MODELING THE FOREST HYDROLOGY OF WETLAND-UPLAND ECOSYSTEMS IN FLORIDA

JOURNAL OF TIlE AMERICAN WATER RESOURCES ASSOCIATIONVOL. 34, NO.4 AMERICAN WATER RESOURCES ASSOCIATION AUGUST 1998

MODELING THE FOREST HYDROLOGY OFWETLAND-UPLAND ECOSYSTEMS IN FLORIDA'

Ge Sun, Hans Riekerk, and Nicholas B. Comerford2

ABSTRACT: Few hydrological models are applicable to pine flat-woods which are a mosaic of pine plantations and cypress swamps.Unique features of this system include ephemeral sheet flow, shal-low dynamic ground water table, high rainfall and evapotranspira-tion, and high infiltration rates. A FLATWOODS model has beendeveloped specifically for the cypress wetland-pine upland land-scape by integrating a 2-D ground water model, a Variable-Source-Area (VAS)-based surface flow model, an evapotranspiration (ET)model, and an unsaturated water flow model. The FLATWOODSmodel utilizes a distributed approach by dividing the entire simula-tion domain into regular cells. It has the capability to continuouslysimulate the daily values of ground water table depth, ET, and soilmoisture content distributions in a watershed. The model has beencalibrated and validated with a 15-year runoff and a four-yearground water table data set from two different pine flatwoodsresearch watersheds in northern Florida. This model may be usedfor predicting hydrologic impacts of different forest managementpractices in the coastal regions.(KEY TERMS: forest hydrology; Florida; ground water hydrology;pine flatwoods; modeling; wetlands.)

INTRODUCTION

The southeastern United States has always been acenter for soft wood timber production due to its idealclimatic and soil water conditions for tree growth(Sabine, 1994). More than one million acres in thecoastal area are classified as pine flatwoods (Cubbageand Flather, 1993). Important values of this ecosys-tem include timber production, wildlife habitat, andgroundwater recharge (Brandt and Ewel, 1989). Ashuman population and wood demands continue to risein the southeastern United States, forestry activitieson pine flatwoods have become increasingly intensi-fied (Riekerk and Korhnak, 1984). Environmental

concerns about forest management practices in thecoastal areas include impacts on water quality, wet-land hydrologic functions, resultant influences onwildlife habitat, and long-term cumulative impacts onsoil productivity. Although the effects of forest man-agement on upland watershed hydrology have beenwell studied in the past century (Bosch and Hewlett,1982; Swank and Crossley, 1988), little information isavailable for the low-land and forested wetland land-scape (Riekerk, 1989; Shepard et al., 1993).

Hydrologic computer simulation models are becom-ing essential tools for scientists as well as land man-agers in the decision making processes (Lovejoy etal., 1997). Although the building procedures and com-ponents of existing hydrologic simulation models aresimilar, one model may be very different from anotherin its capability and applicability. For example, mostof the available hydrologic models developed for hillyregions are not applicable to Florida's coastal condi-tions (Heatwole et al., 1987; Capece, 1994). Also, mod-els developed for agricultural watersheds often needsignificant modification before they can be applied toforests (Thomas, 1989; McCarthy et al., 1992). Severalefforts have been made in modeling flatwoods hydrol-ogy. Heatwole et al. (1987) modified the CREAMSmodel (Knisel, 1980) into CREAMS-WT to better rep-resent the storage-based flatwoods hydrologic systemof agricultural watersheds in southern Florida. Basedon the agricultural drainage model DRAINMOD(Skaggs, 1984), a forest hydrology version DRAIN-LOB was developed to model water managementeffects on the hydrology of loblolly pine plantation inNorth Carolina (McCarthy et al., 1992). The DRAIN-MOD model was also modified to a new model, Field

1Paper No. 97120 of the Journal of the American Water Resources Association. Discussions are open until April 1, 1999.2Respectively, Research Associate, North Carolina State University/USDA Forest Service, 1509 Varsity Dr., Raleigh, North Carolina

27606; Associate Professor (retired), School of Forest Resources and Conservation, University of Florida, Gainesville, Florida 32611; and Pro-fessor, Department of Soil and Water Science, 2169 McCarty Hall, University of Florida, Gainesville, Florida 32611 (E-Mail/Sun:[email protected]).

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION 827 JAWRA

Sun, Riekerk, and Comerford

Hydrologic And Nutrient Transport Model(FHANTM), to simulate the hydrology and phospho-rus movement on flatwoods fields (Tremwel andCampbell, 1992; Campbell et al., 1995). Limited suc-cess was found in directly applying an upland event-based forest hydrological model (VSAS2) ( Bernier,1982) to a pine flatwoods watershed (Guo, 1989). Itwas suggested that the flat topography and largecypress wetland storage significantly reduced storm-flow under most situations. Hydrologic cycles areincluded in most of the forest ecological models forslash pine plantations, but water movement processesare not physically and explicitly modeled (Golkin andEwel, 1985; Ewel and Gholz, 1991).

In recent years, there has been a tendency to devel-op more physically based distributed models (Beven,1989; Jensen and Mantoglou, 1992; Bathurst andO'Connell, 1992; Wigmosta, 1991) and couple hydro-logic processes with biological processes (Al-Soufi,1987; Band et al., 1993; Wigmosta et al., 1994). Thisdevelopment has been driven by the need for compre-hensive large-scale (e.g., basin, global) ecosystemstudies, in which the hydrology is one of the mostimportant components (Swank et al., 1994; Pierce etal., 1987), and has been accelerated by the increase ofcomputation power. The advance of Geographic Infor-mation Systems (GIS) technologies makes it possibleto develop large-scale databases and interpret com-plex model outputs (Maidment, 1993).

