When pipes fill, the water still needs a destination
Urban dual drainage represents two connected water pathways: the below-ground drainage network and the above-ground routes used by water that bypasses inlets or returns to the surface. A pipe model is incomplete for a surface-flood question if overflow simply disappears from the account.
Inlet capacity, street geometry, depressions, downstream water levels and pipe pressure can change where water goes. A larger pipe may not solve an inlet bottleneck, while an apparently adequate inlet may be unable to discharge into a full receiving network.
This guide uses the US EPA Storm Water Management Model documentation to ground the distinction, then develops original simplified examples. It is an explanation of modelling logic, not a drainage design approval, a flood-safety map or advice to inspect operating drains during storms.
Hydrology subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Draw the connected system
Separate rainfall, runoff and hydraulic routing
Map streets and pipes as connected networks
Give bypass flow a route
Follow capacity and exchange
Separate inlet limitation from pipe limitation
Distinguish surcharge from surface flooding
Represent ponding and surface storage physically
Test the simulation
Choose the surface representation for the question
Set downstream and initial conditions deliberately
Test mass continuity and time-step sensitivity
Interpret improvements and limits
Validate internal behaviour, not only the outlet
Compare interventions as system changes
Prepare a reproducible dual-drainage review
CHAPTER 1 OF 12
1. Separate rainfall, runoff and hydraulic routing
Rainfall is an atmospheric input. Runoff generation determines how much water leaves surfaces and stores toward drainage pathways. Hydraulic routing determines how that water moves through streets, channels, pipes and structures. A model can be wrong at any stage, and a good fit at one outlet does not prove every internal process is represented correctly.
Start with an original one-hectare teaching area. A uniform 20-millimetre rain supplies 200 cubic metres because 0.020 metres × 10,000 square metres = 200. If a deliberately simplified accounting assumption sends 60 percent to the drainage system during the chosen period, generated runoff is 120 cubic metres. That assumption is a scenario, not a universal runoff coefficient for urban land.
The 120 cubic metres does not specify a peak flow. Releasing it uniformly over ten minutes gives 0.20 cubic metres per second; over twenty minutes it gives 0.10. Real runoff varies through time. The hydrograph shape matters to inlet capture, temporary storage and pipe pressure even when total volume is unchanged.
Draw three separate boxes: rainfall and losses, runoff pathways, and hydraulic network. Label transfers between them. This prevents the common error of placing runoff directly into pipes while also applying the same water to the streets. The model must account for each generated volume once.
Transfer check: keep rainfall total fixed but shorten its intense portion. The runoff-generation response and hydraulic bottlenecks can both change. Do not assume that equal storm totals imply equal flooding, or that a pipe-capacity calculation can recover the missing rainfall chronology.
CHAPTER 2 OF 12
2. Map streets and pipes as connected networks
A below-ground network contains junctions, pipes, storage, pumps, controls and outfalls as relevant. The surface system contains streets, channels, low points, barriers, openings and other overland routes. Connections between them occur at inlets, manholes or structures under specific hydraulic rules.
The EPA SWMM manual describes dual-drainage representation using street and sewer conduits connected through inlet capture. This is more specific than adding a generic overflow number at each node. The model needs to know where bypass and backflow go, and what controls their movement.
For a fictional street, draw upper junction A, low point B and downstream junction C. Place a pipe network beneath it and identify which inlet sends water to which sewer node. A small error in that connection can transfer water to the wrong catchment even if every pipe diameter is correct. Verify connectivity independently of the map’s appearance.
Use common reference levels for surface and pipe information. Ground elevation, pipe invert, rim elevation and depth are related but different. Confusing an invert elevation with a pipe depth can create a physically impossible connection. Record the vertical datum and units before importing several survey sources.
Transfer check: a kerb reconstruction blocks the former route from B to C. The pipes may be unchanged, but the surface network has changed. A model that updates only underground assets can miss the new ponding mechanism. Surface connectivity is infrastructure information, not merely a backdrop behind the drainage map.
