BASINS Technical Note 8: Sediment Parameter and Calibration Guidance for HSPF


United States	Office of Water

Environmental Protection	4305	January 2006

Agency

<&EPA BASINS Technical Note 8

Sediment Parameter and Calibration
Guidance for HSPF

-------
ACKNOWLEDGMENTS

This BASINS Technical Note No. 8 was prepared by AQUA TERRA Consultants under Work
Assignment No. 4-01 of EPA Contract No. 68-C-01-037. Mr. Paul Cocca was the EPA Work
Assignment Manager (WAM) for this effort for most of the project schedule, and the individual
who recognized the need for developing this Technical Note to make this information available
to the BASINS user community. Mr. James Carleton assumed the role of WAM for this effort as
the final draft was nearing completion.

For AQUA TERRA, Mr. Tony Donigian and Mr. Brian Bicknell were the primary authors, with
significant input and review provided by Mr. Jason Love and Mr Paul Duda. The sediment
calibration and parameter guidance contained in this document was derived from numerous other
HSPF modeling and source documents (e.g. ARM/NPS user manuals, USLE reports, ASCE
Sedimentation Engineering) and based on many years of sediment modeling with HSPF.
However, the material contained in this document should be viewed as general guidance which
will continue to evolve and change as the sediment modeling technology and data bases expand
and improve with time.

ii

-------
CONTENTS

Introduction	1

Sediment Calibration Overview	1

Sediment Erosion Calibration	3

Instream Sediment Erosion Calibration	3

Pervious Land Erosion (SEDMNT) Parameters	6

Impervious Land Washoff (SOLIDS) Parameters	12

Instream Transport (SEDTRN) Parameters 	 14

References	19

Parameter and Value Range Summary Tables	21

Appendix	24

in

-------
BASINS Technical Note 8:

Sediment Parameter and Calibration Guidance for HSPF

Introduction

This technical note provides BASINS users with guidance on how to estimate the input
parameters in the SEDMNT, SOLIDS, and SEDTRN sections of the Hydrological Simulation
Program Fortran (HSPF) watershed model. It also outlines suggested procedures for sediment
calibration, using a variety of graphical and statistical measures. For each input parameter, this
guidance includes a parameter definition, the units used in HSPF, and how the input value may
be determined (e.g. initialize with reported values, estimate, measure, and/or calibrate). Where
possible, the note discusses how to estimate initial values using the data and tools included with
BASINS. Also discussed, where appropriate, is the physical basis of each parameter and the
corresponding algorithms as described in the HSPF Users Manual (Bicknell, et al, 2001) and
earlier literature sources. In addition to the guidance provided below, model users are directed to
other sources, including Sediment Calibration Procedures And Guidelines For Watershed
Modeling (Donigian and Love, 2003) (Reproduced in the Appendix), the ARM Model User
Manual (Donigian and Davis, 1978) and the HSPF Application Guide (Donigian et al., 1984).

Summary tables are attached that provide 'typical' and 'possible' (i.e. maximum 'expected')
ranges for the parameters discussed below, based on both the parameter guidance and experience
with HSPF over the past two decades on watersheds across the U. S. and abroad (Donigian,
1998). The overarching principal in parameter estimation should be that the estimated
values must be realistic, i.e. make 'physical' sense, and must reflect conditions on the
watershed. If the values estimated by the model user and/or derived from the guidance below,
do not agree with the value ranges in the summary table, the user should question and re-
examine the estimation procedures. The estimated values may still be appropriate, but the user
needs to confirm that the parameter values reflect unusual conditions on the watershed.

Another source of parameter information is the HSPF Parameter Database (HSPFParm)
(US EPA, 1999) http://www.epa.gov/waterscience/ftp/basins/HSPFParm . HSPFParm consists
of parameter values from previous applications of HSPF across North America assimilated into a
single database, and with a customized graphical user interface for viewing and exporting the
data. The pilot HSPFParm Database contains parameter values for model applications in over 40
watersheds in 14 states. The parameter values, contained in the database, characterize a broad
variety of physical settings, land use practices, and water quality constituents. The database has
been provided with a simplified interactive interface that enables modelers to access and explore
the HSPF parameter values developed and calibrated in various watersheds across the United
States, and to assess the relevance of the parameters to their own watershed setting.

Sediment Calibration Overview

Sediment is one of the most difficult water quality constituents to accurately represent in current
watershed and stream models. Important aspects of sediment behavior within a watershed

-------
system include loading and erosion sources, delivery of these eroded sediment sources to
streams, drains and other pathways, and subsequent instream transport, scour and deposition
processes.

Sediment calibration for watershed models involves numerous steps in estimating model
parameters and determining appropriate adjustments needed to insure a reasonable simulation of
the sediment sources on the watershed, delivery to the waterbody, and transport behavior within
the channel system. Rarely is there sufficient observed local data at sufficient spatial detail to
accurately calibrate all parameters for all land uses and each stream and waterbody reach.
Consequently, model users focus the calibration on sites with observed data and review
simulations in all parts of the watershed to ensure that the model results are consistent with field
observations, historical reports, and expected behavior from past experience. This type of
'weight-of-evidence' approach is rapidly becoming the standard practice in watershed modeling.

Sediment calibration must be done after the hydrologic calibration is completed, and it is
extremely sensitive to the hydrology, particularly the amount and timing of surface runoff that is
predicted by the model. Calibration of the parameters involved in simulation of watershed
sediment erosion is more uncertain than hydrologic calibration due to less experience with
sediment simulation in different regions of the country. The process is analogous; the major
sediment parameters are modified to increase agreement between simulated and recorded
monthly sediment loss and storm event sediment removal. However, observed monthly sediment
loss is often not available, and the sediment calibration parameters are not as distinctly separated
between those that affect monthly sediment and those that control storm sediment loss. In fact,
annual sediment losses are often the result of only a few major storms during the year.

As noted above, sediment calibration for watershed models involves numerous steps from initial
estimates of model parameters, all the way to mimicking transport behavior within the channel
system and at the watershed outlet. These steps usually include:

1.	Estimating target (or expected) sediment loading rates from the landscape, often as a
function of topography, land use, and management practices

2.	Calibrating the model loading rates to the target rates

3.	Adjusting scour, deposition and transport parameters for the stream channel to mimic
expected behavior of the streams/waterbodies

4.	Analyzing sediment bed behavior (i.e. bed depths) and transport in each channel reach as
compared to field observations

5.	Analyzing overall sediment budgets for the land and stream contributions, along with
stream aggrading and degrading behavior throughout the stream network

6.	Comparing simulated and observed sediment concentrations, including particle size
distribution information, and load information where available

7.	Repeating steps 1 through 6 as needed to develop a reasonable overall representation of
sediment sources, delivery, and transport throughout the watershed system

Parameter guidance is given for each of the modules required to simulate sediment delivery from
the landscape (i.e. SEDMNT, SOLIDS) and instream transport (i.e., SEDTRN) and the
parameters are grouped as required in each UCI table.

2

-------
Sediment Erosion Calibration

Sediment loadings to the stream channel are estimated by land use category from literature data,
local Extension Service sources, or procedures like the Universal Soil Loss Equation (USLE)
(Wischmeier and Smith, 1978) and then adjusted for delivery to the stream with estimated
sediment delivery ratios (SDRs). This delivery adjustment is needed because HSPF, like most
watershed-scale (lumped parameter) models, represents landscape loadings to the stream
channel, which are less than the field-scale estimates from USLE. These estimated loading rates
then become 'calibration targets' for the watershed model.

Model parameters are then adjusted so that model-calculated loadings are consistent with these
estimated 'calibration targets' and loading ranges. The model-calculated loadings are further
evaluated in conjunction with the instream sediment transport calibration (discussed below) that
extends to a point in the watershed where sediment concentration and/or load data are available.
The overall objective is to represent the overall sediment behavior of the watershed, with
knowledge of the morphological characteristics of the stream (i.e. aggrading or degrading
behavior), using sediment loading rates that are consistent with the calibration targets and
modeled concentrations that provide a reasonable match with instream sediment data.

In HSPF, the erosion process on pervious land areas is represented as the net result of
detachment of soil particles by raindrop impact on the land surface, and then subsequent
transport of these fine particles by overland flow. On impervious surfaces (e.g. parking lots,
driveways), soil splash by raindrop impact is neglected and solids washoff is often controlled by
the rate of accumulation of solid materials. The primary sediment erosion parameters are the
coefficients in the soil detachment equation for pervious areas, the coefficients in the sediment
washoff equations for pervious and impervious areas, and the accumulation rate of solids on
impervious surfaces.

In general, sediment calibration involves the development of an approximate equilibrium or
balance between the accumulation and generation of sediment particles on one hand and the
washoff or transport of sediment on the other hand. Thus, the accumulated sediment on the land
surface (i.e., DETS and SLDS) should not be continually increasing or decreasing throughout the
calibration period. Extended dry periods will produce increases in surface sediment
accumulations, and extended wet periods will produce decreases. However, the overall trend
should be relatively stable from year to year. This equilibrium must be developed on both
pervious and impervious surfaces, and must exist in conjunction with the accurate simulation of
monthly and storm event sediment loss, depending on the data available for calibration.

Donigian and Love (2003) (see Appendix) provide additional discussion of sediment calibration
procedures.

Instream Sediment Transport Calibration

-------
be compared with observations. The initial steps in instream calibration involve dividing the
input sediment loads into appropriate size fractions, estimating initial parameter values and
storages for all reaches, and a preliminary model run to calculate shear stress timeseries in each
reach to estimate critical scour and deposition values.

The eroded material is fractionated into sand, silt, and clay prior to entering a model reach using
available soils information; typically, a single fractionation scheme is used for all reaches unless
soils and land surface variations within the watershed support use of reach-specific fractions.
The fractions should reflect the relative percent of the surface material (i.e., sand, silt, clay)
available for erosion in the surrounding watershed, but also should include an enrichment factor
of silt and clay to represent the likelihood of these finer materials reaching the channel.

The sand, silt and clay fractions of total eroded sediment are specified in the MASS-LINK block.
Each unique fractionation scheme will require a separate table in the MASS-LINK block. An
example MASS-LINK table containing the sediment fractionation is shown below, with the sand,
silt, clay fractions shown as 0.05, 0.70, and 0.25, respectively.

MASS-LINK

PERLND
PERLND
PERLND
PERLND
PERLND

END MASS-LINK

1

PWATER PERO
SEDMNT SOSED 1
SEDMNT SOSED 1
SEDMNT SOSED 1
PWTGAS POHT
1

0.0833
0. 05
0.70
0.25
1 . 0

RCHRES

RCHRES
RCHRES
RCHRES
RCHRES

INFLOW IVOL

INFLOW ISED 1
INFLOW ISED 2
INFLOW ISED 3
INFLOW IHEAT

For HSPF, initial sediment parameters, such as particle diameter, particle density, settling
velocity, bed depth and composition, and beginning calibration parameter values can be
evaluated from local/regional data, past experience, handbook values, etc., and then adjusted
based on available site specific data and calibration. Bed composition data are especially
important so that the model results can be adjusted to reflect localized aggradation (deposition)
or degradation (scour) conditions within the stream system.

In HSPF, the transport of the sand (non-cohesive) fraction is commonly calculated as a power
function of the average velocity in the channel reach in each timestep. This transport capacity is
compared to the available inflow and storage of sand particles; the bed is scoured if there is
excess capacity to be satisfied, and sand is deposited if the transport capacity is less than the
available sand in suspension within the channel reach.

For the silt and clay (cohesive) fractions, shear stress calculations are performed by the
hydraulics (HYDR) module, and then in the SEDTRN module they are compared to user-defined
critical, or threshold, values for deposition and scour for each size (shown in Figure 1). When
the shear stress for a timestep is greater than the critical value for scour, the bed is scoured at a
user-defined erodibility rate and transport through the reach occurs; when the shear stress is less
than the critical deposition value, the silt or clay fraction deposits at a settling rate input by the
user for each size. If the shear stress falls between the critical scour and deposition values, the
incoming suspended material is transported through the reach.