The Florida flatwoods landscape is a mosaic ofcypress wetlands and forest uplands; therefore thehydrology of flatwoods is inherently heterogenous andcomplex (Figure 1). Examples are: (1) the slight spa-tial changes in topographic elevation causing signifi-cant changes in the water regime, and (2) obstructivesoil layering because of the spodic and argillic hori-zons in the soil profile. The heterogeneous vegetationcover of wetlands and uplands and associated phenol-ogy may further complicate the interactions betweensurface water and ground water. Due to the complexgeologic formation of flatwoods, the preferential waterpathways under different management conditions ofthis system have not been well documented or under-stood. A new distributed flatwoods forest hydrologicmodel is needed to study the hydrologic processes ofwetland/upland systems and provide a tool to evalu-ate the hydrologic impact of forest harvesting, specifi-cally for this landscape. The South Florida WaterManagement District has proposed a conceptual inte-grated hydrologic model for Dade County in southernFlorida. This area is characterized by flat topography,sandy soils, a shallow ground water table, and welldeveloped canal systems (Yan and Smith, 1994).Unfortunately, the model has not been implementedfor the intended use in regional water supplyplanning. Based on MODFLOW (McDonald and

Harbaugh, 1984) and BROOK (Federer and Lash,1978), the COASTAL model was developed for model-ing large coastal basins (Sun, 1985). Distributedstructure and major hydrologic components of thismodel fit the pine flatwoods hydrologic system, butseveral algorithms have to be rewritten to modelforested landscape. Advantages and disadvantages ofthe COASTAL model have been discussed in detail inSun (1995).

Based on its basic structure, a new forest hydrolog-ical model, FLATWOODS, was developed and validat-ed for pine flatwoods to: (1) predict spatial andtemporal hydrologic effects of forest managementpractices; (2) account for hydrologic heterogeneity andcontinuity of wetland/upland ecosystems and environ-mental variables; and (3) provide a tool for forestwater management and hydrologic research. Thispaper describes model development procedures andpresents model calibration and validation results.

Structure

MODEL DEVELOPMENT

Recognizing the heterogeneity of the flatwoodslandscape, the model imposes a grid over the entirewetland-upland system to discretize the heteroge-neous watershed into different, but homogeneousrectangular cells (Figure 2). The physical properties ofeach cell are assumed to be uniform laterally for eachsoil layer, but non-uniform vertically in different soillayers. Each cell becomes a modeling unit that holdsmathematical equations describing the physical prop-erties. In practice, spatial data for forest lands arerarely available at high resolution with the exceptionof some readily available parameters such as surfaceelevation, vegetation and soil types (wetlands vs.uplands), etc.

Hydrologic Components and Governing Equations

The model consists of four major submodels to sim-ulate the full hydrologic cycle including: (1) evapo-transpiration, (2) unsaturated water flow, (3) groundwater flow, and (4) surface flow (Figures 2 and 3). Themodel simulates the hydrologic processes on a dailytime step. Computer codes were written in one pro-gram in FORTRAN language, which integrates allfour submodels. The flow chart describes the relation-ships among the submodels (Figure 3).

JAWRA 828 JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

Modeling the Forest Hydrology of Wetland-Upland Ecosystems in Florida

Figure 1. One Research Site on a Typical Pine Flatwoods LandscapeShowing the Spatial Relations of Wetlands and Uplands.

W+E

Evapotranspiration

Evapotranspiration (ET) is the most importantcomponent of pine flatwoods water balance (Ewel andSmith, 1992; Sun et al., 1995). Driving forces for thehydrologic system are climatic variables includingrainfall and air temperature (evapotranspiration).The daily rainfall and temperature data as modelinputs of climatic variables are available from actualfield recordings or the local weather station. The ETsubmodel has three components including a rainfallinterception component (In) by forest canopies, evapo-ration from soil/water surfaces, and transpirationthrough plant stomata. The rainfall interceptiondepends on daily rainfall, leaf area index (LAI) andavailable canopy interception storage or dryness ofthe forest canopy.

Rainfall interception is modeled using an empiricalequation:

I=a+bxP (1)

where = rainfall interception (mm/day); P = grossdaily rainfall (mm/day); and a and b = coefficients fit-ted from throughfall measurements.

The maximum interception or canopy storagecapacity Cmax in mm was introduced to make theempirical model (Equation 2) more realistic. Canopystorage capacity was determined by assuming thatcanopy saturation was approximated by a 1.0 mmthick water film over the foliage surfaces:

C5=1.OxLAI (2)


Gator National Forest

• WellContour linePine UplandsClress Wetlands


0

a)•4

a)a)

U)

C0

a)C)a)

a)a)—.C a).OE

o

0

I

a)

0E

CC0

0a)

0- a) -C

-

C

0-C(JD

Cl)

00

C)

0

C

CI)

C')C)


0

a)

0


Then, available canopy storage for rainfall intercep-tion (ASIN) is calculated as:

ASIN = C8 + PET - SIN

where SIN = water stored on the forest canopy theday before the simulation date (mm); PET =potentialevapotranspiration (PET) is estimated by Hamon'smethod (Hamon, 1963; Federer and Lash, 1978).

A PET correction coefficient of 1.3 was found to beappropriate for the Florida climatic conditions as com-pared 1.0 used for Hubbard Brook, New Hampshireand 1.2 for Coweeta, North Carolina (Federer andLash, 1978).

