Environmental Modelling and Software xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Environmental Modelling & Software

journal homepage: www.elsevier.com/locate/envsoft

Modeling framework for simulating concentrations of solute chemicals,

nanoparticles, and solids in surface waters and sediments: WASP8 Advanced
Toxicant Module
Christopher D. Knightesa,∗, Robert B. Ambrose Jr.a, Brian Avanta,c, Yanlai Hana,c, Brad Acreya,c,
Dermont C. Boucharda, Richard Zeppa, Tim Woolb
a
USEPA Office of Research and Development, National Exposure Research Laboratory, Athens, GA, 30605, United States
b
USEPA Water Quality Planning Branch, Data and Information Analysis Section Region 4, Atlanta, GA, 30303, United States
c
Oak Ridge Institute for Science and Education, United States

A R T I C LE I N FO A B S T R A C T

Keywords: Toxicant concentrations in surface waters are of environmental concern due to their potential impacts to humans
Surface waters and wildlife. Numerical models provide system insight, support management decisions, and provide scenario
Numerical modeling testing on the impacts of toxicants. The Water Quality Analysis Simulation Program (WASP) is a widely used
Nanoparticles framework for developing site-specific models for simulating toxicant concentrations in surface waters and se-
Toxicants
diments over a range of complexities and temporal and spatial scales. WASP8, with the Advanced Toxicant
Sediments
module, has been recently released, incorporating a complete architecture redesign for an increased number of
state variables and different state variable types. WASP8 incorporates a new structure for simulating light in-
tensity and photoreactions in the water column, including the distinction of 10 different wavelength bands, and
nanoparticle heteroaggregation to solids. We present a hypothetical case study, using the Cape Fear River, North
Carolina as a representative example for simulating solute chemicals, nanoparticles, and solids to demonstrate
the new and updated capabilities of WASP8.

1. Introduction The Water Quality Analysis Simulation Program (WASP) is a dif-

ferential, spatially-resolved, mass balance, fate and transport modeling
Exposure to toxic contaminants in surface waters may pose a threat framework structured to allow users to simulate concentrations of en-
to the health of humans and wildlife. In the United States (US), the US vironmental contaminants in surface waters and sediments. WASP was
Environmental Protection Agency (EPA) has passed a series of acts to originally developed in the 1980s and has undergone a series of im-
address managing toxic contaminants in rivers and streams, such as the provements and versions over the decades. It is now widely used
Clean Water Act (CWA), the Toxic Substances Control Act (TSCA), the throughout the US and the world. WASP has been applied to a range of
Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA), and the different surface water systems, addressing a range of environmental
Comprehensive Environmental Response, Compensation, and Liability contaminants. WASP has been previously implemented for: nutrient
Act (CERCLA, “Superfund”). The implementation of environmental loading on water quality in Tampa Bay, Florida (US) (Wang et al.,
models is critical to the regulatory decision-making process. Spatial and 1999); mercury fate and transport in the Carson River, Nevada (US)
temporal scales linking environmental controls and environmental (Carroll et al., 2000); nutrients TMDL for Neuse River Estuary, North
quality generally do not allow an observational approach to understand Carolina (US) (Wool et al., 2003); mercury remediation strategies in the
the relationship between economic activity and environmental quality Sudbury River, Massachusetts (US) (Knightes, 2010); impacts of climate
(NRC, 2007). Environmental models are used in the regulatory process, change on water quality of Chungju Lake (South Korea) (Park et al.,
such as assisting in the development of Total Maximum Daily Loads 2013); the fate and transport of multi-walled carbon nanotubes in Brier
(under CWA), evaluating different feasible remediation strategies Creek, Georgia (US) (Bouchard et al., 2017), and as part of an un-
(CERCLA), or evaluating the possible impacts of the release of newly certainty and risk analysis for a margin of safety for a nutrient TMDL for
developed chemicals (review under TSCA). Sawgrass Lake, Florida (US) (Camacho et al., 2018).

∗
Corresponding author.
E-mail address: knightes.chris@epa.gov (C.D. Knightes).

https://doi.org/10.1016/j.envsoft.2018.10.012
Received 18 July 2018; Received in revised form 23 October 2018; Accepted 26 October 2018
1364-8152/ © 2018 Published by Elsevier Ltd.

Please cite this article as: Knightes, C.D., Environmental Modelling and Software, https://doi.org/10.1016/j.envsoft.2018.10.012
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

nanoparticles. This is a departure from the original architecture. Ad-

Software availability
ditional functionality has also been added to incorporate nanoparticle
specific processes, specifically particle attachment kinetics.
Program Title Water Quality Analysis Simulation Program, ver-
In this paper, we present the newly developed, publicly available
sion 8
WASP8 Advanced Toxicant Module (https://www.epa.gov/ceam/
Developers Tim Wool, Robert B. Ambrose, Jr.
water-quality-analysis-simulation-program-wasp, epawasp.twool.com).
Address United States Environmental Protection Agency, Water
The structure of the underlying software and modeling framework and
Quality Planning Branch/Data and Information
how it fits within an overall multi-media system of models is discussed.
Analysis Section Region 4 Atlanta, GA 30303 US Tel:
The general structure of the WASP8 Advanced Toxicant module with
404-562-9260 E-mail: wool.tim@epa.gov First avail-
details on the changes from the previous WASP toxicant modules,
able: July 7, 2017
particularly regarding the change in the structure of simulating state
Hardware 64 Bit Windows, 64 Bit Mac OSC (Yosemite or Higher),
variables and the available state variables, is described, and specific
64 Bit Linux (built on Ubuntu)
state variables and processes are explained. The three main state vari-
Software WASP 8.2 Executable, https://www.epa.gov/ceam/
ables are: solid particles, solute chemicals, and nanoparticles. First, the
water-quality-analysis-simulation-program-wasp,
solids module is addressed, as solids are an important factor for suffi-
http://epawasp.twool.com/
ciently simulating toxicants, particularly regarding contaminants at-
Language Fortran 95, C++
taching to particles and being transported by them. Next, partitioning
Size Executable is 3 MB, Files require 205 MB
and particle attachment are discussed, and the different processes of
equilibrium partitioning, kinetic partitioning, and heteroaggregation.
Light attenuation and phototransformations are then presented, fol-
Most recently, WASP8 (v. 8.2) was released, which incorporates an
lowed by a focus on the processes governing solute chemicals and then
overhaul of the WASP architecture and a restructuring and upgrading of
nanoparticles. Each of these sections discusses the options for different
the toxicant module, which is the focus of this paper. WASP8 updated
processes available to the user, from descriptive to mechanistic.
the underlying code from FORTRAN 77 to Fortran 95 and incorporated
Discussion of the different processes concludes with a hypothetical
a new WASP interface, linkage to a post-processor (Water Resources
application of WASP8 using the Cape Fear River as a case study. The
Database, WRDB v 6.1, www.wrdb.com), and the ability to run on
goal of this application is to demonstrate how the updated WASP can be
Windows, Mac OS, or Linux. WASP8 contains two modules: Advanced
applied for the different state variables and their governing processes.
Eutrophication and Advanced Toxicant. Over the past several years,
The application is structured to simulate flow and four different solids
WASP development has focused on the eutrophication module (Martin
using both the descriptive and mechanistic solids algorithms. To this
et al., 2006; Wool et al., 2013), and much modeling research has fo-
system, there is a hypothetical release of a solute chemical, which
cused on eutrophication (Gargallo et al., 2018; Nguyen et al., 2018;
transforms into another solute chemical, and a nanoparticle, which
Sadeghian et al., 2018). Less work over the years has focused on the
attaches to the different solid phases. The system uses time-varying flow
toxicant module routines, with limited development on the toxicant
conditions simulated using kinematic wave routing. Light is simulated
module since the 1990s. With the updated architecture of WASP8, re-
temporally over the day and year, with attenuation, and photo-
cent work has focused on upgrading the toxicant module routines. One
transformation of the nanoparticle.
driver of this effort was interest in emerging contaminants and their
potential exposure concentrations, particularly nanomaterials, once
released into the environment. Previous versions of WASP were not 2. Modeling framework
structured in a way to handle the introduction of these new con-
taminants. Consequently, the WASP8 Advanced Toxicant module WASP is a surface water modeling framework designed to create
completely restructured its state variables as distinct arrays to permit site-specific models for simulating environmental concentrations in the
for simulation of any number of chemical solutes, solid particles, and water column and sediment layers of different aquatic ecosystems over
space and time. The WASP framework is used to develop a water

Fig. 1. WASP8 model framework.