In HSPF, the hydraulic characteristics of a stream reach are represented by a function table
(FTABLE) that includes the relationships between stage, storage (volume), surface area, and

4

-------
discharge. The accuracy of the FTABLE for a specific reach will be a critical factor in
adequately representing the average velocity and hydraulic radius and calculated shear values, as
a function of the stage, or depth of flow. This is especially evident for simulations of flood flows
that exceed bankfull discharges; improper extension of the FTABLES can lead to erroneous
velocities and shears and scour conditions during high flow events, and have major impacts on
the model simulations for those events.

-------
The shear stress values are then adjusted more carefully in calibration so that scour occurs during
storm periods and deposition occurs at low flows. Once the timing of scour and deposition
processes is correct, the rate of scour is adjusted in an attempt to match either expected behavior
within each reach, and/or the observed concentrations. During high flow periods, the amount of
scour is adjusted with an erodibility factor for each reach that controls the rate of scour whenever
the actual shear stress is greater than the critical shear stress value for scour. During low flow
periods the silt/clay fall velocity parameter can be adjusted slightly to improve the agreement.
The Donigian and Love (2003) paper in the Appendix provides further discussion of the potential
impact of FTABLEs on the shear stress calculations and resulting sediment concentrations. Note
also: In HSPF the units of shear stress are Vo-force/ft2. Typically, shear stress values are
expressed in the literature as Pascals (Pa) or Newtons/m2 (N/m2). The conversion is 1 Pa = 1
N/m2 = 0.02088 lb/ft2.

The calibration procedure generally involves comparison of model simulations (concentrations
and loads) to available observed data. This is often limited to event mean concentrations of total
suspended solids (TSS) for selected storm events and nonstorm (baseflow) periods, or
pollutographs of TSS concentrations throughout a few events. However, other types of
comparisons are also possible, such as load estimates and sediment rating curves; see the
Appendix for examples of these types of comparisons.

Parameter Guidance

The parameter guidance below is listed in the order of the parameter tables required by each
module section responsible for simulating sediment erosion and solids washoff from pervious
and impervious surfaces, and instream sediment transport, deposition, and scour (i.e. SEDMNT,
SOLIDS, and SEDTRN). The parameters are grouped as required in each UCI table.

Pervious Land Accumulation and Removal of Sediment (SEDMNT) Parameters

The SEDMNT section simulates the production and removal of sediment from a pervious land
segment (PERLND).

SED-PARM1 Table:

CRV Flag to select constant (CRV=0) or monthly-variable (CFV=1) erosion-related

cover, COVER. Monthly values are commonly used to reflect seasonal
variability of the vegetation or other erosional cover, e.g., for agricultural areas to
reflect the timing of cropping and tillage practices.

VSIV Flag to select a constant or monthly-variable rate of net vertical sediment input,
NVSI. Constant annual values (VSIV=0) are commonly used since atmospheric
sources are usually difficult to quantify and are a small fraction of surface storage
for pervious land surfaces. However, monthly values (VSIV=1) can be used to
reflect variability of NVSI as impacted by seasonal land surface activities, such as
tillage, if data are available to estimate NVSI.

-------
SDOP Flag to select the algorithm used to simulate removal of sediment from the land
surface; choose either the method used in predecessor models (HSPX, ARM, and
NPS) (SDOP=l), or an alternative method as described in the HSPF User Manual
(SDOP=0). Recommendation: set SDOP to 1. This method, used in the
predecessor models is more commonly used, and has been subjected to more
widespread application.

SED-PARM2 Table:

This table includes parameters for estimating the production and reduction of detached sediment
on the pervious land.

SMPF Supporting Management Practice Factor (unitless) {measure/estimate). SMPF is
used to simulate the reduction in erosion achieved by use of erosion control
practices. SMPF is analogous to the P factor in the Universal Soil Loss Equation
(Wischmeier and Smith, 1978), and initial values should be set equal to the P
value for practices such as contouring, terracing, strip-cropping, etc. Table 1
shows values of P for alternative practices and slope conditions. Model users
should note that the practices listed in Table 1 also affect other HSPF parameters,
such as NSUR, UZSN, LSUR, and SLSUR. The impact of different agricultural
practices can only be evaluated with changes in all relevant parameters. Renard et
al., (1997) provide detailed discussions on the physical basis and estimation of
each of the USLE parameters, and as such it is also a valuable source of guidance
for HSPF sediment parameters.

Use of BASINS Data/Tools:

In cases where GIS coverages are available for alternative management practices,
BASINS GIS capabilities can be used to identify land areas with similar practices
each of which may be represented as a separate PERLND with its own P value.
Alternatively, the P values of different practices may be weighted by area
fractions within a single PERLND.

KRER Coefficient in the soil detachment equation (complex) {measure/estimate, then
calibrate as needed to achieve target loading rates). This parameter is related to
the erodibility or detachability of the specific soil type and surface conditions.
Experience indicates that KRER is directly related to the K factor in the USLE
and can be initially estimated as KRER = K. K values can be obtained with
techniques published in the literature or from soil scientists familiar with local soil
conditions. Table 2 shows representative K values by soil texture and organic
matter content. Renard et al (1997) provide extensive discussion on estimation of
K from soil and land surface characteristics. Since adjustments to KRER will
affect the amount of detached sediment that can be delivered to streams, users
should review the detached sediment storage (DETS) to ensure that the
accumulated sediment on the land surface is not continually increasing or
decreasing throughout the simulation period.

-------
Table 1. Values of Support-Pract

ice Factor, P

Land Slope (%):

1.1-2.0

2.1-7.0

7.1-12.

12.1-18.

18.1-24.

Practice

Contouring (Pc)

0.60

0.50

0.60

0.80

0.90

Contour Strip Cropping (Psc)

R-R-M-M (1)

0.30

0.25

0.30

0.40

0.45

R-W-M-M

0.30

0.25

0.30

0.40

0.45

R-R-W-M

0.45

0.38

0.45

0.60

0.68

R-W

0.52

0.44

0.52

0.70

R-0

0.60

0.50

0.60

0.80

0.90

Contour Listing or Ridge Planting

0.30

0.25

0.30

0.40

0.45

Contour Terracing (P,) (2,3)

0.6/Vn

No Support Practice

1.0

(1) R = row crop, W = fall-seeded grain, 0 = spring-seeded grain, M = meadow. The crops are grown in
rotation, and are arranged on the field such that rowcrop strips are always separated by a meadow or
winter-grain strip.

(2) These Pt values estimate the amount of soil eroded to the terrace channels and are used for
conservation planning. For prediction of off-field sediment, the Pt values are multiplied by 0.2

(3) n = number of approximately equal-length intervals into which the field slope is divided by the
terraces. Tillage operations must be parallel to the terraces.

Source: Stewart, et al., 1975

Use of BASINS Data/Tools:

The State Soil (STATSGO) data layer contains data on 'kffaci\ defined as the soil
erodibility factor that is fragment free for use in the USLE. Run the BASINS
State Soil Characteristic Report and select mean estimate, area-weighted, surface
layer, for 'kffactThe TMDL USLE Tool (Hummel et al., 2000; discussed in the
Appendix) also provides guidance in selecting USLE parameters.

JRER Exponent in the soil detachment equation (complex) (initialize with reported

value, then calibrate as needed). JRER approximates the relationship between
rainfall intensity and incident energy to the land surface for the production of soil
fines. Wischmeier and Smith (1978) proposed the following relationship for the
kinetic energy produced by natural rainfall;

Y = 916+ 331 log X

Where Y = kinetic energy, ft/ton/acre/in.

X = rainfall intensity, inches/hr

-------
Table 2. Representative Values o

'the Soil Erodibility Factor, K

Organic Matter Content (%):

<0.5

2.0

4.0

Texture Class

Sand

0.05

0.03

0.02

Fine Sand

0.16

0.14

0.10

Very Fine Sand

0.42

0.36

0.28

Loamy Sand

0.12

0.10

0.08

Loamy Fine Sand

0.24

0.20

0.16

Loamy Very Fine Sand

0.44

0.38

0.30

Sandy Loam

0.27

0.24

0.19

Fine Sandy Loam

0.35

0.30

0.24

Loam

0.38

0.34

0.29

Silt Loam

0.48

0.42

0.33

Silt

0.60

0.52

0.42

Sandy Clay Loam

0.27

0.25

0.21

Clay Loam

0.28

0.25

0.21

Silty Clay Loam

0.37

0.32

0.26

Sandy Clay

0.14

0.13

0.12

Silty Clay

0.25

0.23

0.19

Clay

0.13-0.29

These values are estimated averages of bn
is near the borderline of two texture classe

3ad ranges of sp<
s. use the avcraj.

xific soil values
;e of the two K v

When a texture
alues.

Source: Stewart et al., 1975.

Using this relationship, various investigations have also shown that soil splash is
proportional to the square of the rainfall intensity (Meyer and Wischeimer 1969,
David and Beer 1974). Thus, a value of about 2.0 for JRER is predicted from
these studies. Most HSPF applications have used the default value of 2.0 for this
exponent. Use the default value of 2.0, and adjust only if supported by local data
and conditions.

AFFIX The fraction by which detached sediment storage decreases each day during non-
storm periods (/day) (estimate, then calibrate as needed). AFFIX is a soil
compaction factor that reduces the amount of detached soil particles available for
transport. This parameter attempts to represent the natural aggregation and
mutual attraction of soil particles and the compaction of the surface soil zone
from which erosion occurs. These processes are a complex function of soil
characteristics, meteorologic conditions, and tillage practices for which a detailed
simulation is not possible. Values in the range of .001 to .1 are possible.

Typically, AFFIX is adjusted in combination with the KRER and NVSI
parameters to ensure that the accumulated detached sediment on the land surface

-------
(DETS) does not continually increase or decrease throughout the calibration
period.

COVER The fraction of land surface which is shielded from rainfall (unitless) {measure/
estimate, then calibrate as needed), and is therefore not susceptible to soil fines
detachment by raindrop impact. Seasonal/monthly values are often used.
Overhead photographs at periodic intervals during the year are the most direct
means of estimating the land cover fraction.

COVER values are sometimes estimated as one minus the monthly C factor in the
USLE. For cropland, the C factors for the various stages of crop growth should
be used in estimating COVER.

Tables 5.7 and 5.8 in the ARM User Manual (Donigian and Davis, 1978) pertain
to the evaluation of C on undisturbed lands and have been reproduced from the
paper by Wischmeier (1975). C factors for disturbed lands (croplands,
agriculture, and construction areas) have been published in the USLE Report
(Wischmeier and Smith 1965). The monthly COVER values estimated from C
may need to be reduced since the C factor includes considerations other than crop
canopy and raindrop interception since it represents a soil loss ratio, i.e. the ratio
of soil loss for the current land surface conditions compared to clean-tilled
continuous fallow conditions. Renard et al (1997) provide the most extensive
recent guidance on estimation of the C factor. Users should avoid using COVER
values of 0.98 to 1.0, even for dense forest or crop canopy conditions, since this
will essentially eliminate any soil detachment and subsequent washoff of detached
materials. In these cases, the lower COVER values allow for fringe and boundary
areas of a PERLND that may contribute soil to the stream.

NVSI The rate at which sediment enters detached storage from the atmosphere
(lb/ac/day), {estimate, then calibrate as needed). NVSI is often input with
monthly variation. It represents any detached sediment accumulation processes
not covered by rainfall impact or agricultural tillage operations that are generally
specified using Special Actions. This can include the effects of wind-blown
sediments, or land disturbance activities such as construction or landscaping. Note
that NVSI can be negative, if the effects of the AFFIX parameter are not
sufficient to represent all detached sediment reduction processes.

