge displaces groundwater that would otherwise discharge to Phillips Branch under ambient conditions. At the highest artificial-recharge rates (1,890 and 3,780 m3/day; Tables 3 and 4), the fraction of artificial recharge reaching Phillips Branch decreases due to discharge to other surface-water bodies (Figs. 6i and j, respectively). The fraction of artificial recharge in Phillips Branch baseflow, however, continues along an increasing trend and, under the 3,780 m3/day scenario, constitutes nearly half (∼42%) of Phillips Branch baseflow (Table 4).
Table 4

Fractions of artificial recharge reaching Phillips Branch and corresponding contributions to stream baseflow

Artificial-recharge scenario (m3/day)

Fig. 6f–j

Phillips Branch baseflow (m3/day)

Fraction of artificial recharge reaching Phillips Branch (%)

Artificial recharge in Phillips Branch baseflow (m3/day)

Fraction of artificial recharge in Phillips Branch baseflow (%)

0

f

4,725

100

0

1.7a

473

g

4,810

100

473

9.8

945

h

4,929

100

945

19.2

1,890

i

5,165

87

1,644

31.8

3,780

j

5,635

63

2,381

42.3

aReflects only natural recharge to the area where artificial recharge occurs

The residence times for particles discharging to Phillips Branch are more sensitive to artificial recharge than artificial discharge (Tables 2 and 3). Median residence time declines slightly due to artificial recharge, whereas the 95th percentile residence time decreases more drastically. Under the 3,780 m3/day scenario, 95% of the source volume flushes within ∼28 years or ∼60% of the corresponding flushing time for the reference simulation (Table 3). This considerable decrease in flushing time is due to the elimination of flow of older groundwater from upgradient regions of the reference source area (Figs. 5 and 6j), as well as the increased hydraulic gradient imposed by the artificial-recharge area.

Residence times for artificial-recharge particles to reach Phillips Branch (Table 5) indicate the time required for soluble and conservative contaminants in such recharge to reach the stream and for steady-state concentrations to develop. For the reference simulation it takes 7 years for contaminants to reach the stream and 16.5 years for the development of steady-state concentrations (assuming only advective transport and a constant contaminant concentration in artificial recharge). As the rate of artificial recharge increases, contaminants reach the stream in less time, but median and maximum residence times increase due to the movement of particles along deeper, longer flow paths. Comparison of maximum residence times for the reference and 3,780 m3/day scenarios, for example, indicates that mounding and flow-path alteration from artificial recharge cause more than a two-fold increase in the time required for steady-state concentrations to develop along the stream channel (Table 5).
Table 5

Residence time statistics for particles released under the artificial recharge area. P percentile

Artificial-recharge scenario (m3/day)

Fig. 6f–j

Residence time statistics for artificial-recharge water (years)

Min

P 25

P 50

P 75

Max

0

f

7.0

8.7

9.7

11.0

16.5

473

g

6.3

8.8

11.3

14.9

20.8

945

h

5.8

9.5

14.0

21.0

26.6

1,890

i

4.8

8.8

17.8

22.3

34.1

3,780

j

3.7

8.0

16.7

20.9

36.3

Summary and conclusions

Based on the Phillips Branch model, the Mid-Atlantic Coastal Plain is a region where simple topographic watershed models that assume coincident surface water and groundwater boundaries may not be sufficient to characterize groundwater flow and discharge. Simulations of groundwater/surface-water interactions using a three-dimensional, steady-state, groundwater-flow model with particle tracking show that stream source areas within this region can be markedly different from their watershed counterparts and that the size and shape of source areas can vary with changes in artificial (and potentially natural) recharge and discharge. Artificial recharge and discharge can also modify groundwater travel time in source volumes and, therefore, the rates at which soluble contaminants may be flushed through aquifer systems.

This work has extensive application to water-resource management in hydrogeologic settings involving sand aquifers incised by baseflow-dominated streams. The findings are particularly applicable in the coastal watersheds of southern Delaware and elsewhere in the Mid-Atlantic Coastal Plain, where intensive agricultural land use has lead to high levels of nitrogenous contamination in groundwater and the symptoms of eutrophication in local estuarine receiving waters. Although this work was focused on a single subwatershed of the larger Inland Bays watershed, the same discrepancies between topographically defined watersheds and stream/estuarine source areas may also occur at the large watershed scale. Detailed studies and field verification of regional and local water-table topography and groundwater flow paths are advisable before watershed management practices are put into place to mitigate the effects of excess nutrient loads on receiving waters.

Acknowledgements

The authors acknowledge the Delaware Department of Natural Resources and Environmental Control (DNREC), Delaware Geological Survey (DGS), US Geological Survey (USGS), and Department of Geological Sciences at the University of Delaware for providing data and financial and in kind support to J.W. Kasper, while a graduate student. A.S. Andres (DGS), S.W. Ator (USGS), and S.C. Cooper (USGS) are gratefully acknowledged for reviewing the draft manuscript and providing valuable suggestions for its improvement. Two anonymous reviewers provided constructive comments that greatly improved the final manuscript.

Copyright information

© Springer-Verlag 2010