• Anderson, M. P., , and W. W. Woessner, 1992: Applied Groundwater Modeling: Simulation of Flow and Advective Transport. Academic Press, 381 pp.

  • Belden, A. C., , and S. G. Pallardy, 2009: Successional trends and apparent Acer saccharum regeneration failure in an oak-hickory forest in central Missouri, USA. Plant Ecol., 204, 305322, doi:10.1007/s11258-009-9593-4.

    • Search Google Scholar
    • Export Citation
  • Bentrup, G., 2008: Conservation buffers: Design guidelines for buffers, corridors, and greenways. U.S. Department of Agriculture Forest Service Southern Research Station General Tech. Rep. SRS-109, 136 pp.

  • Burt, T., , G. Pinay, , and S. Sabater, 2010: What do we still need to know about the ecohydrology of riparian zones? Ecohydrology, 3, 373377, doi:10.1002/eco.140.

    • Search Google Scholar
    • Export Citation
  • Chinnasamy, P., , and J. A. Hubbart, 2014: Measuring and modeling shallow groundwater flow between a semi-karst border stream and Ozark forested riparian zone in the central USA. J. Sci. Res. Rep., 3, 844865, doi:10.9734/JSRR/2014/7711.

    • Search Google Scholar
    • Export Citation
  • Critchfield, H. J., 1966: General Climatology. Prentice-Hall, 420 pp.

  • Dottori, F., , M. L. V. Martina, , and E. Todini, 2009: A dynamic rating curve approach to indirect discharge measurement. Hydrol. Earth Syst. Sci., 13, 847863, doi:10.5194/hess-13-847-2009.

    • Search Google Scholar
    • Export Citation
  • Ensign, S. H., , and M. W. Doyle, 2006: Nutrient spiraling in streams and river networks. J. Geophys. Res.,111, G04009, doi:10.1029/2005JG000114.

  • Fellows, C. S., , H. M. Valett, , and C. N. Dahm, 2001: Whole-stream metabolism in two montane streams: Contribution of the hyporheic zone. Limnol. Oceanogr.,46, 523–531, doi:10.4319/lo.2001.46.3.0523.

  • Fetter, C. W., 2001: Applied Hydrology. Prentice Hall, 598 pp.

  • Freeze, R. A., , and J. A. Cherry, 1979: Groundwater. Prentice Hall, 604 pp.

  • Gooseff, M. N., , S. M. Wondzell, , R. Haggerty, , and J. Anderson, 2003: Comparing transient storage modeling and residence time distribution (RTD) analysis in geomorphically varied reaches in the Lookout Creek basin, Oregon, USA. Adv. Water Resour., 26, 925937, doi:10.1016/S0309-1708(03)00105-2.

    • Search Google Scholar
    • Export Citation
  • Harbaugh, A. W., , and M. G. McDonald, 1996: User's documentation for MODFLOW-96, an update to the US Geological Survey modular finite-difference ground-water flow model. U.S. Geological Survey Open-File Rep. 96-485, 63 pp.

  • Harbaugh, A. W., , E. R. Banta, , M. C. Hill, , and M. G. McDonald, 2000: MODFLOW-2000, the U.S. Geological Survey modular ground-water model—User guide to modularization concepts and the ground-water flow process. U.S. Geological Survey Open-File Rep. 00-92, 127 pp.

  • Harrington, G. A., , G. R. Walker, , A. J. Love, , and K. A. Narayan, 1999: A compartmental mixing-cell approach for the quantitative assessment of groundwater dynamics in the Otway basin, south Australia. J. Hydrol., 214, 4963, doi:10.1016/S0022-1694(98)00243-1.

    • Search Google Scholar
    • Export Citation
  • Harvey, J. W., , and B. J. Wagner, 2000: Quantifying hydrologic interactions between streams and their subsurface hyporheic zones. Streams and Ground Waters, J. B. Jones and P. J. Mulholland, Eds., Academic Press, 3–44.

  • Haycock, N., , and G. Pinay, 1993: Groundwater nitrate dynamics in grass and poplar vegetated riparian buffer strips during the winter. J. Environ. Qual., 22, 273278, doi:10.2134/jeq1993.00472425002200020007x.

    • Search Google Scholar
    • Export Citation
  • Hefting, M., , B. Beltman, , D. Karssenberg, , K. Rebel, , M. van Riessen, , and M. Spijker, 2006: Water quality dynamics and hydrology in nitrate loaded riparian zones in the Netherlands. Environ. Pollut., 139, 143156, doi:10.1016/j.envpol.2005.04.023.

    • Search Google Scholar
    • Export Citation
  • Hill, A. R., 1996: Nitrate removal in stream riparian zones. J. Environ. Qual.,25, 743–755, doi:10.2134/jeq1996.00472425002500040014x.

  • Hynes, H. B. N., 1983: Groundwater and stream ecology. Hydrobiologia, 100, 9399, doi:10.1007/BF00027424.

  • Jacobs, T., , and J. Gilliam, 1985: Riparian losses of nitrate from agricultural drainage waters. J. Environ. Qual., 14, 472478, doi:10.2134/jeq1985.00472425001400040004x.

    • Search Google Scholar
    • Export Citation
  • Kasahara, T., , and S. M. Wondzell, 2003: Geomorphic controls on hyporheic exchange flow in mountain streams. Water Resour. Res., 39, 1005, doi:10.1029/2002WR001386.

    • Search Google Scholar
    • Export Citation
  • Lautz, L. K., , and D. I. Siegel, 2006: Modeling surface and ground water mixing in the hyporheic zone using MODFLOW and MT3D. Adv. Water Resour., 29, 16181633, doi:10.1016/j.advwatres.2005.12.003.

    • Search Google Scholar
    • Export Citation
  • Levia, D. F., , D. E. Carlyle-Moses, , and T. Tanaka, 2011: Reflections on the state of forest hydrology and biogeochemistry. Forest Hydrology and Biogeochemistry: Synthesis of Past Research and Future Directions, D. F. Levia et al., Eds., Springer, 729–733.

  • Luo, Y., , and M. Sophocleous, 2010: Seasonal groundwater contribution to crop-water use assessed with lysimeter observations and model simulations. J. Hydrol., 389, 325335, doi:10.1016/j.jhydrol.2010.06.011.

    • Search Google Scholar
    • Export Citation
  • Marzolf, E. R., , P. J. Mulholland, , and A. D. Steinman, 1994: Improvements to the diurnal upstream–downstream dissolved oxygen change technique for determining whole-stream metabolism in small streams. Can. J. Fish. Aquat. Sci.,51, 1591–1599, doi:10.1139/f94-158.

  • McDonald, M. G., , and A. W. Harbaugh, 1984: A modular three-dimensional finite-difference ground-water flow model. U.S. Geological Survey Open-File Rep. 83-875, 539 pp.

  • McDonald, M. G., , and A. W. Harbaugh, 1988: A modular three-dimensional finite-difference ground-water flow model. Techniques of Water-Resources Investigations of the United States Geological Survey A1, 588 pp.

  • Moriasi, D. N., , J. G. Arnold, , M. W. Liew, , R. L. Bingner, , R. D. Harmel, , and T. L. Veith, 2007: Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans. ASABE, 50, 885900, doi:10.13031/2013.23153.

    • Search Google Scholar
    • Export Citation
  • MSDIS, cited 2012: Missouri spatial data information service. [Available online at http://msdis.missouri.edu/data.]

  • Mulholland, P. J., , E. R. Marzolf, , J. R. Webster, , D. R. Hart, , and S. P. Hendricks, 1997: Evidence that hyporheic zones increase heterotrophic metabolism and phosphorus uptake in forest streams. Limnol. Oceanogr.,42, 443–451.

  • Nash, J. E., , and J. Sutcliffe, 1970: River flow forecasting through conceptual models part I—A discussion of principles. J. Hydrol., 10, 282290, doi:10.1016/0022-1694(70)90255-6.

    • Search Google Scholar
    • Export Citation
  • Pallardy, S., , T. Nigh, , and H. Garrett, 1988: Changes in forest composition in central Missouri: 1968-1982. Amer. Midl. Nat., 120, 380390, doi:10.2307/2426010.

    • Search Google Scholar
    • Export Citation
  • Peterjohn, W. T., , and D. L. Correll, 1984: Nutrient dynamics in an agricultural watershed: observations on the role of a riparian forest. Ecology, 65, 14661475, doi:10.2307/1939127.

    • Search Google Scholar
    • Export Citation
  • Pollock, D. W., 1994: User's Guide for MODPATH/MODPATH-PLOT, version 3: A particle tracking post-processing package for MODFLOW, the U.S. Geological Survey finite-difference ground-water flow model. U.S. Geological Survey Open-File Rep. 94-464, 464 pp.

  • Reed, S. E., 2010: Ambrosia beetle habitat use, Host use, and influence on early wood colonizing microbes in an oak-hickory forest. Ph.D. thesis, University of Missouri, 531 pp.

  • Rochow, J. J., 1972: A vegetational description of a mid-Missouri forest using gradient analysis techniques. Amer. Midl. Nat., 87,377396, doi:10.2307/2423570.

