Vegetation zonation and tidal hydrology are basic attributes of intertidal salt marshes, but specific links among vegetation zonation, plant water use, and spatiotemporally dynamic hydrology have eluded thorough characterization. We developed a quantitative model of an intensively studied salt marsh field site, integrating coupled 2-D surface water and 3-D groundwater flow and zonal plant water use. Comparison of model scenarios with and without heterogeneity in (1) evapotranspiration rates and rooting depths, according to mapped vegetation zonation, and (2) sediment hydraulic properties from inferred geological heterogeneity revealed the coupled importance of both sources of ecohydrological variability at the site. Complex spatial variations in root zone pressure heads, saturations, and vertical groundwater velocities emerged in the model but only when both sources of ecohydrological variability were represented together and with tidal dynamics. These regions of distinctive root zone hydraulic conditions, caused by the intersection of vegetation and sediment spatial patterns, were termed “ecohydrological zones” (EHZ). Five EHZ emerged from different combinations of sediment hydraulic properties and evapotranspiration rates, and two EHZ emerged from local topography. Simulated pressure heads and groundwater dynamics among the EHZ were validated with field data. The model and data showed that hydraulic differences between EHZ were masked shortly after a flooding tide but again became prominent during prolonged marsh exposure. We suggest that ecohydrological zones, which reflect the combined influences of topographic, sediment, and vegetation heterogeneity and do not emphasize one influence over the others, are the fundamental spatial habitat units comprising the salt marsh ecosystem.
 Coastal salt marshes serve critical functions valuable to both natural and human systems. Marshes' complex biogeochemical role in the coastal zone has inspired particularly active research in recent years, since the nutrient outwelling hypothesis by Odum  (as discussed by Childers et al. ). However, it has been difficult to generalize dynamic land-ocean interactions, whether marshes are a source or sink for coastal zone constituents, because of the lack of a comprehensive hydrologic model accounting for intertidal infiltration, reaction, transport, discharge, and mixing of salt marsh groundwater and tidal surface water. A better understanding of salt marsh ecohydrological system dynamics is required before connections with other fields such as coastal biogeochemistry can be strengthened [Rodriguez-Iturbe et al., 2007].
 Salt marsh groundwater flow is represented in the literature by four somewhat conflicting conceptual models: (1) Flow in interior marsh sediments is entirely vertical because of alternating phases of evapotranspiration and infiltration governed by the tides [Hemond and Fifield, 1982; Dacey and Howes, 1984]. (2) Vertical flow in the marsh interior feeds deep, slow flow paths beneath the marsh that contribute to submarine groundwater discharge in the coastal zone [Wilson and Gardner, 2006]. (3) Flow is horizontal and restricted to the few meters near the channel banks affected by rapid drainage; there is no lateral flow in the marsh interior because of low sediment permeability and zero groundwater head gradient [Harvey et al., 1987; Nuttle, 1988; Montalto et al., 2007]. (4) Plant root water uptake near channel banks controls local water table position and unsaturated flow [Howes and Goehringer, 1994; Ursino et al., 2004; Wilson and Gardner, 2005; Li et al., 2005; Marani et al., 2006; Tosatto et al., 2009]. Each of these conceptual models clearly addresses marsh groundwater hydrology at a different scale, yet each discounts the possibility of lateral groundwater flow in the marsh interior and leaves open the question of the effects of different plant species on salt marsh groundwater flow [Silvestri et al., 2005].
 This study integrated salt marsh vegetation zonation, interspecific differences in plant water use, three-dimensional variably saturated groundwater hydrology, sediment heterogeneity, and tidal flooding into a numerical model of the salt marsh ecohydrological system. Our goal was to quantify the relative and combined effects of these multiple system components on marsh hydrology by comparing a complex model including all these components to simpler models that included spatial variations in only sediment properties or evapotranspiration, or neither. We also contrasted flooding and nonflooding tidal regimes.
2. Review of Literature on Intertidal Salt Marsh Groundwater Modeling
 Models of intertidal salt marsh groundwater flow have not reconciled the contrasting scales of tidal and groundwater flow and plant-water interactions thus far, since they have not accounted for the natural three-dimensionality of the intertidal system nor for spatially variable plant-water interactions. Intertidal salt marsh groundwater numerical modeling studies are summarized in Table 1. Zero-dimensional water balance models cannot represent spatial variability and multiscale hydrological processes. One-dimensional models have focused either on vertical infiltration (1-D-Z models) or on the extent of the perpendicular propagation of the harmonic tidal signal away from major tidal creeks (1-D-X models). Most salt marsh groundwater modeling has examined two-dimensional domains that are vertical slices perpendicular to major tidal channels (2-D-XZ models). Empirical investigations of groundwater flow and transport have also focused on this 2-D-XZ geometry [Chapman, 1938a; Agosta, 1985; Harvey and Odum, 1990; Howes and Goehringer, 1994; Hayden et al., 1995; Osgood and Zieman, 1998; Montalto et al., 2006]. Few models or experiments have examined lateral (2-D-XY) or three-dimensional (3-D) salt marsh groundwater hydrology. Notable exceptions are the groundwater tracer experiments by Tobias et al.  and Jordan and Correll  in 2-D-XY and 3-D, respectively, and the 3-D archetypal marsh-and-channel model by Xin et al. .
Table 1. Summary of Intertidal Salt Marsh Groundwater Flow Numerical Models
spatially variable: from ET field data from major vegetation zones
 Many marsh groundwater models have included unsaturated flow; some have included sediment layering; few have included sediment heterogeneity or macroporosity (see Table 1). Evapotranspiration has been omitted from many models or represented as a constant, spatially uniform surface flux. This flux was based on field or laboratory data in only a few cases (see Table 1). Very few models have distributed evapotranspiration demand through a finite rooting depth [Hemond and Fifield, 1982; Hughes et al., 1998; Marani et al., 2006] and none have considered spatially variable evapotranspiration, despite the characteristic zonation and water use variations of salt marsh plant species.
 No salt marsh groundwater model has yet integrated interspecific differences in plant water use, salt marsh vegetation zonation, intertidal plant-water interactions, 3-D variably saturated groundwater hydrology, and tidal flooding into a more complete ecohydrological framework for the intertidal salt marsh system. In this study, we develop such a framework via numerical simulations of an intensively studied field site.
3. Numerical Model Governing Equations
 A numerical model of fully coupled surface water and groundwater dynamics was developed using HydroGeoSphere [Therrien and Sudicky, 1996; Therrien et al., 2008]. Variably saturated groundwater flow was simulated as a 3-D boundary value problem. At the same time, 2-D, depth-averaged surface water flow was simulated in a surface domain spatially coincident with the top (ground) surface of the groundwater flow domain. Exchange between the two domains was solved for, driven by the difference between surface water and groundwater heads at the sediment surface. Capturing rapid tidal dynamics required very fine, adaptive model time steps. The computational cost of fine time steps was justified by the ability to precisely solve for the transient positions of surface water bodies and groundwater seepage faces in three dimensions, a unique feature of this salt marsh modeling study.
