2001 Book Chapter The WEPP Model (Flanagan Et Al, 2001)

Chapter 7
The Water Erosion Prediction Project (WEPP) Model
Dennis C. Flanagan, James C. Ascough II, Mark A. Nearing, John M. Laflen

United States Department of Agriculture - Agricultural Research Service
1. INTRODUCTION
Soil erosion by water continues to be a serious problem throughout the

world, and models play an increasingly critical role in conservation and
assessment efforts. Improved soil erosion prediction technology is needed to
provide land managers, conservationists and others with tools to examine the
impact of different land management decisions on on-site soil loss and off-site
sediment yield and determining optimal land use. Additionally, soil erosion
prediction technology allows policymakers to assess the current status of land
resources and the potential need for enhanced or new policies to protect soil and
water resources.
On-site monitoring of sediment loss is difficult, expensive and time-
consuming. Also, soil erosion events occur intermittently and long-term records
are needed to well-characterize erosion and sediment loss from any particular
site. Thus, soil erosion models are in most cases the primary tools for making
assessments. For example, as noted by Wischmeier and Smith (1978), the U.S.
Department of Agriculture (USDA) Natural Resources Conservation Service
(NRCS) has conducted periodic national resource inventories for many years
using the Universal Soil Loss Equation, USLE.
145
146 Flanagan, Ascough, Nearing, and Laflen
Erosion prediction models are also critical in conservation planning

activities. Models assist land managers in evaluating the impacts of different
management practices (e.g., tillage and cropping systems) on possible soil loss,
and if a particular set of practices can meet a soil loss tolerance goal. Engineers
can also use erosion models to determine the expected sediment delivery during
design of structures such as sediment retention ponds and reservoirs. A design
change may be warranted if land use surrounding a reservoir is projected to
change in the future and impact possible sediment delivery.
This chapter provides information on the development, science, testing, and
status of a process-based erosion prediction model created by the USDA-
Agricultural Research Service (ARS), NRCS, the U.S. Forest Service (FS), and
the U.S. Department of Interior-Bureau of Land Management (BLM). The
Water Erosion Prediction Project (WEPP) model is a distributed parameter,
continuous simulation model with a daily time step, implemented on personal
computers, and designed to simulate small field-sized watersheds and hillslope
profiles within those watersheds. WEPP has been successfully used for many
types of land management, including croplands, rangelands, forest roads and
harvest areas, and construction sites.
2. WEPP MODEL DEVELOPMENT HISTORY
WEPP began in the early 1980’s with the recognition by the soil erosion
research and user community that the USLE technology was not currently
meeting soil erosion prediction needs, and would not do so in the future. The
USLE had many limitations - it ignored many interactions, did not estimate
erosion from irrigation, and did not represent snowmelt erosion. The soil erosion
process was well described and the USLE did not adequately describe these
processes. However, the major problem always was, and still is, to develop a
technology that could be used by the field practitioner. W. Wischmeier, the
developer of the USLE, recognized all these shortcomings, but indicated
(personal communication) that they were extremely difficult to bring into a
practical tool for field use.
Meyer and Wischmeier (1969) published a paper implementing the rill-
interrill concept of soil erosion in a model. The CREAMS (Chemicals, Runoff,
and Erosion from Agricultural Management Systems) model (Knisel, 1980) had
embodied similar concepts with considerable expansion of the science. Foster
(1982) published a paper where modeling concepts for erosion prediction were
presented. In a symposium in 1983, both Foster et al. (1985a) and Laflen (1985)
7. The Water Erosion Prediction Project (WEPP) Model 147
made the case for replacing the USLE. It was felt by many scientists that
supporting technology was, or would soon be, available so that the concepts
enunciated by Ellison in 1947 (a,b) could be implemented to predict soil erosion
on the world's lands.
At the symposium in 1983, Foster et al. (1985a) set forth requirements for a
USLE replacement, and major features such a replacement should have. Initial
requirements included:
• The application to more situations than the USLE, and be more
"powerful" than the USLE;
• Be accurate for prediction of both individual storm response and long-
term soil loss;
• Be fundamentally-based by having separate components for interrill
erosion, rill erosion, sediment transport, and deposition;
• Estimate both erosion and deposition along a non-uniform slope
profile, and provide an estimate of sediment yield from a field sized
watershed;
• Estimate deposition in small impoundments; and
• Be practical for field use by conservation technicians.
Some of the specific features mentioned that the replacement should have
included:
• A set of equations rather than a single equation;
• Programmed on a battery powered, portable computer (not a
programmable calculator);
• Hydrologically based and driven by rainfall and runoff variables from
a climate generator and a hydrology model;
• Include a simple crop growth model;
• Use storm amounts and peak intensities in its computations and not
storm hydrographs; and
• Contain a degree of complexity somewhere between the USLE and
CREAMS models.
In 1985, a workshop in Lafayette, Indiana was hosted by the ARS and
National Soil Erosion Research Laboratory (NSERL). At that meeting it was
agreed that the USLE would be revised (RUSLE) and that a major effort would
be initiated to replace the USLE technology. Major federal agencies agreed to
pursue the development of the replacement technology-including ARS, the Soil
Conservation Service (SCS, now the NRCS), the Forest Service and the Bureau
of Land Management. A team of 12 ARS, NRCS, FS and BLM employees was
formed to provide leadership of the various facets of the project.
The WEPP Core Team developed the objectives of the project, target users
were identified, and a project time frame created. A set of user requirements and
an experimental research program were planned. The main components and
features of the model were also identified. An initial prototype model for
application to hillslope profiles was released in 1989, and a complete validated
and documented model for application to profiles and small watersheds was
released in 1995.
2.1. User Requirements
One of the major milestones in the project was the development of a set of
User Requirements (Foster and Lane, 1987). These requirements detailed what
the users expected to be included in the technology to be produced.
The process followed in developing the User Requirements was to visit each
NRCS technical center and meet with interested parties to discuss the project,
and to agree on what was needed. These meetings were held at Portland
(Oregon), Lincoln (Nebraska), Fort Worth (Texas), and Chester (Pennsylvania).
Erosion prediction committees, technical steering committees and individuals in
all the cooperating agencies provided major input to the User Requirements.
Discussions at WEPP Core Team Meetings resolved differences between user
expectations and the project’s ability to meet potential requirements. The User
Requirements were approved by the Administrators of the Agricultural Research
Service (T. B. Kinney), Soil Conservation Service (W. Scaling), Forest Service
(R. M. Peterson) and Bureau of Land Management (R.F. Burford).
The User Requirements covered areas related to the development,
implementation and use of the model. Almost all areas related to WEPP were
covered. These included: objectives, target users, technology to be developed,
ease of use, areas of applicability, and how the technology might be delivered,
implemented and used. Major emphasis was on developing a practical tool for
the user. There was considerable discussion related to how the user agencies
would implement the model, and what kind of management, conservation
practices and surfaces the model would be applied to.
The User Requirements were clear that the technology would be
implemented in a process-based computer software program. The model would
include hydrology, winter processes, plant growth, hydraulics, erosion and
management components. The technology was to be applied to cropland and
rangeland field-sized areas, perhaps as large as a square mile. It was not to apply
to incised channels, such as classical gullies, but was to apply to waterways,

ephemeral gullies, and terrace channels.
WEPP technology was to include sheet-rill erosion and deposition, and be
able to compute sediment delivery and sediment size distributions of delivered
materials. There were to be three versions - hillslope, watershed and grid. The
model was to represent the major factors in soil erosion - climate, topography,
soils, management and conservation practices. It was to apply to the entire
United States, including Alaska, Hawaii and Puerto Rico. Procedures used in the
model algorithms were to be physically process-based. Inputs were to be readily
available and easily understood. WEPP was to be robust and where possible be
similar to other models.
The technology was to be:..…“sufficiently accurate to lead to the planning
and assessment decision that would be made in the large majority of cases when
full information is available.” The model was to be validated and results
documented. It was to be valid when:
• it served its intended purpose as defined by the User Requirements;
• it was based on scientific principles and represented a reasonable
expression of current scientific understanding of erosion processes;
• it gave expected responses that appeared reasonable;
• it applied to situations not appropriate for the USLE and gave results
that were more useful for agency program objectives than those given
by the USLE;
• it provided a reasonable representation of data covering the range of
conditions of the key situations described above;
• its results were as good as those from the USLE, and judgments of
goodness of fit were based on data sets as a whole; and,
• the model stood up in public hearings of management plans and
assessments (Foster and Lane, 1987).
The User Requirements also described the implementation of the computer
program development procedure. This included model structure, model
components, coding, documentation, and similarities with other models.
2.2. WEPP Experimental Research Program
A range of experimental studies were conducted as part of the WEPP project

(Table 1). While there have been many experimental projects related to WEPP,
this is a list of studies that were officially part of the project and that were
critical to development of the WEPP technology.
Table 1. Major WEPP experimental projects.
Study Objectives Publication(s)

Soil erodibility for forest conditions, hydraulics of overland Burroughs et al. (1992),
flow. Katz et al. (1995)
To develop relationships for: (a) partitioning flow between rills;
(b) estimating rill width from flow rate information; (c)
Gilley et al. (1990),
predicting hydraulic roughness coefficients within rills; and (d) Gilley and Finkner (1991)
estimating hydraulic roughness coefficients on interrill areas
from random roughness information.
To develop relationships for predicting the influence of crop Gilley et al. (1991, 1992),
residues, standing crops, and gravel and cobble materials. Gilley and Kottwitz (1994)
To relate rill & interrill erodibility and critical hydraulic shear to Laflen et al. (1991)
soil properties for cropland soils. Elliot et al. (1989)
Predicting effects of time after tillage on soil erodibility. Morrison et al. (1994)
To relate rill and interrill erodibility and critical hydraulic shear Simanton et al. (1987),
to soil properties for rangeland soils. Simanton et al. (1991)
Predicting effect of cropping systems on interrill and rill soil West et al. (1991, 1992)
erosion.
There was no technology to predict rill and interrill erodibility and critical
hydraulic shear for cropland and rangeland soils when WEPP was initiated.
Furthermore, little was known about erosion processes on forestlands. The team
also recognized that information on the hydraulics of rill flow in the presence of
residue and vegetation was incomplete, and improvements in this area were
needed to successfully implement WEPP. A major knowledge gap was
estimation of rill width and spacing, and hydraulic roughness. Data were also
quite limited for predicting the effects of time after tillage and of management
system effects on rill and interrill erodibility and critical hydraulic shear. The
experiments listed here were intended to overcome some of these shortcomings.
These studies, as well as others, provided many of the answers needed to
successfully implement WEPP.
Of particular importance were the studies by Laflen on 33 cropland soils
(Laflen et al., 1991) and Simanton on 18 rangeland sites (Simanton et al., 1991)
to provide the parameter values for erodibility, as well as to develop equations to
estimate erodibility from soil and other site characteristics. The parameter
estimation techniques provided a starting set of erodibility values for users to
apply the model.
3. WEPP HILLSLOPE MODEL COMPONENT
3.1. Introduction
Application of the WEPP model to hillslope profiles is similar in function to

applications of the USLE, with minor differences. A profile needs to be selected
to represent the slope, soil, and cropping/management conditions of a field or
region of a field. Normally a profile is chosen starting at the top of a hill and
continuing down through any depositional regions until a concentrated flow
channel or the end of the field is reached (applications of USLE would only be
made on the steeper portions of a slope and end at deposition areas). Regions of
homogeneous soils and cropping management on a hillslope profile are treated
as unique overland flow elements (OFE), on which the model keeps track of
water balance, plant growth, residue levels, surface roughness, and infiltration
and erodibility parameter values. Multiple OFEs allow simulation of regions of
different soils with different infiltration and detachment characteristics, as well
as strip cropping management on a hillslope (Figure 1).
Figure 1. Hillslope applications of WEPP can simulate nonuniform slope shapes, soils and
cropping scenarios on multiple Overland Flow Elements.
This section of the chapter will describe the various components of WEPP
that comprise the hillslope model, including weather generation, irrigation,
hydrology, water bala nce, plant growth, residue decomposition, overland flow
hydraulics, and erosion.
3.2. Weather Generation
WEPP is a continuous simulation model with a daily time step, and can use
either observed or generated climatic inputs to drive the runoff and erosion
processes. The CLIGEN (Nicks et al., 1995) weather generator was developed
specifically to create daily climate inputs for WEPP, based upon long-term
weather station statistics. CLIGEN is a stand-alone FORTRAN program that is
run separately from the WEPP FORTRAN program.
CLIGEN generates daily precipitation depth, storm duration, rainfall
intensity characteristics, minimum and maximum daily temperatures, dew point
temperature, solar radiation, and wind direction and speed. Precipitation events
are predicted with a two-stage Markov chain, using the conditional probabilities
of a wet day following a dry day and a dry day following a wet day. Monthly
values for these probabilities are derived from long-term weather station data.
Using the initial condition that the previous day was either wet or dry, CLIGEN
determines if precipitation will occur on a given day by comparing a generated
random number to the wet-dry probability value.
Precipitation depth is determined by sampling a skewed normal distribution.
Storm duration is predicted from an exponential relationship that uses monthly
mean half-hour precipitation depth information. CLIGEN contains a
disaggregation procedure to provide inputs to WEPP to generate time-rainfall
intensity (breakpoint) data for a storm event. Peak storm intensity (rp ) is
computed using:
rp = -2 P ln(1-rl), (1)
where P is the total precipitation depth (mm) and rl is a dimensionless

parameter from a gamma distribution of the monthly mean half-hour
precipitation depths (Arnolds and Williams, 1989; Nicks et al., 1995).
Time to peak storm intensity is predicted by sampling the accumulated
distribution of time to peak information that can be obtained from 15 minute
rainfall intensity data. CLIGEN provides a nondimensional time to peak
intensity and peak intensity rate to the WEPP model, which then simulates a
single-peaked storm by creating breakpoint data internally assuming a double
exponential function for storm intensity. Nicks et al. (1995) provide much more
detail on the rainfall disaggregation procedures.
Daily minimum air temperature, maximum air temperature, dew point
temperature, and wind speed are predicted by CLIGEN assuming normal
distributions with the equation:
Wpred = Wmean obs + STobs (vsn ), (2)
where Wpred is the predicted weather variable of interest, Wmean obs is the
observed long-term monthly mean value for the variable of interest, STobs is the
standard deviation of the observed long-term weather data, and v sn is the
standard normal deviate. Solar radiation is predicted using a function similar to
Equation (2), but substituting a standard normal variate for the standard normal
deviate there. Wind direction for a day is selected based upon random sampling
of wind direction information for a station.
CLIGEN was originally released with about 1000 stations parameterized.
Recently, the FS has reanalyzed data from Nicks et al. (1995) and enhanced and
expanded the available stations to a list of over 2600 (see
http://forest.moscowfsl.wsu.edu/fswepp).
3.3. Irrigation
WEPP has the capability to simulate either stationary sprinkler or furrow