SED-PARM3 Table:

This table contains parameters for estimating the sediment removal from the pervious land by

washoff and gully erosion processes.

KSER Coefficient in the soil washoff or transport equation (complex), {estimate, then
calibrate as needed). It is an attempt to combine the effects of slope, overland
flow length, sediment particle size, and surface roughness on the sediment
transport capacity of overland flow into a single parameter. Consequently,

-------
calibration is the major method of evaluating KSER. Terracing, tillage practices,
and other agricultural management techniques will have a significant effect on
KSER. Experience to date has indicated a possible range of values of 0.01 to 5.0.
However, variations from this can be expected.

JSER Exponent in the soil washoff equation (complex), {initialize with reported value,
then calibrate as needed). JSER approximates the relationship between overland
flow intensity and sediment transport capacity. The vast majority of HSPF
applications have used the default value of 2.0 for this exponent. Use the default
value of 2.0, and adjust only if supported by local data and conditions.

KGER Coefficient in the matrix soil equation, which simulates gully erosion (complex),
{estimate, then calibrate as needed). Unless there is evidence of gully erosion
occurring in the watershed, KGER should be set to 0.0.

JGER Exponent in the matrix soil equation, which simulates gully erosion (complex),
{estimate, then calibrate as needed). The vast majority of HSPF applications
have used the default value of 2.0 for this exponent, but few applications even
include gully erosion. Use an initial value of 2.5, since JGER is expected to be
greater than JSER, and adjust if supported by local data and conditions.

Monthly Input Parameter Tables:

In general, monthly variation in selected parameters, such as COVER and NVSI should be
included with the initial parameter estimates. The monthly values represent the variable on the
first day of each month; the values other days are then interpolated from the values for the
current and following months. All monthly values can be adjusted to calibrate for seasonal
variations.

MON-COVER Table:

Monthly values for the fraction of land surface which is shielded from rainfall. Monthly values
are often used, since COVER is primarily a function of the seasonal and vegetation canopy
changes.

MON-NVSI Table:

Monthly values for the rate at which sediment enters detached storage from the atmosphere.

-------
Impervious Land Accumulation and Removal of Solids (SOLIDS) Parameters

The SOLIDS section simulates the accumulation and removal of solids from an impervious land

segment (IMPLND).

SLD-PARM1 Table:

VASD Flag to select a constant (VASD=0) or monthly-variable (VASD=1) rate of solids
accumulation, ACCSDP. Monthly values are commonly used to reflect
variability of the accumulation as impacted by climate and seasonal land surface
activities.

VRSD Flag to select a constant (VRSD=0) or monthly-variable (VRSD=1) rate of solids
removal, REMSDP. Monthly values are commonly used to reflect variability of
the removal rate as impacted by seasonal land surface activities.

SDOP Flag to select algorithm used to simulate removal of sediment from the

impervious surface; choose either the method used in predecessor models (HSPX,
ARM, and NPS), (SDOP=l), or the alternative method as described in the HSPF
User Manual (SDOP=0). Recommendation: Set SDOP to 1; this method is more
commonly used, and has been subjected to more widespread application.

SLD-PARM2 Table:

KEIM Coefficient in the solids washoff equation (complex), {estimate, then calibrate as
needed). This parameter is an attempt to combine the effects of slope, overland
flow length, sediment particle size, and surface roughness on the sediment
transport capacity of overland flow into a single parameter. Consequently,
calibration is the major method of evaluating KEIM. Experience to date has
indicated a possible range of values of 0.01 to 5.0. However, variations from this
can be expected.

JEIM Exponent in the solids washoff equation (complex), {estimate, then calibrate as
needed). Values in the range of 1 to 2.5 are reasonable, with most models using a
value of 1.6 to 2.0. Use an initial value of 1.8, and then adjust if supported by
local data and conditions.

ACCSDP The rate solids accumulate on the impervious land surface (tons/ac/day),

{estimate, then calibrate as needed). Data from street surfaces suggests values in
the range of 0.0005 to 0.1, with most data in the range of 0.001 to 0.02. Note that
ACCSDP is in units of 'tons/ac/day', whereas the corresponding rate for pervious
surfaces (NVSI) is in lbs/ac/day.

REMSDP The fraction of solids storage which is removed each day when there is no runoff
(per day) {estimate, then calibrate as needed). These removal processes include

-------
wind, air currents from traffic, aggregation to larger, less transportable particles,
and street cleaning activities. Values should range from 0.001 to 0.1, with typical
values in the range 0.001 to 0.07. The effects of street cleaning can be estimated
as:

R = P*(E/D)
where R = sediment removal by street cleaning

P = fraction of impervious area where cleaning is performed
E = efficiency of cleaning
D = frequency of cleaning

For example, if cleaning is performed every seven days on 40% of the area with
an efficiency of 80%, then

R = (.4) (,8)/(7) = 0.046

If wind removal is estimated as 0.02, then REMDSP would be approximately
0.066. Typically, removal rates are evaluated in conjunction with accumulation
rates to establish a limit to the total sediment accumulation that can occur. This
limit is given by ACCSDP/REMSDP. Consequently, joint calibration of
accumulation and removal rates is recommended.

Monthly Input Parameter Tables:

In general, monthly variation in selected parameters, such as SACCUM and REMOV should be
included with the initial parameter estimates. As noted above, the monthly values represent the
variable on the first day of each month; the values for other days are then interpolated from the
values for the current and following months. The monthly values can be adjusted to calibrate for
seasonal variations.

MON-SACCUM Table:

Monthly values for the rate of solids accumulation on the land surface (ACCSDP).
MON-REMOV Table:

Monthly values for the fraction of solids storage which is removed each day when there is no
runoff (REMSDP).

-------
Instream Sediment Transport (SEDTRN) Parameters

The SEDTRN section simulates the transport, deposition, and scour of inorganic sediment from a

free-flowing reach or mixed reservoir (RCHRES).

SANDFG Table:

SDFG Flag to select the method that will be used for sandload simulation.

1. Toffaleti - developed for wide rivers where hydraulic radius is
approximately equal to depth; not often used due to lack of calibration
parameters

2. Colby - developed for wide rivers where hydraulic radius is
approximately equal to depth; not often used due to lack of calibration
parameters

3. User Specified Power Function - most frequently used at the current
time

SED-GENPARM Table:

BEDWID The effective width over which bed sediment is deposited; the BEDWID is

constant regardless of stage, top width, etc. (ft), (
-------
values near 0.45. For beds with more cohesives, the values can be higher, with
values up to 0.9 for newly deposited muds and peat. As consolidation occurs,
particularly if the channel dries out for extended periods, the porosity would
decrease to values closer to 0.4 - 0.5.

SED-HYDPARM Table:

This table is only required if section HYDR is not active. If HYDR is active, as it is in most

simulations, these quantities are specified in the HYDR section.

LEN Length of the stream reach (miles), {measure). Length is used in the computation

of auxiliary variables, including hydraulic radius, flow velocity, and shear stress,
which are used to simulate sediment transport in SEDTRN.

Use of BASINS Data/Tools:

This is populated automatically by BASINS during model initialization.

DELTH Change in elevation from the upstream end of the stream reach to the downstream
end (feet), (measure). DELTH is used to compute channel slope for (1)
calculation of shear stress for cohesives (silt and clay), and (2) if sandload
transport capacity is computed using either the Toffaleti or Colby method in the
SEDTRN Block.

Use of BASINS Data/Tools:

This is populated automatically by BASINS, during model initialization, from the
DEM by selecting the upstream and downstream elevations for the HSPF reach
boundaries. Thus the accuracy of this slope calculation depends on the resolution
and accuracy of the DEM.

DB50 Median diameter of the bed sediment (inches), (estimate/measure). DB50 is used
to calculate: (1) the bed shear stress if the reach is a lake; and (2) the rate of sand
transport if the Toffaleti or Colby method is used. Note: DB50 is not connected
with the sand particle diameter (D) input in the SAND-PM table. Sediment
diameter values can be obtained from texts on sedimentation/hydrology and from
particle size analysis of the sediments in the watershed. Sand particles have
diameters typically ranging from 0.05 - 2 mm (0.002 - 0.08 in). Table 3,
reproduced from ASCE (1975) shows a sediment grade scale.

SAND-PM Table:

D The effective diameter of the sand particles (in), (,measure, estimate). Sediment

diameter values can be obtained from texts on sedimentation/hydrology and from
particle size analysis of the sediments in the watershed. Sand particles have
diameters typically ranging from 0.05 - 2 mm (0.002 - 0.08 in); see Table 3
above. Note: D is not used in the calculations; it is included in this table for
consistency with the cohesive sediment data inputs. The Colby and Toffaleti

-------
calculations for sand transport use the DB50 parameter, which is entered in the
HYDR-PARM2 table. Therefore, enter a nominal value of 0.01 inches.

Table 3. Sediment Particle Diameters and Fall Velocities in Still Water

Class Name

Diameter1
(mm)

Fall Velocity2
(cm/sec)

Very coarse sand

2.0- 1.0

20.

Coarse sand

p
L/i

12.

Medium sand

0.5-0.25

Fine sand

0.25-0.125

2.2

Very fine sand

0.125 - 0.062

0.75

Coarse silt

0.062-0.031

0.16

Medium silt

0.031 -0.016

0.04

Fine silt

0.016-0.008

0.01

Very fine silt

0.008 - 0.004

0.0027

Coarse clay

0.0040 - 0.0020

0.0006

Medium clay

0.0020-0.0010

0.00015

Fine clay

0.0010-0.0005

0.00004

Very fine clay

0.0005 - 0.00024

0.00001

Notes:

1. Source: ASCE, 1975

2. Fall velocity in still water; for diameters <0.125 mm, estimated based on
Stokes Law; assumed: median diameter from column 1, temperature = 24
degC, and density = 2.65 g/cm3. For larger particles, where Stokes Law does
not apply, used estimated data for sand particles from Rouse (1937).

W The fall velocity of the sand particles in still water (in/sec), (measure, obtain

from literature, estimate). Note: for sand transport, W is only used in the
Toffaleti method; therefore, if the Colby method or power function method is
being used, enter a nominal value of 0.4 in/sec.

Particle velocities in still water can be estimated using simple equations such as
Stokes' Law for the terminal velocity of a spherical particle. See texts on
sedimentation (e.g., SCS, (1983) and ASCE, (1975)) or screening procedure
reference texts, for example Mills et al., (1985). Table 3 shows sediment settling
velocities in cm/s. Sand settling velocities will typically be in the range of 0.1 to 4
in/sec.

RHO The density of the sand particles (g/cm3), (measure, obtain from literature,

estimate). Sediment densities are used to calculate the bed depth along with the
porosity (POR). Sediment density values can be obtained from texts on

-------
sedimentation/hydrology and from laboratory analysis of the sediments in the
watershed. Typical values of RHO range from 1.5 to 2.8 g/cm3. If local data are
not available, use 2.6 g/cm3.

KSAND The coefficient in the sandload power function, should be included if SDFG = 3.

(complex) {calibrate). The sand transport power function is based on velocity.
This equation will produce reasonable results if the computed velocites, which are
determined from the volume and discharge columns in the FTABLE, are
reasonable. This coefficient is a calibration parameter; start with a value of 0.1
and adjust, in concert with EXPSND, to improve the comparison between
simulated and observed sand concentrations.

EXPSND The exponent in the sandload power function, should be included if SDFG = 3.

(complex) {calibrate). See the discussion for KSAND above. Begin with a value
of EXPSND of 2.0 and adjust slightly, in concert with KSAND.

SILT-CLAY-PM Table:

This table should be entered twice. The first occurrence provides parameters for silt; the second
contains the clay parameters.

D The effective diameter of the silt or clay particles (in), {measure, estimate).