 Variably saturated groundwater flow was represented by Richards' equation:
Variables are the 3-D gradient (∇) of groundwater flux (q), exchange flux with the surface (Γo), and boundary fluxes (Q). Fluid mass was conserved by temporal (t) changes in sediment saturation (Sw) and fluid pressure (ψ), according to the sediment specific storage (Ss) and porosity (φ).
 Constant fluid density, fluid viscosity, and sediment porosity were assumed. Intertidal groundwater and surface water at the field site that was the focus of this investigation were both highly saline and of comparable density. Fluid density, viscosity, and saturated hydraulic conductivity values were representative of saline fluid (Table 2).
Table 2. Simulation Parameters Common to All Model Scenarios
simulated heads following 7 days of gravity drainage, as if a nonflooding (neap) tide period without tidal oscillations
Nontidal boundary conditions
zero flux through model base and model sides AB-BC (Figure 2)
 Groundwater flux (q) was calculated as the gradient of pressure (ψ) and elevation (z) heads, scaled by the unsaturated sediment hydraulic conductivity (K · kr):
where K was the saturated hydraulic conductivity tensor and kr the relative permeability given the degree of water saturation.
 Saturation-pressure (Sw(ψ)) and relative permeability–saturation (kr(Sw)) relationships were described using the Mualem–van Genuchten method [van Genuchten, 1980]:
 Variables are saturation (Sw), residual saturation (Swr), pressure head (ψ), effective saturation (Se), and empirical parameters (α and β; see Table 2 and Figure 1).
 Surface water flow was represented by the 2-D diffusion wave approximation of the Saint Venant equation for unsteady shallow water flow, assuming depth-averaged flow velocities, hydrostatic surface water heads, and negligible inertial terms:
Variables are fluid sources and sinks (Qo), exchange with the subsurface (Γo), a rill storage factor (φo: 0 at surface and 1 at top of rills), isotropic surface conductance (Ko), and the surface water head (ho) equal to the depth of water (do) above the local surface elevation (zo). Surface conductance (Ko) was defined by Manning's formula, via the direction of maximum surface water velocity gradient (s) and a friction coefficient (n, Table 2):
 The 2-D surface water domain was identical to the top (ground) surface of the 3-D unsaturated groundwater domain. The two domains were coupled by the first-order exchange coefficient (dual-node) approach. The volumetric groundwater–surface water exchange flux (Γo) is
Variables are vertical saturated subsurface hydraulic conductivity (Ksz), groundwater head (h), surface water head (ho), an exchange coefficient (lex, Table 2), and surface relative permeability (krs). The relative permeability (krs) for flow from the subsurface to the surface was given by equation 4. The relative permeability (krs) for flow from the surface to the subsurface was equal to one, permitting free infiltration.
4. Field Site and Model Setup
4.1. Hydrologic Setting
 An intensively studied field site guided the development of the numerical models. The site was in the Palo Alto Baylands Nature Preserve (37°27′54″N, 122°6′58″W), San Francisco Bay, California, USA. The central field site that was the focus of the study consists of a flat marsh plain bounded by one intertidal channel and bisected by a smaller channel (Figure 2). Local tidal harmonics are mixed semidiurnal. Low tides recede beyond the marsh to the mudflats and subtidal San Francisquito Creek channel. The higher of two daily spring tides generally overtops the tidal channels and floods the marsh plain. Neap tides may not exceed bankfull channel capacity and so may not flood the marsh for a few days at a time.
 The near-surface sediments at the site are texturally clay, with 61.82% clay, 35.54% silt, and 2.64% sand, on average (based on 23 core samples 0–30 cm depth; online supplement to Moffett et al. ). Regional sediments consist of at least 3 m of low-permeability estuarine mud overlying the uppermost of a series of alluvial aquifers interbedded with marine clay [Hamlin, 1983].
 Unlike in many conceptual models of coastal hydrogeology, there is no fresh groundwater discharge from inland to the salt marshes of Palo Alto, California. Helley and Lajoie  described four mechanisms of saline intrusion into local shallow aquifer systems, but no groundwater discharge to the near surface. At our field site, a levee separated the high intertidal marsh plain from the subsided, nontidal inland area occupied by the Palo Alto Airport. A well installed on the inland (airport) side of this levee perpetually recorded total heads lower than those in the salt marsh over a period of three years, supporting a conceptualization of the site in which there is no shallow regional groundwater flow to the salt marsh from inland.
4.2. Model Domain
 The numerical model's finite element surface mesh was constructed of ∼1 m wide triangular elements. Elements in locations of greater than 15° local topographic slope (channel banks and levee edges) were refined to ∼0.5 m wide elements. Nodes were assigned interpolated values of surface elevations. Surface topography within the central field site was represented by a kriged model of 742 surveyed surface elevations of centimeter accuracy [Moffett, 2010, Figure 6–15]. The central field site, enclosed by tidal channels and levees, had an average elevation of 1.02 ± 0.06 m above mean sea level (μ ± 1σ) over a 0.96 hectare area. Tidal channels around the central field site were located by walking the banks with a continuously recording GPS unit and channel bank heights were measured at regular intervals.
 To enable logical prescription of boundary conditions, the model domain was extended beyond the central field site to cover a total area of 2.2 hectares. The topography in the areas external to the central field site (levees, mudflat, and adjacent marsh) was based on 1 m lidar data of coarse vertical accuracy (10–20 cm [TerraPointUSA, 2005]), resulting in a rougher appearance in Figure 2. The LIDAR data were registered to the local mean sea level datum via regression with collocated survey data.
 The 3-D finite element mesh for groundwater simulation was extended from the surface to 10 m below mean sea level for a total depth of approximately 11 m beneath the marsh plain. This large model depth was chosen to ensure the influence of the bottom boundary on near-surface hydraulic dynamics would be negligible. Refined detail in the near-surface depths of interest was provided by node sheets parallel to local topography at depths of 0.1, 0.2, 0.3, 0.5, and 1.0 m. The model totaled 6 layers, 43,443 nodes, and 86,242 elements.
 The surface and subsurface flow equations were solved using adaptive model time steps. Nodal saturation changes greater than 5% and nodal head changes greater than 1 cm between time steps were prohibited to avoid overshooting the nonlinear hydraulic dynamics in the unsaturated zone during a time step. The time step duration was decreased until these incremental changes were within allowed levels.