CHAPTER 3 OF 12
3. Give bypass flow a route
An inlet captures some approaching surface flow under its geometry and hydraulic conditions. The remainder bypasses it. If the model routes every drop into the sewer automatically, it removes the very inlet limitation a dual-drainage analysis may need to examine.
Use a fictional steady approach of 0.10 cubic metres per second for five minutes. Suppose the first inlet captures 0.06 while 0.04 bypasses to a second inlet. The first capture is 18 cubic metres and the bypass is 12. If the second inlet captures 0.03, it takes 9 cubic metres, leaving 3 to continue downstream. The original 30 cubic metres is accounted for as 18 + 9 + 3.
Those fixed capture rates are teaching assumptions. Real capture depends on inlet design, depth, approach flow, street slope, local depression and receiving conditions. An on-grade inlet with flow passing it differs from an inlet at a sag where water accumulates. Use an appropriate hydraulic relationship rather than a fixed percentage selected to obtain a desired flood extent.
The second inlet receives the first inlet’s bypass, not the original approach flow again. Counting its capture from the original full flow would duplicate available water. This simple sequential calculation is an excellent test of a more complicated network model.
Transfer check: clog the first inlet so it captures only 0.03. Repeat the ledger. The second may receive more water but cannot necessarily capture all of it. Identify where the additional bypass travels and which low point gains storage. The purpose is to reveal a shifted pathway, not merely to lower one capacity number.
CHAPTER 4 OF 12
4. Separate inlet limitation from pipe limitation
Street flooding can occur because water cannot enter a drainage system fast enough even when downstream pipe capacity remains available. It can also occur because the receiving network is full or pressurised. These are different bottlenecks and can require different responses.
Imagine an inlet able to capture at most 0.04 cubic metres per second under the scenario while its receiving pipe could convey 0.10. With 0.08 approaching on the street, half bypasses despite unused nominal pipe capacity. Enlarging that pipe alone does not remove the stated inlet restriction.
Now reverse the example. An inlet could capture 0.08, but the receiving node cannot accept that rate because downstream conditions constrain the sewer. Treating inlet capacity as a guarantee would force too much water underground. A coupled representation needs to respect both the surface approach and receiving hydraulics.
Inspect results that distinguish capture, bypass, sewer level, pipe flow and surface depth. A single maximum pipe utilisation percentage cannot diagnose all surface flooding. Likewise, a flooded-node summary does not reveal whether the mechanism began with inlet bypass, sewer backflow or an external surface inflow.
Transfer check: a maintenance action restores inlet opening but a high downstream water level persists. Explain why surface performance may improve less than expected. Then lower the downstream level while leaving the inlet blocked. The paired test separates mechanisms more effectively than changing several parameters at once until a model appears satisfactory.
CHAPTER 5 OF 12
5. Distinguish surcharge from surface flooding
A sewer can become pressurised without water necessarily emerging onto the ground at every location. Surcharge and flooding are related but not identical. The interpretation depends on hydraulic heads, conduit and node definitions, rim levels and the model’s treatment of excess volume.
Think of a closed pipe connecting two nodes. If the water level rises above the pipe crown but remains below the surface opening, the pipe may be surcharged without surface discharge at that node. If head reaches an accessible opening and the exchange rule permits outflow, water can enter the surface system. The actual geometry and hydraulic representation determine the transition.
The EPA manual’s dual-drainage approach returns relevant sewer flooding to the street network through specified inlet connections rather than simply removing it. This lets surface water continue moving and potentially return to the sewer later. A different configuration may handle flooding differently, so inspect the actual model settings and version.
Do not artificially raise rim or surcharge settings merely to eliminate a flooding report without a physical basis. That can trap water numerically in a system that would discharge at the surface. Conversely, treating every pressurised conduit as a surface flood can exaggerate the visible consequence.
Transfer check: seal one opening while leaving another lower opening connected. Water may emerge at the other location rather than vanish. The network response is not purely local. This is why asset modifications should be tested across the connected surface and pipe systems rather than judged by a single node’s improved summary.
CHAPTER 6 OF 12
6. Represent ponding and surface storage physically
When water enters a low point, depth depends on stored volume and geometry. A constant ponding area is a simplifying relationship, not a surveyed flood extent. Real depressions can expand with depth, spill across thresholds or connect to another flow path.
