You are on page 1of 10

Advances in Water Resources 112 (2018) 214–223

Contents lists available at ScienceDirect

Advances in Water Resources


journal homepage: www.elsevier.com/locate/advwatres

Stochastic modeling of wetland-groundwater systems T


⁎,a,b a,c d e
Leonardo Enrico Bertassello , P. Suresh C. Rao , Jeryang Park , James W. Jawitz ,
Gianluca Botterb
a
Lyles School of Civil Engineering, Purdue University, West Lafayette, IN 47907-2051, USA
b
Department of Civil, Architectural and Environmental Engineering, University of Padua, Padua I-35100, Italy
c
Agronomy Department, Purdue University, West Lafayette, IN 47907-2054, USA
d
School of Urban and Civil Engineering, Hongik University, Seoul, 121-791, South Korea
e
Soil and Water Science Department, University of Florida, Gainesville, FL 32611, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Modeling and data analyses were used in this study to examine the temporal hydrological variability in geo-
Crossing time graphically isolated wetlands (GIWs), as influenced by hydrologic connectivity to shallow groundwater, wetland
Probability density function bathymetry, and subject to stochastic hydro-climatic forcing. We examined the general case of GIWs coupled to
Eco-hydrology shallow groundwater through exfiltration or infiltration across wetland bottom. We also examined limiting case
Hydrologic regimes
with the wetland stage as the local expression of the shallow groundwater. We derive analytical expressions for
Hydro-period
the steady-state probability density functions (pdfs) for wetland water storage and stage using few, scaled,
physically-based parameters. In addition, we analyze the hydrologic crossing time properties of wetland stage,
and the dependence of the mean hydroperiod on climatic and wetland morphologic attributes. Our analyses
show that it is crucial to account for shallow groundwater connectivity to fully understand the hydrologic dy-
namics in wetlands. The application of the model to two different case studies in Florida, jointly with a detailed
sensitivity analysis, allowed us to identify the main drivers of hydrologic dynamics in GIWs under different
climate and morphologic conditions.

1. Introduction are embedded into the landscape in various spatial patterns, and not
permanently connected to stream networks. Isolated wetlands can in-
Wetlands are key hydrological elements of landscapes and their teract through permanent or episodic connections to regional shallow
spatiotemporal dynamics strongly influence surface water storage, groundwater, depending on the hydro-geologic setting of the specific
runoff, groundwater recharge/discharge processes, and evapo- landscape. We do not consider inter-connectivity among wetlands
transpiration (Rains et al., 2016). They also play a key role in de- during flooding events, when ephemeral stream networks are formed.
termining the temporal variability of catchment biogeochemical pro- Our motivation for exploring stochastic modeling approaches to
cesses (e.g., nutrient and sediment accretion, transformations, etc.) dynamics of isolated wetlands is based on related research on ground-
(Cheng and Basu, 2017; Marton et al., 2015) and ecological processes water-dependent ecosystems (Laio et al., 2009; Tamea et al., 2009) and
(e.g., aquatic habitat suitability and integrity for unique flora and our recent work by Park et al. (2014). Our overall goal is to develop
fauna) (Euliss et al., 2008). generalized analytical solutions, defined by few key parameters, for
Of the diverse types of inland freshwater wetlands, our interest here quantifying long-term dynamics of wetlands representing a range of
is to examine geographically isolated wetlands (GIWs), defined as wet- likely scenarios for linkage between the GIWs and surrounding uplands
lands that are completely surrounded by upland. While there may be no (vadose zone) or shallow groundwater located near the wetland
accepted scientific definition of an isolated wetland, it is clear from the bottom. When a sufficiently thick low-permeability layer occurs along
hydrologic and ecological literature (Leibowitz, 2015; Tiner, 2003) that the pool bottom, as a result of accretion of fine sediment or organic
interactions between such wetlands and other water bodies can indeed material or presence of confining units (e.g., clay), the wetland bottom
occur. Here, we use the term GIWs to simply represent single depres- regulates water exchange between wetland and groundwater
sions with topographically delineated boundaries, which form natural (Rains et al., 2016) (Section 2.3). As the hydrologic resistivity of the
contributing area, for surface water flow to each wetland. Such GIWs wetland bottom decreases, the pool water can readily exfiltrate to


Corresponding author at: Lyles School of Civil Engineering, Purdue University, West Lafayette, IN 47907-2051, USA.
E-mail address: lbertass@purdue.edu (L.E. Bertassello).

https://doi.org/10.1016/j.advwatres.2017.12.007
Received 27 January 2017; Received in revised form 12 September 2017; Accepted 8 December 2017
Available online 13 December 2017
0309-1708/ © 2017 Elsevier Ltd. All rights reserved.
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

surrounding uplands, and exchange with shallow groundwater in re- or imbalance between water inputs and water losses. The configuration
sponse to the hydrologic potential gradients (Section 2.4). Conversely, of the model is depicted in Fig. 1. Water inputs are precipitation over
as the hydrologic resistivity of the wetland bottom increases, the wet- the contributing area of the wetland and groundwater infiltration.
land pool dynamics become increasingly independent of the shallow Water losses from the wetland are dominated by evapotranspiration
groundwater, and upland as well (Supplementary Material, Section S- (ET) and exfiltration. This continuous exchange of water between
1). shallow groundwater and GIW requires evaluation of the coupled dy-
The approach represents an extension of Park et al. (2014), where a namics of the wetland-groundwater system.
stochastic model was presented for GIWs connected to uplands, but In our analysis, shallow groundwater was modeled as a second
connectivity to groundwater was not explicitly considered. In that storage, subject to inputs from ET-censored rainfall, and losses modu-
work, groundwater was not considered as a source of input into the lated through first-order hydrologic recession.
wetland, and water losses from the wetland were only driven by eva- On the basis of the foregoing assumptions, the water balance
potranspiration and exfiltration as a function of the water storage vo- equations for GIW coupled with shallow groundwater can be written as:
lume in the wetland. Here, we extend this approach to explicitly ac-
dW
count for the effect of a dynamic water table (Section 2.3). = ξ (λp , αp Ac ) − ETA (W ) − K1 [h (W ) − z ] S (W )
dt (1)
We provide analytical solutions for the probability density functions
(pdfs) of the volume and stage of a wetland where shallow groundwater dWgw
= ξ (λ, αp Agw ) − K2 Wgw
dynamics influence wetland stage fluctuations. The analytical pdfs we dt (2)
present here are completely described by five non-dimensional para-
In Eqs. (1) and (2), W is the water volume stored in the wetland and
meters defining the hydro-climatic forcing, wetland bathymetry and
Wgw the water volume stored in shallow groundwater. The first term on
hydrologic interactions with the surrounding landscape (upland,
the right-hand side of Eqs. (1) and (2) represents the stochastic water
shallow groundwater). These analytical pdfs enable direct comparison
inputs due to a sequence of rainfall (effective rainfall) events. In par-
between coupled wetland-groundwater systems with widely varying
ticular, following Rodriguez-Iturbe et al. and Botter et al. (2007) the
characteristics. We also analyze the crossing time properties of wetland
term ξ(λp, αpAc) is modeled as a stationary Poisson process where the
stage (volume) to understand how hydroperiod varies according to
events occur at random times with mean frequency λp [T −1], and
multiple factors (Tamea et al., 2011). Hydrologic thresholds are im-
random, exponentially distributed depths, with mean αp [L] over the
portant for the ecological and habitat suitability for a variety of adapted
wetland contributing area Ac. The term ξ(λ, αpAgw) represents the sto-
aquatic fauna and flora (Cronk and Fennessy, 2001; Snodgrass et al.,
chastic water input from the root zone to the groundwater due to a
2000). Therefore, a better understanding of the main processes af-
sequence of censored rainfall events, over the groundwater contributing
fecting the patterns of exposure and inundation along wetland transects
area Agw (Porporato et al., 2004). Effective rainfall events take place
may bring important implications for ecological processes (Altermatt
with frequency λ < λp and are also represented by a marked Poisson
et al., 2009; Laio et al., 2009; Tamea et al., 2009). The coupling of such
process (Botter et al., 2007).
considerations can represent an important step toward a new method
The frequency λ of these censored episodic events produced by the
for wetland classification, depending on the climatic conditions, the
infiltrating precipitation can be expressed in terms of the underlying
type of landscape, the geometry of the wetlands, and their interaction
rainfall, soil and vegetation properties (Botter et al., 2007; Porporato
with the hydrological elements of the watershed. Such attributes are
et al., 2004). In this study, we follow the model proposed by Porporato
embedded in the dimensionless parameters that determine the shapes of
et al. (2004):
the analytical expressions of the derived pdf and the mean crossing
times. DI γ γ / DI exp(−γ )
λ = λp
γ Γ(γ / DI , γ ) (3)
2. Coupled stochastic modeling of wetland-groundwater systems where DI is Budyko dryness index (Budyko, 1974) and represents the
ratio between the mean potential evapotranspiration ⟨ET⟩ and the long-
2.1. Modeling framework term mean of rainfall ⟨P⟩, given by the product, αpλp. Moreover, γ
defines the ratio between the soil-water storage capacity nZr (θfc − θw )
We conceptualize a GIW as a depression in the landscape within and the mean rainfall depth αp, where n is the porosity of the soil, Zr the
which water can accumulate from a well delineated contributing area root zone depth, and θfc and θw representing respectively, the field
and with its bathymetry defined by the landscape topography. The capacity and the wilting point. In Eq. (3), Γ(γ/DI, γ) represents the
changes in GIW water storage are completely described by the balance lower incomplete gamma function of parameters γ/DI and γ.

