Section 5

Section 5
Construct UPM Tools
From Section 4, Data Collection
Construct modelling
Construct modelling
tools
tools
To Section 6. Assessing Performance

Section 5: Updated August 2018
5. SECTION 5: CONSTRUCT UPM TOOLS
This section is divided into a series of subsections that provide information on the types
of modelling tools that may be required to carry out a UPM study. Each subsection
covers a particular generic type of modelling and identifies typical applications, model
requirements and modelling approaches based on the use of a range of simple to
complex modelling tools for each type of application. Specific modelling tools and
software products are not identified. The issues related to the building, calibration,
verification of each type of modelling tool are considered in relation to ‘fitness for
purpose’. Generic limitations of current modelling tools are considered.
The section is subdivided as follows.
5.1 Introduction to modelling tools.
5.2 Rainfall modelling.
5.3 Sewer flow and quality modelling.
5.4 Sewage treatment works quality modelling.
5.5 River impact modelling.
5.6 Tidal waters modelling.
5.7 Integrated urban pollution modelling.
5.1 Introduction to modelling tools

The application of simulation modelling tools forms the key activity in a UPM study. Modelling
tools are used to predict the performance of urban wastewater systems and to assess the
resulting environmental impact of wet weather discharges against appropriate environmental
criteria, as discussed in Section 2. A range of modelling tools are available to cover all
components of a UPM study. Section 3 considered the selection of appropriate models and
modelling capability, in relation to the requirements of a site specific application. This section
considers the generic types of modelling tools that may be required to implement a study in
terms of wastewater transport and treatment, and environmental impact assessment in river,
estuary and marine receiving waters. These types are:
• rainfall modelling;
• sewer flow and quality modelling;
• sewage treatment works modelling;
• river impact modelling;
• estuary impact modelling;
• marine impact modelling; and,
• integrated urban pollution modelling.
Each subsection considers the issues related to the type of modelling and models available in
relation to the requirements for typical applications of the UPM Procedure. Attention is
focused on making models ‘fit for purpose’ for subsequent modelling scenarios, through the
stages of model building, calibration and verification. Applications issues are also considered
in relation to the selection of boundary conditions and initialisation, where appropriate.
Section 5 - 1
Specific modelling tools and software products are not identified, therefore, data requirements
and modelling approaches are illustrated in general terms. Users should select a particular
model in the light of the requirements of the study identified in Section 3.
5.2 Rainfall modelling
By definition, rainfall is the main driving force affecting the wet-weather performance of the
receiving environment, as the rainfall impacts on the discharges from sewers and STWs
diffuse inputs and also run-off from natural surfaces. The way in that rainfall is represented,
or modelled, is of crucial importance to the correct understanding of wet-weather problems
and the subsequent development and testing of solutions.
This section describes the selection and use of long rainfall time series in UPM studies, and
is subdivided as follows:
5.2.1 Rainfall requirements for UPM modelling;
5.2.2 Historical rainfall time series;
5.2.3 Synthetic rainfall time series;
5.2.4 Event definition.
5.2.1 Rainfall requirements for UPM modelling
Rainfall data are required to drive the models that simulate the wet weather performance of
urban wastewater systems and depending on the amount of detail in the study, the receiving
environment. Long time series of rainfall events are needed to provide as full an account as
possible of the variability in rainfall over the chosen time period and to give the user maximum
flexibility in making different event selections depending on the question being addressed.
The effort required in ensuring the reliable representation of rainfall inputs is likely to be
considerably less than that required for other modelling components. However, rainfall
modelling cannot be overlooked given the crucial importance of rainfall in gaining an accurate
understanding of wet-weather performance.
Variability of rainfall
UPM modelling studies require long series of rainfall data that are representative of the local
catchment. The length of the record is important to ensure that the data do not represent a
particularly wet or dry period. A record of at least ten years duration is recommended for UPM
applications and consideration should be made to extending this to 25 years if there is
evidence of significant variation across a 25 year period. The various elements of UPM
modelling require rainfall in a range of formats. The main requirements are for values in
hourly or shorter timesteps for deterministic modelling. Five minute timestep data has been
used, and some studies have used 15 minute timestep data with no significant detriment to
modelled sewer system performance. Long time series of hourly rainfall data are generally
needed as input for simplified sewer flow models.
Spatial variability of rainfall may require the application of more than one time series for large
catchments. This is more easily made use of by using historic rainfall data if sufficient long
term gauges are available.
Section 5 - 2
5.2.2 Historical rainfall time series

Historic rainfall data can be obtained from either raingauge data or more recently rainfall
radar data.
Digitised rainfall data from raingauges are available for a number of UK sites and may be
obtained from the Meteorological (Met.) Office or some other local sources, such as the
municipality, environmental regulator or local water company. The timestep of the datasets
will range between short timestep data of minutes up to daily rainfall records. For direct use in
deterministic models generally, the timestep of data will need to be short, and not more than
15 minutes. Some simpler lumped catchment models can use data at hourly timesteps.
There are a number of factors that need to be taken into account when using raingauge data,
including:
• Selection of the most appropriate gauges;

• Spatial variability of rainfall across large catchments;
• Length of rainfall record; and,
• Amount of missing data and accuracy of gauge data.
Selection of the most appropriate raingauge site to represent the catchment to be modelled
needs to take account of the distance of the gauge from the catchment in question, the
elevation of the gauge compared to the catchment, and the appropriateness of the annual
rainfall at the site compared to the catchment.
It may be possible to use more than one raingauge to provide rainfall inputs, dependant of
course on the availability of data when undertaking analysis of large catchments,. This will
allow an element of spatial variation of rainfall across the catchment. The advantage of using
historical data for this purpose is that the data is dated.
The rainfall record length needs to be long enough to take account of the inherent variability
of rainfall. Both intensity and duration of rainfall is important in assessing discharges from
CSOs and storm tanks and some discharges are very sensitive to changes in antecedent
conditions. Hence, a wide a range of rainfall data is required over as long a time period as
possible. If a relatively short rainfall record is available, reviews should be undertaken of long
rainfall records to ensure the period of the rainfall record is consistent with long term trends.
This could be done, for example, by comparison with a daily rainfall record. Historical data
should always be screened and checked for missing data values.
Although use of radar rainfall data resolves any spatial variation issues, the duration of the
available rainfall radar data set usually precludes its use in UPM analysis. If there is a
suitable duration of data available, checks should be made on the perceived accuracy of the
data when compared to raingauge data and missing data.
5.2.3 Synthetic rainfall time series

Although there are more short timestep raingauge data available for longer periods, there are
occasions when either there are insufficient suitable rainfall sites, or suitably long records
available for the use of historic data. The use of synthetic rainfall data will be required in these
instances.
There are various stochastic rainfall generators available that will generate long rainfall times
series containing all of the characteristics of historical data for any location in the UK. Input
data requirements to generate a series are generally related to parameters such as average
annual rainfall, grid reference, altitude and distance from the coast. Generally, these can also
be seeded with measured rainfall data.
Section 5 - 3
The stochastic rainfall series generated needs to be at a suitable timestep, generally hourly or
5 minute dependent on the type of modelling tool being used.
Any synthetic rainfall series generated should be checked against local rainfall records. A
comparison between the synthetic series and the local rainfall record should be undertaken
for:
• average and range of annual rainfall totals;

• average and range of monthly rainfall totals; and,
• frequency distribution of daily rainfall totals.
It is important that any future changes are evaluated and incorporated as part of the agreed
environmental planning framework (Section 3), particularly in relation to climate change.
5.2.4 Event definition
Unless continuous simulation modelling is to be employed, a major first step before using any
long rainfall series (either historical or synthetic) is to produce an event file that contains only
the events required for subsequent analyses. The definition of an ‘event’ is based on the dry
period separating events. This is open to some interpretation and should be considered on a
catchment by catchment basis. The definition of a suitable inter-event dry period should be
related to the time taken for the sewerage system to return to baseflow conditions, including
the emptying of any upstream storage. However, event definition will also be driven by how
calculation of failures against standards is analysed. If this is carried out on an individual
event basis, the use of very long rainfall events may cause under-prediction of failures.
During the event definition process, the antecedent conditions and catchment wetness
parameters can be calculated for each discrete rainfall event. These parameters are required
for estimating run-off in sewer flow models.
5.3 Sewer flow and quality modelling
This section addresses the role played by sewer flow and sewer quality models (SFMs and
SQMs) in UPM studies. The main processes that affect the pollutant loads discharged from
sewer systems during wet weather periods are considered. Also, the principles of sewer
quality modelling are discussed.
This section is subdivided as follows.
5.3.1 Hydraulic performance of sewer systems.
5.3.2 Sewer quality performance under wet weather conditions.
5.3.3 Modelling sewer quality performance.
5.3.4 Determinands modelled.
5.3.5 Using SQMs.
5.3.1 Hydraulic performance of sewer systems
An adequate understanding of how a sewer system performs hydraulically is a basic

requirement for any UPM study. It is essential to be able to predict how flows will vary during
Section 5 - 4
storms before any sensible predictions of the spatial and temporal variations in pollutant
concentrations and loads can be made.
a) Detailed sewer flow models (SFMs)
Detailed SFMs require a representation of the pipe network together with information
about areas and populations connected to each pipe. The level of detail of this
representation can be varied depending on the accuracy required. Typically, surface run-
off hydrographs can be produced from a specified rainfall profile and then routed through
the modelled pipe network. Flows and depths are calculated throughout the network at
each timestep and surcharge and flooding at manholes are predicted. In addition,
ancillaries such as CSOs, tanks and pumping stations can also be represented.
The simulated performance must be checked to ensure that it accords with actual
performance before a SFM is used to design major upgrading measures on an existing
system. This checking process is called verification and may involve evaluation against
historical data (for example, flooding records) and against specific field measurements of
flows and depths during storms (CIWEM Urban Drainage Group, (2002)).
A SFM should be used to represent not only the combined sewer system but also any
major separate storm water systems in the urban catchment area. This will help to
ensure that all flows and loads discharged to the river during a storm event are
accounted for.
b) Simple sewer flow models (SFMs)
In certain situations it may be beneficial to represent CSO performance in a sewer

system without building a detailed SFM. Such situations will be characterised by relatively
small catchments where there is little interaction between CSOs and where the
magnitude of the continuation flow can be estimated with some confidence (e.g. where it
is controlled by a pump). Reasonable estimates of spill volumes can be made using
simple tank models, that do not specifically model pipe hydraulics. Instead, the flow
processes are represented by a number of tanks in series and in parallel. Each tank
receives foul flows and run-off from a different subcatchment. One of the principal
characteristics of this type of model is the ability to simulate multiple events or long
chronological rainfall sequences rapidly. The same effect could be achieved by heavily
simplifying a detailed SFM.
5.3.2 Sewer quality performance under wet weather conditions
The pollutant loads carried by (and discharged from) sewer systems during wet weather vary
in a complex manner. This is illustrated by the pollutographs shown in Figures 5.1 and 5.2.
Many different processes are involved. These processes need to be understood qualitatively,
and the most significant processes must be quantified to enable reliable predictions to be
made. Equally important are the dry weather processes that contribute to the build-up of
sediments on catchment surfaces and in sewers. Such deposits form an important source of
pollutants that can be deposited in dry weather and subsequently mobilised by the rainfall and
the resulting higher flows that occur during wet weather.
The main processes influencing the quality of flows in combined sewers are:
• foul inputs;
• build-up and wash-off of sediments from the catchment surface;
• deposition and erosion of sewer sediments;
• sediment transport in sewers;
• sediment partitioning in tanks;
Section 5 - 5
• advection and dispersion of pollutants; and,

• biochemical reactions.
Figure 5.1 Example of observed pollutographs in sewer flow - Site 1
Section 5 - 6
Figure 5.2 Example of observed pollutographs in sewer flow - Site 2
Section 5 - 7
a) Foul inputs
Foul inputs include the domestic, commercial and industrial inputs to sewers. The flows
and pollutant loads vary both spatially across a catchment and temporally (diurnal
variation and variation from day to day). Typical diurnal variation is illustrated in
Figure 5.3.
b) Build-up and wash-off of sediments from the catchment surface
Sediments build up on roads, roofs, pavements and in gully pots during dry weather
periods. The quantity and characteristics of these sediments will depend on many
factors, including the different land uses and on length of the dry weather period. Various
pollutants will be associated with these sediments. These sediments and attached
pollutants are washed off the catchment surfaces during storm events and enter
combined sewer systems and separate storm water systems. The quantities washed off
will depend on the intensity of the rainfall and the erosion capability of the overland flow,
as well as the actual quantities available on the surfaces. The quantity available on the
surfaces will be dependent on the length of time between rainfall events.
c) Deposition and erosion of sewer sediments
Suspended sediments tend to settle out of the sewer flow and deposit on the bed of the
sewer during periods of low flow and at locations where flow velocities are low. The
deposition process depends on many factors, including the size and density of the
sediment particles and the flow regime in different parts of the sewer system. Coarser,
denser sediments, that may be derived from catchment surfaces, tend to deposit more
readily than the finer organic sediments that are derived from foul inputs. The deposited
sediments can have attached pollutants (e.g. BOD) and act as a “store” of pollutants
within a sewer system.
As flows increase, either as part of the normal diurnal dry weather cycle, or during storm
periods, deposited sediments may be eroded. The point at which erosion begins and the
subsequent erosion rate depends, among other things, on the flow velocity, the width of
the sediment bed and the characteristics (in particular the shear strength) of the
sediment deposits. As sediment is eroded, interstitial water and associated pollutants are
released into the sewer flow. Sediments and associated pollutants may be deposited and
re-suspended many times during their passage through a sewer system.
d) Sediment transport in sewers
Once sediment has been eroded and entrained in the flow, it may be transported down a
sewer either in suspension or as bed load. Lighter material tends to travel in suspension
and the heavier material as bed load. The mode of transport is strongly influenced by the
flow velocity and the degree of turbulence, that will vary in space and time in the sewer
system.
Section 5 - 8
Figure 5.3 Typical diurnal profiles for dry weather flow and quality in foul sewers
Section 5 - 9
e) Sediment partitioning in tanks
One of the effects of detention tanks in sewer systems is to reduce the local flow velocity.
This in turn encourages suspended sediment to settle and, as a consequence, the
concentration of suspended sediment tends to be less in the spill flow from a tank than in
the tank itself. This partitioning effect helps to reduce the polluting impact of spills. The
deposited sediments may eventually be re-entrained from the base of the tank and
conveyed to the STW or a downstream tank. This will create a short term peak loading
rate at the end of a storm when storage tanks are full and may result in a high pollutant
concentration spill to the environment.
f) Advection and dispersion
Dissolved pollutants and fine suspended sediments are moved through a sewer system
by the process of advection and dispersion. Advection is the movement of pollutants and
suspended sediments in the same direction and at the same speed as the flow. This is by
far the main pollutant transport process.
Dispersion, on the other hand, is a more random process whereby pollutants and
suspended sediments move in such a way as to reduce concentration gradients along a
sewer. During a storm, and to a lesser extent during dry weather flow, concentration
gradients will build up due to the varying input loads and the resuspension of sediment
deposits. Turbulence within the flow regime will tend to disperse the pollutants so as to
reduce these concentration gradients.
g) Biochemical reactions
Finally, some pollutants can undergo biochemical reactions (for example, the decay of
BOD and ammonia), during their transit through a sewer system. Hence, the load of
some non-conservative pollutants can be reduced within the sewer system prior to
discharge. However, this will reduce the dissolved oxygen level in the sewage that may
increase the environmental impact of a spill.
5.3.3 Modelling sewer quality performance
Various approaches can be used to model the performance of sewer systems in terms of the
pollutant load discharged during wet weather. These approaches can be divided into three
groups:
• simple tank simulation models;

