Ontology-based simulation of water flow in organic soils applied to Florida sugarcane

Abstract

An ontology-based simulation (OntoSim) is a unique data modeling environment where soil-plant-nutrient processes are represented as database objects and the user-defined relationships among objects are used to generate computer code (Java) for running the simulation. The aim of this study was to model hydrologic processes of sugarcane-grown organic soils utilizing OntoSim in the Everglades Agricultural Area (EAA) of South Florida. This OntoSim-Sugarcane model describes the complex hydrology of sub-irrigation and open ditch drainage commonly used on Florida farms. Model calibration was conducted by (i) selecting rectangular farm water management units (<12ha), which are encompassed with farm ditches, from two farms in the EAA, (ii) assembling all relevant input data including water tables (WT) recorded at the monitoring farm well of each unit, and (iii) optimizing the fits between the simulated and observed daily WT during two consecutive water years (WY). By calibrating two site-specific parameters - lateral saturated hydraulic conductivities of soil profiles and vertical saturated hydraulic conductivity of the underlying limestone bedrock - good agreement between simulated and observed daily WT was obtained (Nash-Sutcliffe efficiency coefficient >0.65; coefficient of residual mass <1%) within the units during WY96-97 (May 1995-April 1997). The validation of the model during subsequent WY98-99 at both units also showed Nash-Sutcliffe efficiency >0.55 and coefficient of residual mass <3%. It indicated that OntoSim-Sugarcane is able to simulate daily fluctuations of WT within the farm units and estimate lateral drainage/sub-irrigation and deep seepage that significantly contribute to the water balance at farms in the EAA. Thus, it can be a promising management tool to provide farmers with accurate assessment of water movement in this agricultural area.

Full text

Page 1

Ontology-based simulation of water flow in organic soils applied
to Florida sugarcane
Ho-Young Kwona,*, Sabine Grunwalda, Howard W. Beckb, Yunchul Jungb,
Samira H. Daroubc, Timothy A. Langc, Kelly T. Morgand
aSoil and Water Science Department, Institute of Food and Agricultural Sciences, University of Florida, Gainesville, FL 32611, United States
bAgricultural and Biological Engineering Department, Institute of Food and Agricultural Sciences, University of Florida, Gainesville, FL 32611, United States
cEverglades Research and Education Center, University of Florida, Belle Glade, FL 33430, United States
dSoil and Water Science Department and Southwest Florida Research and Education Center, University of Florida, Immokalee, FL 34142, United States
1. Introduction
The Everglades Agricultural Area (EAA), which is located
between Lake Okeechobee and the Everglades Protection Area
(EPA) in South Florida, has been established since late 1940s when
an extensive flood-control system consisting of canals, ditches and
water control structures was installed to convert wet or
submerged historic Everglades into productive croplands (Snyder
and Davison, 1994). Currently 200,000 ha of the EAA are cultivated
and approximately 82% of the cultivated area is planted to
sugarcane (Saccharum officinarum L.) followed by vegetables, rice
(Oryza sativa L.), and sod (Rice et al., 2002). Besides favorable
environmentalconditions includingfertile organicsoils (Histosols)
and subtropical climate (Baucum et al., 2006), Florida sugarcane
industries in the EAA have been sustained by seepage-based
drainage control systems where farm water tables (WT) are
managed to maintainthe water levelnormally <1 m below the soil
surface (Obreza et al., 1998). Such WT controls can be effective in
the EAA due to the facts that (i) Florida has a flat topography
(Snyder and Davison, 1994), (ii) its organic soils ranging from 0.30
to 3 m depth are underlain by an impermeable limestone bedrock
(Snyder, 1994), and (iii) over half of the average annual rainfall
(about 1360 mm y?1) falls during the wet season from June to
October (Ali and Abtew, 1999).
While the sugarcane industry currently ranks third in Florida
agricultural economy behind the greenhouse/nursery and citrus
industries (Baucum et al., 2006) due to these water control
Agricultural Water Management 97 (2010) 112–122
A R T I C L E I N F O
Article history:
Received 14 January 2009
Received in revised form 25 August 2009
Accepted 29 August 2009
Available online 25 September 2009
Keywords:
Ontology
Simulation
Hydrology
Everglades Agricultural Area
Water table
Drainage
A B S T R A C T
An ontology-based simulation (OntoSim) is a unique data modeling environment where soil–plant-
nutrient processes are represented asdatabase objects and theuser-defined relationships among objects
are used to generate computercode (Java) for running the simulation. The aim of this study was tomodel
hydrologic processes of sugarcane-grown organic soils utilizing OntoSim in the Everglades Agricultural
Area (EAA) of South Florida. This OntoSim-Sugarcane model describes the complex hydrology of sub-
irrigation and open ditch drainage commonly used on Florida farms.
Model calibration was conducted by(i) selecting rectangular farm water managementunits (<12 ha),
which are encompassed with farm ditches, from two farms in the EAA, (ii) assembling all relevant input
data including water tables (WT) recorded at the monitoring farm well of each unit, and (iii) optimizing
the fits between the simulated and observed daily WT during two consecutive water years (WY). By
calibrating two site-specific parameters – lateral saturated hydraulic conductivities of soil profiles and
vertical saturated hydraulic conductivity of the underlying limestone bedrock – good agreement
between simulated and observed daily WT was obtained (Nash–Sutcliffe efficiency coefficient >0.65;
coefficient of residual mass <1%) within the units during WY96–97 (May 1995–April 1997). The
validation of the model during subsequent WY98–99 at both units also showed Nash–Sutcliffe efficiency
>0.55 and coefficient of residual mass <3%. It indicated that OntoSim-Sugarcane is able to simulate daily
fluctuations of WT within the farm units and estimate lateral drainage/sub-irrigation and deep seepage
that significantly contribute to the water balance at farms in the EAA. Thus, it can be a promising
management tool to provide farmers with accurate assessment of water movement in this agricultural
area.
? 2009 Elsevier B.V. All rights reserved.
* Corresponding author. 2169 McCarty Hall, PO Box 110290, Gainesville, FL
32611-0290, United States. Tel.: +1 352 392 1951x239; fax: +1 352 392 3902.
E-mail addresses: hkwon@ufl.edu, kwon.hoyoung@gmail.com (H.-Y. Kwon).
Contents lists available at ScienceDirect
Agricultural Water Management
journal homepage: www.elsevier.com/locate/agwat
0378-3774/$ – see front matter ? 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.agwat.2009.08.019