For a fictional flat-bottomed storage area of 200 square metres, 20 cubic metres produces a mean depth of 0.10 metres under the constant-area assumption. The same volume over 100 square metres produces 0.20. If a surveyed depression widens as it fills, a depth–area or depth–volume relationship is more suitable than either constant area.
Suppose the depression spills to a downstream route after storing 30 cubic metres. A simulation storing 50 indefinitely without a spill connection has changed the physical system. It may conserve total water while misrepresenting depth and location. Continuity is necessary but does not prove the geometry is correct.
Record buildings, walls, kerbs, openings and important thresholds at the scale required by the decision. A coarse terrain surface can smooth a small barrier or create a false connection. A finer grid does not automatically solve the problem if the underlying survey or building representation is wrong.
Transfer check: open a formerly blocked passage at the low point. Maximum local depth may fall while downstream flow rises. State both consequences. Reducing one ponded volume is not automatically a system-wide reduction in risk; water has moved somewhere and its new route needs examination.
CHAPTER 7 OF 12
7. Choose the surface representation for the question
A one-dimensional street network can represent selected overland pathways and their cross-sections. A two-dimensional surface model can represent spreading across a terrain domain. Neither is automatically superior for every task. The appropriate choice depends on geometry, data, flow complexity and the consequence being evaluated.
If water is mostly confined to a known street corridor, a carefully surveyed one-dimensional representation may answer a bounded question efficiently. If water splits around buildings, crosses plazas and changes direction across broad shallow areas, a surface representation capable of resolving those pathways may be needed. The choice should follow the mechanism, not visual appeal.
For coupled models, define the exchange locations and laws. Water transferred out of one system must enter the other with a consistent sign and time basis. The same inlet should not appear as two independent unlimited drains. Exchange can also reverse, so a one-way link may be inadequate for a backflow question.
Use a small synthetic test before a large urban domain. Put a known pulse into a simple street, constrain the receiving sewer and verify where the excess accumulates. Then remove the constraint and check return flow. The test should expose duplicate transfer, lost volume and wrong exchange direction without thousands of distracting assets.
Transfer check: introduce a building that divides the flow into two paths. Ask whether the chosen one-dimensional network explicitly represents both. If not, either rebuild that network appropriately or justify a different surface model. A rendered animation cannot compensate for a pathway absent from the mathematics.
CHAPTER 8 OF 12
8. Set downstream and initial conditions deliberately
An outfall is a boundary condition, not a magic sink. River level, tide, control operation or another receiving condition can influence the drainage network. A free-outfall assumption may be inappropriate for a question involving backwater or high receiving levels.
Consider a fictional drainage system that performs adequately when the outfall level is low. During the same rainfall with a higher receiving level, discharge capacity can be reduced and upstream heads can rise. The rainfall has not changed; the downstream hydraulic condition has. A sensitivity run separates that mechanism from a presumed increase in runoff.
Initial storage matters too. A depression or detention facility already partly full has less available capacity. A dry initial pipe network and an initially wet one can respond differently. Record warm-up or initialisation procedures rather than treating the default initial state as an observed fact.
If pumps or controls are represented, preserve their operating rules, limits and failure scenarios. A model assuming perfect operation cannot establish performance during a power failure or maintenance outage. That does not make the normal-operation model useless; it defines the scenario it answers.
Transfer check: align the rainfall peak with a receiving-water peak, then shift the timing while preserving both maxima. The joint timing can change the outcome. Comparing only separate design maxima loses the chronology and dependence relevant to the connected system.
CHAPTER 9 OF 12
9. Test mass continuity and time-step sensitivity
A hydraulic model should account for input, output and storage changes over the simulation. A small continuity residual is an important numerical check, but a good residual cannot prove correct terrain, inlet placement or rainfall. It answers an accounting question, not every scientific question.
Use an original whole-system ledger: runoff input 120 cubic metres, outfall discharge 70, final increase in pipe storage 10 and surface storage 35. The remaining 5 is an unresolved difference under the stated account. Search for omitted losses, initial storage treatment, reporting boundaries or numerical error before declaring the result complete.
