Although many of the counter-measures against pollution incidents relate to physical security, modeling has a central role to play in identifying the key parts of the system to protect and analyzing what happens once a pollutant has entered the water supply network. A good model can provide support in both developing plans of what to do in the event of a pollution incident, and managing incidents if they occur and as a tool for investigating a past event so as to understand what happened.
The first requirement is to base any pollution analysis on a detailed, calibrated, and validated model that is already in place. Water quality modeling requires a all-mains model, US EPA requirements are the minimum of 50% of all pipe lengths or 75% of all pipe volumes, or perhaps going to individual customer connections. Whichever level is chosen (and there are obvious run-time consequences to the choice), the model must contain all pipes and all valves at this level skeletonized models and broad assumptions will not give accurate results or results detailed to the customer level, which is essential. During the skeletonization process, hydraulic equivalency is preserved in terms of flows and pressures. However, hydraulic equivalency of velocities is not preserved, and consequently water quality results obtained form a skeletonized model might not reflect the reality.
The other requirement is the ability to handle daily, weekly, and monthly variations in water demands, as pollutant tracking simulation often extend over several days or weeks. Demand profiles that are simply averages across the day will not produce sufficiently accurate results. To be able to simulate such a long incidents modeling software must provide a fast calculation engine.
The mechanics of pollution tracing are obvious. The pollutant is introduced into the system at a point, either at one of its sources such as a reservoir, or at a node of the network itself. The pollutant is then carried into and through the water network, changing the concentration as it mixes with the clean water. Its path through the network, the consumers whose water it contaminates, and its concentration at every stage are the key parameters that modeling can establish.
To predict the passage of the pollutant and its concentration at each stage, models need to simulate both the flow of the pollutant and its growth, decay or conservation as it passes through the network. It may react with the water or with the pipes, or it may simply remain as a weakening solution in the water as it flows through the network. However, even a weak solution can be of great concern some poisons are dangerous at concentrations of a few parts per million. Additionally, long exposures to low concentrations of some pollutants can be as dangerous as short exposures to high concentrations.
The answers required from a pollution simulation include which customers will receive contaminated water and at what concentration? How quickly will the contaminant reach them? How can the affected area be isolated? What is the most efficient way to flush the pollutant out of the system? Can water be re-routed around this area of isolation? Are there alternative means of supplying the affected areas or other areas facing shortfall?
To answer these questions the model must represent all mains, all valves, and an accurate demand profile at the time of the incident. It must also be able to take new demand profiles as the impact of the incident changes demand from the normal profiles. Often the response to a pollution incident will drastically change the demand profiles. In some areas people will stop consuming the polluted water and the demands will drop while there will be a surge in demand in other areas as the water is rerouted.
There are two uses for models in tracing pollutants. The first is in formulating emergency response plans, which involves proposing scenarios, assessing alternative responses, and identifying the scale of the impact. From this analysis plans can be compiled to counter and minimize the impact of the various types and levels of threat.
Second, a model can be used in the control room to provide decision support during an incident. In this case, as the incident is managed, the success of the various measures for isolating customers can be simulated and checked against the data coming in from SCADA and telemetry systems. These real-time readings can even be accepted as inputs by the best models. SCADA data is often the best predictor in emergency situations of the current system demands and water distribution paths, both of which can drastically change during the incident.
Another key factor is that the model should link easily with customer data files. Knowing the contact details of the affected properties is essential for customer service and informing them of the nature of the incident and what action should be taken. The capability to run SQL queries across this data is the key to quick identification of the affected customers.
Such a very brief summary of the mechanics of pollution tracing cannot cover all the details of using models to address this critical problem, but it should serve as an outline. However, whatever the details of the mechanics, the essential qualities of the modeling package needed for pollution incidents are similar to those for other modeling applications, namely:
A model that is sufficiently detailed to give accurate answers at the customer level this means a package that can work with large numbers of pipes and customer points
A fast and stable simulation engine that can produce results quickly enough on a detailed model
A model that has already been proven to engineers in previous water supply analyses.
The ability to input a time-varying pattern for the pollutant
Links to customer information database for a quick identification of affected properties
Live SCADA connections to calibrate new demands patterns and validate model results in real-time
Flexible demand management to allow a fast evaluation of different demand scenarios