The DWMP blog – Episode 15. How do we model environmental impact?

If you haven’t already seen the earlier episodes in this series, they are all here (https://tinyurl.com/MartinOsborneArticles) I suggest that you start from Episode 1 (https://tinyurl.com/DWMP-blog).

In this episode I consider how we model compliance with the FIS standards to protect river ecology.?I described the background to these standards in Episode 14.

There are two (or more) components that we have to model to assess the environmental impact of sewerage discharges: what happens in the sewer that affects what is discharged and what happens in the watercourse that affects the impact that it has.

In the sewer

The starting point is a model of the flow of sewage through the system. ?Once we have this we can also model the movement of pollutants with the flow.?However, this is complicated because some of the pollutants are embedded in sediment particles that are deposited at the bottom of the sewer during low flow and eroded and moved on in high flow.?This gives rise to the “first foul flush”, a short flush of polluted flow at the start of a rainstorm.?An important part of managing the impact of sewerage discharges is catching the first foul flush in detention storage so that it can be passed forward to treatment.?A model that assumes that the concentration of pollutants is always the average value (defined as an Event Mean Concentration or EMC) therefore underestimates the benefits of adding detention storage.?This is a problem, as detention storage is often the most cost effective method of reducing impact on the environment.

The alternative to using EMC is to model the movement of sediment through the system.?This is never going to give exactly the right results as it depends on the exact values of depth and velocity of flow, roughness of the sewer surface and size and density of the sediment.?However, it can give a reasonable estimate of how the pollutant concentration varies with time and location in the system.

Calibrating each area of the model against measured data would be good, if it could be done; but it is almost impossible.?To measure all of the pollutants of interest requires an automatic sampler to take samples for analysis in the laboratory.?These typically take a sample every 15 minutes.?The concentration predicted by the model is varying from minute to minute so it is almost impossible to compare the two.?Does the graph show agreement or disagreement between model and measured?

Even if we were convinced that we needed to adjust the calibration, what do we change??The most significant parameters are the size and density of the sediment but this will have different effects in different parts of the model depending on how much sediment is deposited.

We should do some verification of the overall model results by comparisons at the variation of pollutant load at the treatment works and we should do sensitivity testing to show the potential range of results, but attempting to calibrate each part of the model is mostly a waste of time and money.

In the river

The processes that we need to model in the river are:

• Mixing and travel
• Pollutant decay
• Re-aeration

Normally it is assumed that discharges mix instantaneously across the width and depth of the river as a 1D model is used. ?2D models to represent partial mixing can be used for more complex situations such as estuaries.

The simplest type of model is a so called “mass balance” model.?This generally works on a one-hour timestep and mixes the discharges during the hour with the upstream flow to calculate the resulting flow and concentrations downstream.?It then calculates travel time of this plug of flow to the end of the defined reach assuming that it is a constant cross section and gradient and then calculates the oxygen sag caused at the end of the reach by the pollutant decay and reaeration.

An extension of this method uses several reaches one after another when more discharges are mixed in for each reach.?Each reach has a different constant cross section and gradient.

There are two limitations of the mass balance approach that means that it can underestimate the pollutant impact.

1. The first is that it operates at a one-hour timestep to assess compliance with a standard defined at one hour.?It will therefore often split the worst hour of a discharge into two separate hours of river flow and so provide more dilution.?Using a 15 or 30 minute timestep solves this.
2. Secondly it calculates the oxygen sag at the end of the defined reach rather than at the location where the worst-case conditions occur.

A more accurate assessment can be made using a model of the river that represents the changing cross section and gradient and calculates the varying velocity and longitudinal mixing of the pollutants as they travel down the river.?This operates at a shorter timestep and so also correctly captures the impact of short duration discharges.?

A further improvement is for the model to be calibrated against observed data for travel time and optionally for pollutant decay and reaeration rates.

The Storm Overflow Assessment Framework (SOAF)

The recently published UK SOAF sets out four levels of modelling depending on the complexity of the situation.?The levels were intended so that the simpler assessment methods incorporate a larger factor of safety.?Moving to a more complex method should therefore give a reduction in the cost of the required improvements to the sewerage system although an increase in the cost of the assessment.?However, note the comments above about underestimation of the mass balance approach.

The four levels of the SAOF are summarised in the table below.

So, how can the SOAF be improved?

Event Mean Concentration does not represent the benefits gained from the use of storage and should not be used.?The alternative of a verified sewer pollutant model is the opposite extreme in cost and is difficult to justify for all but the most sensitive locations.

The “Statistical” method uses a random selection of flow or pollutant concentration from a statistical distribution based on measured data.?However, the two are selected separately and this does not account for any link between them.?For example, high flows having lower concentrations or vice versa.

The choice of assessment point for dissolved oxygen as the end of an arbitrarily defined fixed length reach does not ensure that the worst case is identified.?The assessment point needs to represent the worst oxygen sag and this varies with flow.?This can be done by directly calculating the worst case irrespective of where it occurs.

River hydraulic and water quality models are now much more readily available and should be used more widely than mass balance models.

Alternative framework

I have taken these limitations into account to develop an alternative framework of modelling levels that will deliver better results and better value.?Guidance on when to apply each level still need to be defined.

I have defined a method for the water quality of treatment works discharges and upstream river flows called Statistical+ where the pollutant concentrations are correlated with the flow but then with an additional random variation.?This is likely to be difficult for defining the upstream river quality because there will be insufficient data on pollutant concentrations to derive the correlation unless a targeted survey is carried out.

The proposed alternative approaches are shown in the table below.

To ensure that the simpler methods have a built-in factor of safety, I recommend that for Level 1 the worst-case ammonia and the worst-case dissolved oxygen are assumed to occur at the same location even though this is unlikely.

Summary

The published SOAF does not necessarily give the best balance between accuracy and effort and a review is recommended that aligns it more closely with the EA consenting guidelines.?A more complex set of rules is required on when to use each of the assessment levels.?This can be based on dilution, flow characteristics and special features such as weirs or protected habitats.