    • Search Google Scholar
    • Export Citation
  • Schaap, M. G., , F. J. Leij, , and M. T. van Genuchten, 1998: Neural network analysis for hierarchical prediction of soil hydraulic properties. Soil Sci. Soc. Amer. J.,62, 847–855, doi:10.2136/sssaj1998.03615995006200040001x.

  • Schaap, M. G., , F. J. Leij, , and M. T. van Genuchten, 2001: ROSETTA: A computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions. J. Hydrol.,251, 163–176, doi:10.1016/S0022-1694(01)00466-8.

  • Schilling, K., , Z. Li, , and Y.-K. Zhang, 2006: Groundwater–surface water interaction in the riparian zone of an incised channel, Walnut Creek, Iowa. J. Hydrol., 327, 140150, doi:10.1016/j.jhydrol.2005.11.014.

    • Search Google Scholar
    • Export Citation
  • Sophocleous, M., 2002: Interactions between groundwater and surface water: The state of the science. J. Hydrol., 10, 5267, doi:10.1007/s10040-001-0170-8.

    • Search Google Scholar
    • Export Citation
  • Storey, R. G., , K. W. F. Howard, , and D. D. Williams, 2003: Factors controlling riffle-scale hyporheic exchange flows and their seasonal changes in a gaining stream: A three-dimensional groundwater flow model. Water Resour. Res., 39, 1034, doi:10.1029/2002WR001367.

    • Search Google Scholar
    • Export Citation
  • Strahler, A., 1952: Hypsometric (area-altitude) analysis of erosional topography. Bull. Geol. Soc. Amer., 63, 11171142, doi:10.1130/0016-7606(1952)63[1117:HAAOET]2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • USDA, cited 2012: Web soil survey. [Available online at http://www.nrcs.usda.gov/wps/portal/nrcs/surveylist/soils/survey/state/?stateId=MO.]

  • Willmott, C. J., 1981: On the validation of models. Phys. Geogr., 2, 184194.

  • Woessner, W. W., 2000: Stream and fluvial plain ground water interactions: Rescaling hydrogeologic thought. Groundwater, 38, 423429, doi:10.1111/j.1745-6584.2000.tb00228.x.

    • Search Google Scholar
    • Export Citation
  • Wondzell, S. M., , and F. J. Swanson, 1996: Seasonal and storm dynamics of the hyporheic zone of a 4th-order mountain stream. II: Nitrogen cycling. J. N. Amer. Benthol. Soc., 15,2034, doi:10.2307/1467430.

    • Search Google Scholar
    • Export Citation
  • Wroblicky, G. J., , M. E. Campana, , H. M. Valett, , and C. N. Dahm, 1998: Seasonal variation in surface-subsurface water exchange and lateral hyporheic area of two stream-aquifer systems. Water Resour. Res., 34, 317328, doi:10.1029/97WR03285.

    • Search Google Scholar
    • Export Citation
  • Young, F. J., , C. A. Radatz, , and C. A. Marshall, 2003: Soil survey of Boone County. Missouri Natural Resources Conservation Service Rep., 318 pp.

  • Zheng, C., , and P. P. Wang, 1999: MT3DMS: A modular three-dimensional multispecies transport model for simulation of advection, dispersion, and chemical reactions of contaminants in groundwater systems; Documentation and user's guide. U.S. Army Engineer Research and Development Rep. SERDP-99-1, 169 pp.

  • View in gallery

    Study sites and instrument locations at Baskett Wildlife Research and Education Center, central Missouri, where S is the stilling well site and PZ is the piezometer site.

  • View in gallery

    Conceptual diagram showing cross section of piezometer study design at Baskett Wildlife Research and Education Center, central Missouri. Soil profile depths are watershed averages (USDA 2012).

  • View in gallery

    Measured precipitation (mm) and hydraulic head at stream (m) and at piezometer (m) locations during WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

  • View in gallery

    Observed vs modeled (MODFLOW) hydraulic heads for piezometers located at PZI (Pz1–Pz6) and at PZII (Pz7–Pz12) over WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

  • View in gallery

    Plan view of MODFLOW estimates of hydraulic head distribution (m) for the months of November, February, May, and August over the WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

  • View in gallery

    MT3DMS estimates of nitrate loading in the shallow aquifer from different segments of the study reach for the WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

  • View in gallery

    MODPATH estimates of lateral extent of surface water–groundwater interactions for the months of November, February, May, and August over the WY 2011 at Baskett Wildlife Research and Education Center, central Missouri. The solid lines indicate flow paths as simulated by MODPATH.

  • View in gallery

    MODPATH estimates of lateral extent (m = meter) of surface water–groundwater interactions from the stream bank and travel time (d = day) of water in flow paths over the WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 134 134 19
PDF Downloads 92 92 17

Potential of MODFLOW to Model Hydrological Interactions in a Semikarst Floodplain of the Ozark Border Forest in the Central United States

View More View Less
  • 1 Department of Forestry, University of Missouri, Columbia, Missouri, and International Water Management Institute, Kathmandu, Nepal
  • | 2 Department of Forestry and Department of Soils, Environmental and Atmospheric Sciences, University of Missouri, Columbia, Missouri
© Get Permissions
Full access

Abstract

Riparian shallow groundwater and nutrient movement is important for aquatic and forest ecosystem health. Understanding stream water (SW)–shallow groundwater (GW) interactions is necessary for proper management of floodplain biodiversity, but it is particularly confounding in underrepresented semikarst hydrogeological systems. The Modular Three-Dimensional Finite-Difference Ground-Water Flow Model (MODFLOW) was used to simulate shallow groundwater flow and nutrient transport processes in a second-growth Ozark border forest for the 2011 water year. MODFLOW provided approximations of hydrologic head that were statistically comparable to observed data (Nash–Sutcliffe = 0.47, r2 = 0.77, root-mean-square error = 0.61 cm, and mean difference = 0.46 cm). Average annual flow estimates indicated that 82% of the reach length was a losing stream, while the remaining 18% was gaining. The reach lost more water to the GW during summer (2405 m3 day−1) relative to fall (2184 m3 day−1), spring (2102 m3 day−1), and winter (1549 m3 day−1) seasons. Model results showed that the shallow aquifer had the highest nitrate loading during the winter season (707 kg day−1). A Particle-Tracking Model for MODFLOW (MODPATH) revealed significant spatial variations between piezometer sites (p = 0.089) in subsurface flow path and travel time, ranging from 213 m and 3.6 yr to 197 m and 11.6 yr. The current study approach is novel with regard to the use of transient flow conditions (as opposed to steady state conditions) in underrepresented semikarst geological systems of the U.S. Midwest. This study emphasizes the significance of semikarst geology in regulating SW–GW hydrologic and nutrient interactions and provides baseline information and modeling predictions that will facilitate future studies and management plans.

Corresponding author address: Pennan Chinnasamy, Department of Forestry, 203-T ABNR Building, University of Missouri, Columbia, MO 65211. E-mail address: pcppf@mail.missouri.edu; hubbartj@missouri.edu

Abstract

Riparian shallow groundwater and nutrient movement is important for aquatic and forest ecosystem health. Understanding stream water (SW)–shallow groundwater (GW) interactions is necessary for proper management of floodplain biodiversity, but it is particularly confounding in underrepresented semikarst hydrogeological systems. The Modular Three-Dimensional Finite-Difference Ground-Water Flow Model (MODFLOW) was used to simulate shallow groundwater flow and nutrient transport processes in a second-growth Ozark border forest for the 2011 water year. MODFLOW provided approximations of hydrologic head that were statistically comparable to observed data (Nash–Sutcliffe = 0.47, r2 = 0.77, root-mean-square error = 0.61 cm, and mean difference = 0.46 cm). Average annual flow estimates indicated that 82% of the reach length was a losing stream, while the remaining 18% was gaining. The reach lost more water to the GW during summer (2405 m3 day−1) relative to fall (2184 m3 day−1), spring (2102 m3 day−1), and winter (1549 m3 day−1) seasons. Model results showed that the shallow aquifer had the highest nitrate loading during the winter season (707 kg day−1). A Particle-Tracking Model for MODFLOW (MODPATH) revealed significant spatial variations between piezometer sites (p = 0.089) in subsurface flow path and travel time, ranging from 213 m and 3.6 yr to 197 m and 11.6 yr. The current study approach is novel with regard to the use of transient flow conditions (as opposed to steady state conditions) in underrepresented semikarst geological systems of the U.S. Midwest. This study emphasizes the significance of semikarst geology in regulating SW–GW hydrologic and nutrient interactions and provides baseline information and modeling predictions that will facilitate future studies and management plans.