Page 2

systems, numerous studies have also shown their adverse effects
on water quality in the EAA and downstream hydrologic units (e.g.
LOTAC II, 1990; Izuno et al., 1991). Since Histosols in the EAA have
been formed within a low-nutrient oligotrophic ecosystem that
contains low quantities of essential plant nutrients except
nitrogen, the soils often require fertilization for agriculture
production (Luo, 2004). This agricultural practice combined with
water control systems has contributed to nutrient enrichment of
the Everglades by releasing phosphorus (P) enriched farm drainage
water from the EAA to the EPA (Izuno and Bottcher, 1994; Sievers
et al., 2003). Thus, there is a need to better optimize water
managementinthe EAAconsideringitscomplexhydrologysoas to
reduce the environmental impacts associated with P fertilizer
application and P-enriched farm drainage water. Several studies
have been conducted to understand the hydrology of the EAA by
using mathematical modeling (Bottcher et al., 1998; Zhang and
Gornak, 1999), which is now widely employed by agronomists and
environmental scientists to understand complex soil–plant-
nutrient processes in various agroecosystems because unlike
empirical approaches using statistical analyses of historical
records it can be further applied to test management scenarios
and predict responses to future changes in climate.
Currently, most mathematical simulations are implemented in
aparticularprogramminglanguagelikeFORTRAN,C++,orJava,and
provide graphical user interfaces to run simulations. However,
there are multiple limitations to this traditional approach of
simulation modeling. First, mastery of a programming language is
required to develop and/or modify a simulation model. Second,
documentation is physically separate from the model implemen-
tation and sometimes does not accurately describe the model
implementation. Third, often written documentation of a simula-
tion model does not capture all details of program code so that
ultimately it is necessary to read computer code in order to truly
understand how the model works (Beck et al., 2008). To overcome
these limitations ontologies have received much attention for
implementing mathematical models and building simulation
systems in order to explicitly describe knowledge of a simulation
model (Lacy and Gerber, 2004; Cuske et al., 2005).
Ontologies are formal representations of the concepts and their
interrelationship within a particular domain (Beck et al., 2008). An
application of ontologies to modeling and simulation results –
OntoSim – has been presented by Beck et al. (2008). OntoSim
provides a unique data modeling environment where the processes
related to plant growth, soil moisture, and nutrient uptake are
representedasdatabaseobjectsandrelationshipsamongobjectsare
used to generate computer code (Java) for running the simulation.
OntoSim was originally developed to help Florida citrus growers
optimizing yields in an economically effective manner (Beck et al.,
2008). However, OntoSim can provide a platform to develop any
conceptual and mathematical simulation model.
This paper presents the use of OntoSim to model hydrology of
sugarcane farms on organic soils in the EAA of South Florida. Main
emphasis was made on modeling vertical and lateral drainage flux
inthesaturatedzonetofacilitatesimulationofWTcontrolsystems
commonly used on Florida sugarcane farms with sub-irrigation
and open ditch drainage. Calibration and validation of these
processes were conducted using farm data collected during a four-
year period, followed by a discussion about future applications of
OntoSim to other modeling environments.
2. Materials and methods
2.1. Ontology-based simulation applied to Florida sugarcane
Mathematical concepts and algorithms documented in forms of
publications and simulation models were utilized to build