4.3. Initial Conditions and Fixed Boundary Conditions
 Initial conditions for all simulations were the hydraulic heads resulting from a simulation of 7 days of gravity drainage initiated with the following conditions: a fully saturated domain, a spatially uniform specified head of 3 m, and a specified head boundary of −0.25 m along model sides CD-DA (see Figure 2). The initial head of 3 m above mean sea level placed about 2 m of water on the marsh surface and so ensured complete flooding and saturation prior to the development of partially drained conditions. This approach to developing initial conditions is consistent with established groundwater modeling practice [e.g., Loague and VanderKwaak, 2002; Mirus et al., 2011]. This initial simulation period may be intuitively thought of in this case as if simulating a simplified (nonoscillatory), neap tide period immediately prior to testing subsequent model scenarios. This initial simulation was conducted with homogeneous, isotropic saturated hydraulic conductivity and specific storage values, within the ranges of values appropriate for clay sediments (K = 0.204 m d−1 and Ss = 10−4 m−1 [Schaap and Leij, 2000]).
 Starting from this partially drained initial condition, we developed a suite of model scenarios to test the relative influences of tidal regime, sediment hydraulic properties, and zonally distributed variations in plant water use on the groundwater and surface water hydrology of the field site. In each simulation, no vertical groundwater flow was permitted across the model base nor horizontal flow across the model edge symmetry boundaries AB-BC (Figure 2). Other simulation parameters common to all the model scenarios are listed in Table 2.
4.4. Lateral Boundary Conditions: Tidal Scenarios
 Lateral model faces CD-DA (Figure 2) were simulated as subsurface and surface specified head boundaries. Head values were interpolated to match the adaptive time steps of the model from water levels recorded every 10 min at a temporary tide gauge at the field site on 23–24 December 2007 (Figure 3). We compared scenarios simulating a flooding or nonflooding tide because both occur frequently at the field site because of the mixed semidiurnal tidal signal. The tidal boundary condition for the flooding scenario was given by 12 h of data that included a tide that exceeded bankfull stage and flooded the marsh surface. The tidal boundary condition for the nonflooding scenario was given by 12 h of data that included a maximum tidal stage just below bankfull tidal channel capacity.
 Both tidal scenarios permitted 4.5 h of drainage following the bankfull stage of the ebb tide prior to the end of the simulation (Figure 3), at which time results were evaluated. During low-tide periods, the total heads at the tidal boundaries (locations along edges CD-DA, all with surface elevation −0.26 cm or lower) were maintained at −0.25 m to simulate at least 1 cm of persistent surface water along the boundary, as observed in the field.
4.5. Porous Media Hydraulic Properties: Sediment Scenarios
 Most salt marsh groundwater models to date have simulated a marsh of homogeneous sediments (see Table 1). In this study, we compared a simple homogeneous, isotropic sediment scenario to a scenario of heterogeneous sediment structures underlying the field site that were determined via piezometer data and numerical model sensitivity analysis. The sediments of the homogeneous scenario were isotropic with hydraulic conductivity K = 7 × 10−2 m d−1. Sensitivity analysis suggested this value was most suitable for locations that displayed substantial hydraulic responses to tidal forcing (see Appendix A). The sediments of the heterogeneous scenario were arrayed in a layered system of spatially variable hydraulic conductivity values as depicted in Figure 4a.
 The development of the heterogeneous sediment scenario shown in Figure 4a is detailed in Appendix A. In brief, groundwater head responses to tidal forcing were recorded at 14 field locations in piezometer pairs screened at depths of 0.5 m or 1.0 m below the ground surface (Figure 2). Model sensitivity analysis was conducted using multiple realizations of saturated hydraulic conductivity, specific storage, porosity, and layering, with all parameters in ranges appropriate for the clay sediments of the field site. The simulated response of groundwater heads to a flooding tide was found to be sensitive to the specific storage and the vertical hydraulic conductivity of the clay sediments but insensitive to layering, horizontal hydraulic conductivity, and porosity. The sensitivity analysis led to a conceptual model of sediment heterogeneity at the field site (Figure 4a), which was partially based on historical field site geomorphology (Figure 4b), as described further in Appendix A.
 The surface conditions of the model were varied to test the effects of evapotranspiration patterns on marsh hydrologic response. Two scenarios used common methods from other modeling studies and a third accounted for the distinctive vegetation zonation of salt marshes.
 1. For the zero ET scenario, the surface boundary condition was determined by the simulated surface water hydraulics as either a positive pressure head or a seepage face, with zero specified ET flux.
 3. In the zonal ET scenario, various evapotranspiration rates were distributed linearly over various rooting depths according to the evapotranspiration rates and vegetation zonation of dominant salt marsh plant species determined from the field site, as depicted in Figure 5. The development of Figure 5, evapotranspiration rates, and rooting depths is detailed in Appendix B.
5. Model Results
5.1. Salt Marsh Ecohydrological Zones
 Intricate spatial variations in simulated root zone hydrology were produced upon combining in one model, interspecific differences in plant water use, vegetation zonation, plant-water interactions, 3-D variably saturated groundwater hydrology, and variable degrees of 2-D tidal flooding. The intricate spatial variations in root zone hydrology did not appear in simpler model scenarios that lacked some of these components. The multifaceted model, intended to represent conditions as close as possible to those at the field site, we refer to in the remainder of this paper as the complex model scenario: it included heterogeneous sediment properties as in Figure 4a and spatially variable evapotranspiration and rooting depths as in Figure 5 for flooding or nonflooding tidal conditions as in Figure 3.
 Comparison of the complex model to simpler scenarios showed that approximately seven different root zone hydrological environments occurred at the same time in different regions of the complex model and that these different environments were caused by different local combinations of hydraulic and vegetative influences, not by either influence alone (Table 3). We termed these distinct root zone hydrological environments ecohydrological zones. Ecohydrological zones are two or more adjacent subregions of an ecosystem exhibiting substantially different hydraulic conditions in their root zones because the subregions are separated by contrasts in plant water use, hydraulic properties, or both. The hydraulic conditions noted in this study that differed substantially between ecohydrological zones were root zone pressure head, saturation, and vertical groundwater velocity. Contrasts in hydraulic properties that contributed to ecohydrological zonation in this study were provided by sediment hydraulic conductivity heterogeneity and variations in topography and tidal influence. Contrasts in vegetation water use in this study were provided by spatially variable, species-specific evapotranspiration rates and rooting depths. The remainder of this section presents the ecohydrological zonation in the models and field data due to these vegetative and geological sources of heterogeneity. Additional potential influences on ecohydrological zonation are discussed in section 6.1.
Table 3. Schematic of Causes and Consequences of Seven Ecohydrological Zones and an Eighth Hydrological State Exhibited for a Limited Time After Tidal Flooding
Causes tested in this study were tidal regime, local topography, evapotranspiration (ET) rate and rooting depth, and saturated sediment hydraulic conductivity (K). Consequences observed were pressure head (ψ), saturation (S), and vertical groundwater velocity (Vz; plus is upward, and minus is downward).
