Six research monitoring wells (3.2 or 5.1 cm nominal outside diameter) with 60 cm screens were emplaced in the surficial aquifer at a depth of ~2 m below the ground surface. Shallow drivepoints (0.95 to 1.9 cm nominal outside diameter) with 1 to 5 cm screens were emplaced in peat and in the underlying organic/marl/sand transitional sediments to depths ranging from 0.1 to 2 m. More details on well and drivepoint construction, installation procedures, exact locations of instruments, and measurement techniques are given in Harvey et al. (2000).

Hydraulic Conductivity and Recharge/Discharge Estimates

Hydraulic conductivity of the sand and limestone aquifer beneath WCA-2A and ENR was estimated by a previous study (Harvey et al. 2000). Hydraulic conductivities of Everglades' peat and the organic/marl/sandy sediments immediately underlying the peat were determined as part of the present study using either a constant-head, pump-out method (Tavenas et al. 1990; Brand and Premchitt 1980) or a bail-test method (Luthin and Kirkham 1949) in piezometers. Slug tests are usually thought to be more sensitive to horizontal hydraulic conductivity, which could bias results if used to compute a vertical flux. We used our slug test results as direct estimates of vertical hydraulic conductivity because, to our knowledge, a more reliable method has not been demonstrated for peat. An alternative approach to estimating vertical hydraulic conductivity, i.e., applying corrections to horizontal hydraulic conductivities in the manner typically used in granular sediments, was judged to be inappropriate for peat, because those approaches do not consider special features such as the possibility of preferred flowpaths created by vertical growth of roots. To be consistent with units of reporting recharge and discharge estimates, all hydraulic conductivity values used in the present study are reported in units of cm/day.

Seepage meters were used to obtain direct estimates of vertical water fluxes (i.e., recharge and discharge) through the peat at sites in ENR. Our seepage meters were designed similarly to the Lee-type meter (Lee 1977), but instead of cutoff drums, we built seepage meters from 0.64 cm thick, high-density polyethylene (HDPE ) by creating rings (76 cm diameter by 30 cm wall height), attaching an HDPE conical dome, and installing a PVC bulkhead fitting with a 1.9 cm port on top of the dome for quick connections and disconnections of a prefilled seepage bag. Simultaneous seepagemeter measurements using replicate meters at a single site had an average uncertainty of ± 50%. A more limited data set on recharge and discharge was collected using seepage meters at sites WCA-2BS (Figure 3b) and WCA-3A-15 (Figure 1). More detailed information on seepage meters, including emplacement and operation, and precision and limit of detection, are given in Harvey et al. (2000).

Limited access to more remote wetland sites in WCA-2A required a method other than seepage meters to estimate vertical recharge and discharge fluxes. Daily-averaged surface and ground water level measurements were combined with estimates of peat hydraulic conductivity to compute recharge and discharge in WCA-2A. Calculations were made by multiplying the average hydraulic conductivity of peat at a site by the vertical hydraulic gradient measured at that site. The vertical gradient was estimated as the difference between the surface water stage and the ground water elevation in the shallowest monitoring well (~2 m below the peat surface). For the denominator in the hydraulic gradient, the thickness of peat was used. This assumed that head changed linearly through the peat and that the head measurement in the well was a good estimate of head at the base of the peat. The sign convention for fluxes was a positive flux and negative hydraulic gradient when discharge occurred (i.e., upward flow from ground water to surface water), and a negative flux and positive hydraulic gradient when recharge occurred (i.e., downward flow from surface water to ground water). Our calculations further assumed that head changes in surface water or ground water were rapidly transmitted through the peat without significant time lag, thus maintaining the linear head distribution assumed by Darcy's law. That approximation was justified by the relative timescales involved, i.e., the timescale for pressure propagation through the peat (minutes) compared with a timescale of days to weeks for changing surface water levels (which control head at the peat surface). We estimated the characteristic time of pressure propagation through the peat, t_p, using the equation t_p = 1/2 X L² X S_s/K_peat, where L is the approximate thickness of the restricting layer (1.5 m), S_s is the specific storativity of peat (0.001/m), and K_peat is the approximate hydraulic conductivity of the restricting layer (0.3 m/day) (Thibodeaux 1996).