• detailed SFMs plus event mean spill concentrations; and,
• detailed dynamic SQMs.
These are discussed in more detail in the CIWEM Urban Drainage Group publication "GUIDE
TO THE QUALITY MODELLING OF SEWER SYSTEMS", Version1 November 2006.
5.3.4 Determinands Modelled
The determinands that are usually modelled by SQMs are BOD, COD, ammonia, and
suspended solids (metals and bacteria can also be modelled). Suspended solids are more
generally referred to as sediment, and include the solids that are deposited in the sewer
system. Sediment is divided into sediment fractions (normally two: fine and coarse) each
characterised by particle size, density and settling velocity. Different sediment stores (e.g.
road, roof, pipe) are described by the sediment types they contain. A sediment type is a
particular mixture of fine and coarse sediment fractions with other characteristics, such as
Section 5 - 10
shear strength, wet bulk density and moisture content. Each sediment fraction in a sediment
type is also given a potency factor that expresses the amount of pollutant attached to a mass
or volume of deposited sediment. When the sediment moves, the pollutant attached to the
sediment also moves.
The pollutants modelled can be assumed to be conservative in that no biochemical changes

will occur during their transport through the system. Alternatively, more complex biochemical
interactions can be modelled to represent the decay of pollutants as they are advected
downstream over long distances.
b) Processes modelled
SQMs typically contain submodules to represent:
• foul inputs;
• surface washoff;
• pollutant and sediment behaviour in pipes; and,
• pollutant and sediment behaviour in tanks.
Most dynamic SQMs estimate foul inputs from domestic sources based on the different
land uses assigned to subcatchment areas in the model. Diurnal variations in flow and
quality are included.
The representation of surface wash-off varies between dynamic models. Generally,

however, the wash-off of sediments and pollutants is a function of rainfall intensity either
for unlimited amounts of sediment or for a set depth of sediment on the surface of the
transport source. Gully pot processes may also be included.
Section 5 - 11
Figure 5.4 Example of spill pollutographs produced by dynamic sewer quality model - Site 1
Section 5 - 12
Figure 5.5 Example of spill pollutographs produced by dynamic sewer quality model - Site 2
Section 5 - 13
Various formulations are available for modelling transport processes (in particular, sediment
transport) in sewers. Generally, it is assumed that pollutants and suspended sediments are
moved in the flow by advection. Dispersion may be ignored, as it is considered to be of little
effect, although some SQMs do include the effect of dispersion. A sediment transport model
is used to determine how much sediment is eroded from a defined bed of sediment and
transported by advection through the system. Settling velocities for the sediment particles are
used to define how much deposition takes place. Pollutants attached to the sediment are
transported at the same rate as the sediment. Dissolved pollutants are advected through the
system at the flow velocity rate.
Sewer tanks can be modelled as a simple mixing chamber where no deposition occurs. An
efficiency factor can be used to enable different pass forward and spill concentrations from
the tank to be modelled as would occur if settling had taken place. Alternatively, the process
of sediment and pollutant settling can be represented in the tank to give different sediment
and pollutant concentrations in the spill and pass forward flows.
5.3.5 Using SQMs
The verified SFM should be used to demonstrate the hydraulic performance of the system
before starting to plan the use of the SQM. This can provide several types of information:
• the likely extent of the pollution problems;

• where the data collection for model calibration/verification is most required,
particularly in relation to sediment depth for model initialisation; and,
• the likely solutions to the existing problems.
The way in which the model is initialised should be considered when applying a SQM. System
initialisation involves defining the initial conditions of all pollutants and sediments within the
system, both above and below ground. It also involves defining the sediment depths on the
catchment surface and in the pipes. The build up of sediments in the pipes is related to the
antecedent dry weather period. It is suggested that the system is initialised by running a few
days of dry weather flow through the system, so that indications are made as to where
sediment is likely to deposit. These indications can then be tied to any available site
measurements or local information.
A suitable run time must be defined once the model is initialised. This will depend upon the
use to be made of the model results. If only the period when the CSOs operate is of interest,
a run time can be investigated using the SFM. If the results from the model are being passed
onto a STW model, the SQM needs to be run so that the effects of the storm pass right
through the system until the STW returns to dry weather flow conditions. Care must be taken
to ensure that any storage facilities have been fully emptied.
As well as using the correct run time, it is necessary to optimise the timestep used in the
model. Using a timestep that is too small can sometimes generate instabilities in a model,
especially at structures. A small timestep will also increase the time taken to run the model
such that it may become impracticable. However, using a timestep that is too large can cause
the model to be inaccurate.
The start time of the storm relative to the diurnally varying dry weather flow pattern must also
be considered. There are favourable times and unfavourable times in terms of pollutant
discharge. The model may be used to present an average situation or a worst case situation.
The storm should be timed to coincide with the highest level of dry weather flow to represent
a worst case situation.
Section 5 - 14
Related to this is the use of point source inputs to the system, such as, industrial discharges.
These inputs may be of a continuous nature or they may be intermittent discharges to the
system. The intermittent discharges may occur at set times during the day or may happen at
any time. When the latter is true it is difficult to model the input of these discharges. The
same approach, as described, for dry weather flow, can be used where either a worst case
situation or an average case situation can be represented.
5.4 Sewage treatment works quality modelling
This section discusses how to represent sewage treatment works flow and quality in
UPM studies.
This section is subdivided as follows.
5.4.1 Sewage treatment works performance under wet weather conditions.
5.4.2 Modelling sewage treatment works.
5.4.1 Sewage treatment works performance under wet weather conditions
Typically, sewage works design is based on the treatment of multiples of DWF to achieve
defined standards of effluent quality. The environmental regulator will specify requirements
for treatment capacity in relation to flows arriving at the STW and the definition of DWF . In
England and Wales the treatment standard required usually varies for flows up to a nominal 3
DWF; for flows between 3 and 6 DWF; and, for flows in excess of 6 DWF. All flows up to
3 DWF pass through the full sequence of treatment processes installed on any particular site.
Beyond this there are several possibilities for how and when flows in excess of 3 DWF are
dealt with. Thus, at sites where primary and secondary treatment processes are installed,
flows in excess of 3 DWF (but less than 6 DWF) may be diverted to storm tanks either before
or after primary treatment (usually the former). At some sites, flows in excess of 6 DWF are
diverted directly to the receiving water following preliminary treatment while at others all flows
above 3 DWF pass through storm tanks before discharge. At small works, storm tanks are
not always installed and full treatment is provided for up to 6 DWF, even at sites where
primary sedimentation is not installed.
Sewage treatment works design has traditionally been based on empirical rules that, by
experience, have produced good results in the past. There is the potential for WwTW to
perform differently in wet weather conditions. While well designed STWs operating within
capacity (e.g. up to 3DWF) may not show a significant response to increased storm flows, it
is implicitly understood that performance, as measured by percentage removal of pollutants,
may deteriorate during storm conditions. Such deterioration is often of a short duration, since
prolonged wet weather flow results in a dilution of the influent sewage arriving at the
treatment plant. There is a risk of decreased effluent quality in storm conditions at treatment
plants. Storm flows are usually associated with a ‘first flush’ of foul sewage containing high
concentration of pollutants. Treatment works can be adversely affected by high flows and
pollutant loads in a variety of ways. The effects are a function of the nature of the catchment,
the size of the catchment, the size of the works, the treatment processes installed at the
works and the effluent quality required. Some treatment processes, particularly those with
long retention times, are relatively insensitive to flow and load increases, others such as
‘small footprint’ processes with short hydraulic retention times are easily upset by sudden
increases in load.
Section 5 - 15
The effects of storm flows on a STW can be generally sub-divided into the following
categories.
• Reduction in settling performance - treatment processes that rely on physical

separation of suspended solids by gravity are adversely affected by high flows.
• Reduction in biological removal efficiency - the removal of organic pollutants
(measured as BOD) and specific nutrients such as ammonia, depends on the contact
time between the micro-organisms and the pollutants. Removal efficiency is reduced
during periods of high flow.
• Washout of solids - this is a particular problem with activated sludge processes that
rely on the efficient separation of suspended biomass from treated effluent in a final
settling tank. The effects are similar to those in the first category with the additional
problem that excessive solids loss can result in reduction in biological treatment
efficiency.
• Mechanical problems - high flows during storm periods, particularly ‘first flush’ events
can inundate preliminary processes such as screens and grit removal plant with sewer
debris etc. that may cause operational difficulties.
5.4.2 Modelling sewage treatment works (STW)
It may be necessary to represent STW flow and quality in the UPM process.
The first UPM studies used calibrated sewage treatment works models to derive water quality
data for the STW. This was not particularly successful due to the following reasons:
• the matches to observed performance were not particularly good;

• they were expensive to build and calibrate;
• they did not give the observed variability in water quality in dry weather conditions;
and,
• the variation in wet weather and dry weather performance was generally within the
accuracy bounds of the modelling process.
The preferred approach is to use a log-normal distribution for water quality, based on an
analysis of STW data either collected for operational purposes, or if this is insufficient,
enhanced by additional specific sampling.
As previously stated there is the potential for STWs to perform differently in wet weather
periods. It is therefore necessary to split the datasets between dry and wet conditions, and
check that there is no significant difference in performance between the two sets. If there is,
there will be a need to produce two distributions, one for wet weather and one for dry periods.
There are other mathematical models that are available to describe STW performance, that
fall into several categories. These are:
• detailed mechanistic;
• reduced-order;
• statistical correlation;
• time series; and,
• improved statistical distributions.
a) Detailed mechanistic models
A considerable amount of work has been carried out in the development of mechanistic
models. These are derived from theoretical equations to describe the biological and
physical processes taking place. Although there are recommended default values for the
Section 5 - 16
calibration parameter values required, most of these models require site-specific

calibration. The theoretical basis to dynamic mechanistic models makes them the most
appropriate models to investigate plant operation outside the normal operating
conditions.. If there is the potential for significant changes in STW performance due to
changes in either operating regime or Capital Improvements it may be worthwhile to
develop a detailed mechanistic model.
b) Reduced-order models
Reduced-order or simplified models are developed by simplification of mechanistic

models. For example, nitrification terms are dropped where the sewage works does not
nitrify, or the solids hydrolysis terms are left out where the sewage retention time is long
(e.g. extended aeration). The resulting models are as valid as the full model, as long as
the processes omitted are known to be insignificant. Reduced-order models have been
developed with STW control as the main aim, the model outputs being used as part of a
decision support system.
c) Time series
Time series models are intended for short-term application, backed up with long-term
data. By keeping the time series up-to-date, predictions for the near future can be made
of expected sewage flows and strengths, and expected STW performance. The more
time series data there are to predict seasonal, diurnal, or other variations, the better (in
theory) is the predictive capability of the model. These models have the built-in
assumption that the future will reflect the past. They are most likely to fail in predicting
effluent quality when there are gross changes in the way the plant is configured or is
being operated, or (to a lesser extent) when there are gross changes to the influent
sewage flow or quality.
d) Enhanced Statistical distributions
These models are based on the assumption that the effluent quality behaves as a
random variable. The underlying random distribution has to be selected, but is commonly
assumed to be based on the log-normal distribution. The choice of probability model may
need to account for cross-correlation between determinands. For example, high solids
are usually associated with high BOD, and high flow is also usually associated with high
solids. The probability model may also need to be modified to include historical effects
that account for the fact that the probability of a determinand taking on a value is not
equally likely at a given time, but is affected by past values, hence low values are more
likely if previous values have also been low. Enhanced statistical distribution models have
not been widely used because of the difficulty of creating an appropriate probability
distribution function that includes the features of known effluent quality.
It is recommended that the STW is included in the sewer hydraulic model and flow data
is taken as an output from the model.
5.5 River impact modelling
This section describes the general effects of intermittent wastewater discharges on

river water quality and the role of river quality impact models (RQIMs) for quantifying
these effects. The stages required to make a RQIM ‘fit for purpose’ are identified.
The section is subdivided as follows:
Section 5 - 17
5.5.1 Effects of intermittent discharges on river quality;
5.5.2 Types of modelling approach;
5.5.3 Description of dynamic RQIMs;
5.5.4 Building, calibrating and verifying RQIMs;
5.5.5 Using dynamic RQIMs.
5.5.1 Effects of intermittent discharges on river quality
Wet weather, resulting in intermittent discharges from CSOs, SWOs and storm tanks, plus
poor quality effluents from STWs, commonly affects water quality in many urban rivers. On
occasions this can result in gross pollution of large stretches of urban watercourses. The
main changes in river water quality due to wet weather discharges can include:
• a reduction of DO as a result of:

• degradation of dissolved BOD;
• degradation of BOD attached to sediments;
• resuspension of polluted bed sediments exerting an additional oxygen
demand; and,
• low DO levels in spilled storm sewage; and,
• a rapid increase in river concentrations of ammonia, bacteria, COD, and suspended
sediments, as well as heavy metals and other toxic substances where trade effluent
discharges are present in sewage.
Processes affecting DO levels are illustrated in Figures 5.6 and 5.7.
Figure 5.6 Processes by which organic loads in CSO spills can affect DO levels in
rivers
Section 5 - 18
Figure 5.7 Additional processes affecting DO levels in rivers
The magnitude of the impact will vary from site to site and will be affected by river
characteristics, such as:
• upstream quality and flow;

• channel slope;
• channel geometry and roughness;
• in-river structures;
• pH;
• temperature;
• ecology (macrophytes, algae, fish and invertebrates); and,
• nature of effluent discharged.
Each of these is considered below.
a) Upstream quality and flow
High DO levels and low concentrations of pollutants in upstream river flows will increase
the capacity of a river to assimilate wet weather discharges. Upstream quality will vary
due to:
• natural diurnal and seasonal variations of flow, DO and temperature;

• channel geometry; and,
• the impact of discharges and abstractions.
b) Channel slope
In general, steeper channels create more turbulence, thereby increasing the rate of
oxygen transfer (re-aeration) across the air/water interface. This leads to higher re-
aeration rates and may, depending on the rate of decay of any BOD, result in increased
levels of DO in the river.
Sediments and any BOD attached to these sediments will tend to settle out on the bed in
rivers with flat sections. This will affect the severity and location of the DO sag, as
illustrated in Figures 5.8 and 5.9.
Section 5 - 19
c) Channel geometry and roughness
The channel cross-section and roughness affect the water depth. In general terms, a
shallower water depth will induce more turbulence by increasing the effect of bottom
roughness and, therefore, allowing greater opportunity for re-aeration to occur.
d) Structures
Structures (weirs, bridges, culverts etc.) will generally reduce flow velocities and
consequently increase water depth directly upstream. The net effect is to reduce the
amount of re-aeration (by increasing depth and reducing the water surface gradient)
occurring immediately upstream of the structure, that may create critical quality
conditions. Conversely, velocities are increased downstream, that will increase re-
aeration. Re-aeration at weirs can cause a significant downstream increase in DO levels.
e) pH
The pH of a river is critical in determining both the amounts of un-ionised ammonia

present and hence, the toxicity to fish. Higher pH levels will increase the proportion of un-
ionised ammonia for a given concentration of total ammonia. However, the toxicity of un-
ionised ammonia is reduced at high pH levels. Further information is presented in
Appendix A.
f) Temperature
Higher temperatures will also increase the levels of un-ionised ammonia for a given
concentration of total ammonia. In addition, higher river temperatures will result in lower
DO saturation concentrations. Also, BOD decay processes will proceed more rapidly at
higher temperatures.
Figure 5.8 Potential DO sag after a CSO discharge where most of the BOD settles to
the river bed
Section 5 - 20
Figure 5.9 Potential DO sag after a CSO discharge where most of the BOD remains
in suspension
g) Ecology
Aquatic vegetation affects DO levels in water by the action of two processes.

Photosynthesis adds oxygen to the water column during daylight hours. During the night,
as oxygen production declines, plant respiration depletes oxygen levels in the water
resulting in diurnal variation in DO levels that may become extreme in eutrophic
conditions.
h) Nature of effluent discharged
The nature of the wastewater discharge is also important.
• BOD decay rates can differ widely depending upon the nature and origin of the
pollutants present causing the BOD and the degree of treatment given to the
discharge. A wastewater with a higher BOD decay rate will remove oxygen from the
river more rapidly than the same load with lower BOD decay rates and result in larger
DO sags for the same BOD load discharged.
• Spills with a high sediment BOD load attached will produce a different impact on DO
levels in the river, compared to those spills that contain predominantly dissolved BOD,
due to sediment deposition and erosion, (Figures 4.18 and 4.19).
• DO levels in the discharge affect the oxygen levels in the river at the point of initial
mixing. DO in spills is often higher in steeper sewer catchments. re-aeration at aprons
of spill, and other structures can result in short-term increased DO levels in the river
after the initial mixing.
Figure 5.10 illustrates the typical impact of a CSO discharge on a small receiving
watercourse. It shows that the impact on DO is severe: DO levels fall to almost 1 mg/l.
Section 5 - 21
Initially, during the spill event, re-aeration is greater than oxygen consumption. Once the
spill ceases, oxygen consumption from the breakdown of deposited organic material
dominates below the CSO.
Figure 5.10 Example of measured DO concentrations in a river just downstream of a

CSO following a storm discharge
5.5.2 Types of modelling approach
a) Generic types
There is a wide range of RQIMs available to assess the impact of wet weather discharges on
receiving waters. They range from simple equations to complex, stochastic or deterministic,
steady state or dynamic models. This section describes the types of model that are available
to demonstrate compliance with long term (percentile standards) and wet weather
(Fundamental Intermittent standards) water quality criteria, as described in Section 2.3, to
support a UPM study. The choice is largely determined by the type of problem to be analysed
and the accuracy required. The best guidance is to select the simplest model to address the
problem under consideration.
Two generic types are considered.
 Stochastic models
In general, stochastic models assume steady state hydraulic conditions and aim to
represent the behaviour of river quality and flow over a long term period. Stochastic
models are designed to be used to estimate percentile criteria to assess the impact on
long term water quality expressed as, for example, 90, 95 or 99 percentiles.
 Deterministic models
Deterministic models are designed to be used to estimate the concentration and
duration of water quality resulting from specific event based inputs of flow and quality
to the modelled river. Such models can have a steady state or dynamic representation
of river and discharge flow and can be used to provide predictions for comparison
Section 5 - 22
with, for example, the Fundamental Intermittent standards. Event based results can
also be converted into statistical criteria for comparison with percentile standards.
Both model types may be used with differing levels of complexity in representing the
effects of in-river chemical and biochemical processes to predict the impact of
intermittent and continuous discharges on receiving water quality. Three levels of
model complexity can be considered.
 Simple mass balance

This involves mixing the predicted wastewater discharge with an appropriate quantity
of river water to give an estimate of the resulting downstream river quality. Mass
balance analysis does not take into account any in-river processes or dispersion of
the input. As such it is most useful for determinands such as ammonia where the
worst impact is most likely to be at the point of initial mixing.
 Mass balance including simple BOD/DO processes

This method allows both the BOD and DO levels to be calculated through time (and
thus at distances downstream) from the point of mixing of the river and effluent flows.
This involves the use of equations to calculate the DO balance resulting from the
decay of BOD and surface re-aeration within the river, using typical BOD decay rates
and re-aeration rates. As these equations are relatively simple, it is possible to derive
analytical solutions or to employ numerical procedures on a spreadsheet to arrive at a
solution. However, many simplifying assumptions have to be made, including:
• an assumption that the flow is steady and the channel is uniform;

• ignoring one of the main oxygen demand processes - either that in the water
column or at the bed; and,
• ignoring nitrogen transformations.
 Water quality process interactions

This level of complexity may be required when in-river processes exert a significant
influence on the impact of an intermittent discharge following initial mixing. These
processes can include pollutant dispersion, as well as advection; biochemical
processing involving BOD and nitrogen transformation; photosynthesis and
respiration; and, sediment attached pollutant interactions. The choice of process
interactions will depend on the nature of the problems being considered and the
characteristics of the river.
UPM studies are likely to require the application of the following river impact modelling
approaches. The specific type of model and degree of complexity required will reflect
the requirements of the environmental regulator, in terms of the type of results and
the degree of confidence required. Simplified, integrated modelling tools, discussed in
Section 4.8 represent a combination of some aspects of these two approaches, with
different levels of complexity in representing water quality processes as required for a
particular application.
b) Steady state, stochastic models
Models based on the mass balance approach are only valid if the data used in the mass
balance equation relate to the same instantaneous period of time. If the instantaneous values
are replaced with distributions of flow and quality observed above a discharge and for the
discharge itself, the mass balance equation can be repeated many times with different values
taken from the specified distributions. This is known as the Monte Carlo Simulation, that
makes it possible to apply correlation between the determinands and specify more
representative distribution shapes. The Monte Carlo Simulation method can be applied to a
single discharge or on a catchment wide basis.
Section 5 - 23
Typically, Monte Carlo Simulation models are calibrated and verified using data for river and
effluent flow and quality collected as part of routine environmental monitoring programmes to
assess river quality and effluent compliance against national and EC criteria. Several years of
this type of infrequent data can be combined to produce model input distribution statistics.
However, these data may not fully represent the effects of infrequent intermittent discharges,
particularly from CSOs. Wastewater system simulation models (Sections 5.3 and 5.4) can be
used to generate distribution statistics to represent future wastewater system performance
scenarios. Upstream boundary conditions for modelling scenarios can be estimated from
observed flow and quality distributions that can be amended to reflect the environmental
criteria specified for the scenario; for example, improved upstream quality or reduced
upstream flow.
c) Deterministic, dynamic river quality impact models (RQIMs)
Complex, dynamic, deterministic RQIMs explicitly simulate the varying flow and quality in a
river in response to storm events. These models can be used to produce results for
comparison with river water quality standards that are concentration, duration and frequency
based (for example, the Fundamental Intermittent standards presented in Tables 2.2 and
2.3). Numerical techniques are employed to solve the differential equations describing the
hydrodynamic and water quality processes. Dynamic RQIMs are considered in detail in
Sections 4.5.3 to 4.5.5. Dynamic models will include, to varying degrees of sophistication, the
following main processes.
 Hydrodynamics
Routing of variable flow through river channels with geometry varying over distance.
 Pollutant routing
Advection, dispersion and mixing of pollutants within the water body.
 Biochemical processes
Significant biochemical degradation processes affecting determinands such as BOD
and ammonia.
 Sediment interactions
Settlement, resuspension and transport of river and sewer-derived sediments, plus
storage and release of associated pollutants.
5.5.3 Description of dynamic RQIMs
a) General description
Several dynamic RQIMs are currently available, all of that have the ability to model a wide
range of varying flow conditions and polluting impacts in complex channel networks. Most of
these models are one dimensional; that is, flows and concentrations are assumed to be
uniform both laterally and vertically within the water column. Two and three dimensional
models are available, but these tend to be used in more specialised cases to model
estuaries/tidal rivers and stratified rivers and lakes. This section discusses one-dimensional
models only. RQIMs generally have more detailed process representations than current
SQMs.
Complex dynamic modelling packages are often expensive and have specific hardware
requirements. Additionally, users need to be trained in using and applying a particular model.
The models also require an extensive amount of data for model building, calibration and
verification for wet weather impact assessment. However, the verified dynamic model
solutions are the most reliable method for making predictions for a wide range of potential
Section 5 - 24
environmental scenarios. The best guidance is to select the simplest model to address the
problem under consideration.
Dynamic RQIMs typically have three main modules. These are:
 Hydrodynamic module
The results from a hydrodynamic module must be available before any advection
dispersion or water quality simulations can be started. RQIMs typically use an implicit
finite difference scheme using the Saint Venant equations to calculate varying flow
conditions. These equations can be solved in a number of ways, depending on the
accuracy and complexity of the model.
 Advection dispersion module

The advection dispersion module is usually based on the one dimensional equation
for conservation of mass of a dissolved or suspended substance. An implicit finite
difference scheme can be used to solve a Fickian advection dispersion equation.
Sediment transport calculations can also be conducted within the advection dispersion
model.
 Water quality module