Corresponding author address: Pennan Chinnasamy, Department of Forestry, 203-T ABNR Building, University of Missouri, Columbia, MO 65211. E-mail address: pcppf@mail.missouri.edu; hubbartj@missouri.edu

1. Introduction

Surface water and groundwater are not isolated components of the hydrologic regime but rather integrate with all other watershed hydrologic processes (Hynes 1983; Sophocleous 2002; Burt et al. 2010; Harvey and Wagner 2000), and yet there remains a paucity of observed surface water (SW)–shallow groundwater (GW) information from many regions. Sophocleous (2002) and Woessner (2000) showed that effective management of water resources requires improved understanding of SW–GW interactions. Multiple physical processes influence SW–GW interactions. The dominant physical process for a given region depends on climate, geology, and topography and hydraulic gradients (Fetter 2001; Harvey and Wagner 2000; Levia et al. 2011; Burt et al. 2010). Therefore, observed data that characterize the lateral extent, volume, and the residence time of SW–GW hydrologic fluxes will improve process understanding and management and provide a basis for modeling (Kasahara and Wondzell 2003).

Accurate spatial and temporal representations of SW–GW interactions are critical for understanding stream nutrient loading processes. Considerable research has shown that riparian zone shallow groundwater often has a lower nitrate concentration relative to that of surface water (Peterjohn and Correll 1984; Jacobs and Gilliam 1985; Haycock and Pinay 1993; Hill 1996). Increased denitrification rates in GW, relative to SW, were shown to be the major reason for low riparian zone GW nitrate concentration levels (Jacobs and Gilliam 1985). Such increases in the denitrification rate are primarily due to biotic and abiotic factors that raise soil moisture content and thus water storage in the riparian soil. Since SW–GW hydrologic connectivity greatly influences the groundwater table and soil water storage, it is useful from a management perspective to quantify spatial and temporal variations in net water flow from the stream to riparian zone GW to improve quantitative understanding and to predict (i.e., model) SW–GW flow and nutrient regimes.

To improve mechanistic understanding of field-based observations, researchers are increasingly using numerical simulation models [e.g., Modular Three-Dimensional Finite-Difference Ground-Water Flow Model (MODFLOW), CPFLOW, Saturated–Unsaturated Transport Model (SUTRA), HYDRUS] to better understand SW–GW interactions (Wroblicky et al. 1998; Woessner 2000; Storey et al. 2003; Gooseff et al. 2003; Kasahara and Wondzell 2003). MODFLOW has been shown to be particularly suited for simulating GW processes (McDonald and Harbaugh 1988). MODFLOW was first released in 1984 (Harbaugh and McDonald 1996; Harbaugh et al. 2000). Harrington et al. (1999) used MODFLOW to show that lateral flow rates ranged from 4 to 38 m yr−1 (average of 19.4 m yr−1) and from 0.4 to 5.5 m yr−1 (average of 1.9 m yr−1) in the Gambier unconfined and the Dilwyn confined basins, respectively, in southern Australia. Wroblicky et al. (1998) simulated the lateral extent of SW–GW interactions and hydrologic flux rates through the riparian zone along two first-order stream channels in Aspen Creek and Rio Calveras (New Mexico). They found that the hydraulic conductivity of alluvium and the variation in recharge rates had the greatest impact on the volume, direction, and spatial distribution of SW–GW interactions. Kasahara and Wondzell (2003) estimated SW–GW hydrologic fluxes and residence times along a mountain stream in the Cascades and showed that channel morphology features strongly controlled SW–GW flow and the residence time of water in the subsurface. Wondzell and Swanson (1996) used MODFLOW to analyze the seasonal and storm dynamics of the SW–GW water flux and reported that the subsurface flux was 79% of the stream discharge at summer low flow, 2% during winter base flow, and 0.7% during storms, illustrating the potential for MODFLOW to provide accurate estimates of temporal trends in SW–GW interactions. In recent years, MODFLOW utilities include solute transport and particle tracking (Harbaugh et al. 2000). In a study conducted in the Netherlands, Hefting et al. (2006) used MODFLOW to quantify nitrate loading in the shallow groundwater of a riparian zone and reported that nitrate loads were high within the forested zone, 87 g NO3 m−2 yr−1, relative to a grassland riparian zone, 15 g NO3 m−2 yr−1. Previous studies have shown the success of MODFLOW for estimating SW–GW interactions and nutrient transport, thus providing validation and building user confidence in model predictions.

Previous studies used a variety of time series to study SW–GW interactions. Recent studies have shown that short time series (from several months to 1 year) are sufficient to study dominant processes in SW–GW interactions. For example, using a 10-day simulation and a 1-day calibration, Lautz and Siegel (2006) used MODFLOW and Modular 3-Dimensional Transport Multi Species (MT3DMS) to show that the movement of SW into GW was predominantly an advective process at Red Canon Creek of the Rocky Mountains. Schilling et al. (2006) used MODFLOW and MT3DMS for a 4-month study period (with a 3-day calibration) to evaluate dilution and denitrification process in riparian zone groundwater at Walnut Creek in Iowa. Wondzell and Swanson (1996) used MODFLOW to quantify SW–GW hydrologic fluxes, during storm events for a 1-yr study period (8-day calibration) in a fourth-order mountain stream at McRae Creek in H. J. Andrews Experimental Forest in Oregon. SW–GW flow rates positively correlated to streamflow during base flow conditions but decreased during storm events because of high infiltration rates in the riparian zone. Given the successful outcomes of the previous studies that identified key processes influencing SW–GW hydrologic and nutrient interactions, the use of high-frequency (weekly) hydrologic and nutrient data to quantify spatiotemporal variations in SW–GW nutrient interactions is needed to better manage underrepresented shallow GW systems.

Despite the growing interest in improving SW–GW mechanistic understanding (Sophocleous 2002) and the availability of accurate groundwater modeling tools, SW–GW flow and nutrient concentration relationships in semikarst ecosystems, such as those in the central United States, have lacked investigation and publication in the literature. The following work used MODFLOW to improve process understanding of SW–GW interactions in an Ozark border forest region of the central United States. Objectives were to 1) assess the ability of MODFLOW to accurately predict shallow aquifer transient hydraulic head distribution; 2) model the spatiotemporal variations in SW–GW hydrologic exchange, particularly during high streamflows; 3) model the seasonal nitrogen flux between surface and shallow groundwater; and 4) use MT3DMS to identify the spatial extent of SW–GW mixing in the riparian zone.

2. Methods

2.1. Study site

This investigation took place in the Thomas S. Baskett Wildlife Research and Education Center (BWREC; Rochow 1972), located in the Ozark border region of south-central Missouri (Pallardy et al. 1988). The BWREC is a second-growth mixed deciduous forest. The forest has not been subject to anthropogenic disturbances in over 60 years. Two reaches of Brushy Creek within the BWREC (Figure 1) served as study reaches for the current work. Brushy Creek is a second-order stream (Strahler 1952) with an average slope of 0.94% draining to Cedar Creek 4 km south of the BWREC after draining a watershed with an area of 9.17 km2.

Figure 1.
Figure 1.

Study sites and instrument locations at Baskett Wildlife Research and Education Center, central Missouri, where S is the stilling well site and PZ is the piezometer site.

Citation: Earth Interactions 18, 20; 10.1175/EI-D-14-0015.1

The BWREC has a humid continental climate (Critchfield 1966). Mean January and July temperatures are −2.2° and 25.5°C, respectively, while mean annual precipitation was 1024 mm from 1971 to 2011, recorded at Columbia Regional Airport, 8 km north of the BWREC. The average annual temperature and average precipitation measured at the onsite AmeriFlux tower from 2005 to 2011 were 13°C and 930 mm, respectively, versus 12.9°C and 1089 mm at Columbia Regional Airport during the same time period. The underlying geology of the BWREC is of Ordovician and Mississippian age. Dominant soils are Weller silt loam and Clinkenbeard clay loam (Rochow 1972). Riparian soils consist of Cedargap and Dameron soil complexes (USDA 2012, soil map unit 66017). The well-drained BWREC soils have an average bulk density of 1.2–1.4 g cm−3 (Young et al. 2003, 25–35). Divided into three layers, riparian soils are composed of a silty surface layer (0–38 cm), a silt loam layer (38–78 cm), and a gravelly layer (78–150 cm). Average depth to bedrock in the riparian zone, for the entire watershed, is approximately 150 cm (USDA 2012). The BWREC’s current land use ranges from second-growth forests in the southern portion to pastures in the northern portion. The watershed consists of 2.6% suburban land use, 17.9% cropland, 33% grassland, 43.2% forest, and 3.3% open water and wetlands (USDA 2012). The BWREC’s vegetation consists of northern and southern division oak–hickory forest species (Rochow 1972) including American sycamore (Platanus occidentalis), American elm (Ulmus americana), and black maple (Acer nigrum) dominated riparian reaches (Belden and Pallardy 2009). Understory vegetation consists of sugar maple (Acer saccharum), flowering dogwood (Cornus florida), and black cherry (Prunus serotina) (Reed 2010).

2.2. Instrumentation and data description

2.2.1. Climate data

Precipitation (Campbell Scientific Inc., TE525 Texas Electronics: error of ±1% for rates up to 2.54 cm h−1) and air temperature data (Vaisala HMP45C-L: error of ±0.2°C from 0° to 60°C and ±0.4°C at −35°C) were obtained for the 2011 water year (WY) from an AmeriFlux tower located at an elevation of 238 m on a forested ridge approximately 100 m outside of the watershed (Figure 1). The 2011 WY is defined from 1 October 2010 to 30 September 2011.

