How a stormwater model turns rain into runoff, routes it through inlets, pipes, channels and streets, and shows where it floods. Covers design rainfall through master plan analysis. Everything is on this page, free, with no sign-up.
A stormwater model follows rain from the moment it lands until it leaves the system. It shows how much runs off, how fast it arrives, whether the inlets, pipes and channels can carry it, and where the rest goes when they cannot. That last question, where the water goes, is what separates a drainage model from a flood model.
Map flood depth, extent and duration for a design storm, and count the structures affected.
Check every inlet, pipe and culvert against the level of service it was meant to provide.
Compare peak flows before and after development, and size the detention that offsets the difference.
Compare alternatives by flood damage avoided per dollar, and phase them into a capital plan.
Every urban drainage system is two systems on top of each other. The minor system is the inlets, pipes and ditches, sized for frequent storms. The major system is the streets, swales, open channels and low ground that carry what the pipes cannot. Good flood modeling represents both, and the exchange of water between them.
Pipe capacity, flood mapping or development review set the tools needed.
LiDAR, storm GIS, soils, land cover, rainfall, complaints.
Catchments, imperviousness, infiltration.
Inlets, pipes, culverts, channels, the 2D surface.
Against gauges, high water marks and known flooding.
Design storms, flood maps, deficiencies.
Pipes, detention, green infrastructure, channels.
Benefit, cost and phasing into a capital plan.
Every stormwater result starts with a rainfall input, and the choice of storm often moves the answer more than any model parameter. A design storm has three parts: how rare it is, how long it lasts, and how its rain is spread over that time.
| Return period | Annual chance | Chance in 30 years | Typically used for |
|---|---|---|---|
| 2-year | 50% | > 99% | Water quality, channel protection, small inlets |
| 10-year | 10% | 96% | Storm sewers and inlets (minor system) |
| 25-year | 4% | 71% | Larger trunks, culverts under local roads |
| 100-year | 1% | 26% | Major system, building protection, floodplain mapping |
A "100-year storm" is not one that happens once a century. It has a 1% chance in any year, and a 26% chance of happening at least once over a 30-year mortgage.
In the US, rainfall depths for each return period and duration come from NOAA Atlas 14 (intensity-duration-frequency data for a point). Pick a duration long enough for the whole catchment to contribute: short, intense storms (1 to 6 hours) stress small urban pipe systems, while 24-hour storms stress ponds, channels and larger basins. Many studies run several durations and keep the worst result at each location, called the critical duration.
Plan for a changing climate. Atlas 14 is based on past records. Many cities now add a climate factor, often 10 to 20% more depth, to design storms for long-lived infrastructure. Agree the factor before any design work begins.
Only part of the rain becomes runoff. Some soaks into the ground, some sits in puddles and on leaves, and the rest flows over the surface toward an inlet or channel. Paving a catchment changes all three, and so changes both how much water runs off and how fast it arrives.
The time of concentration is how long water takes to travel from the farthest point of the catchment to its outlet: sheet flow, then shallow concentrated flow, then flow in a gutter, pipe or channel. A shorter time means a higher design intensity and a higher peak.
SWMM treats each subcatchment as a sloping plane split into pervious and impervious parts. Runoff from each part is computed with Manning's equation as the surface ponds above its depression storage. The key inputs:
| Input | What it controls | How to estimate it |
|---|---|---|
| Area | Total volume of rain caught | Delineate from LiDAR to each inlet or outfall |
| % impervious | How much of that rain runs off quickly | Land cover and building footprints. Directly connected impervious area matters most. |
| Width | How fast the runoff arrives (the shape of the hydrograph) | Area divided by the longest overland flow path; a main calibration parameter |
| Slope | Overland flow speed | Average slope along the flow paths, from the DEM |
| Manning's n | Surface resistance | About 0.011–0.015 for pavement, 0.15–0.4 for grass and woods as overland sheet flow |
| Depression storage | Rain held in puddles before runoff starts | About 0.05–0.1 in impervious, 0.1–0.3 in pervious |
| Infiltration | Losses to the soil on pervious areas | Horton, Green-Ampt or Curve Number, from NRCS soil groups A (sandy) to D (clay) |
SWMM uses the same nodes and links as a sewer model, plus a few elements that matter mostly for stormwater: open channels, culverts, street inlets and green infrastructure.