Direction of near-surface groundwater flow locally depends on relative influence of upward flow by ET extraction and downward flow by drainage.
An eighth hydrological state occurs shortly after a flood event; as a temporal, not spatial, condition and attributable to tidal flooding alone, it is not defined as an ecohydrological zone.
 The first of two topographically influenced ecohydrological zones (EHZ) was indicated by large negative pressure heads in the salt marsh channel banks (abbreviated EHZ 1). According to comparisons of different model scenarios after a nonflooding tide, the extent of this zone away from the tidal channels was sensitive to multiple local hydraulic influences. The width of EHZ 1 surrounding the central field site channel was narrow when the channel was surrounded by the relatively high conductivity sediments of the homogeneous sediment scenario (Figures 6c and 6d) and wider when surrounded by the lower conductivity sediments of the heterogeneous sediment scenario (Figures 6a and 6b). The width of EHZ 1 also varied with evapotranspiration regime: it was narrow given spatially uniform evapotranspiration (Figures 6b and 6d) and wider with spatially variable, zonal evapotranspiration and rooting depths (Figures 6a and 6c).
 A second topographically influenced ecohydrological zone was indicated by ponded infiltration in surface depressions (EHZ 2), characterized by positive pressure heads, saturated sediments, and downward groundwater flow (Table 3). The extent of this zone again appeared sensitive to both sediment and evapotranspiration variations. Ponds were less extensive in the heterogeneous sediment scenario, which included a large central area of low hydraulic conductivity (Figures 6a and 6b), compared to the broad flooded areas that remained after a nonflooding tide in the homogeneous-sediment scenario (Figures 6c and 6d). Although the low hydraulic conductivity region in the heterogeneous-sediment scenario might logically have supported perched surface water, evaporative water loss reduced the amount of water available at the surface, accounting for these results.
 Five distinct ecohydrological zones appeared due to the combined influence of sediment and vegetation heterogeneity in the complex model scenario during the period of marsh exposure after a nonflooding tide (see Table 3). Two of these EHZ emerged among sediments of relatively high hydraulic conductivity. In some areas, high evapotranspiration rates superimposed on high hydraulic conductivity sediments resulted in very large negative surficial pressure head values (Figure 6a), moderate saturation values (Figure 7a), and relatively rapid, upward unsaturated flow (Figure 7b) (EHZ 3, Figure 7c). In other areas, low ET rates among high K sediments produced smaller pressure heads (Figure 6a) and groundwater velocities (Figure 7b) (EHZ 4, Figure 7c).
 Among low hydraulic conductivity sediments, different evapotranspiration rates distinguished three ecohydrological zones. High ET among low K sediments resulted in little groundwater movement (Figure 7b) but very large negative pressure heads (Figure 6a) and saturation values below field capacity (Figure 7a) (EHZ 5, Figure 7c). The especially low pressure heads simulated in this zone were facilitated by the shape of the unsaturated characteristic curves modeled for the clay marsh sediments (Figure 1). These constitutive relationships caused a relatively large change in pressure head for a small change in saturation: the soil remained 99% saturated at a pressure head of ψ = −0.05 m, 98% saturated at ψ = −0.1 m, and 94% saturated at ψ = −0.25 m. The hydraulic influences of low K sediments and moderate ET offset each other in some locations, resulting in large negative pressure heads (Figure 6a), moderate soil saturations (Figure 7a), and root zone stagnation with approximately zero vertical root zone fluid velocity (black in Figure 7b) (EHZ 6, Figure 7c). Finally, very low ET among low K sediments induced only small negative pressure heads (Figure 6a) and hardly changed the high saturation and low downward groundwater infiltration rates remaining from prior saturated conditions (cyan in Figure 7b) (EHZ 7, Figure 7c).
 The differentiation of these distinct ecohydrological zones in the complex model scenario after a nonflooding tide was qualitatively validated by tensiometer data collected in the field. Prior to the modeling study, tensiometers attached to a stabilizing stake were pushed into the root zone to 10 cm depth (ceramic cup spanning 5 to 15 cm; locations in Figure 2) and monitored manually during the contrasting hydrological conditions of neap and spring tides. At the end of a neap tide period, during which the marsh was not flooded for 8 days, tensiometer pressure heads were lowest in EHZ 3, 5, and 6 (Figure 8a). Tensiometer pressure heads after this prolonged, neap tide exposure were slightly higher in EHZ 4, and highest in EHZ 7. These field data corroborated the ecohydrological zonation framework summarized in Table 3.
5.2. Postflooding Hydrological Conditions
 In addition to the seven ecohydrological zones identified from nonflooding tidal conditions, an eighth type of hydrological environment temporarily prevailed in the salt marsh for a limited time after tidal flooding. According to the tensiometer field data, spatial variations in near-surface pressure heads were nearly eliminated after a flooding tide (Figure 8b). According to the model simulations, tidal flooding temporarily overwhelmed any significant spatial variations in root zone hydraulic conditions that might have been induced by variations in sediment hydraulic conductivity and evapotranspiration, leaving the marsh surface in a relatively uniform state (Figure 9), as indicated by the tensiometer data. In simulations, soon (4.5 h) after a flooding tide, variations in sediment hydraulic conductivity had only a slight effect on spatial variations in surficial pressure heads in the marsh interior and variations in evapotranspiration had very little effect (Figure 9).
5.3. Temporal Development of Ecohydrological Zones Following Flooding
 Since tidal flooding masked underlying ecohydrological zonation: How do the various root zone hydraulic habitats associated with ecohydrological zonation reappear following a flooding tide? To better understand the temporal development of EHZ after flooding, we conducted a simulation of extended marsh exposure during neap tide. This extended model was based on the previous complex model scenario: it included heterogeneous sediment hydraulic properties and spatially variable evapotranspiration and rooting depths (as in Figures 4a and 5). The model spanned a 6 day period beginning with a flooding tide, followed by 5.5 days of nonflooding tides: the neap tide period recorded at the field site on 16–21 October 2007 (Figure 10a). Root zone hydraulics at six locations within each of the spatially extensive EHZ (2–7) were monitored for the duration of the simulation. Model element discretization was coarsened away from these observation regions to preserve model runtime efficiency during the extended simulation period.
