You are on page 1of 50

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

241

MODELING SURFACE HYDROLOGY, SOIL


EROSION, NUTRIENT TRANSPORT, AND
FUTURE SCENARIOS WITH THE
ECOHYDROLOGICAL SWAT MODEL IN
BRAZILIAN WATERSHEDS AND RIVER BASINS
Nadia Bernardi Bonum(1), Jos Miguel Reichert(2), Miriam Fernanda Rodrigues(3),
Jos Alberto Fernandez Monteiro(4), Jeffrey G. Arnold(5) & Raghavan Srinivasan(6)

Introduction ................................................................................................................................................ 242


Hydrosedimentological models ................................................................................................................ 243
Chemical transport models: phosphorous ............................................................................................... 245
The Soil and Water Assessment Tool - SWAT ......................................................................................... 246
Hydrology and sedimentology ............................................................................................................. 246
Chemical transport ............................................................................................................................... 250
Management practice components ...................................................................................................... 252
Model evaluation: Sensitivity analysis, calibration, and validation .................................................... 252
Running the swat model ........................................................................................................................... 253
Input data .............................................................................................................................................. 253
Topographic data .............................................................................................................................. 253
Land use and agricultural management data .................................................................................. 253
Soil data ............................................................................................................................................ 254
Hydrologic response units (HRUs) ................................................................................................. 254
Weather data .................................................................................................................................... 254
Hydrologic discharge data ............................................................................................................... 254
Model evaluation ................................................................................................................................... 255
Statistics for model evaluation ....................................................................................................... 255
Parameter sensitivity analysis, and calibration and validation ..................................................... 256
Brazilian hydrographic regions and data for SWAT .................................................................................. 257
Hydrographic regions in Brazil .............................................................................................................. 257
Data availability/acquisition ................................................................................................................. 259
Drainage basins/watersheds under study in Rio Grande do Sul ........................................................ 260
(1)

Adjunct professor of the Sanitation and Environmental Engineering Department of the Universidade Federal de
Santa Catarina - UFSC. Trindade. CEP 88040-970, Florianpolis, SC, Brazil. E-mail: nadia.bonuma@ufsc.br
Full professor of the Soil Department of the Universidade Federal de Santa Maria - UFSM. Av. Roraima, 1000. CEP
97105-900, Santa Maria, RS, Brazil. CNPq scholar. E-mail: reichert@ufsm.br
(3)
Doctoral Student in the Graduate Program in Forestry Engineering of UFSM. Av. Roraima, 1000. CEP 97105-900,
Santa Maria, RS, Brazil. E-mail: miriamf_rodrigues@yahoo.com.br
(4)
Visiting researcher in the Applied Limnology Laboratory, Universidade Federal de So Joo del Rei, Minas Gerais,
Brazil. E-mail: jose.monteiro@gmx.ch
(5)
Researcher of the USDA-ARS, Grassland, Soil and Water Research Laboratory. Zip code 76502. Temple, TX, USA.
E-mail: jeff.arnold@ars.usda.gov
(6)
Professor of the Ecosystem Sciences and Management and Biological and Agricultural Engineering Departments
of Texas A&M University - TAMU.1500 Research Parkway. Zip code 77845.College Station, TX, USA. E-mail:
r-srinivasan@tamu.edu
(2)

Tpicos Ci. Solo, 9:241-290, 2015

242

Nadia Bernardi Bonum et al.

Hidrology and erosion in a hilly watershed in southern Brazil: a case study ......................................... 261
Watershed description .......................................................................................................................... 261
Hydrology ............................................................................................................................................. 267
Water balance ....................................................................................................................................... 273
Erosion .................................................................................................................................................. 275
Phosphorus ........................................................................................................................................... 276
Scenarios of land use change and crop management ........................................................................ 277
Conclusions .......................................................................................................................................... 279
Research and model development opportunities .................................................................................... 279
References ................................................................................................................................................ 281

INTRODUCTION
Assessing the factors responsible for soil and water degradation on the
watershed scale is an important land management tool for effective water
erosion control. Hydrosedimentology studies are conducted to understand
and clarify the origin of agricultural pollution, using soil erosion and water
quality models to analyze the impacts of land use and climate changes on
water balance, sediment yield, and water quality.
For agricultural and forest watersheds, the Soil and Water Assessment
Tool - SWAT (Arnold et al., 1998) stands out as a continuous-time model
developed to predict the impact of land management practices on watersheds
with diverse soil, land use, and management conditions. SWAT was developed
based on an extensive soil and vegetation database, agricultural management
practices, and climate data from the United States. The model was originally
designed to use easy-acquisition information, requiring little or no calibration
when used in North American watersheds (Arnold et al., 2012).
When applied in regions with poor data, and where climate, soils, plants,
and agricultural management practices differ from temperate environments,
parameter calibration becomes necessary (Bieger et al., 2015). One of the
biggest challenges is a lack of field data, which consequently requires the
use of information from databases with characteristics different from those of
Brazilian watersheds.
Our objectives in this review were to describe and discuss:
(i) some hydrosedimentological, ecohydrological, and chemical transport
models;
(ii) the ecohydrological Soil and Water Assessment Tool (SWAT) model
and its components: hydrology and sedimentology, chemical transport,
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

243

components of management practices, and model evaluation - sensitivity


analysis, calibration and validation, and practicalities (input data and how to
obtain them; model evaluation, including statistics and parameter sensitivity
analysis; and calibration and validation);
(iii) watershed studies and data availability/acquisition from Brazilian
watersheds/basins;
(iv) a study on a hilly watershed in southern Brazil as an example of how
SWAT is used, including sensitivity analysis of hydrological parameters,
water balance, sensitivity analysis of erosion parameters, sensitivity analysis
of phosphorus parameters, and testing scenarios of crop management and
change in land use;
(v) challenges in applying SWAT to large Brazilian river basins and scenario
analysis;
(vi) opportunities for research and SWAT model improvement.

HYDROSEDIMENTOLOGICAL MODELS
Watersheds are complex systems where natural processes occur,
including rainfall, evapotranspiration, and surface and underground flow, along
with anthropic activities such as deforestation, agricultural production, and
dam construction. Therefore, a complete representation of every process
associated with the hydrological cycle, erosion, and sedimentation is not
possible (Minoti, 2006). According to the principle of parsimony, a model
should be as simple as possible, but not simpler. Therefore, models that
include both the amount of water and sediment yield and water quality better
represent the complexity of these phenomena in watersheds.
Hydrological modeling is used to predict runoff from land areas, infiltration
into soils, and percolation into aquifers. Rainfallrunoff models are often used
when streamflow gauge data are not available or not reliable, or otherwise
when estimates are needed of the impact that changing land uses and land
covers have on the temporal and spatial distribution of runoff (Loucks et al.,
2005).
Hydrological rainfall-runoff models may be classified by how processes
are represented, by the time and space scales that are used, and by which
equation solution methods are used (Singh, 1995). The main features for
distinguishing the approaches are the nature of basic algorithms (empirical,
conceptual, or process-based), whether a stochastic or deterministic approach
Tpicos Ci. Solo, 9:241-290, 2015

244

Nadia Bernardi Bonum et al.

is taken for input or parameter specification, and whether spatial representation


is lumped or distributed.
Many comprehensive spatially distributed hydrological models have been
developed in the past decade due to advances in hydrological sciences,
such as the Geographical Information System (GIS) and remote-sensing
(Narasimhan, 2004). A comprehensive review of watershed hydrology models
may be found in Singh and Woolhiser (2002).
A review and comparison of the mathematical bases of eleven leading
watershed-scale hydrologic and nonpoint-source pollution models was made
by Borah and Bera (2003). The authors concluded that SWAT is a promising
model for long-term continuous simulations in predominantly agricultural
watersheds.
These models were:
(i) Agricultural NonPoint-Source pollution model or AGNPS (Young et al.,
1989);
(ii) Annualized Agricultural NonPoint-Source model or AnnAGNPS (Bingner
and Theurer, 2005);
(iii) Areal Nonpoint Source Watershed Environment Response Simulation or
ANSWERS (Beasley et al., 1980);
(iv) ANSWERS-Continuous (Bouraoui et al., 2002);
(v) CASCade of planes in 2-Dimensions or CASC2D (Ogden and Julien, 2002);
(vi) Dynamic Watershed Simulation Model or DWSM (Borah et al., 2002);
(vii) Hydrological Simulation Program-Fortran or HSPF (Bicknell et al., 1993);
(viii) KINematic runoff and EROSion model or KINEROS (Woolhiser et al.,
1990);
(ix) European Hydrological System model or MIKE SHE (Refsgaard and
Storm, 1995);
(x) Precipitation-Runoff Modeling System or PRMS (Leavesley et al., 1983);
and,
(xi) Soil and Water Assessment Tool or SWAT (Arnold et al., 1998).
Borah and Bera (2003) compared applications of the SWAT, HSPF, and
DWSM models. They indicated the SWAT model as the most promising
long-term continuous simulation model. The application of SWAT seems
especially advantageous for predicting river discharge when data are scarce
(Xie and Lian, 2013). Nevertheless, for simulating monthly sediment yield,
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

245

albeit acceptable model statistics, SWAT did not perform as well as the
multilayer perceptron (MLP) neural network (Singh et al., 2012). In addition,
Moriasi et al. (2012) offers an extensive overview of watershed modeling and
suggestions for calibration and validation for globally used models.
Algorithms for erosion and sediment transport models, just as for
hydrology models, can be classified as empirical, conceptual, or physicsbased. However, many models are likely to contain a mix of modules from
each of these categories. For example, while the rainfall-runoff component of
a water quality model may be physics-based or conceptual, empirical
relationships may be used to model erosion or sediment transport (Merritt et
al., 2003). For instance, the Universal Soil Loss Equation or USLE
(Wischmeier and Smith, 1978) is an empirical model used worldwide for soil
loss estimation. Its modified version MUSLE (Williams and Berndt, 1972)
was developed to compute soil loss for a single storm event. The USLE was
also revised (RUSLE) and revisited for improvement (Renard et al., 1997).
The Water Erosion Prediction Project or WEPP (Flanagan and Nearing,
1995) is a physically-based model for predicting soil erosion and sediment
delivery from fields, farms, forests, rangelands, construction sites, and urban
areas. The Limburg Soil Erosion Model or LISEM (De Roo et al., 1996) is a
physically-based runoff and erosion model that simulates the spatial effects
of rainfall events on small watersheds. The European Soil Erosion Model or
EUROSEM (Morgan et al., 1998) is a physically-based model for predicting
soil erosion by water from fields and small catchments. ANSWERS (Beasley
et al., 1980) includes a conceptual hydrological process and a physicallybased erosion process. SWAT (Arnold et al., 1998) is a watershed-scale model
developed to predict the impact of land management practices on water,
sediment, and agricultural chemical yields in complex watersheds with varying
soil, land use, and management conditions over long periods of time.
A review of modeling approaches used for predicting soil erosion in
watersheds was made by Zhang et al. (1996), while Merritt et al. (2003)
presented one of the most comprehensive reviews of erosion and sediment
transport models. Furthermore, Aksoy and Kavvas (2005) made a more recent
review of hillslope and watershed scale erosion and sediment transport models.

CHEMICAL TRANSPORT MODELS: PHOSPHOROUS


Models are often a more efficient and feasible means of evaluating
management alternatives in assessing nutrient losses, such as P loss,
Tpicos Ci. Solo, 9:241-290, 2015

246

Nadia Bernardi Bonum et al.

because of the time and costs involved (Sharpley, 2007). These models vary
between
(i) empirical models, including models based on indicators such as the PIndex, used to examine the risk of P transfer to runoff (Djodjic et al., 2002;
Lopes et al., 2007), or export coefficient models, such as the Generalized
Watershed Loading Function (GWLF) model (Haith and Shoemaker, 1987);
and
(ii) conceptual and process based models, such as the Chemicals, Runoff,
and Erosion from Agricultural Management Systems (CREAMS) (Knisel,
1980), AGNPS (Young et al., 1989), ANSWER (Beasley et al., 1980), HSPF
(Bicknell et al., 1993), Erosion Productivity Impact Calculator (EPIC)
(Sharpley and Williams, 1990), CENTURY (Parton et al., 1993), and SWAT
(Arnold et al., 1998) models.
Process-based models typically involve numerical solution of a set of
equations that are mathematical representations of processes, such as
leaching of P, P transport in runoff and sediments, and stream processes
that affect P. Leaching of P involves simulation of adsorption and desorption
processes, which are often collectively described by relating solid-phase
(sorbed) P to dissolved P by a variety of nonlinear equations (McGechan and
Lewis, 2002). Two of the more common equations are the Freundlich and
Lagmuir equations.
Although much progress has been made with P simulation models,
inaccurate estimates can be caused in part by incomplete modeling of the
mechanisms involved, as well as a delay in incorporation of recent scientific
results into models (Cabrera, 2007).