ontological implementations of water flux on sugarcane-grown
farms resulting in a simulation model, called OntoSim-Sugarcane.
The processes for soil organic matter decay were based on the
mathematical framework of the CENTURY model (Parton et al.,
1988). Soil–water uptake and transpiration of sugarcane for
calculating evapotranspiration (ET) were described according a
modified version of the DSSAT/CANEGRO model (Inman-Bamber,
1994; Inman-Bamber and Kiker, 1997). Hydrologic processes for
simulating a perched WT and vertical and lateral drainage flux in
the saturated zone were adapted from DRAINMOD (Skaggs, 1980)
and other studies (Alexander, 1988; Reyes et al., 1993).
All the processes identified for simulating sugarcane growth on
Florida organic soils have been represented as mathematical
equations. Reverse engineering from program code into equations
was conducted in cases where processes were found in the form of
computer code from various models (e.g. DSSAT/CANEGRO). These
equations have been represented as database objects of OntoSim
using SimulationEditor and EquationEditor (Fig. 1a and b). First, to
describe the topological and thematic structure of OntoSim-
Sugarcane,fivecompartments–weather,soilprofileandlayer,and
sugarcane and sugarcane stalk – were created using the structure
editor,whichisthemaininterfaceofthe SimulationEditor(Fig.1a).
This tool provides functionalities which enable modeler to create
and maintain a simulation project by designing the structure of a
system (Beck et al., 2008). Then 195 equations and 247 symbols
associated with the elements in the structure diagram were
created using the EquationEditor (Fig. 1b), which uses an interface
similar to other equation editors such as Microsoft Office Equation
Editor, except that equations and symbols are stored internally as
ontology objects. Then computer code (Java) to run a simulation is
generated automatically from the equations and symbols. Finally,
the simulation model was debugged for errors to run successfully.
2.2. The hydrologic processes for water table control
systems in the EAA
The drainage systems in the EAA feature extensive networks of
farm canals, ditches, and pump stations that are managed by both
the South Florida Water Management District (SFWMD) and
growers. The district manages the public canals and its own pump
stations in the EAA while growers manage water levels on their
farm drainage basins within the EAA basin. Normally, a main farm
canal runs from the farm pump station to the far reaches of the
farm, and sub-mains or farm laterals branch off the main canal at
right angles, generally on 800 m spacing, on section and half
section boundaries (Izuno, 1994). Emanating at right angles from
thefarmlateralsareequallyspacedfieldditches,whichareparallel
and subdivide the farm into rectangular areas with nominal
dimensions of 200 by 800 m. These 16 ha blocks are considered the
basin water management unit where sub-irrigation or open ditch
drainage practices are accomplished by either raising or lowering
field ditch water levels (Izuno, 1994).
In this work, such concept of farm water management unit was
adaptedandasmallfarmunit(<12 ha)wasselectedfromanentire
farm area as the hydrologic modeling unit where the hydrologic
processes, including continuous calculation of soil moisture
contents, actual evapotranspiration (AET), upward water flux,
infiltration, and drainage loss according to the water table
simulated for each process, were simulated.
2.2.1. Distribution of soil moisture
Water tables on fields underlain by a restrictive layer have been
simulated by various hydrologic models such as DRAINMOD,
ADAPT (Agricultural Drainage and Pesticide Transport) (Alexander,
1988), and GLEAMS-WT (Groundwater Loading Effects of Agri-
cultural Management Systems-Water Table) (Reyes et al., 1993).
H.-Y. Kwon et al./Agricultural Water Management 97 (2010) 112–122
113