| Element | Represents | Key inputs | Common mistake |
|---|---|---|---|
| Junction | A manhole, catch basin or channel confluence. | Invert, maximum depth to the rim or ground, ponded area. | No ponded area, so water that leaves a node is lost from the model instead of returning when the pipes drain. |
| Inlet | A street inlet that captures gutter flow into the pipe below (SWMM 5.2 and later). | Type (grate, curb opening, combination, slotted, drop), size, clogging factor. | Assuming every drop of street flow reaches the pipe. Real inlets bypass flow, especially on steep streets. |
| Conduit: pipe | A storm sewer. | Shape, size, length, Manning's n (0.012–0.015), offsets. | Pipe n used for corrugated metal, which is much rougher (0.022–0.027). |
| Conduit: open channel | A ditch, swale or stream. | Cross-section transect from survey or LiDAR, n for bed and banks. | A LiDAR cross section that misses the water below the surface at the time of the flight. |
| Conduit: culvert | A pipe or box under a road. | Size, shape, and a culvert code for its inlet shape, so SWMM checks inlet control. | Leaving out the culvert code, so the model ignores inlet control and overstates capacity. |
| Storage unit | A pond, basin or underground vault. | Stage-storage curve from the grading plan or survey. | Counting the permanent pool of a wet pond as flood storage. |
| Orifice and weir | A pond outlet structure: low-flow orifices, overflow weirs, emergency spillway. | Size, invert or crest, coefficient. | Crest heights entered as elevations instead of heights above the node. |
| Outfall | Where the system discharges. | Boundary: free, normal, fixed stage, tidal or a time series of river levels. | A free outfall into a river that is also high during the same storm. |
| LID control | Green infrastructure on a subcatchment. | Type (bioretention, permeable pavement, green roof, rain barrel, infiltration trench) and its layers. | Full performance assumed with no clogging or maintenance. |
In a dual drainage model, each street segment is a conduit drawn above the pipe beneath it, with the street cross section as its shape. Inlets move water from the street to the pipe. When the pipe is full, water stays on the street, or flows back up out of an inlet, and continues downhill. This is how a 1D model shows the major system. Section 5 covers the 2D alternative.
Water in a pipe or a well-defined channel flows in one direction, so a 1D model describes it well. Water spreading across streets, yards and parking lots flows in every direction, and only a 2D model follows it reliably. The right tool depends on where the water goes when the pipes are full.
| Approach | How it works | Use it when |
|---|---|---|
| 1D only (SWMM) | Pipes, channels and, with dual drainage, streets as links. | Pipe capacity studies; flow stays in known paths. |
| 2D only (HEC-RAS 2D) | The ground surface as a mesh of cells. Rain can fall directly on the mesh ("rain on grid"). | Overland flooding, riverine floodplains, areas with no pipe system. |
| Coupled 1D–2D | Pipes in 1D, the surface in 2D, exchanging water at inlets and manholes. | Urban flood mapping, where both pipes and surface flow matter. |
Recent HEC-RAS versions can model pipe networks inside the 2D mesh. Many studies instead couple a SWMM pipe network to a 2D surface model.
A bare-earth LiDAR DEM, typically 1 to 3 ft resolution in urban areas. Burn in channels and culvert openings that LiDAR cannot see; add buildings as obstructions.
Cells of about 10 to 50 ft in urban areas, smaller along streets and channels. Add breaklines along curbs, road crowns, levees and embankments, so cell faces follow the features that steer the water.
Manning's n by land cover: about 0.015–0.02 for pavement, 0.03–0.05 for lawns, 0.08–0.15 for woods. Buildings are blocked out or given very high roughness.
Rain on the mesh or inflow hydrographs from 1D catchments, and downstream water levels where the mesh meets a river or the sea.
Set it from the Courant number: smaller cells and faster flow need shorter steps. An unstable run shows up as oscillating depths or mass balance errors.