Internal transfers should cancel in the combined account. If 30 cubic metres moves from streets into pipes, it is an output from one subaccount and input to another, not a new external supply. Counting captured water as fresh system input would inflate the budget. Keep surface and sewer ledgers separate and then combine them consistently.
Reduce the routing time step and inspect relevant outputs: peak depth, time of peak, exchanged volume, outfall flow and continuity. A stable total volume can coexist with an unstable short peak. Also test reporting intervals; a coarse output interval can hide a rapid maximum even if the internal solver used smaller steps.
Transfer check: make a street shorter and steeper while retaining the old time settings. Reassess numerical behaviour. A setting adequate for a slow network may not resolve a faster pathway. Numerical convergence should be evaluated for the stated decision rather than inferred from a completed run without warnings.
CHAPTER 10 OF 12
10. Validate internal behaviour, not only the outlet
Several different parameter combinations can match an outlet hydrograph while producing different internal surface depths. For a local flood question, outlet agreement is therefore insufficient. Use observations appropriate to the locations and mechanisms being claimed.
Potential evidence includes surveyed levels, reliable timed photographs, water-level records, inlet observations, maintenance records and documented flood extents. Each has limitations. A photograph may show a wet boundary without a precise time or datum. A high-water mark may identify a maximum but not the sequence. Preserve those limits rather than treating all observations as equally exact.
Use one set of events for calibration and another for evaluation where feasible. Include different rainfall patterns and downstream conditions. If every validation event is used to retune inlet capacity, it is no longer independent validation. Record parameter changes and the physical evidence supporting them.
An original diagnostic comparison might show correct outlet volume but excessive low-point depth. Investigate surface geometry and exchange pathways rather than immediately reducing rainfall until the depth matches. That adjustment could destroy the correct water balance while disguising the internal error.
Transfer check: a model calibrated on a clean-inlet event is used during a debris-rich storm. Treat inlet condition as a scenario or observed state, not an unchanging fitted constant. The real-world system and the calibration system may differ in a mechanism central to the outcome.
CHAPTER 11 OF 12
11. Compare interventions as system changes
A useful intervention comparison states what physical property changes and what remains fixed. Inlet maintenance, pipe enlargement, surface storage, rain gardens and controlled overland routes act through different mechanisms. They should not be treated as interchangeable percentages of flood reduction.
For a fictional low point, compare three actions separately: restore a clogged inlet, add ten cubic metres of surface storage and enlarge a downstream pipe. Track local depth, bypass, sewer head, outfall discharge and downstream consequences. One action may improve the low point while increasing a downstream peak; another may mainly delay flow.
Keep rainfall, initial conditions and receiving-water scenarios common across the alternatives unless the scenario itself is being tested. Use multiple relevant events rather than selecting only the storm where one option wins. Include maintenance and failure states when those affect the intended benefit.
Avoid equating reduced modelled surface volume with verified safety. Consequences depend on depth, velocity, duration, access, buildings and people. A model does not authorise walking or driving through water. Actual emergency decisions belong to responsible authorities using current conditions.
Transfer check: add a rain garden but allow overflow to a different street. Test the new pathway and storage state. Nature-based measures still require explicit hydraulic connections and maintenance assumptions; their label does not remove the need to account for water under exceedance conditions.
CHAPTER 12 OF 12
12. Prepare a reproducible dual-drainage review
A concise review package should include the model version, surface and pipe schematics, vertical datum, rainfall basis, runoff-generation assumptions, inlet definitions, exchange rules, storage geometry, downstream boundary, initial state, controls and numerical settings. Add observed evidence and a list of unresolved assets or levels.
Reproduce the 30-cubic-metre sequential-inlet calculation and the 120-cubic-metre system ledger from this guide. These exercises catch double counting and disappearing overflow before the model becomes visually complicated. Then alter one mechanism at a time: clogging, receiving level, surface barrier or spill threshold.
Inspect maps together with time series. A maximum-depth map combines maxima that may occur at different times; it is not necessarily a single simultaneous water surface. Use event chronology when explaining how water moves from an upstream street to a downstream low point.