THE SOIL AND WATER ASSESSMENT TOOL SWAT


Hydrology and sedimentology
The SWAT model (Arnold et al., 1998) was developed by the United
States Department of Agricultures Agricultural Research Service (USDAARS) and Texas A&M University. SWAT may predict long-term impacts of
land management practices on water, sediment, and agricultural chemical
yields in complex watersheds with varying soil, land use, and management
conditions (Neitsch et al., 2005).
The application of SWAT has expanded worldwide over the past years.
Most of these applications are to U.S. and Europe watersheds (Gassman et
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

247

al., 2007). Extensive examples of application of SWAT can be found in Arnold


and Fohrer (2005) and in Gassman et al. (2007).
For watersheds and river basins in Brazil, limited studies include
hydrological and sedimentological simulation and delivery of nutrients to
springs in rural watersheds, and impact assessment of different agricultural
scenarios on sediment yield and water quality (Barsanti et al., 2003; Machado
and Vettorazzi, 2003; Machado et al., 2003; Moro, 2005; Minoti, 2006; Neves
et al., 2006; Baltokoski et al., 2010; Santos et al., 2010; Silva et al., 2011;
Uzeika et al., 2012; Lelis et al., 2012; Andrade et al., 2013; Arago et al.,
2013; Strauch et al., 2013; Lubitz et al., 2013; Britto et al., 2014; Da Silva et
al., 2014b; Franz et al., 2014; Galharte et al., 2014; Pereira et al., 2014). From
the last International SWAT Conference held in Porto de Galinhas, Brazil,
several contributions may be expected in forthcoming publications. The water
budget of the Amazon (Ribeiro et al., 2014; Sanchez-Perez and Sauvage 2014;
Strauch and Volk, 2014), Tocantins (Monteiro et al., 2014), and So Francisco
Basins (Creech et al., 2014; Koch et al., 2014; Taffarelfo et al., 2014) are
sought by different groups, while others are applying SWAT to climate and
land use change (Bonum et al., 2014a; Bressiani et al., 2014; Da Silva et al.,
2014a). Applications in developing countries include Africa (Schuol et al., 2008;
Setegn et al., 2010; van Griensven et al., 2012; Dile et al., 2013; Faramarzi et
al., 2013; Ali et al., 2014), China (Ouyang et al., 2008; Yang et al., 2008; Wu
and Chen, 2009), India (Mishra et al., 2007), Uruguay (von Stackelberg et al.,
2007), and Iran (Hassani et al., 2008; Abbaspour et al., 2009; Faramarzi et
al., 2009; Akhavan et al., 2010; Jajarmizadeh et al., 2014).
Two essential components are needed to set up a SWAT model:
(i) a GIS to support storage and display of relevant maps, to perform terrain
analysis to delineate watersheds, and to identify stream reaches and
associated sub-basins, among others, and
(ii) a component that generates all files needed by SWAT, partly from input
maps and analyses and partly by manual editing (George and Leon, 2007).
SWAT can be set up using the ArcSWAT interface, an upgrade of AVSWATX (Di Luzio et al., 2004), a software system that links ArcGIS software and
the SWAT model. As an alternative to ArcSWAT, MapWindow SWAT
(MWSWAT) is an open source GIS system (George and Leon, 2007).
Watershed delineation with GIS interface is based on a D-8 algorithm.
This model also incorporates a parameter calculation function (Neitsch et
al., 2005). Each cell in a Digital Elevation Model (DEM) is assumed to flow
to one of eight neighboring cells according to steepest-slope direction. SWAT
simulates a watershed by dividing it into multiple sub-basins, which are further
Tpicos Ci. Solo, 9:241-290, 2015

248

Nadia Bernardi Bonum et al.

divided into hydrologic response units (HRUs). These HRUs are the product
of overlaying soil, land use, and slope classes.
Components of the SWAT model include weather, hydrology, soil
temperature, plant growth, erosion/sedimentation, nutrients, pesticides, and
land management (Figure 1). A detailed theoretical description of SWAT and
its major components is documented in Neitsch et al. (2005). Major hydrology
components of SWAT include precipitation, interception, evapotranspiration,
infiltration, percolation, and runoff. The SWAT model uses two phases of the
hydrologic cycle, one for land processes and the other for channel processes.
The hydrologic cycle is simulated in two phases, the land phase and the
routing phase. Land-phase hydrology controls the amount of water, sediment,
nutrient, and pesticide loadings. The routing phase consists of defining
movement of water, sediments, nutrients, and pesticides through the
watershed channel network (Neitsch et al., 2005). Once SWAT determines
water, sediment, nutrient, and pesticide loadings to the main channel, these
loadings are routed through the watershed stream network. As water flows
downstream, a fraction may be lost due to evaporation and transmission
through the channel bed. Another potential loss of water is through agricultural
or human use. The flow may be supplemented by rainfall directly into the
channel and water addition from point-source discharges. Flow is routed

Figure 1. SWAT and its components.


Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

249

through the channel using the variable storage routing method or Muskingum
method. In large sub-basins with a retention time longer than one day, only a
portion of surface runoff and lateral flow will reach the main channel the day it
is generated. SWAT incorporates a storage function to delay a portion of surface
runoff and lateral flow and the nutrients they transport (Neitsch et al., 2005).
The land phase of the hydrologic cycle is based on the water balance
equation:
(1)
where SWt is final soil water content (mm), SW0 is soil water content available
for plant uptake (initial water content - permanent wilting point water content),
t is time in days, PREC is amount of precipitation (mm), SURQ is amount of
surface runoff (mm), ET is amount of evapotranspiration (mm), PERCO is
amount of percolation (mm), and BF is amount of baseflow (mm).
Actual plant transpiration and actual soil evaporation are estimated based
on potential evapotranspiration and additional soil and land use parameters.
SWAT offers three methods to estimate potential evapotranspiration:
Priestley-Taylor (Priestley and Taylor, 1972), Hargreaves (Hargreaves and
Samani, 1985), and Penman-Monteith (Allen et al., 1989).
Surface runoff may be estimated from daily or sub-daily rainfall. For daily
rainfall, the modified Soil Conservation Service (SCS) curve number method
(Mishra and Singh, 2003) is used. The SCS curve number parameter (CN2)
is a function of land use, soil permeability, and antecedent moisture conditions.
Predictions of peak runoff rate are made with a modification from the rational
method. Channel routing is simulated using either the variable-storage method
or the Muskingum method.
After sediment yield is evaluated using the MUSLE equation, SWAT
further corrects this value considering snow-cover effect and sediment-lag in
surface runoff. SWAT also calculates the contribution of sediment to channel
flow from lateral and groundwater sources. Eroded sediment that enters
channel flow is simulated in to move downstream by deposition and
degradation (Neitsch et al., 2005).
Erosion caused by rainfall and runoff is computed with the Modified
Universal Soil Loss Equation (MUSLE) (Williams, 1975), which has an implicit
delivery ratio built into it that is a function of the peak runoff rate, which in
turn is a function of drainage area:
(2)
Tpicos Ci. Solo, 9:241-290, 2015

250

Nadia Bernardi Bonum et al.

where sed is sediment yield on a given day (t), Qsurf is surface runoff volume
(mm ha-1), qpeak is peak runoff rate (m3 s-1), areahru is area of the HRU (ha), K
is the USLE soil erodibility factor (t h MJ-1 mm-1), C is the USLE cover and
management factor, P is the USLE support practice factor, LS is the USLE
topographic factor, and CFRG is the coarse fragment factor.
Slope length and slope steepness parameters used in the MUSLE
topographic factor (LS-factor) calculation are sensitive factors that greatly
affect SWAT sediment yield predictions. The ArcSWAT interface calculates
slope length and slope steepness from the Digital Elevation Model (DEM).
However, slope length calculation does not always succeed when slopes
are steep. When a slope length is not calculated, the interface defaults to a
slope length of 50 m, which is appropriate for relatively flat watersheds, but
in watersheds with steep average slopes (> 25 cm m-1), SWAT will simulate
excessive sheet erosion (EPA, 2004).
The USLE length-slope factor is a measure of sediment transport capacity
of runoff from the landscape, but it fails to fully account for hydrological
processes that affect runoff and erosion (Moore and Burch, 1986b).
Topographic factors have a physical basis (Moore and Burch, 1986a); thus,
they generally work correctly in planar and convex hillslopes. However, their
ability to account for transport capacity effects on sediment delivery does
not extend to situations where transport capacity decreases with downslope
direction (Kinnell, 2008a,b). After sediment yield is evaluated using the MUSLE
equation, SWAT further corrects this value considering sediment lag in surface
runoff. The SWAT model also calculates the contribution of sediment to
channel flow from lateral and groundwater sources (Chaubey et al., 2007).
The channel sediment routing equation uses a modification of Bagnolds
sediment transport equation (Bagnold, 1977), which estimates transport
capacity as a function of velocity. The model either deposits excess sediment
or re-entrains sediment through channel erosion depending on the sediment
load entering the channel.
Sediment yield modeled by SWAT is calculated for each unique HRU in
the watershed, regardless of its position within each sub-watershed or subbasin. There is currently no option to include upslope contributing area while
defining HRUs (White, 2009).

Chemical transport
Transport of chemicals in a watershed depends on the transformations
compounds undergo in the soil environment. SWAT models nutrient cycles
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

251

for N and P. The latter is added to soil by fertilizer, manure, or residue


application. In SWAT, P is removed from the soil by plant uptake and erosion,
and bound to soil through sorption processes. Unlike N, which is highly
mobile, P solubility is limited in most environments.
SWAT monitors six different pools of P in the soil. Three pools are
inorganic forms of P, while the other three pools are organic forms of P. Fresh
organic P is associated with crop residue and microbial biomass, while active
and stable organic P pools are associated with soil organic matter and
materials. Organic P associated with organic matter is partitioned into two
pools to account for variation in the availability of humic substances to
mineralization. Soil inorganic P is divided into solution, active, and stable
pools (Neitsch et al., 2005).
Solution P is actually labile P (Chaubey et al., 2007), in accordance with
the original EPIC version of the P module as described in Jones et al. (1984)
and Sharpley et al. (1984). Labile P was defined by Sharpley et al. (1984) as
P extractable from soil using anion exchange resin and therefore represents
solution P plus weakly sorbed P. Transformations of soil P among these six
pools are regulated by algorithms that represent mineralization,
decomposition, and immobilization. The labile pool is in rapid equilibrium
(several days or weeks) with the active pool, which is in slow equilibrium with
the stable pool.
SWAT makes all nutrient calculations on a mass basis, even though all
nutrient levels are model input as concentrations. Nutrient concentration
(mg kg1) is converted to content (kg ha1 P) by multiplying it by the depth of
the soil layer and soil bulk density (SOL_BD), performing appropriate unit
conversions.
The inorganic P pool, originating either from mineralization of organic P
or P applied directly as inorganic fertilizer, is simulated considering plant
uptake and conversion to active and stable forms of inorganic P. Movement
of P between labile and active mineral pools is estimated using the following
equilibrium equations (Neitsch et al., 2005; Chaubey et al., 2007). P
transported with sediment to the stream is simulated with a loading function,
and baseflow P concentrations can be set to simulate lateral subsurface
flow and groundwater contributions to river loads (Radcliffe et al., 2009).
SWAT model have an option to include or exclude in-stream processes
in simulations. When the in-stream component is included, the model routes
state variables through additional algorithms adapted from QUAL2E, a steadystate stream water-quality model developed by Brown and Barnwell (1987).
Tpicos Ci. Solo, 9:241-290, 2015

252

Nadia Bernardi Bonum et al.

Management practice components


SWAT incorporates detailed information on agricultural and urban land
and water management into a simulation. General agricultural management
practices include tillage, planting, fertilization, pesticide application, grazing,
harvest, kill, and filter strips. These management practices are incorporated into
the model through various input data and parameters affected by the practices.

Model evaluation: Sensitivity analysis, calibration, and


validation
The ability of a watershed model to adequately predict constituent yields
and streamflow for a specific application is evaluated through sensitivity
analysis, model calibration, and model validation (White and Chaubey, 2005).
Sensitivity analysis provides a better understanding of which particular input
parameters have greater effect on model output (Feyereisen et al., 2007).
This analysis also identifies the most sensitive parameters, which ultimately
dictate the set of parameters to be used in the subsequent calibration process
(Kannan et al., 2007).
Model calibration is the process of estimating model parameters by
comparing model predictions (output) for a given set of assumed conditions
with observed data for the same conditions (Moriasi et al., 2007). Calibration
should be performed by a hierarchical process, beginning with hydrology,
followed by sediment transport, and finally pollutant transport, because errors
in the current component will be transferred to and magnified in all the following
components (Santhi et al., 2001).
Sensitivity, calibration, and uncertainty analyses are vital and interwoven
aspects of applying SWAT and other models (Gassman et al., 2007). As
SWAT is a complex model with many parameters that may complicate manual
model calibration (Green and van Griensven, 2008), complex automated
calibration procedures have been successfully used for hydrological modeling
with SWAT (van Griensven and Bauwens, 2003; van Griensven et al., 2006;
Green and van Griensven, 2008).
Since no simulation model is intended merely to show how well it fits the
data used for model development, performance characteristics derived from
the calibration data set are insufficient evidence for satisfactory model
performance (Klemes, 1986). Thus, the set of calibrated model parameters
should be validated against a set of independently measured data. Validation
procedures are similar to calibration procedures in that predicted and
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

253

measured values are compared to determine if the objective function is met


(White and Chaubey, 2005). Good validation results support the usefulness
of the model in predicting future conditions under alternative land use and
management scenarios and future climates.

RUNNING THE SWAT MODEL


Input data
SWAT requires topographic, land use, agricultural management, soil
parameter, and weather data. Digital maps (topographic, land use, slope,
and soil type maps) are processed with a GIS preprocessing interface to
create the input files required by the model. These data and maps are
described below.

Topographic data
Topographic data may be obtained from radar data, such as the Shuttle
Radar Topography Mission (SRTM) or the Advanced Spaceborne Thermal
Emission and Reflection Radiometer (ASTER), or by digitizing contour lines
and the drainage network from a topographic map. The digitized contour
vectors are used to create the Triangular Irregular Network (TIN) for generating
the Digital Elevation Model (DEM) with an appropriate spatial pixel resolution.
The DEM and digitized drainage network are used to delineate and partition
the watershed into sub-watersheds and reaches. Jha et al. (2004) examined
the effect of basin subdivision on simulation results and they suggest that
the optimal size of sub-watersheds is 26 % of the simulated area. The
slope map is divided into different slope classes. Information extracted and
calculated from the DEM includes overland slope, slope length, and elevation
corrections for precipitation and evapotranspiration.