All the water quality calculations are based on empirical equations to calculate the
chemical and biochemical process iterations at each time and distance step. The
water quality module is normally run simultaneously with the advection-dispersion
module.
Details of the specific biochemical and sediment transport processes included in

individual models should be found in the relevant software documentation.
Determinands that can be simulated include:
• DO;
• temperature;
• ammonia;
• nitrate;
• BOD;
• conservative substances;
• sediments and BOD attached to sediments;
• bacteria;
• nutrients;
• chlorophyll-a; and,
• toxic pollutants.
Modelling erosion, transport and deposition of non-cohesive and cohesive sediments

with attached BOD allows the correct simulation of the delayed oxygen demand
exerted by polluted bed sediments. This is particularly important when modelling the
impact of intermittent discharges from CSOs.
b) Results output and interfacing with other models
The results from a dynamic RQIM should be easily presented and interpreted with regard to
appropriate environmental criteria. Results processing should allow determinand
concentrations to be compared with the appropriate duration/threshold criteria. In addition,
models should be able to interface with appropriate sewer and sewage treatment works flow
and quality models to support the concept of an integrated modelling approach.
Section 5 - 25
5.5.4 Building, calibrating and verifying RQIMs
Considerable effort is involved in the data collection, building and calibration stages of model
application. However, once completed, the model can be used as a predictive tool provided
that it has been verified over a range of appropriate conditions and is ‘fit for purpose’. The
following subsections provide a brief review of the type of work that is typically involved.
a) Specifying the modelling requirements
Once the decision has been taken to use a model to address the problem of wet weather
discharges, the user has to be sure that the modelling software selected can fully address the
problem. At this point, some appraisal of the modelling requirements is needed. This includes
defining the approximate number of boundary conditions, inputs, structures, etc. This should
help in identifying any potential modelling problems and help decide modelling priorities,
boundaries and limitations that may, subsequently, affect the data requirements.
The modelling requirements will also identify the complexity of model needed; whether to use
a simple BOD/DO model or more complex water quality simulation including sediment
transport. Increasing model complexity increases both the data requirements and model run
times, but potentially increases the accuracy of the model.
b) Identifying data requirements
Calibration and verification involves the collection of substantial amounts of data. It is,
therefore, vital that the data collection programme is targeted to obtain the most appropriate
and reliable data in a cost effective manner. The model builder should play a major role in
defining the needs of a data collection programme. A site visit by the modeller prior to the
data collection programme should identify any sites that may require additional data and
identify any possible modelling problems.
Data requirements can be divided into three types:
• Input data
The data needed to build a model include cross-sections, general channel geometry,
roughness estimates, details of bridges, culverts and other in-river structures and any
tributaries. Careful specification of location and definition of cross-sections in the initial
stages is important to produce a realistic and stable model; considerable problems
can occur during model building if this is not done.
 Calibration data
These are either available from historical records or are collected during short term
field surveys under dry weather conditions and storm events. The data will include the
following:
• flow and quality for all significant discharges to the modelled river;
• flow and quality at selected sites along the river. The minimum requirement is
three sites; one upstream of all major urban inputs, one as close as possible to
where the worst impact is anticipated and one located well downstream of the
main impact at a location where the river is recovering;
• flow measurements at the upstream and downstream boundary for different river
flows, that may require a longer period than that covered by the short term
survey;
• time of travel data at different flows; and,
• dispersion data, at different flows.
Section 5 - 26
The data must be robust; that is, they should adequately represent the conditions
under that the model will be applied; for example, summer low flows.
 Verification data
These data will be similar to the calibration data but are used to evaluate the
performance of the calibrated model, under dry and storm conditions.
In practice, the data collected during a field survey may in some circumstances be a
compromise between that that is desirable to allow optimal calibration and verification
of the model under the conditions for that it is to be based, and that that is feasible
given the time and resource constraints of the data collection programme. When the
available data are not fully comprehensive, the idealised model calibration and
verification processes described in 5.5.4 (d) and (e) may need to be pragmatically
modified to recognise the shortcomings of the dataset. This may dictate that separate
calibration and verification exercises are inappropriate and that a model that is “fit-for-
purpose” is produced via an integrated calibration and verification process using all of
the relevant data.
c) Model building
The first step in model building is conceptualisation, in that the catchment is defined and river
channels and junctions are identified. Decisions are taken about what to include within the
model and where to draw the boundaries. The upstream boundary of the main river channel
should be above all major urban inputs. Tributaries that are largely urban streams can be
modelled as inputs to the main channel with loads and flows calculated by the sewer models
(see Section 5.3). This is a simpler representation than explicitly including the tributaries
within the RQIM, and will improve the model run times. The downstream boundary of the
main river channel should be far enough downstream to be sure that the worst impact area is
included within the model. Sometimes this downstream boundary will be obvious, for
example, the confluence with a major river that provides ample dilution, or a major weir that
ensures replenishment of any oxygen deficit. On a flat sluggish river the worst impact area is
likely to be quite close to the main discharge point. On a more fast flowing river the worst
impact may not be felt until the river flattens out some distance downstream of the discharge
point, possibly behind a weir. These factors and any local knowledge, about water quality
problems during wet weather periods, should be used to determine the downstream boundary
of the model.
The individual channel and reach details are then defined together with details of any
structures such as culverts and bridges. Finally, the boundary conditions are specified as
described in Section 5.5.5. The model should be checked for stability, to ensure that the
simulated flows and quality results produced are sensible for a range of conditions.
d) Calibration
Calibration is the process of adjusting the various process rates and parameters within the
hydrodynamic and water quality models, in order to match model predictions with observed
data. This process should be completed for both dry weather and storm events. Both
hydrodynamic and quality modules will require calibration. There are three basic stages in
calibrating a model.
 Stage 1 - Time of travel

The first stage of calibration compares the predicted time of travel with survey data.
Adjustments to the channel roughness coefficients will allow the modelled time of
travel to broadly match the observed data. This process may be completed over a
range of river flows. Roughness values also affect water depth so that, where
available, water level data should also be used to check the model's performance.
Section 5 - 27
 Stage 2 - Advection dispersion

Stage two involves checking the dispersion characteristics in the advection dispersion
module. The shape of dye tracer “clouds” measured in the survey should be
reasonably predicted by the model. This is usually an iterative process of adjusting
dispersion coefficients and factors until a reasonable match is achieved at a range of
flows.
 Stage 3 - Water quality

In stage three, water quality predictions are compared with observed water quality
time series or spot sample data. Calibration can proceed by adjusting the various
water quality process rates until a good fit with the observed data is achieved. It is
best to calibrate temperature first, followed by BOD, ammonia and nitrate and finally
DO. It is important that the process rates are not altered beyond realistic bounds. In
general, process rates that are known (from laboratory studies) should be adjusted
first. Next the processes with the most physical meaning should be adjusted before
adjusting other process rates and values.
Difficulties in obtaining a good calibration may result from unknown inputs or from un-
modelled processes (e.g. if algal growth has not been represented). In this case, an
alternative modelling approach may have to be adopted, although this should have
been identified in the initial assessment of the problem.
The acceptability of water quality calibration is very subjective. Calibration objectives

are easier to specify for hydraulics models, where volumes, depths and velocities can
be used and matched to within set percentages or calibration criteria, i.e. 5, 10 or 20%
of the total. However, it is not as easy to do this with quality data, due to its variability.
Often, calibration by eye is as good as any statistical test. But there are some factors
that must be considered. For example, how good is the peak to peak match for
pollutant plug or DO sag? This will give an indication of the load balance. Is the timing
of the pollutant plug or DO sag matched adequately? This will confirm that the
appropriate river channel roughness is being used.
e) Verification
Verification refers to the process of confirming the reliability or accuracy of the model, for
both dry weather and storm events, by comparing model predictions with observed data from
events that were not used for calibration, to ensure that the model is ‘fit for purpose’ for the
likely range of simulation conditions.
Trying to match the predicted and observed flow and quality data is often very difficult.
Numerous sources of errors can occur that could affect the outcome. In an attempt to
rationalise these errors, the model builder should try to identify mismatches between
observed and predicted data and categorise them as resulting from:
• inappropriate process rates and other model parameters;

• processes not included in the modelling capabilities of the software; or,
• mismatches caused by unmodelled polluting inputs or abstractions.
Inappropriate process rates may be rectified by further justifiable adjustment of the relevant
rate coefficients, i.e. recalibration. If this is necessary the model will not be strictly `verified'
unless additional independent data are used.
It may be possible to allow for processes not included in the model either by adjusting
process rates beyond expected ranges, that will limit the model's usefulness, or by simply
allowing for the error in any results, that is normally the more appropriate approach.
Section 5 - 28
In the event of mismatches caused by unknown inputs. Such discharges may be known to
occur but are unknown in terms of both flow and quality, or the mismatch between observed
and predicted data may be so significant that it indicates that unmodelled discharges are
occurring. For either case, the model should not be made to fit the data and the mismatch
should be explained in any accompanying documentation. An attempt should be made to
quantify the unknown input or abstraction, that may necessitate further data collection, to
increase confidence in subsequent decisions based on model results.
Verification is completed when the model has been tested against a minimum of one dry
weather ‘event' and one storm event. A satisfactory match between observed and predicted
river quality should be achieved. If this is not possible, but adequate explanation can be given
for the mismatch, a model can be said to be verified within the limits specified.
5.5.5 Using dynamic RQIMs
Once a model has been built, calibrated and verified it can be used to investigate any number
of scenarios. However, the model user should always exercise care and be satisfied with the
model’s applicability. This will help in identifying problems in the model and assessing its
reliability. The user should ensure that, as a minimum:
• conditions against that the model was verified have not changed; i.e., structures,
process rates, etc.;
• all modelling assumptions and accepted margins of error are made known to the
decision maker; and,
• process rates have not been altered beyond acceptable margins.
The model can only be fully applied in a predictive mode once the user is satisfied with the
basis of the model and the associated data.
a) Boundary conditions for upstream flow and quality
It is important to be able to assign appropriate upstream river and tributary flow and quality
conditions to individual model runs, as the boundary conditions will determine the dilution
capacity of the river; affect the re-aeration rates; and, at high flows, can increase pollutant
load by resuspending sediments.
The river flow and quality conditions at the time of a storm will depend on many factors.
These include:
• the nature of the river catchment, its site, land use and geology;
• upstream discharges and abstractions;
• the time of year; and,
• the rainfall pattern over the previous days; or, for catchments that are
predominantly groundwater fed, the rainfall pattern over the previous months.
Many of these factors are difficult to quantify without detailed river catchment modelling. As a
result, traditional planning approaches have tended to take a conservative assumption that
upstream flows will be low at the time of a storm event. A more rational method,
recommended for UPM studies, is to use a statistical approach to select upstream flow and
quality conditions. This can be further refined depending on the degree of complexity of the
river impact problems and processes being addressed and the type of impact modelling being
carried out.
Boundary conditions define the hydraulic and water quality volumes and concentrations at the
boundaries of the model. These require a time series of data for the duration of the
Section 5 - 29
simulation. The initial conditions are those specified for all points within the model at the first
timestep. The boundary time series for flow and quality can take the form of either steady
state (constant) inputs or a dynamic (changing) input. For most cases, the upstream
boundary conditions can be assumed to remain constant over an event. Catchments with a
rapid upstream response to events or significant upstream diurnal variations in quality and
possibly flow, will require the use of dynamic boundary conditions to initialise the model prior
to the input of wet weather discharges. The initial conditions for running a particular event or
scenario can be selected by using a Monte Carlo Simulation approach to select values from
upstream flow and quality distributions, that can take account of correlations between rainfall,
flow and river quality, as described in Section 4.5.2. Multiple simulations using different input
values can be used to test for sensitivity to upstream conditions.
The following sections describe how suitable river flow and quality time series can be derived
to represent event boundary conditions.
 Upstream river flow

Two approaches are available to identify upstream flow conditions.
– Flow duration curves

Flow duration curves can be used to identify steady state river flow conditions for an
individual event. River flow frequency distributions are normally expressed by flow
duration curves that give the daily mean flows that are exceeded for different
proportions of time. In some instances, the environmental regulator may be able to
provide suitable data. However, there will be situations where the available data are
inadequate and it will be necessary to estimate the flow duration curve by either using
regionalised statistical models (Gustard et al, 1992) or measured flows over a suitable
period of record. If the latter approach is adopted, it is common practice to infer a long
period flow duration curve by relating the measured flows to those at a long term
gauging station. In both approaches it is important to account for the potential impacts
of artificial influences upon the flow regime. Methods for achieving this are described
by Watts et al (1995) (see UPM2 References, Section 1.5.2).
Normally, summer flow conditions will be the most critical for intermittent pollution
events. This is because flows are low (providing less dilution) and temperatures are
high (that reduces DO levels, increases BOD decay rates and increases un-ionised
ammonia). Thus, it is important that the river flow duration curves developed at this
stage are representative of summer conditions.
In many cases it will be possible to characterise a river flow by a shifted log-normal

distribution and to define that distribution by summary statistics; for example, the
mean, the minimum standard deviation and the minimum flow, based on historical
observations.
– Hydrological modelling
Hydrological modelling allows a long dynamic river flow time series to be matched to a
rainfall time series. This provides an excellent basis for continuous simulation of the
impacts of wet weather discharges over a long period of time. Alternatively, the
information provided by the river flow time series can be analysed statistically to
estimate the distribution of river flows at the time of summer storms and to examine
any correlations between storms and river flow. This information should provide a
better basis for river flow estimation for use with dynamic event modelling than the
flow duration curves described for steady-state event modelling.
Conceptual rainfall-run-off models can create a time series of river flows based on a
long rainfall time series and evaporation data. These models use a simple
representation of the main physical processes that govern water flow in a river
Section 5 - 30
catchment. These processes include interception, evapotranspiration, transfers

between soil, groundwater and channel storages and time of travel. Some existing
concurrent rainfall and river flow data are essential to calibrate the models before they
can be used in a predictive mode. Such data will not always be available and this will
limit the applicability of this approach. In addition, calibration is not straightforward and
should only be applied by specialist staff.
Once such a model is calibrated it can be used to generate a realistic dynamic event
time series of river flows using rainfall data. Such river flows should properly reflect
the antecedent rainfall conditions that precede any particular storm being investigated
in the urban catchment.
– Modelling rapid flow response