Fig. 1. Schematic representation of the wetland-


groundwater coupled model. The water inputs are
represented by the rainfall and water losses are
driven by the ET from wetland surface. The exchange
term can be either an additional input (infiltration
from groundwater to wetland) or an additional loss
(exfiltration from wetland to shallow groundwater.

215
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

The second term on the right-hand side of Eq. (1) stands for the Brooks and Hayashi (2002), and modified by Park et al. (2014) for
volumetric water loss through evapotranspiration. In particular ET [L/ normalized variables (W*, A*, and h*) to simplify the derivation of the
T] represents the temporally averaged ET rate, A(W) [L2] the water pdf:
surface area which changes over time as a function of water volume. In
A* = (h*) β /1 − β (6)
Eq. (2), ET losses are accounted for in the censoring of the rain fre-
quency from λp to λ (Eq. (3)). W * = (h*)1/1 − β (7)
The second term on the right hand side of Eq. (2) expresses shallow
groundwater losses due to hydrologic recession dynamics. Accordingly, where A* = A/ Amax , W * = W / Wmax and h* = h/ h max . Amax , Wmax and
K2 [1/T] represents the linearized rate for shallow groundwater reces- hmax are respectively the maximum value for area, volume and stage. β
sion. The last term on the right hand side of Eq. (1), K1 [h (W ) − z ] S (W ), is an empirical parameter that determines the shape of wetland slope.
describes the rate of water exchange between wetland and shallow While the theoretical range of β is (0,1), according to field survey data
groundwater through a wetted surface area, S(W). This process is presented in Hayashi and van der (2000) and Brooks and
modeled following a generalized Darcy’s law, where the hydraulic Hayashi (2002), it generally ranges from 0.20 to 0.65.
gradient is defined by the difference between the wetland pool level (h) As shown in Eq. (1), the wetland-groundwater exchange is modu-
and groundwater stage (z), and K1 [1/T] is the recession of the wetland lated by the wetted area in the pool Sw, which is calculated from the
bottom layer, which depends on hydraulic conductivity and the mean integration of the circular perimeter of the wetland with radius rw along
thickness of the bottom layer. This gradient term can be positive or its depth h. The wetted area in the pool can be expressed as:
negative, depending on the wetland and groundwater levels. When the h 4 π Wmax
groundwater depth, z, is below the pool stage in the wetland, h, the Sw = ∫0 2πrw dh =
(2 − β ) Amax
(W *)(1 − β /2)
(8)
term represents exfiltration from the wetland to upland and ground-
water. When groundwater depth is higher than the pool level, the ex- Eq. (8) shows that the wetted area in the pool Sw is a non-linear function
change term represents the infiltration from the groundwater into the of normalized water volume W*.
wetland. When the sediment layer is completely impermeable (K1=0),
there is no exchange with groundwater. In this case, pool stage dy- 2.3. Steady-state pdf for wetland-groundwater systems
namics are defined only by the rainfall and the ET. Since the goal of the
paper is to investigate the impact of the groundwater on wetland hy- The goal here is to determine a form for the analytical pdfs for the
drology, the solution for this scenario is presented in Supplementary water volume inside the wetland. Methodologically, the dynamics of
Material (Section S-1). the coupled bivariate non-linear system, Eqs. (1), (2) and (4), can be
The exchange term, K1 [h (W ) − z ] S (W ), should appear also in solved only numerically. In order to derive analytical expression for the
Eq. (2); however, since the groundwater volume is much larger com- pdf, we considered the groundwater level as fixed around its long-term
pared to that of a single wetland (Wgw ≫ W), the volume of water ex- mean level ⟨z⟩, which can be either above or below wetland bottom,
changed between the two hydrologic storages can be neglected in thus exerting hydrologic influence on pool dynamics. Thus, influence,
Eq. (2). in the mean, of shallow groundwater on wetland stage dynamics is
The solution of Eq. (1) requires an explicit link between the volume considered, but temporal variability of such effects are not considered
of groundwater and the water table z, which is identified as: in the analytical model. In order to derive an analytical expression for
⟨z⟩, the expectation of the mass balance Eq. (2) is taken, ending up
Wgw
z= with:
Agw n (z ) (4)
λαp Agw
In particular, the drainable porosity, n(z), is strongly affected by the soil Wgw =
K2 (9)
properties and soil water content profile, and strongly varies when the
water table is close to the ground surface (Laio et al., 2009). For this The robustness of this assumption is analyzed in Section 4.2 with the
reason, we modeled n as a function of the water level z following the aid of numerical simulations.
approach developed by Acharya et al. (2012): To calculate ⟨z⟩ from 〈Wgw〉, the dependence of the drainable
porosity on the groundwater level (Eq. (4)) needs to be accounted for.
⎧ −⎛ b + 1 ⎞ ⎫
⎜ ⎟
However, the steady-state pdf for the water volume is derived by con-
n (z ) = (θs − θr ) 1 − (1 + (η (z − H ))b) ⎝ b ⎠ sidering that the groundwater is fixed around its mean level (⟨z⟩).
⎨ ⎬
⎩ ⎭ (5) Therefore, the drainable porosity is not subjected to any variation
(Supplementary Material, Section S-2). Under this assumption, from
where θs and θr are respectively the saturated water content and the
Eq. (9) we have:
residual water content, while η and b are soil-specific parameters of the
modified van Genuchten retention model (van Genuchten, 1980) in- λαp
z =
troduced by Troch et al. (2003). Note that in Eq. (5), the suction at the K2 n ( z ) (10)
soil surface is modeled for a hydrostatic pressure distribution. However,
Eq. (4) is now circular since it defines z in terms of drainable porosity, which has to be solved numerically, or by assuming an effective value
which is function of z. Thus, the groundwater level can be evaluated for n.
only by solving numerically, Eq. (4), or by assuming a constant, effec- The pdf for wetland volume consists of writing the mass balance
tive value for the drainable porosity. The impact of the time variant Eq. (1) scaled with maximum volume Wmax :
drainable porosity on the dynamics of groundwater and wetland stage dW *
= −ρ* (W *) + ξ (λp , r2 αp)
is further discussed in the Supplementary Material (Section S-2). dt (11)
where r2 is the ratio between the contributing area of the wetland Ac
2.2. Wetland bathymetry and Wmax and W * = W / Wmax is the normalized pool water storage. In
Eq. (11), the function ρ*(W*) includes both the volumetric water loss by
It is necessary to define the relationships between the depression ET and the wetland-groundwater exchange term. Therefore, the func-
area (A), volume (W) and depth of water (h) for a wetland. We use a tion ρ*(W*) can be expressed as:
simple power-function model for wetland volume-area-depth
(W − A − h ) relationships, from Hayashi and van der (2000) and ρ* (W *) = ETr1 (W *) β + p1 (W *)2 − 3β /2 + p2 (W *)1 − β /2 (12)