Land use and agricultural management data


Land use may be determined by analyzing satellite or aerial imagery
and by field surveys, assisted by a geographic positioning system (GPS)
with GIS software. A detailed list of agricultural management operations carried
out in the watershed with dates and type of operation (planting of crop, tillage,
and harvest) must be created. In SWAT, the SCS curve number parameter
(CN2) is updated for each management operation. The date of operation
may vary annually depending on the cumulative days exceeding the minimum
Tpicos Ci. Solo, 9:241-290, 2015

254

Nadia Bernardi Bonum et al.

(base) temperature for plant growth. The potential heat units for the crops
must be calculated and the values added to the management input file (.mgt
file). This calculation is particularly important for southern hemisphere
conditions, such as in Brazil.

Soil data
The soil map should be derived from a soil survey made in the watershed.
If a local soil survey is absent, soil information can be obtained from the
Brazilian Agricultural Research Corporation (EMBRAPA), the FAO Soil Map
of the World, or the ISRIC World Soil Information. Some key hydraulic
properties, such as water content at different tension values (available water
capacity) and saturated hydraulic conductivity, can be estimated from soil
physical properties, such as soil particle size (i.e. sand, silt, and clay) and
bulk density, using pedotransfer functions. Soil information related to the
soil map must be added to the SWAT user soil databases (.usersoil file).

Hydrologic response units (HRUs)


The number of HRUs is limited by the precision of digital input maps. A
realistic combination of land uses, soil types, and slope classes should
consider a realistic threshold or percentage of land use.

Weather data
Rainfall data may be obtained from automatic meteorological stations or
from rain gauges installed within the watershed or, in their absence, from
nearby weather stations. When the Penman-Monteith potential
evapotranspiration method is used, solar radiation, air temperature, wind
speed, and relative humidity are required as input. Daily maximum and
minimum temperature, solar radiation, wind speed, and humidity values are
also obtained from an automatic meteorological station. For Brazilian
watersheds and river basins, gaps in climate data might be completed with
information from Brazilian National Institute of Meteorology (INMET) and
National Water Agency (ANA) stations adjacent to the watershed under study.

Hydrologic discharge data


Parshall flumes or other gauging stations in the watershed outlet may
be equipped with automatic water level sensors. Flow rates may be calculated
from a stage-discharge relationship developed using in-situ manual velocity
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

255

measurements at the stream cross section where the water level sensor is
located. In the case of daily simulation, the x-minute flow rates may be
integrated to obtain daily outflow rates. Streamflow data at the watershed
outlet are used for model sensitivity analysis, calibration, and validation.

Model evaluation
Statistics for model evaluation
SWAT performance may be evaluated using graphical comparison and
statistical analysis to determine the quality and reliability of predictions when
compared to measured values. Summary statistics include mean and standard
deviation (SD), where the latter is used to assess data variability. Goodnessof-fit measures include the coefficient of determination (R2) and the NashSutcliffe efficiency value (NSE, Nash and Sutcliffe, 1970).
The R2 ranges from 0 to 1, with higher values indicating less error variance.
NSE ranges from - to 1.0, where a value of 1 indicates a perfect fit and describes
the amount of variance for observed values over time accounted for by the model.
Goodness-of-fit is quantifiable using percent bias (PBIAS) and ratio of
the root mean square error to the standard deviation of measured data (RSR,
Moriasi et al., 2007). PBIAS assesses the average tendency of simulated
data to exhibit under- or overestimation (positive or negative PBIAS values,
respectively, Gupta et al., 1999).
The RSR incorporates the benefits of error index statistics and includes
a normalization factor so that resulting statistical and reported values may
be applied to various constituents. The RSR is the ratio of the root mean
square error to the standard deviation of measured data (Moriasi et al., 2007).
The RSR varies from the optimal value of 0, which indicates zero RMSE or
residual variation and therefore perfect model simulation, to a large positive
value. The lower the RSR, the lower the RMSE is, and the better the model
simulation performance is (Moriasi et al., 2007).
To assess how well the model performed, Green et al. (2006), Green and
van Griensven (2008), and Wu and Chen (2009) used standards of NSE >
0.4 and R2 > 0.5, whereas Santhi et al. (2001) assumed monthly NSE > 0.5
and R2 > 0.6 as indicating acceptable model performance when calibrating
for hydrology. Furthermore, Moriasi et al. (2007) suggested that model
simulation may be judged as satisfactory if NSE > 0.50 and RSR 0.70,
and if PBIAS is 25% for streamflow for a monthly time step. Nevertheless,
when watershed models are evaluated on a daily time step, the ratings can
be less strict than for longer time steps (Moriasi et al., 2007).
Tpicos Ci. Solo, 9:241-290, 2015

256

Nadia Bernardi Bonum et al.

Parameter sensitivity analysis, and calibration and validation


To directly analyze the effect of model parameters on model output and
on model performance, parameter-sensitivity analysis should be performed
(van Griensven et al., 2006). The relative ranking of which parameters most
affect the output is determined by error functions calculated for daily or monthly
flow measured in the watershed outlet gauge.
Measured data from the watershed outlet gauge should be compared to
SWAT output during calibration and validation. Predicted total flow for monthly
and daily calibration and validation is calculated from the FLOW_OUT model
output for the appropriate sub-basin in the main channel output file from
SWAT (output.rch file).
To calibrate daily streamflow, an automated digital filter technique (Arnold
and Allen, 1999) may be used to separate baseflow from the measured
streamflow. Baseflow is an important component of streamflow, and must be
calibrated before the model is fully calibrated for streamflow and for other
components (Jha et al., 2007).
As SWAT is a complex model with many parameters, which complicates
manual model calibration, auto-calibration algorithms can be used. Up to
the SWAT 2009 Version, sensitivity analysis and auto-calibration procedure
tools were embedded in the ArcSWAT interface. The auto-calibration
procedure was based on a multi-objective calibration, and incorporated the
Shuffled Complex Evolution Method algorithms (SCE-UA).
A combination of manual and auto-calibration procedures (Figure 2) is
presented in Green and van Griensven (2008):
(i) first, parameters are manually calibrated until model simulation results
are acceptable as per NSE, R2, RSR, and PBIAS values;
(ii) second, final parameter values that were manually calibrated are used as
initial values for the autocalibration procedure. Maximum and minimum
parameter value limits are used to keep output values within a reasonable
value range;
(iii) third, an autocalibration tool is run with optimal fit values to provide the
best fit between measured and simulated data as determined by the NSE
values and by how reasonable the values are;
(iv) finally, autocalibrated parameter values are then adjusted to ensure they
are reasonable. For validation, the model is run using input parameters
determined during the calibration process for other periods.
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

257

Figure 2. Flowchart of sensitivity analysis and calibration with SWAT version


2009 (left) and with SWAT version 2012 and SWAT-CUP (right).

With SWAT Version 2012, sensitivity analysis, auto-calibration, validation,


and uncertainty analysis can be carried out using the SWAT-CUP software
(Abbaspour et al., 2007). The software allows users to choose among five
optimization techniques: Generalized Likelihood Uncertainty Estimation
(GLUE), Parameter Solution (ParaSol), Sequential Uncertainty Fitting
algorithm (SUFI-2), a Bayesian framework implemented using Markov chain
Monte Carlo (MCMC) and Particle Swarm Optimization (PSO). SWAT-CUP
should first be run with few simulations for sensitivity analysis, and should
then be run again for calibration and/or uncertainty analysis.

BRAZILIAN HYDROGRAPHIC REGIONS AND DATA


FOR SWAT
Hydrographic regions in Brazil
With an area of 8,515,767 km2, Brazilian territory is equivalent to
approximately 84% of the total area of Europe, and is divided into twelve
hydrographic regions, namely Amaznica (Amazonas), Tocantins-Araguaia,
Tpicos Ci. Solo, 9:241-290, 2015

258

Nadia Bernardi Bonum et al.

Atlntico Nordeste Ocidental (Western Northeast Atlantic), Parnaba, Atlntico


Nordeste Oriental (Eastern Northeast Atlantic), So Francisco, Atlntico
Leste (Eastern Atlantic), Paraguai, Paran, Atlntico Sudeste (Southeast
Atlantic), Atlntico Sul (South Atlantic), and Uruguai (Figure 3; ANA, 2006).
Variables used in environmental quality studies have been monitored by
several public and private institutions, and the detail of the information depends
on the interest in a particular area of knowledge, area of coverage, and financial
resources available.

(a)

SP
PR

PARAGUAI
SC

SC

ARGENTINA
RS

RS

URUGUAI

(b)

URUGUAI

(c)

Figure 3. Brazilian hydrographic regions (a), highlighting the Uruguai (b) and
Atlntico Sul (c) hydrographic regions in Rio Grande do Sul (RS).
Source: Adapted from ANA (2006).

Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

259

Data availability/acquisition
Proper management of natural resources depends on knowledge regarding
processes that define resource behavior. Studies and research carried out in
higher education and research institutes to assess water quantity and quality
are rare for watersheds, limited in scale, and with short monitoring periods.
The hydrological monitoring network on water resources in Brazil was
planned and structured to promote electricity generation and water availability
projects, which are implemented in large basins. The river monitoring network
in Brazil, managed by ANA (National Water Agency), promotes the
development of projects on medium and large rivers.
A national monitoring program for small agricultural watersheds does
not exist. Such a program is needed to define public policy management
and use of water resources. This information is currently restricted to isolated
and short-term initiatives of universities and research institutions.
Availability is scarce for models requiring intense input data. General
information is available for watersheds/basins from sources such as the ANA,
Army, CPRM, Embrapa, IBGE, and INMET (Table 1). Except for some specific
monitoring and modeling efforts, the required data are not available at an
adequate scale. For small and medium watersheds, no water and sediment
discharge data are available from public sources, thus hindering validation of
models.
As for weather information, INMET data from ~300 stations are available
online. In addition, ANA relies on a larger database of >2,400 rain gauges
also available through the agency website. Nevertheless, for a country as
large as Brazil (~8.5 106 km2), it is not uncommon for a researcher to find an
area of study lacking long-term and complete meteorological data series. As
an alternative to observed weather data, there are several weather reanalysis
products, such as the Climate Forecast System Reanalysis (CFSR; Saha et
al., 2010) or the ERA-Interim (Dee et al., 2011). Based on weather data of
diverse natures (e.g., terrestrial stations, remote sensing, meteorological
balloons) that feed weather forecast models, reanalyses are offered as a dense
grid of stations with global coverage in extensive time series with high temporal
resolution (6-hourly to daily). The use of reanalysis data in SWAT projects has
been considered very satisfactory (Dile and Srinivasan, 2014), and a specific
evaluation of application in Brazil is under way (Monteiro et al., 2014).
To calibrate and validate the model, the user will need time series of flow
and other variables of interest of at least one station close to the main outlet.
Tpicos Ci. Solo, 9:241-290, 2015

260

Nadia Bernardi Bonum et al.

Table 1. Data acquisition/availability for SWAT at different watershed/drainage


basin scales
Watershed/Drainage
Basin

DEM(1) Land use(2)

Soil(3)

Weather(4)

Area
(km2)

Size

<3

Small

Survey,
Army
map

Survey

Medium

Survey,
Army
map

Survey,
Google
Earth,
Satellite

Large

Army
map,
Emprapa
map

Google
CPRM map,
INMET
Earth,
Embrapa map,
Satellite, IBGE map
IBGE map
ANA

3 10

> 10

and
Management

Discharge(5)
Water

Sediment

Survey,
Station,
CPRM map,
INMET
Embrapa map

Flume

Sampling

Survey,
Station,
CPRM map,
INMET
Embrapa map

Flume

Sampling

Flume, Sampling,
State
ANA
agencies,

1
DEM: Brazilian Army topography map 1:50,000, and Embrapa DEM 1:250,000, both for the entire
country. (2)Soil: Embrapa soil map 1:750,000, CPRM (Servio Geolgico do Brasil) soil map 1:250,000,
and (2)IBGE (Instituto Brasileiro de Geografia e Estatstica) soil map 1:1,000,000, all for the entire
country. For some specific watersheds, maps at more detailed scales are available. (3)Land use: IBGE
map 1:1,000,000 of vegetation and land use, for five only states. (4)Weather: Weather station network
of the INMET (Instituto Nacional de Meterologia). (5)Discharge: Water and sediment discharge
measurements by ANA (Agncia Nacional das guas).

ANA provides daily and monthly series for river discharge from over 1,100
river gauges. Water quality variables are provided from more scattered stations
and have very low temporal resolution (i.e., four annual observations).
In addition to Brazilian sources, maps of soil type and vegetation cover
are also available from global sources at coarser resolution. The United States
Geological Survey provides several vegetation and land cover maps (http://
landcover.usgs.gov/landcoverdata.php). Furthermore, the website of the
WaterBase project hosts vegetation and soil maps with their respective SWAT
database, as well as the free MWSWAT interface that runs on the MapWindow
software (http://waterbase.org/).