irrigation water applications. Stationary sprinkler irrigation assumes that
irrigation water is applied via nozzles from above at a uniform rate across a
single OFE. Furrow irrigation simulations assume that inflow water is applied to
the top of furrows (or rills) running up and down an OFE. There are also three
types of irrigation scheduling permitted: fixed date, depletion level, or a
combination of the two. In depletion level scheduling, the model continually
checks the status of the soil moisture level; if it becomes too dry an irrigation
water application is initiated.
Sprinkler irrigation water is handled identically to that from natural rainfall,
and the model uses all of the same infiltration, runoff and soil loss computations.
However, in simulating the infiltration of water into a furrow, a Kostiakov-
Lewis infiltration function is used to predict two-dimensional flow into the
furrow walls (Kottwitz, 1995). Conservation of mass and kinematic wave theory
are used to solve for the peak runoff rates for a furrow channel. The four stages
of flow in a furrow that are addressed by WEPP are advance, continuing,
depletion, and recession. For furrow events, the important parameters passed to
the erosion component include the peak flow rate and the total runoff depth.
Only rill erosion is computed by the model during a furrow irrigation event.
3.4. Hydrology
3.4.1. Infiltration and Runoff
The hydrology components of WEPP are critical towards providing rainfall

and runoff information to the erosion component. Infiltration, surface runoff,
and soil water balance processes are simulated. The four ma in parameters
required from the hydrology component for erosion prediction are duration of
rainfall excess, effective rainfall intensity, runoff depth, and peak runoff rate.
Also, in addition to providing the values to estimate event soil loss, any water
predicted to infiltrate is used to update the soil moisture content in a continuous
simulation.
A Green-Ampt Mein-Larson model (Mein and Larson, 1973) as modified for
unsteady rainfall (Chu, 1978) is used to predict the cumulative infiltration depth.
Infiltration is divided into two distinct stages: (1) a stage where the ground
surface is ponded (covered with water), and (2) a stage without ponding. During
unsteady rainfall, infiltration can change multiple times between the two stages.
When a soil surface is ponded, infiltration occurs at a rate independent of the
rainfall distribution, while when the soil is not ponded, all rainfall is predicted to
infiltrate.
Depressional storage is estimated as a function of random roughness and
slope steepness (Onstad, 1984). Rainfall water can infiltrate or act to fill the
depressional storage. When all depressional storage has been filled, any
additional rainfall excess goes into runoff generation (Stone et al., 1995).
Peak runoff rate is a very important parameter in WEPP, as it is used in
calculations to estimate flow depth and ultimately flow shear stress. WEPP uses
either a semi-analytical solution of the kinematic wave model (Stone et al.,
1992) or an approximation of the kinematic wave model to determine the peak
runoff rate (Stone et al., 1995). For hillslopes composed of multiple OFEs,
runoff depth is divided by the peak runoff rate from the terminal OFE to
determine the effective duration of runoff for an event, and then peak runoff
rates on all preceding OFEs are determined by dividing the individual OFE
runoff depths by the effective runoff duration.
Figure 2 depicts the rainfall, infiltration, and runoff rates for a hypothetical
storm event simulated by WEPP. One can notice in this figure that the runoff
rate used for erosion predictions is a constant rate at the level of the predicted
peak runoff rate.
Figure 2. Rainfall, infiltration, and runoff for a single storm event as simulated in the WEPP
model.
3.4.2. Water B alance
The water balance component of WEPP (Savabi and Williams, 1995) is

based on similar components in the SWRRB (Simulator for Water Resources in
Rural Basins) model (Williams and Nicks, 1985; Arnold et al., 1990).
Modifications have been made to improve some of the predictions of rainfall
interception, percolation, and soil evaporation parameters. The water balance
estimates the daily snow depth and snow evaporation and melt, the potential
evapotranspiration, soil evaporation, plant transpiration, soil water content in the
root zone, and percolation of infiltrated water down through the soil layers
(Figure 3).
WEPP uses either the Penman equation (Penman, 1963; Jensen, 1974) or the
Priestly-Taylor (1972) function to predict the potential evapotranspiration for a
simulation day, depending upon whether wind inputs are available to the model.
Information from the climate (temperature, solar radiation, wind), plant growth
(leaf area index, root depth), and residue decomposition (residue cover) model
routines are used in the computation of soil evaporation, plant transpiration, and
extraction of moisture from the soil layers. A deficit of moisture can cause plant
moisture stress that will reduce predicted biomass production. Also, low soil
moisture values can trigger the application of irrigation water under depletion
level scheduling.
Figure 3. WEPP hillslope profile (that can be from within a small watershed) hydrology processes
include precipitation, infiltration, runoff, plant transpiration, soil evaporation and percolation
(from Savabi and Williams, 1995).
The model tracks soil moisture to a maximum depth of 1800 mm, and any
water that moves below the root zone is not traced further. The top of a soil
profile is divided internally in the model into two layers, each 100 mm thick,
and any layers below the top two are 200 mm thick. Water in excess of field
capacity in an upper soil layer is percolated to the next lower soil layer using
storage routing techniques. Percolation of water downward can be restricted by
lower soil layers that are at or near saturation. WEPP also contains a component
to simulate subsurface lateral flow and flow to drainage tile and ditches (Savabi
et al., 1995).
3.5. Soil Component
The soil component of WEPP is responsible for the estimation of baseline

soil infiltration and erodibility properties, as well as adjustments to these
parameters during execution of continuous model simulations. The four major
hydrologic parameters estimated are random roughness, ridge height, bulk
density, and effective hydraulic conductivity. The three erosion parameters are
interrill erodibility, rill erodibility, and critical hydraulic shear stress. Many
factors can affect all of these parameters, but the most important factor in many
cropland situations is the impact of tillage operations. Only effective
conductivity and erodibility parameters will be discussed in this section. The
reader is referred to Alberts et al. (1995) for full details on all soil parameter
calculations.
3.5.1. Effective Hydraulic Conductivity
Effective hydraulic conductivity is a critical parameter in WEPP model

simulations, as this value and any adjustments to it directly impact the amount
and rates of infiltration and runoff. The model can be applie d using either
constant values for conductivity, or temporally-varying ones. Alberts et al.
(1995) provide information on how to estimate time-invariant values for
cropland and rangeland.
However, to best represent the impacts of management (especially on
croplands), a better approach is for the user to input a baseline effective
conductivity value for their soil, and have the model continuously adjust the
daily value based upon the management, crop, soil, and residue status. Since
users likely would not be able to measure the baseline conductivity value for
their soil, extensive parameterization work was conducted using runoff plot data.
For croplands, baseline effective conductivity values were developed through
model optimization runs on 43 soils using measured and curve number
predictions under continuously tilled fallow management (Alberts et al., 1995).
The resulting equations within WEPP that can be used to predict conductivity in
cropland simulations are:
Kb = -0.265 + 0.0086 SAND1.8 + 11.46 CEC-0.75, (3)
for soils having clay content less than or equal to 40 percent, and with:
Kb = 0.0066 e (244/CLAY) , (4)
for soils with greater than 40 percent clay content, where Kb is effective
hydraulic conductivity in mm⋅h-1 , SAND is percent sand content in the surface
soil, CLAY is percent clay content in the surface soil, and CEC is the cation
exchange capacity in the surface soil in meq/100g (Alberts et al., 1995; Flanagan
and Livingston, 1995).
Several adjustments for row crops were developed using data from natural
rainfall studies on fallow, row-cropped, and perennial-cropped plots (Risse et
al., 1994; Zhang et al., 1995a, 1995b). A major adjustment is for the effects of
soil crusting and tillage, which are affected by the amount of surface cover, the
soil random roughness, the cumulative rainfall kinetic energy since the last
tillage operation, a soil stability factor, and a crust factor defined by Rawls et al.
(1990). In row-cropped situations, the adjustments to conductivity are a function
of effective canopy cover, residue cover, and the storm rainfall amount. For
perennial crops, analysis of 88 plot-years of measured data under perennial
crops showed that on the average the effective conductivity value is
approximately 1.8 times greater than that from corresponding row crop
conditions. Additionally, hydraulic conductivity is also adjusted due to frozen
soil conditions. More details on hydraulic conductivity adjustments are provided
in Alberts et al. (1995).
For rangela nd conditions on which the rill surface cover is less than 45
percent, the effective conductivity for rangelands is predicted using:
Kerange = 57.99 - 14.05 ln (CEC) + 6.20 ln (ROOT10) - 473.39 BASR2 + 4.78 RESI , (5)
while for rangeland conditions where rill cover is greater than or equal to 45
percent,
Kerange = -14.29 - 3.40 ln (ROOT10) + 0.3783 SAND + 2.0886 ORGMAT +

398.64 RR- 27.39 RESI + 64.14 BASI , (6)
where Kerange is effective rangeland hydraulic conductivity in mm⋅h-1 , CEC is

cation exchange capacity in meq/100g, ROOT10 is root biomass in the top 10
cm of the soil in kg⋅m-2 , BASR is the product of the fraction of basal surface
cover in rill areas and total basal surface cover, RESI is the product of the
fraction of litter surface cover in interrill areas and the total litter surface cover,
SAND is percent sand content of the surface soil, ORGMAT is percent organic
matter in the surface soil, RR is soil surface random roughness in m, and BASI is
the product of the fraction of litter surface cover in interrill areas and the total
basal surface cover (Alberts et al., 1995).
3.5.2. Soil Erodibility
Soil erodibility parameters directly impact the rates of detachment from

interrill and rill erosion areas. Field experimentation conducted on 33 cropland
soils (Elliot et al., 1989) and 18 rangeland sites (Simanton et al., 1987) provided
information that allows the parameters to be estimated from site-specific soil
properties. For cropland situations, a set of baseline erodibility values must be
input to the WEPP model that represent a freshly-tilled soil with minimal
residue cover. Adjustments are then made on a daily basis to alter the erodibility
values based upon a range of cover, residue, and other factors.
For cropland soils with a sand content of 30 percent or more, the erodibility
estimation equations are:
Kib = 2728000 + 192100 VFS, (7)
Krb = 0.00197 + 0.00030 VFS + 0.03863 e -1.84 ORGMAT , (8)
and
τc = 2.67 + 0.065 CLAY - 0.058 VFS, (9)
while for cropland soils having less than 30 percent sand, the equations are:
Kib = 6054000 - 55130 CLAY, (10)
Krb = 0.0069 + 0.134 e -0.20 CLAY , (11)
and
τc = 3.5 (12)
(Flanagan and Livingston, 1995; Alberts et al., 1995), where Kib is baseline
interrill erodibility in kg⋅s⋅m-4 , Krb is baseline rill erodibility in s⋅m-1 , τc is
baseline critical shear stress in Pa, VFS is percent very fine sand in the surface
soil, CLAY is percent clay in the surface soil, and ORGMAT is percent organic
matter in the surface soil.
On cropland, interrill erodibility is adjusted daily for a large number of
factors, including canopy cover, ground cover, roots, sealing and crusting, and
freezing and thawing. Rill erodibility is adjusted for incorporated residue, roots,
sealing and crusting, and freezing and thawing effects. Critical shear stress is
adjusted daily for the effects of random roughness, sealing and crusting, and
freezing and thawing.
For rangeland situations, erodibility parameters are predicted using the

following equations (Flanagan and Livingston, 1995; Alberts et al., 1995):
Kirange = 1810000 - 19100 SAND - 63270 ORGMAT - 846000 θfc , (13)
Krrange = 0.0017 + 0.000024 CLAY - 0.000088 ORGMAT