2.2.2. Stream stage and hydraulic head measurements

Four in-stream stilling wells were installed (referred to as SI–SIV) that supplied stage data to estimate stream discharge entering and leaving each study reach (Figure 1). Each stilling well was equipped with a Solinst Levelogger Gold pressure transducer (error of ±0.003 m) with stream stage recorded at 5-min intervals. Streamflow rating curves were determined from measured stage–discharge relationships using the stream cross-sectional method (Dottori et al. 2009) with a flowmeter (Marsh-McBirney Flo-Mate flowmeter with an error of ±2%).

Between SI and SII, six piezometers were installed in transect [piezometer site I (PZI)] extending 9 m from the stream bank into the riparian zone (Figure 1). PZI was located at an elevation of 177 m along an east–west stream reach approximately 90 m long and 15 m wide at bankfull. Similarly, piezometer site II (PZII) was located between SIII and SIV, at an elevation of 174 m, along an approximate north–south stream reach 157 m long and 10 m wide at bankfull. Each 3.58-m drive-point piezometer had a 4-cm inner diameter and a 76-cm slotted screen at the end (Figure 2) and was instrumented with a Solinst Levelogger Gold pressure transducer that recorded hydraulic head at 5-min intervals. To compensate for elevation differences between wells, water level was adjusted from the datum by adding gravitational head [calculated from the difference in depth of wells from a reference datum (343 cm)].

Figure 2.
Figure 2.

Conceptual diagram showing cross section of piezometer study design at Baskett Wildlife Research and Education Center, central Missouri. Soil profile depths are watershed averages (USDA 2012).

Citation: Earth Interactions 18, 20; 10.1175/EI-D-14-0015.1

2.3. Numerical modeling

2.3.1. Model description and assumptions

Shallow groundwater flow was modeled using MODFLOW 2000 (McDonald and Harbaugh 1984) and a three-dimensional finite difference model, distributed with a graphical user interface by Aquaveo [Groundwater Modeling System (GMS)]. Solute transport and the lateral extent of surface water–groundwater mixing were modeled using the MODFLOW extensions, MT3DMS and a Particle-Tracking Model for MODFLOW (MODPATH).

The ground surface elevation assigned to the model layers was acquired from a 5-m 2007 Missouri Spatial Data Information Service (MSDIS) digital elevation model (DEM; downloaded from MSDIS 2012). The model area was 1227 m × 3855 m, with eastern, western, and southern boundaries defined by Brushy Creek watershed boundaries and the northern boundary following the groundwater flow line defined by the elevation at that point. The initial grid spacing across the model domain was 10 m × 10 m and was later refined to 2 m × 4 m, which was the smallest cell size feasible, constrained by computing power. Refined grid dimensions were sufficient to capture individual piezometer well locations and stream channel.

Natural Resources Conservation Service and USDA soil maps (USDA 2012) were used to identify the three dominant soil layers of each study site resulting in a three-layer model with regular centered blocks and finite difference grids with a 0.1-m cell length. The bottom layer was 0.74 m thick, with hydraulic conductivities of 86.4, 8.64, and 8.64 m day−1 for Kx, Ky, and Kz, respectively. The hydraulic conductivity (K) was estimated using Rosetta, a built-in computer program in HYDRUS–1D (Schaap et al. 2001) that estimates pedotransfer functions (PTFs) by supplying the textural class and two groundwater head values as input data (Schaap et al. 1998). The bottom layer had a gravel texture and 44.6%, 32.6%, and 22.8% of sand, silt, and clay, respectively. The Kz was set an order of magnitude lower in order to reflect the anisotropy commonly observed in such systems as per USDA (2012) and Storey et al. (2003). The second layer, with a silt loam texture and 24.4%, 57.4%, and 18.2% of sand, silt, and clay, respectively, was 0.46 m thick with three-dimensional hydraulic conductivities of 0.18 m day−1 (Kx and Ky) and 0.018 m day−1 (Kz). The top layer, an alluvium deposit with a silt loam texture and 18.3%, 63.1%, and 18.6% of sand, silt, and clay, respectively, was 0.38 m thick with hydraulic conductivities of 8.64, 0.864, and 0.864 m day−1 for Kx, Ky, and Kz, respectively. The assigned K values were validated with slug tests performed on site (see section 2.3.2) and with the values mentioned in Freeze and Cherry (1979). For soil layers 1–3, soil porosity was set to 0.4, 0.5, and 0.3, respectively, as per the values published in Anderson and Woessner (1992). Similarly, specific yield and specific storage were 0.2, 0.09, and 0.3 and 1 × 10−4, 1 × 10−3, and 1 × 10−5 m−1 (Fetter 2001; Anderson and Woessner 1992; Freeze and Cherry 1979) for layers 1–3, respectively.

Brushy Creek was represented with a specific head arc with nodes representing the observed daily stream stage. Stream elevation was quantified by a DEM (5-m resolution). Observed daily precipitation data were used to define recharge per day and were applied as a constant recharge flux to the top of the most active layer. The bottom of the third layer (gravel) was assigned a no-flow boundary condition due to presence of confining layer and/or bedrock (USDA 2012). The model was initially forced at steady state, following which the model variables (specifically, Kx, Ky, Kz, the width and the depth of the stream, the alluvial layer, and the model solver package) were each parameterized independently. In particular, K values were parameterized according to the range specified in Freeze and Cherry (1979).

2.3.2. Model calibration and validation

As per methods used by Storey et al. (2003), MODFLOW was forced at steady state and specific parameters (Kx, Ky, Kz, the alluvial layer, the time step, and the model solver package) were selected to reflect conditions of Brushy Creek riparian zone as listed in the USDA (2012) soil maps. The model was parameterized with observation wells (n = 12) corresponding to piezometer locations (Figures 1 and 2), enabling comparison of modeled data to observed data. MODFLOW was calibrated for a 3-month period (April–June 2010, with a total of 91 time periods). During the calibration period, soil physical and hydrological (Kx, Ky, Kz, specific storage, porosity, longitudinal dispersivity, and specific yield) parameters were adjusted as per the methods of Wondzell and Swanson (1996), Lautz and Seigel (2006), and Schilling et al. (2006). During the calibration period, the estimated error interval was set to ±0.01 m, with a confidence interval of 95%. The residual (difference between observed and modeled head) was calculated to assess the performance of the model. Thus, following calibration, a residual near zero was achieved and the model was validated from July to September 2010.

To quantify model bias, observed and modeled head values were evaluated using the Nash–Sutcliffe (NS) efficiency parameter (Nash and Sutcliffe 1970), the root-mean-square error (RMSE) (Willmott 1981), the mean difference (MD), and the standard regression method (r2). NS parameter values range from −∞ to 1.0, where 1.0 indicates that the model is in perfect agreement (Moriasi et al. 2007; Luo and Sophocleous 2010). RMSE values closer to zero indicate better model performance (Moriasi et al. 2007). The equation of the best-fit regression line (the r2 coefficient of determination) can indicate the agreement between the modeled and observed head, provided that modeled and observed heads vary linearly (Luo and Sophocleous 2010). The equations needed to calculate the aforementioned statistics are as follows:
e1
e2
e3
where vo is the variance of observed values, N is the number of data points, xi is the observed value, yi is the corresponding predicted value, and is the average observed value for the study period.

2.3.3. Groundwater flow modeling

Model parameters used in this study are listed in Table 1. Once calibrated, the model was forced in steady state mode to estimate the starting heads. The observed hydraulic heads at each cell were used as initial heads for transient simulations. A more detailed discussion of the parameterization of the MODFLOW code is contained in the MODFLOW user’s manual (Harbaugh et al. 2000).

Table 1.

Model parameter values used in MODFLOW.

Table 1.

The model was then implemented with piezometer transects. The cells for which daily hydraulic heads were known (i.e., cells overlapping the stream and the piezometers) were isolated and head data were entered. Daily stream hydraulic head values (n = 4 sites; Figures 1 and 2) were input into the model. MODFLOW estimates the head values at other locations along the stream by interpolation (McDonald and Harbaugh 1984). Hydraulic head values for the start and end of the stream were assigned to the elevation of the stream at that point as per Anderson and Woessner (1992).

2.3.4. Nitrate transport modeling

Groundwater flow results obtained from the MODFLOW simulation were used as inputs to the MT3DMS model (Zheng and Wang 1999). Active cells in the MT3DMS transport model were identical to those in the flow model, as the MT3DMS was built on the existing MODFLOW grid. The stilling well locations (SI–SIV) and the piezometer locations were set as specific concentration nodes. Weekly observed nitrate concentration data (mg L−1) for WY 2011 (n = 52) collected at stream stage monitoring sites SI–SIV and in riparian zone shallow groundwater wells were used as specific concentration boundary conditions for the model (Figure 2). It was assumed there were negligible nitrate inputs from external sources (e.g., drainage, leakage, and fertilizer applications). The MT3DMS model was forced at daily time steps with MODFLOW to assess spatiotemporal variations in aquifer nitrate loading (kg m−2) for WY 2011. A more detailed discussion of the fundamentals of the MT3DMS module is located in the MT3DMS user’s manual (Zheng and Wang 1999).