Drainage basins/watersheds under study in Rio Grande


do Sul
The state of Rio Grande do Sul, with an area of 281,748 km2, is part of
the South Atlantic and Uruguay hydrographic regions (Figure 3; ANA, 2006).
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

261

Since the state has a relatively large area (about three times larger than
Portugal and equivalent to approximately 79 % of German territory), water
resources are divided into 25 river basins. Studies on hydrosedimentological
process modeling were performed in river basins and watersheds at different
scales according to the data or resources available for collecting the
information needed. The drainage basins under study are transboundary
basins or watersheds draining to the Atlantic Ocean, and are thus of
paramount importance for water resource management, including international
waters.
Published dissertations and theses by Hydrographic Region and their
respective rural watersheds in the state of RS, regardless of the scale of the
watershed and sub-watershed, were organized by Rodrigues et al. (2014).
There were 91 (ninety-one) studies, covering 52 watersheds of different scales,
where only 12 river basins have studied in rural watersheds (Table 2).
Information on the environment, hydrosedimentology, climate,
geomorphology, geology, soil type and use, and conceptual and mathematical
models used for simulation and prediction of hydrological and sedimentological
processes, quality assessment of management practices, management of
natural resources, and models for assigning value and charging for the use
of natural resources were studied and are presented in Table 2. This information
is available and can be used for hydrosedimentological modeling studies.
The largest number of published studies were on the Arroio Lajeado
Ferreira (11 publications) and Arroio Lino (7 publications) watersheds.
Seventeen different models were used in all, and SWAT was studied in only
two watersheds (Rodrigues et al., 2014). SWAT use in the Arroio Lino
watershed is described in detail later is this paper.

HIDROLOGY AND EROSION IN A HILLY WATERSHED


IN SOUTHERN BRAZIL: A CASE STUDY
Watershed description
Intensively cultivated agricultural lands are prone to changes in water
balance and may be sources of environmental contamination. Due to shortage
of suitable areas for cropping, farmers cut and clear the forest of steep lands
to grow tobacco under conventional soil tillage. Steep land combined with
inadequate cultivation practices have caused rapid degradation of natural
resources, worsening the poverty cycle (Merten and Minella, 2006).
Tpicos Ci. Solo, 9:241-290, 2015

Watershed

Sub-watershed

Area (km2)

LU

Hydrology

P
Ibirub-Quinze
de Novembro

x
x

x
x

x
x

Brao direito
Brao esquerdo

0.75
0.73
18.70
3.32
1.64
1.68

Bacia principal
Sub-bacia

0.94
0.39

x
x

PBH 140
PBH 80

1.45
0.80

Arroio Caador
Arroio Tiririca
Arroio Ju
Arroio Caracol

Drenagem 1
Drenagem 2
Arroio Jacutinga
Arroio Lino

Horto Florestal
Terra Dura
Assentamento
Alvorada
Arroios Caador
e Tiririca
Arroios Ju
e Caracol
Sistema Salto
Rio dos Sinos
Arroio Morungava
Rio Pardo

WB WQ E/SL SSC

x
x

x
x

x
x

x
x

x
x

x
x

x
x

x
x

21.97
16.29

x
x

x
x

x
x

x
x

x
x

0.13
0.25
52.50
3 746.68
40.20
3 658.34

x
x

x
x

x
x
x

x
x
x

x
x
x
x

x
x

Century

1900, 1956, 1977,


1988, 2002
07/2010-06/2011
04/2011-11/2011
01/2009-02/2011

USLE, SWAT

x
x

x
x

x
x

USLE, LISEM

x
x

x
x

x
x

2006-2013

2009-2013
2009-2013
07/2006-06/2007
07/2006-06/2007

x
x

Study period

SY

x
x
x
x
x

Model

USLE,MGB, IPH
USLE
10/1984-08/2001

Continued...

Nadia Bernardi Bonum et al.

835.00

Erosion and
sediments

262

Tpicos Ci. Solo, 9:241-290, 2015

Table 2. Watersheds in the state of RS, Brazil and their main characteristics as studied and published in theses and dissertations

Watershed

Sub-watershed

Area (km2)

Rios VacacaVacaca Mirim

LU

11,077.34
353.59

x
x

x
x

x
x
x
x

x
x
x
x

x
x
x

Alto da Colina

0.76
0.96
1,145.70
1.90

Menino Deus I

18.00

Menino Deus II
5.03
Menino Deus IV
18.67
Rancho do Amaral 4.50
Reservatrio
DNOS
33.00

x
x
x

x
x
x

Stio do Tio Pedro

x
x

x
x

Arroio Grande
Horto Florestal
Ponta das Canas
Fazendda Alvorada
Rio Vacaca Mirim

Taquari-Antas
Arroio Lajeado

0.39
26,491.82

Hydrology

x
x

x
x

x
x
x
x

x
x
x
x

x
x
x

x
x
x

Erosion and
sediments

WB WQ E/SL SSC

x
x

x
x
x

x
x
x

x
x

x
x
x

Study period

SMAP
USLE

1996-2004
1987, 1999-2002,
10/2005-03/2006

SY

x
x

Model

LISEM
2012-current
LISEM
2012-current
USLE, MUSLE
Kineros2
06/2001-03/2002,
2008
AGNPS2001,
MUSLE, USLE,
fingerprinting 08/1996-12/1997,
1987-2009, 2011
MUSLE, IPH II 10/2011-07/2012
10/2011-07/2012
2008

x
x

x
x

x
x

Kineros2
IPH-MGBq

1972, 1997,2001,
2003-2005
2003-2004
09/1993-09/2002

263

Continued...

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

Tpicos Ci. Solo, 9:241-290, 2015

Table 2. Cont.

264

Watershed

Sub-watershed

Ferreira

Vila Maria

Sub-bacia
Sub-bacia
Sub-bacia
Linha 16-17
Linha 18

Rio Conceio
Rio Guapor
Rio Pelotas
Arroio Santa Brbara
2006
Sanga da Inocncia
Rio Ibicu

Arroio Lajeado
Grande
Rio Ibirapuit
Rio Itu
Rio Iju
Arroio Lajeado Atafona
Trs Negrinhos

Area (km2)

LU

1.19

x
x
x
x
x
83.00

x
x
x
x
x
x

x
x

x
x
x
x

x
x
x
x

0.63
35,495.38

x
x

33.19

5,976.00
2,809.61
10,703.78
41.80
10.33

x
x

x
x

x
x
x

x
x
x

0.57
0.15
0.04
4.00
7.60
800.00
2,000.00
909.00

x
x

x
x

Hydrology

Erosion and
sediments

WB WQ E/SL SSC
x

x
x
x

x
x

x
x
x
x

Model

SY
x

x
x

USLE, SWAT,
fingerprinting,
WEPP, LISEM,
Century,
137Cs
2002-current

05/2010-06/2011
05/2010-06/2011
RUSLE
2011-current
SWAP, ANIMO 2003-current

1953, 1965, 1995,

MAGICC/CENGEN,
MGB-IPH, CRUZ,
OutorgaLS
1960-2005
x

x
x

Study period

MUSLE

12/2004-06/2007,
2009-2010
1997-1998

MGB-IPH-C
x

1990-1991

Continued...

Nadia Bernardi Bonum et al.

Tpicos Ci. Solo, 9:241-290, 2015

Table 2. Cont.

Watershed

Sub-watershed

Ponte Mstica

Rio Potiribu
1989-current
Arroio Taboo
Arroio Turcato
Arroio Donato
Arroio Piraizinho
Rio Quara
Estncia Tarum
e Fazenda
So Carlos

Rio Inhandava

Estncia Tarum
Fazenda
So Carlos

Area (km2)

9,450.00
Santo ngelo
5,440.00
Colnia Mousquer 2,160.00
Passo do
Faxinal
1,940.00
Ponte Nova
do Potiribu
609.00
660.00

84.00
19.50
0.20
220.33
4,500.00

x
x
x
x
x

0.92
0.21
11.96

LU

Hydrology

Erosion and
sediments

WB WQ E/SL SSC

Model

x
x

x
x
x

x
x

x
x

x
x

x
x

x
x

SY

x
x

1989-1994

x
x

Study period

MGB-IPH-C, SWAP, ANIMO

Century, USLE
05/01/2005-05/
01/2007

2008-2013
2008-2013

265

Variables - S: soil; LU: Land use; C: Climate; P: Precipitation; Q: water flow; WB: water balance; WQ: water quality; E/SL: Erosion/Soil loss; SSC: suspended sediment
concentration; SY: sediment yield. Models - AGNPS2001: Agricultural Non-Point Source Model; Century: Soil Organic Matter Model Environment; CRUZ: a model of optimal
water balance, which consist of water allocation (grant) based on water balance; Fingerprint: Identification technique of producing fonts sediment; IPH II: Rainfall flow
Mathematical model of the Hydrologic Research Institute; Kineros2: Kinematic runoff and Erosion Model; LISEM: Limburg Soil Erosion Model; MAGICC/CENGEN: Model for
the assessment of Greenhouse gas Induced Climate Change/scenario GENerator; MGB - IPH: Great Basin Model of the Hydrologic Research Institute; MGB-IPH-C: Great
Basin Model of the Hydrologic Research Institute for modeling the flow of C; IPH-MGBq: Hydrologic Research Institute - Great Basin Model of quality; MUSLE: Modified
Universal Soil Loss Equation; OutorgaLS: generalized platform for analysis of grants for water abstraction and effluent discharge; RUSLE: Revised Universal Soil Loss
Equation; SMAP: Soil Moisture Accounting Procedure; SWAT: Soil and Water Assessment Tool; USLE: Universal Soil Loss Equation; WEPP: Water Erosion Prediction Project;
137Cs: Erosion and deposition rates and sediment budget model; SWAP: Soil, Water, Atmosphere, Plant Environment; ANIMO: Agricultural, Nutrient Model.

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

Tpicos Ci. Solo, 9:241-290, 2015

Table 2. Cont.

266

Nadia Bernardi Bonum et al.

Most tobacco in Southern Brazil is produced on small farms on land


with low agricultural potential (Merten and Minella, 2006). Unsuitable
agricultural practices for land use capability and high fertilizer and pesticide
application rates make tobacco growing an activity with a high risk of
contamination of water resources (Bortoluzzi et al., 2006; Becker et al.,
2009; Kaiser et al., 2010) and propensity to intense erosion and
sedimentation (Minella et al., 2009) in agricultural watersheds.
Understanding of hydrological processes is crucial for adequate
assessment of soil and water degradation. A key hydrological factor is surface
runoff, which is primarily responsible for the transport of sediments, nutrients,
and other contaminants throughout the watershed. However, long-term
watershed monitoring data are rare due to the expenses involved (Santhi et
al., 2006), and thus the lack of streamflow measurements hinders evaluation
of water availability.
The Arroio Lino watershed (Figures 4 and 5), used herein as a case
study, covers 4.8 km2 and is located in Agudo, state of Rio Grande do Sul,
Brazil (29.1 S, 67.1 E). The Arroio Lino creek is a tributary of the Jacu
River, with a drainage area characterized by intensive use of the land for
agriculture and livestock production.

Figure 4. Location of the Arroio Lino watershed.


Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

(a)

(b)

267
(c)

Figure 5. Watershed delineation and land use on July 28, 2009 (a); March 16,
2011 (b); and December 9, 2011 (c). From: Image 2014 DigitalGlobe, 2014
Google Earth.

The watershed belongs to the Serra Geral Formation, which has basalt
hillsides and localized outcroppings of Botucatu sandstone (Pellegrini et al.,
2009). Chernossolos (Mollisols) predominate, but Neossolos (Entisols) are
found on steeper slopes (USDA, 1999; Dalmolin et al., 2004). Vegetation is
composed of remnants of seasonal deciduous forests in different stages of
succession (Pellegrini et al., 2009).
The climate is humid subtropical (Cfa type), according to the Kppen
classification, with an average temperature of more than 22 C in the hottest
month and between - 3 and 18 C in the coldest month (Alvares et al., 2013).
Rainfall is well distributed, ranging from 1,300 to 1,800 mm year-1 (Kaiser et
al., 2010; Alvares et al., 2013). A Walter and Lieth climate diagram (1960) for
the closest INMET station to the Arroio Lino watershed is shown in Figure 6.
Annual crops occupy almost 30 % of the Arroio Lino watershed area,
and native forest covers more than 50 %. Approximately 90 % of cropland is
dedicated to tobacco production (Pellegrini et al., 2009). The tobacco crop
is grown with conventional tillage, causing environmental degradation due to
intense agricultural exploitation.

Hydrology
Sensitivity analysis (Bonum et al., 2013) was carried out using 27
parameters of the SWAT model suggested as being the most sensitive to
streamflow simulation (van Griensven et al., 2006). The parameter whose
variation had:
(i) the highest sensitivity was initially the SCS Curve Number II value (CN2).
CN2 is a key parameter of the SCS method; increased values of CN2 imply
an increase in surface runoff;
Tpicos Ci. Solo, 9:241-290, 2015

268

Nadia Bernardi Bonum et al.

Figure 6. Walter and Lieth (1960) climate diagram for Santa Maria, the closest
INMET station to the Arroio Lino watershed. At the top of the panel, from
left to right: observed period, mean annual temperature, and mean annual
precipitation; the red line stands for mean monthly temperature (left
axis), and black numbers beside the axis are the mean daily maximum
and mean daily minimum temperature of the warmest and coldest month,
respectively. The blue line stands for precipitation, right axis show
precipitation at the rate of 2 mm C-1 up to 100 mm, and 20 mm C-1 from
100 mm upward. The dashed area indicates the dry season (not present
in this region) and the striped area indicates the moist months. Below
the y-axis, months with probable frost are indicated in blue. Note: for the
Southern Hemisphere, the diagram starts in July to show summer in the
central zone.