 During the neap tide simulation, each of the EHZ followed a distinct trajectory in developing its characteristic hydraulic conditions. In topographic depressions (EHZ 2), the root zone remained saturated for at least two days after flooding (2.5–3 days into the simulation), despite ongoing evaporation (black lines in Figure 10). The rest of the marsh quickly developed unsaturated conditions, although via various trajectories. High K sediments were saturated by the flooding tide and then exhibited moderate rates of pressure head and saturation decline at 10 cm depth (cyan and magenta lines, EHZ 3 and 4, Figures 10b and 10d). In contrast, low K sediments maintained unsaturated conditions at 10 cm depth during surface flooding (red, blue, and green lines, EHZ 5–7, inset in Figure 10d) but eventually developed the lowest pressure heads and saturations, with their relative conditions influenced by the high, medium, or low imposed ET (Figures 10b and 10d). Deeper in the root zone, the relative conditions in low K and high K regions were reversed, with greater desaturation occurring in high K regions than in low K regions at 20 cm depth over the neap tide period (Figures 10c and 10e).
 Each EHZ's distinctive unsaturated conditions were accompanied by different shallow vertical groundwater flow regimes. In high K regions and surface depressions, there was a short spike of infiltration during the initial flooding tide (EHZ 2–4, inset in Figure 10f) but the vertical specific discharge then increased precipitously to meet the ET demand. In low K regions, the initial flooding tide did not perturb the slight upward flux to the surface induced by medium to high ET rates (EHZ 5–6, inset in Figure 10f) nor the slight downward flux at the surface in the low K, low ET zone (EHZ 7, inset in Figure 10f), though the flux direction in the latter zone (EHZ 7) reversed about half a day after flooding.
 Since spatially variable sediment properties and evapotranspiration greatly influenced local marsh hydraulics: Did they also influence cumulative groundwater–surface water exchange? To answer this question we compared two limiting case scenarios: the complex model with heterogeneous sediments and spatially variable evapotranspiration (HetK/VarET, as in Figures 4a and 5), and simple model with homogeneous sediments and no evapotranspiration (HomK/NoET). Absolute model results may have been influenced by the overestimation of infiltration by Richards' equation [Li et al., 2005; Wilson and Gardner, 2006; Tosatto et al., 2009], but comparison between models was still informative.
 Cumulative simulated surface water–groundwater exchange rates were primarily influenced by sediment hydraulic properties, with lower infiltration and exfiltration rates during high tides in the complex HetK/VarET scenario (Figures 11c–11f); evapotranspiration had no apparent effect. In contrast to the exchange rate, both sediment hydraulic properties and evapotranspiration influenced the exchanged volume and the volumetric magnitudes of infiltration and pore water flushing (Figures 11g–11j). The simple HomK/NoET scenario showed considerable pore water flushing, indicated by the high maximum infiltrated volume reached and the large net exfiltration at the end of the simulation. In contrast, the complex HetK/VarET scenario showed only modest infiltration, yet some of that infiltrated water was retained as a net positive infiltration volume at some marsh locations at the end of the tidal period, replacing water removed by evapotranspiration. Continued exfiltration after the start of tidal infiltration is expected in salt marshes if 3-D marsh geometry is accounted for [Harvey et al., 1987; Xin et al., 2011]. The rates of change of volumetric exchange were more gradual in simulations of a nonflooding tide than in simulations of a flooding tide (Figures 11h and 11j versus 11g and 11i).
 Vertical profiles of groundwater head and discharge taken from the same two model scenarios provided more detailed insight into the distribution of, and controls on, groundwater–surface water exchange (Figure 12). Groundwater discharge from the channel banks appeared anomalously large in the simple HomK/NoET scenario but more subtle in the complex HetK/VarET scenario. Four notable aspects of the shallow groundwater flow paths and groundwater discharge in the complex HetK/VarET scenario are described in the following list and labeled in Figure 12 by corresponding Roman numerals (i–iv).
 In the locations labeled i in Figure 12, depression storage (EHZ 2) supplied water to adjacent high K, high ET zones (EHZ 3).
 Groundwater discharge to tidal channels was suppressed in the locations labeled ii in Figure 12 by locally high ET, which reduced the heads around the channel (in EHZ 1, Figure 12b).
 Groundwater discharge was suppressed in the locations labeled iii in Figure 12 by low K outcrops. For example, discharge was suppressed around an island in the main tidal channel because the channel beds intersected low K sediments below 0.5 m elevation (Figure 12f). At the same time, the ET “discharge” from the top of the island was enhanced in an effect similar to that in location i in Figure 12, only in this case the water “source” was suppressed drainage. Discharge was also suppressed where low K sediments outcropped around the levee and at the shoreline in Figure 12h.
 In addition to the effects of heterogeneous sediments and ET, the influence of rooting depth on root zone hydraulics was apparent in the locations labeled iv in Figure 12. For example, strong vertical head gradients were induced in the root zone by regions of high ET and shallow (10 cm) roots among low K sediments (EHZ 5) in Figures 12d and 12h.
6.1. Ecohydrological Zonation
 This modeling study demonstrated how ecological and hydrogeological heterogeneity can combine to create diverse root zone hydraulic habitats in even a small salt marsh area, despite the largely homogeneous influence of tidal flooding on a flat marsh plain. Considering heterogeneity in either vegetation or sediments, alone, resulted in very little variability among root zone hydraulic conditions. The intricate mosaic of root zone hydraulic conditions that emerged in the complex model due to the intersection of heterogeneous vegetative and hydrological influences constituted ecohydrological zonation. Although coastal hydrology and vegetation have separately been recognized as influencing salt marsh groundwater dynamics for nearly a century [e.g., Johnson and York, 1915; Chapman, 1938a; Mahall and Park, 1976c; Hemond and Fifield, 1982; Dacey and Howes, 1984; Harvey et al., 1987; Nuttle, 1988; Howes and Goehringer, 1994], a spatially explicit conceptual model that integrates both hydrogeological and ecological influences on salt marsh groundwater dynamics has not previously been proposed, which is what is captured in the concept of ecohydrological zonation.
 Ecohydrological zonation encapsulates both visually obvious, above-ground wetland patterning and hidden, below-ground hydraulic patterning, which are each integral parts of overall wetland ecohydrology. Since this study examined only spatial variations in evapotranspiration, rooting depth, sediment hydraulic conductivity, and topography, it is likely that our definition of ecohydrological zonation (see section 5.1) captured only some components of a broader suite of intersecting spatial patterns that contribute to wetland organization. Additional relevant system attributes, the overlay of which might produce a richer picture of the three dimensional basis for salt marsh habitat complexity, include: heterogeneity in sediment storage properties, maps of macroporosity or bioturbation, root zone biogeochemical gradients (e.g., inorganic or microbial or due to root exudates), and spatial patterns in meteorological influences (e.g., prevailing wind direction, average solar angle). Whatever intersecting system attributes ultimately contribute to a full definition, we suggest that ecohydrological zones are the fundamental habitat units comprising the salt marsh ecosystem. This perspective contrasts with a century-long focus on visually obvious vegetation patterns as the major unit of habitat variation in salt marshes. The ecohydrological zonation concept also contrasts with previous conceptual models of wetlands that treat vegetation as a consequence of, not also a contributing cause of, spatial and temporal variations in the physical variables used to define the wetland hydrological system [e.g., Brinson, 1993].