2
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

quality model for a specific system of interest and contaminants of simulating water quality in Yongdam Lake, Korea (Seo et al., 2009).
concern and is structured using a series of different module compo- WASP currently contains two kinetics modules: Advanced
nents. Fig. 1 shows the overall WASP modeling framework. Generally, Eutrophication and Advanced Toxicants. Example state variables
the user interacts with WASP through the interface to create the WASP available in both modules are presented in Table 1. The Advanced
input file (WIF). WASP reads the WIF to run a simulation. WASP has a Eutrophication module has explicitly incorporated state variables (e.g.,
series of hydrodynamic modules incorporated into its architecture (e.g., nitrate, ammonia, orthophosphate) (Martin et al., 2006) (Wool et al.,
flow routing, kinematic wave, diffusive wave, and dispersive wave) 2013) and has been reviewed in the context of other eutrophication
(Ambrose and Wool, 2017) and can import data files from other hy- models (Sharma and Kansal, 2013). The Advanced Toxicant module is
drodynamic models (e.g., EFDC) via a hydrodynamic file (HYD file). structured differently than the Advanced Eutrophication module by
The science modules are shared libraries for kinetic parameters and having state variable classes (e.g., solute chemicals, nanoparticles, so-
transformation processes for the specific state variables. Stored data lids). This provides the user with the flexibility of constructing a WASP
files include data base time series information for time functions (e.g., model designed for a range of different contaminants of interest. By
solar radiation, air temperature), boundary concentrations, inflows, and defining and parameterizing governing processes, the user effectively
loads. Once the WASP simulation is performed, an OUT file (a text file defines that state variable (e.g., a solute chemical is defined as atrazine
of parameters and outputs) and a BMD (binary data) file are produced. by its parameterization). Each state variable and process has different
The BMD is readable by WRDB, which is a tool that works with WASP to options for how it is described.
manage data and provide graphical outputs. WRDB can also be used to Governing equations for concentrations follow the advection-dis-
generate figures or files for other programs to read (e.g., CSV) for fur- persion-reaction equation [e.g., GEM (Little et al., 2018), CE-QUAL-W2
ther data analysis. WASP can also be used in the greater context of (Sadeghian et al., 2018), CE-QUAL-ICM (Cerco et al., 2010)]. The sci-
multi-media modeling, taking output from watershed models [e.g., ence module and the transport module are solved using differential
SWAT (Nietsch et al., 2011), HSPF (Bicknell et al., 2001)] and feeding mass balance equations for each segment for each time step. The user
forward as input into bioaccumulation models [e.g., BASS (Barber, may select different solution techniques, but generally the forward
2006), Gobas model (Arnot and Gobas, 2004)], as was done to link Euler numerical approximation approach is used with varying time
atmospheric deposition to fish tissue concentrations (Knightes et al., steps designed to minimize numerical instability.
2009), and can be a part of a multi-media modeling framework like the
Framework for Risk Analysis in Multimedia Systems (FRAMES)
(Johnston et al., 2011; Whelan et al., 2014). 3. WASP8 Advanced Toxicant Module
WASP can be structured with different spatial resolutions as re-
quired by the user, to simulate zero-dimensional lakes and ponds; one- The Advanced Toxicant Module has been updated as part of WASP8
dimensional streams, with or without branching; two-dimensional to simulate an increased number of specific state variables. Table 2
rivers and stratified lakes; and three-dimensional estuaries or large presents some of the changes from WASP7 to WASP8 with respect to the
lakes. The WASP model domain is constructed with WASP segments: Advanced Toxicant module. Originally, the WASP Toxicant module had
surface water or subsurface water (water segments), and surface a single class of state variables, and WASP kept track of whether it was a
benthic or subsurface benthic (sediment segments). Segments are as- solid type or a dissolved chemical type. This legacy architecture made it
sumed to be uniformly mixed, whether they are a water column seg- challenging to incorporate increased numbers of state variables as well
ment or a sediment segment. Within each segment, each state variable as different types of state variables. Using a new architecture, state
is simulated for every time step. The user defines how the segments are variables have been separated into their own classes, with up to 5 or 10
laid out in space through the interface. This allows for three-dimen- state variables per state variable class; these numbers can be increased
sional model domains, with multiple sediment layers as needed. by recompiling after adjusting a compile parameter.
WASP incorporates simple hydrodynamic modules (Ambrose and Now that WASP8 has separate state variables, the structure of the
Wool, 2017), which include general constant flow routing, stream Advanced Toxicant module has been redesigned, and additional pro-
routing using hydrogeometry, kinematic wave routing using slope and cesses and options have been added. The new design also facilitates
roughness, and dynamic wave for tidal influences [based on DYNHYD5 updates in the science and algorithms moving forward with WASP fu-
(Ambrose et al., 1993)]. These modules are best suited for one-di- ture development. Table 3 presents an overview of the processes for
mensional systems, including branching. For larger lakes and estuaries, different state variables in WASP8. As part of the development of the
a hydrodynamic model can be used to provide details on flow, which WASP8 Advanced Toxicant Module, model verification was performed.
can be entered manually using flow routing or via hydrodynamic model The solids, light, particle attachment, and nanomaterial and chemical
linkage. EFDC can be used to simulate the hydrodynamics of the solute reaction algorithms were isolated and simple simulations were
system, which is then linked to the WASP model for simulating con- performed and compared to analytical solutions to verify that the al-
taminant fate and transport, which has been demonstrated successfully gorithms and modules were performing properly (Ambrose et al.,
for a nutrient TMDL in the Neuse River, NC, USA (Wool et al., 2003) for 2017).
algal blooms caused by increased residence time by construction of
instream structures in the Nakdong River, Korea (Seo et al., 2012) and

Table 1
WASP8 Modules and Environmental Constituents. The two modules available in WASP8 and the state variables along with example types of environmental
constituents that can be simulated.
WASP8 Modules Example Environmental Constituents

Advanced Eutrophication Dissolved Oxygen, Nitrogen (Nitrate, Ammonia), Orthophosphate, Periphyton, Phytoplankton, Carbonaceous Biochemical Oxygen Demand,
Temperature, Salinity, Alkalinity
Advanced Toxicant Solute Chemicals: PAHs, PCBs, Pesticides, Metals;
Solids: Sand, Silt, Clay, Particulate Organic Matter;
Nanoparticles: Graphene Oxide, Carbon Nanotubes;
Mercury: Elemental Mercury, Divalent Mercury, Methylmercury;
Dissolved Organic Carbon; Pathogens; Salinity; Temperature

3
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

Table 2
Overview comparison of advances in WASP8 advanced toxicant module from WASP7.
WASP7 - Toxicant WASP8 – Advanced Toxicant

State Variables 1-3 Chemicals 1-10 + Solute Chemicals

1-3 Solids 1-10 + Solids
1-10 + Nanoparticles
1-5+ Dissolved Organic Carbon
1-5+ Pathogens
Temperature
Salinity
Particle Attachment Equilibrium Partitioning Equilibrium Partition,
Kinetic Sorption,
Heteroaggregation kinetics
Light Attenuation Single wavelength, constant attenuation coefficient per 10 bands of wavelengths,
segment Attenuation a function of suspended solids, dissolved organic carbon (DOC), and algae by
wavelength
Photoreactions Rate x light intensity Rate x light intensity for each wavelength band
Daughter Products Chemicals can break down into other chemicals Chemicals can form nanoparticles and other chemicals; nanoparticles can form chemicals and
other nanoparticles

Table 3 particle size classes have been shown to affect the distribution and
Overview of processes for state variables in WASP8 advanced toxicant module. accumulation of simulated nanomaterials in surface waters (Quik et al.,
State Variable Processes
2015). Each solid type state variable is defined by the way the user
constructs the model and parameterizes the solids state variable. Solids
Solids Settling, Deposition, Erosion, Resuspension, Burial, in the Advanced Toxicant module can either be organic materials (e.g.,
Bedload, Decay, Growth plankton, algae, detritus) or inorganic (e.g., sand, silt, clay). The solids
Solute Chemicals Oxidation, Reduction, Biodegradation, Particle
Attachment, Ionization, Volatilization, Photoreaction
state variable captures the processes for all these solids types in a
Nanoparticles Decay, Photoreaction, Particle Attachment general framework. The user parameterizes each solid type to reflect its
Dissolved Organic Decay, Complexation characteristics and properties (Fig. 2).
Carbon There are four options for simulating solids in WASP: solid flow
fields, descriptive, mechanistic van Rijn equations (van Rijn, 1984), or
mechanistic Roberts equations (Roberts et al., 1998). For all options, all
4. Solids module
solids move via advection. Dispersion may also move solids if the ex-
change option is turned on and dispersion rates are defined between
Solids are an important environmental component of water quality
surface water segments. Particle density is set for each particle type.
modeling. Suspended solid concentrations can be of concern on their
own, by reducing light penetration or blanketing benthic spawning
areas, or because environmental contaminants attach to the particles 4.1. Solid flow fields
and their transport becomes linked to the movement of the particles. In
previous WASP releases, up to three solid particles could be simulated, The Solid Flow Fields option is the legacy option used in WASP7 and
e.g., sand, fines (clay/silt), and particulate organic matter. In WASP8, earlier versions. This option treats each solid state variable type sepa-
there can be up to 10 solids types as defined by the user, such that clays rately. Movement of solids from one segment to another is explicitly
and silts can now be separate state variables. This allows for increased defined by using a flow field of a specific flow velocity and the cross-
refinement in solids modeling, with different settling and resuspension sectional area of the interface. Flow fields are generic, and the user
rates as well as different partition coefficients. Increased numbers of establishes whether it is resuspension or settling via the direction of the

Fig. 2. Overview of processes for simulating solids in WASP8. (A) Descriptive and (B) Mechanistic.