ASIN = C8

I = a + b x P

I, = ASIN

If PET> SIN

If I < ASIN

If I > ASIN

I = 0 If ASIN = 0

Throughfall (Tn), T = P - I, represents rainfallthat passes through all vegetation layers and becomesavailable for infiltration into the ground. Variation ofthe leaf area index (LAI) of cypress wetlands andslash pine plantations as a function of Julian day (t)were derived from measurements (Liu, 1996). For theslash pine forest at the Gator National Forest site(GNF),

LAI(t)= 2. 25+0. 25xsin(--t+ L 1t365

(3)

For mature slash pine plantations, a maximum LAIand transpiration rate were assumed (Gholz et al.,1991):

MLAI(t)= 5.0+0.83xsinI--t+it' 365

(4)

For the cypress wetlands at the research site (Liu,1996):

LAI(t) = 1.2 If t < 60

LAI(t) = 1.2 + (3.93 - 1.20)!(121 - 61) t = 1.2 + 0.0455 t

LAJ(t)=L5+4.4xsin(.Tht+1.97cJ

If 60< t <121

Ift> = 121

Figure 3. FLATWOODS Model Flow Chart Showingthe Relationships Between the Submodels.

Cypress forests at the sites of the present studywere assumed to be in the mature stage with themaximum leaf area index.

Evaporation of intercepted water on forest canopieshas first demand on PET. Residual potential evapo-transpiration (RET) is the difference between PETand intercepted water evaporation from plant sur-faces. Actual evaporation (AE) from soil!water sur-faces is dependent on atmospheric demand, soil waterconditions, and is affected by forest canopy shading.Actual transpiration (AT), which involves physical



F1 = 1.0

CT

F• = i_10i,f_0i')IOi,191,w)

and physiological processes, is the most difficult com-ponent to model. In the FLATWOODS model, ATextracted from each of the soil layers is a function ofpossible realized transpiration (PRT), soil water con-ditions, and root density. The concept of PRT isdefined as the maximum transpiration that a crop canhave for a certain atmospheric condition and LA!.PRT is a function of the residual potential evapotran-spiration (RET), and stage of plant development indi-cated by LAI and root density.

Solar energy is first consumed by evaporation ofthe rain water intercepted on the forest canopy beforereaching the ground. Residual potential evapotranspi-ration (RPET) has been defined as the differencebetween PET and SIN. Evaporation from the soil sur-face is also reduced by forest canopy shading(McCarthy et al., 1992). Field data suggest that evap-oration from a floating pan under the cypress wetlandcanopies of the study area was only about 30 percentof the standard pan evaporation (Liu, 1996).

PE = RPET x exp [-KE1 x LAI(t)] (5)

where PE = potential evaporation (mm/day); RPET =residual potential evapotranspiration (mm/day) =PET - SIN; LAI(t) = leaf area index on Julian day t;and KE1 = soil evaporation reduction coefficient.

Actual evaporation (AE) is modeled as a function ofwater table depth (WDEP) in two phases:

Phase 1: Atniösphere DependentAE=PE IfWDEP<35cm

Phase 2: Water Table DependentAE = PE x KEC x [(100h)/(10035)JKE2

If 35 cm <WDEP < 100 cm

Phase 3: No EvaporationAE=0 IfWDEP>lOOcm

where KEC and KE2 are parameters.The assumption was made that transpiration will

not be limited as long as the soil moisture content ishigher than the field capacity, and plant roots of bothcypress and slash pine plantations do not stopextracting water when under inundation (Fisher andStone, 1990). Potential transpiration (PT) is the dif-ference between RPET and AE, representing the max-imum available energy for plant transpiration.

PRT = PTx LAI(t) x TRD(6)

MLAI(t) MTRD

where PRT = potential realized transpiration asexplained in the previous section (mm/day); TRD =

root density of the entire soil profile of the modeledforest (m/m2); and MTRD = root density when the for-est reaches an age with maximum transpirationcapacity (m/m2).

The actual transpiration (AT) from each of thethree layers depends on PRT, root density, and thesoil moisture content of each layer:

AT=EAT i=1,2,3

AT1 WEIGHT(i) x PRT x F

WEIGHT (i) = RD11TRD

where RD1 = root density of the layer i (m/m2);WEIGHT1 = weighing factor that distributes the totaltranspiration among the soil layers according to theroot density distribution; and

If 9> =

_______ Ife <

where F1 = coefficient to reflect the effects of soil mois-ture content, ranging from 0-1; CT = empirical param-eter; O = soil moisture content in layer i; O,f = soilmoisture content at field capacity in layer i; andsoil moisture content at the wilting point in the layer

Unsaturated Water Flow

Due to the high infiltration rate of sandy soils,rainfall that is not intercepted by forest canopies infil-trates rapidly into the soils without much overlandflow. A maximum of three unsaturated soil layershave been used to simulate subsurface unsaturatedwater flow. The first layer (0-40 cm) represents the Ahorizon where most plant roots reside. The secondlayer (40-65 cm) represents the spodic horizon (Bh)where soil properties (soil porosity and saturatedhydraulic conductivity) are distinct from the top layer.The third layer ranges from the 65-cm depth to thewater table. The thickness of each layer variesthroughout the simulation depending on the watertable depth. For example, if the water table comes tothe soil surface, the whole soil profile is saturated andthe number of unsaturated soil layers becomes zero; ifthe water table depth is greater than 40 cm but lessthan 65 cm, two unsaturated soil layers are stipulatedwith the first layer as 0-40 cm and the second as 40cm to the water table level. Drainage representing



downward unsaturated water flow from an upperlayer to a lower layer is estimated by Darcy's equationassuming unit total potential gradient. Upward waterflux represents water flow from a lower layer to anupper layer, driven by the water potential gradientsinduced by ET. This component is calculated in directproportion to the ET flux in the layer. While evapora-tion from the soil surface takes place only from thefirst layer, plant roots extract water from all threeunsaturated layers and the saturated zone. Soil mois-ture content in each layer is calculated by the waterbalance for each time step. For some areas, such aswetlands, the soil profile may be fully saturated dur-ing part or all of the year. Percolation from the bottomof the third layer of the unsaturated zone becomes theinput (source) to the underlying saturated subsystem.