Page 3

These models have a common feature that the soil moisture
distribution within the soil profile is assumed to be at a ‘‘drained to
equilibrium’’ condition. In this case, the moisture content for each
soillayercanbedefinedbyanempiricalfunctionthatdescribesthe
soil–water characteristic relationship between soil–water content
(u)andthecorrespondingsoilmatrixpotential(c).Theequationof
Van Genuchten (1980) was adapted to estimate moisture
distribution within the soil profile:
uðcÞ ¼ urþ ðus?urÞ
1
1 þ jacnj
??m
(1)
where urand usare the residual and saturated soil–water content
(cm3cm?3), respectively;c is the soil matrix potential or capillary
Fig. 1. The system structure of the OntoSim-Sugarcane and equations and symbols created using the SimulationEditor (a) and EquationEditor (b).
H.-Y. Kwon et al./Agricultural Water Management 97 (2010) 112–122
114

Page 4

pressure head (cm) that is equal to the distance from soil layer to a
water table under hydrostatic equilibrium; and the empirical
parameters a (cm?1), n and m are estimated by the fitting of this
function and the data measured for the soil–water retention
relationship. Here it is assumed that m = 1 ? 1/n. An example in
Fig. 2a shows the distribution of soil moisture (curve A1–B1–C1)
corresponding to the depth of WT below the soil surface (WTD1) in
cm using Eq. (1).
2.2.2. Calculation of actual evapotranspiration and upward flux
The soil moisture distribution calculated is allowed to deviate
from steady-state due to the removal of water from the profile by
ET, resulting in a root zone depleted of water. This deviation
implies that an upward gradient is induced between the water
table and the depleted root zone and thus water may move
upwards in the soil profile by upward flux in response to this
gradient. This upward movement of water is estimated by
assuming that a steady-state condition exists between the water
table and an evaporating surface. A similar approach as presented
by Martinez (2006) using the algebraic solution of Anat et al.
(1965) was used to estimate maximum upward flux (UFLUXin cm):
UFLUX¼ VKSAT 1 þ1:886
h2þ 1
??h
RD
hb
???h
(2)
where VKSATis vertical saturated hydraulic conductivity (cm h?1);
h = 2 + 3n(1 ? 1/n); RD is root zone depth (cm) that changes with
time during the growing season and is determined by using the
bottom of a root zone as the upper boundary; and hbis often
referred to as bubbling pressure (cm) and determined from the
relationship between effective saturation and c (Skaggs, 1980).
The actual upward flux is taken as the smaller one of either
UFLUXor ET calculated as a function of potential ET [the FAO (Food
and Agriculture Organization of the United Nations)-56 Penman–
Monteith equation] (Allen et al., 1998) and model-simulated leaf
area index and root length density (Jones et al., 2003).
If UFLUXmeets ET demand, actual upward flux is equaled to ET.
This results in the increase of soil air volume (or the decrease of
soil–water volume) by the upward flux within a soil profile,
causing a new WT to recede from the previous WT. The new WT is
estimatedusingthe relationshipbetween soilairvolume withinan
entire soil profile and the new WTD. For example, the soil air
volume for each layer, which can be calculated using Eq. (1) at
specific WTD, is summed up within the entire soil profile to give
specific total air volume at the specific WTD. By repeating this
procedure, the relationship between soil air volume and various
WTDs can be derived and modeled using a simple equation that
can be used to estimate fluctuation of WT according to
correspondingchangeofsoil–watervolumeduetoETanddrainage
loss:
AirV ¼ a1WTD2þ a2WTD(3)
where AirV is the air volume (cm) in the entire soil profile, and a1
and a2are coefficients.
Accordingly, a soil moisture distribution is updated based on
the new WTD and the moisture in the root zone can be still under a
steady-statecondition.ThecurveA2–B2–C2ofFig.2billustratesthe
case when UFLUXmeets ET demand and thus WTD1becomes WTD2.
If UFLUXfails to meet ET demand, actual upward flux is equaled