4
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

flow field. As each solid flow field is linked to a solids type and is from captures the fact that all settling particles deposit onto the sediment
the legacy structure, the flow fields are limited to three. layer when shear stress is low. As shear stress increases, fewer particles
deposit, until flow conditions reach a regime where the shear stress is so
4.2. Descriptive solids transport high that no particles deposit. This allows the system to shift between
deposition or no deposition.
The Descriptive Solids Transport (Fig. 2A) option uses constant
settling and resuspension velocities, which do not change with stream 4.3.3. Non-cohesive erosion
velocity. For this option, the settling velocity, vs [m s−1], and re- The model domain does not differentiate the boundary layer
suspension velocity, vr [m s−1], are entered for each solid type for each (Fig. 2B) as a separate WASP8 segment from the overlying water
segment. As part of the model domain set up in the segment definition, column. The boundary layer is used conceptually to separate the pro-
the user enters which segment is below a given segment. WASP then cesses of erosion and resuspension in WASP8. When the sediment is
knows for a given segment where both settling and resuspension is acting non-cohesively (fine fraction is less than critical cohesive frac-
going to and coming from. Therefore, if a model has sand, silt, and clay, tion), all particles act non-cohesively. In cohesive sediment segments,
the settling and resuspension velocities for descriptive transport are only sands and larger particles (> 0.1 mm diameter) are subject to non-
determined by the user and entered directly into the interface. cohesive erosion and resuspension. WASP8 calculates the non-cohesive
erosion velocity for each solids type using equations either developed
4.3. Mechanistic solids transport by van Rijn or Roberts, depending on the solids option.
The van Rijn equation calculates the gross deposition rate, vs [m
The Mechanistic Solids Transport (Fig. 2B) options simulate dy- s−1], where ve = Evs , using a proportionality constant, E, where
namic settling, deposition, erosion, and resuspension velocities based
D −0.2 η
on a set of parameters and the simulated flow velocity of the surface E = 0.015γe Rd τ∗
ks (3)
water. Settling is the process of solids moving downward through the
water column and is a function of particle size and density. Deposition, and γe is a user-specified multiplier (default = 1), η is a user-specified
erosion, bedload, and resuspension are functions of the shear stress on exponent (default = 1.5), and τ∗ is the non-dimensional shear stress,
the stream bed. Deposition is when a solid particle moves from the given by:
water column to the surface sediment. Erosion is when solids move τ b − τc, e
from the surface sediment layer into the overlying water boundary ⎧ τc, e
τb ≥ τc, e
τ∗ =
layer. Resuspension is when the solid particle leaves the boundary layer ⎨ 0 τb < τc, e
⎩ (4)
and enters the water column. Erosion and resuspension additionally
−2
depend on whether the sediment layer is acting cohesively or non-co- where τc, e is the critical shear stress for erosion [N m ] (See Appendix,
hesively. For a non-cohesive sediment layer, all particles act in- Eqn. A-3).
dependently. When a sediment layer is cohesive, all particles < 0.10 For the Roberts equation,
mm diameter (fines) erode as a single unit. The user defines the critical ve = γe Aρbm τbn (5)
cohesive sediment fraction for which the sediments act cohesively (the
−3
default is 0.2, so when fines comprise 20% or more of the bed, the bed where ρb is the bulk density [kg m ] and coefficient, A, and exponents
acts cohesively). The inclusion of the mechanistic solids transport al- m and n are determined experimentally for different particle sizes from
lows for erosional events during high flows, and for depositional events fine silt (< 5.7 nm) to coarse sand (> 1.25 mm) (Roberts et al., 1998).
and zones for slower flows.
4.3.4. Non-cohesive resuspension
4.3.1. Settling Particles that erode from the sediment layer enter the boundary
For the mechanistic option, WASP8 calculates the settling velocity layer. At that point, particles can resuspend and enter the overlying
for different ranges of particle diameter, D [m], as a function of particle water segment or travel downstream via bedload. Eroded particles
density, ρs [kg m−3], water density, ρw [kg m−3], and absolute visc- move in the boundary layer via bedload when the shear stress is less
osity, μ [kg m−1 s−1] (van Rijn, 1984). than the critical shear stress for resuspension, τc, r [N m−2] (See
R
d
Appendix).
⎧ 18 g′D D ≤ 100 μm When τb > τc, r , particles resuspend with a resuspension velocity,
⎪ 10
vs = ( 1 + 0.01Rd2 − 1) g′D 100 μm < D ≤ 1 mm vr = fr ve [m s−1], where
⎨ Rd
⎪1.1 g′D D > 1 mm 0 τ b < τc , r
⎩ (1) ⎧
⎪ ln ⎛ u∗ ⎞ − ln ⎛ u∗, c, r ⎞
⎪ ⎝ vs ⎠ ⎝ vs ⎠
Where Rd is the sediment particle densitometric Reynolds number, fr = τ b ≥ τc , r
Rd = D g ′D /(μ/ ρw ) [unitless], and g ′ = g (ρs / ρw − 1) [m s−2], where g ⎨ ln (4) − ln ⎛ u∗, c, r ⎞
⎪ ⎝ vs ⎠
is the acceleration of gravity, 9.807 m s−2. ⎪ 1 u∗ ≥ 4vs (6)
⎩
−1
4.3.2. Deposition and u∗ = τb/ ρw is the shear velocity [m s ] and u∗, c, r = τc, r / ρw is the
For the mechanistic option, WASP8 calculates the deposition velo- critical shear velocity for resuspension [m/s] (Jones and Lick, 2001;
city, vd [m s−1] as vd = vs αd , where αd [unitless] is the probability that a HydroQual, 2007).
solid will deposit and is a function of bottom shear stress as well as the
lower and upper critical shear stress thresholds, τc, d,1 [N m−2] and τc, d,2 4.3.5. Non-cohesive bedload
[N m−2] (Krone, 1963), where Bedload is the transport of non-cohesive solid particles downstream
through the boundary mobile layer. When τb > τc, e , bedload may begin.
⎧1 τb ≤ τc, d,1 The bedload flux per unit width, gb, l [g m−1 s−1] is given as:
⎪ τc, d,2 − τ b γD
αd = ( )
⎨ τc, d,2 − τc, d,1
τc, d,1 < τb ≤ τc, d,2
⎛ ⎞
η

⎪ 0 τb > τc, d,2 D 1.2 u − u c, e

⎩ (2) gb, l = αb, l ρs uH ⎛ 50 ⎞ ⎜ ρs − ρ w
⎟
⎝ H ⎠ ⎜ gD50 ⎟
and γD is a dimensionless exponent (default = 1). This relationship ⎝ ρw
⎠ (7)