- (0.00088 BD dry/1000) - 0.00048 ROOT10 , (14)
and
τcrange = 3.23 - 0.056 SAND - 0.244 ORGMAT + (0.9 BDdry/1000) , (15)
where Kirange is baseline interrill erodibility in kg⋅s⋅m-4 , Krrange is baseline rill

erodibility in s⋅m-1 , τcrange is baseline critical shear stress in Pa, BDdry is the dry
soil bulk density in kg⋅m-3 , θfc is the volumetric water content of the soil at 0.033
MPa in m3 ⋅m-3 , and SAND, CLAY, ORGMAT, and ROOT10 are as previously
defined.
Adjustments are made to rangeland interrill erodibility for ground cover and
freezing and thawing effects. Freezing and thawing adjustments are also made to
the rangeland rill erodibility and critical shear stress values.
3.6. Plant Growth
Simulation of plant growth is very important in a continuous simulation

erosion model, since the production of biomass and residue, and the interaction
of management and environmental factors can greatly impact predicted soil loss.
At present, cropland and rangeland plant growth are simulated within WEPP
using somewhat different approaches.
For cropland plant growth, WEPP uses approaches similar to the EPIC
(Erosion Productivity Index Calculator) model (Williams, 1995), in which crop
growth is a function of daily heat unit accumulation. Potential crop growth for a
day is based on the product of a biomass energy conversion ratio multiplied by
the photosynthetic active radiation for the day. Canopy cover, canopy height,
and leaf area index are functions of the predicted amount of vegetative biomass.
Potential biomass produced can be decreased by either water or temperature
stress factors. User inputs allow description of senescence periods in which
biomass and canopy cover values decline. Cropland plant growth options
include annual crops, perennial crops, grain harvest, silage, hay harvest, and
grazing.
For rangeland situations, the model uses either a unimodal or a bimodal

potential growth curve that is meant to describe the growth characteristics for a
plant community (Arnold et al., 1995). Input information is needed on the
maximum potential biomass production and when the peaks occur. The user
must provide information on the grasses, shrubs, and trees present. Management
for rangeland situations includes grazing (biomass removal/conversion), burning
(residue removal/change in plant community make-up), and herbicide
application (change in plant community make-up).
3.7. Residue Decomposition and Management
Plant residues are often the most cost-efficient way to achieve soil
conservation goa ls. Thus, a continuous simulation erosion model must be able to
predict the creation and loss of residue materials due to crop production, residue
decomposition, and residue management operations.
Residue decomposition is due to the activity of microorganisms that are
present in most environments and use the plant residues as food. The WEPP
model currently uses somewhat different approaches to estimate residue
decomposition in cropland and rangeland simulations, though it is hoped that at
some point in the future the two procedures can be merged into one.
On croplands, WEPP uses a decomposition day approach in which there is an
optimal decomposition rate computed assuming ideal conditions (no
constraints). This optimal rate can then be decreased due to environmental
(temperature, moisture), soil fertility, and residue particle size factors (Stott et
al., 1995). At present, only the water and temperature environmental stress
factors are active in WEPP. Thus the main equation used is:
M t,j = M t-1,j eENVINDj ⋅ORATEj , (16)
where t is a day counter and j is a residue type counter, M t,j is the mass of
residue today in kg⋅m-2 , M t-1,j is the mass of residue the previous day in kg⋅m-2 ,
ORATEj is the decomposition rate constant in kg⋅m-2 ⋅d-1 for a given residue type,
and ENVINDj is the environmental factor that determines the fraction of a
decomposition day that has occurred during day t . The environmental index is
computed using:
ENVIND = Minimum (WFC, TFC), (17)

where WFC is the daily water factor and TFC is the daily temperature factor.
Stott et al. (1995) provide details on how the water and temperature factors are
computed for different residue materials (standing, flat, buried) as a function of
precipitation, soil moisture, and air temperatures.
Decomposition of residue materials in rangeland simulations uses an older
set of equations based on the work of Ghidey et al. (1985):
Rg t = Rg t -1 (1 - [αf (Smi Tavg)/Cn ]2 ) - Bc , (18)
where Rgt is the residue mass today in kg⋅m-2 , Rgt-1 is the residue mass the
previous day in kg⋅m-2 , α f is a litter decay coefficient, Smi is the total rainfall
depth in the past 5 days in m, Tavg is the average daily temperature in o C, Cn is
the carbon-to-nitrogen ratio of the residue, and Bc is the daily disappearance of
residue due to insects and rodents in kg⋅m-2 (Stott et al., 1995).
The WEPP model allows for considerable flexibility in terms of residue
management. Every tillage operation has a tillage intensity value that can be
used to estimate standing to flat residue conversion, as well as burial of flat
residue. There are also special residue management options in cropland
simulations that allow for burning, shredding/cutting, residue removal, and
residue addition.
3.8. Overland Flow Hydraulics
Surface runoff is represented in WEPP in two ways: (1) broad uniform sheet
flow is assumed and the overland flow hydrograph is calculated using hydraulic
roughness values that are weighted averages of the rill and interrill areas, and (2)
flow is partitioned into broad uniform sheet flow for interrill erosion
calculations and concentrated flow for rill erosion calculations. Field tests by
Gilley et al. (1990) at eleven sites found rills to form on average at about 1.0 m
spacings on cropland soils, and this is used as the model default. There is also an
option in cropland simulations to input a user-specified value for rill spacing.
For rangeland simulations, the rill spacing is confined within a range of 0.5 to
5.0 m, and the model determines the value based upon the number of range
plants growing along a 100 m transect. Rill width used in the model for a storm
event is either a constant value input by the user, the rill width from a previous
larger storm event, or a new rill width computed using:
wrill = 1.13 ( Qrill )0.303, (19)

where wrill is the rill width in meters, and Qrill is the flow discharge rate in m3 ⋅s-1
(Gilley and Weltz, 1995).
WEPP assumes that rills have rectangular cross sections, and uses the Darcy-
Weisbach equation to describe the flow characteristics. The model computes
total Darcy-Weisbach friction factors for cropland rills that are a function of rill
surface roughness, residue, and live plants in the rills. Friction factors on
cropland interrill areas are a function of interrill surface roughness, interrill
surface cover, smooth bare soil roughness, and live plants on the interrill areas.
Friction factors for rangeland situations are impacted by many of the same
things as for those on cropland, but include additional effects such as those due
to rocks and cryptogams. See Gilley and Weltz (1995) for complete details.
An iterative solution of the uniform flow equation is used to determine the
depth of flow in the rill channels, and flow shear stress can then be computed
with the equation:
τ = γ R sin(arctan(Srill)), (20)
where τ is flow shear stress (Pa), γ is the specific weight of water in kg⋅m-2 ⋅s-2 , R
is the hydraulic radius in m, and S rill is the bed slope in the rill in m⋅m-1 . The
total shear stress is then partitioned between that acting on the soil and that
acting on other roughness elements using a ratio of the Darcy-Weisbach friction
factors, and the shear stress acting on the soil is used in the soil detachment and
sediment transport computations (Foster et al., 1995).
3.9. Soil Erosion
In WEPP hillslope profile simulations, the model separates the upland

erosion processes into those caused by excess flow shear detachment in rill
channels, and interrill detachment caused by raindrop impact and shallow rain-
impacted flows.
The soil erosion computations in the WEPP hillslope profile erosion
component use solutions to a steady state sediment continuity equation to
estimate values of sediment load and net detachment or deposition at points
down a profile. The governing equation is:
dG/dx = Df + D i , (21)
where G is sediment load in kg⋅s-1 ⋅m-1 , x is distance downslope in m, Df is rill

erosion rate in kg⋅s-1 ⋅m-2 , and Di is interrill sediment delivery rate in kg⋅s-1 ⋅m-2
(Foster et al., 1995).
Interrill sediment delivery to rills is predicted in WEPP using the equation:
Di = Kiadj Ie σir SDRRR Fnozzle (Rs /wrill), (22)
where Kiadj is the adjusted interrill erodibility factor in kg⋅s⋅m-4 , Ie is effective

rainfall intensity in m⋅s-1 , σir is the interrill runoff rate in m⋅s-1 , SDRRR is a
sediment delivery ratio that is a function of random roughness, row side-slope
and the interrill particle size distribution, Fnozzle is an adjustment factor to
account for sprinkler irrigation nozzle impact energy variation (value of 1.0 for
natural rainfall conditions), Rs and wrill are the respective rill spacing and rill
width in m.
Effective rainfall intensity is estimated during the infiltration computations
using the equation:
Ie = (1/te) ∫ Idt, (23)
where Ie is the effective rainfall intensity in m⋅s-1 , I is breakpoint rainfall

intensity in m⋅s-1 , t is time in seconds, t e is the total time over which rainfall rate
exceeds infiltration rate, and the integral is evaluated over time t e.
The interrill sediment delivery ratio, SDRRR, is determined through three
steps. First, an interrill roughness factor is calculated based upon a functional
representation of Table 8.4 in Foster (1982) using the equation:
RIF = -23 RR + 1.14, (24)
where RIF is the interrill roughness factor and RR is random roughness in m of

the current overland flow element. Random roughness values used in this
function are limited between 0.0061 and 0.0496 m (RIF va lues of 1.0 to 0.0,
respectively).
The second step of the procedure is to compute a delivery ratio (DRi ) for
each of the WEPP particle size classes using the interrill roughness factor and
the fall velocity of each size class. The five WEPP particle size classes,
predicted using relationships from Foster et al. (1985b) are primary clay,
primary silt, primary sand, small aggregate and large aggregate. For particles
with fall velocities (vfi) less than 0.01 m⋅s-1 the relationship used is:
DRi = az (RIF) bz, (25)
where
az = e (0 . 0672 + 6 . 59 v fi )
, (26)
and
bz = 0.1286 + 2209 v fi , (27)
and for particles with fall velocities greater than or equal to 0.01 m⋅s-1 the
relationship used is:
DRi = 2.5RIF - 1.5. (28)
The subscript i represents the individual particle size class, and delivery ratio
values are constrained to between 0 to 1 within the model.
The final step to determine the interrill sediment delivery ratio is to take a
weighted average of the sediment delivery ratio for each particle size class,
weighted by the mass fraction of sediment in each class:
SDRRR = ∑ fdeti (DRi ) (29)
from i=1 to 5, where fdeti is the fraction of each size class predicted with the
Foster et al. (1985b) equations. Additionally, the fraction of sediment in each
size class that is delivered from the interrill areas to the rills (fideli ) is calculated
as:
fideli = f deti (DRi / SDRRR ). (30)
These values are used to update the flow sediment size classes at the end of each
detachment region and start of each deposition region in the rills.
Rill erosion rate may be either positive in the case of detachment or negative
in the case of deposition. Rill detachment in WEPP is predicted when the flow
sediment load is below transport capacity, and flow shear stress acting on the
soil exceeds critical shear stress. In that case, Df is predicted with:
Df = Kradj (τ - τcadj )[ 1 - (G/Tc)], (31)

where Kradj is the adjusted rill erodibility factor in s⋅m-1 , τ is flow shear stress in
Pa, τcadj is adjusted critical shear stress of the soil in Pa, G is sediment load in
the flow (kg⋅s-1 ⋅m-1 ), and Tc is flow sediment transport capacity in kg⋅s-1 ⋅m-1 .
Sediment transport capacity is computed using a simplified function of shear
stress raised to the 3/2 power, times a coefficient that is determined through
application of the Yalin (1963) equation at the end of the slope profile (Finkner
et al., 1989).
Deposition in rills is predicted when flow sediment load exceeds sediment
transport capacity. In this case the model predicts the rill erosion rate using:
Df = (β veff / q) [Tc - G] , (32)
where β is a raindrop-induced turbulence factor, v eff is an effective fall velocity

for the sediment in m⋅s-1 , q is flow discharge per unit width in m2 ⋅s-1 , and Tc and
G are as previously defined. Currently, β is assigned a value of 0.5 for rain-
impacted flows, and a value of 1.0 for other cases such as snow melt or furrow
irrigation erosion.
WEPP solves the sediment continuity equation for total sediment load using
normalized equations at 100 points down each overland flow element. In cases
in which sediment load is below transport capacity, a Runge-Kutta numerical
method is used to solve for sediment load at progressive points down an OFE
using Equations (21), (22), and (31). When sediment load exceeds transport
capacity, a closed-form solution is used with Equations (21), (22), and (32).
The erosion component of WEPP also contains equations that are used to
estimate the sediment particle sorting in rill channels due to selective deposition.
An analytic solution of the normalized sediment continuity equation is used to
estimate the particle size distribution reaching the end of a deposition region,
assuming that the total sediment load computed previously with Equation (32) is
correct. See Foster et al. (1995) and Flanagan and Nearing (2000) for details on
the procedure.
4. WEPP MODEL WATERSHED COMPONENT
4.1. Introduction
The WEPP model watershed component was developed as an extension of
the WEPP model hillslope component for predicting erosion effects from
various management practices; and to accommodate spatial and temporal
variability in topography, soil properties, and land use conditions within

agricultural and non-agricultural small watersheds. Figure 4 depicts an example
watershed containing the three primary elements of the watershed component -
hillslopes, channels, and impoundments. The channel and impoundment
elements encompass the watershed application (estimation of sediment yield
from small watersheds) of WEPP. This chapter section describes the WEPP
model watershed component conceptual framework and presents mathematical
representations of the processes simulated by the channel hydrology and erosion
elements. The processes simulated by the impoundment element are not
described here, however, impoundment effects on watershed component channel
peak discharge and time of concentration calculations are discussed.
- Flow Direction
OFE 1
Hillslope 1
Hillslope 2
OFE 2
I1 Channel 1
Impoundment C Hillslope 4
ha
nn
el F
low
I2 Channel 2
Hillslope 3 OFE 2 Hillslope 6

OFE 1 Channel 3
I3
low
dF
Watershed Outlet
an
erl
Hillslope 5
Ov
Outlet
Hillslope 7
Figure 4. Example of WEPP watershed model flow routing using hillslope, channel and
impoundment elements.
4.2. Watershed Component Development
4.2.1. Conceptual Framework
The hillslope hydrologic and erosion information required by the WEPP

model watershed component is stored in a pass file and includes: (1) storm
duration; (2) overland flow time of concentration; (3) a parameter α that
expresses the proportion of total rainfall occurring during overland flow time of
concentration; (4) runoff depth; (5) runoff volume; (6) peak runoff rate; (7) total
sediment detachment at the end of the hillslope; (8) total sediment deposition at
the end of the hillslope; (9) sediment concentration by particle size class at the
end of the hillslope; and (10) the fraction of each particle size in the eroded
sediment. Watershed configurations are represented by the manner in which
hillslope, channel, and/or impoundment elements feed watershed (channel and
impoundment) elements, and how the channels and impoundments are fed
(either from the top or laterally from the left or right). This is illustrated by the
watershed configuration shown in Figure 5. This watershed is similar to the
example watershed in Figure 4, except that the hillslope, channel, and
impoundment elements are isolated and numbered. The WEPP model hillslope
component calculates hydrologic and erosion information for hillslope elements
1 through 7 and creates a hillslope-to-watershed master pass file. Watershed
flow routing begins at the highest upstream watershed element (impoundment
element 8, fed by hillslope element 4). Flow routing then continues to the next
downstream element (channel element 9, fed by hillslope elements 3 and 5, and
impoundment element 8), and proceeds downstream through all of the
remaining watershed elements (impoundment elements 10 and 12, and channel
elements 11 and 13) until the watershed outlet is reached.
The direction from which upstream elements drain into a channel is always
relative to the direction of flow in the channel element. For an impoundment, it
is relative to the direction of flow in the next downstream channel. Some
restrictions apply to watershed element configuration, including: (1) hillslopes
are fed by nothing, and may feed channels and impoundments; (2) channels are
fed by hillslopes, other channels, and impoundments, and may feed other
channels and impoundments; and (3) impoundments are fed by channels and
hillslopes, and may only feed channels. Further explanation of watershed
configuration restrictions can be found in Flanagan and Livingston (1995).
Additional information necessary to run the channel element includes the
channel slope, soil, management, climate, and the channel hydraulic
characteristics (watershed channel) files. The channel slope, soil, management,
and climate files are nearly identical to the corresponding hillslope component
input files. Information required by the impoundment element includes an
impoundment structures inventory file, and a file containing impoundment
characteristics and stage-area-length relationships. A complete description of all
watershed component input files is given by Flanagan and Livingston (1995).
- Hillslope Elements
5 - Impoundment
4 Elements
8 9 - Channel Elements
6
10 11
3 7
2 12
13
1
- Flow Direction Watershed
Outlet
Figure 5. Example of WEPP watershed model flow routing using hillslope, channel and
impoundment elements.
4.2.2. Watershed Processes
The WEPP computer program modeling approach is a combination of