to UFLUXand a new WTD is calculated using the same procedure as
used for the previous case. But soil moisture in the root zone,
deviates from a steady-state condition by the difference between
ET and UFLUX, creating a water depleted root zone. Fig. 2b shows a
new moisture distribution in such case (curve D–E–B2–C2)
corresponding to the new WTD (WTD2).
2.2.3. Infiltration
Infiltration rates are calculated by the equation of Green and
Ampt (1911) simplified as
f ¼A
Fþ B
(4)
A ¼ VKSATðuS?uÞhb
B ¼ VKSAT
where f is the infiltration rate which is equal to the downward flux
(cm h?1); F is the cumulative infiltration (cm h?1); the parameters
A and B depend on the soil properties, initial water content and
distribution, and surface condition; and u is soil–water content at
the corresponding WTD in this case. Thus, the relationship
between A and B and WTD can be derived, resulting in estimated
infiltration rates according to WTD.
Once infiltration rate is estimated, the amount of infiltration is
compared with the depth of a depleted root zone within the soil
profile. If it is greater than the infiltration, all soil layers are
completely replenished to drained volume soil–water content and
WTD is updated as a new WTD (WTD3, Fig. 2c) and the distribution
of soil moisture follows hydrostatic equilibrium with the WTD.
Otherwise, infiltration fills up the first soil layer and so on until no
water is left for infiltration.
h
h ? 1
??
(5)
2.2.4. Water flux
Lateral drainage (water flux from farm to ditch), or sub-
irrigation (water flux from ditch to farm), is calculated using WT of
farm water management unit and farm ditches (Fig. 3).
QL¼4KeðWTfarm? WTditchÞ½2ðWTditch? dÞ þ WTfarm? WTditch?
S2
(6)
where, QLis either lateral drainage or sub-irrigation (cm h?1); Keis
effective lateral saturated hydraulic conductivity (cm h?1) that is a
function of lateral saturated hydraulic conductivity of the soil
profiles (LKSAT) (Skaggs, 1980); WTfarmand WTditch(cm) are the WT
atthefarmwatermanagementunitandditchabovemeansealevel
(AMSL);S isthe distance fromthe middle ofa farmunitto the ditch
(cm);anddisthedistancefromthebottomofditchtothemeansea
level (cm).
Deep seepage (QV, cm h?1) from soil profiles to limestone
bedrock, which is estimated by a straightforward application of
Darcy’s law, can be present if there are solution holes and fractures
in the limestone bedrock.
QV¼VKSAT?bedrockðWTfarm? WTditchÞ
dbedrock
(7)
where VKSAT?bedrockis vertical saturated hydraulic conductivity of
the underlying limestone bedrock (cm h?1) and dbedrock is its
thickness (cm).
Fig. 2d shows the new water table (WTD4) receding from the
previous water table as drainage decreases water content within
the soil profile.
2.3. Model testing
2.3.1. Dataset
Two farms in the EAA where comprehensive best management
practice (BMP) research has been conducted to monitor BMP
implementation and related P load parameters from 1992 to 2002
(Daroub et al., 2009) were selected for this study. One farm,
UF9202A, has an area of 130 ha under sugarcane monoculture. The
average ground surface elevation is 388 cm AMSL and the average
soil depth of its Lauderhill mucks (Euic, hyperthermic, shallow
H.-Y. Kwon et al./Agricultural Water Management 97 (2010) 112–122
115

Page 5

Fig. 2. Flow chart of the hydrologic processes calculated in the OntoSim-Sugarcane: (a) the distribution of soil moisture (curve A1–B1–C1) corresponding to WTD1under
‘‘drained to equilibrium’’ condition; (b) the new moisture distribution (curve D–E–B2–C2) calculated when UFLUXfails ET demand (soil moistures in the root zone are deviated
from a steady-state condition by the difference between ET and UFLUX, creating a depleted root zone); (c) the moisture distribution when all soil layers are completely
replenished to soil–watercontentof hydrostatic equilibriumwiththeWTD3;(d)themoisture distributioncorresponding toWTD4receding fromWTD3asdrainagedecreases
water content within the soil profile.
H.-Y. Kwon et al./Agricultural Water Management 97 (2010) 112–122
116