The Storm Water Management Model (SWMM) is a dynamic rainfall–runoff model developed and maintained by the United States Environmental Protection Agency (EPA) for simulating the quantity and quality of runoff, principally in urban areas. It represents how precipitation becomes surface runoff and how that runoff moves through pipes, channels, storage facilities, pumps, and flow-control structures. SWMM can simulate a single rainfall event or continuous operation over extended periods, supporting the analysis and design of stormwater, combined-sewer, and sanitary-sewer systems.
Development and scope
SWMM was first developed in 1971 through an EPA-sponsored effort involving several organizations. Successive versions expanded its treatment of runoff generation, drainage hydraulics, and pollutant transport. SWMM 5, released in 2005, introduced a substantially rewritten computational engine and an integrated graphical interface.
EPA distributes the model and its source code without charge. Its computational engine is written in C, while EPA’s desktop graphical interface is designed for Windows. The engine can also be incorporated into other software environments. Commercial and open-source applications may use the SWMM engine while supplying different interfaces, mapping tools, or additional analytical capabilities; those additions are not necessarily features of the standard EPA distribution.
Although developed primarily for urban drainage, SWMM can represent nonurban catchments as well. Its distinguishing emphasis is the interaction between runoff-producing land surfaces and engineered drainage infrastructure, rather than comprehensive simulation of every process within a large natural watershed.
Model structure
A SWMM project combines a representation of the land surface with a network that carries and stores water.
Subcatchments
The modeled area is divided into subcatchments, each representing a portion of a drainage basin that drains toward a designated outlet. A subcatchment is characterized by parameters such as area, slope, representative flow width, surface roughness, and impervious fraction.
SWMM distinguishes between pervious and impervious surfaces. Pervious areas allow infiltration, while impervious areas generate runoff once available surface storage is filled. The model can route runoff between surface types within a subcatchment or toward another subcatchment or drainage-network node.
Subcatchments are conceptual units rather than necessarily exact representations of individual parcels. Their delineation and parameterization determine which spatial variations the model preserves and which it averages.
Nodes and links
The drainage system is represented as a connected network:
- Nodes include junctions, outfalls, storage units, and flow dividers.
- Links include conduits, pumps, orifices, weirs, and controlled outlets.
Conduits can represent closed pipes or open channels with several cross-sectional shapes. Outfalls specify downstream discharge conditions, including fixed or time-varying water levels. Storage units describe facilities whose available volume changes with water depth.
External inflows can represent dry-weather wastewater, imported flow records, or rainfall-derived infiltration and inflow entering a sewer network.
Meteorological and operational inputs
Rainfall can be supplied through rain-gage time series. Other inputs support processes such as evaporation and snow accumulation and melt. Operating rules can change pump status or regulator settings in response to simulation time, water levels, and other modeled conditions.
The temporal resolution of these inputs matters. A rainfall record suitable for annual runoff accounting may be too coarse to reproduce rapid peak flows in a small urban catchment.
Hydrologic processes
Surface runoff
SWMM treats each subcatchment surface as a nonlinear reservoir. Water accumulates through precipitation and runon, while infiltration, evaporation, and runoff remove water.
A simplified water balance is
where is surface water storage and each is a volumetric rate.
Runoff begins when surface water exceeds the depression-storage threshold. Its rate is calculated using a form of Manning’s equation that relates flow to surface roughness, slope, representative width, and water depth. Depression storage represents water retained in small surface irregularities or otherwise unavailable for immediate runoff.
Infiltration and subsurface flow
SWMM provides alternative infiltration formulations, including Horton, Green–Ampt, and curve-number methods. These approaches differ in their physical assumptions, data requirements, and treatment of infiltration capacity or soil moisture.
An optional groundwater component represents exchanges involving unsaturated and saturated subsurface storage. It can simulate groundwater contributions to the drainage system. This is a conceptual catchment-scale representation, not a substitute for a spatially detailed groundwater-flow model.
Rainfall-derived infiltration and inflow into sewers can also be represented using unit-hydrograph methods. This process is distinct from infiltration of surface water into soil: it describes additional wet-weather water entering the sewer network through pathways such as defects and connections.
Continuous simulation
Continuous simulation tracks changing conditions between storms, including soil wetness, surface storage, groundwater storage, and pollutant accumulation. Consequently, storms with identical rainfall can produce different responses depending on antecedent conditions.
This capability is important when evaluating annual runoff volumes, the frequency of sewer overflows, or the performance of storage and infiltration facilities across many events.
Hydraulic routing
SWMM offers three principal routing methods with different assumptions and computational demands.
Steady-flow routing
Steady-flow routing transfers flow through the network without representing time-dependent hydraulic storage and wave propagation in detail. It is useful for relatively simple analyses but cannot reproduce many interactions found in constrained or surcharged drainage systems.
