Flood Modelling Software Explained: 1D, 2D and Coupled Models

Flood modelling software solves the equations that describe how water moves, then reports the answer as levels, depths, velocities and extents. Which package is used matters far less than how many dimensions the model solves in: one, two, or the two coupled together. That one decision sets what the model can physically represent, what it costs to build, and whether a regulator will accept the result.

The Environment Agency has never published a list of approved flood modelling software. What it requires instead is evidence that the package you chose was suitable for the question being asked.

This explainer covers what the software calculates and how 1D, 2D and coupled 1D/2D flood modelling differ in practice. Three flood modelling techniques account for nearly all UK work, and the type of flood model chosen decides most of the rest. It then sets out which packages are in use on UK projects in 2026, what became of the older names still in circulation, and what the Environment Agency expects of any hydraulic model put forward as evidence. Where a site needs more than national mapping can give, Unda's consultants build bespoke 1D, 2D and coupled flood models to fit it.

What flood modelling software actually calculates

Flood modelling software does two jobs, and one word usually hides both of them. Hydrological modelling estimates how much water arrives and when, as a flow hydrograph or a rainfall profile. Hydraulic modelling then routes that water across the ground and through channels, culverts and pipes, solving for depth and velocity at every point and every time step. The output is a set of flood levels. The map everyone looks at is a rendering of those levels, one step removed from the calculation. The numbers come first.

The hydraulic half rests on the shallow water equations, a simplification of fluid motion which assumes water is far wider than it is deep. That holds for rivers and floodplains, which is why the same mathematics sits under every package in common use. What separates them is the numerical scheme, the geometry it is solved over, and how much of the physics each one keeps.

Some schemes keep the full equations, including the inertial terms that let a flood wave accelerate. Others drop those terms and solve a diffusion wave approximation. It runs faster and behaves well on gentle floodplains. It understates momentum wherever water moves quickly: through a breach, down a steep street, around a tight bend. HEC-RAS offers both and makes the modeller choose. That tells you how much it matters.

• Terrain. Usually LiDAR, at a resolution fine enough to capture the kerbs, walls and embankments that steer water. A model cannot represent a feature its grid never sampled. On constrained sites a topographical survey fills the gap.
• Hydrology. Design flows or rainfall for the return periods being tested. In the UK these usually come from Flood Estimation Handbook methods and ReFH2 rather than from the hydraulic package itself, with the current climate change allowance applied across the development's lifetime.
• Calibration. Observed flood records the model is tuned against, so its output carries evidential weight in front of a reviewer.

A sophisticated model fed poor data does not give a more reliable answer. It gives a more convincing-looking one. That is worse. The confidence it invites has not been earned.

1D, 2D and coupled 1D/2D flood modelling: what actually differs

The difference is how many directions water is allowed to move in. In a 1D model the watercourse is a line of surveyed cross-sections, and the solver works out water level and flow along that line. In a 2D model the ground is a grid or mesh of cells, and the solver works out depth and velocity in every one, so water spreads in any direction. A coupled 1D/2D model runs both at once and passes flow between them, solving the channel in 1D and the land around it in 2D.

What matters in practice is what each one cannot do. A 1D model cannot say where water goes once it leaves the bank, because it has no floodplain except as storage hung off a cross-section. A 2D model handles a culvert or a bridge soffit poorly, because those are sub-grid structures a mesh cannot resolve. Coupling removes both limits at once, which is why most detailed river work in the UK is now done that way, and why the phrase turns up in so many Environment Agency scoping responses.

Coupling is not free, and licence fees are the smallest part of the cost. The two domains exchange flow at a defined boundary, usually the bank line or the top of a structure. That exchange has to conserve mass without going unstable at the link. Get it wrong and the model still runs. The numbers still look plausible. They are wrong at exactly the location the assessment cares about.

A coupled model is not automatically more accurate than a 1D one. It is more capable, which is a different property, and being more capable gives it more places to fail.