(ii) the second greatest effect was the soil evaporation compensation factor
(ESCO). Kannan et al. (2007) noticed that a change in the value of the ESCO
affects all the water balance components; and
(iii) the third highest sensitivity was the baseflow alpha factor (ALPHA_BF).
Similar analysis made in other watersheds has suggested that CN2 and
ALPHA_BF have great importance in water quality simulation (van Griensven
et al., 2006).
Both manual and auto-calibration procedures were required to correct
these simulation errors. To calibrate and validate base and surface runoff
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

269

flows, total flow was separated into two components. An automated digital
filter technique (Arnold and Allen, 1999) was used to separate baseflow from
the measured total flow (Figure 7).
Simulated surface flow was increased through calibration of the
parameters listed in Table 3 (Bonum et al., 2013).
The Soil Conservation Service runoff curve number for the moisture
condition II (CN2) parameter was originally set to values recommended by
the USDA SCS National Engineering Handbook (USDA, 1972) for each
hydrologic group. The ICN and CNCOEF parameters were used to account
for soil moisture in addition to the SCS runoff curve number (Green and van
Griensven, 2008). The soil evaporation compensation factor (ESCO) is a
calibration parameter and not a directly measurable property. As ESCO
increases, the depth to which soil evaporative demand can be met decreases,
which limits soil evaporation and reduces the simulated value for
evapotranspiration (ET) (Feyereisen et al., 2007). The soil evaporation
compensation factor (ESCO) was adjusted to decreasing actual
evapotranspiration. Available soil water capacity (SOL_AWC) was reduced,
which resulted in an increase in surface flow. Stormflow is inversely
proportional to SOL_AWC; the two variables exhibit a straight-line relationship
throughout the range of values for SOL_AWC. Reducing SOL_AWC results
in the soil profile filling sooner, with more runoff, less ET, and increased
baseflow (Feyereisen et al., 2007).

Figure 7. Results of baseflow separation from streamflow hydrograph.


Bonum et al., 2013.

Tpicos Ci. Solo, 9:241-290, 2015

270

Nadia Bernardi Bonum et al.

Table 3. The SWAT model parameters included in the final calibration and their
initial and final ranges
Parameter
ALPHA_BF
CN2

CNCOEF
ESCO
FFCB

GW_DELAY
ICN
PHU

REVAPMN

SOL_AWC
SOL_K
SURLAG

Description

Range

Baseflow alpha factor


0.0 to 1.0
Initial SCS runoff curve
number for moisture
condition II
25 %
Curve number coefficient
0.5 to 2.0
Soil evaporation
compensation factor
0.0 to 1.0
Initial soil water storage
expressed as a fraction of
field capacity water content
0.0 to 1.0
Groundwater delay time
0 to 500
Daily curve number
calculation method
0 or 1
Potential heat unit
(used for tobacco)
1,000 to 2,000
Potential heat unit
(used for corn)
1,000 to 2,000
Potential heat unit
(used for beans)
1,000 to 2,000
Threshold depth of water
in the shallow aquifer
for revap to occur.
0 to 500
Available water capacity
of the soil layer
25 %
Saturated hydraulic
conductivity of the soil
25 %
Surface runoff lag coefficient
0 to 4

Initial
Value

Calibrated Value

0.048

30 to 100
0

+ 10 %
0.5

0.95

0
31

1
5

1,800

1,000

1,800

1,450

1,800

1,350

300

Default

-5 %

Default
4

+5 %
1

As simulated baseflow values with SWAT were significantly lower in


relation to baseflow estimated from measured streamflow, groundwater
parameters were adjusted to improve the subsurface response. The threshold
depth of water in the shallow aquifer for revap to occur (REVAPMN) was
increased and the time for water leaving the bottom of the root zone to reach
the shallow aquifer (GW_Delay) was reduced. Saturated hydraulic conductivity
(SOL_K) of the first soil layer was increased, which resulted in an increase
in baseflow.
Lastly, the temporal distribution of the flow and the shape of the hydrograph
were improved through calibration of the stormflow lag time (SURLAG) and
the baseflow recession constant (ALPHA_BF).
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

271

Monthly observed and simulated streamflow matched well during both


calibration (2005) and validation (2004) periods (Figure 8) at the watershed
outlet (Bonum et al., 2013). Monthly calibration and validation R2 values
were 0.90 and 0.86 (> 0.6). Based on Moriasi et al. (2007), model
performance was very good for the calibration period. This is supported
by an NSE of 0.87 (> 0.75), RSR of 0.35 ( 0.50), and PBIAS of -8 %
(modular value < 10 %). Similarly, for the validation period , model
performance was good since the NSE was 0.76, RSR was 0.49, and
PBIAS was -13.3 % (10 % < PBIAS < 15 %). Since validation assesses
the performance of the calibrated model parameter set against a set of
independent measured data, it is typically more difficult to get good
validation performance in comparison to calibration.
Special attention was given to the magnitude of peak flows and the
shape of recession curves on the daily time scale (Figures 9 and 10). For
the daily calibration period, the R2 value was 0.78, whereas for the validation
period, the R2 value was 0.59. Daily calibration NSE and RSR were 0.56
and 0.66, respectively, while validation NSE and RSR were 0.20 and 0.97,
respectively.
To evaluate daily streamflow variability, measured and simulated daily
streamflow data were converted to flow duration curves (FDC, Figure 11).
FDC derived from the simulated hydrographs indicated an overestimation of
peak flows and an underestimation of low flows by the calibrated SWAT

Figure 8. Monthly flow calibration and validation results.


Bonum et al., 2013.

Tpicos Ci. Solo, 9:241-290, 2015

272

Nadia Bernardi Bonum et al.

model. That model simulations could not capture runoff peaks well in the
daily flow record may be due to uncertainty in the modified curve number
method of the Soil Conservation Service (Mishra and Singh, 2003) used to
estimate surface runoff. When watershed concentration time is less (shorter)
than one day, uncertainty in estimated surface runoff from daily rainfall is
even higher.

Figure 9. Daily streamflow calibration results.


Bonum et al., 2013.

Figure 10. Daily streamflow validation results.


Bonum et al., 2013.

Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

273

Figure 11. Flow duration curves derived from measured and simulated data
from the Arroio Lino watershed.

A significant variation between measured and simulated curves can be


observed for estimation of low flows (Figure 11), and this error may be attributed
in part to inadequacy of hydrograph recession simulations of the complete
time series (Figures 7 and 8). Measured flows had flatter recessions after
main events than simulated ones. The consequence of estimation of steeper
recessions is that the model tended to under-simulate low flows for this
watershed. Thus, daily peak and hydrograph recession characteristics are
critical for model predictions of watershed streamflow. The analysis of FDC
also indicated that runoff peaks, which are of great importance in sediment
and nutrient transport simulation, were better estimated than low flow.

Water balance
Average annual values for hydrologic components, such as surface runoff
(SURQ), lateral runoff (LATQ), groundwater contribution to streamflow (GW),
percolation (PERCO), soil water storage (SW), evapotranspiration (ET), and
water yield (WY), were obtained from SWAT outputs (Figure 12, as an
example of the watershed studied) and compared to calculated values, based
on precipitation (PREC) and streamflow measurements in the Arroio Lino
watershed.
Tpicos Ci. Solo, 9:241-290, 2015

274

Nadia Bernardi Bonum et al.

Figure 12. Average annual simulated water balance for the Arroio Lino
watershed.

Potential evapotranspiration (PET) computed by the SWAT model using


the PenmanMonteith equation (Monteith, 1965) was compared with PET
data from the INMET weather station (class A pan). The PET data computed
with SWAT follows the same temporal trend as the PET data measured,
confirming the accuracy of the Penman-Monteith approach in the study site
(Medeiros, 1998). PET was corrected for land cover based on simulated
plant growth to determine actual evapotranspiration (ET, Neitsch et al., 2005),
where 34 and 41 % of annual precipitation is lost by evapotranspiration in
the watershed during the calibration and validation periods, respectively.
Furthermore, monthly evapotranspiration equals or exceeds monthly
precipitation in four months of the year.
For simulated and measured runoff volumes on an average annual basis,
the statistical results were better in the calibration period than in the validation
period also because annual rainfall in the calibration period was 34 % higher.
Overall SWAT seems to simulate wet years better than dry years (Green et
al., 2006; Setegn et al., 2010). Surface runoff contributes 47 and 52 % of the
water yield during the calibration and validation periods, respectively, whereas
groundwater contributes 12 and 9 % of the water yield during the calibration
and validation periods, respectively (Bonum et al., 2013).
Average annual simulated surface runoff (396 mm) is 90 % of the
measured average data value (435 mm). A baseflow index (BFI) (baseflow/
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

275

total streamflow) was estimated from daily streamflow records using a


recursive digital filter method. This approach estimated that the BFI of measured
data is 0.51, whereas the simulated BFI is 0.52. Therefore, the calibrated
model was considered acceptable for generating baseflow predictions.
Although there is no well-established rainy season for this region, higher
runoff occurred during the months of September and October. Seasonal trends
can be depicted by plotting measured and predicted monthly streamflow
values against time. The largest measured and simulated monthly runoff
volume (October 2005) had values of 210.13 mm and 233.52 mm, respectively,
indicating a difference of only 10 %. For the six measured events with runoff
greater than 40 mm, SWAT overestimated runoff four times. This indicates
that there was no clear trend for over or underestimation. The error associated
with monthly-measured runoff is estimated at an average of 5 to 15 % (Bonum
et al., 2013).

Erosion
Sediment yield was overpredicted, with a relative error of + 190 % (Bonum
et al., 2014b). Simulated sediment yield was decreased through calibration
of the following parameters: USLE support practice factor (USLE_P), initial
SCS runoff curve number for moisture condition II (CN2), average slope
steepness (Slope), and average slope length (Slsubbsn).
Even after calibration, simulated sediment yield was very high, with a
relative error of + 84 %, compared with observed values. For sediment yield
simulation in a small steep watershed (1.19 km) in Southern Brazil, Uzeika
et al. (2012) also did not find satisfactory results with the SWAT model,
possibly due to limitations in the sediment load equation (MUSLE) or sediment
propagation in the channel. Sediment deposition was seen as responsible
for the overprediction of sediment yield since large volumes of sediment
were deposited in depressions in fields near the alluvial channel, thus
indicating that not all eroded soil from hillslopes reaches the stream (Bonum
et al., 2014b).
The SWAT model overpredicts sediment yield in steep slope watersheds
possibly because of its inability to capture landscape variations; i.e., as of
45 cm m-1 slope, the landform is depressed and sediment deposits there,
but SWAT still maintains the sediment as in flow, thus overestimating sediment
load at the stream outlet.
By introducing a new algorithm in SWAT (Bonum et al., 2014b) in an
attempt to simulate landscape unit routing of sediment, the results were
Tpicos Ci. Solo, 9:241-290, 2015

276

Nadia Bernardi Bonum et al.

promising, but calibration of transport capacity parameters (ktc) in the new


sediment routine has yet to be adequately solved. Therefore, further research
is needed to address the uncertainties involved, and needs to be applied and
evaluated using other input datasets, especially in areas where reliable spatial
sediment transport patterns and spatially distributed depositional data is
available.

Phosphorus
Sensitivity analysis was carried out using nine parameters of the SWAT
model suggested as the most sensitive for P simulation (van Griensven et
al., 2006). Among these parameters, the SWAT model was the most sensitive
to the P soil partitioning coefficient (PHOSKD), P percolation coefficient
(PPERCO), N soil partitioning coefficient (NPERCO), and deep aquifer
percolation fraction (RCHRG_DP). Simulation using default values
underestimated P loads. Model calibration was required to correct these
simulation errors. The simulated P loads were increased through calibration
of the following parameters: PHOSKD, PPERCO, NPERCO, RCHRG_DP, P
enrichment ratio for sediment loading (ERORGP), and the P availability index
(PSP).
To evaluate the spatial distribution of P transfer, P measured loads were
compared with SWAT simulation results in five monitoring points in the Arroio
Lino watershed. NSE values were below 0.5, and RSR values were slightly
above 0.7 in three sub-watersheds. It was very difficult to correctly evaluate
model performance as water sampling for P analysis was not systematically
performed in storm events. Harmel et al. (2006) suggested that the uncertainty
of measured data must be considered to appropriately evaluate watershed
models.
Although predicted P loads are in the order of magnitude of measured
ones, the model was not as accurate when predicting P loads in three of the
sub-watersheds. One possible reason is that calibration of streamflow and
sediment parameters was made only in the watershed outlet, which limited
the calibration of P related parameters in the sub-watersheds. This means
that a large amount of measured data is necessary for more precise model
calibration.
SWAT underestimated the high P loads in the Arroio Lino watershed
possibly because its desorption/adsorption model component was not able
to simulate the low P retention capacity of the soil. SWAT utilizes a linear
isotherm for desorption/adsorption based on the simplified model (EPIC)
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

277

developed by Jones et al. (1984) and Sharpley et al. (1984). According to


Vadas et al. (2006), this simplified model underestimated soil P desorption,
and the authors suggested replacing EPICs constant sorption and desorption
rate factor with more dynamic rate factors. Since soils differ in clay mineralogy,
Fe, Al, and Ca contents, and pH, and all these factors affect sorption/
precipitation, they differ substantially in their P retention capacity. The use
of non-linear isotherms and sorption parameters (Rossi et al., 2012) that
consider soil properties, such as clay mineralogy and Fe content, could
improve the SWAT P predictions.
SWAT predicted a total P load of 10,500 kg year-1 P. A large portion of
the Arroio Lino watershed was forestland (~60 %), followed by tobacco crop
(~25 %), pastures (~9 %), other crops (4 %), and roads (~3 %). Although
cropland (tobacco and other crops) occupies only 29 % of the total land
cover, it contributed almost 80% of the soluble P transported by surface
runoff, 76 % of the mineral P sorbed to sediment, and 67 % of the organic P
transported with sediment into the reach. One explanation is the high initial
labile P levels in cropland soils. In addition, the application of fertilizer
increases P loss since surface application increases P availability for surface
runoff and sediment transport.