process-based modules (components) and physically-based empirical
relationships. A general description of primary watershed component elements
is presented below; channel hydrology and erosion elements are discussed later
in more detail. The modeling approach employed to represent watershed
systems in WEPP is briefly described by the following sources
• Channel hydrology and water balance (Ascough et al., 1997), as
represented by calculations for infiltration, evapotranspiration, soil
water percolation, canopy rainfall interception, and surface
depressional storage in the same manner as does the hillslope
hydrology component for overland flow areas. Rainfall excess is
calculated using a Green-Ampt Mein-Larson infiltration equation
(Mein and Larson, 1973). The peak runoff rate at the channel (sub-
watershed) or watershed outlet is calculated using either a modified
Rational equation or an equation used in the CREAMS model (Knisel,
1980).
• Channel erosion (Ascough et al., 1997), with the assumption that

watershed sediment yield is a result of detachment, transport, and
deposition of sediment on overland (rill and interrill) flow areas, and
permanent channel (limited to grassed waterways, terrace channels or
similar sized channels) or ephemeral gully flow areas. Flow depth and
hydraulic shear stress along the channel are computed by regression
equations based on a numerical solution of the steady state spatially
varied flow equation. The movement of suspended sediment (i.e.,
detachment, transport, and deposition) on rill, interrill, and channel
flow areas is based on a steady state erosion model (Foster and Meyer,
1972) that solves the sediment continuity equation.
• Impoundment trapping of incoming sediment, thus reducing
sediment yield at the watershed outlet. Impoundments represented in
the watershed component include terraces, farm ponds, and check
dams. Outflow hydrographs and sediment concentration are calculated
for various types of outflow structures suitable for both large (e.g.,
farm ponds) and small (e.g., terraces) impoundments including
culverts, filter fences, straw bales, drop and emergency spillways, and
perforated risers. Deposition of sediment in impoundments is
calculated assuming complete mixing and later adjusted to account for
stratification, non-homogeneous concentrations, and impoundment
shape. A continuity mass balance equation is used to predict sediment
outflow concentration. The impoundment element performs both
hydraulic and sedimentation simulations; a complete description of
these simulation processes may be found in Lindley et al. (1998a;
1998b).
4.2.3. Range of Application
The WEPP model watershed component was originally intended for use on
field-sized areas and conservation treatment units, with a maximum size field
limitation of roughly a section (~260 ha) (Foster and Lane, 1987). It was also
anticipated that the watershed component could be applied on rangeland
watersheds of up to 800 ha (Foster and Lane, 1987), but Baffaut et al. (1997)
recommended that it not be used on watersheds larger than 40 ha and that
hillslope lengths should not exceed 100 m. The watershed model has also been
used for non-agricultural applications to predict sediment yields from surface
mine watersheds (Elliot et al., 1993), and large forested areas (Elliot et al.,
1996). The watershed component is not applicable to areas containing classical

gullies or stream channels which may have the following hydrologic or erosion
processes: (1) headcut erosion; (2) sloughing of gully sidewalls; (3) seepage
effects on erosion in concentrated flow channels; (4) perennial stream channels;
and (5) partial area hydrology. The watershed component does not contain a
baseflow estimation component so it cannot be used for stream channel erosion
prediction; furthermore, it cannot be used for classical gully erosion prediction
because a failure mechanism component for gully sidewall sloughing is la cking.
However, the watershed component is applicable to constructed waterways (e.g.,
terrace channels and grassed waterways) and to concentrated flow and cropland
ephemeral gullies. In range and forestland applications, fields can include gullies
up to the size of typical concentrated flow gullies occurring in cropland fields
(channels ranging from one to two meters in width and up to one meter in
depth).
4.3. Channel Hydrology Processes
4.3.1. Runoff Volume
Surface runoff entering a channel is assumed to be the sum of: (1) lateral
inflow from hillslopes or impoundments; (2) flow into the channel inlet from an
upstream hillslope or impoundment; and (3) flow into the channel inlet from
upstream channels, and can be written as:
runoff v = runoff l + runoff I , (33)
where: runoff v = total channel inflow volume; runoff l = lateral inflow volume
from hillslopes or impoundments; and runoff i = channel inflow volume from
upstream hillslopes, impoundments, or channels, with all volumes in m3 . The
total channel inflow volume in m3 , runoffv , is divided by the physical channel
area in m2 to obtain the channel inflow runoff depth, runoff d in m.
The storm (event) duration for the channel, durc in s, is taken to be the
maximum duration of: (1) the storm duration of any watershed element
(hillslope, impoundment, or channel) that contributes surface runoff to the
channel; (2) the storm duration for the channel itself; or (3) the duration of any
sprinkler irrigation event occurring on the channel. Once the channel inflow
volume (runoff v) and depth (runoff d) are known, channel infiltration,
depressional storage, rainfall excess, and transmission losses are calculated. If
there is a precipitation event (rainfall, snow melt, or sprinkler irrigation) for the
current day, the precipitation statistics are passed to the disaggregation routines.
Cumulative channel infiltration is computed using an implementation of the
Green-Ampt Mein-Larson (GAML) model (Mein and Larson, 1973), as
presented by Chu (1978) for the case of unsteady rainfall and multiple times to
ponding. The basis for this implementation can be found in Stone et al. (1995).
Infiltration parameters for the channel are calculated and an average rainfall
excess rate for an interval is computed. Rainfall excess is the amount of rainfall
that does not infiltrate when rainfall intensity exceeds the infiltration rate.
Before the total rainfall excess amount is calculated, the volume is adjusted for
soil saturation conditions and depressional storage. The total rainfall excess
amount is then computed using the GAML model and treated as the preliminary
or initial channel runoff depth, rci in m.
Following the calculation of rci, there are four general cases which can arise
on a channel that determine the calculation of the final channel runoff depth, rcf:
• Case I: rci > 0; runoff d > 0. The first case occurs when there is
rainfall excess from both the upstream contributing watershed
elements and the channel itself. In this case, the channel inflow
depth, runoffd , is simply added to the initial channel runoff depth, rci.
• Case II: rci > 0; runoffd = 0. The second case occurs when there is no
channel inflow, but rainfall excess, rov c in m3 , is produced on the
channel itself. For Cases I and II, the final channel runoff volume, rov f
in m3 , and depth, rcf , are computed by subjecting rci to runoff volume
reduction caused by infiltration during hydrograph recession.
• Case III: rci = 0; runoff d > 0. The third case occurs when there is
channel inflow, but a precipitation event results in no rainfall excess
produced on the channel itself. This could occur if channel inflow was
due to irrigation and no precipitation or irrigation water is applied
directly to the channel. In this case, the channel runoff depth can also
be reduced through channel transmission losses.
• Case IV: rci = 0; runoff d = 0. The fourth case occurs when there is no
channel inflow or rainfall excess on the channel itself. In this case rcf
and rov f are set equal to zero, and no further calculations are
necessary.
4.3.2. Channel Water Balance
Channel water balance calculations are performed after channel inflow and
outflow have been computed. The channel water balance and percolation
routines are identical to those used in the WEPP hillslope model component.
Input from the climate, infiltration, and crop growth routines are used to
estimate soil water content in the root zone, soil evaporation, plant transpiration,
interception, and percolation loss below the root zone. A complete description of
WEPP model hillslope and watershed component water balance and percolation
routines is given by Savabi and Williams (1995).
4.3.3. Channel Peak Runoff Rate
The peak runoff rate entering a channel depends on the contributing

hillslope, channel, and impoundment ele ments. One hillslope, one
impoundment, or up to three channels may contribute runoff to a channel
element. The peak runoff rate calculations are performed only if the final
channel runoff volume (rov f) is greater than 0.001 m3 . Otherwise, the peak
runoff rate and the runoff duration are set equal to zero and calculations are
continued for the next downstream channel or impoundment element. If only
one watershed element contributes runoff to a channel, the peak runoff rate
entering the channel is set equal to the peak runoff rate leaving the contributing
element. For example, if a hillslope is the only watershed element contributing
runoff to a channel, then the peak runoff rate entering the channel is the peak
runoff rate leaving the final hillslope OFE. The same would be true for a single
impoundment contributing runoff to a channel.
The SCS (Triangular) Synthetic Hydrograph method (Huggins and Burney,
1982) is used when runoff from hillslopes, channels and impoundments merges
onto a channel or into an impoundment. The time-discharge hydrographs for
each watershed element contributing to the channel or impoundment are first
calculated. The time-discharge relationship for the combined-element flow
hydrograph is then calculated by taking the maximum base time for all element
hydrographs and superimposing the hydrographs together over that time period.
Finally, the peak runoff rate entering the channel or impoundment is set equal to
the largest discharge value on the superimposed hydrograph.
The watershed component channel element contains two methods for
estimating the peak runoff rate at the channel (sub-watershed) or watershed
outlet: (1) the modified Rational equation; and (2) the CREAMS peak runoff
equation. The modified Rational equation is recommended for estimating peak
runoff. The CREAMS peak runoff equation (Smith and Williams, 1980) was
statistically derived using data from watersheds much larger than the 40 ha
maximum watershed area recommended for watershed model applications. It is
expected that for applications to watersheds smaller than 40 ha the modified
Rational equation will produce more accurate peak discharge results, although
no formal study has been conducted to verify this assumption.
4.3.3.1. Modified rational equation
Implementation of the modified Rational equation in the channel element

closely follows the methodology used in the EPIC model (Williams, 1995), with
the exception that in WEPP the equation is used to calculate the peak runoff rate
at each channel outlet rather than at the watershed outlet as in EPIC. The
Rational equation can be written as:
q po = α rovf / (3600 tc ), (34)
where: qpo = peak runoff discharge in m3 ⋅s-1 at the channel or watershed outlet;
rovf = final channel runoff volume in m3 ; t c = time in hours of concentration at
the channel or watershed outlet; and 3600 = time conversion constant.
The dimensionless parameter α expresses the proportion of total rainfall that
occurs during t c, and is calculated for the final hillslope OFE, and for each
channel and impoundment watershed element. A generalized equation for the
channel or watershed outlet time of concentration can be estimated by adding
the overland, channel, and impoundment flow times in hours over the slowest
flow path and is given by:
tc = tcc + tcs + tci, (35)
where: t cc = average channel travel time; t cs = time of concentration for overland

flow; and tci = time of concentration for impoundments.
If the channel is a first order channel, and one or more hillslopes contribute
runoff to it, the time of concentration is computed with the equation:
tc = tcc + tcs max, (36)
where t cs max is the largest time in hours of concentration from the contributing
hillslopes.
If the channel is a first order channel, and one or more impoundments

contribute runoff to it, then a check is made for the maximum time of
concentration of any hillslope contributing to the impoundment(s) which
contribute to the channel and the time of concentration of the impoundment
itself:
tci = max ( tcs max , tci ), (37)
Finally, t ci is compared to the time of concentration for other hillslopes that

may contribute runoff to the channel. If t ci is larger than the t cs of these
hillslopes, then it is assumed to control the time of concentration. The channel
time of concentration is then computed with the equation:
tc = tcc + tci. (38)
The program tracks the flow routing network having the largest time of
concentration throughout the watershed for all watershed elements. For higher-
order channels, this is accomplished by finding the upstream watershed element
(hillslope, channel or impoundment) contributing runoff to the channel inlet that
has the largest time of concentration. This time of concentration is then
compared to the time of concentration of any watershed element (hillslope or
impoundment) that may contribute lateral runoff (inflow) to the channel. The
average channel travel time, t cc, is then calculated as for first order channels,
with the exception that values for channel physical properties (e.g., Manning’s n
and slope) may or may not be spatially averaged depending on flow routing
network characteristics. The generalized time of concentration for higher order
channels is given by Equation (35). If a hillslope controls the time of
concentration, tci will be zero. If an impoundment controls the time of
concentration, t cs will be zero. The average channel travel time, t cc, must be
calculated for each channel.
4.3.3.2. The CREAMS equation
The CREAMS peak runoff equation (Smith and Williams, 1980) is the second
method for calculating the peak runoff rate at the channel outlet in the WEPP
model watershed component. The equation was statistically derived using data
from watersheds with areas ranging from 70 to 6200 ha. The peak discharge at
the channel or watershed outlet is calculated with the equation:
q po = (0 . 00071 )A 0w .7 s 0 .159 ( 39 . 37 v ) 0 . 71764 ( A w0 . 0166 )

lw − 0 . 187
, (39)
where: q po = peak discharge at the channel outlet in m3 ⋅s-1 ; Aw = watershed area

contributing to the channel in m2 ; s = average surface slope in m⋅m-1 ; v =
average runoff depth at the channel outlet in m; lw = dimensionless watershed
length to width ratio; and 0.0007172 and 39.37 = unit conversion constants.
The CREAMS peak runoff equation may produce acceptable results on
watersheds approaching the 40 ha maximum area recommended for watershed
component applications, however, the Rational equation should be used until
modifications are made to the hillslope component rill erosion equations that
will allow longer overland flow lengths, and consequently larger watershed
areas to be modeled.
4.3.4. Effective Runoff Duration

After the peak discharge at the channel or watershed outlet is calculated,
effective runoff duration is calculated as:
rov f
durro = , (40)
q po
where durro is the effective runoff duration in seconds. The effective runoff
duration is used in the erosion calculations (discussed in the next section) to
determine inflow and sediment loading rates.
4.4. Channel Erosion Processes
The watershed component channel erosion element has been adapted and
modified from the CREAMS model channel erosion element (Knisel, 1980). It is
similar to the WEPP model hillslope component erosion element with major
differences being: (1) the flow shear stress is calculated using regression
equations developed by Foster et al. (1980) which approximate the spatially
varied flow equations (Chow, 1959); and (2) only entrainment, transport, and
deposition by concentrated flow are simulated. The channel element is used to
represent flow in terrace channels, diversions, major flow concentrations where
topography has caused overland flow to converge, grass waterways, row
middles or graded rows, tail water ditches, and other similar channels. The
channel element does not describe classical gully or large stream channel
erosion.
Channel erosion is based on a steady state sediment continuity equation.