Sediment diameter values can be obtained from texts on sedimentation/hydrology
and from particle size analysis of the sediments in the watershed. Table 3 contains
a sediment grade scale. Silt particles have diameters typically ranging from 0.005
- 0.05 mm (0.0002 - 0.002 in). Clay particles range from 0.0002 to 0.004 mm
(8.E-6 - 0.00015 in). In the absence of measured data, use 0.0006 inches for silt
and 0.0001 inches for clay.

W The fall velocity of the silt or clay particles in still water, (in/sec) {measure,

obtain from literature, estimate). Particle velocities in still water can be
estimated using simple equations such as Stokes' Law for the terminal velocity of
a spherical particle, with adjustments for drag for non-spherical particles. See
texts on sedimentation or screening procedure reference manuals, for example
Mills et al., (1985). Table 3 contains estimated values of W (cm/s) as a function of
particle diameter and density. In the absence of detailed data and calculations, use
a value of 0.0005 in/sec for silt and 0.00005 in/sec for clay.

RHO The density of the silt or clay particles, (g/cm3), (measure, obtain from literature,
estimate). Sediment densities are used to calculate the bed depth along with the
porosity (POR). Sediment density values can be obtained from texts on
sedimentation/hydrology and from laboratory analysis of the sediments in the
watershed. Typical values of RHO range from 1.5 to 2.8 g/cm3. If local data are
not available, use 2.3 g/cm3 for silt and 2.0 for clay.

-------
TAUCD The critical bed shear stress for deposition (lb/ft2), (calibrate). Initial values of
TAUCD and TAUCS should be estimated on a reach-by-reach basis by
examining graphs of the simulated shear stress timeseries (TAU) graphed at the
timestep of the simulation. Assign values of TAUCS that are below the
maximum of the curve, and TAUCD that are above the minimum of the curve.
Calculated shear stress values should also be inspected to verify they are
reasonable. The FTABLE volume and discharge columns determine the shear
stress along with the simulated discharge, and unreasonable shear stress values
should first be corrected by adjusting the FTABLE. Generally, TAUCS values
will be greater than TAUCD, and the values of both parameters for silt will be
greater than or equal to those for clay. Adjust TAUCD and W (particle fall
velocity) to calibrate the timing and magnitude of silt and clay concentrations.
Increasing TAUCD will result in increasing the occurrence and magnitude of
deposition and vice versa.

TAUCS The critical bed shear stress for scour (lb/ft2), (calibrate). See the discussion for
TAUCD above to set initial values of TAUCS. Adjust TAUCS in concert with
the erodibility coefficient (M) to calibrate the timing and magnitude of silt and
clay concentrations. Increasing TAUCS will result in reductions in the occurrence
and magnitude of scour and vice versa.

Procedures are available to calculate critical shear stress values from Shields'
equation using bed and channel properties, as follows:

Tc = 0 (ys - Y) D

where 9 is the dimensionless Shields parameter for entrainment of a sediment
particle of size D, Ys is the unit weight of bed sediment, and y is the unit weight of
water. Donigian and Love (2005) have used these procedures to estimate tc values
and assess channel stability issues in urbanizing watersheds using HSPF. These
same calculations can be used to develop initial TAUCS values which would
subsequently be adjusted in calibration.

M The erodibility coefficient of the sediment (lb/ft2.d), (calibrate). This coefficient

is entirely a calibration parameter. Set it to 0.01 and then adjust it (in concert with
TAUCS, and with consideration of the balance between land sediment loading
and channel scour) to result in reasonable silt and clay concentrations during
scour conditions in the reach.

-------
| REFERENCES

ASCE. 1975. Sedimentation Engineering, ASCE Manuals and Reports on Engineering Practice,
No. 54, V.A.Vanoni, ed., American Society of Civil Engineers, New York.

Bicknell, B.R., J.C. Imhoff, J.L. Kittle Jr., A.S. Donigian, Jr., T.H. Jobes, and R.C. Johanson.
2001. Hydrological Simulation Program - FORTRAN, User's Manual for Version 12.
U.S. EPA, National Exposure Research Laboratory, Athens, GA.

David, W.P. and C.E. Beer. 1974. Simulation of Sheet Erosion, Part 1. Development of a
Mathematical Erosion Model. Iowa Agriculture and Home Economics Experiment
Station. Ames, Iowa. Journal Paper No. J-7897.

Donigian, A.S. Jr., and J.T. Love. 2005. The Use of Continuous Watershed Modeling to address
Issues of Urbanization and Channel Stability in Southern California. ASCE World Water
and Environmental Resources Congress 2005. May 16-19, 2005. Anchorage, AK. (paper
also attached in Appendix).

Donigian, A.S. Jr., and J.T. Love. 2003. Sediment Calibration Procedures and Guidelines for
Watershed Modeling. WEF TMDL 2003, November 16-19, 2003. Chicago, Illinois.

Donigian, A.S., Jr. 1998. Personal communication, 1998.

Donigian, A.S. Jr., J.C. Imhoff, B.R. Bicknell and J.L. Kittle. 1984. Application Guide for

Hydrological Simulation Program Fortran (HSPF), prepared for U.S. EPA, EPA 600/3
84 065, Environmental Research Laboratory, Athens, GA.

Donigian, A.S., Jr. and H.H. Davis, Jr. 1978. User's Manual for Agricultural Runoff
Management (ARM) Model, U.S. Environmental Protection Agency, EPA_
600/3_78_080.

EPA, 1999. HSPFParm: An Interactive Database of HSPF Model Parameters, Version 1.0.

EPA_823_R_99_004. U.S. EPA, Office of Water, Washington, DC. Available from the
BASINS web site http://www.epa.gov/waterscience/ftp/basins/HSPFParm .

Hummel, P.R., J.C. Imhoff, R. Dusenbury and M. Gray. 2000. TMDL USLE, A Practical Tool

for Estimating Diffuse Sediment Source Loads within a Watershed Context. Prepared for:
U.S. EPA National Exposure Research Laboratory, Athens, GA.
(http://www.epa.gov/ceampubl/swater/usle).

Meyer, L.D. and W.H. Wischmeier. 1969. Mathematical Simulation of the Processes of Soil
Erosion by Water, Trans. Am. Soc. Agric. Eng. 12(6):754-762.

Mills, W. B., D. B. Porcella, M. J. Ungs, S.A. Gherini, K.V. Summers, L. Mok, G.L. Rupp, and
G.L. Bowie. 1985. Water Quality Assessment: A Screening Procedure for Toxic and

-------
Conventional Pollutants in Surface and Groundwater. Vols. I and II. EPA/600/6-85/002.
NTIS PB86-12249 6, U.S. EPA, Athens, Georgia.

Renard, K.G., G.R. Foster, G.A. Weesies, D.K. McCool, andD.C. Yoder, coordinators. 1997.
Predicting Soil Erosion by Water: A Guide to Conservation Planning With the Revised
Universal Soil Loss Equation (RUSLE). U.S. Department of Agriculture, Agriculture,
Handbook No. 703, 404 pp.

Rouse, H. 1937. Nomogram for the Settling Velocity of Spheres, Division of Geology and
Geography Exhibit D of the Report of the Commission on Sedimentation, 1936-37,
National Research Council, Washington, DC.

Soil Conservation Service. 1983. National Engineering Handbook. Section 3 - Sedimentation.
United States Department of Agriculture, Washington, DC.

Stewart, B. A., D. A. Woolhiser, W. H. Wischmeier J. H. Caro, and M. H. Frere. 1975. Control
of Water Pollution from Cropland, Volume 1, A manual for guideline development. U.S.
EPA and USD A, Report Nos. EPA-600/2-75 026a and ARS H-5-1, Washington, DC.

Wischmeier, W.H. 1975. Estimating the Soil Loss Equation's Cover and Management Factor for
Undisturbed Areas, pi 18-124 in: Present and Prospective Technology for Predicting
Sediment Yields Sources, USDA ARS-S-40.

Wischmeier, W.H., and D.D. Smith. 1965. Predicting Rainfall Erosion Losses from Cropland
East of the Rocky Mountains: Guide for selection of practices for soil and water
conservation. U.S. Department of Agriculture, Agricultural Handbook No. 282.

Wischmeier, W.H., and D.D. Smith. 1978. Predicting rainfall erosion losses - a guide to
conservation planning. The USDA Agricultural Handbook No. 537.

-------
VALUE RANGES FOR HSPF SEDIMENT EROSION AND SOLIDS WASHOFF
PARAMETERS

RANGE OF VALUES

NAME

DEFINITION

UNITS

TYPICAL

POSSIBLE

FUNCTION OF ...

COMMENT

MIN

MAX

MIN

MAX

PERLND

SED-PARM2

SMPF

Management Practice (P)
factor from USLE

none

0.0

1.0

0.0

1.0

Land use, Ag
practices

Use P factor from USLE

KRER

Coefficient in the soil
detachment equation

complex

0.15

0.45

0.05

0.75

Soils

Estimate from soil erodibility
factor (K) in USLE

JRER

Exponent in the soil
detachment equation

none

1.5

2.5

1.0

3.0

Soils, climate

Usually start with value of 2.0

AFFIX

Daily reduction in detached
sediment

per day

0.03

0.10

0.01

0.50

Soils, compaction, ag
operations

Reduces fine sediments following
tillage

COVER

Fraction land surface
protected from rainfall

none

0.0

0.90

0.0

0.98

Vegetal cover, land
use

Seasonal/monthly values often
used

NVSI

Atmospheric additions to
sediment storage

Ib/ac-dy

0.0

5.0

0.0

20.0

Deposition, activities,
etc.

Can be positive or negative

SED-PARM3

KSER

Coefficient in the sediment
washoff equation

complex

0.5

5.0

0.1

10.0

Soils, surface
conditions

Primary sediment Calibration
parameter

JSER

Exponent in the sediment
washoff equation

none

1.5

2.5

1.0

3.0

Soils, surface
conditions

Usually use value of about 2.0

KGER

Coefficient in soil matrix
scour equation

complex

0.0

0.5

0.0

10.0

Soils, evidence of
gullies

Calibration, only used if there is
evidence of gullies

JGER

Exponent in soil matrix scour
equation

none

1.0

3.0

1.0

5.0

Soils, evidence of
gullies

Usually use value of about 2.5

IMPLND

SLD - PARM2

KEIM

Coefficient in the solids
washoff equation

complex

0.5

5.0

0.1

10.0

Surface conditions,
solids charac.

Primary solids Calibration
parameter

JEIM

Exponent in the solids
washoff equation

none

1.0

2.0

1.0

3.0

Surface conditions,
solids charac.

Usually use value of about 1.8

ACCSDP

Solids accumulation rate on
the land surface

Ib/ac-dy

0.0

2.0

0.0

30.0

Land use, traffic,
human activities

Calibration, primary source of
solids from impervious areas

REMSDP

Fraction of solids removed
per day

per day

0.03

0.2

0.01

1.0

Street sweeping,
wind, traffic

Usually start with value of about
0.05, and calibrate

-------
HSPF HYDRAULIC SEDIMENT PARAMETERS AND VALUE RANGES

RANGE OF VALUES

NAME

DEFINITION

UNITS

TYPICAL

POSSIBLE

FUNCTION OF ...

COMMENT

MIN

MAX

MIN

MAX

RCHRES

SANDFG

SDFG

Indicates Method Used for
Sandload Simulation

none

Type of stream;
user experience.

1 - Toffaleti, 2 - Colby,
3 - Power Function

SED-GENPARM

BEDWID

Width of cross-section
over which HSPF will
assume bed sediment is
deposited

500

1000

Reach\

Waterbody

morphology

Constant regardless of stage,
top-width, etc

BEDWRN

Bed depth which, if
exceeded (i.e., through
deposition) will cause a
warning message to be
printed

0.5

Reach\
Waterbody
morphology, User
Needs

Only affects when warning
messages will be printed about
high bed depth/deposition.
Lakes/reservoirs will have
higher values.