Water flow in the unsaturated zone is essentiallyvertical due to the low topographical gradient of flat-woods. Lateral unsaturated hydraulic gradients mayincrease at the pond margin areas, where surfacewater/ground water interactions take place.

The procedure for soil moisture routing and unsat-urated water flow is based on the soil water balance:

o-'= +(D:_l_E:_t_D:)xAt (7)

where O4.1 = unknown water content expressed in thelayer i, day t+1 (cm); O = known water content at thelayer i, day t (cm); D1 = drainage rate at the layeri-i, day t; if layer i is the top layer, D1 = infiltration(cm/day); D = drainage rate at layer i, day t (cm/day);

= if the layer i is the first layer, it represents theevaporation rate from layer 1 — otherwise, it is theupward flux from the layer i to i-i due to water poten-tial differences between the two layers (cm/day); T =transpiration rate from the layer i (cm/day); and i\t =time step (one day).

The expression for the drainage rate (Di) may bederived from Darcy's Law (Equation 8), where h is thematric potential uniquely related to the water content0:

au (ahq = —K(h)— =—K(h)I —+ i) =—K(h)

az(8)

Assuming the water matric potential is uniform foreach layer, then ah/az = 0. The drainage rate or perco-lation rate can then be approximated as unsaturatedhydraulic conductivity K(h). The flux of q can also beapproximated by K(h) corrected by a coefficient ofL5. By combining the relationship of k(h)—.h and therelationship between Se, 0, and h that are developedby Van Genuchten (1980), a new equation that relates

soil moisture content and percolation rate [D(9)] wasderived as in Equation (9).

D(o)=K(e)=KsS2[1_S ls_h/m_m 2

ee i) ](9)

where Se = effective degree of water saturation (%);K8 = saturated hydraulic conductivity (cm/day); m =1-un, where n is a constant determined by fitting thesoil moisture release curve; and = volumetric soilmoisture content (%).

The upward flux, E or exfiltration rate in layer i iscontrolled by the vaporation at the soil surface andthe transpiration from the layer i and those above it.In the present model, E is modeled by:

E =CETx7't (10)

where CET = model input parameter; and T = tran-spiration from the layer i.

Ground Water Flow

The base of the unsaturated zone becomes theupper boundary of the saturated zone. The bottom ofthe saturated zone is the restricting clay layer about2 m from the soil surface. Below this flow-restrictingclay layer with hydraulic conductivity < iO rn/day,which may be discontinuous in extent, often liesanother intermediate aquifer composed of sands andsandy barns. Vertical flow (leakage) through the bot-tom of the saturated zone is estimated using anempirical function. Within the saturated zone, watermoves horizontally from one cell to the surroundingfour cells governed by a 2-D groundwater flow modelwith Dupuit assumptions (Bras, 1990). The twoimportant parameters in the groundwater flow equa-tion (Equation 11), specific yield and hydraulic con-ductivity, are not constant but vary depending on theposition of the water table in the soil profile. A watersource for the submodel is water percolation from theunsaturated zone. A sink of the ground water flowmodel includes ET extracted from the aquifer, exfil-tration (upward flux) from the saturated zone to theunsaturated zone and/or surface flow from those cellswhere the water table is above a critical elevation,and deep seepage from the bottom of the saturatedzone.

The 2-D model for ground water flow in an uncon-fined aquifer was adopted to represent the saturatedflow system and constitutes the core of the FLAT-WOODS model.



a(Kh)a(Kh)ax a — at

where K, K = hydraulic conductivity along the hori-zontal X axis and Y axis, calculated as the average ofthe aquifer thickness and varying with water tableelevation (rn/day); h = hydraulic head (m); W = waterflux representing sources (e.g., drainage from unsatu-rated layers) and sinks (e.g., ET, surface runoff, deepseepage)(m3/day); t = time (day); and S,, = specificyield of the aquifer — it is not a constant, but varieswith water table elevation.

To solve the above equation, boundary conditionsand initial conditions must be first specified. For aclosed system, such as a watershed, the boundaryconditions at watershed ridges can be set as a no-flowboundary. For a plot without lateral physical bound-aries, a constant gradient boundary may be appropri-ate, especially for flat homogeneous terrain. Initialconditions included spatial distributions of the soilmoisture content and corresponding hydraulic headfor each cell. The initial soil moisture content wasautomatically estimated by a generalized relationshipbetween the water table depth and the soil moisturecontent of the unsaturated layers (Sun, 1995).

Since Equation (11) is only applicable to the groundwater flow in porous media, problems may arise whenit is applied in the wetland-upland landscape undercertain circumstances.

Case 1: Surface Water - Ground Water Interac-tion. This case can be found at the margins of acypress pond when the area of a grid cell used in themodel is smaller than a certain specific wetland/pondarea. Under this condition, Darcy's law still holds forwater flow between adjacent cells. However, the spe-cific yield of the cell in the wetland may be set to 1.0.

Case 2: Surface Water - Surface Water Interac-tion. This case applies to the situations when thesizes of two adjacent grid cells are smaller than a cer-tain wetland/pond area. In this case, the surface flowdominates the entire pond system and the groundwater flow model is not valid. However, anapproximation is justified by setting S,, = 1.0 and Ks =50 rn/day (Walters and Bengtsson, 1994).