Sediment load in the channel is a function of the incoming upstream load (from
hillslopes, channels, and impoundments), the incoming lateral load (from
adjacent hillslopes and impoundments), and the ability of the flow to detach and
transport channel bed material or soil particles. The flow detachment rate is
proportional to the differences between the flow shear stress exerted on the bed
material and the critical shear stress, and depends on the transport capacity of
the flow and the sediment load. Net detachment occurs when the flow shear
stress exceeds the critical shear stress of the soil or channel bed material and the
sediment load is less than the transport capacity. Net deposition occurs when
sediment load is greater than transport capacity.
A non-erodible layer having an initial depth and width is assumed to exist at
some depth below the bottom of the channel. Within ephemeral gullies,
detachment is assumed to occur initially from the channel bottom until the non-
erodible layer (usually the primary tillage depth) is reached. Once the channel
encounters the non-erodible layer it starts to widen and the erosion rate
decreases with time until the flow is too shallow to cause detachment. The
ephemeral gully cross-sectional geometry is updated after each precipitation
event that causes detachment in order to calculate channel hydraulics for
subsequent events.
The channel element allows for modeling of deposition in a backwater area
at a field outlet by taking into account conditions where the friction slope does
not equal the bed slope. Such deposition is not uncommon, and is important in
estimating sediment yields associated with the enrichment of fine sediment
during deposition. Field outlet controls are accounted for through solution of a
series of spatially varied flow equations (Foster et al., 1980), and thus can be
used to simulate backwater effects on sediment deposition.
4.4.1. Effective Channel Length
The general case for concentrated flow in a field situation is a channel of

length lch with an upstream inflow rate qt and a lateral inflow rate ql along the
channel reach. The upstream inflow rate, qt, is equal to the peak runoff rate
(discharge), qpo , of the upstream contributing watershed element(s). The
upstream and lateral inflow rates correspond to the peak discharge at steady
state, and are treated as steady state spatially varied flow with increasing
discharge along the length of the channel. The effective channel length, leff, is
the length of channel required to produce the channel outlet discharge, qpo ,
given the lateral inflow rate. That is, leff is the length of the channel if it is
extended upslope to where discharge would be zero with the given lateral inflow
rate. If there is lateral inflow to the channel, leff is computed as:
 q 
l eff = l ch  1. 0+ t  (41)
 ql 
The difference between the actual and effective channel lengths, ltop, is then
proportionally added to each channel computational segment length. If there is
no lateral inflow to the channel, leff and ltop are set equal to zero. Next, the
discharge rate at the channel inlet is calculated. If there is lateral inflow, the
upper discharge rate is computed as:
l top
qu = q po , (42)
l eff
where qu is the discharge at the channel inlet in m3 ⋅s-1 .

The effective lateral inflow rate, qlat in m2 ⋅s-1 , used for the lateral sediment
inflow calculations, is then calculated as:
q po
qlat = , (43)
l eff
If the initial lateral inflow rate, ql , is zero, then qu is set equal to qpo and qlat
is set equal to zero. After the initial calculations for qu and qlat are performed,
the discharge rate at the lower end of each computational segment, qls in m3 ⋅s-1 ,
can be calculated as:
x
qls = q po , (44)
l eff
where x is the segment downslope distance from the top of the channel in m.
The erosion computations proceed down the length of the channel through
the computational segments. The procedure used in the channel element is to:
(1) set qu for the downslope segment equal to the upslope segment qls; (2) solve
the spatially varied flow equations for a channel of length leff to produce flow
depth, velocity, and shear stress along each channel computational segment; and
(3) apply the transport and detachment capacity equations segment-by-segment
along the original length of channel, lch, to compute sediment yield for the
channel.
4.4.2. Sediment Load
Sediment load is assumed to be limited by either the amount of sediment

made available by detachment or by transport capacity. A quasi-steady state is
assumed and sediment movement downslope obeys continuity of mass as
expressed by the equation (Foster et al., 1980):
dq sed
= DL + D F , (45)
dx
where qsed is the channel sediment load in kg⋅m-1 ⋅s-1 , DL is the lateral sediment
inflow in kg⋅m-2 ⋅s-1 ; and DF is detachment or deposition by flow in kg⋅m-2 ⋅s-1 .
The assumption of quasi-steady state allows deletion of time terms from
Equation (45). All sediment load (detachment, transport, and deposition)
calculations are done for each particle size class. Similar to the hillslope erosion
element, the default number of particle size classes for the channel erosion
element is five. Each class is represented by a particle diameter and particle
density. Because the channel erosion equations use a single lateral sediment
inflow rate, the sediment discharges from the lateral contributing watershed
elements (adjacent hillslopes and impoundments) are combined into a single
value. A weighted average, based upon the relative runoff volume from the left
and right channel banks, is used to compute the average sediment flux in
entering the channel laterally on a length basis. If there is no lateral inflow, the
lateral sediment flux is set equal to zero.
For each computational segment, the channel element computes an initial
potential sediment load which is the sum of the sediment load from the
immediate upslope segment plus that added by lateral inflow within the
segment. If this potential load is less than the flow transport capacity,
detachment occurs at the lesser of the detachment capacity rate or the rate that
will just fill transport capacity. When detachment by flow occurs, soil particles
are added to the flow having the same particle size distribution for detached
sediment given as input. These concepts are explained in greater detail in the
following section.
4.4.3. Sediment Detachment/Transport/Deposition
If the sediment load of all particle classes at the upper boundary is less than
the transport capacity of the respective classes, then the potential rate at which
concentrated flow detaches soil particles from the soil matrix and potential
sediment load at the lower boundary of the channel segment are computed. The
detachment capacity, Dc in kg⋅m-2 ⋅s-1 , is described with the equation:
Dc = K ch( τ − τ cr ) , (46)
where Kch is an erodibility factor in s-1 , τ is the average shear stress in Pa,
and τ cr is the critical shear stress below which channel erosion is negligible in
Pa.
Until the channel reaches the non-erodible layer, an active channel of
rectangular shape is assumed to erode at the rate:
E ch = wc K ch ( τ − τ cr ) , (47)
where Ech is the soil loss per unit channel length in kg⋅m-1 ⋅s-1 , and wc is the
channel width in m.
It should be noted that Equations (46) and (47) are not the CREAMS
equations for detachment capacity and channel eros ion, but rather are similar to
the WEPP hillslope component rill erosion equations. Once the channel reaches
the non-erodible layer it starts to widen and the erosion rate decreases with time
until the flow is too shallow to cause detachment. Foster et al. (1980) describe
the equations used for channel widening after the non-erodible layer is reached.
The sediment transport capacity for each particle size class, based upon the
potential sediment load, is computed using the Yalin sediment transport
equation (Yalin, 1963). A complete description of the transport capacity
calculations is presented by Foster et al. (1980). If the sediment load of all
particle classes is greater than the transport capacity then deposition is assumed
to occur at the rate of:
Dr = α er ( Tc − q sed ) , (48)
where: Dr = deposition rate in kg⋅m-2 ⋅s-1 ; α er = a first order reaction coefficient

based on particle fall velocity and unit width discharge in m-1 ; Tc = transport
capacity in kg⋅m-1 ⋅s-1 ; and qsed = sediment load in kg⋅m-1 ⋅s-1 .
The potential sediment load and transport capacity at the lower boundary of
the segment is then computed. Net detachment or net deposition may occur,
meaning that within each channel segment four different detachment-deposition
limiting cases are possible:
• Case I: Net deposition at the upper boundary and net deposition at

the lower boundary (deposition may occur over the entire segment).
• Case II: Net deposition at the upper boundary and net detachment
by flow at the lower boundary may occur when transport capacity
increases within the segment.
• Case III: Net detachment by flow at the upper boundary and net
deposition at the lower boundary may occur when transport capacity
decreases in a segment.
• Case IV: Net detachment by flow at the upper boundary and net
detachment by flow at the lower boundary (detachment by flow may
occur over the entire segment).
4.5. Watershed Component Summary
The WEPP model watershed component was developed to predict erosion

effects from various management practices and to accommodate topographic,
soil type, and land use variability within small watersheds. Overland flow
hydrologic and erosion output (e.g., runoff volume, peak runoff rate, and
sediment concentration) is linked to channel and impoundment elements,
allowing water and sediment from one or more hillslopes to be routed through a
field-scale watershed system. The watershed component is capable of: (1)
identifying zones of sediment transport, deposition and detachment within
constructed channels (e.g., grassed waterways or terraces) or concentrated flow
(ephemeral) gullies; (2) simulating backwater flow conditions for channels with
heavy vegetation or for channels with a restricted outlet such as a weir or ridge;
(3) accounting for the ability of impoundments such as farm ponds, filter fences,
and check dams to trap incoming sediment, thereby reducing sediment yield at
the watershed outlet; and (4) representing spatial and temporal variability in
erosion and deposition processes as a result of landscape management practices.
An additional important feature of the watershed component is a process-based
description of hillslope and channel water balance, plant residue decomposition,
and crop growth, including daily updating of explicit relationships between
surface conditions (e.g., roughness, canopy and residue cover, etc.) and
infiltration/erosion parameters (e.g., Green-Ampt effective hydraulic
conductivity and interrill/rill erodibility). It is applicable at the small watershed
scale (up to 40 ha) where the sediment yield at the outlet is significantly
influenced by hillslope and channel processes.
5. MODEL VALIDATION STUDY RESULTS
Thorough evaluation and testing of the WEPP model is critical to acceptance

of the technology. An important aspect of the evaluation process involves
comparisons of model results to measured hydrologic and erosion data,
comparisons to previous accepted erosion models, sensitivity and error
propagation analyses, and detailed evaluation of all components of the
simulation such as water balance, plant growth, residue decay, infiltration, frost
layers, etc. WEPP is a conservation planning tool, therefore, land management
decisions, which always have associated monetary and social costs, will be
based in part on model output results. The material presented here focuses on
comparison of WEPP model results to measured erosion data from plots and
small watersheds. Recently, a series of studies have been conducted to compare
erosion model predictions of soil loss to measured data for the WEPP model.
For comparison, we also review briefly similar studies on hillslopes recently
conducted on the USLE and RUSLE.
Table 2 lists some of the natural runoff and erosion plot data used in a large
validation study of the WEPP hillslope model (from Zhang et al., 1996). Data
from these plots were taken on a storm-by-storm basis over a period of several
years under natural rainfall and typical crop management conditions.
The WEPP model (v95.1) was applied to the data in Table 2 with no
calibration of the model inputs, and results of predicted versus observed average
annual soil loss are shown in Figure 6. Measured on-site weather data were used
as climate inputs in these analyses. Model predictions of average annual soil loss
were very good, with a coefficient of determination (r2 ) of 0.85, and a Nash-
Sutcliffe efficiency parameter of 0.88. The N-S efficiency parameter is a measure
of variance from the one-to-one prediction line (Nash and Sutcliffe, 1970).
Risse et al. (1993) applied the USLE to 1700 plot years of data from 208
natural runoff plots. Average for the observed soil loss on an annual basis was
3.51 kg⋅m-2 . Using the USLE, annual values of predicted soil loss averaged 3.22
kg⋅m-2 with an average magnitude (absolute value) of error of 2.13 kg⋅m-2 , or
approximately 60 percent of the mean. Rapp (1994) applied the RUSLE model
to the same set of data as Risse et al. For RUSLE, annual values of predicted soil
loss averaged 3.16 kg⋅m-2 . The average magnitude (absolute value) of error was
not reported, but it is apparent that the two models performed similarly overall
in terms of soil loss prediction. Zhang et al. (1996) applied the WEPP hillslope
model to 290 annual values and obtained an average of 2.18 kg⋅m-2 for the
measured soil loss, with an average magnitude of error of 1.34 kg⋅m-2 , or about
Table 2. Natural runoff and erosion plot validation data used in developing Figure 6.
Location
Crop Management System Reps. Years Events
/Soil
Holly Springs, 1. fallow 2 1961-68 208
Mississippi, 2. conv. corn, spring TP 2 " 163
Providence sil 3. bermuda-corn-bermuda 2 1962-68 127
4. conv. soyb. 70-73, 78-80, conv. 2 1970-80 406
Corn silage 74-77
5. no-till soybean 70-73, conv. corn 2 " 406
74-77, reduced-till soybean 78-80
6. no-till corn & soyb. 70-73, no-till 2 " 405
corn 74-77, no-till soyb. 78-80
7. no-till corn and soybean rotation 2 1970-76 267
70-73, no-till corn silage 74-77
Madison, 1. fallow 3 1962-70 59
South Dakota, 2. conv. corn, spring TP 3 " 48
Egan sicl 3. cons. corn, no TP 3 " 50
4. continuous oats, no TP 3 1962-64 15
Morris, 1. fallow 3 1962-71 67
Minnesota, 2. conv. corn, fall TP 3 " 67
Barnes l 3. bromegrass-corn-oats 3 " 41
Presque Isle, 1. fallow 3 1961-65 65
Maine, 2. continuous potato 3 " 64
Caribou gr sil 3. potato-oats-meadow 3 " 46
Watkinsville, 1. fallow 2 1961-67 147
Georgia, 2. conv. corn, spring TP 2 " 97
Cecil scl 3. conv. cotton, spring TP 2 " 112
4. corn-bermuda-bermuda 2 " 83
Bethany, MO, 1. alfalfa 1 1931-40 83
Shelby sil 2. brome grass 1 " 79
Geneva, 1. fallow 1 1937-46 97
New York, 2. summer fallow, winter rye 1 " 77
Ontario l 3. conv. soybean, spring TP 1 " 45
4. red clover 1 1937-41 19
5. bromegrass 1 1937-46 30
Guthrie, 1. fallow 1 1942-56 170
Oklahoma, 2. conv. cotton, sp ring TP 1 " 140
Stephensville 3. bermuda grass 1 " 96
fsl 4. wheat -clover-cotton 1 " 124
conv., conventional till; TP, turn-plow; cons., conservation till
61 percent of the mean. In all cases the relative errors tended to be greater for
the lower soil loss values. All three studies were conducted without model
calibration, and model input parameter values were not adjusted for the specific
data used in the comparisons.
100
WEPP Predicted Soil Loss (kg/m2)
80 1:1
60
40
20
0
0 20 40 60 80 100
Measured Soil Loss (kg/m2)
Figure 6. WEPP model predicted average annual soil loss versus measured soil loss from natural
runoff plot study.
What is reported above for plot data is obviously a very "broad brush"
picture of the performance of the three erosion models, but in essence, the
results indicate that for the prediction of soil loss, the three models appear to
perform approximately on the same level of accuracy. However, there are a
couple of important points to be considered. In the first place, all three models
do predict soil loss, but only the WEPP model is specifically designed to predict
sediment yield. Thus, if prediction of average soil loss on the hillslope is the
goal, one might conclude from the studies that any of the three models work
equally well. However, if one needs to know the deposition rates in the toe-slope
of the hill, how much sediment might be transported off-site, sediment load from
a channel area, or the distribution of erosion along the hillslope, then only
WEPP will provide that information. Also, with regard to RUSLE vs. the USLE,
one should note that, although uncalibrated, the data used in this study was the
same or quite similar to the data used to develop the USLE. Risse et al. (1993)
and Rapp (1994) actually discuss this point in their papers and delineate the
consequence of the issue. It turns out that the USLE and RUSLE perform equally
well on the portion of the data used to develop the models as on those data not
used. RUSLE, however, was largely a response to a need to better predict soil
loss in regions or situations not well represented in the data used in the Risse et
al. (1993) and Rapp (1994) studies, such as semi-arid rangelands, no-till crops,
and for the erosivity factor, the entire western United States. In those situations,
certainly, one might expect that RUSLE will perform better than the USLE.
Table 3. Total runoff and sediment yields for the WEPP small watershed studies.
Sediment Yield # of # of
Watershed Runoff (mm) (t/ha) Years Select
Meas.* Pred.* Meas.* Pred.* Record Events
Chickasha C5, OK 320 309 4.27 3.81 4 34