Kinematic-wave routing
Kinematic-wave routing represents the movement and attenuation of flow under assumptions that simplify the momentum equation. It is suitable where gravity-driven flow predominates and downstream conditions do not substantially influence upstream water levels.
It does not represent backwater, flow reversal, or pressurized pipe flow in the way required for complex sewer-flooding analysis.
Dynamic-wave routing
Dynamic-wave routing solves a one-dimensional form of the Saint-Venant equations, representing conservation of mass and momentum. It can simulate backwater, flow reversal, surcharge, pressurized conduits, and interactions among storage, pumps, and hydraulic controls.
For a conduit without lateral inflow, continuity can be expressed as
where is flow cross-sectional area, is discharge, is distance along the conduit, and is time.
Dynamic-wave routing is the most general of the standard routing options, but it also requires greater attention to network geometry, initial conditions, time-step selection, and numerical stability. Greater hydraulic detail does not compensate for inaccurate input data.
Water-quality simulation
SWMM can track user-defined pollutants through runoff generation and drainage-network transport. Its water-quality functions include:
- Pollutant buildup on land surfaces during dry periods.
- Pollutant washoff during runoff.
- Different buildup and washoff characteristics for different land uses.
- Pollutant concentrations in external inflows.
- Transport and mixing within the drainage network.
- Specified decay and treatment relationships.
These functions support estimates of pollutant loads, combined-sewer discharges, and the relative effects of management measures. A model may, for example, represent suspended solids washed from paved areas and routed through a detention facility.
The standard model does not provide a comprehensive mechanistic description of sediment transport, aquatic ecology, or every chemical reaction. Detailed receiving-water effects or wastewater treatment processes generally require additional models or explicitly defined approximations.
Low-impact development controls
SWMM includes representations of low-impact development controls, also associated with green stormwater infrastructure. Available control types include bioretention cells, rain gardens, green roofs, infiltration trenches, permeable pavement, rain barrels, rooftop disconnection, and vegetative swales.
These controls are represented through combinations of surface, soil, pavement, storage, drainage, and drainage-mat layers, depending on the control type. The model calculates water balances among the applicable layers.
This allows comparison of alternatives such as:
- Reducing runoff volume through infiltration and evapotranspiration.
- Delaying discharge through temporary storage.
- Diverting roof runoff away from directly connected impervious surfaces.
- Reducing peak inflow to an existing sewer network.
Performance depends on design assumptions, soil properties, underdrain settings, and antecedent wetness. Detailed flow and volume calculations do not automatically imply equally detailed predictions of pollutant removal.
Applications and workflow
SWMM is used in drainage design, sewer-capacity assessment, overflow analysis, flood investigation, and urban planning. It can compare pipe upgrades with storage facilities, test pumping or gate-control rules, and assess how land-cover changes affect runoff.
A typical modeling workflow comprises:
- Defining the question. Peak flood depth, annual runoff volume, overflow frequency, and pollutant load require different data and levels of detail.
- Building the physical representation. Catchment boundaries, pipe dimensions, invert elevations, structures, and outlet conditions establish the system geometry.
- Preparing forcing data. Rainfall, evaporation, external inflows, and downstream water levels define the conditions being simulated.
- Checking model behavior. Connectivity, water balances, hydraulic plausibility, and numerical diagnostics are examined.
- Calibrating and validating. Simulations are compared with observations before alternative scenarios are evaluated.
Geographic information systems are commonly used to prepare spatial data. SWMM can also be connected to external optimization tools, operational software, or research workflows through its engine interfaces and supporting libraries.
Calibration, uncertainty, and limitations
Calibration adjusts uncertain parameters to improve agreement with observations such as discharge, water level, runoff volume, or pollutant concentration. Validation evaluates performance against independent observations rather than only the data used for fitting.
Different parameters can produce similar outputs, especially when observations are sparse. Matching a peak discharge alone does not establish that runoff volume, timing, infiltration, and hydraulic storage are all correctly represented.
Important limitations include:
- Input uncertainty. Rainfall variability, undocumented sewer connections, sediment deposits, and inaccurate elevations can substantially affect results.
- Spatial simplification. Subcatchments aggregate surface characteristics, while sewer conduits are predominantly represented in one dimension.
- Surface-flooding representation. Standard SWMM is not a general two-dimensional overland-flow model. Water exceeding node capacity may be treated as flooding loss or stored through specified ponding assumptions. Detailed flood-path and inundation mapping may require coupling to a two-dimensional surface model.
- Numerical sensitivity. Abrupt transitions, short conduits, rapidly changing controls, and unsuitable time steps can cause oscillation or convergence difficulties.
- Scenario uncertainty. Future rainfall, land use, maintenance, and operating conditions are assumptions, not predictions produced by SWMM itself.
Continuity reports indicate whether the numerical simulation adequately conserves water and pollutant mass. A small continuity error is a useful diagnostic, but it is not evidence by itself that the model accurately represents the real drainage system.