There will be situations where the upstream river catchment, or a significant part of it,
is known to respond rapidly to a rainfall event because a large proportion of the
upstream catchment is urbanised. As a result, river flows increase rapidly and can
provide extra dilution at the time of storm sewage discharges.
These situations are best handled by including the part of the river catchment that is
responsible for the rapid response within either the sewer flow model or the river
impact model itself. In particular, it is most important to include any urban surface
water catchments that will have rapid run-off characteristics within the sewer flow
model.
It should be noted that most hydrological models will only estimate the natural
response of the river catchment to rainfall. Hence, any significant upstream
abstraction, discharge or reservoir control may need to be accounted for by using an
additional method or by assuming flow variations based on local observations.
 River quality
Two choices are possible to identify upstream quality conditions. The upstream
boundary quality can remain unchanged during an event (steady state) or quality can
change during an event due to upstream catchment responses and/or diurnal
variation. Existing upstream quality data can be used to estimate statistical
distributions to be sampled by Monte Carlo simulation to identify the initial conditions
for a specific modelling scenario. As with river flows, it will be normally more important
to concentrate on summer data. Accuracy in estimating upstream quality conditions is
usually less significant than in estimating river flows or effluent flow and quality. This is
because upstream quality is generally much better than the thresholds used to assess
the significance of river impacts (Section 2.3). Guidance on selection of boundary
conditions where rivers are polluted and fail to achieve these thresholds above the
study area can be found in Section 2.3.
The selected initial quality input conditions can remain constant over the event
simulation or can be varied to reflect observed or assumed variations in the course of
an event.
b) Initialisation
All RQIMs should be initialised correctly before use. Initialisation ensures that the correct flow
and quality values are used at all points in the model and not just at the boundaries. For
example, the correct diurnal variation of DO is only established when the model has been run
long enough for data describing the quality at the boundaries to reach all points in the model.
Figure 5.11 shows a typical observed summer DO timeseries produced by the diurnal effects
of photosynthesis and respiration in a river. Generally model results for DO should replicate
the observed diurnal pattern before attempting to introduce the effects of intermittent
discharges. However, as noted in Section 2.3.2 (d), it may be appropriate in certain
Section 5 - 31
circumstances (for example, eutrophic rivers) to remove the effect of diurnal DO fluctuations
from modelling runs to assess compliance with environmental standards.
Figure 5.11 Typical diurnal DO concentrations observed in a river
c) Sensitivity checks
It is good practice to complete a few sensitivity checks to assess how the predicted impacts
change when sensible changes are made; for example, to process rates, temperatures, the
time of day of the wet weather event, etc. This knowledge will help the user to understand the
limitations of the model and to interpret the results more reliably.
d) Problems and limitations
Two types of limitation are apparent with dynamic RQIMs. These can be categorised as
either limitations associated with a particular piece of licensed software (model specific) or
limitations associated with dynamic models in general.
 Model specific problems and limitations

Most models have specific limitations and difficulties peculiar to that piece of software.
These limitations may be in the hydrodynamic, advection-dispersion or quality
modules. Some may be rectified with future model developments and others may be a
fundamental limitation of the model software. Most users who are familiar with a piece
of software will be aware of that model's particular limitations, and new users are
advised to join the relevant users' group or contact other users to identify these
limitations, as frequently they are not identified in the model’s User Guide.
 General problems and limitations

– Hydraulics
The results from a water quality simulation will only be as accurate as the results from
the hydrodynamic model. Any errors in the hydraulics will be automatically transferred
to the quality model.
Section 5 - 32
– Initialisation
All models require correct initialisation and specification of boundary conditions. The
model should be in a stable condition prior to any input of pollutants. This ‘stable'
condition will, however, allow for diurnal variations, especially in DO and temperature.
– Stability
It is often the case that a total eradication of instability in dynamic models is not
possible. The model should, however, be free from any numerical instability that
affects the accuracy of the results to any significant degree.
– Modelling low flows

Problems can occur when modelling water quality in rivers with low flows. The
calculations can become unstable and the results erroneous. This can be a particular
problem when modelling low flows in steep catchments.
Modelling tributaries (input as either boundary conditions or via sewer model

interfaces) can cause the model to crash if the tributary flow is significantly larger than
the flow in the river, such as a final effluent from a large STW discharging into a small
river. Normally, this can be partly overcome by reducing either the number of
calculation points or the timestep.
– Calculation points/timesteps
Some complex models require a small distance and timestep when introducing a large
polluting discharge. This is especially the case when modelling steep and complex
channel networks, or sediment and BOD interactions. This results in longer run times.
– Range of applicability
The model will be most reliable for conditions that are similar to those used for
calibration and verification. However, it will normally be necessary to make predictions
for more extreme events. Potential inaccuracies as a result of using the model outside
its verification range should be quantified.
e) Interpretation of results
All results should be viewed with reference to the model's expected accuracy and limitations
of applicability. An associated error band is inevitable within any model. An assessment of
this error band, that will be related to the input data and how the model was built, is useful.
The most widely used measure of model adequacy remains direct comparison with observed
data. Limitations within the model should also be assessed and quantified where possible.
Suitable safety margins must be used when comparing the model's result against standards.
These margins should reflect the accuracy expected from the model, with larger margins for
more inaccurate models.
5.6 Tidal waters modelling

Tidal waters modelling covers the use of environmental models that describe waters where
tidal movements are significant to pollutant transport. It includes both coastal and estuary
environments, and this section of the manual covers both aspects. The complete discipline
of tidal waters modelling is very broad and, as with the remainder of this manual, the focus of
this section is on the wet weather water quality impacts of discharges from the sewerage
network and treatment works.
Section 5 - 33
Depending on the circumstances, the modelling approach adopted can range from simple
spreadsheet approaches to complex, dynamic 3D models. It is therefore important to
consider what the most appropriate modelling approach is for the specific questions being
asked. Where studies involve several organisations, it is important that the overall approach
is agreed before detailed work commences.
Modelling of tidal waters generally has two distinct phases:
 Representation of the tidal movement, so that the models accurately describe the
transport, advection and dispersion of conservative substances.
 A water quality component that allows for decays, interactions and other transformations
of the substances of interest
In some large bacterial modelling studies, where there are complex inputs from a wide variety
of sources and the volume of data and associated time taken for processing would be
prohibitive, a third post processing unit impact approach may also be adopted.
Accounting for tides can often make the use of simple approaches more challenging, and the
outputs less useful, but there will be situations where a simplified approach is reasonable and
appropriate. In some circumstances, especially where there is significant uncertainty, even
after a study has been completed, it may be considered more appropriate to use surrogate
approaches instead of modelling.
Studies will need to be supported by field data sufficient to calibrate, validate and, in some
cases verify the models. This data can also assist in an initial identification and
characterisation of the sources of pollution, a process known as a source budget study.
In some instances, models of tidal waters may have been previously developed by other
organisations, for other purposes, such as navigation. Whilst such models may provide a
useful starting point for the hydrodynamic component of the modelling study, it is worth being
aware that significant additional work may be required to ensure that they are also fit for
purpose for water quality modelling.
The relevant standards that may apply to tidal waters are defined in section 2.3. In estuaries
particularly, it should be noted that several different types of standard may be used. As set
out in section 3.5, it is key that the use objectives of all receiving waters, and the
environmental standards which the environmental regulator requires compliance with, are
identified at this stage.
This section is subdivided as follows:
5.6.1 Introduction;
5.6.2 Effects of Intermittent Discharges on Environmental Quality;
5.6.3 Types of Tidal Waters Impact Models;
5.6.4 Building, Calibration and Validation of Tidal Waters Models;
5.6.5 Using Tidal Waters Impact Models;
5.6.6 Interpretation of Results.
Section 5 - 34
5.6.1 Introduction
Tidal models can be used for a wide range of applications, both in terms of the type of
assessment, and in terms of the range of complexity that can be dealt with. They provide a
means of understanding environmental conditions (e.g. tides, winds, differing times of
discharge) and the potentially complex interaction of discharges (intermittent, continuous and
diffuse) which may need to be considered together to develop a useful understanding of
impact.
In common with inland modelling, the scoping stage of the study is important, and time
should be taken at the start of each investigation to understand the problem, the aims of the
study, and the tools and method that will be required. Why a model is needed, and the type of
model or the modelling approach needs to be understood. There may be a need to use a
combination of approaches to protect receiving waters depending on the importance and
vulnerability of the receiving water concerned.
The strategic value of expanding a model, for example, to cover a large stretch of coastline
covering several bathing waters, may need to be considered.
5.6.2 Effects of intermittent discharges on environmental quality
The main changes in tidal water quality due to wet weather discharges are:
 A rapid and potentially large increase in the concentration of pathogenic organisms

(bacteria and viruses);
 A reduction of Dissolved Oxygen (DO) (this is particularly relevant to estuarine areas);
 A rapid increase in concentrations of ammonia, biochemical oxygen demand (BOD), and

suspended sediments.
Bacterial impacts affect bathing waters and shellfish waters. The presence of raised levels of
pollution can increase health risks to bathers or to consumers of shellfish, where pathogenic
organisms are present. As pathogenic populations are transient, unpredictable, often
sporadic and present in relatively low concentrations, indicator species are used which are
accepted to identify the potential for exposure to pathogenic organisms. These Faecal
Indicator Organisms (FIOs) are the principal determinands for both bathing waters and
shellfish waters assessments. If there is a requirement for direct assessment of pathogens
(for example to understand a virus outbreak) then the same modelling approach can be used.
Both FIOs and pathogens are subject to death or a change to non-viable status in the
environment (as both are conditioned to survive and flourish in a host, for example a
mammalian gut). Temperature, salinity and other factors all play a part in the inactivation of
pathogens and FIOs, but the principal factor is exposure to ultra-violet light (sunlight). The
impact of UV is modified by factors which protect these organisms from exposure, the most
important being turbidity and the potential to adsorb to sediment particles.
Pollutants such as ammonia, BOD and suspended solids will tend to have an environmental
impact (often on DO levels) as compared to the potential health impacts of bacteria and
viruses. Ecology can be adversely affected in both the long and short term, by restricting or
preventing organisms from occupying their normal habitats or providing preferential
conditions for other species to colonise a habitat or ecosystem. For example, DO levels have
an impact on the ability of migratory fish to pass through estuaries.
Pollutants such as ammonia, BOD and suspended solids will tend to have an environmental
impact (often on DO levels) as compared to the potential health impacts of bacteria and
Section 5 - 35
viruses. Ecology can be adversely affected in both the long and short term, by restricting or
preventing organisms from occupying their normal habitats or providing preferential
conditions for other species to colonise a habitat or ecosystem. For example, DO levels have
an impact on the ability of migratory fish to pass through estuaries.
Where there is a complex interaction of sources, impacts may not necessarily always be
intuitive, and there is a need to understand how the different sources interact with one
another under different discharge scenarios.
The modelling approach selected must have the potential to generate data which can be
used to evaluate performance against the relevant standards, over a given period, and
provide confidence that all contributing sources are represented adequately.
Source apportionment is the concept of using modelling to identify individual sources and the
proportion of impact they contribute under different circumstances, and is vital in
understanding how adverse impacts are generated. It is very difficult to apply interventions or
mitigations in an effective and efficient manner if the sources of pollution are poorly
understood.
Source apportionment is a description often also used for source budget studies. Source
budget studies use sampling, monitoring and other field studies to identify key pollutant
sources impacting an area, including the scale and range of loads from each source, and are
often a pre-cursor to detailed modelling. When a source apportionment project or phase is
being referred to it is important to clarify whether this is a modelling approach or a sampling
approach.
For other impacts, the most commonly required water quality model is for predicting DO. The
processes that affect dissolved oxygen in the water column are:
• Temperature
• Salinity
• Sediment Oxygen Demand;
• Decay of BOD;
• Decay of ammonia;
• Reactions with decaying phytoplankton;
• Respiration;
• Photosynthesis;
• Re-aeration.
The process of photosynthesis adds dissolved oxygen to the water during daylight hours.
During the night, when photosynthesis does not occur, plant respiration may deplete oxygen
levels in the water resulting in diurnal variation in DO levels that may become extreme in
eutrophic conditions. The physical process of re-aeration describes the transfer of dissolved
oxygen across the water surface. Whilst these reactions directly affect dissolved oxygen
concentration in the water, many other reactions may need to be described in the model
because they affect the BOD, ammonia or phytoplankton. For example, in an estuary, the
settling of particulate BOD on the bed at times of low currents and subsequent resuspension
during stronger currents alters the distribution of the BOD. Since ammonia is hydrolysed
from organic nitrogen, that too may need to be described in the model. If photosynthesis is
considered significant, it may be necessary to model the availability of nutrients and
phytoplankton.
In addition, the adsorption and desorption of substances, exchanging between the dissolved
and particulate phases, can be important for substances such as heavy metals and
pesticides. It is important to take a pragmatic a view as possible regarding complex
interactions, without compromising the value of model outputs, to avoid the exercise
Section 5 - 36
becoming too complicated. The complexities involved in understanding water quality

processes may mean that those processes need to be parameterised or represented more
simplistically. In such cases, the reasoning, justification and approach undertaken should be
clearly set out and agreed. Sensitivity testing may help to understand the implication of these
simplifications. Whatever situation is being studied, use should be made of all information,
including verified subjective information.
5.6.3 Types of tidal water quality impact models
The aim of any tidal waters hydrodynamic model is principally to represent the movement of
water (i.e. to accurately characterise the movement of the tides and any currents) such that
pollutants introduced to the model are correctly transported. A water quality model will then
represent the dispersion and decay of those pollutants.
 Issues specific to Estuaries

In general, currents in estuaries oscillate with the tidal period (about 12.5 hours) and can
be stronger than in rivers, especially where the estuary is long, shallow at low water, of
reasonably uniform width and has a tidal range of more than a metre.
The two main physical features that distinguish estuaries from rivers are:
 Tidal oscillating flow; and,