216
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

−1
where the three terms on the right hand side represent respectively loss 1 1 1⎤
driven by ET (with r1 = Amax / Wmax ), exfiltration from the wetland to C=⎡ ∫0 I (W *) dW * +
⎢ ETr1 λp ⎥
⎣ ⎦ (22)
groundwater, and infiltration from groundwater into the wetland. The
coefficients p1 and p2 which describe these processes can be expressed Then the solution in Eq. (20) is expressed as:
as:
1 ϕ*r
pW (W *) = I (W *) + δ (W *)
4 π ∫
1
I (W *) dW * + ϕ*r ∫
1
I (W *) dW * + ϕ*r
p1 = hmax K1 0 0
(2 − β ) Amax (13)
(23)
4 π Eq. (23) is, in fact, a generalized solution for water storage dynamics in
p2 = (hgw − z ) K1
(2 − β ) Amax (14) a non-linear reservoir subject to shot noise (Park et al., 2014) and
where hgw is the thickness of the aquifer. coupled with the groundwater.
Using this loss/gain function and following the approach developed
by Botter et al. (2009), the steady-state pdf for W* can be expressed as: 2.4. Groundwater-dominated wetlands

C W*
pW (W *) =
ρ* (W *)
exp ⎧ −

+ λp ∫ ρ* (W 1*) dW * ⎫⎬ + C
δ (W *) As the hydrologic resistivity of the wetland bottom approaches zero
(i.e. K1 ≫ 0), the time-lag between wetland stage and the shallow
⎩ αp r2 ⎭ λp
(15) groundwater table diminishes, such that the wetland is simply surface
expression of the groundwater level. In such cases, wetland stage and
where the first term on the right-hand side expresses the continuous part groundwater depth are deterministically linked at all times, and tem-
of the pdf when 0 < W* ≤ 1, while the second term is the atom of poral variability of the wetland is described entirely by the dynamics of
probability in W * = 0 (i.e., fraction of time when W * = 0 ) and there is the shallow groundwater. The groundwater-wetland coupled model
no open water in the wetland. The term C is a normalization constant (Section 2.3) tends to converge to model governed only by the
1
that ensures ∫0 pw (u) du = 1. groundwater as K1, approaches infinity. However, this convergence
Because of the nature of the inverse of the loss/gain function in cannot be easily shown analytically using Eq. (23) since the loss term
Eq. (15), a closed form expression for pW (W *) can not be obtained. inside the steady state pdf approaches infinity for K1 → ∞. Therefore,
Hence, pW (W *) needs to be calculated numerically. the dynamics of a wetland with low bottom resistance have to be in-
The shape of the pdf is controlled by the wetland shape parameter,
vestigated by reformulating the water balance for a groundwater
β, and the following scaled parameters:
system recharged by effective rainfall.
ETr1 ETAmax Here, we investigate how well this approximation represents ob-
ϕ* = =
αp λp r2 αp λ p A c (16) served stage variations in a case study (Section 4.3). In this model we
no longer need a separate wetland mass balance Eq. (1) and we only
r = αp r2 (17) conceptualize the groundwater as a linear reservoir, whose dynamics
p1 follow the mass balance in Eq. (2) with a constant drainable porosity. If
ɛ1 = this mass balance is expressed in terms of the groundwater stage z, we
ETr1 (18)
obtain:
p2
ɛ2 = dz αp
ETr1 (19) = −K2 z + ξ ⎛λ, ⎞
dt ⎝ n⎠ (24)
These parameters can be physically interpreted as follows; ϕ* is a
normalized aridity index, which quantifies the ratio of ET rate occur- According to Eq. (24), the groundwater dynamics are completely de-
ring from the maximum water surface relative to the mean water input scribed in terms of stage. Assuming that wetland stage is at hydrologic
rate from rainfall events occurring over the contributing area; r is the equilibrium with the shallow groundwater level, the only parameter
average volumetric input per event, scaled to the maximum volume of needed is wetland bottom elevation. Note that given the groundwater
the wetland; ε1 is interpreted as the ratio of exfiltration to evapo- level at any time, it is possible to know when the wetland starts to fill.
transpiration losses when W* = 1; ε2 is interpreted as the ratio of in- This happens when the groundwater is close to and above the wetland
filtration to evapotranspiration losses when W* = 1. The additional bottom. In terms of pdf, this can be quantified considering the ex-
parameter, ε2, compared to Park et al. (2014) model, accounts for the ceedance probability P[z > zb] where zb is the threshold that corre-
possible infiltration contribution from shallow groundwater into the sponds to the wetland bottom evaluated from the reference point of the
wetland. groundwater. Furthermore, when there is a known maximum wetland
The steady state pdf for wetland storage can be expanded using stage above the ground surface (zmax ), such as from surface drainage
these dimensionless parameters as: outlets, it is possible to confine the groundwater stage also to this upper
boundary.
C ⎡ 1 ⎤ exp ⎧ 1 If we shift the datum at the bottom of the wetland, zb, and we scale
pW (W *) =
ETr1 ⎢ β
⎣ (W * ) + ɛ1 (W *)
(2 − 3β /2) + ɛ (W *) (1 − β /2) ⎥
2 ⎦ ⎨
⎩r the mass balance Eq. (24) to the maximum depth of the wetland
(h max = zmax − z b ), we can describe the dynamics of the wetland nor-
ETr1
− ⎡−W * +
⎢ ϕ*
∫ ρ* (W 1*) dW * ⎤⎥ ⎫⎬ + C
λp
δ (W *) malized stages h* = (z − z b)/ h max . When h* approaches zero, the wet-
⎣ ⎦⎭ (20) land is almost completely dry, while a value of h* close to 1, indicates a
By letting, full wetland. Thus, the mass balance given in Eq. (24), is expressed in
terms of h* as:
1 1
I (W *) = exp ⎧ dh* αp ⎞
(W *) β + ɛ1 (W *)(2 − 3β /2) + ɛ2 (W *)(1 − β /2) ⎨
⎩r = −K2 h* + ξ ⎛λ, ⎜ ⎟

dt ⎝ hmax n ⎠ (25)
ETr1
− ⎡−W * +
⎢ ϕ*
∫ ρ* (W 1*) dW * ⎤⎥ ⎫⎬ The steady state solution for the pdf of the wetland normalized
⎣ ⎦⎭ (21)
stages h* is obtained following Porporato et al. (2004) and
we can compute the normalizing constant C by knowing that Botter et al. (2007). The general solution for the mass balance Eq. (25)
1
∫0 pw (W *)) dW * = 1. Thus, can be expressed by a gamma distribution:

217
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

Fig. 2. Site map of Florida with the location of the two studied wetlands. Case Study 1 represents the wetland located within the boundaries of the SFWMD and is located in Lake
Okeechobee watershed. Case Study 2 represents the wetland located within the boundaries of the SWFWMD and is located in Tampa Bay watershed. In this inset, which is taken from
SWFWMD website, it is shown that the wetland is located below the Starkey well boundary.

γH(λ / K2) 2.5. Threshold crossing times


ph (h*) = exp(−γH h*)(h*)(λ / K2− 1)
Γ(λ / K2) (26)
The duration of water stage excursion above and below some
where the parameter γH = h max n/ αp represents the mean normalized threshold level is of crucial importance for the management of the GIW
stage increment due to incoming rainfall. as viable aquatic habitats or to perform some critical biogeochemical
From Eq. (26), it is evident that the pdf of the normalized wetland functions. One possible way to represent these characteristics is to
stage is completely described by three parameters: γH, λ and K2. In analyse the crossing properties at some specific levels of water volume
particular, the ratio λ/K2 is the key parameter which defines the shape (or stage) that are important for the system.
of the pdf. When (λ/K2) > 1 the pdf is a bell shape, typical of a wet In general, for a stochastic variable x that obeys the Langevin
regime, while when (λ/K2) < 1, the shape switches to monotonically equation:
decreasing, indicating a dry regime (Botter et al., 2013). Thus,
dx
(λ / K2) = 1 serves as an important threshold for separating dry (erratic) = −ρ (x ) + ξ
dt (30)
and humid (persistent) hydrologic regimes for wetlands in equilibrium
with the groundwater. However, since we are considering a domain where ξ is a marked Poisson noise and ρ(x) the loss function, the time
bounded at the bottom (h* = 0 ) and at the top of the wetland (h* = 1), it interval, τI, elapsed between an upcrossing and a subsequent down-
is necessary to truncate this pdf and account for the atoms of probability crossing of a threshold ξ is a new stochastic variable. The mean of this
that quantify the probability to observe a completely dry wetland variable, TI(ξ), represents the average interarrival between an up-
(h* = 0 ) and the full-pool stage (h* = 1). In order to fully characterizes crossing of the threshold from below and the subsequent downcrossing
these scenarios it is necessary to define two atoms of probability: from above (Tamea et al., 2011):

p0 (h* = 0) = P [z < z b] δ (z ) = P [h* < 0] δ (h*) (27) 1 − PE (x )


TI =
ρ (x ) p (x ) (31)
p1 (h* = 1) = P [z > z max ] δ (h* − 1) = P [h* > 1] δ (h* − 1) (28)
where the terms PE(x), p(x) and ρ(x) represent respectively, the cumu-
Therefore, the full analytical expression for the pdf of the wetland lative non-exceedance probability, the pdf and the loss function of the
normalized stages h* can be defined as: relevant variable. Similarly, the time interval, τE, elapsed between a
downcrossing and a subsequent upcrossing is a stochastic variable,
γH(λ / K2) whose average value is:
ph (h*) = exp(−γH h*)(h*)(λ / K2− 1) H [1 − h*] H [h*] + p0 (h* = 0)
Γ(λ / K2) PE (x )
TE =
+ p1 (h* = 1) (29) ρ (x ) p (x ) (32)
where H represents the Heaviside step function. In this case, the shape In our analysis, the variable x could be either the normalized volume of
of the pdf expressed in Eq. (29) is not only influenced by the ratio λ/K2, the wetland W*, as in Eq. (11), or the normalized stage of the wetland
but is also affected by the truncation due to the atoms of probability, h*, as in Eq. (25).
which may transform a hump-shaped pdf into a monotonically de- To compute the observed exposure times, TE, a threshold is selected
creasing. The expression for the steady-state pdf of the normalized stage for either the normalized volume of the wetland (W*) or the normalized
h*, given by Eq. (29) is useful since it represents a general approach for stage of the wetland (h*). Then, the upcrossing and downcrossing
modeling long-term dynamics of GIW with permeable bottom, and in events are identified, and the interarrivals between a downcrossing and
equilibrium with shallow groundwater. the following upcrossing are calculated. Finally, the mean value of all

218
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

Table 1
Parameters of the model for the two case studies.

Model parameter Description Lake Okeechobee Tampa Bay Source

Measured
hmax [m] Max depth 0.5 0.8 Nilsson et al. (2013); Park et al. (2014)
Amax [ha] Max surface area 1.8 3.2 Min et al. (2010); Nilsson et al. (2013)
Ac [ha] Contributing area 3.0 Park et al. (2014)
Wmax [m3] Max volume 4230 Park et al. (2014)
β Wetland profile coefficient 0.53 Park et al. (2014)
ET [mm/d] Evapotranspiration 3.6 2.7 Abtew (2001); Bidlake et al. (2006)
αp [mm] Mean rainfall depth 9.1 12.5
λp [1/d] Mean rainfall frequency 0.41 0.29
Estimated
hgw [m] Aquifer thickness 0.5
n Drainabel porosity 0.09 0.3
Zr [cm] Root Zone 25 25
θfc Field capacity 0.45 0.5
θw Wilting point 0.15 0.15
Derived
λ [1/d] Mean effective frequency 0.27 0.16 Eq. (3)
⟨z⟩ [m] Mean groundwater level 0.65 Eq. (4)
ϕ* Aridity index 0.58 Eq. (16)
r Average volumetric input 0.06 Eq. (17)
ε1 Exfiltration rate 1.85 Eq. (18)
ε2 Infiltration rate -0.76 Eq. (19)
Calibrated
K1 [1/d] Sediment layer recession 1.56
K2 [1/d] Groundwater recession 0.045 0.014

these interarrivals was computed. This procedure is then repeated for the NPE is 14.3 m. Therefore, the difference between these values
different thresholds. (0.8 m) is the estimated maximum depth.