Direct rainfall modelling sits slightly outside this scheme. Rather than introducing water at an inflow boundary, rainfall falls on every cell in the 2D domain and runs off. It is the standard technique for surface water assessment, where there is no single channel to route flow down. It is also far more sensitive to cell size and roughness than a channel model is. Mapping and interrogating those outputs is a GIS job as much as a modelling one.

How to tell which flood model a site needs

Start with the flood sources and the flow routes they create. Not with the software. Environment Agency guidance is explicit that dimensionality follows from knowing where water goes. Commit to a package before that, and the model often cannot answer the question it was built for.

1. Identify every source in play. A river, surface water, groundwater, sewers, the sea: each behaves differently, and some cannot be represented in one dimension at all.
2. Ask whether flow stays confined. Water that remains in the channel for the events being tested can be solved in 1D. Water that leaves the bank and spreads needs 2D.
3. Check the geometry against the flow. A long, narrow reach behaves one-dimensionally. A short, wide floodplain does not, and where reach length is only a few times the floodplain width, one dimension will not carry it.
4. Look for sub-grid structures. Bridges, culverts, weirs and pipes are represented properly in 1D and crudely in 2D, which argues for coupling, not for picking a side.
5. Test the data against the ambition. A 2D or coupled model needs terrain good enough to justify its resolution and inflows good enough to justify its runs. Without both, the extra dimension buys cost and false precision.
6. Build in acceptance from the start. The Environment Agency and the lead local flood authority judge the approach as well as the result. The justification for the dimensionality is part of what gets delivered.

Proportionality runs the other way too. Since the Environment Agency added banded surface water depth data to the Flood Map for Planning in May 2026, a number of routine sites no longer need a model of their own. Knowing when not to model is as much a professional judgement as knowing how, and it is one of the questions a flood risk assessment for planning settles early. In Wales the same judgement sits inside a flood consequence assessment instead.

The flood modelling software used on UK projects in 2026

Six packages cover most hydraulic modelling and flood analysis carried out on UK flood risk work, and they divide fairly cleanly by what each one was originally built to solve. HEC-RAS, from the US Army Corps of Engineers, is free and strongest on rivers. TUFLOW and MIKE+ are built around 2D. InfoWorks ICM is built around integrated catchments where pipes and surface water interact. Flood Modeller pairs a long-established 1D river solver with 2D. InfoDrainage handles site-scale drainage design, not flood mapping.

Version numbers matter here. More than they look as though they should. Environment Agency guidance asks that models be built and run in the latest versions of software, or that robust justification and testing be supplied where they are not. A model built two releases behind carries an explanatory burden its author may not have priced in.

HEC-RAS 7.0.1, released in June 2026, is the version for production work. The rebuilt HEC-RAS 2025 remains in public beta and its own documentation still carries a beta caveat.

Cost splits the field far more sharply than capability does. HEC-RAS is public domain, with no licence fee of any kind, which is why it is the usual starting point for a river-only study and for learning the discipline. EPA SWMM is likewise free and open source. TUFLOW, Flood Modeller, InfoWorks ICM and MIKE+ are commercial and licensed by module or edition, so what a project pays turns on how much of the software it actually uses.

Open-source and academic tools cover narrower ground than their licences suggest. LISFLOOD-FP, developed at the University of Bristol, is released under the GNU General Public Licence for non-commercial use, which excludes most of the consultancy work this article describes.

What happened to ISIS, MIKE 11 and MicroDrainage

Flood Modeller was originally called ISIS. The 1D river package was attributed jointly to Halcrow and Wallingford Software. It was renamed Flood Modeller at version 4.0, across 2014 and 2015, by which time Halcrow had been bought by CH2M. Jacobs acquired CH2M in 2017 and owns the product today. Several other familiar names have been retired the same way, which is why a model handed over five years ago can name software that no longer exists.

None of this is cosmetic housekeeping. DHI decommissioned MIKE 11, MIKE FLOOD and MIKE HYDRO River at Release 2024 and folded their functions into MIKE+; MIKE 21 survives, though its Classic Grid variant does not. InfoWorks RS conversion was supported in InfoWorks ICM up to version 2023.0 and withdrawn from 2023.1. A legacy RS model now needs a route through an older ICM release before it can be opened at all.