2.3.5. Surface water–groundwater lateral interaction extent modeling

MODPATH was forced with groundwater flow results from MODFLOW. MODPATH is a particle-tracking postprocessing module for computing the lateral extent of SW–GW interactions and flow paths (Pollock 1994). MODPATH also computes the travel times associated with each particle. MODPATH tracer particles, which are imaginary water particles (Pollock 1994), were placed upstream of SI and SIII to compute the lateral extent of the flow paths that begin in the stream, pass through the shallow aquifer, and then rejoin the stream. Monthly variations in flow paths were assessed by running MODPATH along with monthly MODFLOW results. As per study results of Storey et al. (2003) stream morphological characteristics such as meanders, woody debris, and the presence of boulders were omitted. Quantifying streambed heterogeneities and evapotranspiration that can affect net SW–GW flux volume (Woessner 2000; Sophocleous 2002) were also beyond the scope of the current work but provide impetus for future investigations. Since the current work focused primarily on quantifying lateral extent of SW–GW interactions, neglecting streambed heterogeneities and evapotranspiration did not affect GW flow lateral extent calculations.

3. Results and discussion

3.1. Climate during the study period

Climate in the BWREC during WY 2011 was warmer and drier than normal with mean air temperature of 12.5°C and total precipitation of 647 mm. The maximum precipitation for a single day was 44 mm on 31 December 2010. Seasonal precipitation (winter, spring, summer, and fall) was 170 mm from December to March, 250 mm from March to June, 135 mm from June to September, and 94 mm from September to December. Annual precipitation during the water year was approximately 21% lower than the 30-yr average (1971–2012) of 816 mm measured at the Columbia Regional Airport (6 km from BWREC). The annual-mean air temperature was approximately 11% cooler relative to the 30-yr average air temperature of 14.1°C.

3.2. Hydraulic head of stream stage monitoring wells

Observed 5-min interval hydraulic heads at the four stilling wells (SI–SIV) and the 12 piezometers (Pz1–Pz12) were reduced to daily average values for WY 2011 (Figure 3). During WY 2011, SI had the highest average stage (178.36 m), followed by SIII, SIV, and SII with 178.56, 178.49, and 177.03 m, respectively (Table 2). The winter and spring seasons had higher stage (average = 178.69 m) relative to the summer (178.45 m) and fall (178.42 m). The difference in stage measurements at the streambed between the upstream (SI) and downstream locations (SIV; at 830 m apart) was greater during fall (0.15 m) and summer (0.13 m) relative to winter (0.12 m) and spring (0.11 m) and presumably attributable to variations in seasonal baseflow conditions (Chinnasamy and Hubbart 2014). Annual average daily groundwater head was higher at PZII (178.37 m) relative to PZI (177.18 m), even though similar vegetation was present at both sites. Similar to the observed stage, the groundwater head was higher during the winter and spring (178.59 m at PZI), followed by summer (178.21 m at PZI) and fall (178.12 m at PZI). The stage and groundwater head values were higher in the winter and spring following higher precipitation compared to summer and fall, indicating riparian zone groundwater recharge with precipitation. The average depth to groundwater was 69.70 cm at PZI and 92.32 cm at PZII during spring months (32% difference) and 253.41 cm at PZI and 231.30 cm at PZII during fall months (8% difference). During the dry season, depth to groundwater was 214.9 and 197.61 cm at PZI and PZII, respectively, and water level in the piezometers dropped below average level (126.62 and 150. 93 cm at PZI and PZII, respectively).

Figure 3.
Figure 3.

Measured precipitation (mm) and hydraulic head at stream (m) and at piezometer (m) locations during WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

Citation: Earth Interactions 18, 20; 10.1175/EI-D-14-0015.1

Table 2.

Descriptive statistics of hydraulic heads and error analysis between observed and modeled heads for the 2011 WY at Baskett Wildlife Research and Education Center, central Missouri.

Table 2.

3.3. Model calibration and validation

After achieving a residual (see section 2.3.2.) of 0.001 and 0.000 m at PZI and PZII, respectively, model parameters were used for model validation (July–September 2010). The residuals during the validation period were 0.003 and 0.001 m at PZI and PZII, respectively. Nash–Sutcliffe values, for the calibration period, ranged from −0.54 to 0.47 (Table 2). During the calibration period, the RMSE values ranged from 0.25 to 0.61, while the MD and r2 values ranged from −0.08 to 0.46 m and from 0.77 to 0.54, respectively. The average NS, RMSE, MD, and r2 values at PZII were 0.29, 0.29 m, −0.17 m, and 0.75, respectively; these values were better than the average NS, RMSE, MD, and r2 values of −0.31, 0.56 m, −0.35 m, and 0.59 at PZI, respectively. Therefore, during model validation, relative to PZI, the MODFLOW hydraulic head predictions were closer to actual head values for the piezometers located at PZII (Figure 4), which was located in the center of the modeling domain. Model results during periods of lower head (e.g., August–October 2011) or when there was no streamflow (August–October 2011) were confounding. It is possible to improve model predictions, for low-flow conditions, with accurate boundary conditions for the entire watershed and not for only the floodplain. However, those data were not available for the current work.

Figure 4.
Figure 4.

Observed vs modeled (MODFLOW) hydraulic heads for piezometers located at PZI (Pz1–Pz6) and at PZII (Pz7–Pz12) over WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

Citation: Earth Interactions 18, 20; 10.1175/EI-D-14-0015.1

3.4. Groundwater flow simulations

For WY 2011, the entire study reach was, on average, a losing stream with 1988 m3 day−1 lost to GW. On an annual average, the study reach between SI and SII (160 m in length) was a losing reach (1201 m3 day−1); the study reach between SII and SIII (543 m in length) was also a losing reach (1129 m3 day−1); and the study reach between SIII and SIV (149 m in length) was a gaining reach (343 m3 day−1). Thus, on average, the study reach lost more water to the shallow aquifer during summer (2405 m3 day−1) relative to water lost during the fall (2184 m3 day−1), spring (2102 m3 day−1), and winter (1549 m3 day−1) seasons. This result is reasonable considering that the majority of rainfall occurs from early spring to midsummer months in central Missouri.

These results are similar to other central U.S. studies. For example, Marzolf et al. (1994) reported average streamflow gain (from GW) of 17.28 m3 day−1 in a study conducted at Walker Branch Creek (reach length = 62 m) in Tennessee. Marzolf et al. (1994) results were 0.9% of the streamflow gain observed at Brushy Creek with reach length = 830 m. The higher flow in Brushy Creek relative to the Walker Branch Creek study is explained in part by larger drainage area and study reach length. Mulholland et al. (1997) indicated that the Hugh White Creek (reach length = 78 m) in North Carolina was, on average, a gaining reach with GW input of 140 m3 day−1, which is 7% of the volume of water lost per day in Brushy Creek. Presumably, the presence of semikarst geology in BWREC could result in greater SW water loss to the GW aquifer relative to the geological composition of Walker Branch Creek and Hugh White Creek. At BWREC, the entire study reach lost 2331 m3 day−1 to the aquifer and gained 343 m3 day−1 from the aquifer (annual average). Monthly plan view maps (Figure 5) of water flow vectors showed that shallow groundwater primarily flows along layer 3 (the gravel layer) because of a higher hydraulic conductivity relative to other layers and because of the presence of the shallow bedrock (low depth to bedrock in the floodplain). A visual inspection of flow vectors at every layer indicated that during higher streamflow periods, a significant amount of water movement occurred from the stream toward the subsurface aquifer in the top and second layers.

Figure 5.
Figure 5.

Plan view of MODFLOW estimates of hydraulic head distribution (m) for the months of November, February, May, and August over the WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

Citation: Earth Interactions 18, 20; 10.1175/EI-D-14-0015.1

3.5. Surface water–groundwater interactions

The MODFLOW generated flow budget showed that flux rates at the stream reach alternated between positive, where water entered the aquifer (study reaches SI–SII, SII–SIII, and SI–SIV), and negative, where water entered the stream (study reach SIII–SIV). Alternating between a losing or gaining stream indicates that stream water lost to the aquifer may reemerge at downstream locations, depending on the flow path and residence time. The groundwater flux per unit stream reach length was similar for the three study reaches, with values of 2.08, 2.30, and 2.47 m3 day−1 m−1 for reaches SII–SIII, SIII–SIV, and SI–SIV, respectively. This result indicates that the water flow (7%) and geomorphological and soil physical properties did not differ greatly between study reaches. The high groundwater flux rates observed at BWREC in the current work indicate high SW–GW connectivity was at least somewhat attributable to semikarst geology. Mulholland et al. (1997) reported a groundwater flux rate of 1.80 m3 day−1 m−1 at Hugh White Creek (reach length = 72 m), North Carolina, while Marzolf et al. (1994) and Fellows et al. (2001) reported groundwater flow rates of 0.80 and 0.04 m3 day−1 m−1 at Walker Branch Creek (reach length = 62 m) in Tennessee and Rio Calaveras (reach length is 110 m) in New Mexico, respectively. These studies concluded that the lateral extent of SW–GW hydrological connectivity was proportional to the groundwater flux rate. Based on previous results, Brushy Creek could have a greater lateral SW–GW extent than that observed in other regions of the United States, likely attributable to semikarst flow paths. Improving mechanistic understanding of shallow groundwater flow and residence time provides impetus for future investigations.