For students, connect the explanation to first principles: conservation of water, geometric storage, capacity constraints and connected pathways. Ask why a wider pipe does not always solve a flooded street and why lowering local depth can move risk elsewhere. The answer should trace water rather than repeat a software feature list.
The final conclusion should identify the scenarios for which the model is supported, the controlling bottlenecks, the outputs that remain sensitive and the observations that would change the decision. A bounded explanation is more useful than a claim that a model proves the whole neighbourhood cannot flood.
Work one final storage-and-return example. A fictional low point receives 24 cubic metres of sewer backflow while its outlet is blocked. Assume it occupies a constant 120-square-metre area and has no other flow during that phase. Its mean depth rises by 0.20 metres. When the receiving sewer recovers, suppose the inlet admits 0.02 cubic metres per second for ten minutes. It returns 12 cubic metres underground, leaving 12 on the surface and a mean depth of 0.10 metres under the same geometry. The water has not all disappeared when the sewer stops surcharging.
Now add a surface spill after depth reaches 0.15 metres. At the stated constant area, storage at that threshold is 18 cubic metres. If the same 24 arrives and the spill removes the excess promptly under this simplified scenario, 6 cubic metres travels downstream rather than remaining at the low point. After the later 12-cubic-metre return to the sewer, 6 remains locally. Track all three destinations: 12 returned, 6 spilled and 6 stored. Their sum remains 24.
These two alternatives share the same backflow input and the same later inlet return but give different local depths and downstream volumes. A storage node without the spill connection cannot answer the second scenario correctly merely because it conserves mass. The exercise separates the continuity check from the topology check.
As a final transfer, replace constant area with an expanding depression. The stored volumes still balance, but the depths must be recomputed from the new geometry. Do not carry the 0.20-metre result into the new case. This is the same first-principles discipline used throughout the guide: preserve the conservation relationship, rebuild the geometric or hydraulic relationship that changed, and explain which conclusion survives. It turns a software output into a reasoned account that another person can inspect.
Worked comparison
| Stage | Approach or stored volume | Captured or discharged | Remaining |
|---|---|---|---|
| First inlet over five minutes | 30 m³ | 18 m³ | 12 m³ bypass |
| Second inlet, same five minutes | 12 m³ | 9 m³ | 3 m³ downstream |
| Whole teaching route | 30 m³ | 27 m³ captured | 3 m³ bypass |
| Independent system ledger | 120 m³ input | 70 m³ outfall + 45 m³ storage | 5 m³ unresolved |
Questions and practical checks
Does surcharge always mean water is on the street?
No. A pipe can be pressurised while the hydraulic head remains below a surface opening. Flooding depends on geometry, heads and the exchange representation.
Why does a larger pipe sometimes fail to fix ponding?
The controlling restriction may be inlet capture, a surface barrier, downstream water level or another bottleneck. Diagnose the pathway before selecting an intervention.
Can a small continuity error validate a flood map?
No. It supports numerical accounting. The geometry, boundary conditions, runoff and internal water levels still require evidence.
Is a two-dimensional surface model always required?
No. Choose the representation needed for the actual flow paths and decision, with adequate data and validation.
What should a learner do first?
Draw both networks and every exchange. Work the inlet ledger, then show where uncaptured and backflow water goes. No volume should disappear because it is inconvenient.
Continue with the connected subjects
Flood risk engineering — connect hydraulic outcomes to exposure and consequences
Urban water-sensitive design — follow interventions through their water pathways
Urban rain-garden management — preserve function through maintenance
Rainfall time steps and equal storm totals
Sources and scope
US EPA SWMM — official model documentation and reference manuals
US EPA SWMM 5.2 user manual — streets, inlets, dual drainage and routing
Primary research — coupled Iber and SWMM urban pluvial flood modelling
The numerical cases in this guide are original fictional teaching examples, not measurements from a real site. This is an educational explanation, not a field safety instruction, engineering approval or an official certification. For a real project, use qualified practitioners, current product documentation and site-specific evidence.