Surface Flow

Surface flow may occur from a grid cell underextreme wet conditions when the soil profile is satu-rated and the water table elevation is higher than thespecified critical elevation. This situation happened

in both wetlands and uplands areas of the study sites.The critical elevation is defined as the surface eleva-

(11) tion above which overland flow occurs. For uplands, itwas the topographical elevation. For cells in cypressponds, the critical elevation was the minimum eleva-tion of the palmetto line at the flow outlet. Palmettoline indicates an abrupt vegetation change from pal-metto to cypress at the edge of a cypress wetland. Thetotal daily runoff from the watershed was first simu-lated as a function of the average water table level,then runoff leaving a specific flooded cell is modeledas a function of the total saturated area in the entirewatershed and the water table elevation of that cell.

Based on the Variable Source Area Concept(Hewlett and Hibbert, 1965), the surface runoff leav-ing a flooded cell is modeled proportionally to thetotal saturated area of the entire watershed. The totaldaily runoff from the watershed was first simulatedas a function of the average water table level, thendistributed proportionally to each cell that had over-land flow.

Sj(t) = Ki x exp [h(t)- K3 ]K2 If h(t) � CWT

Sf(t) = Ri x [h(t)-K3}/[CWT-K3] If K3 <h(t) < CWT

S1(t) = 0 Otherwise

where Sf(t) = daily total surface runoff from the entirewatershed (mm/day); Ri, K2 = parameters calibratedby measured data; h(t) = average water table eleva-tion (m); CWT = critical water table elevation Cm) =Average Topographical Elevation of the watershed -Critical Soil Depth = ATOPO - CSD; CSD = the soildepth above which the runoff occurs at the outlet ofthe watershed (m); and K3 = the water table elevationbelow which no runoff occurs from the entire water-shed (m).

Surface runoff from each flooded cell, a sink termin Equation (ii), was modeled by:

SC1 (t) = 51(t) x AITSA (12)

where A1 = the surface area of the cell Cm2); andT5A(t) = total saturated area (variable source area)Cm2).

The deep seepage component, one of the sink termsin Equation (ii), is defined as the water that leaksfrom the bottom of the water table aquifer to thelower and sometimes artesian aquifer. The deep seep-age loss was assumed to follow a power function of thethickness of the surficial aquifer:

DS(t) = Kd x [h(t) - BOT]2 (13)



where Kd = a model parameter; and BOT = bottomelevation of the surficial aquifer of a cell (m).

MODEL INPUTS AND OUTPUTS

Compared to lumped models, distributed modelslike FLATWOODS require detailed spatial informa-tion about a watershed. Such information is generallyacquired through GIS. There are 10 input filesrequired to run the FLATWOODS model and it gener-ates two output files. Detailed information regardingmodel input requirements and outputs are listed inTable 1 and Table 2, respectively.

MODEL CALIBRATION AND VALIDATION

The FLATWOODS model was calibrated and vali-dated with hydrologic data collected from the GNF(Figure 1) and the Bradford Forest study sites (Fig-ure 4). Detailed information about experimentalinstallation and measurement data from the two siteshave been documented in Sun (1995). A combinationof statistical and graphical methods was employed toquantifr differences between simulated and measuredvariable series. The Pearson Correlation Coefficientand the objective function (index of disagreement)were used to evaluate the model performance andobtain optimum parameters. Soil parameters such as

TABLE 1. A List of Inputs Required to Run the FLATWOODS Model.

Model Inputs

Climatic Data

Watershed Configuration

Soil Parameters

Vegetation Parameters

Initial Conditions

Variables/Parameters

— Daily rainfall— Daily average air temperature— Latitude of the study site— Correction factor for calculating potential evapotranspiration by Hamon's method— Empirical function for rainfall interception

— Grid size, width and length, and area— Landuse type (cypress wetlands, pine upland, harvested wetland, and harvested upland— Topographical elevation and bottom elevation of each cell; the critical elevation of each cell— Boundary condition identifiers

o = no flow1 = variable hydraulic head

-1 = constant gradient boundary

— Soil moisture characteristic curves by Van Genuchten's method— Saturated hydraulic conductivity of each soil layer in each cell;— Specific yield of each soil layer in each cell— Relation between soil moisture content and water table level for different soil layers— Empirical parameters used in the ET submodels— Empirical parameters used to calculate surface runoff and deep seepage

— Leafarea index functions for each forest type and their change with time— Root density for each forest type in each soil layer (m rn-2)

— Hydraulic head (water table elevation) in each cell throughout the watershed at thebeginning of a simulation.

TABLE 2. A List of Outputs From the FLATWOODS Model.

— Daily water table elevation in each grid cell and its average over the entire watershed

— Daily rainfall interception, evaporation, and transpiration from each cell

— Daily total surface runoff and groundwater flow across the boundaries from the entire watershed

— Dailywater drainage (percolation) from the unsaturated zone to the saturated zone

— Daily soil water storage in each soil layer

— Daily deep seepage from the surficial aquifer to the underlying second aquifer

— Statistical analysis on model performance

JOURNAL OF THE AMERICAN WATER RESOURCES AssociATioN 835 JAWRA


r'I,

a

(I)

0

0

0I—

0N

8Q

0 >.

:eo(0 't) - .- 0


(0

tN

t)

0

c1

N

(0

a0

Cl

Cl

Cl)-C

0•Cl

'2)

>,U)

-Cl

-ClaNClC

(0>< a

ClCl

C00r..)

00 O N0

(0 u) P' C1 .- 0W3SIS PJO — A


specific yield (S7) and coefficients for evapotranspira-tion are the major calibrated parameters.