Hydrogeologic Simulation

Factors affecting recharge and discharge were examined using a simple hydrogeologic model of ground water flow for a leaky aquifer overlain with a thin aquitard adjacent to a canal. Barlow and Moench (1998) provided a solution to the problem based on a one-dimensional (horizontal) flow assumption through the aquifer with uniform hydrogeologic properties, and with vertical leakage across the aquitard (envisioned as Everglades' peat in our case). Because of one-dimensional flow, the head at the left boundary of the aquifer (in contact with the canal) was equal to the canal water level and was constant with depth. That boundary condition represented the hypothetical situation where the canal fully penetrated the aquifer, which was not the situation in WCA-2A where the surficial aquifer was ~60 m thick and the canal was ~4 m in depth. Despite the fact that the canal only partially penetrates the aquifer, the fully penetrating assumption was judged sufficient for our purposes because of its simplicity and because it successfully has been used in the past as a first order approximation of boundary conditions for situations that were in reality more complicated. The governing equations for the hydrogeologic model are as follows:

2h	=	Ss	h	+ q'         (1)
x2		Kx	t

2h'	=	S's	h'	for the domain b < z < (b + b')         (2)
z2		K'	t

q' = -	K'	(	h'	)	z=b	(3)
	Kxb		z

where h and h' are hydraulic heads in the aquifer and aquitard (peat) (m), respectively; x is horizontal distance from a canal boundary (m) in the domain x_o < x < ; S_s and S'_s are specific storage of the aquifer and aquitard, respectively (1/m); K_x and K' are the horizontal hydraulic conductivity of the aquifer and vertical hydraulic conductivity of aquitard, respectively (m/s); z is vertical distance (m); b and b' are the thicknesses of the aquifer and aquitard, respectively (m); and q' is the volumetric flux to or from the aquifer per unit volume of aquifer divided by the aquifer hydraulic conductivity (Barlow and Moench 1998). The initial conditions for the model are

h(x, 0) = h_i

h'(z, 0) = h_i

where h_i is the initial head in the aquifer. Boundary conditions in the aquifer are

h(0, t) = h_o

h(, t) = h_i

where h_o is the new head at the canal-aquifer interface achieved after an instantaneous step change. Boundary conditions in the aquitard are

h' (x, z = b, t) = h(x,t)

h' (x, z = b + b', t) = h_i

An analytical solution for these equations exists, but we chose to solve our problem using the numerical code called STLK1 provided by Barlow and Moench (1998).

The simple hydrogeologic simulation described here was used to try to isolate the effect of the levee boundary on discharge in the wetland. The stress applied to the model was a sudden 1 m increase in head at the left boundary (representing an increase in the water level of a canal that is separated from the wetland by a levee). The model ignored other possible influences, such as climatic factors, surface water pumping, and operation of water-control structures, as well as the slight slope of the wetland ground surface and water level surface. Assuming that the surface water level in the wetland is constant presumes quick drainage away from the levee of recently discharged ground water. Constant surface water level was implemented using the source bed option of the STLK1 model, which holds the hydraulic head constant at the top of the restricting layer. Aquifer and restricting layer thicknesses, aquifer hydraulic conductivity, and head change at the boundary were set on the basis of field estimates and held constant for all simulations. Hydraulic conductivity of the restricting layer (peat) was initially set to 30 cm/day, an intermediate value of K_peat that was representative of vertical hydraulic conductivity in both ENR and WCA-2A. A value of specific storage for both peat and aquifer (0.001 m) was selected based on literature values (Anderson and Woessner 1992). Other parameter values used in model simulations were aquifer depth 60 m, peat depth 1 m, and K_aquifer 3000 cm/day. For the purpose of testing sensitivity to the value of K_peat, two additional simulations were run using values of K_peat = 0.3 and 3000 cm/day, respectively.