 The idea of a salt marsh being segmented into many ecohydrological zones remains to be tested further with new field studies. Of the seven ecohydrological zones we identified, only one has been studied in detail: the channel bank environment (EHZ 1) [e.g., Howes et al., 1981; Howes and Goehringer, 1994; Ursino et al., 2004; Marani et al., 2006; Tosatto et al., 2009]. Howes and Goehringer  illustrated this zone's principle characteristics: substantial groundwater drainage, relatively high groundwater flow rates, and comparatively low soil saturation between high tides. Although this zone is generally thought to extend 1–2 m from the channel bank in muddy sediments [Hughes et al., 1998], our models showed that its width depends on multiple local factors including near-channel sediment hydraulic properties and evapotranspiration magnitude and distribution.
6.2. Limitations of the Models
 Although we believe ecohydrological zonation to be a generally useful concept, the details of the zonation proposed by this study are subject to some limitations. To capture the topographic, geologic, and vegetative complexity of a natural salt marsh, our models were necessarily site-specific. Much of the ecohydrological zonation that might be expected in intertidal high-marsh settings may have been represented among the many combinations of evapotranspiration rates, rooting depths, and sediment properties we tested, but the number of zones and their specific characteristics are likely to vary from site to site.
 In general, salt marshes that are slowly accreting or subsiding, such as low-energy lagoon or estuarine systems, are likely to share the fine sediment texture and low topographic relief of our field site. Reported saturated hydraulic conductivity values similar to our 0.07–0.0007 m d−1 values include an Australian estuarine marsh, 0.01–0.17 m d−1 [Hughes et al., 1998], and a Venice lagoon marsh, 0.00017 m d−1 (core) and 0.086–0.00086 m d−1 (models) [Cola et al., 2008]. Regarding topography, the use of a kriged surface in the central portion of our models did not capture microtopographic variations in surface elevation that might affect very local hydraulics. However, such microtopography is generally considered to occur at meter scales or less [Gray and Scott, 1977; Tessier et al., 2002; Morzaria-Luna et al., 2004; Varty and Zedler, 2008] and so would have been below the resolution of our finite element mesh even if such fine scale data had been available. The central field site, on which we focused, was less than one hectare in area and was physically surveyed at 742 points with centimeter accuracy. The remaining outer portions of the model that were represented using lower-accuracy lidar data were physically and hydraulically separated from the central site by large tidal channels or levees. These outer portions of the model were not used to draw conclusions and were only included to permit logical prescription of boundary conditions away from the central field site. Therefore, it is unlikely that the surface topography in these outer areas substantially influenced the results in the central field site area.
 Even our most complex model was not utterly ecohydrologically comprehensive; not all potentially relevant processes were included. As a first approximation, we represented plant water use as if the plants were anisohydric or hydrolabile: plants that continue using water at a relatively constant rate and do not exhibit water uptake limitation for extended periods. This assumption of temporally constant evapotranspiration is a good approximation for halophytes able to use osmotic regulation to extract water even at very low soil water contents [Marani et al., 2006]; it was also consistent with the constant rate of salt marsh plant water uptake observed in lysimeters [Dacey and Howes, 1984] and with most previous modeling approaches (see Table 1). In reality, each species, and even each individual plant, may experience water and/or salt limitation at a different time, to a different degree, under different conditions. The simulation of such dynamic plant-water feedbacks was beyond the scope of our preliminary models. Most of our analysis was drawn from 12 h simulation periods including only 4.5 h following the bankfull stages of ebbing high tides, during which short time periods the lack of dynamic vegetation feedback is unlikely to have substantially affected the results. However, the lack of dynamic plant-water feedbacks in the extended neap tide simulation means that the lowest pressure head and saturation values shown in Figure 10 might not actually be achieved in the field since the curves would flatten out if water limitation were reached. Still, the extended simulation, as is, provides a bound on the maximum divergence in behavior among EHZ, given no limitation of evapotranspiration.
 The likely effects of processes not accounted for in the model, such as plant water use limitation, can be qualitatively interpreted by comparing the model results in Figure 10b to the tensiometer data in Figure 8. A short time after a flooding tide, almost all the pressure head measurements in the field were in the relatively narrow 20 cm range of −40 to −60 cm (Figure 8b). Simulated pressure heads (in EHZ 3–7) remained within 20 cm of each other until about 2.5 days after the flood tide, at which time they also spanned the range −40 to −60 cm. The longer time required by the model to reach the −40 to −60 cm range might be attributed to the lack of air entrapment processes in the model: persistent entrapped air is known to help maintain low pressure heads in the field [Chapman, 1938a]. If so, the low pressure heads and saturations in our nonflooding model scenarios (Figures 6 and 7) are conservative results, as additional entrapped air would further lower these values. A longer time after a flooding tide, the pressure heads modeled in high K zones (EHZ 3–4) remained roughly consistent with field measurements, for at least a few days. Among low K sediments, however, the model developed more pronounced unsaturated conditions faster than were indicated by the field data (EHZ 5–7). The omission of plant water uptake limitation from the model likely explains the more extreme pressure head declines in the low K regions of the model, since the ability of plants to extract water from low K sediments likely does become limited in the field sometime after flooding.
6.3. Significance of Ecohydrological Zonation in Coastal Hydrology
 Are the transient and heterogeneous plant-water interactions that we have categorized as ecohydrological zones ecologically and hydrologically significant? We examine this question from three angles: first, in terms of root zone aeration; second, in terms of marsh groundwater dynamics; and third, in terms of marsh-estuary exchange.
6.3.1. Root Zone Aeration
 The potential significance of ecohydrological zonation for salt marsh root zone aeration and vegetation productivity may be understood by examining the most prominent ecohydrological zone that emerged from our simulations. This zone, EHZ 5, was characterized by very large negative surficial pressure heads, very low root zone saturations, and small upward groundwater flow in low K, high ET regions. EHZ 5 reached negative pressure heads of up to 25 cm at the marsh surface a short time after a nonflooding tide (Figure 6a), and more over a longer, neap tide period (Figure 10b), barring substantial plant water limitation. However, given the unsaturated hydraulic response of the clay soils (Figure 1), even a pressure head of just −25 cm locally lowered the marsh surface to 94% saturation, providing 6% aeration of the near-surface root zone pore volume.