POR

Porosity of the bed
(volume voids/total
volume)

none

0.3

0.6

0.25

0.9

Reach \ Sediment
Bed

Characteristics

Only affects bed depth
calculation. Can set to 0.5 if no
data are available.

SED-HYDPARM

LEN

Length of the RCHRES

miles

0.1

1.0

0.01

100

Topography,

stream

morphology

If very large lengths are
calculated, reach should be
subdivided.

DELTH

Drop in water elevation
from upstream to
downstream extremities of
the RCHRES

0.1

100

Topography,

stream

morphology

If large drops are calculated,
the reach should be subdivided
into multiple separate reaches.

DB50

Median diameter of bed
sediment (assumed
constant)

0.01

0.02

0.001

1.0

Channel bed
properties

Only used for lake shear stress
and Toffaleti/Colby methods

SAND-PM

Effective diameter of the
transported sand particles

.002

0.08

.0005

0.2

Sediment
properties

Not used in calculations. Set to
0.01 in.

Fall velocity of transported
sand particles in still water

in/sec

0.2

0.1

10.

Particle diameter
and density

Used for Toffaleti method.

RHO

Density of sand particles

g/cm3

2.2

2.7

1.5

3.0

Sediment
properties

Used for calculating bed depth.

KSAND

Coefficient in sandload
power function formula

complex

0.01

0.5

0.001

10.

Sand properties
and hydraulics

Calibration. Affects sand
concentration.

EXPSND

Exponent in sandload
power function formula

complex

1.5

3.5

1.0

6.0

Sand properties
and hydraulics

Calibration. Affects sand
scour. Usually start with 2.0

SILT-CLAY-PM

Effective diameter of silt,
or clay particles

.0002
.00001

.0025
.00015

.0001
.000005

.004
.00025

Sediment
properties

Used for calculating bed depth.

-------
w

Fall velocity of transported
silt or clay particles in still
water

in/sec

.0001

0.01

0.0

0.1

Particle
diameter and
density

Affects concentration during low
flow.

RHO

Density of silt or clay
particles

g/cm3

1.8

2.7

1.5

3.0

Sediment
properties

Used for calculating bed depth.

TAUCD*

Critical bed shear stress for
deposition

lb/ft2

0.01

0.3

0.001

1.0

Silt/clay
properties and
hydraulics

Calibration. Affects timing &
magnitude of deposition. Initial
values based on computed
shear stress.

TAUCS*

Critical bed shear stress for
scour

lb/ft2

0.05

0.5

0.01

3.0

Silt/clay
properties and
hydraulics

Calibration. Affects timing &
magnitude of scour. Initial
values based on computed
shear stress.

Erodibility coefficient

lb/ft2.d

0.01

0.001

5.0

Silt/clay
properties and
hydraulics

Calibration. Affects magnitude
of scour.

- Minimum values for lakes and other waterbodies may be 2 to 3 orders of magnitude lower than the
table ranges in order to represent reasonable sediment trapping efficiency.

-------
APPENDIX

SEDIMENT CALIBRATION PROCEDURES AND GUIDELINES
FOR WATERSHED MODELING

A. S. Donigian, Jr.

J. T. Love
AQUA TERRA Consultants
2685 Marine Way, Suite #1314
Mountain View, CA 94043

Presented at Water Environment Federation TMDL 2003 Conference,

Chicago, IL November 16-19, 2003

ABSTRACT

Sediment is one of the most difficult water quality constituents to accurately represent in current
watershed and stream models. Important aspects of sediment behavior within a watershed
system include loading and erosion sources, delivery of these eroded sediment sources to
streams, drains and other pathways, and subsequent instream transport, scour and deposition
processes.

Sediment calibration for watershed models involves numerous steps in estimating model
parameters and determining appropriate adjustments needed to insure a reasonable simulation of
the sediment sources, delivery, and transport behavior within the channel system. Rarely is there
sufficient observed local data at sufficient spatial detail to accurately calibrate all parameters for
all land uses and each stream and waterbody reach. Consequently, model users focus the
calibration on sites with observed data and review simulations in all parts of the watershed to
ensure that the model results are consistent with field observations, historical reports, and
expected behavior from past experience.

This paper explores a 'weight of evidence' approach for sediment calibration as part of overall
watershed model calibration, using both graphical and statistical measures, based on recent
experience with the U. S. EPA Hydrological Simulation Program - FORTRAN (HSPF). Model
parameterization and calibration procedures are described, using sample model results, to
demonstrate recommended graphical and statistical procedures to assess model performance for
sediment loadings, concentrations, and budgets within a watershed modeling framework.
Although the results are specific to the EPA HSPF model, the approach and procedures for
sediment calibration are applicable to other watershed models that represent sediment processes
and behavior at the watershed scale.

KEYWORDS

HSPF, erosion, sediment delivery, GIS, sediment rating curves, shear stress, TMDL, USLE

-------
INTRODUCTION

Sediment is a primary constituent of concern for many watershed assessments and Total
Maximum Daily Load (TMDL) studies being performed across the country. In addition to issues
related to sediment impacts on stream habitats, sediment is also a carrier of many other
pollutants, including metals, phosphorus, organics, and bacteria. Unlike many other pollutants,
eliminating all sediments from the stream is not a solution since that will ultimately lead to
channel scour and/or bank failures, as the stream attempts to reach a stable, dynamic equilibrium.

Sediment is also one of the most difficult water quality constituents to accurately represent in
current watershed and stream models. Important aspects of sediment behavior within a
watershed system include loading and erosion sources, delivery of these eroded sediment sources
to streams, drains and other pathways, and subsequent instream transport, scour and deposition
processes.

A 'weight-of-evidence' approach is rapidly becoming the standard practice in watershed
modeling. Model performance and calibration/validation are evaluated through qualitative and
quantitative measures, involving both graphical comparisons and statistical tests. For flow
simulations where continuous records are available, all these techniques are often employed, and
the same comparisons are performed, during both the calibration and validation phases. For
water quality constituents, including sediment, model performance is often based primarily on
visual and graphical presentations as the frequency of observed data is often inadequate for
accurate statistical measures beyond basic metrics (e.g., mean). However, consistency checks
with expected value ranges for loading rates and stream morphology and behavior are critical
when spatially distributed field data are limited.

This paper explores the 'weight of evidence' approach for sediment calibration as part of overall
watershed model calibration based on recent experience with the U. S. EPA Hydrological
Simulation Program - FORTRAN (HSPF) (Bicknell et al., 2001). Model parameterization and
calibration procedures are described, using example applications and sample model results, to
demonstrate recommended graphical and statistical procedures used to assess model performance
for sediment loadings, concentrations, and budgets within a watershed modeling framework.
Although the results are specific to the EPA HSPF model, the approach and procedures for
sediment calibration are applicable to other watershed models that represent sediment processes
and behavior at the watershed scale.

SEDIMENT CALIBRATION OVERVIEW

Sediment calibration follows the hydrologic calibration and must precede water quality
calibration. Calibration of the parameters involved in simulation of watershed sediment erosion
is more uncertain than hydrologic calibration due to less experience with sediment simulation in
different regions of the country. The process is analogous; the major sediment parameters are
modified to increase agreement between simulated and recorded monthly sediment loss and
storm event sediment removal. However, observed monthly sediment loss is often not available,
and the sediment calibration parameters are not as distinctly separated between those that affect
monthly sediment and those that control storm sediment loss. In fact, annual sediment losses are
often the result of only a few major storms during the year.

-------
Sediment calibration for watershed models involves numerous steps in estimating model
parameters then determining appropriate adjustments needed to ensure a reasonable simulation
of the sediment sources, delivery, and transport behavior within the channel system. These steps
usually include:

1. Estimating target (or expected) sediment loading rates from the landscape, often as a
function of topography, land use, and management practices

2. Calibrating the model loading rates to the target rates

3. Adjusting scour, deposition and transport parameters for the stream channel to mimic
expected behavior of the streams/waterbodies

4. Analyzing sediment bed behavior (i.e. bed depths) and transport in each channel
reach as compared to field observations

5. Analyzing overall sediment budgets for the land and stream contributions, along with
stream aggrading and degrading behavior throughout the stream network

6. Comparing simulated and observed sediment concentrations, including particle size
distribution information, and load information where available

7. Repeating steps 1 through 6 as needed to develop a reasonable overall representation
of sediment sources, delivery, and transport throughout the watershed system

Rarely is there sufficient observed local data at sufficient spatial detail to accurately calibrate all
parameters for all land uses and each stream and waterbody reach. In fact, for sediment
modeling, users are often limited to observed data for monthly or storm periods at only selected
sites within the watershed. Consequently, model users focus the calibration on those sites with
observed data, and then must review simulations in all parts of the watershed to ensure that the
model results are consistent with field observations, historical reports, and expected behavior
from past experience. This is especially critical for sediment modeling due to the extreme
dynamic behavior of sediment erosion and transport processes.

Below we journey through each of the above steps to provide model users with general guidance
and recommendations for modeling watershed-scale sediment processes in a logical and
reasonable fashion. Although the specific parameter definitions and discussions and sample
model results are based on the HSPF model, the overall procedures should prove useful to users
of other watershed scale sediment model codes.

SEDIMENT EROSION CALIBRATION

Model parameters are then adjusted so that model-calculated loadings are consistent with these
estimated 'calibration targets' and loading ranges. The model-calculated loadings are further

-------
evaluated in conjunction with the instream sediment transport calibration (discussed below) that
extend to a point in the watershed where sediment concentration and/or load data are available.
The objective is to represent the overall sediment behavior of the watershed, with knowledge of
the morphological characteristics of the stream (i.e. aggrading or degrading behavior), using
sediment loading rates that are consistent with the calibration targets and modeled
concentrations that provide a reasonable match with instream sediment data.

Step 1: Estimating Sediment Loadings from the Landscape

Sediment concentrations measured at a particular gage reflect the combined affects of nonpoint
source contributions from multiple land uses, any point sources upstream from the gage, and
instream processes (e.g., deposition, scour, bank erosion). Consequently, the calibration target
sediment loading rates need to be developed to help guide the calibration effort and ensure that
the simulated erosional rates from each land use category are reasonable, and thereby provide a
sound basis for calibrating the instream processes.

Erosion is primarily a function of the amount of soil exposed directly to rainfall and surface
runoff, which in turn is affected by rainfall, land cover, land slope, soil disturbance, and transport
properties of the soil. The USLE is an empirical equation commonly used to estimate erosional
rates as a function of these factors. The USLE formula is expressed as follows:

A = R* K* L* S* C* P

A = annual soil loss in tons per acre per year
R = rainfall erosivity factor
K = soil erodibility factor
L = slope length factor
S = slope gradient factor
C = cover management factor
P = erosion control practice factor

The R factor is typically obtained from a national or regional isoerodent map, readily available
in many soil engineering handbooks (e.g. Renard et al, 1997), and accounts for the amount and
intensity of rainfall and runoff typical of a region.

The K factor is a measure of the susceptibility of soil particles to detachment and transport by
rainfall and runoff. Texture is the principal factor affecting K, but structure, organic matter and
permeability also contribute. Determining K values can be performed either from handbook
tables, or acquisition of accurate geo-spatial soils data from the U.S. Department of Agriculture
(USDA), Natural Resource Conservation Service (NRCS) if a GIS approach is employed. The
most common forms of these data are the STATSGO and SSURGO databases and/or GIS
coverages.

The L factor is very closely associated with the S factor, where S is the slope gradient factor and
the L is the length of that slope. The USLE was created to predict soil erosion delivered to the
base of a 22-meter agricultural plot with a uniform slope of 9 percent. The S and L factors are
typically combined, defined as the topographic factor LS, to account for site specific conditions

-------
relative to the standard plot.