Storm GIS is usually the weakest data a utility owns: inverts are missing, ditches are not mapped, and culverts are drawn where they were planned rather than where they were built. Plan for field verification from the start.
| Data | Typical source | Used for |
|---|---|---|
| Terrain | LiDAR DEM (USGS 3DEP or local flights) | Catchment delineation, slopes, 2D surface, rim elevations |
| Drainage network | Storm GIS, as-builts, development plans | Inlets, pipes, culverts, ditches, outfalls |
| Field survey | Culvert and channel surveys, inlet inventories | Inverts, sizes and shapes the GIS lacks |
| Land cover | Impervious surfaces, building footprints, aerial imagery | % impervious, roughness, building obstructions |
| Soils | NRCS SSURGO hydrologic soil groups | Infiltration parameters, curve numbers |
| Rainfall and flow | Rain gauges, radar rainfall, stream gauges, level loggers | Calibration storms |
| Flood evidence | Complaints, high water marks, photos, insurance claims | Validation and problem areas |
A master plan usually models pipes 15 to 18 inches and larger, every culvert, and every channel. Smaller pipes are added where flooding is known.
Confirm flow direction, connect pipes to inlets and outfalls, fill missing inverts by interpolating between known points, and flag every guess for field checks.
Use the DEM to draw the area draining to each modeled inlet or channel point. Adjust by hand where curbs, walls and roads steer water in ways the DEM misses.
Overlay impervious cover, soils and slope to get % impervious, infiltration, slope and width for each subcatchment.
Survey the critical ones. Enter culvert codes, channel transects, and stage-storage curves with their outlet structures.
Outfalls into rivers or tidal water need a realistic tailwater: a stage that rises with the storm, not a free outfall.
Either a dual drainage street network or a 2D mesh linked to inlets and manholes (Section 5).
Keep SWMM's continuity errors under about 1 to 2%. Look for nodes flooding in small storms, which usually point to data errors rather than real problems.
Stormwater models rarely have as much measured data as water or sewer models. Many are calibrated against a few gauges and validated against what people saw: high water marks, photos and complaint records. Use every source; each one tests a different part of the model.
Temporary level loggers in pipes and channels, and USGS stream gauges. They test runoff volume, peak and timing.
Debris lines and stains surveyed after a flood. They test flood depths and the 2D surface.
Where and how often flooding happens. They test whether the model floods in the right places.
Use the nearest gauges, or gauge-adjusted radar for each subcatchment. Rainfall error is often the largest error in the model.
In small storms almost all runoff comes from directly connected impervious area. Adjust % impervious until runoff volumes match.
In larger storms pervious areas start to contribute. Adjust infiltration and depression storage.
Adjust subcatchment width and overland n until the hydrograph rises and peaks at the right time.
Compare with high water marks. Large misses point to a missing culvert, a blocked inlet, a wrong tailwater or terrain the DEM got wrong.
Run storms and flood events not used in calibration.
| Measure | Common target |
|---|---|
| Peak flow, per storm | +25% to −15% (CIWEM Urban Drainage Group) |
| Runoff volume, per storm | +20% to −10% |
| Depth in pipes and channels | Within about ±0.1 m (4 in); surcharged +0.5 m to −0.1 m |
| Continuous flow record | Nash-Sutcliffe efficiency above 0.5 is often called satisfactory |
| High water marks | Typically within 0.5 to 1 ft, allowing for the uncertainty of the marks themselves |
| Reported flooding | The model floods where people reported flooding, and not where they did not |
Run the design storms and compare results against level-of-service criteria. Stormwater criteria are less about pipes being full and more about where the water ends up: on the road, across a driveway or inside a building.
| Criterion | Typical value | Design storm |
|---|---|---|
| Storm sewers | Carry the flow with the hydraulic grade below the inlets or rims | 2- to 10-year |
| Street spread | Keep at least one travel lane open on collectors; both lanes on arterials | 10-year |
| Road flooding | Depth no more than about 6 in at the crown for emergency access | 100-year |
| Culverts | Headwater no more than 1.5 times the culvert height (HW/D ≤ 1.5); no road overtopping | 10- to 100-year by road class |
| Buildings | No water above finished floor; often 1 ft of freeboard below it | 100-year |
| New development | Post-development peak no higher than pre-development | 2-, 10- and 100-year |
These are common values, not a standard. Use your city's drainage criteria manual, state DOT rules for roads, and FEMA rules for mapped floodplains.