 Combination of fresh and salt waters.
Strong currents imply high dispersion, but a tidal oscillation gives a wide variety of impact
scenarios, especially for an intermittent discharge:
 Low water discharge - when the current velocity turns from seawards to landwards,
would result in a 'hot spot' because of poor dilution in relatively shallow water. This
could give a high impact on water quality in the short term. The incoming tide would
then carry the 'hot spot' of pollutant upstream for some six hours, possibly with little
dispersion.
 High water discharge - though again at a time of slack current, would be into deeper
water initially and it would be carried seawards and possibly out of the estuary into the
sea.
 Intermediate between these two extremes - wet weather spills would be into tidal
currents that could give high dilution by spreading a spill over several kilometres in the
estuary.
It should be noted that in some locations, local conditions can result in deformed tides, which
can result in either unequal flood and ebb tide times, or double peaked tides. These
situations should be identified accordingly.
It is important to realise that estuaries can be complex areas to model, and although simple
approaches can be used, there is soon a need to understand complexities. The most obvious
are tidal oscillation or reciprocation (the tendency for a pollutant to be retained in the estuary
due to small residual movement across tides), and the superimposition of pollutant plumes
arising from tidal movement. Considering discharges from more than one source, and the
need to represent diffuse sources, means that in many cases, simple river-based approaches
may not provide the required resolution. Where they do, they will deliver a much quicker and
simpler approach for basic assessments.
Tidal oscillation in a long estuary may imply a long residence time for the material discharged,
so for an Environmental Quality Standard (EQS) for dissolved oxygen, this may be difficult to
meet as pollutants can exert their oxygen demand over many days.
Section 5 - 37
A further complication stemming from an oscillating flow occurs when the estuary is
branched. On the ebb tide, water will flow from tributaries into the main channel and out
towards the sea. After low water, flow will return up tributaries, although the phasing of
outflows and inflows can lead to considerable mixing between tributaries.
In a hypothetical non-tidal estuary, the seawater would tend to form a semi-static wedge as a
lower layer with the entering fresh water moving out across the surface as a thin layer
(stratification). Vertical mixing between the two layers would be slow as the density difference
suppresses vertical movement (and mixing). This may amplify polluting effects or restrict
pollutants to one of these layers.
Most estuaries in the UK are of the well-mixed variety, where fresh and seawater mix
completely, though some deep estuaries may be stratified, with a gravitational circulation
predominantly landwards near the bed and seawards at the surface. The inflow near the bed
may vary in strength from a small fraction of the freshwater flow in shallow well-mixed
estuaries to some ten times the freshwater flow in deep estuaries. In such cases a 3D
modelling approach will be required. It should again be noted that there may be situations
where shallow estuaries can exhibit stratification.
An aspect of both estuaries and some coastal areas is the potential for channels and flats to
change position and shape. The model should be developed on the best possible
information, and consideration should be given to critically reviewing the potential for channel
migration to alter results and predictions. It may be that an alternative bathymetry should be
tested, although it is recognised that it is not practical to undertake any validation for a
situation which cannot be measured in the field. A clear agreement of scope and approach
for these issues should be gained at the start of a project.
It should be borne in mind that the resuspension of sediments in estuaries may have water
quality implications, including on bacteria, oxygen demand and hazardous substances. A
dynamic sediment model that includes the settlement, resuspension and transport of
sediments plus the storage and release of associated pollutants may be required if these
impacts are significant.
 Tidal Waters Modelling
For all the reasons described above, in complex water quality modelling cases as a
minimum a 2D dynamic model will be appropriate for estuary modelling. As this is the
same approach required for ‘marine’ modelling, it is appropriate to consider both
issues together (as ‘tidal waters’ modelling).
The application of models for estuarine and marine situations may differ, but the
principles behind model use remain the same. Overall, for tidal waters, the situation
may be complex, with both near and far sources interacting (and in different ways at
different times of tide, weather, etc.). Impacts also superimpose due to tidal
movement taking pollutants away from a receiver on one tide and bringing them back
on the return tide. Whilst this leads to challenges in model build and application,
models are very important to understanding impacts. It is therefore vital that input data
is accurate and calibration and validation is robust and clear.
There are several types of model, and modelling approaches to consider. Figure 5.1
provides a process through which the modelling approach can be determined.
Where the impact is simple (for example from one source, or very close to a sensitive
receiver), and volumes or concentrations are likely to be small, then ‘simple’
approaches are likely to be useful. In addition, they offer the benefit of significantly
less time and cost to run than dynamic modelling approaches.
Section 5 - 38
Where a standard is required in the initial mixing zone from a discharge (i.e. beyond
the ‘boil’ of the discharge reaching the surface - the ‘near-field’), then an initial dilution
approach is appropriate. If an understanding is required of further dispersion and
dilution (but within one tide), then a model which can incorporate simple secondary
dispersion can be employed.
Neither of these approaches will be accurate where the tide takes pollutant away from
the area, but then it returns on the next tide (superimposing of pollutants). Current
data (speed and direction) and water depth data are required (these might be
obtained from data or an existing coastal model), as are details of pollutants and
concentrations. Initial dilution may not be appropriate where the discharge is exposed
for part of the tide, and initial dilution standards are likely to be failed (as dilution is
zero when the discharge is exposed).
Where the transport is complex (returning on subsequent tides) or multiple sources,

multiple (and possibly distant) receivers and complex discharge scenarios will be
encountered, then dynamic or complex models are likely to be required. These are
likely to be mid and far field type situations, as these complex models cannot deliver
assessment in the near field. Where an understanding of both near and mid field
quality is required (for example to ensure that a discharge meets basic aesthetic and
pollution control standards, and that more distant receivers are compliant) then both
simple and complex approaches would be employed – that is a near-field study
followed by a mid or far field study.
Note that in tidal waters modelling, several definitions of near- and far- field modelling
can be used. The one adopted in this document is based on the description provided
below.
Section 5 - 39
Does
Doesinitial
initialAssessment
Assessmentsuggest
suggestthatthatthe
thewater
waterquality
qualityproblem
problemisissimple?
simple?
● Exigent for one tide or less;
●
Exigent for one tide or less;
● Unlikely
Unlikelytotohave
haveimpact
impactatatdistance
distancefrom
fromsource;
●
source;
● One source or a few sources easily disaggregated.
●
One source or a few sources easily disaggregated.
No Yes
Yes
Dynamic
Dynamicmodelling
modelling Simple
Simplemodelling
modelling
Approach
Approach assessment
assessment
Issue
Issueonly
onlyininthe
the
Simple
Simplewell
wellmixed
mixedEstuary
Estuary Near
Nearfield?
field?
1D
1DModelling
ModellingApproach
Approach
More
Morecomplex,
complex,well
wellmixed
mixed Yes No
Estuary, Yes No
Estuary, multiplesources,
multiple sources,
Large
Largeextent.
extent.
2D
2D modellingapproach
modelling approach
Initial
Initialdilution
dilutionmodelling
modelling
Complex
Complexestuary,
estuary,poorly
poorly
Mixed
Mixedororsubject
subjecttotostratification
stratification
3D
3Dmodelling
modellingapproach.
approach.
Simple
Simplemodelling
modelling
Assessment
Assessmentinvolving
involving
initial
initialdilution
dilutionand
and
Assess
Assessavailable
availabledata simplified
Adequate
data simplifiedsecondary
secondary
Adequatefor
formodelling
modellingstudy
study dispersion
dispersion
Yes
Yes No
No
Model
Modelbuild,
build, Implement
Implementfield
field
Calibration,
Calibration, data
datacollection
collection
Validation
Validationand
and programme
programme
Operation.
Operation.
Figure 5.12 Process for detemining type of model to use
 Near-Field Modelling
Near field modelling represents the plume as it is discharged and is moved through
the water column. It seeks to describe the initial process of turbulence and
entrainment of the discharge in the surrounding water, and the mixing and dilution of
the discharge as this occurs. Used largely to determine aesthetic standards for
continuous discharges (especially historically when lower levels of wastewater
Section 5 - 40
treatment were prevalent) it is an important tool where standards must be met over
very short distances or where discharges are made directly into designated receiving
waters.
 Steady-State Near-Field/Simple-Secondary Dispersion Modelling
This type of model is generally a spreadsheet-based, steady-state model using

current and depth data to deliver near-field output, coupled with basic secondary
dispersion. Such models do not generally represent tidal movement well, (for
example the superimposition of pollutant due to tidal flux) but can be useful,
particularly where there is a requirement to meet a standard over a short distance (for
example, mixing zones for a substance with an Environmental Quality Standard).
They are therefore most applicable to individual outfalls.
 Mid-field Modelling
This refers to the representation of the transport of pollutants beyond the initial mixing
zone, and typically over a timeframe of a few to several days. Transport of pollutant is
principally by advection (tidal or current driven movement) and dispersion (mixing at
the boundaries of the pollutant plume). Initial dilution (the ‘boil’ and the initial mixing
zone) is effectively ignored.
Bacterial transport and impact modelling is included in mid-field modelling. This is

undertaken with a dynamic model capable of correctly representing tides and current
movement, and capable of integrating any wind induced effects. The model used is
typically a 2D (depth-averaged) model, but there are situations where a 1D or 3D
model approach may be appropriate (more often associated with estuarine areas).
The mid-field approach can integrate many sources, and can be used to model
multiple receivers simultaneously.
The domain of the model needs to be sufficient to cover all the sources and receivers,
and to ensure that pollutant is not lost over the boundary of the model.
Pollutant, or water quality modelling, can be undertaken either using an advection-

dispersion or particle tracking approach. Considerations for selecting the most
appropriate approach include:
• The nature of the environment to be modelled. For example, particle tracking may
not be appropriate for estuarine situations with large inter-tidal extents;
• The resolution of the model (cell size) may have a bearing, when accounting for
numerical dispersion in the model;
• Timescale may be a consideration; particle tracking approaches will generally be

much quicker than advection-dispersion (AD) modelling;
• Ability to represent the associated water quality processes.
If used incorrectly, both approaches can lead to misrepresentation of impacts. For

example, particle tracking in inter-tidal areas can lead to particle trapping in drying
cells, and inaccurate predictions. An advection-dispersion approach where
numerical dispersion is significant can also lead to inaccurate predictions.Use of a
particle tracking approach will also require consideration of the number of particles
that can be represented in the model. Each particle is associated with a certain
pollutant load and therefore particle numbers should not be so low that one, or a
Section 5 - 41
few, particle(s) could cause a significant change in predicted concentration.

Conversely, there may be a point when too many particles will be required to
preserve resolution, and at this stage particle tracking may not be the best option.
 Far-Field Modelling
Far-field modelling uses the same basic principles as mid-field modelling, but refers to
pollutants which persist and potentially accumulate over a longer period
(days/weeks/months). The same modelling tools as for mid-field modelling may be
used, but the types of pollutants are those that persist, or that interact with other
substances that continue to have a potentially polluting effect. This could refer to
nutrients, for instance, or potentially to the modelling of viruses, which typically have
much longer survivability than bacteria.
The domain of the model needs to be considered over the longer time periods
involved, to ensure that pollutant is not lost over the boundary of the model.
A decision to build a ‘regional’ domain for a model, as a strategic long-term tool,

generally means that all the potential considerations for modelling are covered.
 Compliance Modelling Tools
To understand the impact of discharges on the potential of a receiver (for example a

bathing water) to meet a given standard, the modelling approach needs to be able to
produce a prediction of the standard that can be attained. For example, for bathing
waters, this could mean predicting a 90 th or 95th percentile value and comparing this
with the reference value of the standard. Table 2.8 of this manual shows how the risk
of non-compliance is taken into consideration when these bathing water standards are
used in the UK. Where detailed compliance assessment predictions are carried out
then these are usually based on at least ten years (or ten bathing seasons), such that
a representative set of rainfall/wet weather spill conditions can be assessed.
Discharges from intermittent sources can occur at any time; they are not regular or
predictable, as they are determined by independent rainfall events. In complex cases,
to produce predictions that account for all potential impact scenarios, model runs over
all stages of tide (as spills can occur at any stage of tide), wind, rainfall, and possibly
river flows need to be undertaken. Model runs should account for all known sources,
represented in a way that is suitable for the study being undertaken. Transparency in
how estimates or assumptions have been calculated for sources with limited data is
very important, as it allows sensitivity testing to be developed, and a reasonable
interpretation of results.
 Unit impact approach
Undertaking assessments of the impacts of intermittent discharges on tidal water

quality directly within the water quality modelling software is likely to have run-time
and data generation issues. It is also likely to be inflexible when considering scenario
changes, sensitivity, or in responding to input errors. To address this, so that detailed
compliance studies can be modelled and analysed within feasible timescales, a unit
impact approach is normally adopted.
A unit impact approach describes the modelling of arbitrary flows and concentrations,
over a specified period, to establish the impact relationship between a source of
pollution (for example a CSO) and a receiver (for example a bathing water). These
unit runs are undertaken for all scenarios of tide, wind, etc. This generates a
Section 5 - 42
database of unit impacts, across all potential impact scenarios. By using information
on actual spill events under specified conditions (for example, the output from a
sewerage network model and a specified rainfall time series), the unit run impacts are
scaled for actual spill events.
As with all modelling approaches, consideration should be given as to whether this

approach is appropriate for the study being considered.
5.6.4 Building, Calibrating and Validating Tidal Water Models
This section discusses building, calibrating and validating tidal water models. There should
always be an agreed set of criteria and threshold values upon which the model should be
developed against, and assessed. These would include, for example, current speed and
direction, tidal level, as well as water quality elements such as dye patches. Demonstrable
meeting of those agreed criteria would then deem the model ‘fit-for-purpose’. A UK example
of such criteria is 'A framework for marine and estuarine model specification in the UK’ (WRc,
1993).
During Calibration, the internal model parameters are adjusted to find the best fit between
model predictions and field data. For example, the parameters adjusted might include bed
roughness (hydrodynamics) and dispersion coefficients (water quality). Adjustments should
only be made within reasonable bounds, based on physical reality and evidence based. This
is not a force-fitting process.
During Validation, the model predictions are tested, with the model parameters unchanged,
against an independent field dataset. In coastal models, calibration can be undertaken with
data taken on a neap tide, and validation undertaken with spring tide data. These two
datasets are independent.
Where detailed compliance modelling is carried out then Verification is the comparison of
the output of compliance assessments undertaken with calibrated and validated tidal waters
models to field or sampling data, without adjusting model parameters. For example, the
comparison of bathing water monitoring data to model predictions.
Once an acceptable level of calibration, validation and where appropriate verification have
been achieved, the model will be ‘fit for purpose’ (that is, a validated model which is
demonstrated as meeting the criteria agreed for use in assessments), if there are no gross
changes in the area likely to significantly affect hydrodynamics and/or dispersion. For each
study undertaken, there is a process of ensuring the model is fit for purpose for a specific
investigation. This may not be the case, for example, if all the hydrodynamic and dispersion
field data were collected far from the new investigation site. In this case there may be a
requirement for additional field data collection, or it may be that previous studies and/or
existing data will provide sufficient information. This should be appraised on a case-by-case
basis. Conversely, the fact that a model has been verified using bacterial data does not
mean that it is unsuitable for use to appraise other pollutants, although some further
calibration and validation or verification may well be necessary.
Calibration and validation can be, and usually should be, a staged process, being carried out
first on the hydrodynamic model and then on the water quality model. This reduces the
number of independent variables to be considered at any one time and provides a closer
focus.
 Building a detailed tidal waters model
The first step in building any detailed 2D or 3D tidal waters model is to build the
Section 5 - 43
underlying hydrodynamic model, a major task starting with a decision on the

dimensionality to be adopted.
A summary of the key components of the hydrodynamic model build is set out below:
DOMAIN – Model extent needs to be sufficient to cover all points of concern and to
retain all pollutant inside the domain. The boundaries should also be sufficiently far
from the area of interest to minimise the impact of any boundary errors. The use of a
strategic regional model versus a local domain should be considered. Broadly, a
large domain will include more areas of interest, and allow them to be modelled
together, at the expense of run-time and the generation of larger output datasets.
GRID RESOLUTION – Cell size is a balance of resolution and computational time.