Gator National Forest (GNF) Site

More than 130 shallow wells were installed at theGNF site and the water table levels in each well weremeasured bi-weekly from 1992 to 1995 (Crownover etal., 1995). Harvesting treatments were imposed fromApril 5 to May 31, 1994. Therefore, two years of pre-harvest and one year of post-harvest data are avail-able for model testing. The arithmetic average of allwater table elevations from each measurement wasused for the model calibration. The model was cali-brated with water table data for the wet year of 1992and a dry year 1993, which had a 170 mm surplusand 230 mm deficit of rainfall compared to normalyears (rainfall =1330 mm), respectively. Satisfactoryresults were obtained for those two years (Figure 5a).During May-October of 1993, 6-93 percent of theobservation wells were dry due to extremely low rain-fall. Using both a dry and a wet year for the modelcalibration enhanced the generality of the model forfuture prediction. Model validation with data fromJanuary 1, 1994, to May 31, 1994, during the pre-treatment period also showed good model perfor-mance (Figure 5a). The Pearson CorrelationCoefficient was 0.91 and 0.96 for the calibration peri-od and the validation period, respectively.

Harvesting in May 1994 was followed by doublebedding activities in the fall of 1994. The southeastblock was totally clear cut, including both cypresswetlands and uplands, but only the wetlands wereclear cut in the northwest block of the research area.The harvested upland areas in the southeast blockwere planted with slash pine seedlings in January of1995 and the cypress wetlands was left for naturalregeneration. Apparently, the most significant effectof the forest harvesting on model parameters was thereduction of LA!. This was assumed to be reduced to0.5 for harvested wetlands and 0.1 for uplands underclear-cut conditions. Soil structure of the first layeralso might have been altered due to compaction bythe mechanical operations, but the change was diffi-cult to parameterize quantitatively and thus wasneglected. Under these assumptions, the model wascalibrated and validated with post-treatment data col-lected during June 1, 1994-May 31, 1995 (Figure 5b).The Pearson Correlation Coefficient was 0.88 and0.82 for model calibration and verification periods,respectively.

Bradford Forest Watershed

The Control Watershed at the Bradford Forest inBradford County, Florida has been monitored since1978 as one of the most comprehensive efforts of for-est hydrological studies in the southeastern UnitedStates (Pratt, 1979; Riekerk, 1989). Spatial watertable distribution data were not available for thisstudy. Instead, a long-term data series of runoff atthe watershed outlet was used for model calibrationand validation. The FLATWOODS model was cali-brated with five years runoff data (1978-1982) on ayear by year basis (Sun, 1995). Several water tabledata sets from individual wells were used as a guidefor calibrating the average water table elevationacross the watershed. The year 1978 was a wet year(1453 mm rainfall) while the year 1981 was a very dryyear (916 mm rainfall). The model was validated withrunoff data from 1983 to 1992 using the same param-eters for the five-year calibration (Figure 6). Themodel did not predict the pike flows very accurately inwet years (1983 and 1992) with low Pearson Coeffi-cient of 0.61 and 0.62, respectively. Two reasons havebeen hypothesized: (1) the boundary and outlet ditch-es in this artificially created watershed may generatehigher peak flows during the wet seasons, especiallyin extreme years (Iritz et al., 1994); and (2) the FLAT-WOODS model used a single runoff-water table levelrelationship independent of time to predict runoff,and no cell-by-cell surface flow routing procedures.However, the assumption made in the runoff-ground-water table relationship proved effective since themodel fitted the low flow reasonably well during thetwo-year drought period of 1989-1990. Using morephysically based models such as manning's equationand unit hydrograph, or incorporating daily rainfallin the water table-runoff relations may improvemodel performance for stormflow events.

SUMMARY AND CONCLUSIONS

Based on the existing COASTAL model, a new dis-tributed forest hydrologic model, FLATWOODS, wasdeveloped. The FLATWOODS model was specificallydesigned to simulate the hydrologic processes inforested wetland-upland systems, where saturatedand unsaturated soil zones coexist and the surfacewater and subsurface water interact. Vertically, themodel divided a flatwoods hydrologic system intothree connected subsystems: an ET subsystem, anunsaturated water flow subsystem, and a groundwater flow subsystem. Laterally, the model dividesthe entire watershed into small rectangular cells


30

29.5

29

04-.Co>a)a)a)-aCo4-.I-

27.5

28.5

1994 Calibration

Post-treatment

1995 Validation

>.

I

Figure 5. FLATWOODS Model Calibration and Validation During the Pro-Harvest (a)and Post-Harvest (b) Periods for the GNF Site.



Pre-treatment

28

>.

I-

27

1992MJJASONDJFMAMJJASONDJFMAM

19941993

30

29.5

29

28.5

28Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May

0

>Co

E

0C

which serve as the simulation units. The presentmodel employs deterministic equations to calculatethe water balances of each subsystem on a daily timestep with readily available climatic data and measur-able parameters. The FLATWOODS model has thecapability to predict daily groundwater table levels,evapotranspiration, soil water storage in up to threelayers, and total daily runoff from a watershed sub-ject to various boundary conditions.

Model validation using both daily runoff and watertable suggests the model can simulate the generalhydrologic processes on flatwoods landscape with suf-ficient accuracy. The FLATWOODS forest hydrologi-cal simulation model provides an alternate tool tostudy the hydrology of pine flatwoods ecosystems. Byusing this model, impacts of different forest manage-ment scenarios such as various harvesting schemesmay be evaluated. Also, this model can be used tosimulate the dynamics of long-term wetland hydro-period.

As any other existing watershed scale hydrologicalmodel, the predictability of the present model is heav-ily dependent on field testing with measured data.Empirical equations that simulate surface runoff inthe model may be further validated with newresearch data. Predictions of spatial features are alsoneeded to compare with measured data. Computer

simulation models cannot replace the field experimen-tation, but they are a complementary means toachieve the same goals. Field investigations areessential for model improvements. With the help ofGIS, model parameterization and interpretation ofmodel outputs become easier tasks for distributedmodel such as FLATWOODS (Maidment, 1993; Sunet al., 1998).

ACKNOWLEDGMENTS

This study was supported by the Intensive Management Prac-tice Assessment Center (IMPAC) at the University of Florida,Gainesville and the National Council for Air and Stream Improve-ment of the Paper Industry, Inc. (NCASI).