Coshocton 109, OH 25 26 1.99 1.02 11 4
Coshocton 130, OH 49 30 0.036 1.11 7 6
Coshocton 191, OH 20 20 0.055 0.035 11 3
Holly Springs 1, MS 3409 2820 64.7 153.7 8 237
Holly Springs 2, MS 3576 2658 65.9 121.8 8 241
Holly Springs 3, M S 2858 2600 94.0 141.6 8 241
Riesel SW-12, TX 1086 940 ------- 3.88 6 57*
Riesel W-12, TX 833 860 15.77 9.61 6 117
Riesel W-13, TX 879 920 10.38 8.05 6 83
Tifton Z, GA 403 332 6.67 8.31 8 46
WatkinsvilleP-1, GA 596 567 53.9 67.6 11 33
+
Sediment data not available for the SW-12 watershed; *Meas. = measured, *Pred. = predicted
Table 3 shows the comparison of measured and WEPP predicted total runoff
and sediment yields for several small watersheds (Liu et al., 1997). On-site
observed weather data were used as climate inputs to WEPP in these analyses as
well. WEPP over-predicted sediment yield by a factor of approximately two for
the Holly Springs watersheds. This may be due to the fact that at Holly Springs
in several years the corn was cut for silage. When the silage option in WEPP
was used the model assumed a 95 percent biomass removal, whereas the actual
removal rates were much lower at Holly Springs (Keith McGregor, personal
communication). Also, weed growth after harvest at Holly Springs was
substantial, and WEPP does not have a specific weed growth option. Another
possibility for the prediction bias for the Holly Springs application could be a
problem with erodibility parameterization. However, the results from the study
of Zhang et al. (1996), which included the application of WEPP to plot data
from Holly Springs, do not bear this out. In that study, the fallow plot had a
measured erosion rate of 170 kg⋅ha-1 year, and WEPP predicted 161 kg⋅ha-1 year.
The row-cropped plots from Holly Springs, however, were overpredicted by a
factor of 2 to 4 (Zhang, personal communication). These results would indicate a
problem with the WEPP application at Holly Springs associated with cropping
routines rather than soil parameters. The evaluations at Holly Springs pointed to
a need for model improvement for silage options, as well as simulation of weed
growth in further evaluations there.
The Coshocton watersheds produced small amounts of sediment (Table 3). In
general, experience has shown that other erosion models including the USLE
(Risse et al., 1993), RUSLE (Rapp, 1994), and WEPP hillslope (Zhang et al.,
1996; Nearing and Nicks, 1998) tend to produce large errors on a percentage
basis for low erosion rates. The reason for this is that there tend to be fewer
events measured at such sites and that there is more natural variability in terms
of relative amounts for small events. For example, the Coshocton watersheds
had only 3 to 6 measured runoff events for the three watersheds. The absolute
values of the errors for these two watersheds are within the range of absolute
errors for the remainder of the watersheds, and the total erosion for the three
Coshocton watersheds are the least of all the watersheds for both the measured
and predicted case. The conclusion was reached that WEPP performed as well
as could be expected for the Coshocton site, given the nature and quantity of the
data used (Liu et al., 1997).
In the case of watersheds, of course, the USLE and RUSLE are not applicable
because they do not address the issue of deposition and sediment yields, but are
only useful for obtaining soil loss over the area of the hillslope that experiences
a loss.
Several other evaluation studies of the WEPP model have been reported
recently. Amore et al. (1999) successfully applied WEPP to a very large basin in
Sicily. Renschler et al. (2000) applied WEPP using a Geographic Information
System (GIS) interface to two moderate -sized watersheds in Iowa. Nearing et al.
(1998) successfully applied WEPP to rainfall simulator data from the Ukraine,
and Savabi et al (1996a) applied the model to watersheds in Austria. In addition
to traditional validation studies, Savabi et al. (1996b), Favis-Mortlock and
Savabi (1998), and Doerring et al. (1999) have used WEPP to estimate the
potential changes in erosion rates under global climate change.
6. DATA AND MODEL UNCERTAINTY: IMPACTS ON

MODEL EVALUATION AND AP PLICATION
Uncertainty plays an important role in both evaluating soil erosion models and
in using them for prediction purposes. Much emphasis has been placed, and for
good reason, on uncertainty in model predictions as a function of input parameter
variance. The propagation of input uncertainty through a model to its output is an
important problem because it is both a major source of prediction error and because
it is the source of error over which we have the greatest control. It is also the easiest
to study. A more difficult problem is that of model structural error. To date there
have been no major, systematic studies of model structural errors. Data uncertainty
has been studied to some degree, but the primary problem in this regard is lack of
sufficient measurement replications to adequately characterize data variance. Also,
the implications of data uncertainty on model application have not been extensively
addressed (Nearing et al., 1999).
The first, and most basic, test of a model is sensitivity analysis (Nearing et al.,
1990; Tiscareno-Lopez et al., 1993, 1994; and Baffaut et al., 1997). Sensitivity
analysis is conducted for two basic reasons: a) to determine if the model is
responding to changes in input parameters in basic, sensible ways, and b) to
determine the most important (sensitive) of the model inputs relative to model
response. A sensitivity analysis should always be conducted prior to evaluation of
model uncertainty. This is important because the model sensitivity to specific
variables varies depending on the conditions modeled. The model can be sensitive
to a particular variable in one case and relatively insensitive in another. For
example, one often finds that soil loss results are usually quite sensitive to the
amount of precipitation, rainfall intensity, rill erodibility, interrill erodibility,
critical hydraulic shear, incorporated residue, and ground cover (Nearing et al.,
1990). However, for a case where the dominant erosion process is interrill, rill
erodibility might have little impact on the erosion rate.
Another important point that needs to be made with regard to model sensitivity
is the occasional lack of sensitivity of the model to certain inputs that are
“expected” to be related. For example, one expects that soil erodibility will
decrease, and hence erosion will decrease, as the soil bulk density increases when a
soil consolidates after tillage. WEPP erosion rates are not specifically sensitive to
bulk density in this way, however, the process of consolidation and its effect on
erodibility is captured by the model not via a specific tie to the bulk density value,
but rather to the time of consolidation after tillage. In this respect, it must be said
that lack of sensitivity of output to a specific input variable does not necessarily
imply the lack of a process. A careful reading of the WEPP technical
documentation (Flanagan and Nearing, 1995) may show that the process has been
captured in a different manner.
Uncertainty in model predictions as a function of input variance is typically
studied using either a Monte-Carlo or First-Order Error Analysis. Quinton (1994)
studied the EUROSEM model using a Monte-Carlo technique and found that the
output uncertainty bands could be quite high. Deer-Ascough (1995) performed a
similar type of analysis on WEPP. Quinton (1997) subsequently used two different
methods in an attempt to reduce the level of prediction uncertainty and met with
limited success.
So how does one evaluate erosion models in the context of such enormous
problems in both model and data uncertainty, and what are the implications of
uncertainty for model use?
Quinton (1994) suggested a methodology for erosion model validation that
uses a two-part process: corroboration and evaluation. Corroboration involves
one-way and two-way sensitivity analysis, and the comparison of the model
response to critical experimental data in order to examine the fundamental
hypotheses imbedded in the model structure. Evaluation, in Quinton’s scheme,
involves the definition of the application and the selection of a success/failure
criterion such as a certain percentage of error allowed, based on the intended
application of the model. Quinton (1994) proposes that the model be applied to
the type of data to be used in the application, and that confidence limits be
established using ranges of the input values. Finally the model results are
compared to the observed erosion data to determine the coincidence of the
model output bands with the observed values. Quinton (1994) used this
approach in his evaluation of the EUROSEM model.
A quantitative method for evaluating model effectiveness has been proposed
by Nearing (2000). The purpose of that study was to analyze a large number of
replicated plot data and describe methodology that allows the model evaluator to
take natural, within-treatment variability of erosion plots into account when
models are tested. A large number of data from pairs of replicated erosion plots
were evaluated and quantified. The thesis proposed for defining an evaluation
criteria for an erosion simulation model was that if the difference between the
model prediction and the measured value lies within the population of
differences between the measured data pairs, then the model reasonably reflects
the erosion for that population. Another way of looking at this concept is that the
replication of an individual plot may be considered as a “real-world” physical
model of that plot. The question, then, of whether or not a simulation model
prediction is “good” is made relative to how well that simulation model
performs as compared to the physical model as represented by the replicated
plot. The basis for the evaluation method presented was that if the difference
between the model prediction and a measured plot data value laid within the
population of differences between pairs of measured values, then the prediction
was considered “acceptable.” A model “effectiveness” coefficient was defined
for studies undertaken on large numbers of prediction versus measured data
comparisons. This method provides a quantitative criterion for taking into
account natural variability and uncertainty in measured erosion plot data when
that data is used to evaluate erosion models.
There is another, simpler concept that can be used to evaluate the effectiveness
of a model in the presence of model uncertainty that also follows along the logic of
using the replicated plot as a basis for comparison. Pairs of measured erosion
values for 3007 replicated plot pairs from several different erosion experiments are
shown in Figure 7 (Nearing, 1998). The Nash-Sutcliffe efficiency parameter (Nash
and Sutcliffe, 1970) for this dataset was 0.7. If one reasonably does not expect that
their erosion model will out-perform the physical model of the replicate plot, one
would not expect a better fit than 0.7 when making model evaluations.
What are the implications for uncertainty in predictions of erosion? Govers
(1996) suggested that the problems associated with process-based models were so
overwhelming that a return to the empirical model for prediction purposes is the
practical alternative. However, as seen above, results of the process-based WEPP
model compare well to the USLE and RUSLE results, and we already have
discussed the fact that process-based models provide us with a different and
broader range of results than do empirical soil loss models.
It is true that on a case-by-case basis erosion predictions can be very much in
error. Yet erosion models have proven to be very effective tools for conservation
planning and erosion assessment. In the case of conservation planning, often the
goal is not so much to accurately predict erosion on individual fields as it is to
either choose a best management practice from various alternatives or to design an

effective management strategy. For these purposes, process-based erosion models
can be highly effective. For making regional erosion surveys, we are not as
interested in the erosion rates on individual fields as we are on integrated erosion
rates across areas. Again, erosion models have been shown to be effective in
obtaining reasonable estimates on average, if not on the specific field, and should
be effective as regional erosion assessment tools. As for regulatory purposes, it has
been suggested that knowledge of how poorly a model might perform on an
individual field may open up opportunities for individual land managers and
owners to question the use of erosion models in regulating activities on their land.
However, a properly implemented and applied model can rapidly provide an
impartial estimate of soil erosion, based upon current scientific understanding.
Even if the prediction results are imperfect (which results from models always will
be), use of them is much better than the alternative - land management with no
information on potential erosion danger. Prediction uncertainty should not prevent
the use of models for regulatory purposes.
Nash-Sutcliffe E = 0.7 1:1

"Predicted" Soil Loss (kg/m 2 )
0
0 2 4 6 8
"Measured" Soil Loss (kg/m 2 )
Figure 7. Pairs of measured erosion values for 3007 replicated plot pairs from several different
erosion experiments from Nearing (1998).
7. WEPP MODEL STATUS AND CURRENT