Flexible mesh type models can provide increased resolution in key areas but are
constrained by the size of the smallest element in the mesh, and the overall number of
elements should be considered. Resolution should not lead to issues with numerical
dispersion. Numerical dispersion is the process of pollutant being distributed through
a model cell and is particularly an issue where grid cells are large and so
concentration of pollutant may quickly be inaccurately lowered.
BATHYMETRY – The bathymetry of the study area is required. A key consideration is

the mobility of the seabed, and whether any historical data is still representative of
current conditions. Where bathymetry is mobile, then consideration should be given
as to how this is accounted for as part of the overall modelling investigation.
HYDRODYNAMIC DATA – The applicability of historical measured data should be

considered together with the collection of new data.
Examples of data that may be required for a hydrodynamic model include:
 Salinity;
 Temperature;
 Wind data;
 Current direction data;
 Current meter data (magnitude);
 Water level data;
 Drogue data;
 Acoustic Doppler Current Profiler (ADCP) data.
BOUNDARY DATA – This needs to be suitable for the purposes of the model.
Boundary data can be generally considered to be in two forms, the outer tidal
boundaries and point source boundaries such as rivers and large discharges. If tidal
boundaries are extracted from larger scale models, then it should be ensured that the
validation of these models is acceptable and appropriate for the study being
considered. For rivers and other large sources boundary information can be obtained
from hydrology model or recorded data. In estuaries it is especially important to
consider if river sources will affect the hydrodynamics/hydrology. This is considered
unlikely in many cases, but there will be circumstances where this could be important.
In these cases, these sources must be integrated into the hydrodynamic model, and
not left for inclusion at the water quality modelling stage.
It is important to stress the need to check and test assumptions (for example domain and
grid resolution). Checking and testing these key issues at this stage minimises the risk of
the model not meeting all user requirements at a more critical stage of the investigation.
Section 5 - 44
Calibration and validation of the hydrodynamic model requires, as a minimum, tidal level
and salinity observations over tidal cycles at several points in an estuary. For 2D and 3D
models, comparisons of predicted and observed velocities is usually also expected.
Temperature may also be a factor for 3D models.
Once the hydrodynamic model is completed dispersion coefficients are identified. Initially
default model coefficients can be applied which will be subsequently adjusted to provide a
best fit with available field data.
The classic method of determining dispersion coefficients is the use of readily measured
tracers such as Rhodamine WT dye. These have the advantage of being directly
measurable in the field, allowing the creation of a dye contour or dye patch which can
then be compared with model output.
Dye studies should be designed with care, and should cover a number of scenarios in
order to provide data on a wide range of conditions and potential differences in
dispersion. High water, low water releases on both spring and neap tides as a minimum
should be considered. Experienced contractors with expertise in following, tracking and
measuring dye patches are also important in obtaining useful data. Licensing issues,
potential restrictions on use and the amount of dye required should all be considered
carefully. Dye can also be affected by local conditions. Spot releases can be instructive
as they show the general dispersion pattern quickly and relatively cheaply. However, they
can be less important in charecterising the dispersion of a continuous discharge or
dischrges of several hours duration. Thus, continuous tracer releases over at least one
hour should also be considered.
The use of alternative tracers, for example bacteriophage or spore tracers, can provide
similarly useful data and may provide additional information on the connectivity between
sources and receiving waters. These types of tracers must be sampled and analysed in a
laboratory, but in a robustly designed field programme can provide further calibration and
validation information.
A complimentary approach is the use of salinity data in instances where there is

measured data of changing salinity in different tidal stages (for example where a river
influences a bathing water, or in an estuary). Salinity data is often regularly taken with
bathing water samples, and there may therefore be a long-term dataset available.
Although not an approach available in areas where there are no freshwater inputs, it
should be used where available and where the data is appropriate.
The importance of salinity (and the use of tracers) is their conservative nature and their
predictably. Where freshwater enters seawater, there is generally a well understood and
replicable mixing of waters.
Salinity changes in the vertical profile is an important element for 3D models, and data
should be collected accordingly.
The rate of bacterial die off (T 90) is generally the key consideration for bacterial modelling.
T90 values should be set sensibly since adjusting T90 without consideration for reality will
provide any answer required. Experience and reference to previous studies and
academic literature will help in the selection of appropriate T 90 rates. The scale of the
study will affect choices regarding a varying decay rate (for example the modelling of a
specific scenario) or a representative value (for example compliance modelling of bathing
waters over several years).
 Other Water Quality Modelling
Although bacterial modelling has been the principal interest in many previous studies,
Section 5 - 45
modelling of other water quality parameters may also be required. For example, to model
DO accurately, sources of oxygen demand need to be identified and represented. A key
process in many estuaries is Sediment Oxygen Demand (SOD), and if it is omitted, DO
values are likely to be over-estimated.
Ammonia is a potentially acute pollutant with toxicity issues, but as part of the nitrogen
cycle is also a factor in nutrient enrichment.
For complex issues, the modeller may be required to represent relatively complex
interactions. There is therefore a consideration regarding parameterisation (simplified
representation of processes), probably coupled with sensitivity, to be had.
 Water Quality Verification
Where detailed bacterial compliance modelling is being carried out, using the calibrated
and validated hydrodynamic and water quality model, water quality processes (such as
decay or chemical interaction) are then incorporated for the study for which the model is
to be employed. Bacterial sampling data would be used for source characterisation, spill
volume and spill rates would be obtained from sewer models or measurements, and
estimates or assumptions used for decay rates (using available local data or published
research evidence). With correctly characterised sources, when the water quality model
is run, the model output will be comparable with sample data. This is essentially a process
of historical verification, using monitoring data. It is important to analyse the monitoring or
historical data to ensure that there have been no step changes in source character that
will affect model comparisons.
During this verification process it is sometimes necessary to adjust factors such as

decay, without considering a re-calibration of the transport model. T 90 is best represented
as a narrow band of values (to initially test). Anything outside this range will need further
justification and consideration of the specific application, the available field source
apportionment data or rarely further model calibration. Bacterial decay is dominated by
exposure to ultra-violet light, and this will be affected by turbidity, insolation, latitude, and
salinity. All these factors need to be considered when determining an appropriate range
of T90 values.
Where assumptions or estimates are used, then the sensitivity of the model to these
assumptions should be tested, where it is considered appropriate.
It is therefore at this stage that the accuracy of the prediction of bacterial pollutant
impacts is assessed. It is undertaken at this stage because it is only at this stage that a
full characterisation of sources has been undertaken, and so it is reasonable to expect
the model to represent bacterial or other water quality parameter sampling undertaken in
the field.
A ‘fit for purpose’ model must be validated or verified for the range of weather or
operational conditions that are being investigated. For example, for a compliance
investigation of a bathing water, a full range of weather conditions would need to be
accounted for in a bathing season. For a shellfish study, the full range across a year (or
years). For a direct modelling assessment of, for example, a specific bathing water
failure, then the environmental conditions (weather, tide, river flow, etc) at the time of the
failure would need to be accounted for.
 Potential Issues with Calibration, Validation and Water Quality Verification
There are several issues which can arise during the calibration/validation/verification
process. A number are considered below, to provide general guidance for commonly
encountered issues. Again, a familiarity with the general process of calibration and
Section 5 - 46
validation is assumed, and so issues directly associated with the application of tidal
models to the UPM process are discussed. Issues with observed data are often key
sources of error.
It should be remembered that all calibration processes are inherently iterative. Testing
and adjusting are part and parcel of the process.
 Verification of the water quality impacts not matching
Issues to consider when modelled and observed information do not match include:
 Checking the performance of other model inputs (river, sewer);
 Examining the output data, to see whether the data suggests a missing source, or that
the characterisation of identified sources is incorrect;
 Ensuring that key model areas function correctly – not just one point. Conversely,
there may be areas of the model where accurate performance is not required and
therefore this does not need to be focussed on;
 Whether significant changes may have occurred to the bathymetry in the area being
studied over the period being studied.
 Reliability of field data/collection or analysis errors, etcUseful outputs that can inform
overall process
 Knowledge of residuals and residence/oscillations (estuaries) – drogue releases in
model, for instance, as well as field data;
 Including supporting evidence/data along with comparison of fitness criteria;
 Demonstrate understanding of limitations of both models and fitness criteria, and field
data.
UPM models are management tools which should deliver accurate outputs regarding the
transport and impact of pollutants. As with all complex modelling, the degree of
representation of all processes in detail is a balance which needs to be achieved in a
pragmatic manner.
As with all modelling, the degree of fit should be assessed considering that the tool and the
study is not necessarily designed to the same level of detail as, for example, an academic
study seeking to understand in detail processes and relationships. The concept of
parameterisation and the representation of detail by proxy or aggregation and estimate
should not be discounted, if it is clearly explained and documented, and checks, such as
sensitivity testing, are delivered to understand the implications of assumptions.
5.6.5 Using Tidal Waters Impact Models
Broadly, there are two approaches to consider:
Compliance modelling: This is the generation of data to calculate the statistical distribution
and estimate of compliance covering a period of several years or more. This will produce a
compliance prediction directly comparable to the given percentile standard over the complete
range of potential environment impact scenarios. It generally requires the integration of
impacts from all identifiable sources on a given receiver, and, taking a bathing waters study
as an example, would typically be undertaken over at least ten years.
Section 5 - 47
This is achieved using a combination of tidal model, sewer model (and potentially river
models) together with a means of integrating multiple unit impacts to collate and analyse the
data from the water quality modelling assessment.
Direct modelling: Modelling of a shorter period - a discrete event or scenario, directly within
the tidal waters model. Typically to test the performance of a given source or sources under
prescribed conditions. While, equally applicable to single or multiple source testing, this
approach is more difficult to use to produce estimates of compliance covering a long period.
 Source identification
There are several potential sources to be considered, depending on the precise requirements
of the study:
Continuous discharges
Intermittent discharges (Surface water discharges are included in this category.)
Diffuse sources
Source identification is a critical stage, and as many stakeholders as is practicable should

be consulted. This will include operational staff in water companies (knowledge of how
assets perform), Regulator staff (may be particularly important for diffuse sources), and
may include other sources such as Internal Drainage Boards, Universities, industrial
bodies, etc.
 Supporting Data
An important consideration in any study is the amount of good data which exists, and the
level of new data collection that will be required. This is almost certainly a project specific
consideration, and the following points and issues will need to be considered to establish
specific requirements:
 The aims of the study;

 The degree of accuracy, or resolution between sources required (for example,
whether there is a requirement solely for the provision of a ‘river’ source, or
additionally for a breakdown of sources within that river);
 Any step change in source characterisation (for example asset investment) which may
render historical data obsolete;
 Sources of local data which may be of use – academic studies or PhD research can
be a valuable source of data.
As a rule, a study that attempts to represent a complex situation will require
comparatively complex supporting data - a good level of knowledge of local sources
and their characterisation will be essential. In many cases, this may be met by
existing data. Where it is not, then applicable field surveys will be necessary.
 Considerations of Model Data
In addition, and to some extent dependent on the modelling tools being used, there will be
a need to specify extraction points or areas in the model domain. Extraction points are
those points or co-ordinates where data is output from the model for further detailed
analysis. In some software, extraction points need to be identified prior to model runs. In
these cases, the agreement and selection of appropriate extraction points will be critical.
Extraction of data will generally be:
Section 5 - 48
Single extraction point, e.g. a Designated Sampling Point at a bathing water.
Multiple extraction points, e.g. several designated sampling points if several bathing
waters are modelled concurrently.
Areal - To understand overall compliance or impact over a designated area, e.g.