3.6. Nitrate transport and loading

Temporal variations in nitrate concentrations between SW and GW were significant (p < 0.01) during the study period. Spatial variations between SW nutrient concentrations were also significant (p = 0.001). Groundwater nutrient concentrations, however, were not significantly different (p > 0.05) between sites. Nitrate concentrations within the entire study reach were highest during winter (0.994 mg L−1), followed by the spring (0.346 mg L−1), summer (0.200 mg L−1), and fall (0.113 mg L−1) seasons of WY 2011. The MT3DMS model results indicated a net annual nitrate loss of −328 kg day−1 from the study reach (830 m) to the aquifer. Study reach SI–SII (distance of 160 m) lost, on average, 54 kg day−1 to the aquifer (and/or to the riparian zone), while SII–SIII (distance of 543 m) lost 268 kg day−1 and SIII–SIV (distance of 149 m) lost 6 kg day−1 to the aquifer (and/or to the riparian zone). Nitrate loading to the aquifer increased with stream reach length (r2 = 0.95; Figure 6). Relative to other study reaches, the SIII–SIV reach had the lowest nitrate loading. This result could be attributable to the relative length of the SIII–SIV reach, which was 11% shorter than other study reaches. In addition, observed data indicated that there was more water input from the shallow aquifer to the stream at SIII–SIV, which may have resulted in nitrate dilution. Study reaches SI–SII and SII–SIII had the highest nitrate losses during the winter season (107 and 594 kg day−1, respectively), while SIII–SIV had the highest nitrate losses during spring (10 kg day−1). Study reaches SI–SII, SII–SIII, and SIII–SIV had the lowest nitrate losses during the fall season, with a net average loss of 13, 71, and 1 kg day−1, respectively (Figure 6). MT3DMS results indicated a lateral extent of 70 m up to which surface–groundwater hydrologic and nutrient mixing occurred.

Figure 6.
Figure 6.

MT3DMS estimates of nitrate loading in the shallow aquifer from different segments of the study reach for the WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

Citation: Earth Interactions 18, 20; 10.1175/EI-D-14-0015.1

The daily average nitrate loading from the stream to the aquifer (328 kg day−1) was lower than the average amount of nitrate uptake by second-order streams (10 000–100 000 kg day−1) as reported by Ensign and Doyle (2006). According to their metric, nitrate loading in BWREC was lower than the potential amount of nitrate processed by a second-order stream. The aquifer nitrate loading observed between stream reach sites in Brushy Creek was significantly different (p < 0.001), indicating that the rate of in-stream nitrate-cycling processes could be different between headstream locations and downstream locations (Ensign and Doyle 2006). On this basis, management plans to regulate nutrient loading may need to vary between headwaters and downstream reaches.

The current study results indicated that Ozark border streams, such as Brushy Creek, could be a source of nitrate to riparian zone shallow aquifers during shallow groundwater recharge periods. Peterjohn and Correll (1984) showed that a deciduous forest can have an average of 20, 50, and 60 kg ha−1 yr−1 of total nitrogen from precipitation, upslope groundwater sources, and leaf litter decomposition, respectively. Applying Peterjohn and Correll’s (1984) estimated rates to BWREC, the nitrogen received from Brushy Creek surface water of 85.5 kg ha−1 yr−1 indicates that approximately 40% of the annual deciduous forest riparian zone nitrogen budget is received from surface water loading.

Rates of denitrification and plant nutrient uptake can improve MODFLOW predictions. However, denitrification and plant uptake rates were not quantified in the current work. Future investigations should focus on vegetation cover and the rate of soil nutrient mobilization to groundwater and the delivery to surface water. While not directly studied in the current work, it was assumed that nitrate concentrations were reduced (by more than 90%) in the riparian zone by geochemical processes (Levia et al. 2011). Future studies should quantify specific biogeochemical transform rates in the riparian zone (e.g., the nitrification and denitrification of nitrates and the volatilization of ammonium), thereby increasing understanding regarding hydrologically mediated biogeochemical processes in the Ozark border region of the central United States.

3.7. Spatiotemporal variations in the lateral extent of surface water–groundwater interactions

MODPATH revealed significant spatiotemporal variations (Table 3 and Figures 7 and 8) between sites PZI and PZII (p < 0.05) in subsurface flow path and travel time, ranging from 213 m and 3.6 yr to 197 m and 11.6 yr. The temporal variations in lateral extent of SW–GW mixing (10–75 m from the stream bank) indicates that forest managers should be aware that SW–GW connectivity may stretch well beyond current recommended buffer widths. The annual average flow-path distances at PZI and PZII were 196 and 189 m, respectively. The annual average travel time was 106% higher at PZII (4326 days) than at PZI (1330 days), indicating that nutrients transported by water had more time to undergo nutrient transformation in the subsurface at PZII, relative to PZI. Extended residence times also indicate that water is stored longer in the soil matrix at PZII and therefore available for plant uptake. The computed near-streamflow paths from Brushy Creek appeared, disappeared, contracted, and expanded in response to seasonal hydrologic changes (Figures 7 and 8). Modeling results indicated identical behavior on the unmonitored side of the riparian zone; therefore, management implications of the results of this work should be applied with confidence to either side of the stream reach. During drier months (May in particular), flow paths did not extend into the piezometer transect but were restricted to the stream bank and within the stream channel. During such periods, surface–groundwater interactions were negligible. Thus, even during low-flow periods, the stream reach was, on average, losing and therefore hydrologically well connected to GW. During no-flow periods, SW–GW hydrologic connectivity was not quantified because of a lack of monitoring wells in the streambed.

Table 3.

Seasonal and spatial MODPATH lateral flow pathlength and travel time results for study sites at Baskett Wildlife Research and Education Center, central Missouri.

Table 3.
Figure 7.
Figure 7.

MODPATH estimates of lateral extent of surface water–groundwater interactions for the months of November, February, May, and August over the WY 2011 at Baskett Wildlife Research and Education Center, central Missouri. The solid lines indicate flow paths as simulated by MODPATH.

Citation: Earth Interactions 18, 20; 10.1175/EI-D-14-0015.1

Figure 8.
Figure 8.

MODPATH estimates of lateral extent (m = meter) of surface water–groundwater interactions from the stream bank and travel time (d = day) of water in flow paths over the WY 2011 at Baskett Wildlife Research and Education Center, central Missouri.

Citation: Earth Interactions 18, 20; 10.1175/EI-D-14-0015.1

The baseline results from this study provide a foundation for the development of future studies and management plans that should include consideration of water residence time and storage in the riparian zone. Monthly MODPATH results indicated spatiotemporal variations in flow pathlength and travel time. In particular, increases in lateral flow pathlength results in increased travel time. Annual average travel times at PZI and PZII were estimated to be 3.6 yr (1329 days) and 11.8 yr (4326 days), respectively. Annual average nutrient concentration at PZII (0.009 mg L−1) was 18% lower than that in PZI (0.011 mg L−1). In-stream nutrient processing could be the reason for the observed lower nutrient loading in the downstream extents (SIV) of Brushy Creek. Presumably attributable to increased precipitation (80%) during the fall season, the model predicted increased flow pathlengths and corresponding increases in the travel time (Figures 7 and 8).

MODPATH results indicated 37, 31, 3, and 37 distinct flow-path lines (Figure 7) for November, February, May, and August of WY 2011, respectively. The specific location of flow-path lines is important for management plans in terms of buffer strip planning (Bentrup 2008). According to Pollock (1994), information about dominant subsurface flow-path lines that originate from the stream are useful to predict future stream meandering patterns. Wroblicky et al. (1998) reported an annual average of 9 and 6 distinct flow-path lines for study sites located at Aspen Creek and Rio Calaveras, in New Mexico, respectively. More distinct flow-path lines observed along Brushy Creek could be due to the presence of semikarst geology in the study area. The variability in the number of distinct flow-path lines during WY 2011 corroborates other evidence that the stream reach alternates between a losing and gaining system, as described in Wroblicky et al. (1998). Lautz and Siegel (2006) reported 17 distinct flow-path lines with varying lengths in Red Canyon Creek in Wyoming. They further reported that the residence time ranged from several hours to 10 years, relative to an average of 7 years observed at BWREC. The average flow pathlength at Red Canyon Creek was also shorter (150 m) relative to that at BWREC (205 m). The shorter average residence times at BWREC relative to that at Aspen Creek, Rio Calaveras, and Red Canyon could be due to the relatively high hydraulic conductivity (i.e., semikarst geology) at BWREC. The nutrients transported in semikarst flow-path lines have shorter time to undergo biochemical transformations relative to those provided in other studies (e.g., Lautz and Siegel 2006). However, the nutrients can be lost to deeper aquifers or karst flow paths below subsurface flow paths (Hill 1996). Therefore, management plans should consider no thinning policies in forested riparian zones in order to maximize plant nutrient uptake from shallow GW before nutrients are lost to deep aquifers to mitigate nutrient GW pollution.