ACTIVITIES
Since 1995, the WEPP project has been in a maintenance and

implementation phase, with major work in development of improved user
interfaces and databases, and minimal work on development of expanded
scientific components. A major impediment to acceptance of the model by field
NRCS users was lack of sufficiently easy-to-use interfaces. To address this
need, work began in 1996 on development of a computer interface for field
conservation users, which would be developed initially for Windows 95/98/NT
platforms. Also in 1997, a joint effort with the other ARS erosion projects
began to develop a common interface program known as MOSES (MOdular Soil
Erosion System) began that would allow use of the WEPP, RUSLE, WEPS
(Wind Erosion Prediction System), and RWEQ (Revised Wind Erosion Equation)
models through common screens and databases.
Prototype beta versions of the WEPP Windows interface were released in
1998 (Flanagan et al., 1998) and 1999, with a complete hillslope and watershed
program becoming available in 2000. Also, work on development of an Internet
Web-based WEPP interface began in 2000 to provide a simple tool that will
allow users interested in the model to determine if it can meet their needs, and if
they may wish to obtain the fully functional stand-alone program. An initial
prototype of the MOSES interface was delivered in October 2000.
Three areas of current work on WEPP model science testing and
enhancement are winter hydrology, furrow irrigation, and impoundment
simulations. These efforts are being conducted through cooperative efforts with
ARS scientists in Pullman, Washington, and Kimberly, Idaho, and FS scientists
in Moscow, Idaho, respectively. An initial stage of the impoundment component
enhancement has been completed, and suggested changes have been evaluated
and incorporated into the official WEPP model code. Improvements to the
WEPP winter component should be completed in 2001. Additional testing of the
WEPP furrow irrigation routines across some typical locations and soils in the
U.S. are planned to determine if changes may be required in the model sediment
transport relationships for these low slope and low flow conditions.
Other efforts at the NSERL have focused on linkage of the WEPP watershed
model with (GIS). Recent work has developed procedures and interfaces to
automatically delineate watershed boundaries, channels, hillslope regions, and
representative slope profiles from Digital Elevation Model (DEM) data. This
work has shown that automatic techniques can be successfully used to rapidly
set up accurate WEPP model simulation topographic inputs, potentially reducing

the work required of users conducting watershed simulations (Cochrane and
Flanagan, 1999). Figure 8 shows how application of the model to flow paths
generated from a DEM in a watershed can be used to display areas of high
erosion rates. Efforts are underway to transfer the research findings and software
to the Windows interface programs. Also, a recent field study collected
differential global positioning system (DGPS) data for a watershed using several
typical commercial units (Renschler et al., 2001), and the usefulness of this type
of topographic information for WEPP model simulations is under evaluation.
Figure 8. Application of WEPP model to Treynor, Iowa watershed #2, using flow paths generated
from a digital elevation model (DEM) as slope profile inputs, then running WEPP simulations on
all flow paths.
The WEPP team is also working with the WEPS modeling group in a joint
effort to ultimately develop a single process-based model that can be used for
either water or wind erosion simulations (Fox et al., 2001; Retta et al., 2001).
This common model would assure that all computations related to hydrology,
water balance, crop growth, residue decomposition, effects of tillage, etc. would
be identical for the wind or water simulations when applied with the same slope,
soil, management, and climate. Consistency between predictions is very
important for action agencies, such as the NRCS, that want to apply both models
in conservation planning activities. Plans are to develop a common wind and
water model within the next few years.
An updated version of the WEPP scientific model and interface programs is
provided on approximately an annual basis, and all WEPP materials are
provided free of charge and distributed from the NSERL Internet site. Inquiries
concerning the WEPP project and model may be directed to the email address
wepp@ecn.purdue.edu. The WEPP model software, documentation, and
information are available from the USDA-ARS National Soil Erosion Research
Laboratory web site: http://topsoil.nserl.purdue.edu.
REFERENCES
Alberts, GA, Nearing, MA, Weltz, MA, Risse, LM, Pierson, FB, Zhang, XC, Laflen, JM, and
Simanton, JR, 1995, Soil component: in USDA-Water Erosion Prediction Project: Hillslope
Profile and Watershed Model Documentation (DC Flanagan and MA Nearing, eds.), NESRL
Rept. No. 10, USDA-ARS National Soil Erosion Research Laboratory, West Lafayette,
Indiana.
Amore, E, Santoro, C, Modica, C, and Nearing, MA, 2000, Application of two soil erosion models
to a large Sicilian basin: Proc. 3rd Int. Cong. Eur. Soc. Soil Conser. March 28-April 1, 2000.
Arnold, JG, and Williams, JR, 1989, Stochastic generation of internal storm structure: Trans. Am.
Soc. Agric. Eng., 32: 161-166.
Arnold, JG, Williams, JR, Nicks, AD, and Sammons, NB, 1990, SWRRB: A basin scale simulation
model for soil and water resource management, Texas A&M Univ. Press, College Station,
Texas. 142 pp.
Arnold, JG, Weltz MA, Alberts EE, and Flanagan, DC, 1995, Plant growth component: in USDA-
Water Erosion Prediction Project: Hillslope Profile and Watershed Model Documentation
(DC Flanagan and MA Nearing, eds.), NSERL Report No. 10, USDA-ARS National Soil
Erosion Research Laboratory, West Lafayette, Indiana.
Ascough II, JC, Baffaut, C, Nearing, MA, and Liu, BY, 1997, The WEPP watershed model: I.
hydrology and erosion: Trans. Am. Soc. Agric. Eng., 40: 921-933.
Baffaut, C, Nearing, MA, Ascough II, JC, and Liu, BY, 1997, The WEPP watershed model: II.
sensitivity analysis and discretization on small watersheds: Trans. Am. Soc. Agric. Eng., 40:
935-943.
Burroughs, ER, Luce, CH, and Phillips, F, 1992, Estimating interrill erodibility of forest soils:
Trans. Am. Soc. Agric. Eng., 35: 1489-1495.
Chow, VT, 1959, Open-Channel Hydraulics, McGraw-Hill, New York.
Chu, ST, 1978, Infiltration during an unsteady rain: Water Resour. Res., 14: 461-466.
Cochrane, TA, and Flanagan, DC, 1999, Assessing water erosion in small watersheds using WEPP
with GIS and digital elevation models: J. Soil Water Conserv., 54: 678-685.
Deer-Ascough, LA, 1995, A Framework for Uncertainty Analysis of Complex Process-Based
Models, Ph.D. Thesis, Purdue Univ.
Doerring, O, Habeck, M, Lowenberg-Doboer, J, Pfeifer, R, Randolph, JC, Southworth, J,
Mazzocco, M, and Nearing, M, 1999, Global climate change: Implications of extreme events
for conservation strategies, Abstracts, 10th Meeting of the Int. Soil Conserv. Org., May 23-28,
1999, Purdue Univ., West Lafayette, Indiana.
Elliot, WJ, Liebenow, AM, Laflen, JM, and Kohl, KD, 1989, A Compendium of Soil Erodibility
Data from WEPP Cropland Soil Field Erodibility Experiments 1987 & 1988: NSERL Report
No. 3, USDA-ARS National Soil Erosion Research Laboratory, West Lafayette, Indiana.
Elliot, WJ, Qiong, W, and Elliot, AV, 1993, Application of the WEPP model to surface mine
reclamation, Am. Soc. Surface Mine Reclam., Paper presented at Challenge of Integrating
Diverse Perspectives in Reclamation, 10th National Meeting, Spokane, Washington.
Elliot, WJ, Luce, C, and Robichaud, P, 1996, Predicting sedimentation from timber harvest areas
with the WEPP model: Paper presented at 6th Federal Interagency Sedimentation Conf., Las
Vegas, Nevada.
Ellison, WD, 1947a, Soil Erosion, Soil Sci. Soc. Am. Proc., 12: 479-484.
Ellison, WD, 1947b, Soil erosion studies-parts 1-6, Agric. Eng., 28: 145-146; 28: 197-201; 28:
245-248; 28: 297-300; 28: 349-351; 28: 402-405, 408.
Favis-Mortlock, DT, and Savabi, MR, 1998, Shifts in rates and spatial distribution of soil erosion
and deposition under climate change: in Advances in Hillslope Processes (MG Anderson and
S Brooks, eds.), Wiley, London.
Finkner, SC, Nearing, MA, Foster, GR, and Gilley, JE, 1989, A simplified equation for modeling
sediment transport capacity: Trans. Am. Soc. Agric. Eng., 32: 1545-1550.
Flanagan, DC, and Livingston, SJ (eds.), 1995, USDA-Water Erosion Prediction Project: WEPP
User Summary, NSERL Report No. 11, USDA-ARS National Soil Erosion Research
Laboratory, West Lafayette, Indiana.
Flanagan, DC, and Nearing, MA (eds.), 1995, USDA-Water Erosion Prediction Project: Hillslope
Profile and Watershed Model Documentation, NSERL Report No. 10, USDA-ARS National
Soil Erosion Research Laboratory, West Lafayette, Indiana.
Flanagan, DC, Fu, H, Frankenberger, JR, Livingston, SJ, and Meyer, CR, 1998, A Windows
interface for the WEPP erosion model, ASAE Paper No. 98-2135, Am. Soc. Agric. Eng., St.
Joseph, Michigan, 14 pp.
Flanagan, DC, and Nearing, MA, 2000, Sediment particle sorting on hillslope profiles in the
WEPP model: Trans. Am. Soc. Agric. Eng., 43: 573-583.
Foster, GR, 1982, Modeling the erosion process: in Hydrologic Modeling of Small Watersheds
(CT Haan, ed.), ASAE Monograph No. 5, Am. Soc. Agric. Eng., St. Joseph, Michigan: 295-
380.
Foster, GR, and Meyer, LD, 1972, A closed-form soil erosion equation for upland areas: in
Sedimentation: Symposium to Honor Professor H.A. Einstein, (HW Shen, ed.), Ft. Collins,
Colorado.
Foster, GR, Lane, LJ, Nowlin, JD, Laflen, JM, and Young, RA, 1980, A model to estimate
sediment yield from field-sized areas: development of model: in CREAMS: A Field-Scale
Model for Chemicals, Runoff, and Erosion From Agricultural Management Systems, Vol. II:
User Manual (WG Knisel, ed.), Conserv. Res. Report No. 26. USDA-Sci. Educ. Admin.: 193-
281.
Foster, GR, Laflen, JM, and Alonso, CW, 1985a, A replacement for the Universal Soil Loss
Equation (USLE): Proc. Natural Resources Modeling Symp. (DG DeCoursey, ed.), Pingree
Park, CO, October 16-20, 1983, USDA- ARS, 30: 468-472.
Foster, GR, Young, RA, and Neibling, WH, 1985b, Sediment composition for nonpoint source
pollution analyses: Trans. Am. Soc. Agric. Eng., 28: 133-139.
Foster, GR, and Lane, LJ, (eds.), 1987, User Requirements: USDA-Water Erosion Prediction
Project (WEPP): USDA-ARS, NSERL Rept. No. 1.
Foster, GR, Flanagan, DC, Nearing, MA, Lane, LJ, Risse, LM, and Finkner, SC, 1995, Hillslope
erosion component: in USDA-Water Erosion Prediction Project: Hillslope Profile and
Watershed Model Documentation (DC Flanagan and MA Nearing, eds.), USDA-ARS,
NSERL Report No. 10.
Fox, FA, Flanagan, DC, Wagner, LE, and Deer-Ascough, L, 2001, WEPS and WEPP science
commonality project: Proc. ASAE Symp. on Soil Erosion Res. for the 21st Century (JC
Ascough II and DC Flanagan, eds.), Honolulu, Hawaii, January 3-5, 2001: 376-379.
Ghidey, F, Gregory, JM, McCarty, TR, and Alberts, EE, 1985, Residue decay evaluation and
prediction: Trans. Am. Soc. Agric. Eng., 28: 102-105.
Gilley, JE, Kottwitz, ER, and Simanton, JR, 1990, Hydraulic characteristics of rills: Trans. Am.
Soc. Agric. Eng., 33: 1900-1906.
Gilley, JE, and Finkner, SC, 1991, Hydraulic roughness coefficients as affected by random
roughness: Trans. Am. Soc. Agric. Eng., 34: 897-903.
Gilley, JE, Kottwitz, ER, and Wieman, GA, 1991, Roughness coefficients for selected residue
materials: J. Irrig. Drain. Eng., Am. Soc. Civil Eng., 117: 503-514.
Gilley, JE, Kottwitz, ER, and Wieman, GA, 1992, Darcy-Weisbach roughness coefficients for
gravel and cobble surfaces: J. Irrig. Drain. Eng., Am. Soc. Civil Eng., 118: 104-112.
Gilley, JE, and Kottwitz, ER, 1994, Darcy-Weisbach roughness coefficients for selected crops:
Trans. Am. Soc. Agric. Eng., 37: 467-471.
Gilley, JE, and Weltz, MA, 1995, Hydraulics of overland flow: in USDA-Water Erosion
Prediction Project: Hillslope Profile and Watershed Model Documentation (DC Flanagan and
MA Nearing, eds.), USDA-ARS, NSERL Report No. 10.
Govers, G, 1996, Soil erosion process research: a state of the art, Academie voor Wetenschappen,
Letteren en Schone Kunsten van Belgie, Klasse der Wettenschappen, 58: 1.
Huggins, LF, and Burney, JR, 1982, Chapter 5. Surface runoff, storage, and routing: in Hydrologic
Modeling of Small Watersheds (CT Haan, ed.), ASAE Monograph No. 5, Am. Soc. Agric,
Eng., St. Joseph, Michigan: 167-225.
Jensen, ME, (ed.), 1974, Consumptive Use of Water and Irrigation Requirements, Report Tech.
Comm. Irrig. Water Requirements, Irrig. Drain. Div., ASCE, 227 pp.
Katz, DM, Watts, FJ, and Burroughs, ER, 1995, Effects of surface roughness and rainfall impact
on overland flow: J. Hydr. Div., Am. Soc. Civil Eng., 121: 546-553.
Knisel, WG, (ed.), 1980, CREAMS: A field-scale model for Chemicals, Runoff, and Erosion from
Agricultural Management Systems, USDA-Science and Education Admin. Cons. Res. Rept.
No. 26, 640 pp.
Kottwitz, ER, 1995, Irrigation component: in USDA-Water Erosion Prediction Project: Hillslope
Profile and Watershed Model Documentation (DC Flanagan and MA Nearing, eds.), USDA-
ARS, NSERL Report No. 10.
Laflen, JM, 1985, Tillage and residue effect on erosion from cropland: Proc. Natural Resources
Modeling Symp (DG DeCoursey, ed.), Pingree Park, CO, October 16-20, 1983, USDA-ARS,
30: 438-441.
Laflen, JM, Elliot, WJ, Simanton, R, Holzhey, S, and Kohl, KD, 1991, WEPP soil erodibility
experiments for rangeland and cropland soils: J. Soil Water Conserv., 46: 39-44.
Lindley, MR, Barfield, BJ, Ascough II, JC, Wilson, BN, and Stevens, EW, 1998a, Hydraulic
simulation techniques incorporated in the surface impoundment of WEPP: App. Eng. Agric.,
14: 249-256.
Lindley, MR, Barfield, BJ, Ascough II, JC, Wilson, BN, and Stevens, EW, 1998b, The surface
impoundment element for WEPP: Trans. Am. Soc. Agric. Eng., 41(3): 555-564.
Liu, BY, Nearing, MA, Baffaut, C, and Ascough II, JC, 1997, The WEPP watershed model: III.
Comparisons to measured data from small watersheds: Trans. Am. Soc. Agric. Eng., 40: 945-
951.
Mein, RG, and Larson, CL, 1973, Modeling infiltration during a steady rain: Water Resour. Res.,
9: 384-394.
Meyer, LD, and Wischmeier, WH, 1969, Mathematical simulation of the process of soil erosion
by water: Trans. Am. Soc. Agric. Eng., 12: 754-758, 762.
Morrison, JE Jr., Richardson, CW, Laflen, JM, and Elliot, WJ, 1994, Rill erosion of a vertisol with
extended time since tillage: Trans. Am. Soc. Agric. Eng., 37: 1187-1196.
Nash, JE, and Sutcliffe, JV, 1970, River flow forecasting through conceptual models 1. A
discussion of principles: J. Hydrol., 10: 282-290.
Nearing, MA, Deer-Ascough, LA, and Laflen, JM, 1990, Sensitivity analysis of the WEPP
hillslope profile erosion model: Trans. Am. Soc. Agric. Eng., 33: 839-849.
Nearing, MA, and Nicks, AD, 1998, Evaluation of the Water Erosion Prediction Project (WEPP)
model for hillslopes: in Modelling Soil Erosion by Water (J Boardman and DT Favis-
Mortlock, eds.), Springer-Verlag NATO-ASI Series I-55, Berlin: 45-56.
Nearing, MA, 1998, Why soil erosion models over-predict small soil losses and under-predict
large soil losses: Catena, 32: 15-22.
Nearing, MA, Bulygin, SY, and Kotova, MM, 1998, Tentative verification and adaptation of the
WEPP model for the Ukrainian condition: problems, solutions, prospects: Pochvovedenie, 31:
96-99.
Nearing, MA, Govers, G, and Norton, LD, 1999, Variability in soil erosion data from replicated
plots: J. Soil Sci. Soc. Am., 63: 1829-1835.
Nearing, MA, 2000, Evaluating soil erosion models using measured plot data: Accounting for
variability in the data: Earth Surf. Proc. Landf., (accepted for publication).
Nicks, AD, Lane, LJ, and Gander, GA, 1995, Weather generator: in USDA-Water Erosion
MA Nearing, eds.), USDA-ARS NSERL Report No. 10.
Onstad, CA, 1984, Depression storage on tilled soil surfaces: Trans. Am. Soc. Agric. Eng., 27:
729-732.
Penman, HL, 1963, Vegetation and hydrology: Tech. Com. No. 53, Commonwealth Bureau Soils,
Harpenden, UK: 125 pp.
Priestly, CHB and Taylor, RJ, 1972, On the assessment of surface heat flux and evaporation using
large scale parameters: Mon. Weath. Rev., 100: 81-92.
Quinton, JN, 1994, Validation of physically based erosion models with particular reference to
EUROSEM: in Conserving Soil Resources: European Perspectives (RJ Rickson, ed.), CAB
International, Wallingford, UK: 300-313.
Quinton, JN, 1997, Reducing predictive uncertainty in model simulations: a comparison of two
methods using the European Soil Erosion Model (EUROSEM): Catena 30: 101-117.
Rapp, JF, 1994, Error assessment of the Revised Universal Soil Loss Equation using natural runoff
plot data, MS Thesis, Univ. Arizona.
Rawls, WJ, Brakensiek, DL, Simanton, JR, and Kohl, KD, 1990, Development of a crust factor for
the Green-Ampt model: Trans. Am. Soc. Agric. Eng., 33: 1224-1228.
Renschler, CS, Flanagan, DC, and Nearing, MA, 2000, Spatially distributed soil erosion
assessment with commonly available data – GIS-based applications with WEPP: Proc. 3rd Int.
Cong. Euro. Soc. Soil Conserv., March 28-April 1, 2000, Valencia, Spain.
Renschler, CS, Flanagan, DC, and Engel, BA, 2001, Data accuracy issues in spatially distributed
soil erosion modeling: what does decision-making gain?: Proc. Soil Erosion Res. for the 21st
Century Symp. (JC Ascough II and DC Flanagan, eds.), Honolulu, Hawaii, January 3-5, 2001,
ASAE: 509-512.
Retta, A, Deer-Ascough, LA, Wagner, LE, Flanagan, DC, and Armbrust, DV, 2001, Common
plant growth component for WEPP and WEPS: Proc. Soil Erosion Res. for the 21st Century
Symp. (JC Ascough II and DC Flanagan, eds.), Honolulu, Hawaii, January 3-5, 2001, ASAE:
380-383.
Risse, LM, Nearing, MA, Nicks, AD, and Laflen, JM, 1993. Assessment of error in the Universal
Soil Loss Equation: J. Soil Sci. Soc. Am., 57: 825-833.
Risse, LM, Nearing, MA, and Savabi, MR, 1994, Determining the Green-Ampt effective hydraulic
conductivity from rainfall-runoff data for the WEPP model: Trans. Am. Soc. Agric. Eng., 37:
411-418.
Savabi, MR, and Williams, JR, 1995, Water balance and percolation: in USDA-Water Erosion
Savabi, MR, Skaggs, RW, and Onstad, CA, 1995, Subsurface hydrology: in USDA-Water Erosion
Savabi, MR, Klik, A, Grulich, K, Mitchell, JK, and Nearing, MA, 1996a, Application of WEPP
and GIS on Small Watersheds in the U.S. and Austria: Proc. "HydroGIS 96", Int. Conf. on
Application of GIS in Hydrology and Water Resources Management., April 16-19, 1996,
Vienna: 469-476.
Savabi, MR, Nearing, MA, Norton, LD, Arnold, J, Rawls, W, and Nicks, AD, 1996b, Global
Change and Agriculture: Soil, Water, and Plant Resources. Vol. II. Climate and Hydrological
Systems: 112.
Simanton, JR, West, LT, Weltz, MA, and Wingate, GD, 1987, Rangeland experiments for Water
Erosion Prediction Project, ASAE Paper No. 87-2545, Am. Soc. Agr. Eng., St. Joseph,
Michigan, 10 pp.
Simanton, JR, Weltz, MA, and Larsen, HD, 1991, Rangeland experiments to parameterize the
Water Erosion Prediction Project model: vegetation canopy cover effects: J. Range Mgmt., 44:
276-282.
Smith, RE, and Williams, JR, 1980, Simulation of the surface water hydrology: in CREAMS: A
Field-Scale Model for Chemicals, Runoff, and Erosion From Agricultural Management
Systems, Vol. II: User Manual (WG Knisel, ed.), Conserv. Res. Report No. 26. USDA-Sci.
Educ. Admin: 165-192.
Stone, JJ, Lane, LJ, and Shirley, ED, 1992, Infiltration and runoff simulation on a plane, Trans.
Am. Soc. Agric. Eng., 35: 161-170.
Stone, JJ, Lane, LJ, Shirley, ED, and Hernandez, M, 1995, Hillslope surface hydrology: in USDA-
(DC Flanagan and MA Nearing, eds.), USDA-ARS NSERL Report No. 10.
Stott, DE, Alberts, EE, and Weltz, MA, 1995, Residue decomposition and management: in USDA-
(DC Flanagan and MA Nearing, eds.), USDA-ARS NSERL Report No. 10.
Tiscareno-Lopez, M, Lopes, VL, Stone, JJ, and Lane, LJ, 1993, Sensitivity analysis of the WEPP
watershed model for rangeland applications, I. hillslope processes: Trans. Am. Soc. Agric.
Eng., 36: 1559-1672.
Tiscareno-Lopez, M, Lopes, VL, Stone, JJ, and Lane, LJ, 1994, Sensitivity analysis of the WEPP
watershed model for rangeland applications, II. channel processes: Trans. Am. Soc. Agric.
Eng., 37: 151-158.
West, LT, Miller, WP, Langdale, GW, Bruce, RR, Laflen, JM, and Thomas, AW, 1991, Cropping
system effects on interrill soil loss in the Georgia Piedmont: J. Soil Sci. Soc. Am., 55: 460-
466.
West, LT, Miller, WP, Bruce, RR, Langdale, GW, Laflen, JM, and Thomas, AW, 1992, Cropping
system and consolidation effects on rill erosion in the Georgia Piedmont: J. Soil Sci. Soc.
Am., 56: 1238-1243.
Williams, JR, and Nicks, AD, 1985, SWRRB, a simulator for water resources in rural basins: an
overview: in Proc. Natural Resources Modeling Symp (DG DeCoursey, ed.), Pingree Park,
Colorado, October 16-20, 1983, USDA-ARS-30:17-22.
Williams, JR, 1995, The EPIC model: in Computer Models of Watershed Hydrology (V.P. Singh,
ed.), Littleton, Colorado: Water Resour. Pub.: 909-1000.
Wischmeier, WH and Smith, DD, 1978, Predicting rainfall erosion losses - A guide to
conservation planning, Agricultural Handbook No. 537, U.S. Dept. Agr., Washington, DC.
Yalin, YS, 1963, An expression for bedload transportation: J. Hydraul. Div., Proc. ASCE,
89(HY3): 221-250.
Zhang, XC, Nearing, MA, and Risse, LM, 1995a, Estimation of Green-Ampt conductivity
parameters: part I. row crops: Trans. Am. Soc. Agric. Eng., 38: 1069-1077.
Zhang, XC, Nearing, MA, and Risse, LM, 1995b, Estimation of Green-Ampt conductivity
parameters: part II. perennial crops: Trans. Am. Soc. Agric. Eng., 38: 1079-1087.
Zhang, XC, Nearing, MA, Risse, LM, and McGregor, KC, 1996, Evaluation of runoff and soil loss
predictions using natural runoff plot data: Trans. Am. Soc. Agric. Eng., 39: 855-863.