LITERATURE CITED

Al-Soufi, R., 1987. A Physically-Based Model for the AgrohydrologicProcesses. J. of Hydrology 93: 199-2 19.

Band, L. E., P. Patterson, R. Nemani, and S. W. Running, 1993.Forest Ecosystem Processes at the Watershed Scale: Incorporat-ing Hillslope Hydrology. Agricultural and Forest Meteorology63:93.126.

Bathurst, J. C. and P. E. O'Connell, 1992. Future of DistributedModeling: The Systeme Hydrologique Europeen. HydrologicalProcesses 6:265-277.



14

12

10

8

6

4

2

01978 79 80 81 82 83 84 85 86 87 88 89 90 91 92

Figure 6. FLATWOODS Model Calibration (1978-1982) and Validation (1983-1992)Using 15 Years Runoff Data From the Bradford Forest Site.


Bernier, P. Y., 1982. VASA2: A Revised Source Area Simulator forSmall Forested Basins. Ph.D. Dissertation, University of Geor-gia, Athens, Georgia, p. 143.

Beven, K. J., 1989. Changing Ideas in Hydrology —The Case ofPhysically-Based Models. Journal of Hydrology 105:157-172.

Bosch, J. M. and J. D. Hewlett, 1982. A Review of CatchmentExperiments to Determine the Effect of Vegetation Changes onWater Yield and Evapotranspiration. J. of Hydrology 55:3-23.

Brandt, K. and K. C. Ewel, 1989. Ecology and Management ofCypress Swamps: A Review. Florida Cooperative Extension Ser-vice, WAS, University of Florida, Gainesville, Florida, 20 pp.

Bras, R. L., 1990. Hydrology — An Introduction to Hydrologic Sci-ence. Addison-Wesley Publishing Company, p. 631.

Campbell, K. L., J. C. Capece, and T. K. Tremwel, 1995. Surface!Subsurface Hydrology and Phosphorus Transport in the Kissim-mae River Basin, Florida. Ecological Engineering 5:301-330.

Capece, J. C., 1994. Hydrology and Contaminant Transport on Flat-woods Watersheds. Ph.D. Dissertation, University of Florida,Gainesville, Florida, p. 225.

Crownover, S. H., N. B. Comerford, and D. G. Neary, 1995. WaterFlow Patterns in Cypress/Pine Flatwoods Landscapes. Soil andSci. Soc. Amer. J. 59:1199-1206.

Cubbage, F. W. and C. H. Flather, 1993. Forested Wetland Area andDistribution: A Detailed Look at the South. J. For. 91:35-40.

Ewel, K. C. and H. L. Gholz, 1991. A Simulation Model of the Roleof Belowground Dynamics in a Florida Pine Plantation. ForestScience 37:397-438.

Ewel, K. C. and J. E. Smith, 1992. Evapotranspiration From Flori-da Pond Cypress Swamps. Water Resources Bulletin 28:299-304.

Federer, C. A. and D. Lash, 1978. BROOK: A Hydrologic Simula-tion Model for Eastern Forested. Water Resources ResearchCenter, University of New Hampshire, Durham, New Hamp-shire, Research Report 19, Mpp.

Fisher, H. M. and E. L. Stone, 1990. Air-Conducting Porosity inSlash Pine Roots From Saturated Soils. Forest Sciences 36:18-33.

Gholz, H. L. S. A. Vogel, W. P. Cropper, Jr., K. McKelvey, and K. C.Ewel, 1991. Dynamics of Canopy Structure and Light Intercep-tion in Pinus elliottii Stands, North Florida. Ecological Mono-graphs 61:33-51.

Colkin, K. R. and K. C. Ewel, 1984. A Computer Simulation of theCarbon, P, and Hydrologic Cycles of a Pine Flatwoods System.Ecological Modeling 24:113-136.

Guo, W., 1989. Simulation of Stormflow From a Pine FlatwoodsWatershed. M.S. Thesis, Univ. of Florida, Gainesville, Florida,p.29

Hamon, W. R., 1963. Computation of Direct Runoff Amounts FromStorm Rainfall. mt. Assoc. Sci. Hydrol. Pub. 63:52-62.

Heatwole, C. D., K. L. Campbell, and A. B. Bottcher, 1987. ModifiedCREAMS Hydrology Model for Coastal Plain Flatwoods.TRANSACTIONS of the ASAE 30:1014-1022.

Hewlett, J. D. and A. R. Hibbert, 1965. Factors Affecting theResponse of Small Watersheds to Precipitation in Humid Areas.In: International Symposium on Forest Hydrology, W. E. Sopperand H.W. Lull (Editors). Pergamon Press, Elmsford, New York,pp. 275-290.

Iritz, L., B. Johansson, and L. Lundin, 1994. Impacts of ForestDrainage on Floods. Hydrological Sciences 39:637-661.

Jensen, K. H. and A. Mantoglou, 1992. Future of Distributed Mod-elling. Hydrological Processes 6:225-264.

Knisel, W. G. (Editor), 1980. CREAMS: A Field-Scale Model forChemicals, Runoff, and Erosion From Agricultural ManagementSystems. Conservation Research Report No. 26, USDA, Wash-ington, D.C., p. 643.

Liu, 5. 1996. Evapotranspiration From Cypress (Taxodium ascen-dens) Wetlands and Slash Pine (Pinus elliottii) Uplands inNorth-Central Florida. Ph.D. Dissertation, University of Flori-da, Gainesville, Florida, p. 258.

Lovejoy, S. B., J. G. Lee, T. 0. Randhir, and B. A. Engel, 1997.Research Needs for Water Quality Management in the 21st Cen-tury: A Spatial Decision Support System. J. of Soil and Conser-vation 52(1):18-22.