3. Case studies, hydrological data and calibration 3.2. Hydrological data

3.1. Case studies Florida wetland BW1 was already analyzed by Park et al. (2014),
therefore, for this wetland, data are available in terms of both stages
Two karst plain depressional GIWs in different parts of Florida and volume storages of the wetland. In order to normalize these data,
(USA) are considered as representative cases studies in this paper: one the elevation of the ditch bottom has been used to set hmax , while the
in the Lake Okeechobee watershed and one in the Tampa Bay wa- maximum water volume recorded during the entire period of observa-
tershed (Fig. 2). tion, was used as Wmax . In addition, the time series of the upland
Min et al. (2010) monitored water levels in four GIWs in the Lake groundwater is also available (Min et al., 2010). In this study, we ex-
Okeechobee basin, and we used their data for the wetland labeled BW1 amined wetland and groundwater dynamics during the wet season from
(27°24′34.00′′ N, 80°56′48.76′′ W) because the stage dynamics of this August to October 2003. Wetland dynamics during the dry season,
wetland were least affected by artificial drainage. This wetland is when the groundwater stage us far below the wetland bottom, have
characterized by a maximum surface area of about 1.8 ha, and the only been examined by Park et al. (2014). Rainfall data for the wet period
ditch present in the wetland is located at a height of 0.5 m from the under consideration were downloaded from the South Florida Water
bottom, which imposes the maximum stage of the wetland. Note that Management District (SFWMD) hydrologic database. These daily pre-
for the major part of the depressional wetlands present in the region the cipitation data were measured at the Basset station (27°24′41.10′′ N,
depth is never larger than one meter and they typically exhibit a bowl 80°55′16.2′′ W), 2.3 km from to the studied wetland.
or elongated bowl-shape. Bhadha and Jawitz (2010) studied a similar In the second case study, wetland stage level was monitored daily by
wetland in the area and found that the soil immediately below wetlands South West Florida Water Management District (SWFWMD) over a
(0 − 2m ) is characterized by a higher organic matter content with an period of twelve years, from 2003 to 2015. Rainfall data were down-
order of magnitude lower hydraulic conductivity compared to upland loaded from the National Oceanic and Atmospheric Administration
soils. Thus, the wetland bottom restricts rate of upland-wetland water (NOAA) website for the station located in the Tampa International
exchange. Below this 2 m, a clay-rich soil layer is present. Airport for the same period. Groundwater monitoring data from a well
The second wetland selected for this study is located south of the near the Tampa Bay site were also obtained from the SWFWMD hy-
Starkey well-field boundary (Nilsson et al., 2013) and classified as a drologic database. However, these are at lower temporal resolution
cypress wetland. Wetland planar area (A=3.2 ha) was calculated from (monthly) compared to daily data for stage measured in a well located
the Geographic Information System (GIS) Land Use 1999 (SWFWMD in the middle of the wetland.
2007) and it corresponds to the wetland surface extent. Additionally,
wetland dry bed elevation (DBE) and normal pool elevation (NPE) have 3.3. Estimation of model parameters
been previously identified by South West Florida Water Management
District (SWFWMD) staff. The wetland DBE was evaluated as the dry The wetland-groundwater model developed in Section 2.2 was ca-
reading mark in the staff gauge and assumed to be the lowest point in librated based on the observation data from the study wetland during
the wetland (Nilsson et al., 2013). The wetland NPE corresponds to August 2003-October 2003 from the Lake Okeechobee wetland. The
physical and vegetative markers, i.e., inflection points on the buttresses mean rainfall depth, αp, was calculated as the observed mean daily
of cypress trees (Carr et al., 2006). In this wetland, the DBE based on depth during wet days, mean rainfall frequency, λp, was calculated as
the National Geodetic Vertical Datum of 1929 (NGVD) is 13.5 m, and the relative proportion of wet days, while monthly average ET rate is

219
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

the recession constant of the groundwater, K2, was calibrated on the


basis of its lowest mean square error (MSE) between the numerical si-
mulations and the observed time series. The calibration was performed
using both a time varying drainable porosity, n(z), following
Acharya et al. (2012), and also using a constant value for this variable.
Table 1 reports only the parameters obtained from the latter method.
The alternative calibration procedures are reported in the Supplemen-
tary Material, and they provide similar results.
Calibration to estimate wetland recession constant (K1) was based
on the match between the observed wetland stage time series and the
model simulations (see Table 1). In particular, the time series simulated
by the model through Eq. (1) is an expression of the coupled ground-
water-wetland dynamics.
For the Tampa Bay wetland, αp and λp were calculated from the data
available during the period 2003–2015. The two key parameters of the
groundwater-dominated model are λ and K2. In this case the variation
of the drainable porosity can be neglected since we are considering only
the time periods during which the groundwater level is above the
wetland bottom. Therefore, the porosity is constant and its value is
reported in Table 1. The estimation of the recession constant K2 is
performed following the method developed in Basso et al. (2015). From
the water stage record we considered only those recessions that happen
above the dry bed. Then, every recession was estimated through a
linear regression between the temporal derivatives of z (dz/dt) and the
corresponding observed z. We selected the median value as the reces-
sion constant from all estimated values.

4. Results

4.1. Sensitivity analysis

The dependence of the pdf shapes for normalized water volume, p


(W*), (Eq. (23) and mean exposure times, TE, (Eq. (32) on morphologic
and hydro-climatic attributes, was explored by varying the five scaled
parameters ϕ*, ε1, ε2, r and β. The detailed results of the sensitivity
analysis are reported and commented in Supplementary Material
(Section S-3).
This analysis showed that four scaled parameters (ϕ*, r, ε1, ε2) have
largest impact on the shape of p(W*) and TE: two are related to the
hydroclimatic forcing (ϕ*) and wetland bathymetry (r), and the other
two define hydrologic recession of wetland bottom (ε1) and shallow
groundwater (ε2).
Scaled aridity index, ϕ*, which represents the ratio of potential ET
rate to the water input due to rainfall, has a large impact on the shape of
p(W*) and TE. As ϕ* increases from low to high values, the pdf shifts
from a monotonically increasing distribution, with its mode con-
Fig. 3. (A) Comparison between temporal patterns of rainfall and normalized water vo- centrated near the maximum volume, towards a monotonically de-
lume in the Lake Okeechobee wetland (BW1). The blue line represents the model pattern, creasing pdf with high probability of low levels of the volume of water
while red circles represent the observations collected by Min et al. (2010). (B) Compar-
inside the wetland (Figure S2 A-B). The shapes for p(W*) and TE are also
ison between the observed pdfs generated from data of Min et al. (2010) (bar histograms)
influenced by the parameter ε1 (Figure S2 C-D), which is a function of
with the steady state pW (W *) (Eq. (23)). (C) Comparison between the observed mean
the wetland bottom recession constant, K1, and ε2 (Figure S2 E-F),
exposure time generated from the data (red circles) and the analytical model presented in
Eq. (32) for the same period (blue solid line). Fifty-years Monte Carlo simulation are also
which is controlled by the groundwater recession constant K2 and the
reported for both the pdf and exposure times (blue circles). (For interpretation of the drainable porosity n. Larger values of ε1 and positive values of ε2 shift
references to colour in this figure legend, the reader is referred to the web version of this the mean of p(W*) towards lower normalized volume because the ex-
article.) filtration process dominates. The pattern of the mean exposure times
reflects this behavior. Under these conditions the water losses are re-
provided by Carr et al. (2006) and is reported in Table 1. Based on the levant and so the time spent by the system at lower volumes is large. On
hmax value, we evaluated the contributing area of the site using the the other hand, when ε2 is negative, the infiltration from the ground-
relation Ac = Amax /0.6 following the stage-area relationship presented water drives the mean of pdf of the normalized volume towards larger
in Min et al. (2010). The wetland profile coefficient β (Table 1) was volumes. The shape of p(W*) and TE is also sensitive to the parameter r
estimated from survey data reported in the same paper. (Figure S2 G-H) that is related to the maximum volume of the wetland.
The groundwater table parameters were estimated using the ob- When this parameter is large, the wetland tends to be smaller and
served data. The mean frequency for the effective rainfall events re- therefore more sensitive to the fluctuations of the volume; the variance
charging groundwater, λ, was obtained from precipitation and ET data of p(W*) is large and the mean exposure times are small. When r de-
by means of Eq. (3), using the n, Zr, s1 and sw values in Table 1. Then, creases the variance of p(W*) decreases and converges to a Gaussian
distribution, while the exposure times are longer. Instead, the shapes