5
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

where D50 is the mean particle diameter [m], αb, l is a fitted coefficient, u WASP time step). For the surface sediment layer, solids may be de-
is stream velocity [m s−1], H is water depth [m], η is a fitted exponent, posited so that total mass in the segment increasing, or eroded, re-
and uc, e is the critical stream velocity for erosion (See Appendix). sulting in mass increasing.
Suggested values for the fitted parameters are αb, l = 15 and η = 1.5 For the Static option, the volume is held constant, so the density of
(van Rijn, 2007). the segment changes to account for the change in mass, and solids aren't
moved between segment layers. For the Dynamic option, the volume
4.3.6. Cohesive resuspension changes and the density is held constant. If there is only one layer (one
When the surface sediment layer is in the cohesive regime, all co- segment) in the sediment, then deposition can cause the layer to in-
hesive solids erode at the erosion velocity, ve [m s−1]. The cohesive crease in depth and volume indefinitely, and there will be no burial (no
erosion flux, Ecoh [g m−2 s−1], is given by the excess shear stress power sediments are pushed out the bottom of the segment). Conversely,
law formulation (Lick et al., 1994), Ecoh = fcoh Mτ∗n , where M is the shear erosion can result in loss of volume under the Dynamic option; how-
stress multiplier [g m−2 s−1], n is the shear stress exponent, fcoh is the ever, erosion is halted after segment volume reaches 5% of the initial
fraction of the surface bed that is cohesive, and τ∗ is the excess shear value.
stress [N s−1] defined in Section 4.3.3. The shear stress multiplier When there are multiple sediment layers in the Dynamic option, the
varies between 0.1 and 100 [g m−2 s−1], with a default value of 5. The surface sediment segment volume and depth are reset to initial values at
shear stress exponent varies between 1.6 and 4, with a default value of each benthic time step. WASP8 calculates the segment volume that
3. The critical shear stress for erosion varies between 0.5 and 8 [N needs to move, Vb [m3] and its associated depth, db [m], to return to the
m−2], with a default value of 2. WASP8 is structured so that the shear initial volume and depth. The mass of contaminants and solids move
stress multiplier, exponent, and the critical shear stress for erosion can with vb [m d−1], the burial rate. Burial fluxes Fb, i [g m−2 s−1] of solid
vary spatially, if desired. or pollutant, i, are given by Fb, i = Ci vb , where Ci [g m−3] is the con-
centration of the constituent of interest in the sediment layer. Negative
burial rates reflect erosion. Under erosion conditions, the burial rate is
4.3.7. Organic solids production and dissolution
Additional processes are available if the user wants to classify a reversed, and WASP8 recruits the necessary volume, Vb, from the se-
diment layer below. The mass moved then reflects the concentration of
solid particle as an organic solid, such as algae or particulate organic
matter (POM). For any given solid, the net production rate, Rp [g m−3 the sediment layer below the segment of interest.
In the subsurface sediment layers, at each benthic time step, mass is
d−3], is given as
either moved upward or downward. Under deposition conditions, the
Rp = (Rp, s + Rp, t ) θpT − 20 (8) subsurface sediment segment receives solids from the overlying sedi-
−3 −1
ment segment at each time step. If the subsurface sediment segment has
where Rp, s [g m d ] is the spatially-variable production rate (by a higher bulk density, Vb and db are compressed and pore water is pu-
WASP segment), Rp, t [g m−3 d−1], is the time-variable production rate shed upward. The compressed volume is then passed downward to the
(defined by providing a WASP Time Function), and θp [-] is the tem-
next sediment layer or out of the system. Solids and pollutants in this
perature correction coefficient.
volume are similarly passed downward, maintaining the initial bulk
Organic matter may also decay and leave behind ash and dissolved
density or porosity. Under erosion conditions, the sediment segment
organic matter. Using a similar functionality, the loss rate of an organic
receives the erosion volume from the underlying sediment layer, along
solid particle via dissolution is given by the dissolution rate constant, kd
with the associated solids and pollutant masses. This process is con-
[d−1],
tinued downward until all underlying sediment segments have been
k d = (k d, s + k d, t ) θdT − 20 (9) adjusted.
−1
where kd, s [d ] is the spatially-variable dissolution rate (by WASP 5. Toxicant partitioning and attachment to particles
segment), kd, t [d−1] is the time-variable dissolution rate (defined by
providing a WASP Time Function), and θd [-] is the temperature cor- One of the most important processes governing the fate and trans-
rection coefficient. When an organic solid type dissolves, a fraction of port of toxicant chemicals is their interaction with particles. WASP8
the particle leaves behind residue, described as the “ash dry weight currently incorporates equilibrium partitioning between solute chemi-
residue.” The mass associated with the fraction of residue then is added cals and solid particles, kinetic sorption, and heteroaggregation kinetics
to a different solid type, as identified by the user. Similarly, the organic of nanoparticles and solid particles. Previously, WASP only allowed for
fraction may dissolve into one of the DOC state variables. For example, equilibrium partitioning. This section presents the modeling structures
as POM dissolves, a fraction may form silt, and another fraction may available for simulating the interactions of both solute chemicals and
form DOC. The Advanced Toxicant module has these simple equations nanoparticles with particles.
for organic solids, if a more rigorous representation of organic matter is
desired, then the Advanced Eutrophication module can be run to si- 5.1. Equilibrium partitioning
mulate nutrients and algae growth of 3 types of phytoplankton. The
output can then be pulled directly into the Advanced Toxicant as an For organic chemicals partitioning onto sediment organic matter,
input file (Wool et al., 2013). partitioning kinetics are usually faster than the time steps simulated in
WASP. Under these conditions or as a simplifying assumption, in-
4.3.8. Solids burial and erosion stantaneous equilibrium partitioning is often used in water quality
WASP segments are classified as either water or sediments. Beneath modeling of toxicant contaminants. For environmental concentrations
the water column, the user can structure one or multiple sediment (< 10−5 M or one-half of water solubility), equilibrium sorption is
segments. For each sediment layer segment, the initial solids con- linear with dissolved concentration (Karickhoff, 1984) as:
centration in each segment establishes the reference bulk density [g Cs, i = K d, i Cw, i , for a pollutant, i, and a particular solid, s. Cs, i [mg L−1] is
m−3] and porosity [m3 void space m−3]. As the simulation progresses, the chemical concentration on the solid phase, Cw, i [mg L−1] is the
solid particles are depositing, eroding, growing, and dissolving. For concentration in the water phase, and K d, i [L kg−1] is the partition
each sediment segment, WASP8 conducts a solids mass balance, with coefficient. When there are multiple solid phases present, this re-
two approaches for representing the sediment bed volume: Static lationship holds for all constituents, and the overall fraction of each
(constant volume) and Dynamic (constant bulk density). The Dynamic chemical on each phase is given by solving all equations simulta-
runs using the benthic time step, Δtb [d] (operates distinctly from the neously, so that:

6
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

K d, ij Sj Pathogens and viruses may be deactivated by different wavelengths of

fi, j = n
1 + ∑i = 1 K d, ij Sj (10) light. To effectively simulate phototransformations requires the light
intensity that reaches the surface of the water body, the amount of light
where fi, j is the fraction of a solute chemical, i, on the given solid phase, that penetrates through the water column, and the photoreaction itself.
j (e.g., silt, sand, clay), n is the number of solid phases present, j, and Sj The amount of light reaching the surface depends on cloud cover,
[mg L−1] is the specific solid concentration. Complexation with DOC is shading, and the solar zenith angle. Solar radiation is comprised of
simulated using the same functionality. different wavelengths. Each wavelength affects different environmental
constituents in different ways, attenuates differently as it passes
5.2. Kinetic sorption through the water column, and comprises a different fraction of solar
radiation. WASP8 accounts for ultraviolet (295–379 nm), visible
In some cases, equilibrium partitioning is not a good representation (380–749 nm) and infrared (750–2500 nm). Ultraviolet and visible light
of the system. For example, solute chemicals may exhibit sorption and are both divided into five wavelength bands (See Appendix). This sec-
desorption hysteresis, where kinetic sorption may be a more appro- tion provides details on the different options for incorporating input
priate representation. Because the system is dynamic, WASP8 uses se- total radiation, attenuation of solar radiation before reaching the water
parate state variables to represent the freely-dissolved chemical, Cw, i surface, light attenuation in the water column, and phototransforma-
[mg L−1], and the sorbed chemical, Cs, i [mg L−1]. This model structure tions.
requires n+1 state variables, where n is the number of solids present in
the system, and S [mg L−1] is the concentration of the specific solid 6.1. Input total solar radiation
type. The sorption process is described as:
dCw, i WASP8 has 3 different options for quantifying solar radiation, Iinput
= −k for Cw, i S + k rev Cw, i [W m−2 h−1], reaching the water surface: internally calculated diel
dt (11)
−1 −1
light; user input diel light, or user input daily light with calculated diel
where kfor [L mg d ] is the forward reaction rate constant (sorption) light. For the first option, WASP8 will calculate Iinput based on the site
and krev [d−1] is the reverse reaction rate constant (desorption). latitude and longitude varying by date and time of day. For the second
option, Iinput as one or more (up to four) time functions is entered, so
5.3. Nanoparticle heteroaggregation kinetics different locations can have different solar radiation at the surface as a
function of time. For the third option, the sum of all radiation for a
Attachment of nanoparticles to solid particles is simulated using given day [W m−2 h−1] is entered, then WASP8 calculates Iinput over
equations based on colloidal theory using heteroaggregation kinetics the course of the day and year.
(Arvidsson et al., 2011; Praetorius et al., 2012). The overall hetero-
aggregation rate is defined by: 6.2. Solar radiation to water surface
khet , i = αk coll, i, j Ns, j (12)
As solar radiation passes through the atmosphere, it attenuates due
Where, α [unitless, varies from 0 to 1] is the attachment efficiency, to different processes. If WASP8 is calculating Iinput, atmospheric at-
which is a user-entered segment-specific parameter; k coll, i, j [m3 d−1] is tenuation is not automatically incorporated. If solar radiation is en-
the rate of collision, calculated internally by WASP8, between the na- tered, then the included attenuation processes needs to be known so the
noparticle, i, and a solid particle, j; and Ns, j [# m−3] is the number others can be accounted for. WASP8 incorporates four attenuation
concentration of suspended particles, an explicitly simulated state processes specifically: cloud cover, canopy shading, water surface re-
variable. The collision rate is comprised of three components: Brownian flectance, and attenuation in ice. Cloud cover and canopy shading are
motion, fluid motion, and differential settling. incorporated by using time functions, and reflectance is a constant.
2kB Tw (rNP + rS )2 4 These factors adjust Iinput as:
k coll = + G (rNP + rS )3 + π (rNP + rS )2 vNP, s − vs, s
3μ w rNP rS 3 Io = 0.65∗Iinput (1 − Cloud )2 (1 − Shade )(1 − Reflectance ) (14)
(13)
and the development of ice on the surface will further attenuate solar
where kB [1.38 × 10−20 m2 g s−2 K−1] is the Boltzmann constant and radiation as:
μw [g m−1 s−1] is the dynamic viscosity of water calculated as a
Io, ice = Io (1 − α )2 (1 − β ) e −γh (15)
function of temperature. Tw [Kelvin] is the temperature of water, G
[s−1] is the shear rate, rNP [m] is the radius of the nanoparticle, rS [m] where α is albedo, β is surface absorption, γ is ice extinction coefficient,
is the radius of the solid particle, vNP, s [m s−1] is the settling velocity of and h [m] is the ice thickness (Edinger et al., 1974; Cole and Buchak,
the nanoparticle, and vs, s [m s−1] is the settling velocity of the solid 1994). Ice thickness can be calculated by WASP8, by simulating water
particle. The kcoll is calculated internally in WASP8 for each segment at temperature, or included by using a time function.
each time step. T can be directly simulated or entered for each segment
or as a time function. The user enters rNP and rS. The vNP, s is user-defined 6.3. Light attenuation through water column
per segment. The vs, s is defined based on the solids option the user
chooses (Section 4). The shear rate, G, is calculated as G = u/H. WASP8 WASP8 simulates light attenuation through the water column fol-
requires n+1 nanoparticle state variables for simulating hetero- lowing the Beer-Lambert equation: Iz = Io e−K e, λ z , where Ke,λ is the light
aggregation for each nanoparticle type, where n is the number of solid extinction coefficient [m−1] for a given wavelength index, λ, and z is
particles simulated in the system. depth below the surface [m]. WASP8 allows for a segment-specific,
general light extinction coefficient or for a WASP-calculated light ex-
6. Light module and phototransformations tinction coefficient. WASP8 calculates the light extinction coefficient
specific to each wavelength, as a summation of contributions from
Solar radiation is an important factor in simulating environmental background water, algal chlorophyll a (chl) [μg L−1], raised to a
constituents in surface waters. Solute chemicals (Schwarzenbach et al., chlorophyll exponent, aλ [unitless], DOC [mg L−1], and total suspended
1993), metals (Stumm and Morgan, 1981; Morel and Hering, 1993), solids (TSS) [mg L−1] as:
and nanoparticles (Hou et al., 2015) may undergo photochemical re-
K e, λ = Kw, λ + K chl, λ (chl )aλ + KDOC, λ DOC + Ksolid, λ TSS (16)
actions, resulting in their degradation and/or transformation.

7
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

Additional information on the values for these wavelengths, divi- L−1], KE [mg L−1] is the half-saturation coefficient for an environ-
sions of wavelength by class, and the division of wavelengths into wave mental concentration, and CE [mg L−1] is the environmental con-
bands as a function of latitude are provided in the Appendix. centration of interest (e.g., oxygen). If Monod substrate kinetics are
used, then krate = μmax Xmonod_subs Xmonod_env [mg L−1 d−1]. Time func-
6.4. Photoreactions tions can be used to apply time varying concentrations of reactants.

Solute chemicals and nanoparticles may undergo photoreactions. 8. Additional processes for nanoparticles
WASP8 simulates photoreactions using first-order kinetics with the
photoreaction rate constant kphoto [d−1], which accounts for photo- Nanoparticles are new state variables in WASP8, routinely defined
reactions due to ultraviolet and visible light. The wavelength specific as materials sized between 1 and 100 nm. Engineered nanoparticles
reaction rate constant, kλ [d−1 W−1 m2], due to each wavelength band have been applied in all areas of our daily lives, and their production
is multiplied by the associated wavelength-specific radiation in the has increased appreciably in recent years. Such rapid expansion of
segment, Iλ [W m−2] and summed to provide the total rate constant as: nanoparticle production increases the likelihood of nanoparticles being
10
kphoto = ∑λ = 1 kλ Iλ . released into the environment. Besides the heteroaggregation particle
attachment process in surface waters described in Section 5, nano-
7. Additional solute chemicals processes particles may undergo phototransformations, sulfidation, oxidation,
and dissolution (See Fig. 4). Similar formulations describing
Distribution and concentrations of dissolved contaminants in sur-
face waters are significantly influenced by interactions between con-
taminants and the physical and chemical components of aquatic en-
vironments. These fate processes must be fully assessed during the
evaluation of contaminant fate and transport to accurately simulate the
behavior of contaminants in surface waters. WASP8 includes biode-
gradation, oxidation, reduction, and photochemical reactions. The
previous WASP version (WASP7) can simulate these four chemical re-
actions for up to three chemical state variables, while WASP8 allows
simulation of up to 10 variables, including transformation of one che-
mical into another, a chemical into nanoparticles, as well as nano-
particles forming chemicals. These advances allow for increasing the
speed of modeling multiple contaminants at once (e.g., simulating 4
metals at once), as well as for simulating more complex biodegradation
or transformation pathways (e.g., simulating an organic chemical with
4 daughter products). Fig. 3 presents an overview of the processes
available for solute chemicals in WASP8.
Oxidation, reduction, and biodegradation reactions are simulated
use a generic transformation reaction rate:
k reaction = k rate Xtemp Xseg Xsegtype Xphase Xmonod _ subs Xmonod _ env (17)
constructed by multiplying the base reaction rate constant, krate [d−1],
by a series of unitless factors. The temperature correction factor, Xtemp ,
allows for the reaction rate to change as a function of temperature, as
Xtemp = θT − 20 where θ [unitless] is the temperature-activity coefficient
and T [Celsius] is temperature. The segment multiplication factor, Xseg ,
allows each individual segment to have a different rate constant by
multiplying it by this factor, with a different value for each reaction
process (i.e., biodegradation, oxidation, reduction). The segment type
multiplication factor, Xsegtype , is set to affect all segments of a general
type (i.e., water column, surface sediment, subsurface sediment). In this
way, reactions can be set binarily to either happen or not in the water or
sediments, or to be faster or slower depending on the segment type. The
3
phase multiplication factor, Xphase , is described as: Xphase = ∑i fi x i
where fi is the mass fraction of the contaminant in phase i of three
possible phases: the freely dissolved phase, DOC-complexed phase, or
sorbed onto a solid phase, and x i are phase multiplication factors as-
sociated with each phase. The phase multiplier effectively allows con-
trol for when a reaction takes place for dissolved, DOC-complexed, or
sorbed species.
Previously, WASP only allowed for first order kinetics. WASP8 now
includes first order, second order, and Monod kinetics to represent
different kinetics formulations [e.g., Schnoor (1996)]. Monod kinetics
can be incorporated with the rate expressed as a product of a maximum
rate constant (μmax) and rate-limiting factors for the substrate con-
centration and additional environmental concentrations (e.g., electron
acceptors). These factors are given by Xmonod _ subs = 1/(Ks + C ) and
Xmonod _ env = CE /(KE + C ) where Ks [mg L−1] is the half-saturation
coefficient for the substrate, C is the concentration of the substrate [mg Fig. 3. Overview of governing processes for solute chemicals in WASP8.