The Cfactor is the crop/vegetation and management factor. It is used to determine the relative
effectiveness of soil and crop management system or vegetation in terms of preventing soil loss.
The C factor is a ratio comparing the soil loss from land under a specific crop and management
system to the corresponding loss from continuously fallow and tilled land.

The,Pfactor is the support practice factor and reflects the effects of practices that will reduce the
amount and rate of the water runoff and thus reduce the amount of erosion. The factor represents
the ratio of soil loss by a support practice to that of straight-row farming up and down the slope.
The most commonly used supporting cropland practices are cross slope cultivation, contour
farming, and strip cropping. In cases where the conservation practice factor is not relevant, it is
set as 1.0 for all areas, which does not negatively or positively influence the output of the model.

As the USLE was developed at a field scale, depositional processes that occur in overland flow
prior to reaching a stream channel are not included. Therefore, it is necessary to reduce the gross
erosion by a fraction. This fraction or portion of sediment that is available for delivery is
referred to as the Sediment Delivery Ratio (SDR). This ratio is then multiplied by the predicted
or gross erosion rate to estimate the percent of eroded material to reach a specific point or
location (e.g., outlet, waterbody, channel). There is no generally accepted procedure to estimate
the SDR, which is affected by numerous factors including sediment source, texture, nearness to
the stream, channel density, basin area, slope, land use/cover, and rainfall-runoff factors;
however, several empirical formulas exist (e.g. see Greenfield et al., 2002).

Under EPA funding, AQUA TERRA Consultants developed a Visual Basic spreadsheet tool
named TMDL USLE (Hummel et al., 2000) for use in sediment associated TMDL studies
(http://www.epa.gov/ceampubl/swater/usle). This spreadsheet is useful for estimating the
expected relative magnitude of land surface sediment loadings (tons per year) from different land
use types within a watershed, and thereby can be used to develop sediment calibration targets for
watershed models. Maps, recommended value tables for USLE factors, and other information
useful in deriving appropriate values for the USLE and delivery ratios are provided, to the extent
that it is practical, throughout the U.S. The tool includes an on-line tutorial and active links to
Internet web sites containing supplemental information that can assist users in evaluating USLE
factors.

Figure 1 shows the computation screen for the TMDL USLE program, with highlighted
comment 'bubbles' identifying the source of values and information in each cell.

As an alternative to a spreadsheet based approach, using a GIS platform allows the USLE to be
applied on a cell-by-cell basis, using watershed specific information, for a more spatially
accurate use of the equation and model land use specific estimates of erosional rates. The USLE
can be applied in a grid-based GIS environment where map algebra can be performed with the
GIS layer values.

-------
Control buttons:

hold cursor on button for tool tip:
use beginner's "help" for
-.detailedbutton functions

Provide by usa- ^
knowledge of values
derived for previous
local USLE appli-
cations or by using
estimation aid in
on-line "help"

Updates each time
str earn loa din g for
new segment is
computed

Provide by user
knowledge of local
sediment b ehavior or
by estimation aid sin
on-line "help"

Total Edge of Stream Sediment Load (tons/yr): h 36OI '

Figure 1 - TMDL USLE tool

Typically, in a GIS environment the S and L factors are combined together and defined as the
topographic factor (LS) with the slope length normalized to a standard plot length of 22 meters
and the slope normalized to 9 percent. Numerous empirical formulas exist such as the one
below.

LS = [0.065 + 0.0456(slope) + 0.006541 (slope)2] * [resolution / normlength] "

slope = slope steepness (%)

resolution = cell resolution (meters or feet)
normlength = 72.5 ft or 22.1 meters

n = function of slope (see table below)

slope

< 1

1 < slope < 3

3 < slope < 5

> 5

0.2

0.3

0.4

0.5

The Kfactor can be obtained from a descriptor or attribute, referred to as 'kffaci' within the
STATSGO or SSURGO databases. The attribute 'kffaci' defines the soil erodibility factor that is
fragment free for use in the USLE.

Within a GIS, the SDR can be calculated in a manner to try and account for all the
aforementioned controlling factors, or in a simplified manner based on the drainage area of the
channel segment as defined by the model setup. This simple approach is referred to as a
watershed area-based method. The equation below was converted from a curve presented in the
National Engineering Handbook produced by the Soil Conservation Service in 1983 (USDA-
NRCS, 1983).

SDR = 0.417762 * A 0 134958 _ 0.127097

A = drainage area (sq. miles)

-------
Numerous additional empirical formulas exist, including formulas and tables provided within the
TMDL USLE tool. Ultimately, once the gross erosional rates are adjusted by the SDR, it is
possible to use the GIS to summarize the range of erosional rates on a model land use specific
basis. Figure 2 graphically depicts the process of taking the input data (i.e., DEM, soils, land
use/cover), calculating USLE factors, and developing estimates of the erosional rates.

DEM

SOILS

LAND USE \ COVER

GROSS EROSION

SEDIMENT DELIVERY

X SDR =

Figure 2 - GIS Framework for USLE

Calibration targets developed from the TMDL USLE spreadsheet, a GIS-based approach, or
other procedures should be used only to define approximate ranges of loading rates to help guide
the calibration for the watershed or region of concern. There will often be extreme variation in
the calculated rates from year to year, and site to site, and users should only expect that the
model rates will be consistent with the targets, not necessarily equal to them. Table 1 shows
typical ranges of sediment loading rates for various land categories.

Table 1 - Typical Ranges of Expected Erosion Rates

Tons/ac

Tonnes/ha

Forest

0.05-0.4

0.1 -0.9

Pasture

0.3 -1.5

0.7-3.4

Conventional

o
o

2.2 - 15.7 (crop dependent)

Tillage

Conservation

o
^r

LT)

1.1 - 9.0 (crop dependent)

Tillage

Hay

0.3 -1.8

r-
o

Urban

0.2 -1.0

0.4-2.2

Highly Erodible

> ~ 15.0

> ~ 33.6

Land

-------
Step 2: Sediment Erosion Calibration

Each of the calibration steps identified in the overall procedures involve first a parameterization
component followed by the actual calibration, or parameter adjustment, component to improve
agreement between model values and various field observations. Clearly, the specific parameters
to adjust for soil erosion calibration will depend on the specific model being used. In HSPF, the
erosion process on pervious land areas is represented as the net result of detachment of soil
particles by raindrop impact on the land surface, and then subsequent transport of these fine
particles by overland flow. On impervious surfaces (e.g. parking lots, driveways), soil splash
by raindrop impact is neglected and solids washoff is often controlled by the rate of
accumulation of solid materials. The primary sediment erosion solids parameters are as follows:

KRER - Coefficient in soil detachment equation (pervious areas)

KSER - Coefficient in sediment washoff equation (pervious areas)

KEIM - Coefficient in impervious area solids washoff equation

ACCSDP - Accumulation rate of solids on impervious surfaces

Although a number of additional parameters are involved in sediment erosion and solids
calibration, such as those related to vegetal cover, agricultural practices, rainfall and overland
flow intensity, etc., KRER and KSER are the primary ones controlling sediment loading rates.
KRER is usually estimated as equal to the erodibility factor, K, in the USLE (noted above), and
then adjusted in calibration, while KSER is primarily evaluated through calibration and past
experience. For impervious surfaces, the rate of washoff is controlled by the KEIM parameter,
but the net washoff is most often limited by the accumulation rate, ACCSDP. Table 2 lists the
sediment and solids washoff parameters in HSPF, along with typical and possible minimum and
maximum ranges based on application experience over the past 20 years. In addition, the
HSPFParm database (U. S. EPA, 1999) provides calibrated parameter values for numerous
watersheds across the US.

In general, sediment calibration involves the development of an approximate equilibrium or
balance between the accumulation and generation of sediment particles on one hand and the
washoff or transport of sediment on the other hand. Thus, the accumulated sediment on the land
surface should not be continually increasing or decreasing throughout the calibration period.
Extended dry periods will produce increases in surface sediment accumulations, and extended
wet periods will produce decreases. However, the overall trend should be relatively stable from
year to year. This equilibrium must be developed on both pervious and impervious surfaces, and
must exist in conjunction with the accurate simulation of monthly and storm event sediment loss,
depending on the data available for calibration. Additional guidance in sediment erosion
calibration is provided in the HSPF Application Guide (Donigian et al., 1984).

-------
Table 2 - Value Ranges for HSPF Sediment Erosion and Solids Washoff Parameters

RANGE OF VALUES

NAME

DEFINITION

UNITS

TYPICAL

POSSIBLE

FUNCTION OF ...

COMMENT

MIN

MAX

MIN

MAX

SED-PARM2

SMPF

Management Practice (P)
factor from USLE

none

0.0

1.0

0.0

1.0

Land use, Ag
practices

Use P factor from USLE

KRER

Coefficient in the soil
detachment equation

complex

0.15

0.45

0.05

0.75

Soils

Estimate from soil erodibility
factor (K) in USLE

JRER

Exponent in the soil
detachment equation

none

1.5

2.5

1.0

3.0

Soils, climate

Usually start with value of 2.0

AFFIX

Daily reduction in detached
sediment

per day

0.03

0.10

0.01

0.50

Soils, compaction, ag
operations

Reduces fine sediments following
tillage

COVER

Fraction land surface
protected from rainfall

none

0.0

0.90

0.0

0.98

Vegetal cover, land
use

Seasonal/monthly values often
used

NVSI

Atmospheric additions to
sediment storage

Ib/ac-dy

0.0

5.0

0.0

20.0

Deposition, activities,
etc.

Can be positive or negative

SED-PARM3

KSER

Coefficient in the sediment
washoff equation

complex

0.5

5.0

0.1

10.0

Soils, surface
conditions

Primary sediment Calibration
parameter

JSER

Exponent in the sediment
washoff equation

none

1.5

2.5

1.0

3.0

Soils, surface
conditions

Usually use value of about 2.0

KGER

Coefficient in soil matrix
scour equation

complex

0.0

0.5

0.0

10.0

Soils, evidence of
gullies

Calibration, only used if there is
evidence of gullies

JGER

Exponent in soil matrix scour
equation

none

1.0

3.0

1.0

5.0

Soils, evidence of
gullies

Usually use value of about 2.5

SLD - PARM2

KEIM

Coefficient in the solids
washoff equation

complex

0.5

5.0

0.1

10.0

Surface conditions,
solids charac.

Primary solids Calibration
parameter

JEIM

Exponent in the solids
washoff equation

none

1.0

2.0

1.0

3.0

Surface conditions,
solids charac.

Usually use value of about 1.8

ACCSDP

Solids accumulation rate on
the land surface

Ib/ac-dy

0.0

2.0

0.0

30.0

Land use, traffic,
human activities

Calibration, primary source of
solids from impervious areas

REMSDP

Fraction of solids removed
per day

per day

0.03

0.2

0.01

1.0

Street sweeping,
wind, traffic

Usually start with value of about
0.05, and calibrate

-------
INSTREAM SEDIMENT TRANSPORT CALIBRATION

Parameterization of Instream Sediment Transport Processes

Once the sediment loading rates are calibrated to provide the expected input to the stream
channel, the sediment calibration then focuses on the channel processes of deposition, scour, and
transport that determine both the total sediment load and the outflow sediment concentrations to
be compared with observations. In practice, instream calibration involves steps 3, 4 and 5 as
listed and discussed above; these steps involve both initial parameterization, to establish initial
parameter values, and a subsequent adjustment process. For HSPF, the initial parameterization
includes the following:

• Divide input sediment loads into appropriate size fractions

• Estimate initial parameter values and storages for all reaches

• Run HSPF to calculate shear stress in each reach to estimate critical scour and
deposition values

Since the sediment load from the land surface is calculated in HSPF as a total input, it must be
divided into sand, silt, and clay fractions for simulation of instream processes. Each sediment
size fraction is simulated separately, and storages of each size are maintained for both the water
column (i.e. suspended sediment) and the bed.