3.8. Model limitations

The results from this work are novel due to the lack of such information from the Ozark border region. Results therefore provide a great deal of new baseline information to the scientific literature and land managers. A high-frequency (5-min interval) 1-yr time series of hydrologic data (surface water and groundwater) was used to identify dominant semikarst geology governed hydrologic processes and flow paths. A longer time series would be useful to improve understanding of climate variability mediated alterations to observed hydrologic process of the current work. However, that was not the objective of this paper. The current study was conducted in a forest conservation area that has remained largely unchanged for approximately the past 100 years to provide critical and timely baseline process information to scientists and land managers in the region. The current work also tested the applicability of a physically based groundwater model in a unique semikarst Ozark border region, in which hydrologic data and related modeling remain scarce.

There were some modeling limitations in the current work. For example, the accuracy of subsurface water flux estimates, which are proportional to the hydraulic conductivities (Kx, Ky, and Kz), were limited to the accuracy of K values (Wondzell and Swanson 1996), which is difficult to measure and often limited in the published literature (Freeze and Cherry 1979; Harvey and Wagner 2000; Fetter 2001). The absence of borehole data, for quantifying the model stratigraphy, could also introduce simulation errors. Because of the absence of nitrate concentration data at the model boundaries, study results were limited to the identification of nutrient loading within the aquifer from surface water to the last piezometer located 9 m from the stream bank into the floodplain. Since the current study focused only on advection and dispersion processes in the cycling of nitrate, biological processes that affect the nitrate-cycling rate (i.e., denitrification) were omitted from the analyses.

4. Conclusions

Surface water–groundwater hydrologic and nutrient interaction modeling using MODFLOW, MT3DMS, and MODPATH was effective for determining spatiotemporal variations in magnitude and extent of SW–GW interactions in semikarst geology of the U.S. Midwest. The current study approach is novel with regard to the use of transient flow conditions in underrepresented semikarst geology of the Midwest. Transient simulations were possible because of the availability of high-frequency observed water quality (weekly nutrient concentration) data and water quantity (stream and shallow groundwater flow) monitoring networks.

MODFLOW modeling provided numerical approximations of hydrologic head that were statistically comparable to field observations (NS = 0.47, r2 = 0.77, RMSE = 0.61 cm, and MD = 0.46 cm). The average NS, RMSE, MD, and r2 values at PZII were 0.29, 0.29 m, −0.17 m, and 0.75, respectively, and were better than the average NS, RMSE, MD, and r2 values of −0.31, 0.56 m, −0.35 m, and 0.59 at PZI, respectively. MODFLOW results indicated that the study reach was, on average, a losing stream (82% of the length) with significant seasonal variations (p < 0.05). For the study period, the study reach between SI and SII (160 m) was a losing reach (1201 m3 day−1); the study reach between SII and SIII (543 m) was a losing reach (1129 m3 day−1); and the study reach between SIII and SIV (149 m) was a gaining reach (343 m3 day−1). Thus, on average, the study reach lost more water to the shallow aquifer during summer (2405 m3 day−1) relative to water lost during the fall (2184 m3 day−1), spring (2102 m3 day−1), and winter (1549 m3 day−1) seasons.

The shallow groundwater flux per unit length was 2.47 m3 day−1 m−1 and was not significantly different (p > 0.05) between study sites. Given this result, it was inferred that geomorphological and soil physical properties did not differ between study reaches. The MT3DMS model results indicated a net annual nitrate loss of −328 kg day−1 from the study reach to the GW. However, even with a high nitrate loading, GW nitrate concentrations were low relative to that of SW. Results indicated high nitrate concentrations of surface water relative to groundwater (greater than 90%) over the study period. Most nitrate transport to the subsurface aquifer from the stream occurred because of advection processes from the stream (i.e., physical processes) that vary spatially and temporally. The stream also serves as a nitrate source to the riparian zone for 80% of the stream length. MODPATH revealed significant spatiotemporal variations, between sites PZI and PZII (p < 0.05), in subsurface flow path and travel time, ranging from 213 m and 3.6 yr to 197 m and 11.6 yr. The annual average travel time was 106% higher at PZII (4326 days) than at PZI (1330 days), indicating that the nutrients transported by water had more time to undergo biochemical transformation in the riparian zone subsurface at PZII, relative to PZI.

The use of MT3DMS along with MODPATH in MODFLOW increased modeling confidence for estimating the lateral extent of SW–GW mixing (75 m) in semikarst geology of the Midwest. The lateral extent of SW–GW interaction was not uniformly distributed along the study reach but exhibited temporal and spatial variations (i.e., disappeared, expanded, contracted, and/or reappeared). Results from the current study are likely representative of similar semikarst and karst geologic setting environments across the central United States. Since understanding hydrologic processes is requisite for estimating biogeochemical processes, the current work serves as a basis for future spatiotemporal investigations of biogeochemical processes and enhanced management strategies in the central United States and other karst hydroregions.

Acknowledgments

Funding was provided by the Environmental Protection Agency (EPA) Region 7 (Grant CD-97701401-0). Results presented may not reflect the views of the EPA and no official endorsement should be inferred. Gratitude is extended to multiple members of the Interdisciplinary Hydrology Lab (IHL) of the University of Missouri School of Natural Resources. Sincere gratitude is extended to Stephen Pallardy and Kevin Hosman for climate data and valuable information about the BWREC and to multiple reviewers, whose comments greatly improved the quality of the manuscript.

References

  • Anderson, M. P., , and W. W. Woessner, 1992: Applied Groundwater Modeling: Simulation of Flow and Advective Transport. Academic Press, 381 pp.

  • Belden, A. C., , and S. G. Pallardy, 2009: Successional trends and apparent Acer saccharum regeneration failure in an oak-hickory forest in central Missouri, USA. Plant Ecol., 204, 305322, doi:10.1007/s11258-009-9593-4.

    • Search Google Scholar
    • Export Citation
  • Bentrup, G., 2008: Conservation buffers: Design guidelines for buffers, corridors, and greenways. U.S. Department of Agriculture Forest Service Southern Research Station General Tech. Rep. SRS-109, 136 pp.

  • Burt, T., , G. Pinay, , and S. Sabater, 2010: What do we still need to know about the ecohydrology of riparian zones? Ecohydrology, 3, 373377, doi:10.1002/eco.140.

    • Search Google Scholar
    • Export Citation
  • Chinnasamy, P., , and J. A. Hubbart, 2014: Measuring and modeling shallow groundwater flow between a semi-karst border stream and Ozark forested riparian zone in the central USA. J. Sci. Res. Rep., 3, 844865, doi:10.9734/JSRR/2014/7711.

    • Search Google Scholar
    • Export Citation
  • Critchfield, H. J., 1966: General Climatology. Prentice-Hall, 420 pp.

  • Dottori, F., , M. L. V. Martina, , and E. Todini, 2009: A dynamic rating curve approach to indirect discharge measurement. Hydrol. Earth Syst. Sci., 13, 847863, doi:10.5194/hess-13-847-2009.

    • Search Google Scholar
    • Export Citation
  • Ensign, S. H., , and M. W. Doyle, 2006: Nutrient spiraling in streams and river networks. J. Geophys. Res.,111, G04009, doi:10.1029/2005JG000114.

  • Fellows, C. S., , H. M. Valett, , and C. N. Dahm, 2001: Whole-stream metabolism in two montane streams: Contribution of the hyporheic zone. Limnol. Oceanogr.,46, 523–531, doi:10.4319/lo.2001.46.3.0523.

  • Fetter, C. W., 2001: Applied Hydrology. Prentice Hall, 598 pp.

  • Freeze, R. A., , and J. A. Cherry, 1979: Groundwater. Prentice Hall, 604 pp.

  • Gooseff, M. N., , S. M. Wondzell, , R. Haggerty, , and J. Anderson, 2003: Comparing transient storage modeling and residence time distribution (RTD) analysis in geomorphically varied reaches in the Lookout Creek basin, Oregon, USA. Adv. Water Resour., 26, 925937, doi:10.1016/S0309-1708(03)00105-2.

    • Search Google Scholar
    • Export Citation
  • Harbaugh, A. W., , and M. G. McDonald, 1996: User's documentation for MODFLOW-96, an update to the US Geological Survey modular finite-difference ground-water flow model. U.S. Geological Survey Open-File Rep. 96-485, 63 pp.

  • Harbaugh, A. W., , E. R. Banta, , M. C. Hill, , and M. G. McDonald, 2000: MODFLOW-2000, the U.S. Geological Survey modular ground-water model—User guide to modularization concepts and the ground-water flow process. U.S. Geological Survey Open-File Rep. 00-92, 127 pp.