8
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

Fig. 4. Overview of governing processes for nanoparticles in WASP8.

Fig. 5. Site Map of the Cape Fear River, North Carolina, US. The model domain is presented with 45 river reaches for the 90 km length of the system from USGS
gage 021055500 to USGS gage 02105769, with the associated watershed.

9
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

Table 4 reactions, WASP8 uses a general formulation, using the same general
Physical-chemical properties of simulated solute chemicals and nanomaterial. chemical reaction (eqn. (17)) that is used for solute chemicals, as is
Parameter Chemical 1 Chemical 2 Units described in Section 7. A general, first-order decay rate process is also
available. A nanoparticle can be set to transform into a daughter pro-
Molecular Weight 350 300 g mole−1 duct, which can be either a solute chemical or another nanoparticle,
Henry's Law Constant 0.02 0.001 atm mole−1 m−3
with the yield of each given by a user-defined yield coefficient.
Partition Coefficient, Sand 10 100 L kg−1
Partition Coefficient, Silt 1000 10000 L kg−1
Partition Coefficient, Clay 1000 10000 L kg−1 9. Additional state variables
Partition Coefficient, POM 500 5000 L kg−1
Biodegradation Rate Constant 0.2 0.01 d−1 In addition to the state variables and processes already discussed,
the WASP8 Advanced Toxicant module also has DOC, salinity, tem-
Nanomaterial Units
perature, pathogens, and mercury as available state variables. In earlier
Collision Efficiency 0.0002 unitless versions of WASP, DOC was a segment-specific parameter, now WASP8
Settling Velocitya 0.02 m d−1 permits explicit simulation of 1–5+ DOC types, which may oxidize or
Particle Diameter 100 nm
decay as first-order processes, and organic solids that may decay to
Photoreaction Rate Constant 0.001 to 0.01 d−1
form DOC. Salinity, temperature, and pathogens were originally avail-
a
WASP assumes settling and deposition of nanoparticles from the water able in a separate WASP module (HEAT), with the option of setting the
column to the sediments are negligible; the settling velocity is solely used to output from HEAT to be linked to the Eutrophication or Toxicant
calculate the collision rate constant. module. Mercury was previously available in WASP7, as an overlay
using the early toxicant module kinetics module (TOXI) with a separate
nanoparticle fate and transport processes have been presented mercury module (MERC). MERC simulated elemental mercury, divalent
(Praetorius et al., 2012; Quik et al., 2015). mercury, and methylmercury explicitly with specific transformation
Photoreactions are modeled as described in Section 6.4. For other processes outlined for these species. MERC has been incorporated into
the WASP8 Advanced Toxicant module, with the additional state

Fig. 6. Simulation Results of Solids. Sand, silt, clay, and POM (particulate organic matter) concentrations are presented over time for segment 20 (midstream) and
40 (downstream) for the water column (top row), surface sediment layer (middle row), and deep sediment layer (bottom row). Descriptive Solids option are
presented on the left two columns, and Mechanistic Solids are presented in the right two columns. Concentrations are presented using area plots, so that the total
concentration of solids is presented as the sum of all the concentrations.

10
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

Fig. 7. Light Simulation at Segment 20. Light intensity in the water column at noon throughout the simulation for each wavelength band and specific photo-
reaction rate constants based on each wavelength's intensity.

variable of recalcitrant divalent mercury, so that historically con- is in the Piedmont region of NC. This section of the Cape Fear River is in
taminated mercury sites can explicitly simulate the difference between the coastal plain region and is used as a source of drinking water at
labile and recalcitrant mercury. several locations, and there are no substantial tributaries.
Flow routing is simulated using the kinematic wave functionality,
10. Example WASP8 model application: hypothetical case study of and depth, width, velocity, and volume are simulated per time step for
the Cape Fear River each water column segment. USGS 02105500 is used as the upstream
inflow condition. Details on simulating flow for the Cape Fear River are
The goal of this paper is to present the details on the advances presented in the Appendix. WASP8 has different methods for simulating
implemented in WASP8. To provide details on a WASP8 model to de- flow in surface waters, and for simulating simple hydrodynamics in 1-
monstrate the new and updated capabilities, an example WASP8 model dimensional systems. WASP8, itself, is not a hydrodynamics model and
was constructed using a hypothetical case study of the Cape Fear River, relies on other, more rigorous models for more complex hydro-
North Carolina (NC), US (Fig. 5). First, the model is developed to si- dynamics. Here, a 1-dimensional kinematic wave is appropriate. Using
mulate flow and the simulated flow at the outflow is compared to ob- the upstream boundary and the parameters, we compared the outflow
servations. Next, hypothetical simulation of solids concentrations is to the matching USGS Gage (02105769), with an R2 = 0.92 and a Nash-
performed, using realistic, but arbitrary boundary concentrations. The Sutcliffe Efficiency = 0.91. The simulated average travel time is 5 days,
solids state variables are: sand, silt, clay, and particulate organic matter with a range of 2–40 days. Further details on simulating flow are pre-
(POM). Then, a load of a solute chemical and then a nanomaterial are sented in the Appendix, further details on using WASP8 for stream
added, representing a hypothetical release from a waste water treat- routing are available in Ambrose et al. (2017).
ment plant or industrial outflow. The solute chemical transforms into a Once stream flow is running, solids and DOC are simulated. Solids
second state variable, and the nanomaterial attaches to the different are simulated using two different approaches: the descriptive option
solids forming additional state variables (i.e., nano-silt, nano-clay, using constant settling and resuspension, and the mechanistic option,
nano-POM). Additionally, light intensity in the water column and DOC using dynamic settling, deposition, erosion, and resuspension.
are simulated explicitly. Physical-chemical properties of solute and Suspended solids have an upstream boundary condition of 6 mg L−1
nanomaterial properties are presented in Table 4. Further details on silt, 4 mg L−1 clay, 2 mg L−1 POM, and 8 mg L−1 DOC. Initial con-
solids, solutes, and nanomaterials are available in Appendix Tables A7 – centrations of solids are zero in the water column, and sand of
A9. 472,540 g m−3, silt of 250,030 g m−3, clay 356,500 g m−3, and POM
The model domain consists of a 90 km stretch of the Cape Fear River 30,860 g m−3 in the sediment layers. Once the solids are running, a
downstream from USGS Gage 02105500 to USGS Gage 02105769, load of 0.1 kg d−1 of the nanoparticle is added to the first segment. The
consisting of 45 water column segments, each 2 km long and 92 m wide, nanoparticle is set to attach to silt, clay, and POM particles via het-
overlying 45 surface sediment segments (0.05 m depth), and 45 sub- eroaggregation kinetics, forming new state variables: nano-silt, nano-
surface sediment segments (0.15 m depth). That is, each 2 km reach of clay, and nano-POM. These particle-attached nanoparticles then follow
the river consists of three stacked layers: water column, surface sedi- the processes of the solid to which they are attached. The non-attached
ment, and subsurface sediment. Upstream of this section, the Cape Fear nanoparticle photo-degrades. Light is simulated using the option in

11
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

Fig. 8. Simulation Results for Nanoparticles. Concentrations of nanoparticles in different phase (Freely dissociated nanoparticles (Free Nano), nanoparticle
associated with silt (Nano-Silt), nanoparticle associated with clay (Nano-Clay), and nanoparticle associated with POM (Nano-POM) presented over time for segment
20 (midstream) and 40 (downstream) for the water column (top row), surface sediment layer (middle row), and deep sediment layer (bottom row). Descriptive Solids
model option are presented on the left two columns, and Mechanistic Solids model option are presented in the right two columns. Concentrations are presented using
area plots, so that the total concentration of nanoparticles is presented as the sum of all the concentrations.