shellfish water, the whole bathing water area or estuary for water quality issues.
The amount of data that needs to be extracted may have an impact on model run times,
and the volume of output data generated. For some model approaches, extraction points
may need to be identified prior to the running of the tidal waters model. If further points
are required, then the model must be run again. It is therefore crucial that requirements
are identified at the outset of the project.
 Ensuring sources and receivers are modelled for all stages of the assessment
For all tidal models, the cell that a discharge occurs in, or a data extraction point is
identified by, must be ‘wet’. That is, the cell must have the minimum depth of water which
has been identified by the modeller as the minimum depth to indicate ‘wetness’ (i.e. for
transport and pollutant processes to be modelled). When water levels fall below this
depth, the cell is effectively turned off for that timestep. Discharges will not occur, and
impacts in that cell will not be calculated. The model will not necessarily indicate when
and where this is happening. For discharges which occur in the intertidal area (and many
intermittent discharges fall into this area), there is a risk that for part of a tide, there will be
no discharge from these sources, when there should be. This discharge will therefore be
under-represented in any assessment. For extraction points, impacts will be under-
represented if the cell becomes dry at any point during an impact phase.
One approach for overcoming this is to use the concept of a ‘wandering discharge’ or
‘wandering extraction point’. The points move to the next wet cell as the tide advances
and recedes, ensuring that discharge or impact occurs in a cell that is both ‘active’ in the
model and as close to the actual discharge or extraction point as possible. The track of
this wandering line should be assessed and decided according to local information,
experience and expert judgment. An alternative less useful approach is to alter the
location of the outfall or extraction point in the model to a cell that is wet all the time.
Some sensitivity testing may be required.
 Environmental data – wind and rainfall
It is likely that wind data will be required for any assessment. Wind data will be required
to test the sensitivity of a given situation to wind effects impacting on the direction and
rate of transport of pollutant (advection).
The wind data should be representative of the local conditions, cover a sufficient period,
and should be used in a manner appropriate to the study being undertaken.
If rainfall is not integrated into network or river model output for use in the tidal waters
assessment, then appropriate rainfall data may be required. Often this will be accounted
for in other models.
In all studies where rainfall is required, it is essential that all elements of the assessment
(tidal waters modelling, sewer modelling, river modelling) use rainfall data on a common
basis. This may mean that a tidal waters modelling assessment will require the
production of an upland rainfall series, where a whole catchment is being assessed.
Historical rainfall is preferable to stochastic datasets, as it can help understand temporal
and spatial extent and variability. To provide confidence in the integration of sewer model
Section 5 - 49
and river model output used in the tidal waters model, it is useful to calculate annual spill
volumes and spill frequency for each discharge for checking against monitored sewer
discharge and river modelling data.
It is essential that sewer, river and tidal modellers work together to ensure that a full
understanding of the issues exist and all parties understand what is required of each
other.
 Sewerage network data
It is likely that sewer network model data will often be used as part of a coastal modelling
investigation, and particularly for compliance assessments. To ensure that the network
model output is fit for purpose it is recommended that the following are checked:
Duration – at least ten years of contemporary data is likely to be required.
Timestep – the timestep of the output should be appropriate to the tidal waters modelling
being undertaken. It is probable that a timestep of 15 minutes will be acceptable.
Source – it is preferable to use an historical rainfall series where possible, as opposed to

synthetic series.
 River Data
River flow and quality data may be the product of river model output, or it may be the
result of field data collection and the use of local gauging data.
There is likely to be a requirement for both flow and quality data. There needs to be a
sufficient resolution to capture flow changes that arise from the wet weather events of
interest. We would suggest that daily average flows are not suitable for example..
Attention must be paid to the fact that decay of pollutants will be different in freshwater
and saline environments (for example bacterial decay is longer in rivers).
It is essential that river and tidal modellers work together to ensure that a full
understanding of the issues exist and all parties understand what is required of each
other.
5.6.6 Interpretation of results
Tidal models, especially when used to deliver compliance results, offer the opportunity to gain
large amounts of information regarding the transport and impact mechanics of pollution and
the importance of specific sources. Used in a direct mode, they provide clear and
unequivocal evidence of impact and significance – provided they have been set up and used
correctly, with the right degree of experience and expertise involved at each stage.
The other key element is their use in solution evaluation. The outputs can be used to direct
solution development, and then to test solution options. Models offer the potential to deliver
optimised solutions, using post processing or other complimentary analytical tools. The
optimisation process can be applied over multiple sources. There are several issues to
consider in the interpretation of modelling results:
 Sensitivity
Processing power, data availability and experience in modelling studies has improved the
understanding of how to apply coastal models to impact analyses, and as such the default
Section 5 - 50
position that models are likely to be inaccurate, and therefore large amounts of sensitivity
or conservative assumptions are required are outdated. There is of course, an important
role for sensitivity analyses, especially when the potential assumptions required for the
characterisation of sources is considered. Some sensitivity testing is essential because
the modelling system is, by definition, a simplified representation of reality.
A reasonable starting point for sensitivity testing is to consider those assumptions or

estimates that could reasonably introduce uncertainty into the modelling study.
Sensitivity tests should therefore be included as a part of the whole modelling study, with
agreement as to which parameters should be tested, or the conditions in which sensitivity
testing will be required. Common considerations for sensitivity testing would be:
 Bacterial decay rate;

 Pollutant concentration;
Other parameters could be considered as necessary. There may be situations where the
bathymetry of the model may be subject to sensitivity. This is more likely in areas of
mobile mudflats or sandbanks, where there can be gross changes in the channel
morphology, and as such the transport of pollutant can be significantly changed. Models
can still be useful in such situations, but the validated model bathymetry will only
represent one set of conditions, and tests should be undertaken to investigate the
significance of channel migration.
 Source Apportionment
Output from the models can isolate each source, as well as combining all the sources to
predict overall impact. This means that the models can provide insight into the
proportional significance of each source to performance at or in a sensitive receiver. An
understanding of the proportional significance can help in improving performance or
environmental quality. Apportionment can also demonstrate the relative importance of,
for example, point and diffuse sources.
The maximum value will arise from this analysis where sources are resolved to an
individual source level, but there is value in a study delivering data regarding the overall
contributions of assets, private sources and diffuse, for example (where it may be that
diffuse pollution is represented as a single representative source).
 Solution Evaluation
Where discharges are identified as contributing to or causing a breach of a water quality

standard, then mitigation solutions will need to be considered. Dependent on the nature
of the impact and the key sources, these might entail:
 Enhanced levels of wastewater treatment;

 Improvements to intermittent discharges (for example by increased storage to reduce
spills);
 Better operational interventions;
 Catchment improvements (for diffuse sources).
In developing solutions, tidal waters models provide the initial starting point of designs, as
these models have determined either the level of non-compliance, the key sources or
both. Where detailed compliance modelling is done then the models provide predictions
of the amount of pollutant which needs to be removed from the environment to deliver
compliance.
Section 5 - 51
 Where spill frequency or spill duration solutions are a regulatory requirement,

modelling studies may still offer several benefits:
 Such studies help identify those intermittent discharges that are significant to
compliance;
 They can test the extent to which solutions based on meeting different types of
standards will provide compliance;
 They also present a means of testing potential design solutions, and therefore
solutions can be optimised.
 Providing underpinning evidence to understand, articulate and maximise the benefits
versus costs of potential improvements.
5.7 Integrated urban pollution modelling

This is the final section on modelling tools, and discusses the need to develop and build a
tool that can combine the results from models of the individual components. The preparation
of an integrated tool forms a discrete step within the process. Such a tool is required to allow
the rapid simulation of many rainfall events so that compliance with the standards may be
tested, in Section 6 of the UPM Procedure.
This section is subdivided as follows:
5.7.1 The need for an integrated modelling tool;
5.7.2 Audit trails;
5.7.3 Transferability of results between models;
5.7.4 Simplification.
5.7.1 The need for an integrated modelling tool
In carrying out a UPM study, planners/modellers must select the various components
described in the previous sections, tailor them to the particular situation being studied and
use them in an integrated way. The strength of the UPM approach lies in this holistic
approach that seeks to take account of all the interacting factors that can contribute to wet
weather pollution problems.
The complexity and multiplicity of the UPM tools is also a potential weakness of the approach
and this needs to be recognised and addressed. Specifically, what is needed is a tool or
procedure that allows the integration of the different UPM components so as to promote their
efficient use in testing the compliance of proposed solutions with standards in Section 6.
Individuals and organisations approaching this issue will undoubtedly develop their own
procedures and tools for this purpose - there is no prescriptive way of achieving efficient
integration. However, the following points need consideration in developing such procedures
and tools.
5.7.2 Audit trails
One of the complexities in UPM modelling is the large number of runs with different event
details, different model components and different model configurations. Results data are
often transferred from one model to become input data for another model and the results
from, say, the river impact model for a large number of events are finally processed to reach
a conclusion about compliance or non-compliance.
Section 5 - 52
It is essential that a sound system is developed to record run and version details so that a
clear audit trail exists. This is needed to track back from the final compliance assessment
calculation to the event details that have been simulated to support the assessment.
5.7.3 Transferability of results between models
Spill volumes and loads generated by sewer models often become input data for river impact
or marine impact models. There are similar transfer interfaces between sewer and STW
models and between STW and river models. Problems can arise at these transfers because
of different file formats, different units and different time steps. In addition, there can be
incompatibility between the determinands used in different models.
Sound procedures are needed to identify and resolve these issues early in a study. The aim
is to achieve efficient transfer of data between models and also to fully appreciate the
implications of any differences in the way critical quality processes are represented.
5.7.4 Simplification
Many of the models described in earlier sections are detailed deterministic simulation models
designed to provide an accurate representation of system performance under wet weather
conditions. In theory, it is possible to run these models with large numbers of rainfall events
and different background conditions to demonstrate the performance of sewer/STW systems
and to compare this performance with the appropriate standards. Current performance can
be assessed, and, by repeating these runs with, for example, extra sewer storage, potential
solutions can be evaluated. However, this procedure can be very onerous because of the
time and effort involved in multiple runs with complex models. Some compromises must be
madeWith increased computing power, these problems will diminish.
An alternative approach, that can overcome the current limitation, is to create a simplified
model of the urban system, calibrate this against a small number of detailed model results
and then use the simplified model for multiple runs, and possibly for continuous simulation.
This approach also involves compromises, as some accuracy is lost in the description of
individual events. However, this loss of accuracy is more than compensated for by the greater
range of event simulations that are possible and, hence, the greater overall confidence in
performance assessment.
Using a simplified urban pollution model allows greater account to be taken of the variability
in rainfall, river conditions, marine conditions and foul flow quality and how the interaction of
these factors affects the probability of exceeding the critical standard thresholds. In some
situations it may be appropriate to simplify just one component of the whole system. For
example, a very detailed sewer flow model may be grossly simplified and yet retain good
hydraulic representation at critical overflows. Dynamic river impact models with multiple
cross-sections and complex water quality processes can also be greatly simplified once the
critical aspects have been identified. Often it is possible to assess performance using steady-
state runs rather than dynamic runs.
In other situations, it may be useful to simplify all the elements of the urban system (sewers,
STW and river) into one integrated model. Potentially, a good integrated model will take
account of all the data transfers between different elements and will provide a sound basis for
establishing audit trails.
A simplified model can be very focused as a design tool in the task of checking compliance
with specific standards. In developing such a model, a planner needs to have a good
appreciation of what is important and what is not important for the task in hand. This will
Section 5 - 53
involve judgements based on a very good understanding of the water quality processes in
operation in the study area.
Section 5 - 54

Section 5

Uploaded by

Copyright:

Available Formats

Section 5

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Section 5

Uploaded by

Copyright:

Available Formats

Section 5

Construct UPM Tools

From Section 4, Data Collection

To Section 6. Assessing Performance

5. SECTION 5: CONSTRUCT UPM TOOLS

The section is subdivided as follows.

5.1 Introduction to modelling tools.

5.2 Rainfall modelling.

5.3 Sewer flow and quality modelling.

5.4 Sewage treatment works quality modelling.

5.5 River impact modelling.

5.6 Tidal waters modelling.

5.7 Integrated urban pollution modelling.

5.1 Introduction to modelling tools

5.2 Rainfall modelling

5.2.1 Rainfall requirements for UPM modelling;

5.2.2 Historical rainfall time series;

5.2.3 Synthetic rainfall time series;

5.2.4 Event definition.

5.2.1 Rainfall requirements for UPM modelling

5.2.2 Historical rainfall time series

• Selection of the most appropriate gauges;

5.2.3 Synthetic rainfall time series

• average and range of annual rainfall totals;

5.2.4 Event definition

5.3 Sewer flow and quality modelling

This section is subdivided as follows.

5.3.1 Hydraulic performance of sewer systems.

5.3.2 Sewer quality performance under wet weather conditions.

5.3.3 Modelling sewer quality performance.

5.3.4 Determinands modelled.

5.3.5 Using SQMs.

5.3.1 Hydraulic performance of sewer systems

An adequate understanding of how a sewer system performs hydraulically is a basic

a) Detailed sewer flow models (SFMs)

b) Simple sewer flow models (SFMs)

In certain situations it may be beneficial to represent CSO performance in a sewer

5.3.2 Sewer quality performance under wet weather conditions

• advection and dispersion of pollutants; and,

Figure 5.1 Example of observed pollutographs in sewer flow - Site 1

Figure 5.2 Example of observed pollutographs in sewer flow - Site 2

b) Build-up and wash-off of sediments from the catchment surface

c) Deposition and erosion of sewer sediments

d) Sediment transport in sewers

e) Sediment partitioning in tanks

f) Advection and dispersion

5.3.3 Modelling sewer quality performance

• simple tank simulation models;

5.3.4 Determinands Modelled

The pollutants modelled can be assumed to be conservative in that no biochemical changes

SQMs typically contain submodules to represent:

The representation of surface wash-off varies between dynamic models. Generally,

5.3.5 Using SQMs

• the likely extent of the pollution problems;

5.4 Sewage treatment works quality modelling

This section is subdivided as follows.

5.4.1 Sewage treatment works performance under wet weather conditions.

5.4.2 Modelling sewage treatment works.

5.4.1 Sewage treatment works performance under wet weather conditions