Fractionating the Eroded Material

The eroded material is fractionated into sand, silt, and clay prior to entering a model reach using
available soils information; typically, a single fractionation scheme is used for all reaches.
However, if the resolution of the data and spatial diversity of the soils warrant alternate schemes,
it is possible for each reach to use separate fractions. The fractions should reflect the relative
percent of the surface material (i.e., sand, silt, clay) available for erosion in the surrounding
watershed, but also should include an enrichment factor of silt and clay to represent the
likelihood of these finer materials reaching the channel. Thus, the sand particles are more likely
to be deposited in the overland flow plane, in swales, ditches, depressions, etc. and therefore the
sand would be somewhat transport limited, compared to the silt and clay. For example, if
surface soils demonstrate a 40/50/10 distribution for sand/silt/clay, the fractionation for input to
the reach might be 15/55/30. Investigation of the bed material composition will also help to
provide insight into appropriate fractionation values.

Estimate Initial Parameter Values And Storages For All Reaches

For HSPF, initial sediment parameters, such as particle diameter, particle density, settling
velocity, bed depth and composition, and beginning calibration parameter values can be
evaluated from local/regional data, past experience, handbook values, etc., and then adjusted
based on available site specific data and calibration. Bed composition data are especially
important so that the model results can be adjusted to reflect localized aggredation (deposition)
or degradation (scour) conditions within the stream system.

In HSPF, the value of bed depth represents the amount of material (calculated from input values
for bed width and porosity) that can be scoured from the stream reach; in effect it provides a

-------
limit so that the model will inform the user, through a warning message, when the channel has
been completely scoured so that the user can make appropriate parameter changes if needed. We
often set initial bed depths to 2.0 to 5.0 feet for the natural (i.e. non-channelized) stream
segments, and 0.5 to .05 feet for the channelized segments to allow a reasonable amount of
scour in the upstream natural channel and limit the scour to scattered localized deposits in
channelized sections.

Setting Initial Critical Scour and Depositional Shear Stresses

In HSPF, the transport of the sand (non-cohesive) fraction is commonly calculated as a power
function of the average velocity in the channel reach in each timestep. This transport capacity is
compared to the available inflow and storage of sand particles; the bed is scoured if there is
excess capacity to be satisfied, and sand is deposited if the transport capacity is less than the
available sand in suspension within the channel reach. For the silt and clay (cohesive)
fractions, shear stress calculations are performed by the hydraulics (HYDR) module and are
compared to user-defined critical, or threshold, values for deposition and scour for each size.
When the shear stress for a timestep is greater than the critical value for scour, the bed is scoured
at a user-defined erodibility rate; when the shear stress is less than the critical deposition value,
the silt or clay fraction deposits at a settling rate input by the user for each size. If the calculated
shear stress falls between the critical scour and deposition values, the suspended material is
transported through the reach. After all scour and/or deposition fluxes have been determined, the
bed and water column storages are updated and outflow concentrations and fluxes are calculated
for each timestep. These simulations are performed by the SEDTRN module in HSPF, complete
details of which are provided in the HSPF User Manual (Bicknell et al., 2001).

In HSPF, if the model reach being simulated is a stream or river, the bed shear stress is
determined as a function of the slope and hydraulic radius of the reach, as follows:

TAU = SLOPE*GAM*HRAD

where:

TAU = stream bed shear stress (Ib/ft2 or kg/m2)

SLOPE = slope of the RCHRES (-)

GAM = unit weight, or density, of water (62.4 Ib/ft3 or 1000 kg/m3)

HRAD = hydraulic radius (ft or m)

The hydraulic radius is calculated as a function of average water depth (AVDEP) and mean top
width (TWID):

HRAD = (AVDEP*TWI D)/(2.*AVDEP + TWID)

Average depth is computed as: AVDEP = VOL/SAREA

The mean top width is found using: TWID = SAREA/LEN

where:

LEN = length of the RCHRES, supplied by the user

-------
If the stream reach is a lake, alternative methods are used for the shear calculations (see Users
Manual for details).

In HSPF, the hydraulic characteristics of a stream reach are represented by a function table
(FTABLE) that includes the relationships between stage, storage (volume), surface area, and
discharge. From the equations shown above, it is clear that the accuracy of the FTABLE for a
specific reach will be a critical factor in adequately representing the hydraulic radius and
subsequent shear values, as a function of the stage, or depth of flow. This is especially evident
for simulations of flood flows that exceed bankfull discharges; improper extension of the
FTABLES can lead to erroneous shear and scour conditions during high flow events, and have
major impacts on the model simulations for those events.

As part of the sediment parameterization, the model is run with the initial parameter estimates
and shear stress values are output for each stream reach. For the silt and clay size particles, the
critical shear stress parameters (one for scour and one for deposition) for each size are adjusted
so that the model calculates scour during high flow events, deposition and settling during low
flow periods, and transport with neither scour nor settling for moderate flow rates; this is shown
schematically in Figure 3. In general, the values are set so that scour of clay occurs at lower
shear values than for silt (i.e. clay scours before silt), and deposition of silt occurs at higher shear
values than clay (i.e. silt deposits before clay).

-------
Reach Tau Plot

0 1 1 1 1 1 0

1996 1997 1998 1999 2000

Year

Figure 4 - Example calculations for setting critical shear values for HSPF

Step 3: Adjust Instream Scour, Deposition, and Transport parameters

Step 4: Analyze Bed Behavior and Transport Fluxes

Step 5: Analyze Overall Sediment Budgets and Stream Behavior

These 3 steps are listed together as they normally are performed while reviewing the same
tabulations of model results; bed behavior and sediment budgets need to be reviewed to establish
the basis for parameter adjustments on a reach-by-reach basis. Table 3 shows an example
tabulation of reach-by-reach model results for a small eastern US watershed. The watershed
includes 2 tributary reaches and 6 mainstem reaches, and the tabulations include sediment
erosion (nonpoint) loads, point loads, upstream and total inflow loads, total outflow loads, and
both cumulative and reach trapping efficiencies. For example, note that the tributaries
demonstrate net scour behavior (deposition/scour column values are negative), while the
mainstem is depositional throughout. This information can be compared with historical accounts
or field observations to identify the 'expected' behavior for those stream segments. If this
information is contrary to the model representation, i.e., the model simulates deposition when the
reach is primarily being scoured, reach parameters and/or inflows need to be adjusted to correct

-------
the simulated behavior.

Table 3 - Example Tabulation of Stream Sediment Fluxes and Behavior

Nonpoint

Point
Source

Upstream
In

Total
Inflow

Outflow

Deposit (+)
Scour (-)

Cumulative
Point/NonPt

Cumulative
Trapping
Efficiency

Reach
Trapping
Efficiency

Reach
Segment

(tons)

(%)

Mainstem 1

212.5

107.4

6,453.7

6,785.3

6,186.3

599.7

10,566.9

41.5

8.8

Mainstem 2

68.8

0.0

6,186.3

6,255.0

5,384.8

870.6

10,635.7

49.4

13.9

Tributary 1

102.4

0.0

102.2

125.0

-22.7

102.2

-22.0

Mainstem 3

5.8

0.0

5,509.8

5,515.6

4,916.3

599.9

10,744.0

54.2

10.9

Tributary 2

281.1

0.0

280.5

352.6

-72.1

280.5

-25.5

Mainstem 4

215.4

0.0

5,268.9

5,483.9

4,269.8

1,215.1

11,240.4

62.0

22.1

Mainstem 5

54.1

0.0

4,269.8

4,323.8

3,507.1

826.2

11,294.5

68.9

18.9

Mainstem 6

93.9

0.0

3,507.1

3,600.8

2,190.8

1,421.3

11,388.4

80.8

39.2

Table 4 shows the corresponding detailed behavior of bed depth and sediment fractions in
selected reaches within the watershed (Tributary 1 and Mainstem reaches 2 and 3). The
tabulations in Table 4 include, annual inflow loads, outflow loads, and deposition/scour in the
reach, along with the composition of these loads and the bed behavior and composition
throughout the simulation period. Field observations of bed depth changes, expected deposition
rates, bed sediment composition fractions, etc. can be used to assess the validity of the model
results and identify needed changes.

These results demonstrate the types of analyses performed as part of the sediment calibration
effort. In this example, sand comprises a small fraction of the total sediment concentration and
discharge, and thus the sand parameters are set to values to reflect this small contribution. In
other watersheds, the non-cohesive (sand) fractions may be more critical and thus require greater
focus and calibration effort.

The primary focus of this example calibration is the silt and clay parameters. As noted above,
the shear stress values are adjusted so that scour occurs during storm periods and deposition
occurs at low flows. Once the timing of scour and deposition processes is correct, the rate of
scour (i.e. erodibility factor in the model) is adjusted in an attempt to match either the expected
behavior within each reach, from review of the type of information shown in Tables 3 and 4,
and/or the observed concentrations (discussed below in the next step). During high flow periods,
the amount of scour is adjusted with an erodibility factor for each reach that controls the rate of
scour whenever the actual shear stress is greater than the critical shear stress value for scour.

The need to analyze the model simulations on a reach-by-reach basis is mandated by the extreme
variability in sediment processes. If upstream reaches are depositing more than expected, then
inflows to downstream reaches will be less than what really occurs, requiring parameter
adjustments that may not be reasonable for the downstream reaches. The opposite would occur
if upstream reaches are eroding much more than expected; inflows to downstream reaches will
be too large, resulting in more deposition than would be expected. If the reach parameters are set
so that deposition does not occur, then the upstream eroded load will be transported and will
subsequently impact other downstream reaches. Thus, an upstream to downstream analysis, on a
reach-by-reach basis, is required to adequately assess model simulations.

-------
Table 4 - Exam

)le Bed and Stream Reach Sediment Simulations

Mainstem 2

Final

Average

Arithmetic

I nit Val

1995

1996

1997

1998

1999

2000

Fraction

Average

Bed Depth (ft)

2.00

2.10

2.20

2.30

2.20

Bed Storage (tons)

Sand

0.79

34,684.7

34,775.7

34,795.8

34,805.4

34,828.3

34,840.5

0.70

0.73

34,788.4

Silt

0.15

7,623.7

8,780.0

9,310.2

9,867.1

10,581.5

11,573.3

0.23

0.20

9,622.6

Clay

0.07

3,112.9

3,165.8

3,181.9

3,201.7

3,222.9

3,255.5

0.07

3,190.1

Total

45,421.3

46,721.6

47,287.9

47,874.2

48,632.8

49,669.2

47,601.2

1 nf low (tons)

Sand

152.6

273.5

115.0

133.0

130.6

219.8

0.03

170.7

Silt

4,705.1

5,024.8

2,503.7

2,571.7

3,357.7

4,437.0

0.61

0.60

3,766.7

Clay

3,152.5

3,270.2

1,354.6

1,528.3

1,939.2

2,638.8

0.36

0.37

2,313.9

Total

8,010.2

8,568.5

3,973.3

4,232.9

5,427.4

7,295.6

6,251.3

Dep(+)/Scour(-) (tons)

Sand

4.7

-0.4

22.9

12.1

25.6

14.9

13.3

Silt

1,038.4

1,135.4

530.0

556.7

714.9

991.9

827.9

Clay

39.8

44.4

16.0

19.8

21.3

32.5

29.0

Total

1,082.8

1,179.4

568.9

588.5

761.8

1,039.4

870.1

Outflow (tons)

Sand

148.0

273.8

92.2

120.9

104.9

204.9

0.03

157.5

Silt

3,668.1

3,889.2

1,974.0

2,015.0

2,642.8

3,445.1

0.55

2,939.0

Clay

3,113.9

3,225.5

1,339.0

1,508.6

1,917.8

2,606.2

0.42

0.43

2,285.2

Total

6,929.9

7,388.5

3,405.2

3,644.4

4,665.6

6,256.2

5,381.6

Tributary 1

Final

Average

Arithmetic

I nit Val

1995

1996

1997

1998

1999

2000

Fraction

Average

Bed Depth (ft)