2001 Book Chapter The WEPP Model (Flanagan Et Al, 2001)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2001 Book Chapter The WEPP Model (Flanagan Et Al, 2001)

Uploaded by

Copyright:

Available Formats

Chapter 7

The Water Erosion Prediction Project (WEPP) Model

Dennis C. Flanagan, James C. Ascough II, Mark A. Nearing, John M. Laflen

Soil erosion by water continues to be a serious problem throughout the

Erosion prediction models are also critical in conservation planning

2. WEPP MODEL DEVELOPMENT HISTORY

2.1. User Requirements

to incised channels, such as classical gullies, but was to apply to waterways,

2.2. WEPP Experimental Research Program

A range of experimental studies were conducted as part of the WEPP project

Table 1. Major WEPP experimental projects.

Study Objectives Publication(s)

3. WEPP HILLSLOPE MODEL COMPONENT

Application of the WEPP model to hillslope profiles is similar in function to

3.2. Weather Generation

where P is the total precipitation depth (mm) and rl is a dimensionless

Wpred = Wmean obs + STobs (vsn ), (2)

WEPP has the capability to simulate either stationary sprinkler or furrow

3.4.1. Infiltration and Runoff

The hydrology components of WEPP are critical towards providing rainfall

3.4.2. Water B alance

The water balance component of WEPP (Savabi and Williams, 1995) is

3.5. Soil Component

The soil component of WEPP is responsible for the estimation of baseline

3.5.1. Effective Hydraulic Conductivity

Effective hydraulic conductivity is a critical parameter in WEPP model

Kb = -0.265 + 0.0086 SAND1.8 + 11.46 CEC-0.75, (3)

Kb = 0.0066 e (244/CLAY) , (4)

Kerange = -14.29 - 3.40 ln (ROOT10) + 0.3783 SAND + 2.0886 ORGMAT +

where Kerange is effective rangeland hydraulic conductivity in mm⋅h-1 , CEC is

3.5.2. Soil Erodibility

Soil erodibility parameters directly impact the rates of detachment from

Kib = 2728000 + 192100 VFS, (7)

Krb = 0.00197 + 0.00030 VFS + 0.03863 e -1.84 ORGMAT , (8)

τc = 2.67 + 0.065 CLAY - 0.058 VFS, (9)

Kib = 6054000 - 55130 CLAY, (10)

Krb = 0.0069 + 0.134 e -0.20 CLAY , (11)

For rangeland situations, erodibility parameters are predicted using the

Kirange = 1810000 - 19100 SAND - 63270 ORGMAT - 846000 θfc , (13)

Krrange = 0.0017 + 0.000024 CLAY - 0.000088 ORGMAT

τcrange = 3.23 - 0.056 SAND - 0.244 ORGMAT + (0.9 BDdry/1000) , (15)

where Kirange is baseline interrill erodibility in kg⋅s⋅m-4 , Krrange is baseline rill

3.6. Plant Growth

Simulation of plant growth is very important in a continuous simulation

For rangeland situations, the model uses either a unimodal or a bimodal

3.7. Residue Decomposition and Management

M t,j = M t-1,j eENVINDj ⋅ORATEj , (16)

ENVIND = Minimum (WFC, TFC), (17)

Rg t = Rg t -1 (1 - [αf (Smi Tavg)/Cn ]2 ) - Bc , (18)

3.8. Overland Flow Hydraulics

wrill = 1.13 ( Qrill )0.303, (19)

3.9. Soil Erosion

In WEPP hillslope profile simulations, the model separates the upland

where G is sediment load in kg⋅s-1 ⋅m-1 , x is distance downslope in m, Df is rill

Di = Kiadj Ie σir SDRRR Fnozzle (Rs /wrill), (22)

where Kiadj is the adjusted interrill erodibility factor in kg⋅s⋅m-4 , Ie is effective

Ie = (1/te) ∫ Idt, (23)

where Ie is the effective rainfall intensity in m⋅s-1 , I is breakpoint rainfall

RIF = -23 RR + 1.14, (24)

where RIF is the interrill roughness factor and RR is random roughness in m of

DRi = az (RIF) bz, (25)

bz = 0.1286 + 2209 v fi , (27)

DRi = 2.5RIF - 1.5. (28)