  • Harrington, G. A., , G. R. Walker, , A. J. Love, , and K. A. Narayan, 1999: A compartmental mixing-cell approach for the quantitative assessment of groundwater dynamics in the Otway basin, south Australia. J. Hydrol., 214, 4963, doi:10.1016/S0022-1694(98)00243-1.

    • Search Google Scholar
    • Export Citation
  • Harvey, J. W., , and B. J. Wagner, 2000: Quantifying hydrologic interactions between streams and their subsurface hyporheic zones. Streams and Ground Waters, J. B. Jones and P. J. Mulholland, Eds., Academic Press, 3–44.

  • Haycock, N., , and G. Pinay, 1993: Groundwater nitrate dynamics in grass and poplar vegetated riparian buffer strips during the winter. J. Environ. Qual., 22, 273278, doi:10.2134/jeq1993.00472425002200020007x.

    • Search Google Scholar
    • Export Citation
  • Hefting, M., , B. Beltman, , D. Karssenberg, , K. Rebel, , M. van Riessen, , and M. Spijker, 2006: Water quality dynamics and hydrology in nitrate loaded riparian zones in the Netherlands. Environ. Pollut., 139, 143156, doi:10.1016/j.envpol.2005.04.023.

    • Search Google Scholar
    • Export Citation
  • Hill, A. R., 1996: Nitrate removal in stream riparian zones. J. Environ. Qual.,25, 743–755, doi:10.2134/jeq1996.00472425002500040014x.

  • Hynes, H. B. N., 1983: Groundwater and stream ecology. Hydrobiologia, 100, 9399, doi:10.1007/BF00027424.

  • Jacobs, T., , and J. Gilliam, 1985: Riparian losses of nitrate from agricultural drainage waters. J. Environ. Qual., 14, 472478, doi:10.2134/jeq1985.00472425001400040004x.

    • Search Google Scholar
    • Export Citation
  • Kasahara, T., , and S. M. Wondzell, 2003: Geomorphic controls on hyporheic exchange flow in mountain streams. Water Resour. Res., 39, 1005, doi:10.1029/2002WR001386.

    • Search Google Scholar
    • Export Citation
  • Lautz, L. K., , and D. I. Siegel, 2006: Modeling surface and ground water mixing in the hyporheic zone using MODFLOW and MT3D. Adv. Water Resour., 29, 16181633, doi:10.1016/j.advwatres.2005.12.003.

    • Search Google Scholar
    • Export Citation
  • Levia, D. F., , D. E. Carlyle-Moses, , and T. Tanaka, 2011: Reflections on the state of forest hydrology and biogeochemistry. Forest Hydrology and Biogeochemistry: Synthesis of Past Research and Future Directions, D. F. Levia et al., Eds., Springer, 729–733.

  • Luo, Y., , and M. Sophocleous, 2010: Seasonal groundwater contribution to crop-water use assessed with lysimeter observations and model simulations. J. Hydrol., 389, 325335, doi:10.1016/j.jhydrol.2010.06.011.

    • Search Google Scholar
    • Export Citation
  • Marzolf, E. R., , P. J. Mulholland, , and A. D. Steinman, 1994: Improvements to the diurnal upstream–downstream dissolved oxygen change technique for determining whole-stream metabolism in small streams. Can. J. Fish. Aquat. Sci.,51, 1591–1599, doi:10.1139/f94-158.

  • McDonald, M. G., , and A. W. Harbaugh, 1984: A modular three-dimensional finite-difference ground-water flow model. U.S. Geological Survey Open-File Rep. 83-875, 539 pp.

  • McDonald, M. G., , and A. W. Harbaugh, 1988: A modular three-dimensional finite-difference ground-water flow model. Techniques of Water-Resources Investigations of the United States Geological Survey A1, 588 pp.

  • Moriasi, D. N., , J. G. Arnold, , M. W. Liew, , R. L. Bingner, , R. D. Harmel, , and T. L. Veith, 2007: Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans. ASABE, 50, 885900, doi:10.13031/2013.23153.

    • Search Google Scholar
    • Export Citation
  • MSDIS, cited 2012: Missouri spatial data information service. [Available online at http://msdis.missouri.edu/data.]

  • Mulholland, P. J., , E. R. Marzolf, , J. R. Webster, , D. R. Hart, , and S. P. Hendricks, 1997: Evidence that hyporheic zones increase heterotrophic metabolism and phosphorus uptake in forest streams. Limnol. Oceanogr.,42, 443–451.

  • Nash, J. E., , and J. Sutcliffe, 1970: River flow forecasting through conceptual models part I—A discussion of principles. J. Hydrol., 10, 282290, doi:10.1016/0022-1694(70)90255-6.

    • Search Google Scholar
    • Export Citation
  • Pallardy, S., , T. Nigh, , and H. Garrett, 1988: Changes in forest composition in central Missouri: 1968-1982. Amer. Midl. Nat., 120, 380390, doi:10.2307/2426010.

    • Search Google Scholar
    • Export Citation
  • Peterjohn, W. T., , and D. L. Correll, 1984: Nutrient dynamics in an agricultural watershed: observations on the role of a riparian forest. Ecology, 65, 14661475, doi:10.2307/1939127.

    • Search Google Scholar
    • Export Citation
  • Pollock, D. W., 1994: User's Guide for MODPATH/MODPATH-PLOT, version 3: A particle tracking post-processing package for MODFLOW, the U.S. Geological Survey finite-difference ground-water flow model. U.S. Geological Survey Open-File Rep. 94-464, 464 pp.

  • Reed, S. E., 2010: Ambrosia beetle habitat use, Host use, and influence on early wood colonizing microbes in an oak-hickory forest. Ph.D. thesis, University of Missouri, 531 pp.

  • Rochow, J. J., 1972: A vegetational description of a mid-Missouri forest using gradient analysis techniques. Amer. Midl. Nat., 87,377396, doi:10.2307/2423570.

    • Search Google Scholar
    • Export Citation
  • Schaap, M. G., , F. J. Leij, , and M. T. van Genuchten, 1998: Neural network analysis for hierarchical prediction of soil hydraulic properties. Soil Sci. Soc. Amer. J.,62, 847–855, doi:10.2136/sssaj1998.03615995006200040001x.

  • Schaap, M. G., , F. J. Leij, , and M. T. van Genuchten, 2001: ROSETTA: A computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions. J. Hydrol.,251, 163–176, doi:10.1016/S0022-1694(01)00466-8.

  • Schilling, K., , Z. Li, , and Y.-K. Zhang, 2006: Groundwater–surface water interaction in the riparian zone of an incised channel, Walnut Creek, Iowa. J. Hydrol., 327, 140150, doi:10.1016/j.jhydrol.2005.11.014.

    • Search Google Scholar
    • Export Citation
  • Sophocleous, M., 2002: Interactions between groundwater and surface water: The state of the science. J. Hydrol., 10, 5267, doi:10.1007/s10040-001-0170-8.

    • Search Google Scholar
    • Export Citation
  • Storey, R. G., , K. W. F. Howard, , and D. D. Williams, 2003: Factors controlling riffle-scale hyporheic exchange flows and their seasonal changes in a gaining stream: A three-dimensional groundwater flow model. Water Resour. Res., 39, 1034, doi:10.1029/2002WR001367.

    • Search Google Scholar
    • Export Citation
  • Strahler, A., 1952: Hypsometric (area-altitude) analysis of erosional topography. Bull. Geol. Soc. Amer., 63, 11171142, doi:10.1130/0016-7606(1952)63[1117:HAAOET]2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • USDA, cited 2012: Web soil survey. [Available online at http://www.nrcs.usda.gov/wps/portal/nrcs/surveylist/soils/survey/state/?stateId=MO.]

  • Willmott, C. J., 1981: On the validation of models. Phys. Geogr., 2, 184194.

  • Woessner, W. W., 2000: Stream and fluvial plain ground water interactions: Rescaling hydrogeologic thought. Groundwater, 38, 423429, doi:10.1111/j.1745-6584.2000.tb00228.x.

    • Search Google Scholar
    • Export Citation
  • Wondzell, S. M., , and F. J. Swanson, 1996: Seasonal and storm dynamics of the hyporheic zone of a 4th-order mountain stream. II: Nitrogen cycling. J. N. Amer. Benthol. Soc., 15,2034, doi:10.2307/1467430.

    • Search Google Scholar
    • Export Citation
  • Wroblicky, G. J., , M. E. Campana, , H. M. Valett, , and C. N. Dahm, 1998: Seasonal variation in surface-subsurface water exchange and lateral hyporheic area of two stream-aquifer systems. Water Resour. Res., 34, 317328, doi:10.1029/97WR03285.

    • Search Google Scholar
    • Export Citation
  • Young, F. J., , C. A. Radatz, , and C. A. Marshall, 2003: Soil survey of Boone County. Missouri Natural Resources Conservation Service Rep., 318 pp.

  • Zheng, C., , and P. P. Wang, 1999: MT3DMS: A modular three-dimensional multispecies transport model for simulation of advection, dispersion, and chemical reactions of contaminants in groundwater systems; Documentation and user's guide. U.S. Army Engineer Research and Development Rep. SERDP-99-1, 169 pp.

Save