which WASP8 calculates solar radiation based on the latitude and ranged from 0.999 to 1.0.
longitude of the Cape Fear location, providing changes in surface light
intensity with time of day and day of year. Next, a load of 10 kg d−1 of
Chemical 1 is added to the first segment; Chemical 1 biodegrades into 10.1. Solids
Chemical 2. Both solute chemicals volatilize, attach to solids assuming
equilibrium partitioning, and complex with DOC. Initial concentrations The Cape Fear River WASP model was run using two different ap-
of solute chemicals, nanoparticles, and DOC are zero everywhere. The proaches for simulating solids: a descriptive approach and a mechan-
simulation was run for 15 years. istic approach, as described in Section 4. This is done to demonstrate
Further details on parameterization are provided in the Appendix. the different results that one would expect with these different ap-
The simulations presented here are not meant to provide a complete proaches. The simulation results are presented in Fig. 6. Due to the high
representation of all WASP8 capabilities; but rather, these example si- settling velocity of sand, there is no sand observed in the water column.
mulations are meant to provide details on some of the processes In the descriptive solids simulation, in the water column, silt, clay, and
available in WASP8 and information on how one could construct a POM fluctuate as the flow and velocity of the segment fluctuate. Clay is
model depending on the questions to be answered and the goals to be the dominant solid type in the water column, as it has the slowest
met. This example, simulates four solid particle types, where WASP7 settling velocity. The solid concentrations in the surface sediment layer
could only simulate 3. Nanoparticles are a completely new state vari- remains relatively constant, with silt slightly increasing as it settles
able, which can form 3 other distinct other state variables as they attach from the water column to the sediment layer. The subsurface sediment
to solid particles. Further, the enhanced light and transformation rou- layer remains relatively constant throughout the duration of the si-
tines are incorporated. As part of its regular simulations, WASP8 per- mulation. For the mechanistic solids, the water column concentrations
forms a mass check to confirm that mass balance is being upheld, which of silt, clay, and POM vary over the course of the simulation. Generally,
WASP outputs live to the screen as it runs as well as part of its output the water column concentrations are lower than for the descriptive
files. A perfect mass balance = 1. For these simulations, the mass check solids option, reflecting a deposition-dominated process where re-
suspension is small or negligible. The spikes in water column solids

12
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

Fig. 9. Simulation Results for Chemicals. Concentrations of the two chemicals presented over time for segments 10 (upstream), 20 (upper midstream), 30 (lower
midstream), and 40 (downstream) for the water column (top row), surface sediment layer (bottom row). The Descriptive Solids model option was used.
Concentrations are presented using area plots, so that the total concentration of chemicals is presented as the sum of both concentrations.

concentration occur during times of high bottom shear stresses, which 10.3. Nanoparticles
result in erosional events. In response to the increased solids con-
centrations in the water column, the sediment layer loses mass of the Fig. 8 presents concentrations of the different phases of nano-
finer particles of silt, clay, and POM. WASP8 then moves solids from the particles in two representative segments, Segment 20 and 40. WASP
deep sediments to replace the lost solids, which results in enriching the simulates concentrations in all segments for all time steps, and these
surface sediment layer with sand. This result is reflected by the jump in two segments are presented for illustration. The descriptive and me-
sand concentration during the largest erosional event. This comparison chanistic solids options are both presented to provide an example for
of the simulated solids concentration trends using the descriptive and how these choices can affect the fate and transport of the nanoparticles.
mechanistic approaches illustrates the fact that the mechanistic ap- In the water column, the total nanoparticle concentrations vary in-
proach results in greater variation in suspended solids concentrations as versely with flow, as the load is held constant and flow varies over time.
a function of time. The nanoparticle is present predominantly in the free, non-attached
form (Free Nano), and concentrations decrease moving downstream as
10.2. Light nanoparticles attach and settle to the sediments. In the sediment layer,
the dominant form is the nanoparticle attached to clay (Nano-Clay),
Fig. 7 shows an example of light simulation in WASP8. The ten followed by nanoparticle attached to silt (Nano-Silt), with a small
different bands of light attenuate differently through the water column. fraction attached to POM (Nano-POM). This is the case for both the
The light intensity at noon is shown for all 10 wavelength bands for descriptive solids and mechanistic solids case. For the descriptive solids,
each day for the entire year for the length of the simulation. The total the concentrations in the sediment steadily accumulate over time. In
light intensity changes over the course of the year, highest in summer the mechanistic solids case, there is less deposition to the sediment
and lowest in winter. The fraction of the total light intensity is de- layer, and thus the concentrations accumulate less. Halfway through
monstrated by the thickness of each band. As the concentration of solids the simulation, there is an erosional event. As seen in Fig. 6, halfway
changes in the water column, the amount of attenuation of each band of through the simulation, there is a spike of solids in the water column
light changes. The total photoreaction rate constant is the sum of the and a jump in the surface sediments solid concentration. At the same
contributions due to each wavelength band. The simulation is para- time, the concentration of Nano-Clay in the surface sediments decreases
meterized to have higher wavelength-specific base rate constants for suddenly. Following this event, the concentration of Nano-Clay in-
UVB with decreasing rate constants as the wavelength increases. This creases. For the deep sediment layer, the descriptive solids option
translates to the pattern observed in Fig. 7, where the shorter wave- shows an accumulation of nanoparticle concentrations, increasing
lengths have a larger contribution to the total photoreaction rate con- slower than in the surface layer. For the mechanistic solids option, there
stant, even though they comprise a smaller fraction of the total light is minimal deposition from the surface sediments to the deep sediment
intensity. For each time step and each location, WASP is calculating the layer in this example application; therefore, no nanoparticles enter this
attenuation of light – based on the concentrations of particles and DOC layer.
and depth – and then calculating the total reaction rate constant, which
is then applied to any state variable with a light driven process.