Maidment, D. R., 1993. GIS and Hydrologic Modeling. In: Environ-mental Modeling With GIS. M. F, Goodchild, B. 0. Parks, andL. T. Steyaert (Editors). Oxford University Press, New York,New York, pp. 147-167.

McDonald, M. G. and A. W. Harbaugh, 1984. A Modular ThreeDimensional Finite-Difference Groundwater Flow Model. U.S.Geological Survey, National Center, Reston, Virginia, 510 pp.

McCarthy, E. J., J. W. Flewelling, and R. W. Skaggs, 1992. Hydro-logic Model for Drained Forest Watershed. J. of Irrigation andDrainage Engineering 118:242-255.

Pierce, R. S., F. H. Bormann, and G. E. Likens, 1987. The Ecosys-tem Approach to Forest Hydrological Research. In: ForestHydrological Resources in China; An Analytical Assessment,P. F. Ffolliott and D. P. Guertin (Editors). MAB Workshop,Harbin City, China, pp. 92-98.

Pratt, D. A., 1978. The Hydrology of a Mixed Pine and CypressCommercial Forest in Bradfor County, Florida. M. S. Thesis,University of Florida, Gainesville, Florida.

Riekerk, H., 1989. Influence of Silvicultural Practices on theHydrology of Pine Flatwoods in Florida. Water ResourcesResearch 25:713-719.

Riekerk, H. and L. V. Korhnak, 1984. Environmental Effects of Sil-viculture in Pine Flatwoods. In: Third Biennial Southern Silvi-cultural Research Conference, E. Shoulders (Editor). U.S. ForestService, General Technical Report, S0-54:528-535.

Sabine, J. B., 1994. Application of Minor Drainage on Non-Industrial Private Lands. In: Proceedings Water Management inForested Wetlands. USDA Forest Service, Southern Region,Atlanta, Georgia, 158 pp.

Shepard, J. P., L. A. Lucier, and L. W. Haines, 1993. Industry andForest Wetlands: Cooperative Research Initiatives. J. of For.91(5):29-33.

Skaggs, R. W., 1984. DRAINMOD — A Water Management Modelfor Shallow Water Table Soils — User's Guide. North CarolinaState University, Raleigh, North Carolina, p. 54.

Sun, C-H., 1985. COASTAL-A Distributed Hydrological Model forLower Coastal Watershed in Georgia. Ph.D. Dissertation, Uni-versity of Georgia, Athens, Georgia.

Sun, G., H. Riekerk, and L. V. Korhnak, 1995. The Hydrology ofCypress Wetland-Upland Ecosystems in Florida Flatwoods. In:Versatility of Wetlands in the Agricultural Landscape, K. L.Campbell (Editor). American Association of Agricultural Engi-neering, pp 489-500.

Sun, G., 1995. Measurement and Modeling of the Hydrology ofCypress Wetlands — Pine Uplands Ecosystems in Florida Flat-woods. Ph.D. Dissertation, University of Florida, Gainesville,Florida, pp. 360.

Sun, G., H. Riekerk, L. A. Arvanitis, and V. Valavanis, 1998. Inte-grating a Hydrological Simulation Model With GIS to StudyEnvironmental Effects of Forest Harvesting. Geolnfo Systems8(4):32-36.

Swank, W. T. and D. A. Crossley, Jr. (Editors), 1988. Forest Hydrol-ogy and Ecology at Coweeta. Ecological Studies. Springer-Ver-lag, New York, Vol. 66, pp. 297-3 12.

Swank, W. T., S. G. McNulty, and L. W. Swift, 1994. Opportunitiesin Forest Hydrology Research From Broad Environmental Per-spectives. In: Proceedings of' the International Symposium onForest Hydrology, T. Ohta (Editor). University of Tokyo, Tokyo,Japan, pp. 19-29.



Thomas, D., 1989. The ANSWERS Model, Forest Hydrology Ver-sion. In: Application of Water Quality Models for Agriculturaland Forested Watersheds, D. B. Beasley and D. L. Thomas (Edi-tors). Southern Cooperative Series Bull. No. 338:39-52.

Tremwel, T. K. and K. L. Campbell, 1992. FHANTM, a ModifiedDRAINMOD: Sensitivity and Verification Results. ASAE PaperNo. 92-2045, American Society of Agricultural Engineers, St.Joseph, Michigan.

Van Genuchten, M., 1980. A Closed-Form Equation for Predictingthe Hydraulic Conductivity of Unsaturated Soils. Soil Sci. Soc.of Amer. J. 44:892-898.

Walters, M. and T. Bengtsson, 1994. Application of a GroundwaterModel to Simulate Water Quality in a Florida Lake. HydrologicBudget for a Florida Marsh. In: Lake and Reservoir Manage-ment, Vol 9, No. 2, 14th Annual International Symposium,Orlando, Florida.

Wigmosta, M. S., 1991. Modeling and Monitoring to Predict Spatialand Temporal Characteristics in a Small Catchment. Ph.D. Dis-sertation, University of Washington, Seattle, Washington. pp.183.

Wigmosta, M. S., L. W. Vail, and D. P. Lettenmaier, 1994. A Dis-tributed Hydrology-Vegetation Model for Complex Terrain.Water Resources Research 30: 1665-1679.

Yan, J. and K. R. Smith, 1994. Simulation of Integrated SurfaceWater and Ground Water Systems — Model Formulation. WaterResources Bulletin 30:879-890.

JOURNAL CF THE AMERICAN WATER RESOURCES AsSOCIATION 841 JAWRA

Documents

MODELING THE FOREST HYDROLOGY OF WETLAND-UPLAND ECOSYSTEMS IN FLORIDA