The change with the nearest deadline sits in drainage design, not flood mapping. Autodesk has set MicroDrainage's end of sale at 31 January 2027 and its end of life at 31 January 2030, naming InfoDrainage as the successor. Causeway Flow is a separate product from an unrelated vendor, and a migration destination some practices choose, not a successor in the lineage.

MicroDrainage has months left on sale. Any drainage model still living in it needs a migration plan, not a renewal.

Old names still turn up in model files, planning conditions and consultants' scopes. Recognising them is not nostalgia: a dated model cited in a flood risk assessment has to be either rebuilt or justified, and knowing which lineage it came from is how you work out which of the two you are facing.

Does the Environment Agency approve flood modelling software?

No, and the question comes up constantly. The Environment Agency publishes no list of approved or accepted flood modelling packages, and none of its guidance endorses one. What it requires is evidence that the software chosen suited the intended use, demonstrated against the standards its river modelling technical standards set out. Responsibility for that sits with the applicant. Unda's guide to what those standards ask of a flood risk assessment covers how they work in practice.

The standards were first published in August 2021 and last updated on 8 May 2026. They ask four things of any hydraulic model offered as planning evidence, set out in the model approach guidance.

• Approach justification. Why this dimensionality and this method, with evidence wherever an alternative approach was taken.
• Calibration. Tuning against observed events where records exist, typically to within 0.15 metres, and against at least three events where that is possible.
• Verification. Running a further observed event through the calibrated model without touching the coefficients, to show the tuning was not circular.
• Sensitivity testing. Varying the parameters the answer depends on, commonly by plus or minus 20 per cent on inflows, roughness, the downstream boundary and structure coefficients.

Reporting is a requirement in its own right, not a formality bolted on at the end. A method statement, a Model User Report, a development log and version control are all expected, and a model passport travels with the model. That is a document to plan for before modelling starts.

One document gets misread more often than any other in this area. The Environment Agency's benchmarking study of 2D hydraulic modelling packages was published in August 2013 and has not been updated since. It is a research report, not an approvals register, and it predates sub-grid sampling, quadtree meshes and every current GPU solver. Citing it as evidence that a package is accepted is an argument that will not survive review. Where the disagreement is with the mapping and not the method, challenging the flood map is a separate exercise with its own burden of proof.

What is changing in flood modelling in 2026

Three shifts are worth tracking this year. Solvers are moving onto graphics hardware and onto Linux. That changes what a model costs to run. It does not change what a model can represent. National mapping has improved to the point where it displaces bespoke modelling on straightforward sites. And machine learning has begun appearing inside mainstream packages. So far it assists; it does not solve.

• GPU and cloud. TUFLOW added native Linux support in its 2026 release, and the HEC-RAS 2025 beta carries a GPU solver. Both cut the cost of large runs. Neither improves the physics.
• Better national data. NaFRA2, released in December 2024, feeds better river, surface water and climate change data into national mapping. The May 2026 Flood Map for Planning update added banded surface water depths on top.
• AI inside the tools. Autodesk added an AI assistant to InfoWorks ICM 2027.1. It helps build and interrogate models. It does not solve them, and neither does the wider question of whether AI can write a flood risk assessment have the answer the marketing implies: what AI can and cannot do in flood risk work remains a narrower question than the marketing suggests.
• Consolidation continues. MicroDrainage, MIKE 11 and InfoWorks RS have all been retired or dated, and what remains is converging on fewer, broader platforms.

None of that changes the question a developer, lender or insurer or homeowner actually needs answered, which is whether this particular site floods, how deeply, and what can be done about it. The software is the instrument. Choosing it, and defending the answer it gives in front of a regulator, is the part that has not automated.

Unda builds 1D, 2D and coupled models in-house. The same models support planning submissions and due diligence on an acquisition. Where mapping alone will not settle the question, a flood modelling service scoped to the site is the route to a defensible answer. Other questions start elsewhere: flood zones and planning policy, or a surface water drainage strategy. Call 01293 214444 or email enquiries@unda.co.uk for a free quote within the hour.

Frequently asked questions