13
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

10.4. Solute chemicals Guide: Supplement to Water Quality Analysis Simulation Program (WASP) User
Documentation. US Environmental Protection Agency, Office of Research and
Development, Washington, DC.
Fig. 9 presents the concentrations of the two chemicals for the Ambrose Jr., R.B., Avant, B., Han, Y., Knightes, C.D., Wool, T.A., 2017. Water Quality
mechanistic solids option, moving downstream in Segment 10, 20, 30, Assessment Simulation Program (WASP8): Upgrades to the Advanced Toxicant
and 40. These figures focus on the mechanistic solids to illustrate how Module for Simulating Dissolved Chemicals, Nanomaterials, and Solids. U.S.
Environmental Protection Agency, Washington, D.C. EPA/600/R-17/326.
the concentrations change moving along the length of the river. In this Arnot, J.A., Gobas, F.A.P.C., 2004. A food web bioaccumulation model for organic che-
scenario, Chemical 1 biodegrades into Chemical 2. Thus upstream, the micals in aquatic ecosystems. Environ. Toxicol. Chem. 23, 2343–2355.
concentration of Chemical 1 is larger than Chemical 2. Moving down- Arvidsson, R., Molander, S., Sandén, B.A., Hassellöv, M., 2011. Challenges in exposure
modeling of nanoparticles in aquatic environments. Hum. Ecol. Risk Assess. 17,
stream, Chemical 2 increases in concentration as Chemical 1 decreases. 245–262.
The model is structured using equilibrium partitioning of the chemicals Barber, M.C., 2006. Bioaccumulation and Aquatic System Simulator (BASS) User's Manual
and the solid phases, with Chemical 1 being more hydrophilic than Beta Test, Ver 2.2. U.S. Environmental Protection Agency, Office of Research and
Development, Athens, GA.
Chemical 2, The difference in hydrophilicity is evidenced in that Che-
Bicknell, B.R., Imhoff, J.C., Kittle Jr., J.L., Jobes, T.H., Donigian Jr., A.S., 2001.
mical 2 increases in concentration in the sediments moving down- Hydrological Simulation Program - FORTRAN (HSPF), User's Manual for Version
stream. 12.0. US Environmental Protection Agency, Athens, GA.
Bouchard, D., Knightes, C.D., Chang, X., Avant, B., 2017. Simulating multiwalled carbon
nanotube transport in surface water systems using the Water Quality Analysis
11. Conclusions Simulation Program (WASP). Environ. Sci. Technol. 51, 11174–11184.
Camacho, R.A., Martin, J.L., Wool, T., Singh, V.P., 2018. A framework for uncertainty and
The WASP8 Advanced Toxicant module has undergone a significant risk analysis in Total Maximum Daily Load applications. Environ. Model. Software
101, 218–235.
redesign and upgrade from previous WASP versions. This upgrade in- Carroll, R.W.H., Warwick, J.J., Heim, K.J., Bonzongo, J.C., Miller, J.R., Lyons, W.B.,
cludes the introduction of more state variables and different types of 2000. Simulation of mercury transport and fate in the Carson River, Nevada. Ecol.
state variables, particularly with the introduction of the nanoparticle Model. 125, 255–279.
Cerco, C.F., Tillman, D., Hagy, J.D., 2010. Coupling and comparing a spatially- and
state variable, and the expansion of how light attenuation and photo- temporally-detailed eutrophication model with an ecosystem network model: an in-
reaction processes are handled, with different options for simulating itial application to Chesapeake Bay. Environ. Model. Software 25, 562–572.
solids, which are a critical aspect of toxicant fate and transport mod- Cole, T.M., Buchak, E.M., 1994. CE-QUAL-W2: a Two-dimensional, Laterally Averaged,
Hydrodynamic and Water Quality Model, Version 2.0, User Manual (Draft
eling. Over the years, WASP has been increasingly used to investigate Instruction Report, Vicksburg, MS). .
and address different contaminant fate and transport problems. We Edinger, J.E., Brady, D.K., Geyer, J.C., 1974. Heat Exchange and Transport in the
hope that with these advances in the toxicant module that WASP8 will Environment. Report No. 14. EPRI Publication No. 74-049-00-34. Electric Power
Research Institute, Palo Alto, CA.
provide a modeling framework for a range of users that is flexible and
Gargallo, S., Solimeno, A., Martín, M., 2018. Which are the most sensitive parameters for
rigorous enough to address the expanding range of problems posed by suspended solids modelling in free water surface constructed wetlands? Environ.
both historic and emerging contaminants. Model. Software 102, 115–119.
Hou, W.-C., Chowdhury, I., Goodwin, D.G., Henderson, W.M., Fairbrother, D.H.,
Bouchard, D., Zepp, R.G., 2015. Photochemical transformation of graphene oxide in
Disclaimer sunlight. Environ. Sci. Technol. 49, 3435–3443.
HydroQual, I., 2007. A Primer for ECOMSED, Version 1.4 (December). .
This paper has been reviewed in accordance with the U.S. Johnston, J.M., McGarvey, D.J., Barber, M.C., Laniak, G., Babendreier, J., Parmar, R.,
Wolfe, K., Kraemer, S.R., Cyterski, M., Knightes, C.D., Rashleigh, B., Suarez, L.,
Environmental Protection Agency's peer and administrative review Ambrose Jr., R.B., 2011. An integrated modeling framework for performing en-
policies and approved for publication. Mention of trade names or vironmental assessments: application to ecosystem services in the Albemarle-Pamlico
commercial products does not constitute endorsement or re- basins (NC and VA, USA). Ecol. Model. 222 2741-2484.
Jones, C., Lick, W., 2001. Contaminant flux due to sediment erosion. In: Proceedings of
commendation for use. The views expressed in this article are those of the 7th International Conference: Estuarine and Coastal Modeling, pp. 280–293.
the authors and do not necessarily represent the views or policies of the Karickhoff, S.W., 1984. Organic pollutant sorption in aquatic systems. J. Hydraul. Eng.
US EPA. 110, 707–735.
Knightes, C.D., 2010. Modeling Mercury Transport and Transformation along the Sudbury
River, with Implications for Regulatory Action. US Environmental Protection Agency,
Acknowledgements Office of Research and Development, National Exposure Research Laboratory,
Athens, GA.
Knightes, C.D., Sunderland, E.M., Barber, M.C., Johnston, J., Ambrose Jr., R.B., 2009.
We would like to acknowledge the efforts of Lourdes Prieto for
Application of ecosystem-scale fate and bioaccumulation models to predict fish
gathering the data necessary for constructing the Cape Fear River WASP mercury response times to changes in atmospheric deposition. Environ. Toxicol.
model and the helpful review and comments by Caroline Stevens and Chem. 28, 881–893.
Craig Barber as well as by an anonymous reviewer. Krone, R.B., 1963. A Study of Rheologic Properties of Estuarial Sediments. U.S. Army
Corps of Engineers.
Lick, W., Lick, J., Ziegler, C.K., 1994. The resuspension and transport of fine-grained
Appendix. Supplementary data sediments in Lake Erie. J. Great Lake. Res. 20, 599–612.
Little, K.W., Parks, A.B., Lillys, T.P., 2018. An open-source, generic, environmental model
for chemical fate and transport analysis in multi-media systems. Environ. Model.
Supplementary data to this article can be found online at https:// Software 103, 90–104.
doi.org/10.1016/j.envsoft.2018.10.012. Martin, J., Ambrose Jr., R.B., Wool, T.A., 2006. WASP7 Benthic Algae - Model Theory and
User's Guide - Draft. U.S. Environmental Protection Agency.
Morel, F.M.M., Hering, J.G., 1993. Principles and Applications of Aquatic Chemistry.
Funding John Wiley & Sons, Inc., New York.
Nietsch, S., Arnold, J., Kiniry, J., Williams, J., 2011. Soil and Water Assessment Tool
This work was funded through the US EPA’s Office of Research and Theoretical Documentation Vesion 2009. Texas Water Resources Institute Technical
Report 406. Texas A&M University System, College Station, TX.
Development’s Chemical Safety for Sustainability Research Program.
Nguyen, T.T., Keupers, I., Willems, P., 2018. Conceptual river water quality model with
This research did not receive any specific grant from funding agencies flexible model structure. Environ. Model. Software 104, 102–117.
in the commercial or not-for-profit sectors. NRC, 2007. Models in Environmental Regulatory Decision Making. National Academic
Press, Washington, DC.
Park, J.Y., Park, G.A., Kim, S.J., 2013. Assessment of future climate change impact on
References water quality of Chungju Lake, South Korea, Using WASP coupled with SWAT. J. Am.
Water Resour. Assoc. 49, 1225–1238.
Ambrose Jr., R.B., Wool, T.A., Martin, J.L., 1993. The Dynamic Estuary Model Praetorius, A., Scheringer, M., Hungerbühler, K., 2012. Development of environmental
Hydrodynamics Program, DYNHYD5. Model Documentation and User Manual. US fate models for engineered nanoparticles: a case study of TiO2 nanoparticles in the
Environmental Protection Agency, Athens, GA. Rhine River. Environ. Sci. Technol. 46, 6705–6713.
Ambrose Jr., R.B., Wool, T.A., 2017. WASP8 Stream Transport - Model Theory and User's Quik, J.T.K., de Klein, J.J.M., Koelmans, A.A., 2015. Spatially explicit fate modelling of
nanomaterials in natural waters. Water Res. 80, 200–208.

14
C.D. Knightes et al. Environmental Modelling and Software xxx (xxxx) xxx–xxx

Roberts, J., Jepsen, R., Gotthard, D., Lick, W., 1998. Effects of particle size and bulk Chemical Equilibria in Natural Waters. John Wiley & Sons, Inc., New York.
density on erosion of quartz particles. J. Hydraul. Eng. 124. van Rijn, L.C., 1984. Sediment transport, Part II: suspended load transport. J. Hydraul.
Sadeghian, A., Chapra, S.C., Hudson, J., Wheater, H., Lindenschmidt, K.-E., 2018. Eng. 110, 1613–1641.
Improving in-lake water quality modeling using variable chlorophyll a/algal biomass van Rijn, L.C., 2007. Unified view of sediment transport by currents and waves. I: in-
ratios. Environ. Model. Software 101. ititaion of motion, bed roughness, and bed-load transport. J. Hydraul. Eng. 649–667.
Schnoor, J.L., 1996. Environmental Modeling: Fate and Transport of Pollutants in Water, Wang, P.F., Martin, J., Morrison, G., 1999. Water quality and eutrophication in Tampa
Air, and Soil. John Wiley & Son, Inc., New York. Bay, Florida. Estuar. Coast Shelf Sci. 49, 1–20.
Schwarzenbach, R.P., Gschwent, P.M., Imboden, D., 1993. Environmental Organic Whelan, G., Kim, K., Pelton, M.A., Castleton, K.J., Laniak, G.F., Wolfe, K., Parmar, R.,
Chemistry. John Wiley & Songs, Inc., New York. Babendreier, J., Galvin, M., 2014. Design of a component-based integrated en-
Seo, D., Kwon, K.H., Lee, Y.S., 2009. 3-D hydrodynamic modeling of Yongdam lake, Korea vironmental modeling framework. Environ. Model. Software 55, 1–24.
using EFDC. Desalination Water Treat. 19. Wool, T.A., Ambrose Jr., R.B., Martin, J., 2013. WASP7 Multiple Algae - Model Theory
Seo, D., Kim, M., Ahn, J.H., 2012. Prediction of chlorophyll-a changes due to weir con- and User's Guide. US Environmental Protection Agency.
structions in the Nakdong river using EFDC-WASP modelling. Environ. Eng. Manag. Wool, T.A., Davie, S.R., Rodriguez, H.N., 2003. Development of three-dimensional hy-
17, 95–102. drodynamic and water quality models to support total maximum daily load decision
Sharma, D., Kansal, A., 2013. Assessment of river quality models: a review. Rev. Environ. process for the Neuse River estuary, North Carolina. J. Water Resour. Plann. Manag.
Sci. Biotechnol. 12, 285–311. 129, 295–306.
Stumm, W., Morgan, J.J., 1981. Aquatic Chemistry: an Introduction Emphasizing

15