2.00

Bed Storage (tons)

Sand

0.9

28,369.1

28,440.4

28,434.4

28,415.9

28,407.7

28,417.6

0.90

28,414.2

Silt

0.05

1,569.4

1,563.8

1,561.0

1,559.1

1,554.2

1,544.3

0.05

1,558.6

Clay

0.05

1,563.0

1,549.4

1,544.4

1,538.6

1,529.9

1,513.4

0.05

1,539.8

Total

31,501.5

31,553.6

31,539.8

31,513.6

31,491.9

31,475.4

31,512.6

I nf low (tons)

Sand

32.1

49.3

24.3

14.9

29.2

64.6

0.35

35.8

Silt

45.9

70.5

34.8

21.4

41.8

92.3

0.50

51.1

Clay

13.8

21.2

10.4

6.4

12.5

27.7

0.15

15.3

Total

91.8

141.0

69.5

42.7

83.6

184.6

102.2

Dep(+)/Scour(-) (tons)

Sand

-0.7

-6.8

-7.3

-19.3

-8.7

9.4

-5.6

Silt

-6.6

-10.0

-2.8

-2.0

-4.8

-9.9

-6.0

Clay

-13.1

-17.8

-5.0

-5.7

-8.7

-16.5

-11.2

Total

-20.4

-34.6

-15.1

-27.0

-22.2

-17.0

-22.7

Outflow (tons)

Sand

32.8

56.1

31.6

34.3

37.9

55.2

0.27

0.33

41.3

Silt

52.6

80.5

37.5

23.3

46.6

102.2

0.51

0.46

57.1

Clay

27.0

39.0

15.5

12.2

21.2

44.2

0.22

0.21

26.5

Total

112.4

175.6

84.6

69.7

105.7

201.6

124.9

Mainstem 3

Final

Average

Arithmetic

I nit Val

1995

1996

1997

1998

1999

2000

Fraction

Average

Bed Depth (ft)

2.00

2.10

2.20

2.10

Bed Storage (tons)

Sand

0.79

27,713.8

28,089.4

28,204.8

28,346.9

28,480.0

28,720.2

0.74

0.75

28,259.2

Silt

0.15

5,710.9

6,354.4

6,552.4

6,789.1

7,078.2

7,578.0

0.19

0.18

6,677.2

Clay

0.07

2,481.1

2,537.1

2,547.4

2,564.5

2,582.8

2,617.0

0.07

2,555.0

Total

35,905.8

36,981.0

37,304.6

37,700.6

38,141.1

38,915.2

37,491.4

I nf low (tons)

Sand

182.5

332.6

125.4

156.1

144.5

263.5

0.04

200.8

Silt

3,723.3

3,973.5

2,013.7

2,039.7

2,691.8

3,552.2

0.55

0.54

2,999.0

Clay

3,141.7

3,265.6

1,355.1

1,521.2

1,939.8

2,651.9

0.41

0.42

2,312.6

Total

7,047.5

7,571.7

3,494.2

3,716.9

4,776.1

6,467.7

5,512.3

Dep(+)/Scour(-) (tons)

Sand

164.8

300.0

116.0

142.4

133.3

240.1

182.8

Silt

480.4

627.7

198.1

236.6

289.1

499.6

388.6

Clay

40.0

49.2

10.2

17.3

18.3

34.2

28.2

Total

685.2

976.9

324.4

396.3

440.7

773.9

599.6

Outflow (tons)

Sand

17.7

32.6

9.3

13.7

11.2

23.4

0.00

18.0

Silt

3,244.6

3,345.3

1,816.1

1,803.0

2,402.7

3,052.6

0.54

0.53

2,610.7

Clay

3,103.1

3,215.9

1,345.6

1,503.9

1,921.4

2,617.7

0.46

0.47

2,284.6

Total

6,365.4

6,593.8

3,171.1

3,320.6

4,335.3

5,693.8

4,913.3

-------
Step 6: Compare Results with Available Data

The remaining step in the calibration procedure is to compare model simulations of
concentrations and loads to available observed data. In many cases, this may be limited to event
mean concentrations of total suspended solids (TSS) for selected storm events and nonstorm
(baseflow) periods, or pollutographs of TSS concentrations throughout a few events. However,
other types of comparisons are also possible, such as load estimates and sediment rating curves;
each of these is discussed below.

Figure 5 shows an example of the conventional storm event comparison for TSS for the same
small eastern US watershed. Clearly, such comparisons need to be made for as many storm
events as there are available data. Figure 5 shows a very good simulation for both flow and TSS
concentrations; most simulations will not be this good, and will show large variations from storm
to storm. Even with this storm, there are inconsistencies demonstrated; the simulated flow peak
is higher than the observed and precedes it, identifying a possible time lag in the peak that is not
well represented in the model. In addition, the flow peak is about 25% higher than observed,
whereas the TSS peak is only about 10% higher. Both of these differences may be entirely
acceptable, considering the uncertainty in the observations and needed accuracy of the model,
but model results need to be viewed with a critical eye toward demanding consistent behavior,
i.e. if flows are over-simulated, TSS should be over-simulated, and vise-versa.

319
£ 255

J 192
E 128
64
0

" oOO

..I Q

i O \ o

12:00 AM

l o

i j

!.. u

- ic

1 1 o
1 1

1 I

1 I o

' 1

' \

' ®

'°

! ,°

yvi

c\ C)A O

Wo. °

V - 2° OOooOOOQ

Day 1

Day 2

Figure 5 - Example comparison of simulated and observed storm flow and TSS
concentrations

-------
Figure 6 shows an annual plot of simulated and estimated sediment loads for the watershed
outlet. In this case the 'Load Estimates' were extrapolations from available TSS data, and were
not the results of continuous, or even daily sampling. So this comparison is really between two
models: a simulation model (HSPF) and a regression model. However, even this type of
comparison is useful, recognizing that differences do not necessarily detract from the validity or
utility of the simulation model. This is simply one additional type of comparison that can be
included in the weight-of-evidence approach to sediment calibration. The average annual values
in Figure 6 indicate a very good simulation even though there are significant differences year to
year. For sediment modeling these types of differences are to be expected, since, as noted
earlier, sediment is one of the most difficult water quality constituents to model accurately.

Figure 6 - Example comparison of simulated and estimated annual TSS loads

Figure 7 shows an example of comparing sediment rating curves, simulated and observed, for a
single site at the watershed outlet. These curves essentially demonstrate the relationship between
flow rate and sediment concentrations, and the concept in this comparison is to evaluate whether
the model and the data demonstrate similar relationships. The top curve shows the flow versus
load relationship, corresponding to the bottom curve of sediment concentration versus flow.
Regression lines have been fitted to both the data and model results, and are shown on the
curves. The log scale is used, as is typical for, and usually required for sediment rating analyses.

-------
Figure 7 - Example of observed and simulated sediment rating curves

It should be noted that the regression curves are fitted to a subset of the observed and simulated
values; the solid dots were excluded from the analyses, corresponding to TSS concentration
values less than 1.0 mg/1. The rationale for this was as follows:

• The observations were performed primarily at moderate to high flow conditions,
so very low concentrations would not be well represented.

• The HSPF model employs a relatively gross channel representation, with long
reach lengths, that tends to eliminate localized turbulence and scour conditions
that would likely contribute to under-simulating the low concentrations.

• The extremely low concentrations contribute to a small fraction of the total annual
sediment load, approximately less than 3 to 5 percent of the annual load in this
watershed.

The overall results in Figure 7 show a relatively good simulation of the flow-sediment
relationship demonstrated in the sediment rating curves. Both the range of concentration and
flow values, and the slopes of the regression lines, demonstrate consistency between the model
and the observed data. If large consistent differences existed, it would justify continued

-------
calibration efforts to minimize such differences.

Step 7: Repeat Calibration Steps, as Needed, To Improve Sediment Representation

Modeling tends to be a circular, or iterative, process. For sediment calibration, Steps 3 through 6
often need to be repeated until all the components of the calibration exercise are in reasonable
balance. In some cases, the process may need to reconsider the target loading rates developed in
Step 1, and then re-calibrate the model rates. This might occur if the surface loadings appear to
dominate unrealistically the overall watershed simulation results.

The iteration process doesn't require that every comparison be brought to the same level of
agreement, but only that the entire process be repeated until the entire 'weight-of-evidence' from
the simulations indicates either that the model is 'as good as it can be' or that it can not meet the
specific needs of the watershed assessment. This should produce sufficient evidence that the
model is either acceptable for the intended purpose, a recommendation that the model or data
input improvements may be needed, or that a different model should be considered.

CLOSURE

This paper explores the 'weight of evidence' approach for sediment calibration as part of overall
watershed model calibration based on recent experience with the U. S. EPA HSPF. The steps in
the overall sediment calibration process are identified and discussed, along with specific issues
related to model parameterization and calibration procedures. Using example applications and
sample model results, we demonstrate recommended graphical and statistical procedures used to
assess model performance for sediment loadings, concentrations, and budgets within a watershed
modeling framework. Although the results are specific to the EPA HSPF model, the approach
and procedures for sediment calibration are applicable to other watershed models that represent
sediment processes and behavior at the watershed scale. Although sediment calibration remains
one of the most difficult components of watershed-scale water quality assessments, it is hoped
that the procedures outlined herein will provide some guidance and assistance to model users
who attempt this often daunting task.

REFERENCES

Bicknell, B.R., J.C. Imhoff, J.L. Kittle Jr., A.S. Donigian, Jr., T.H. Jobes, and R.C. Johanson.
2001. Hydrological Simulation Program - FORTRAN, User's Manual for Version 12. U.S.
EPA, National Exposure Research Laboratory, Athens, GA.

Donigian, A.S. Jr. 2002. Watershed Model Calibration and Validation: The HSPF Experience.
WEF National TMDL Science and Policy 2002, November 13-16, 2002. Phoenix, AZ. WEF
Specialty Conference Proceedings on CD-ROM.

Donigian, A.S. Jr., J.C. Imhoff, B.R. Bicknell and J.L. Kittle. 1984. Application Guide for
Hydrological Simulation Program - Fortran (HSPF), prepared for U.S. EPA,
EPA-600/3-84-065, Environmental Research Laboratory, Athens, GA.

-------
Greenfield, J., T. Dai, and H. B. Manguerra. 2002. Watershed Modeling Extensions of the
Watershed Characterization System. WEF National TMDL Science and Policy 2002,
November 13-16, 2002. Phoenix, AZ. WEF Specialty Conference Proceedings on CD-ROM.

Hummel, P.R., J.C. Imhoff, R. Dusenbury and M. Gray. 2000. TMDL USLE, A Practical Tool
for Estimating Diffuse Sediment Source Loads within a Watershed Context. Prepared for:
U.S. EPA National Exposure Research Laboratory, Athens, GA.
(http://www.epa.gov/ceampubl/swater/usle/index.htm).

U.S. Department of Agriculture Soil Conservation Service. 1983. Sedimentation. Section 3,
Chapter 6. National Engineering Handbook.

USDA-NRCS. 1983. Sediment sources, yields, and delivery ratios. Chapter 6 in National
Engineering Handbook, Section 3, Sedimentation. U.S. Department Agriculture,

Natural Resources Conservation Service formerly Soil Conservation Service (SCS), 6.2-6.19.
U.S. Government Printing Office. Washington, D.C.

U. S. EPA. 1999. HSPFParm: An Interactive Database of HSPF Model Parameters, Version 1.0.
EPA-823-R-99-004. U. S. Environmental Protection Agency, Office of Water, Washington,
DC.

Wischmeier, W.H., and D.D. Smith. 1978. Predicting rainfall erosion losses - a guide to
conservation planning. The USDA Agricultural Handbook No. 537.

-------