 Although 6% aeration might be insignificant in terrestrial environments of drier, coarser soils, it is very significant in intertidal settings. Salt marsh soils may have a field capacity of at least 97.1% [e.g., Bradley and Morris, 1990], so only about 3% soil aeration could be caused by gravity drainage over a few days. The field capacity of our simulated clay soils was over 98%, so soil aeration totaling 6% of the pore volume indicates at least two times more loss of water from the root zone by evapotranspiration (≥4%) than by drainage (≤2%). The capacity for physical drainage is even smaller in some marshes: e.g., a minimum residual saturation of 97.3% to 98.5% (from specific yield of 2–3% and porosity of 75–90% [Dacey and Howes, 1984]). Thus, the 6% aeration we simulated in regions of high evapotranspiration and low soil hydraulic conductivity (EHZ 5) is quite significant relative to prior studies, especially given that infiltration may have been overestimated by Richards' equation [Li et al., 2005; Wilson and Gardner, 2006; Tosatto et al., 2009].
 A related, interesting result from our simulations was that the plant species Spartina foliosa and Salicornia virginica caused the same degree of surficial soil aeration despite being simulated with different evapotranspiration rates (5.6 and 3.6 mm d−1) and different rooting depths (20 and 10 cm). At the marsh surface, the simulated hydraulics of this grass and succulent were indistinguishable, leading to them being combined in the same ecohydrological zone (EHZ 3 if in low K sediments; EHZ 5 if in high K sediments). As much as 22% of total soil aeration has been attributed to the grass species Spartina alterniflora growing in South Carolina salt marshes [Morris and Whiting, 1985], but the same phenomenon has not previously been suggested to occur in the root zone of the succulent Salicornia virginica.
6.3.2. Marsh Groundwater Dynamics
 The potential significance of ecohydrological zonation for salt marsh groundwater movement, and possible redistribution of nutrients within the marsh, is notable from the surprisingly complex groundwater flow patterns induced by the EHZ mosaic. In some areas, the spatial juxtaposition of EHZ's contrasting root zone hydraulic conditions appeared to induce novel horizontal water movement. Because our marsh plain was topographically nearly flat, the spatial variation in surface pressure head in Figure 6a was almost identical to the spatial variation in total hydraulic head (not shown). The largest potential for horizontal flow between EHZ occurred between EHZ 3 and 4 (Figure 6a, complex model, nonflooding scenario), where the instantaneous specific discharge rate at the end of the short simulation was as high as 0.04 cm d−1 near the marsh surface. Although small, this potential vegetation-induced flow was not insignificant compared to potential horizontal flow induced by topography. Topography induced 1–2 cm d−1 instantaneous specific discharge at the marsh surface at the edge of the ponded infiltration regions (edges of EHZ 2) and 8 cm d−1 near the channel banks (in EHZ 1). Evapotranspiration-induced advection toward strongly transpiring vegetation was proposed by Mann and Wetzel , who tested the concept in microcosms with partial success. Harvey et al. , however, noted the opposite effect: infiltration was enhanced in, and pore water moved away from, small hummocks occupied by vegetation, primarily because of hummock macropores. A consistent conceptual model remains to be developed, but our results suggest that the relative applicability of these two prior conclusions depends on the local balance of the influences of evapotranspiration, sediment hydraulic properties, and microtopography. Any vegetation-induced horizontal groundwater flow in clay sediments is bound to be small and further complicated by transient, oscillatory tidal influences. However, we hypothesize that even a small amount of net fluid transport across vegetation zone boundaries (e.g., between the high ET Salicornia or Spartina of EHZ 3 and the low ET Distichlis of EHZ 4) could be biogeochemically and ecologically significant in the long term, at the temporal scale of stable salt marsh vegetation configurations sometimes spanning decades or centuries.
 The models revealed a second phenomenon on the marsh plain that may have important biogeochemical consequences: local stagnation of vertical groundwater flow in the root zone. Stagnation was induced in low K, moderate ET regions by the balance between downward drainage potential and upward ET potential (EHZ 6; Figures 7b and 7c). Stagnation was also induced along boundaries between vegetation zones simulated with different water uptake rates, if those boundaries were among low conductivity sediments (black lines in Figure 7b). Although such stagnation is likely to be a transient condition, even oscillatory root zone pore water flow exhibiting occasional stagnation periods would lessen the local groundwater–surface water exchange magnitude and might strongly influence root zone oxygen and nutrient balances.
6.3.3. Marsh-Estuary Exchange
 The potential significance of ecohydrological zonation in terms of marsh-estuary exchange is based on the premise that marsh groundwater is biogeochemically distinct from coastal surface waters. Salt marshes have long been conceptualized as net suppliers of dissolved and particulate constituents to coastal waters, with perhaps up to half of the supply from groundwater seepage [Jordan and Correll, 1985; Childers et al., 2000]. In our complex model simulations, which included sediment and evapotranspiration heterogeneity, groundwater discharge occurred mainly via the tidal channel network, not via the marsh-estuary shoreline. However, this phenomenon of discharge focused in tidal channels may not apply to all settings: some coastlines include much greater inland groundwater heads than at our field site, or more hydraulically conductive mudflat sediments, which may permit substantially more direct shoreline seepage.
 This study combined salt marsh vegetation zonation, interspecific differences in plant water use, 3-D variably saturated groundwater hydrology, sediment heterogeneity, and transient tidal flooding of different magnitudes in complex models of the salt marsh ecohydrological system. The models demonstrated that superimposed patterns of differential plant water use and sediment hydraulic properties can produce a surprisingly complex mosaic of salt marsh root zone hydraulic environments, despite the largely homogeneous influence of tidal flooding on a flat marsh plain. These substantially different root zone hydraulic environments, or ecohydrological zones, are caused by spatial differences in plant water use, sediment hydraulic properties, and possibly by additional system attributes not examined in this study. On the basis of the ecohydrological zonation observed in our models and field data, we highlight three conclusions.
 1. The well-studied tidal channel bank environment is hydrologically distinct from the less well understood marsh interior. Even so, groundwater dynamics in channel banks may be overemphasized in models not accounting for sediment heterogeneity.
 2. The marsh interior is not a large homogeneous zone of predominantly stagnant groundwater: it may be conceptually divided into distinct ecohydrological zones, each with particular ecohydrological characteristics. This zonation also creates potential for lateral, vegetation-induced exchange among zones.
 3. Tidal flooding temporarily masks the ecohydrological zones' mosaic of diverse root zone hydraulic habitats, but the zones, established during low and neap tides, may serve to define salt marsh ecosystem organization at least as much as variations in high tides' influence.