220
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

for p(W*) and TE are quite insensitive to variations of β (Figure S2 I-J). completely empty wetland, while the value of h* = 1 represents a full
wetland. The pdf of wetland stage is characterized by a bell shape, ty-
4.2. Lake Okeechobee wetland pical of a wet regime. In fact, this is confirmed by ratio (λ/K2) ≃ 15.
Good agreement was observed (Fig. 4A) between analytical and the
Time series of the W* from the data collected by Min et al. (2010) observed p(h*), although the analytical solution tends to slightly
for the Lake Okeechobee wetland (BW1) are compared with the cali- overestimate the probability of high stages. This discrepancy could be
brated numerical simulation in Fig. 3A. These results were calculated explained by a small error in the proposed normal pool elevation. The
by solving Eq. (11) where water inputs to the systems were the actual observed pdf shows that the wetland never exceeded h* = 0.95, corre-
rainfall data measured at Basset station, while the groundwater level z sponding to a height of 14.25 m. However, the value of the normal pool
was obtained from Eqs. (2) and (4) with a constant value of n. The elevation provided by the SWFWMD was 14.3 m. The analytical solu-
results obtained using a drainable porosity dependent on z are discussed tion tends also to underestimate the probability of low normalized
in the Supplementary Material (Section S-2). The numerical model re- stages, h*. As for the Lake Okeechobee wetland, one of the possible
produces the coupled dynamics of the two systems (Fig. 3A). These reasons is related to the assumption of a simple bathymetry, instead of a
results emphasize that the recession of the bottom layer, K1, which is more complex shape. In addition, the simplified model expressed in
embedded in the definition of ε1, is the key parameter governing cou- Eq. (25) assuming groundwater fluctuations to be the dominant factor
pled dynamics between wetland and groundwater. controlling stage dynamics, underestimates the role of second-order
Comparison between the analytical (blue solid line) and observed effects by neglecting the coupled dynamics. This is evident in p(h*)
pdf (red circles) of the W*, shows that the analytical pdf captures the based on long-term time series generated by numerical simulations
observed frequency distribution of W* (Fig. 3B). The observed period is presented in the Supplementary Material (Figure S4 A). Here, the nu-
characterized by a high probability associated with larger volumes, merical pdf better reproduce the lower normalized stages.
typical of a “wet” regime. The analytical mean exposure times trend (solid line) is compared
Q-Q plots were used to estimate the error between analytical and (Fig. 4B) with the observed mean exposure times (circles) obtained
observed pdfs. The procedure consists on building the Q-Q plot by using from the analysis of the stage record. As the threshold increases the
9 quantiles ranging from 0.1 to 0.9 and then calculating the MSE be- mean exposure time increases as well, indicating the tendency that the
tween the simulated quantiles and those achieved from the real data. lower points of the wetland tend to be flooded for a longer period. The
The result confirms the good agreement between the observed and the analytical solution tends to overestimate the exposure time spent at
analytical pdfs, despite the assumption of having considered the higher stages.
groundwater as fixed around its mean level. Observed pdf is strongly
affected by the initial condition, since the available data are restricted 5. Discussion
to only one season of the year. This may be why the analytical pdf
underestimates the lower normalized volumes in the initial days of the The stochastic models we presented in this work and in our earlier
season (Fig. 3A). Another reason that could lead to this underestimation paper (Park et al., 2014), present multiple analytical pdf solutions for
is related to the assumption of considering a simple bathymetry for the the study of geographically isolated wetland hydrologic dynamics. The
wetland (constant value of the shape parameter β), while the actual proposed models are summarized in Table 2 along with the key para-
landscape topography is perhaps more complex. meters that define the shapes of the probability density function and
Fig. 3C compares the analytical mean exposure time obtained by crossing times.
solving Eq. (32) and the observed values derived from field data. The Comparison of short-term observed time series for Lake Okeechobee
general trend (one season), of mean exposure times given by Eq. (32), is wetland with the simulations obtained from Eqs. (1) and (2) showed
similar to the observed pattern. However, the data from good agreement. Despite the limited time period (one season) for which
Min et al. (2010) are for short term, and do not allow a robust analysis data are available, and simplification of considering the groundwater
of mean crossing times for a wide range of thresholds. In particular, the depth as being fixed around its long-term mean level (thus avoiding
wetland time series of Fig. 3A shows that the wetland tends to remain at solving the joint pdf for groundwater stage and wetland volume), the
higher level for much of the time (W * = 1 means a full wetland), analytical pdf of W* captures well the observed values. Because a closed
without significant excursion in the lower part of the state space, that form for the analytical p(W*), which accounts for the coupled dynamics
are essential to better represent the observed dynamics crossing times. between wetland and groundwater could not be derived, in Fig. 3 we
Analyses based on multi-year data were performed with data from the reported also the results in terms of pdf and exposure times TE from a
Tampa Bay wetland, as discussed next. 50-years Monte Carlo numerical simulation, using the numerical model
with time-variable groundwater depths. These simulations were ob-
4.3. Tampa Bay wetland tained by solving the system of Eqs. (1) and (2), thus accounting for the
coupled dynamics between wetland and groundwater. The numerical
For the Tampa Bay wetland, we present the results of the applica- simulations were obtained forcing the systems with 50 years of syn-
tion of the groundwater-dominated wetland model (Section 2.4), in- thetic rainfall evaluated fixing αp and λp. In addition, the non-dimen-
stead of the groundwater-wetland coupled model (Section 2.3) which sional parameters that describe the pdf of W* were kept constant from
was applied to Lake Okeechobee wetland. The results achieved using the results of the calibration procedure.
the latter model show that the analytical pdf is not able to correctly Fig. 3 shows that numerical Monte Carlo simulation tends to better
reproduce the wetland dynamics, primarily because the assumption of reproduce the observed pattern of p(W*), since in this model the
fixed (at steady state level) groundwater is not valid. This statement is groundwater fluctuates and strongly interacts with the wetland pool. In
supported by the stage pdf obtained by solving Eqs. (1) and (2), where addition, especially for larger volume, the analytical model (Eq. 23),
the groundwater is free to fluctuate. In this case, the numerical pdf of tends to overestimate p(W*) and, therefore, underestimate TE. Exposure
normalized stage for the coupled model is more similar to the observed times estimated using the long-term simulations (Monte Carlo) with the
pattern. Further details are reported in the Supplementary Material numerical model (Eq. (1) and (2)) more closely follow the trends pre-
(Section S-4). dicted by the analytical model for the and mean exposure times
Fig. 4 presents the results of the application of the groundwater (Eq. (32)). Despite this, the analytical pdf and analytical TE derived by
dominated models to the Tampa Bay wetland. The analytical pdf solu- assuming a constant value for the groundwater are not significantly
tion expressed by Eq. (29), reproduces the empirical pdf based on h* different from the results of the numerical simulation.
observed in the wetland. Zero value on the x-axis represents a Arguably, the assumption of steady groundwater depth is

