Start with a model purpose
Define the decisions the model must support: growth servicing, pressure compliance, transfer capacity, storage operation or outage response. This determines the network detail, demand scenarios, simulation duration and observations required. A model can be suitable for one purpose and inadequate for another.
Assemble a traceable baseline
- Check pipe connectivity, diameters, roughness assumptions, node elevations and unit systems.
- Represent reservoirs, tanks, pumps, valves and control rules using consistent levels and datums.
- Reconcile allocated customer demand and non-revenue water with the system water balance.
- Record the survey, GIS and operational evidence behind material assumptions.
Calibrate, then validate
Compare simulated flows, pressures and tank levels with time-aligned measurements. Diagnose incorrect connectivity, boundary conditions and demand patterns before using roughness as a universal correction. Validate with a different operating period or event so that agreement is not limited to the calibration dataset.
Read the result in context
Hydraulic head is not pressure head. At a junction, pressure head is hydraulic head minus elevation. Test extended operation, not just a single snapshot, and check whether initial storage conditions influence your conclusion.
EPANET and InfoWorks WS Pro provide different workflows for these assessments.
Separate calibration parameters from known inputs
Treat measured elevations, pipe dimensions, valve states and recorded pump operation as evidence to check. Treat unknown or uncertain parameters as candidates for estimation within plausible bounds. If a valve is closed in reality but open in the model, changing roughness to match one pressure observation can conceal the topology error.
Prepare a parameter register and note which observations can identify each parameter. Several combinations of demand, roughness and source head can sometimes match the same pressure data. Add flow or storage observations and different operating periods to distinguish those explanations. More fitted parameters do not necessarily produce a more credible model.
Align measurements in time and reference
Pressure loggers measure pressure relative to the local gauge reference. Convert to head using the appropriate units, density and instrument convention before comparing with model output. Add the gauge elevation if comparing hydraulic head rather than pressure head. Keep the gauge elevation distinct from a nearby node’s ground level.
Synchronise logger, telemetry and model clocks, including daylight saving and interval definitions. A one-hour offset in a pump cycle can look like an incorrect demand pattern. Record sensor accuracy, calibration history, data gaps and any known operational interventions.
Use a period long enough to include the relevant tank and pump behaviour. Avoid selecting only the hours that agree. If the model is intended for peak-day planning, successful matching under a quiet average-demand period does not establish that it represents the high-flow resistance.
Use residuals as evidence, not just a score
State the variable, units, sites, times and weighting used in a summary statistic.
A positive pressure residual means the model overpredicts pressure under this convention. Inspect residuals by location and time before reducing them to one number. A near-zero mean can result from large positive and negative errors cancelling. A low overall RMSE can hide a poor fit at the few locations governing a design decision.
Plot pressure, flow and storage traces together with pump and valve states. A nearly constant offset suggests a different mechanism from an error that grows with flow. Use that pattern to test hypotheses. Calibration should create an explanation for the agreement, not merely optimise a metric.
Challenge the accepted model
- Validate against a period or event not used to tune the parameters.
- Check mass continuity, storage recovery, pump duty and control switching.
- Test credible uncertainty in demands, boundary heads and poorly known assets.
- Apply the scenario matrix appropriate to growth, outages and service requirements.
- Report where the model is suitable and which conclusions remain sensitive to unresolved inputs.
A model may support a trunk-capacity comparison while lacking the detail needed for local fire-flow assessment. Explicit suitability limits make its results more useful. For implementation, the EPANET guide starts with independent flow and head checks, while WS Pro focuses on the demand evidence behind the solution.
Sources & further reading
- EPANET 2.2 User Manual ↗US Environmental Protection Agency · 2020
External source · Checked 24 September 2026 - InfoWorks WS Pro overview ↗Autodesk · Living product documentation
External source · Checked 24 September 2026 - Best Practice Modelling Guidelines ↗eWater · 2011
External source · Checked 24 September 2026
Source findings are distinguished from editorial interpretation. Apply current local criteria and project evidence when making engineering decisions.