 The central theme of this study recognized that, although the tides, channels, and vegetation are the most visually prominent features of coastal salt marshes, the interactions between tides and vegetation occur most directly through the roots and root zone. The root zone sediments influence tidal infiltration and root water uptake and govern the mass transfer between these processes [Chapman, 1938b]. This perspective, focused on the root zone and its abiotic and biotic hydraulic properties, partially reconciles existing conceptual models of the salt marsh groundwater system as vertical flow due entirely to evapotranspiration and infiltration [Hemond and Fifield, 1982]; slow downward flow in the marsh interior feeding deep submarine groundwater discharge [Wilson and Gardner, 2006]; horizontal flow restricted to the channel banks, with no interior marsh flow [Harvey et al., 1987; Nuttle, 1988; Montalto et al., 2007]; or plant water uptake near channel banks controlling local water table position and unsaturated flow [Howes and Goehringer, 1994; Ursino et al., 2004; Wilson and Gardner, 2005; Li et al., 2005; Marani et al., 2006; Tosatto et al., 2009]. On the basis of the results of our detailed 3-D simulations, we conclude that each of these mechanisms is simultaneously significant in situ in different spatial regions of even a small marsh site and these different spatial regions can be usefully thought of as distinct ecohydrological zones.
Appendix A:: Development of Heterogeneous Sediment Model
 Numerical model sensitivity analysis was used to estimate the sediment structures underlying the field site. Groundwater head responses to tidal forcing were recorded in pairs of piezometers installed to depths of 0.5 m and 1.0 m at 14 field locations (Figure 2). Piezometers were 2 inch plastic pipe with 15 cm of slotted screen above an end cap; casings were tall enough to not be submerged by high tides. Clean, well-sorted sand filled the auger hole around the screen, below native clay material backfill. Piezometers and the ground surface were surveyed by total station. Piezometer and tidal channel water levels were logged by pressure transducers every 10 min over portions of 4 years (Dataflow Systems Pty Ltd, Christchurch, New Zealand). The principle data used in this study were from a typical spring tide day, 23–24 December 2007, chosen to provide 24 h of clean data at as many logged piezometers as possible.
 Simulated porous medium properties were initially based on literature values and laboratory measurements and were then refined on the basis of field measurements and sensitivity analysis. Falling-head laboratory tests of two clay cores from the field site yielded similar results: K = 7 × 10−4 m d−1 [Moffett et al., 2008]. Sensitivity analysis was conducted using multiple values of K, specific storage (Ss), porosity (ϕ), and layering, with parameters in ranges appropriate for the clay sediments of the site. Observation points simulated in the model corresponded to the locations of the 28 field piezometers.
 The simulated response of groundwater heads to a flooding tide was found to be sensitive to Ss and to the vertical hydraulic conductivity (Kz) of the sediments. The response was insensitive to layering, to the horizontal hydraulic conductivity (Kh), and to ϕ. Example results are provided in Figures A1a–A1c and additional results in chapter 6 by Moffett . Model sensitivities to Ss and Kz were comparable (Figures A1a and A1b). Since the plausible values of K for clays span a larger relative range than Ss for clays (4 orders of magnitude compared to 2 [Smith and Wheatcraft, 1993]), we adjusted values of Kz, rather than Ss, obtaining a scenario similar to field observations by narrowing the constraints relative to the range of plausible values.
 A conceptual model of a heterogeneous, layered K field (Figure 4a) was developed: (1) The best fit Kz value for each of the 28 observation points was identified and assigned locally as an isotropic K value (since simulations were insensitive to Kh). (2) A three-layer system was constructed from these 28 scattered K values and is described as follows.
 1. Shallow sediments, down to about 0.5 depth (to 0.5 m above mean sea level, MSL), were inferred from the 0.5 m deep piezometer data. Shallow piezometers with low K values generally fell within a high-elevation area of the central site (Figure 2), which was used to guide the demarcation of a near-surface low K region (Figure 4a).
 2. Deeper sediments, about 0.5 to 1 m depth (0.5 to 0 m above MSL), were inferred from the 1.0 m deep piezometer data. Deep piezometers with low K values surrounded the tidal channel bisecting the field site. It is likely that this bisecting channel is the remnant of a previously much larger tidal channel visible in geographic surveys and aerial imagery of the site from 1857, 1897, and 1921 [Hermstad et al., 2009], but now mostly filled with estuarine sediment (Figure 4b). The channel arc was used as a guide in delineating the shape of a inferred low K region from about 0.5 to 1 m depth.
 3. The bottom layer of sediments, to the base of the model, was assigned a low K value (7 × 10−4 m d−1) on the basis of a conceptual model of increasing sediment compaction and decreasing numbers of burrow and root macropores with depth. The levees surrounding two sides of the central field site in the model were also assigned this low K value to account for sediment compaction during construction.
 Simulations using this heterogeneous sediment conceptual model (Figure 4a) successfully reproduced the observed responses of most of the 0.5 m deep piezometers to tidal forcing (Figure A1d). To not overspecify the heterogeneity of the model to fit the field data, K values in a few small areas were maintained as similar to their surroundings, resulting in some small model data discrepancies, e.g., at location XS (Figure A1d).
Appendix B:: Development of Heterogeneous Evapotranspiration Model
 To develop a conceptual model of zonally distributed evapotranspiration rates and rooting depths, we began with a map of the field site vegetation from a prior study [Moffett et al., 2010]. The relative magnitudes of evapotranspiration (ET) of the dominant species under the conditions prevailing at the field site were obtained in a four-step thermal remote sensing–based procedure [Moffett and Gorelick, 2012]. (1) A diurnal series of time-lapse thermal images of vegetation surface temperature was collected over a patch of each dominant species (Spartina foliosa, Salicornia virginica, and Distichlis spicata). (2) Transpiration was estimated by applying the canopy-level model of Jarvis and McNaughton  to the canopy temperature data, also using measurements of net radiation, ground heat flux, and meteorological conditions collected on the imaging days at the field site; only a portion of the net radiation was partitioned to the canopy, according to a canopy radiation extinction factor. (3) Soil evaporation was estimated using the Priestley and Taylor  model based on the remainder of the net radiation. (4) An average of the combined daytime evaporation and transpiration rates was calculated from each species' simulated diurnal ET time series.
 Most of the field site was dominated by Spartina foliosa, Salicornia virginica, or Distichlis spicata alone and these vegetation zones were assigned their respective ET rates. Some areas apparently codominated by Salicornia and Distichlis were assigned the average of the two rates; small adjacent zones dominated by Frankenia salina or Jaumea carnosa were also included in this generic class. Other small, fragmented zones were amalgamated into larger surrounding zones. The levees around the site were largely covered in Distichlis growing in dry soils, approximated as having half the marsh Distichlis ET rate. Remaining portions of the model external to the central marsh site (adjacent marsh, mudflat) were assigned the generic rate 4 mm d−1 (Figure 5). The ET rates were simulated as constant flux values.
 This work was supported by National Science Foundation grant EAR-0634709 to Stanford University. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. We thank the City of Palo Alto Baylands Nature Preserve for permission to conduct the field studies and numerous colleagues for assistance with field installation, surveying, and monitoring. We thank the San Francisco Estuary Institute for sharing with us the historical coastlines, aerial photography, and lidar data.