221
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

Fig. 4. (A) Comparison between the ob-


served pdf generated from data (bar histo-
grams) with the steady state pdf presented
in Eq. (29) (red solid lines) in Tampa Bay
wetland. (B) Comparison between the ob-
served mean exposure time generated from
the data (blue circles) and the analytical
model presented in Eq. (32) (red solid lines)
for the same period. (For interpretation of
the references to colour in this figure le-
gend, the reader is referred to the web
version of this article.)

particularly effective for Lake Okeechobee wetland because the shallow could readily adapt to a wide range of hydrologic conditions.
groundwater did not frequently cross the wetland stage. Shallow
groundwater rises above wetland pool only 25 days out of 92. In sup-
port to these considerations, the values of the infiltration parameter is 6. Conclusions
ε2 is -0.76, indicating a weak tendency for groundwater to infiltrate into
the wetland. This analysis suggests that the assumption made by The models described in this paper aim to quantify the effect of
Park et al. (2014) in an earlier study, where the wetland was re- temporal hydrologic variability of storage and stage in GIWs, as influ-
presented as affected only by an effective exfiltration process, is justi- enced by stochastic hydro-climatic forcings (rainfall and evapo-
fied. transpiration) with connectivity to surrounding upland or shallow
For the Tampa Bay wetland, stage is just an expression of the groundwater. Starting from the system of Eqs. (1) and (2) and in-
groundwater dynamics, because the wetland is flooded only when the troducing some simplifying assumptions we derived pdf solutions for a
water-table rises above wetland bottom. Differently from the Lake range of scenarios. The first model (Supplementary Material, Section
Okeechobee wetland, in the Tampa Bay wetland scenario the seasonal S1) can be used for a GIW wetland an impervious bottom (K1 ap-
effects of short-term data are not relevant. In fact, here we have a long proaches zero), and the wetland is truly isolated from the surrounding
record (12 years at daily scale) and, therefore we can perform a long- upland and shallow groundwater. The second model (Park et al., 2014),
term analysis, thus comparison of empirical and analytical pdf of stages the wetland bottom has a finite permeability, but the shallow ground-
provides a robust assessment. water level is below the groundwater level such that evaporation plus
The main assumption underlying the groundwater-dominated exfiltration from GIW to upland are the only loss pathways (ε). In the
model is that equilibrium is instantaneously reached between shallow third model, the groundwater is located fixed at an elevation close to
groundwater and the wetland level once groundwater crosses the the wetland bottom, either below or above the wetland stage, with both
wetland bottom. Resistivity of the wetland bottom is assumed to be low exfiltration (ε1) and infiltration (ε2) as the contributing processes
enough that it does not control lag in the hydrologic response of the (Section 2.3). This approximation may underestimate the volume and
wetland. This scenario in which the wetland is just a surface-water stage in the wetland when the groundwater is highly variable and the
expression of the shallow groundwater is useful for the hydrological wetland sediment layer allows the water exchange between the two
modeling of multiple wetlands in a region that is forced by the same systems. For the fourth case, the wetland bottom is highly permeable
rainfall and shallow regional groundwater. Once the relative position of (K1 approaches infinity), enough to offer little hydrologic resistance,
the wetlands bottom with respect to water-table is known, its stage and the shallow groundwater dynamics dominate wetland stage dy-
dynamics can be evaluated based only on groundwater fluctuations. namics (Section 2.4).
For both types of wetlands in Florida, we also derived analytical We also present comparisons of the analytical pdfs to empirical pdfs
expressions for mean crossing times for exceedance of any specified based on Monte Carlo simulations using a numerical model to examine
threshold wetland stage (or volume), and compared them with values adequacy of the simplifying assumptions. Given the limitations on
estimated from the time series data from site monitoring. In concert availability of or access to long-term monitoring data for GIW hy-
with other factors, hydrologic threshold-crossing-times have important drology, we present here comparison to a limited data set from two
implications for determining the suitability of wetlands as aquatic ha- GIWs in Florida. Compiling observational data from GIWs in diverse
bitats, and adaptive strategies specific flora and fauna need to adopt for landscape settings with a range of hydro-climatic forcing (Budyko
coping with variable conditions in the wetlands. Thus, the application Aridity Index) is indeed the focus of our ongoing efforts. Our sensitivity
of the crossing times theory to wetlands can be very useful from an eco- analyses suggest that long-term GIW stage fluctuations are moderated
hydrological perspective to determine habitat suitability. For example, by few scaled parameters as reflected in the shifts of the pdf shapes. The
particular “specialist” plant species could only survive under conditions two recession constants, K1 for wetland bottom and K2 for shallow
with a small range of hydro-periods, while other “generalist” plants groundwater, control the relative importance of coupling of wetland
with upland and groundwater, while the hydro-climatic forcing (ϕ*)

Table 2
Summary of the models considered in investigating wetland hydrology.

Model Description Equation Dimensionless parameters

Impervious GIW Resistivity of wetland bottom approaches infinity (K1 → 0) Eq. (S2) ϕ*, r, β
GIW connected to upland Connection to upland is driven by exfiltration Park et al. (2014) ϕ*, ε, r, β
GIW connected to upland and groundwater Exchange between wetland and steady groundwater Eq. (23) ϕ*, ε1, ε2, r, β
Groundwater Dominated GIW Resistivity of wetland bottom approaches zero (K1 → ∞) Eq. (29) γH, λ/K2