Scenarios of land use change and crop management


After calibration and verification of SWAT model for streamflow,
sediments, and P, different management scenarios were simulated in the
Arroio Lino watershed. Three different management systems were considered
as future scenarios: conventional tillage (CT), minimum tillage or conservation
tillage (MT), and no-tillage cultivation (NT). Model output variables investigated
were surface runoff, baseflow, total water yield, total sediment loading, organic
P, soluble P, and total P.
To predict future impacts of management alternatives, a stochastic
weather generator was used to produce a 30-year period of simulated rainfall
(PREC), potential evapotranspiration (PET), and evapotranspiration (ET) data
over the simulated period. Annual rainfall ranged from 1,145 to 2 196 mm year-1,
with a median and standard deviation of 1 686 and 257 mm year-1, respectively.
Simulated surface runoff (SR) decreased when changing from conventional
tillage to minimum tillage (CT-MT), minimum tillage to no-tillage (MT-NT),
and conventional tillage to no-tillage (CT-NT). However, the baseflow (BF)
increased when decreasing tillage intensity following almost the same order
of magnitude as the increases in surface flow. Hence, the percentage of
Tpicos Ci. Solo, 9:241-290, 2015

278

Nadia Bernardi Bonum et al.

deviation in water yield (WY) is only 6 % due to change from conventional


tillage to no-tillage management.
The biggest decrease in sediment yield (SY) was from conventional tillage
scenarios to no-tillage scenarios (CT-NT, 66 %), followed by conventional
tillage to no-tillage (CT-NT, 39 %) and minimum tillage to no-tillage (MT-NT,
28 %). Pellegrini (2006) analyzed different management scenarios in a plot
field scale in the Arroio Lino watershed and found the same range of variation.
In relation to the soluble P (Psol), organic P (Porg), and total P (Ptot)
loads, the biggest change was due to a decrease (- 50 %) in the fertilizer
application rate (CT-NT), rather than due to the change in management
practices (MT-NT and CT-MT). Lowering the P rate by 50 % in tobacco fields
decreased mean annual P loads by 60 %.
Percentage deviation of modeling results regarding application of
management scenarios on water balance components, nutrients, and
sediment loading are depicted in figure 13. In general, the largest differences
were between conventional tillage scenarios and no-tillage scenarios.

Figure 13. Percentage deviation of modeling results regarding application of


management scenarios on water balance components, nutrients, and
sediment loading. Key: CT is conventional tillage; MT is minimum tillage;
NT is no-tillage; SF is surface runoff; BF is baseflow; WY is water yield;
Psol is soluble phosphorus; Porg is organic phosphorus; Ptot is total
phosphorus.
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

279

Conclusions
Hydrology: The objective of calibrating and validating SWAT to match
the measured water flow within the measurement error is successfully
achieved when proper data are available and processes in the watershed
studied are properly represented. The SWAT model is a promising tool for
evaluating hydrology in subtropical areas, especially on a monthly and
annual basis.
Sediments: Initial simulations may not be satisfactory even after
exhaustive calibration since this is a second step in the modeling process,
though the portrayal of sediment detachment, transport, deposition, and
delivery to channels within the watershed may be more important. Inclusion
of a description of sediment transport capacity or perhaps consideration of a
physically-based erosion model in the SWAT source code could improve
accuracy in modeling sediment yield at the watershed outlet.
Nutrients (phosphorus): Initial simulations may not be satisfactory even
after exhaustive calibration, since this is the third step in the modeling process
and embodies cumulative errors from hydrology and sediment SWAT
components. Furthermore, modeling nutrient behavior as large set of
simplifying equations. Although predicted P loads may be in the order of
magnitude of measured ones, calibration of streamflow and sediment
parameters in the watershed outlet may hinder calibration of P related
parameters in sub-watersheds, particularly if P determination in water is not
performed for storm events. Additionally, algorithms used to estimate P
desorption/adsorption are a critical component.
Management and climate scenarios: Simulation of management
practices and climate scenarios is a promising tool if the SWAT model is
properly calibrated and associated errors are indicated.

RESEARCH AND MODEL DEVELOPMENT


OPPORTUNITIES
Modeling of sediment and nutrient loads is essential for determining or
predicting impacts at the landscape level. Results of these models, along
with existing measurements, provide clues for identifying the origin, nature,
and magnitude of pollution. However, the reliability of SWAT results depends
on the overall availability of large amounts of high quality data, and proper
description of the processes depicted in the model.
Tpicos Ci. Solo, 9:241-290, 2015

280

Nadia Bernardi Bonum et al.

As previously discussed, obtaining good data (digital elevation, soil, land


use and management, and weather data) as basic input for modeling, and
validation data (measured water, sediment, and chemical discharge) are
essential for adequate results in the modeling process.
A national monitoring program for small/medium agricultural watersheds
should be implemented by establishing a network and using uniform
procedures for sampling and data analysis as a contribution to establishing
public policies for water resource use and management, overcoming the
isolated and short-term initiatives of universities and research institutions.
Nevertheless, for large-scale assessments in Brazil, available data seem
sufficient. SWAT was used to model green and blue water availability at the
continental level in Africa, calibrated on 207 gauges (Schuol et al., 2008).
Therefore, the data from over 1,100 river gauges recorded by ANA should
suffice, with the advantage of being all in the same format, making conversion
to SWAT standards easier. For studies assessing the general impact of land
use, i.e., the effect of substitution of natural vegetation by crops or pastures,
general land use maps available (cf. Section 6.2) should provide enough
information.
To have the best spatial variability of weather, a SWAT model requires
data from one meteorological station per sub-basin. Unfortunately, weather
data recorded by federal agencies (INMET and ANA, the latter only for
precipitation) do not cover all of Brazilian territory at an adequate density
and with homogeneous distribution. As an alternative, large scale SWAT
projects could use reanalysis data sets (Dile and Srinivasan, 2014). Climate
Forecast System Reanalysis data (Saha et al., 2010) are available for
download in SWAT format from the hydrological model homepage. The data
has a fair spatial distribution (5 decimal degrees) and a daily spatial resolution
ranging from 1979-2010. Precipitation CFSR data were compared to
observations, and the suitability of this data set was analyzed by Monteiro
et al. (2014). There were considerable differences between both data sets,
especially from 1998 to 2002, possibly related to changes in the assimilation
of satellite data by CFSR (Wang et al., 2011). Therefore Monteiro et al.
(2014) suggest an interpolation method to bring statistically generated data
closer to observations. The ERA-Interim reanalysis (Dee et al., 2011) has
similar spatial resolution and the same time span as CFSR, but its similarity
to Brazilian weather is still unexplored.
The most difficult challenges in modeling large-scale Brazilian watersheds
should arise due to anthropic effects. Reservoirs alter the normal hydrological
behavioral of streams, increasing base flow and decreasing discharge peaks.
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

281

SWAT is able to incorporate dam operation data into the discharge routine,
but a centralized database of dams and reservoirs is still lacking. Furthermore,
water quality data is scarce, and the time series tend to be quite diluted (four
annual measurements for ANA stations). If the objective of the study is related
to sediment budget, the United States Geological Survey has developed a
tool called Loadest (Runkel et al., 2004; Lorenz et al., 2013) available from
the following website (http://water.usgs.gov/software/loadest/). Loadest
selects the best option for modeling sediment load or concentration in function
of discharge from among nine different models. Using the best model selected,
it is possible to generate a daily series of sediment load or concentration
given the daily discharge.
For further improvement of sediment and P load simulation in similar
watersheds using the SWAT model, some suggestions are to (i) apply the
model to studying watersheds after fragmenting them into smaller units (subbasins and HRUs) to have data on different spatial scales; (ii) use measured
data as input parameters for the model to minimize uncertainties; (iii) perform
parameter sensitivity analysis to facilitate further calibration and validation;
(iv) improve flow simulation in sub-watersheds as much as possible, as this
is expected to enhance sediment and P simulation; (v) ensure that total flow
is properly simulated by partitioning its components (surface flow and
baseflow) to be separately modeled; (vi) research further calibration of
transport capacity parameters, since this is a very important issue of the
SWAT sediment routine that has yet to be adequately solved; (vii) use nonlinear isotherms and sorption parameters that consider soil properties, such
as clay mineralogy and Fe content, which could improve SWAT P predictions;
(viii) overcome the lack of sufficient hydrology, sediment, and nutrient data
for validation; (ix) simulate climate change effects and scenarios for land use
and management that are realistic and possible to adopt in a given case
study; (x) use simulation data for predicting loss of soil and transfer of nutrients
due to the adopted management systems, as well for environmental planning;
and (xi) expand work on large river basins.

REFERENCES
Abbaspour KC, Faramarzi M, Ghasemi SS, Yang H. Assessing the impact of climate change
on water resources in Iran. Water Resour Res. 2009;45:1-16.
Abbaspour KC, Vejdani M, Haghighat S. SWATCUP calibration and uncertainty programs for
SWAT. In: Proceedings of the Congress on Modeling and Simulation, 2007; Melbourne, 2007.
Melbourne: Modeling and Simulation Society of Australia and New Zealand; 2007. p.16031609.

Tpicos Ci. Solo, 9:241-290, 2015

282

Nadia Bernardi Bonum et al.

Akhavan S, Abedi-Koupai J, Mousavi SF, Afyuni M, Eslamian SS, Abbaspour KC. Application
of SWAT model to investigate nitrate leaching in HamadanBahar Watershed, Iran. Agric
Ecsyst Environ. 2010;139: 675:88.
Aksoy H, Kavvas ML. A review of hillslope and watershed scale erosion and sediment
transport models. Catena. 2005;64:247-71.
Ali YSA, Crosato A, Mohamed YA, Abdalla SH, Wright NG. Sediment balances in the Blue Nile
River Basin. Int J Sediment Res. 2014;29:316-28
Allen RG, Jensen ME, Wright JL, Burman RD. Operational estimates of evapotranspiration.
Agron J. 1989;81:650-62.
Alvares CA, Stape JL, Sentelhas PC, Gonalves JLM, Sparovek G. Kppens climate
classification map for Brazil. Meteorol Z. 2013;22:711-28.
Agncia Nacional das guas ANA. Regies Hidrogrficas. 2006. Accessed on Dec. 02,
2014Available at: <http://www2.ana.gov.br/Paginas/default.aspx>
Andrade MA, Mello CR, Beskow S. Simulao hidrolgica em uma bacia hidrogrfica
representativa dos Latossolos na regio Alto Rio Grande, MG. R Bras Eng Agric
Amb. 2013;17:69-76.
Arago R, Cruz MAS, Amorim JRA, Mendona LC, Figueiredo EE, Srinivasan VS. Anlise de
sensibilidade dos parmetros do modelo Swat e simulao dos processos
hidrossedimentolgicos em uma bacia no Agreste Nordestino. R Bras Ci Solo. 2013;37:1091102.
Arnold JG, Allen PM. Automated methods for estimating baseflow and groundwater recharge
from streamflow records. J Am Water Resour As. 1999;35:411-24.
Arnold JG, Fohrer N. SWAT2000: current capabilities and research opportunities in applied
watershed modeling. Hydrol Process. 2005;19:563-72.
Arnold JG, Moriasi D, Gassman PW, Abbaspour KC, White MJ, Srinivasan R, Santhi C, Harmel
RD, van Griensven A, van Liew MW, Kannan N, Jha M. SWAT: Model use, calibration, and
validation. Trans ASABE. 2012;55:1491-508.
Arnold, J.G, Srinivasan, R, Muttiah, R.S, Williams, J.R. Large area hydrologic modeling and
assessment part I: model development. J Am Water Resour As. 1998;34:73-89.
Bagnold RA. Bed load transport by natural rivers. Water Resour Res. 1977;13:303-12
Baltokoski V, Tavares MHF, Machado RE, Oliveira MP. Calibrao de modelo para a simulao
de vazo e de fsforo total nas sub-bacias dos rios Conrado e Pinheiro Pato Branco (PR).
R Bras Ci Solo. 2010;34:253-61
Barsanti P, Disperati L, Marri P, Mione A. Soil erosion evaluation and multitemporal analysis in
two Brazilian basins. In: Proceedings of the 2nd International Swat Conference; 2003;BaryItaly,. p.117-30.
Beasley DB, Huggins LF, Monke EJ. ANSWERS: A model for watershed planning. Trans
ASAE. 1980;23:938-44.
Becker AG, Moraes BS, Menezes CC, Loro VL, Santos DR, Reichert JM, Baldisserotto B.
Pesticide contamination of water alters the metabolism of juvenile silver catsh,
Rhamdiaquelen. Ecotoxicol Environ Safe. 2009;72:1734-9.
Bicknell BR, Imhoff JC, Kittle Jr. JL, Donigian Jr. AS, Johanson RC. Hydrologic Simulation
Program FORTRAN (HSPF) Users Manual for Release 10. Athens, Ga.: U.S. EPA
Environmental Research Lab, Report No. EPA/600/R93/174, 1993. 660p.
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

