The role of hydraulic models in Wastewater Based Epidemiology: Opportunities and Limitations

Wastewater analysis for SARS-CoV-2 RNA has been underway in various locations across the world since early 2020 and is a rapidly expanding area of research. There is a limited understanding of the behaviour of viruses within the sewer network. However, it is likely that the virus is subject to a range of biotic and abiotic factors which may influence its persistence and transit time within the network . Understanding this behaviour is key in developing sampling strategies which underpin an effective Wastewater Based Epidemiology (WBE) health surveillance system. Developing this understanding requires answers to a number of core questions, namely:

•Does the virus follow conservative flow regimes?
•Does the virus degrade in the sewer environment?
• How long does the virus persist in the network?
•What are the impacts of re-entrainment after weather events?

A better knowledge of these factors would allow better interpretation sample results in the context of the environmental conditions, which will also create the foundation for being able to triangulate and track the sources of viral outbreaks.

In the UK over 90% of the water and wastewater networks are representing in hydraulic models. These models have been built over the past 10 – 15 years and have been subject to millions of pounds of investment to ensure they are fit for purpose and accurate. These models have the potential to be a key tool in an effective WBE program if used in the right way. To date, development has primarily been focused on flooding, which is the core driver in the UK, and less so on the representation of water quality parameters.

Method

Through this study we worked to assess how well these models could be used to represented viral transport in sewer network, as well as identify any potential limitations.

To do this we worked in two catchments, each with different properties to deploy an enveloped viral marker at known places into the sewer network and then tracking its appearance at various locations. This work was complemented with the conservative flow fluorescent dye tracer, rhodamine-WT. We followed these tracers through the use of an autosamplers in the network over 17 days. Using these results, we then calibrated a hydraulic model (with contaminant transport processes) to understand the key parameters which influence transport (degradation rate, flow rate, rainfall, pipe material and roughness).

The models use the Innovyze Infoworks ICM modelling platform. In all cities the models were 'cut down' to represent only the major network with inflow files at junctions, as in Figure 1. This was done to increase the speed of the models.

In order to ensure accuracy these cut down models are checked against the full model to make sure there are no differences. The conditions on the sampling days were then re-created by adding any rainfall and inputting the markers at the upstream manhole, representing it as coliform.

Results

The different catchments each highlighted different challenges and opportunities. In the first there was a pumping station in between the input location and the sampling point. This meant that initially the model under predicted the travel time by over 4 hours. To counter this the impact of the pump was derived by including its operational controls. After this was done the time of travel matched that of the experiments closely, as seen in Figure 2. In our simulations we ran a number of scenarios with different T90 decay times. In Figure 2 it can be seen that a T90 of 10 – 12 hours most accurately matches the experiments. This influences the design of a sampling strategy or interpretation of the result to incorporate this decay of the virus.

In the second catchment there was a single straight run of pipe, with no significant hydraulic structures. However, we were unable to recreate the travel time accurately, as shown in Figure 3.

The models use a standard Dry Weather Flow day based upon when they were originally verified. We compared velocity measurements taken on the day of the experiments to the validation day and there was a large differential. This may be explained by changing water usage during Covid altering the flow patterns in the sewer, as was suggested by the flow monitor at the outlet point, as shown in Figure 4.

Conclusion

The key conclusion from the work is that it is possible to represent an enveloped virus in the sewer network through the use of hydraulic models. Both in terms of travel time and decay rate. However, using the models 'off the shelf' can lead to large inaccuracies as the models may not be validated/ developed with the intention for use in water quality modelling or water usage has changed during Covid with home working to the point at which the models are no longer representative. In these conditions having flow measurements recorded at the same time as sample extraction can lead to more certainty around the outputs/ results and allow for the models to be adapted to match reality.

If the models are fully validated it is possible to use the water quality modules to represent the virus transport with a T90 decay rate of 10-14 hours. This means that sampler coverage needs to ensure that areas outside of this distance from a sampler need to be covered by additional in-network samplers to represent SARS-CoV-2 prevalence. Or include scaling factors to incorporate the likely decay of the virus.
The transport of the virus is conservative (via advection), as a result if the key parameter of interest is travel time this can be calculated based on only knowledge of the network configuration and the flow rates. However, the models are still required to explore the relationship with decay rates.

Future work

The opportunity to represent other health markers and chemicals is being explored as well as whether these models can be run in real time and integrate real time sample data to track the sources of infection in communities.