222
L.E. Bertassello et al. Advances in Water Resources 112 (2018) 214–223

defines the balance between precipitation inputs and evaporation ephemeral (vernal) forest pools in southern New England. Wetlands 22.
losses. We also investigated wetland stage fluctuations based on mean 247–55.https://dx.doi.org/10.1672/0277-5212(2002)022%5B0247:DAVAHR%5D2.
0.CO;2.
crossing time above (below) a specified threshold, to characterize the Budyko, M.I., 1974. Climate and life. San Diego, California: Academic, pp. p.508.
persistence of certain hydrologic regimes. Carr, D.W., Leeper, D.A., Rochow, T.F., 2006. Comparison of six biologic indicators of
For both wetlands, comparisons of stage (volume) pdfs with model hydrology and the landward extent of hydric soils in west-central Florida, USA cy-
press domes. Wetlands 26, 1012.
simulations suggest that long-term temporal variability of wetland hy- Cheng, F.Y., Basu, N.B., 2017. Biogeochemical hotspots: role of small water bodies in
drology can be reproduced using few scaled parameters that are ex- landscape nutrient processing. Water Resour. Res. 53, 5038–5056. http://dx.doi.org/
pressions of site-specific hydro-climatic conditions, wetland pool 10.1002/2016WR020102.
Cronk, J.K., Fennessy, M.S., 2001. Wetland Plants: Biology and Ecology. CRC Press.
bathymetry and the size, the influence of shallow groundwater. Euliss, N.H., Smith, L.M., Wilcox, D.A., Browne, B.A., 2008. Linking ecosystem processes
Model analyses presented here suggest that different shapes of the with wetland management goals: charting a course for a sustainable future. Wetlands
volume and stage pdfs and patterns of mean exposure time can arise 8, 553–562. https://doi.org/10.1672/07-154.1.
van Genuchten, M.T., 1980. A closed-form equation for predicting the hydraulic con-
under different climatic and geomorphic conditions. These analyses
ductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44 (5), 892–898.
may help in the development of a general approach to classify wetland Hayashi, M., van der, K.G., 2000. Simple equations to represent the volume-area-depth
hydrologic regimes, and in the prediction of the likely ecological re- relations of shallow wetlands in small topographic depressions. J. Hydrol. 237,
sponse of wetland ecosystems to unsteady external forcing. For in- 74–85. https://doi.org/10.1016/S0022-1694(00)00300-0.
Laio, F., Tamea, S., Ridolfi, L., D’Odorico, P., Rodriguez-Iturbe, I., 2009. Ecohydrology of
stance, the proposed framework could allow for a theoretical analysis of groundwater-dependent ecosystems: 1. stochastic water table dynamics. Water
the expected shift in vegetation species in response to climate change Resour. Res. 45, W05419. https://doi.org/10.1029/2008wr007292.
scenarios, or how amphibian population dispersal and survival patterns Leibowitz, S.G., 2015. Geographically isolated wetlands: why we should keep the term.
Wetlands 35 (5), 997–1003. https://doi.org/10.1007/s13157-015-0691-x.
are influenced by current hydro-climatic patterns. Marton, J.M., Creed, I.F., Lewis, D.B., Lane, C.R., Basu, N.B., Cohen, M.J., Craft, C.B.,
2015. Geographically isolated wetlands are important biogeochemical reactors on the
Acknowledgements landscape. BioScience 65, 408–418. https://doi.org/10.1093/biosci/biv009.
Min, J.H., Perkins, D., Jawitz, J., 2010. Wetland-groundwater interactions in subtropical
depressional wetlands. Wetlands 30, 997–1006. https://doi.org/10.1007/s13157-
Financial support for this research was provided in part by NSF 010-0043-9.
Project 1354900, “Plant Adaptation in Variable Environments”. Nilsson, K., Rains, M., Lewis, D., Trout, K., 2013. Hydrologic characterization of 56
geographically isolated wetlands in west-central Florida using a probabilistic method.
Additional funding was provided by the Lee A. Rieth Endowment in the Wetlands Ecol. Manage. 21, 1–14. https://doi.org/10.1007/s11273-012-9275-1.
Lyles School of Civil Engineering, Purdue University. Park, J., Botter, G., Jawitz, J.W., Rao, P.S.C., 2014. Stochastic modeling of hydrologic
variability of geographically isolated wetlands: effects of hydroclimati forcing and
wetland bathymetry. Adv. Water Resour. 69, 38–48. https://doi.org/10.1016/j.
Supplementary material
advwatres.2014.03.007.
Porporato, A., Daly, E., Rodriguez-Iturbe, I., 2004. Soil water balance and ecosystem
Supplementary material associated with this article can be found, in response to climate change. Am. Nat. 164 (5). 625–32. https://doi.org/10.1086/
the online version, at 10.1016/j.advwatres.2017.12.007. 424970.
Rains, M.C., Leibowitz, S.G., Cohen, M.J., Creed, I.F., Golden, H.E., Jawitz, J.W., Kalla, P.,
Lane, C.R., Lang, M.W., McLaughlin, D.L., 2016. Geographically isolated wetlands are
References part of hydrological landscape. Hydrol. Process. 30, 153–160. https://doi.org/10.
1002/hyp.10610.
Rodriguez-Iturbe, I., Porporato, A., Ridolfi, L., Isham, V., Cox, D.,. Probabilistic modelling
Abtew, W., 2001. Evaporation estimation for lake Okeechobee in South Florida. Irrig. of water balance at a point: the role of climate soil and vegetation. Proc. R. Soc.
Drain. Eng. 127. 140-7. https://doi.org/10.1002/hyp.7887. London, Ser. A 455, 3789–3805. https://doi.org/10.1098/rspa.1999.0477.
Acharya, S., Jawitz, J.W., Mylavarapu, R.S.W., 2012. Analytical expressions for drainable Snodgrass, J.W., Komoroski, M.J., Bryan Jr, A.L., Burger, J., 2000. Relationships among
and fillable porosity of phreatic aquifers under vertical fluxes from evapotranspira- isolated wetland size, hydroperiod, and amphibian species richness: implications for
tion and recharge. Water Resour. Res. 48, W11526. https://doi.org/10.1029/ wetland regulations. Conserv. Biol. 14, 414–419. https://doi.org/10.1046/j.1523-
2012WR012043. 1739.2000.99161.x.
Altermatt, F., Pajunen, V.I., Ebert, D., 2009. Desiccation of rock pool habitats and its Tamea, S., Laio, F., Ridolfi, L., Rodriguez-Iturbe, I., 2011. Crossing properties for geo-
influence on population persistence in a Daphnia metacommunity. PLoS ONE 4 (3), physical systems forced by poisson noise. Geophys. Res. Lett. 38, L18404. https://doi.
e4703. https://doi.org/10.1371/journal.pone.0004703. org/10.1029/2011GL049074.
Basso, S., Frascati, A., Marani, M., Schirmer, M., Botter, G., 2015. Climatic and landscape Tamea, S., Laio, F., Ridolfi, L., D’Odorico, P., Rodriguez-Iturbe, I., 2009. Ecohydrology of
controls on effective discharge. Geophys. Res. Lett 42, 8441–8447. https://doi.org/ groundwater-dependent ecosystems: 2. Stochastic soil moisture dynamics. Water
10.1002/2015GL066014. Resour. Res. 45, W05420. https://doi.org/10.1029/2008wr007293.
Bhadha, J.H., Jawitz, J.W., 2010. Characterizing deep soils from an impacted subtropical Tiner, R., 2003. Geographically isolated wetlands of the united states. Wetlands 23,
isolated wetland: implications for phosphorus storage. J. Soils Sediments 10, 494–516. https://doi.org/10.1672/0277-5212(2003)023%5B0494:GIWOTU%5D2.
514–525. https://doi.org/10.1007/s11368-009-0151-4. 0.CO;2.
Bidlake, W.R., Woodham, W.M., Lopez, M.A., 2006. Evapotranspiration from areas of Troch, P.A., Paniconi, C., van Loon, E., 2003. The hillslope-storage boussinesq model for
native vegetation in west-central Florida. U.S. Geological Survey water-supply paper. subsurface flow and variable source areas along complex hillslopes: 1. formulation
pp. 2430. http://pubs.er.usgs.gov/publication/ofr93415ER. and characteristic response. Water Resour. Res. 39, 1316. https://doi.org/10.1029/
Botter, G., Basso, S., Rodriguez-Iturbe, I., Rinaldo, A., 2013. Resilience of river flow re- 2002WR001728.
gimes. Proc. Nat. Acad. Sci. 110. 12925-30. https://doi.org/10.1073/pnas.
1311920110.
Botter, G., Porporato, A., Daly, E., Rodriguez-Iturbe, I., Rinaldo, A., 2007. Probabilistic Further reading
characterization of base flows in river basins: roles of soil, vegetation, and geomor-
phology. Water Resour. Res. 43, W06404. https://doi.org/10.1029/2006WR005397.
Botter, G., Porporato, A., Rodriguez-Iturbe, I., Rinaldo, A., 2009. Nonlinear storage-dis- Milly, P.C.D., 1993. An analytic solution of the stochastic storage problem applicable to
charge relations and catchment streamflow regimes. Water Resour. Res. 45, W10427. soil water. Water Resour. Res. 29. 3755-8. https://doi.org/10.1029/93WR01934.
https://doi.org/10.1029/2008wr007658. Milly, P.C.D., 2001. A minimalist probabilistic description of root zone soil water. Water
Brooks, R.T., Hayashi, M., 2002. Depth-area-volume and hydroperiod relationships of Resour. Res. 37. 457-63. https://doi.org/10.1029/2000WR900337.

223

You might also like