283

Bieger K, Hormann G, Fohrer N. Detailed spatial analysis of SWAT-simulated surface


runoff and sediment yield in a mountainous watershed in China. Hydrol Sci J. 2015;60:784800.
Bingner RL, Theurer FD. AnnAGNPS Technical processes: Documentation Version 3.2.
Oxford: USDA-ARS National Sedimentation Laboratory; 2005.
Bonum NB, Corseuil CW, Rossi CR, Kobiyama M, Zanin PR. Hydrologic assessment in a
Brazilian forest watershed using SWAT model. In: Book of Abstracts International Swat
Conference; 2014, Porto de Galinhas, [Br] 2014a. p.93.
Bonum NB, Rossi CHG, Arnold JG, Reichert JM, Paiva EMCD. Hydrology evaluation of the
soil and water assessment tool considering measurement uncertainty for a small watershed
in Southern Brazil. Appl Eng Agric. 2013;29:189-200.
Bonum NB, Rossi CHG, Arnold JG, Reichert JM, Minella JPG, Allen PM, Volk M. Simulating
Landscape Sediment Transport Capacity by Using a Modified SWAT Model. J Environ Qual.
2014b;43:55-66.
Borah DK, Bera M. Watershed-scale hydrologic and nonpoint-source pollution models:
Review of mathematical bases. Trans ASAE. 2003;46:1553-66.
Borah DK, Xia R, Bera M. DWSM - A dynamic watershed simulation model. In: Singh VP,
Frevert DK. editors. Mathematical models of small watershed hydrology and applications.
Highlands Ranch, Colorado: Water Resources Publications; 2002. p.113-66.
Bortoluzzi EC, Rheinheimer DS, Gonalves CS, Pellegrini JBR, Zanella R, Copetti ACC.
Contaminao de guas superficiais por agrotxicos em funo do uso do solo numa
microbacia hidrogrfica de Agudo, RS. R Bras Eng Agric Amb. 2006; 10:881-7.
Bouraoui F, Braud I, Dillaha TA. ANSWERS: A nonpoint-source pollution model for water,
sediment, and nutrient losses. In: Singh VP, Frevert DK., editors. Mathematical models of
small watershed hydrology and applications. Highlands Ranch, Colorado: Water Resources
Publications, 2002. p.833-82.
Bressiani DA, Srinivasan R, Jones CA, Mendiondo EM. Climate change impacts on the
streamflow of a semi-arid watershed, Northeast Brazil. In: Book of Abstracts International
Swat Conference; Porto de Galinhas, [Br]: 2014. p.86.
Britto FB, Menezes Neto EL, Aguiar Netto AO, Rego NAC. Sustentabilidade hdrica da SubBacia do Rio Sangradouro, Sergipe. R Bras Geogr Fs. 2014;7:155-64
Brown LC, Barnwell TOJ. The Enhanced Water Quality Model QUAL2E and QUAL2EUNCAS documentation and user manual. Washington: U.S. Environmental Protection Agency;
1987.
Cabrera ML. Modeling phosphorus in Runoff: Basic approaches. In: Radcliffe DE, Cabrera
ML, editors. Modeling phosphorus in the environment. Boca Raton, [US]: CRC Press; 2007.
p.65-80.
Chaubey I, White KL, Green CH, Arnold JG, Srinivasan R. Phosphorus modeling in soil and
water assessment tool model. In: Radcliffe DE, Cabrera ML. editors. Modeling phosphorus
in the environment. Boca Raton, [US]: CRC Press; 2007. p.163-88.
Creech C, Brito R, Selegean J, Miller C. Anthropogenic impacts to the sediment budget of
the So Francisco river navigation channel using SWAT. In: Book of Abstracts International
Swat Conference; 2014; Porto de Galinhas, [Br]: 2014. p.59.

Tpicos Ci. Solo, 9:241-290, 2015

284

Nadia Bernardi Bonum et al.

Da Silva JM, Rocha HR, Saad SI, Basilio EM, Lobo GA. The hydrological environmental
services of Permanent Preservation Areas (PPA): A case study with numerical modeling in
the Ribeiro das Posses watershed. In: Book of Abstracts International Swat Conference;
2014; Porto de Galinhas, [Br]: 2014a. p.54.
Da Silva RM, Santos CAG, Silva AM. Predicting soil erosion and sediment yield in the Tapacur
catchment, Brazil. J Urban Environ Eng. 2014b;8:75-82.
Dalmolin RSD, Pedron FA, Azevedo AC, Zago A. Levantamento semidetalhado de solos da
microbacia do arroio Lino-Municpio de Agudo (RS). Santa Maria, RS: FATEC/UFSM; 2004.
84p. (Relatrio Tcnico Programa RS-Rural/SAA-RS)
De Roo APJ, Wesseling CG, Ritsema CJ. LISEM: A single event physically-based hydrological
and soil erosion model for drainage basins: I Theory, input and output. Hydrol Process.
1996;10:1107-17.
Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S, Vitart F. The ERA-Interim
reanalysis: Configuration and performance of the data assimilation system. Quarter J Royal
Meteorol Soc. 2011;137: 553-97.
Di Luzio M, Srinivasan R, Arnold JG. A GIS-coupled hydrological model system for the
watershed assessment of agricultural nonpoint and point sources of pollution. Trans GIS.
2004;8:113-36.
Dile YT, Srinivasan R. Evaluation of CFSR climate data for hydrologic prediction in datascarce watersheds: An application in the Blue Nile River Basin. J Am Water Resour As.
2014;50:1226-41.
Dile YT, Berndtsson R, Setegn SG. Hydrological response to climate change for Gilgel Abay
River, in the Lake Tana Basin - Upper Blue Nile Basin of Ethiopia. PLoS ONE. 2013;8:1-13.
Djodjic F, Montas H, Shirmohammadi A, Bergstrom L, Ulen B. A decision support system for
phosphorus management at a watershed scale. J Environ Qual. 2002;31:937-45.
EPA. White Paper (f) Use of SWAT to simulate nutrient loads and concentrations in California.
2004. Accessed on Dec. 02, 2014. Available at: www.fs.fed.us/digitalvisions/tools/
index.php?shortcat=GridTools.
Faramarzi M, Abbaspour KC, Ashraf Vaghefi S, Farzaneh MR, Zehnder AJB, Srinivasan R,
Yang H. Modeling impacts of climate change on freshwater availability in Africa. J Hidrol.
2013;480:85-101.
Faramarzi M, Abbaspour KC, Schulin R, Yang H. Modelling blue and green water resources
availability in Iran. Hydrol Process. 2009;23:486-501.
Feyereisen GW, Strickland TC, Bosch DD, Sullivan DG. Evaluation of SWAT manual
calibration and input parameter sensitivity in the Little River watershed. Trans ASABE.
2007;50:843-55.
Flanagan DC, Nearing MA. Water erosion prediction project hillslope profile and watershed
model documentation. West Lafayette: USDA-ARS National Soil Erosion Research Laboratory:
NSERL 1995. (Report, 10)
Franz C, Makeschin F, Wei H, Lorz C. Sediments in urban river basins: Identification of
sediment sources within the Lago Parano catchment, Brasilia DF, Brazil - using the fingerprint
approach. Sci Total Environ. 2014;466-467:513-23.
Galharte CA, Villela JM, Crestana S. Estimativa da produo de sedimentos em funo da
mudana de uso e cobertura do solo. R Bras Eng Agric Amb. 2014;18:194-201.
Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

285

Gassman PW, Reyes MR, Green CH, Arnold JG. The soil and water assessment tool:
Historical development, applications, and future research directions. Trans ASABE.
2007;50:1211-50.
George C, Leon LF. WaterBase: SWAT in an open source GIS. Open Hydrol J. 2007;1:19-24
Google Earth. Image 2014 DigitalGlobe, 2014 Google Earth. Accessed on October 25,
2014.
Green CH, van Griensven A. Autocalibration in hydrologic modeling: Using SWAT2005 in
small-scale watersheds. Environ. Modell Software. 2008;23:422-34.
Green CH, Tomer MD, Di Luzio M, Arnold JG. Hydrologic evaluation of the Soil and Water Assessment
Tool for a large tile-drained watershed in Iowa. Trans ASABE. 2006;49:413-22.
Gupta HV, Sorooshian S, Yapo PO. Status of automatic calibration for hydrologic models:
Comparison with multilevel expert calibration. J Hydrol Eng. 1999;4:135-43.
Haith DA, Shoemaker LL. Generalized watershed loading functions for streamflow nutrients.
Water Resour. B. 1987;23:471-8.
Hargreaves GH, Samani ZA. Reference crop evapotranspiration from temperature. Appl
Eng Agric. 1985;1:96-9.
Harmel RD, Cooper RJ, Slade RM, Haney RL, Arnold JG. Cumulative uncertainty in measured
streamflow and water quality data for small watersheds. Trans ASABE. 2006;49:689-701.
Hassani N, Asghari HR, Frid AS, Nurberdief M. Impacts of overgrazing in a long term
traditional grazing ecosystem on vegetation around watering points in a semi-arid rangeland
of North-Eastern Iran. Pak J Biol Sci. 2008;11:1733-7.
Jajarmizadeh M, Lafdani EK, Harun S, Ahmadi A. Application of SVM and SWAT models for
monthly streamflow prediction, a case study in South of Iran. KSCE J Civ Eng. 2014; online:1-13.
Jha M, Gassman F, Secchi S, Gu R, Arnold JG. Effect of watershed subdivision on SWAT
flow, sediment and nutrient predictions. J Am Water Resour As. 2004;40:811-25.
Jha M, Gassman PW, Arnold JG. Water quality modeling for the Raccoon River watershed
using SWAT2000. Trans ASABE. 2007;50:479-93.
Jones GCA, Cole CV, Sharpley AN, Williams JR. A simplified soil and plant phosphorus model.
I. Documentation. Soil Sci Soc Am J. 1984;40:800-5.
Kaiser DR, Reinert DJ, Reichert JM, Streck CA, Pellegrini A. Nitrate and Ammonium in soil
solution in tobacco management systems. R Bras Ci Solo. 2010;34:379-88.
Kannan N, White SM, Worrall F, Whelan MJ. Sensitivity analysis and identification of the best
evapotranspiration and runoff options for hydrological modeling in SWAT-2000. J Hydrol.
2007;332:456-66.
Kinnel PIA. Discussion: Misrepresentation of the USLE in LS sediment delivery a fallacy?
Earth Surf Proc Land. 2008a;33:1627-9.
Kinnel PIA. Sediment delivery from hillslopes and the universal soil loss equation: Some
perceptions and misconceptions. Hydrol Process. 2008b;22:3168-75.
Klemes V. Operational testing of hydrological simulation Models. Hydrol Sci J.
1986;31:13-24.
Knisel WG. CREAMS a field scale model for chemicals, runoff and erosion from agricultural
management systems. Washington: USDA Conservation Research; 1980. (Report, 26)
Tpicos Ci. Solo, 9:241-290, 2015

286

Nadia Bernardi Bonum et al.

Koch H, Liersch S, Azevedo R, Silva A, Hattermann F. Modelling of water availability and


water management for the So Francisco Basin, Brazil. In: Book of Abstracts International
Swat Conference; Porto de Galinhas, [Br]: 2014. p.60.
Leavesley GH, Lichty RW, Troutman BM, Saindon LG. Precipitation- runoff modeling system:
Users manual. Denver: U.S. Geological Survey Water Resources, Investigation; 1983. (Report,
83-4238)
Lelis TA, Calijuri ML, Santiago AF, Lima DC, Rocha EO. Sensitivity Analysis and Calibration of
Swat Model Applied to a Watershed in Southeastern Brazil. R Bras Ci Solo. 2012;36:623-34.
Lopes F, Merten GH, Franzen M, Giasson E, Helfer F, Cybis LF.A. Utilizao de P-Index em
uma bacia hidrogrfica atravs de tcnicas de geoprocessamento. R Bras Eng Agric Amb.
2007;11:312-17.
Lorenz DL, Runkel RL, De Cicco, L. River Load Estimation, v0.2. US Geological Survey, 2013.
Accessed on Dec 02, 2014. Available at: https://github.com/USGS-R/rloadest.
Loucks DP, van Beek E, Stedinger JR, Dijkman JPM, Villars MT. Water Resources Systems
Planning and Management: An Introduction To Methods, Models And Applications. Paris:
UNESCO; 2005. 677p.
Lubitz E, Pinheiro A, Kaufmann V. Simulao do transporte de sedimentos, nitrognio e
fsforo na Bacia do Ribeiro Concrdia, SC. R Bras Rec Hdricos. 2013;18:39-54.
Machado RE, Vettorazzi CA. Simulao da produo de sedimentos para a microbacia
hidrogrfica do Ribeiro dos Marins (SP). R Bras Ci Solo. 2003;27:735-41.
Machado RE, Vettorazzi CA, Xavier AC. Simulao de cenrios alternativos de uso da terra
em uma microbacia utilizando tcnicas de modelagem e geoprocessamento. R Bras Ci Solo.
2003;27:727-33.
McGechan MB, Lewis DR. Sorption of phosphorus by soil, part 1: Principles, equations and
models. Biosyst Eng. 2002;82:1-24.
Medeiros SLP. Avaliao de mtodos de estimativa da evapotranspirao de referncia
para a regio mesoclimtica de Santa Maria-RS. R Bras Agrometeorol. 1998;6:105-9.
Merritt WS, Latcher RA, Jakeman AJ. A review of erosion and sediment transport models.
Environ Modell Software. 2003;18:761-99.
Merten GH, Minella JPG. Impact on sediment yield due to the intensification of tobacco
production in a catchment in Southern Brazil. Ci Rural. 2006;36:669-72.
Minella JPG, Merten GH, Walling DE, Reichert JM. Changing sediment yield as an indicator of
improved soil management practices in southern Brazil. Catena. 2009;228-36.
Minoti RT. Abordagens qualitativa e quantitativa de micro-bacias hidrogrficas e reas
alagveis de um compartimento do Mdio Mogi Superior/SP. [tese]So Carlos: Escola de
Engenharia de So Carlos Universidade de So Paulo; 2006.
Mishra A, Froebrich J, Gassman PW. Evaluation of the SWAT model for assessing sediment
control structures in a small watershed in India. Trans ASABE. 2007;50:469-78.
Mishra SK, Singh VP. Soil Conservation Service Curve Number (SCS-CN) methodology.
Dordrecht, The Netherlands: Kluwer Academic Publishers; 2003. (Water Science and
Technology Library, 42)

Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

287

Monteiro JAF, Gcker B, Srinivasan R. Comparison between Climate Forecast System


Reanalysis (CFSR) weather data and data from meteorological stations in Brazil to evaluate
the suitability of CFSR data for SWAT. In: Book of Abstracts of the International Swat
Conference; Porto de Galinhas, [Br] 2014. p.75.
Monteith J.L. Evaporation and the environment. In The state and movement of water in living
organisms. In: 19th Symposia of the Society for Experimental Biology. London, U.K: University
of Cambridge Press; 1965. p.205-34.
Moore ID, Burch GJ. Physical basis for the length-slope factor in the universal soil loss
equation. Soil Sci Soc Am J. 1986a;50:1294-98.
Moore ID, Burch GJ. Sediment transport capacity of sheet and rill flow: Application of unit
stream power theory. Water Resour Res. 1986b;22:1350-60.
Morgan RPC, Quinton JN, Smith RE, Govers G, Poesen JWA, Auerswald K, Chisci G, Torri D,
Styczen ME. The European Soil Erosion Model (EUROSEM): A dynamic approach for predicting
sediment transport from fields and small catchments. Earth Surf Proc Land. 1998;23:527-44.
Moriasi DN, Arnold JG, van Liew MW, Bingner RL, Harmel RD, Veith TL. Model evaluation
guidelines for systematic quantification of accuracy in watershed simulations. Trans ASABE.
2007;50:885-900.
Moriasi DN, Wilson BN, Douglas-Mankin KR, Arnold JG, Gowda PH. Hydrologic and water
quality models: Use, calibration, and validation. Trans ASABE. 2012;55:1241-7.
Moro M. A utilizao da interface SWAT-SIG no estudo da produo de sedimentos e do
volume de escoamento superficial com simulao de cenrios alternativos. [tese] Piracicaba:
Escola Superior de Agricultura Luiz de Queiroz, 2005.
Narasimhan B. Development of indices for agricultural drought monitoring using a spatially
distributed hydrologic model. [dissertao] College Station: Texas, Texas A & M University;
2004.
Nash JE, Sutcliffe JV. River flow forecasting through conceptual models. Part 1: A discussion
of principles. J Hydrol. 1970;10:282-90.
Neitsch SL, Arnold JG, Kiniry JR, Williams JR. Soil and water assessment tool - theoretical
documentation: Version 2005. Temple: Blackland Research Center; Texas Agricultural
Experiment Station; 2005.
Neves FF, Silva FGB, Crestana S. Uso do modelo AVSWAT-X na avaliao do aporte de
nitrognio (N) e fsforo (P) aos mananciais de uma microbacia hidrogrfica contendo atividade
avcola. Eng. Sanit. Amb. 2006;11:311-7.
Ogden FL, Julien PY. CASC2D: A two-dimensional, physically based, Hortonian hydrologic
model. In: Singh VP, Frevert DK. editors. Mathematical models of small watershed hydrology
and applications, Highlands Ranch, Colorado: Water Resources Publications; 2002. p.69112.
Ouyang W, Hao F, Wang X. Regional non point source organic pollution modeling and critical
area identification for watershed best environmental management. Water Air Soil Poll.
2008;187:251-61.
Parton WJ, Scurlock JMO, Ojima DS, Gilmanov TG, Scholes RJ, Schimel DS, Kirchner T,
Menaut JC, Seastedt T, Moya EG, Kamnalrut A, Kinyamario JI. Observations and modeling of
biomass and soil organic matter dynamics for the grasslands biome world-wide. Global
Biogeochem. Cyles. 1993;7:785-809.

Tpicos Ci. Solo, 9:241-290, 2015

288

Nadia Bernardi Bonum et al.

Pellegrini A. Sistemas de cultivo da cultura do fumo com nfase s prticas de manejo e


conservao do solo. [dissertao] Santa Maria: Universidade Federal de Santa Maria;
2006.
Pellegrini JBR, Rheinheimer DS, Gonalves CS, Copetti ACC, Bortoluzzi EC, Tessier D.
Impacts of anthropic pressures on soil phosphorus availability, concentration, and
phosphorus forms in sediments in a Southern Brazilian watershed. J. Soils Sediments.
2009;10:451-60.
Pereira DR, Almeida AQ, Martinez MA, Rosa DRQ. Impacts of deforestation on water
balance components of a watershed on the Brazilian East Coast. R Bras Ci Solo.
2014;38:1350-8.
Priestley CHB, Taylor RJ. On the assessment of surface heat flux and evaporation using
large-scale parameters. Mon Weather Rev. 1972;100:81-92.
Radcliffe DE, Lin Z, Risse LM, Romeis JJ, Jackson CR. Modeling phosphorus in the lake
allatoona watershed using SWAT: I. Developing phosphorus parameter values. J Environ
Qual. 2009;38:111-20.
Refsgaard, J.C, Storm, B. MIKE SHE. In: Singh VP, editor. Mathematical models of small
watershed hydrology and applications. Highlands Ranch, Colorado: Water Resources
Publications; 1995. p.809-46.
Renard KG; Foster GR, Weesies GA, Mccool DK, Yoder DC. Predicting soil erosion by water:
A guide to conservation planning with the revised Universal Soil Loss Equation (RUSLE).
Washington: USDA; 1997. (Agricultural Handbook, 703)
Ribeiro CBM, Mohanty BP, Bressiani DA, Fernandes JG, Rotunno Filho OC. Parameterization
of physical and climatic characteristics in the Amazon basin for hydrological simulation with
SWAT model. In: Book of Abstracts of the International Swat Conference; Porto de Galinhas,
[Br]: 2014. p.7-8.
Rodrigues MF, Reichert, JM, Vogelmann ES, Awe GO, Minella JPG. Agricultural watershed
studies for socio-economic development in southern Brazil. CapWEM. 2014;4:4-92.
Rossi CG, Bonum NB, Heil DM, Williams JR. Evaluation of the langmuir model in the soil and
water assessment tool for a high soil phosphorus condition. Environ Modell Softwer.
2012;38:40-9.
Runkel RL, Crawford CG, Cohn TA. Load Estimator (LOADEST): A FORTRAN Program for
Estimating Constituent Loads in Streams and Rivers. Reston, Virginia: US Geological Survey
2004. (Techniques and methods book 4)
Saha S, Moorthi S, Pan H-L, Wu X, Wang J, Nadiga S, Goldberg M. The NCEP climate forecast
system reanalysis. B. Am. Meteorol. Soc. 2010;91:1015-57.
Sanchez-Perez J-M, Sauvage S. The role of the alluvial floodplain to modeling water discharge
using SWAT model in the Amazon catchment. In: Book of Abstracts of the International Swat
Conference; Porto de Galinhas, [Br]: 2014. p.95.
Santhi C, Arnold JG, Williams JR, Dugas WA, Srinivasan R, Hauck LM. Validation of the SWAT
model on a large river basin with point and nonpoint sources. J. Am. Water Resour. As.
2001;37:1169-88.
Santhi C, Srinivasan R, Arnold JG, Williams JR. A modeling approach to evaluate the impacts
of water quality management plans implemented in a watershed in Texas. Environ Modell
Softwer. 2006;21:1141-57.

Tpicos Ci. Solo, 9:241-290, 2015

Modeling surface hydrology, soil erosion, nutrient transport, and future scenarios...

289

Santos I, Andriolo MV, Gibertoni RC, Kobiyama M. Use of the SWAT model to evaluate the
impact of different land use scenarios on discharge and sediment transport in the
Apucaraninha River watershed, Southern Brazil. In: Banasik K, Horowitz AJ, Owens PN,
Stone M, Walling DE, editors. Sediment dynamics for a changing future. Wallingford: IAHS
Press; 2010. p.322-8.
Schuol J, Abbaspour KC, Srinivasan R, Yang H. Estimation of freshwater availability in the
West African sub-continent using the SWAT hydrologic Model. J Hydrol. 2008;352:30-49.
Setegn SG, Srinivasan R, Melesse AM, Dargahi B. SWAT model application and prediction
uncertainty analysis in the Lake Tana Basin, Ethiopia. Hydrol Process. 2010;24:357-67.
Sharpley AN, Williams JR. EPIC-Erosion Productivity Impact Calculator: I model documentation.
Washington: USDA; 1990. (Technical Bulletin, 1768)
Sharpley AN. Modeling phosphorus movement from agriculture to surface waters. In: Radcliffe
DE, Cabrera ML, editors. Modeling phosphorus in the environment. Boca Raton, [United
States]: CRC Press; 2007. p.163-88.
Sharpley AN, Jones CA, Cole CV. A simplified soil and plant phosphorus model. II: prediction
of labile, organic, and sorbed phosphorus. Soil Sci Soc Am J. 1984;48:805-9.
Silva VA, Moreau MS, Moreau AMSS, Rego NAC. Uso da terra e perda de solo na Bacia
Hidrogrfica do Rio Colnia, Bahia. R Bras Eng Agric Amb. 2011;15:310-5.
Singh A, Imtiyaz M, Isaac RK, Denis DM. Comparison of soil and water assessment tool
(SWAT) and multilayer perceptron (MLP) artificial neural network for predicting sediment
yield in the Nagwa agricultural watershed in Jharkhand, India. Agric Water Manage.
2012;104:113-20.
Singh VP, Woolhiser DA. Mathematical modeling of watershed hydrology. J Hydrol Eng.
2002;7:270-92.
Singh VP. Computer models of watershed hydrology. Highlands Ranch, Colorado: Water
Resources Publications; 1995.
Strauch M, Volk M. Setting up SWAT for the Upper Amazon. In: Book of Abstracts of the
International Swat Conference; Porto de Galinhas, [Br]: 2014. p.6.
Strauch M, Lima JEFW, Volk M, Lorz C, Makeschin F. The impact of Best Management
Practices on simulated streamflow and sediment load in a Central Brazilian catchment. J
Environ Manage. 2013;127:S24-S36.
Taffarello D, Bressiani DA, Calijuri MC, Mendiondo EM. On contrasting field evidences of
water quality to perform physically-based SWAT simulations in challenging Brazilian biome
under change. In: Book of Abstracts of the International Swat Conference; Porto de Galinhas,
[Brl]: 2014. p.73-4.
United States Department Of Agriculture USDA. Soil Conservation Service, SCS. National
Engineering Handbook, Section 4: Hydrology. Washington: USDA/NHE; 1972. Pt 630.
United States Department Of Agriculture USDA. Soil Survey Staff. Soil taxonomy: A basic
system of soil classification for making and interpreting soil surveys. 2nd ed. Washington:
U.S. Department of Agriculture/Natural Resources Conservations Service, Agriculture
Handbook; 1999.
Uzeika T, Merten GH, Minella JPG, Moro M. Use of the SWAT model for hydro-sedimentologic
simulation in a small rural watershed. R Bras Ci Solo. 2012;36:557-65.

Tpicos Ci. Solo, 9:241-290, 2015

290

Nadia Bernardi Bonum et al.

Vadas PA, Krogstad T, Sharpley AN. Modeling phosphorus transfer between labile and non
labile soil pools: Updating the EPIC Model. Soil Sci Soc Am J. 2006;70:736-43.
van Griensven A, Bauwens W. Multi objective autocalibration for semi-distributed water
quality models. Water Resour Res. 2003;39:1348-56.
van Griensven A, Meixner T, Grunwald S, Bishop T, Di Luzio M, Srinivasan R. A global
sensitivity analysis tool for the parameters of multi-variable catchment models. J Hydrol.
2006;324:10-23.
van Griensven A, Ndomba P, Yalew S, Kilonzo F. Critical review of SWAT applications in the
upper Nile basin countries. Hydrol Earth Syst Sci. 2012;16:3371-81.
von Stackelberg NO, Chescheir GM, Skaggs RW, Amatya DK. Simulation of the hydrologic
effects of afforestation in the Tacuarembo River basin, Uruguay. Trans ASABE. 2007;50:45568.
Walter H, Lieth H. Klimadiagramm weltatlas. Jena, Germany: G. Fischer; 1960.
Wang W, Xie P, Yoo SH, Xue Y, Kumar A, Wu X. An assessment of the surface climate in the
NCEP climate forecast system reanalysis. Clim. Dynam. 2011;37:1601-20.
White ED. Development and application of a physically based landscape water balance in
the SWAT model. [dissertation] Ithaca, NY: Cornell University; 2009.
White KL, Chaubey I. Sensitivity analysis, calibration, and validations for a multisite and
multivariable SWAT model 1. J Am Water Resour As. 2005;41:1077-89.
Williams JR, Berndt HD. Sediment yield computed with universal equation. J Hydrol Eng Div.ASCE. 1972;98:2087-98.
Williams JR. Sediment routing for agricultural watersheds. Water Resour. Bull.
1975;11:965-74.
Wischmeier WH, Smith DD. Predicting rainfall erosion losses; a guide to conservation planning.
Washington: USDA; 1978.
Woolhiser DA, Smith RE, Goodrich DC. Kineros, A Kinematic Runoff and Erosion Model:
documentation and user manual. ARS77. Fort Collins, Colorado: USDA Agricultural Research
Service; 1990.
Wu Y, Chen J. Simulation of nitrogen and phosphorus loads in the Dongjiang River basin in
South China using SWAT. Front Earth Sci China. 2009;3:1-6.
Xie H, Lian Y. Uncertainty-based evaluation and comparison of SWAT and HSPF applications
to the Illinois River Basin. J Hidrol. 2013;481:119-31.
Yang J, Reichert P, Abbaspour KC, Xia J, Yang H. Comparing uncertainty analysis techniques
for a SWAT application to the Chaohe Basin in China. J Hydrol. 2008;358:1-23.
Young RA, Onstad CA, Bosch DD, Anderson WP. AGNPS: A nonpoint-source pollution model
for evaluating agricultural watersheds. J Soil Water Conserv. 1989;44:168-73.
Zhang L, Oneill AL, Lacey S. Modeling approaches to the prediction of soil erosion in
catchments. Environm Software. 1996;11:123-33.

Tpicos Ci. Solo, 9:241-290, 2015

You might also like