Pipes move water faster; storage slows it down. A detention pond does not remove runoff. It holds the peak back and releases it slowly, which is why it can bring the post-development peak below the pre-development one (Section 3).
From the grading, the volume held at each water level.
From the outlet structure, the outflow at each water level. A typical multi-stage riser has a small orifice for the water quality storm, a larger orifice or weir for the 2- and 10-year, and an emergency spillway for the 100-year.
For each time step, inflow minus outflow equals the change in storage. SWMM's storage unit does this automatically; by hand it is the storage-indication method.
Each design storm must meet its peak target, and the 100-year must pass the spillway with freeboard to the top of the embankment.
Watch the timing. A pond that delays its peak can make flooding worse downstream if the delayed peak lands on top of the peak from another catchment. Check ponds at the watershed scale, not only at their own outlets.
| Practice | How it works | Best at |
|---|---|---|
| Bioretention / rain garden | Ponds runoff on a planted soil layer that filters it and lets it soak in | Frequent small storms, water quality |
| Permeable pavement | Runoff passes through the surface into a stone storage layer below | Parking lots and low-traffic roads |
| Green roof | A thin soil layer holds rain and returns it to the air | Volume reduction where there is no ground space |
| Infiltration trench | A stone-filled trench stores runoff and lets it infiltrate | Sandy soils (groups A and B) |
| Rain barrel / cistern | Captures roof runoff for later use | Small storms, reuse |
Green infrastructure is strongest in frequent storms, which carry most of the yearly runoff volume and pollution. In a 100-year storm it usually fills early, so flood control still needs pipes, channels or detention. SWMM models each practice as an LID control with its soil, storage and drain layers.
A stormwater master plan lists the flood problems in a city, the projects that solve them, and the order to build them in. Because flood projects compete for limited money, the plan has to show which projects prevent the most damage per dollar.
Combine model results with complaints, maintenance records and staff knowledge into a single list of problem areas.
Existing conditions, future land use at buildout, and future rainfall with a climate factor. Run each at the design storms set by the level-of-service criteria.
Every inlet, pipe, culvert and channel that breaks a criterion, and every street and building that floods. Group neighboring problems with one cause into a single project area.
For each area, test the options below. Model the whole watershed with each alternative in place, so you can see whether it moves flooding downstream.
Use depth-damage curves for each flooded building to estimate damage in each storm. Integrate across storms to get expected annual damage. The benefit of a project is the expected annual damage it removes.
Capital, operating and maintenance costs over the project life. Rank by benefit-cost ratio, with life-safety problems such as deep road flooding and flooded underpasses placed first regardless of ratio.
Build downstream capacity before upstream improvements that send more water to it. Coordinate with road reconstruction, and look for grant funding, which often requires a benefit-cost ratio of at least 1.0.
| Option | How it helps | Watch for |
|---|---|---|
| Upsize pipes and inlets | More water off the street, faster | Sends more water, sooner, to the next downstream problem |
| Replace culverts | Removes a bottleneck and road overtopping | Loses the storage the old culvert created upstream |
| Regional detention | Cuts peaks for a whole watershed | Needs land; timing with other tributaries |
| Channel improvements | More conveyance in open channels | Environmental permits; erosion downstream |
| Green infrastructure | Less runoff in frequent storms, water quality credit | Limited effect in large storms; long-term maintenance |
| Overland flow paths | Safe routes for water the pipes cannot carry | Needs space between buildings and street grading |
| Floodproofing or buyouts | Protects the few buildings no project can reach economically | Community acceptance |
Most stormwater studies still do these steps by hand: weeks of catchment delineation, one design storm and duration at a time, and flood maps that are out of date as soon as the next development is built. The Stormwater practitioner track teaches you to automate the same workflow on your own watershed.
Build the SWMM and HEC-RAS 2D models from GIS and LiDAR by script, run every storm and duration in the cloud, calibrate against your own gauges and high water marks, and drive it all through an AI agent. Two days, live online, on your own catchment.
Free. We'll send an email when we add sections, new free courses or practitioner session dates. Unsubscribe anytime.