Benchmark with known end states2D · free surfaceCandidate

Filling floodplain hollows

Does every drop of inflow end up stored in the hollows?

Test A2C-EA2
Néelz (2013)
Version 1.0.0

Right answer: Known end states · Néelz (2013)
Results: 10 models · closest under 0.01 % off · see them ↓
Must: store exactly the water let in (within 0.1 %)

Sketch, not to scale.

Terrain
water depth at the end of the run (demo solver) for EA Test 2 — filling of floodplain depressions, with output points and the inflow line in red

Cumulative inflow (exact)
Stored volume, demo solver

In a closed domain the two must coincide at every instant; the gating check is the final value.

Gaps are times when the cell at that point is dry.

What happens

A short flood spreads over a floodplain dotted with 16 shallow hollows and slowly fills them.

Why it matters

Slow, shallow filling is where models with poor wetting and drying lose water.

A good model must

store exactly the water let in (within 0.1 %)

Exact answer and models

Water level at gauge P1 (250 m, 250 m)

There is no exact water level here; the test is scored on stored volume. Levels are shown so the models can be compared with each other.

Point at the chart to read every model's value; click a name to hide or show it.

Water level at gauge P8 (750 m, 1750 m)

There is no exact water level here; the test is scored on stored volume. Levels are shown so the models can be compared with each other.

Point at the chart to read every model's value; click a name to hide or show it.

How models did

In rank order: by each model's percent difference from the right answer, the largest over the test's accuracy checks (lower is better; results within 10 % of each other share a rank). Mass error, timing and NSE are shown beside it, not folded in. Each model runs the same inputs. Compare them in the results viewer (depth maps and animation) →

Rank	Model	Difference	Largest difference in	Run time
1=	ANUGA 4.0.1	<0.01 %	Water conserved	–
1=	LISFLOOD-FP 8.3 · FV1	<0.01 %	Water conserved	1.1 min
1=	LISFLOOD-FP 8.3 · ACC	<0.01 %	Water conserved	32.2 s
1=	LISFLOOD-FP 8.3 · DG2	<0.01 %	Water conserved	–
1=	Demo 2D suite example	<0.01 %	Water conserved	47.1 s
1=	TELEMAC-2D openTELEMAC 9.1	<0.01 %	Water conserved	–
7=	HEC-RAS 6.6 previous release · 2D SWE-EM	0.72 %	Water conserved	3.8 min
7=	HEC-RAS 7.0 2D · SWE-EM	0.73 %	Water conserved	4.0 min
7=	HEC-RAS 6.6 previous release	0.77 %	Water conserved	3.9 min
7=	HEC-RAS 7.0 1D · 2D SWE-ELM	0.79 %	Water conserved	3.3 min

ANUGA

ANUGA 4.0.1 (Geoscience Australia / ANU), default settings on a triangular mesh

#1= under 0.01 % off

Settings: anuga (rectangular_cross mesh, defaults, sampled at cell centres)

Water conserved

• < 0.0001 % off

Scored with method 3.0 · Score card: ANUGA 4.0.1

Run it yourself

ANUGA 4.0.1 · runs on Linux · produced by scripts/run_anuga.py in open-solvers.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
conda create -n anuga -c conda-forge anuga=4.0.1 python=3.12 pip scipy
conda activate anuga
python -m pip install -e ".[dev]"
python scripts/run_anuga.py A2C-EA2

The model files for this run will be published with its next run.

ANUGA builds its mesh and boundaries in Python: the script is the model.

The test itself (inputs, exact answer, checks): A2C-EA2.zip. Check the score card with benchmark1d2d verify-card: its signature by the suite's public development key shows the card has not changed since it was written, not who wrote it.

LISFLOOD-FP FV1

LISFLOOD-FP 8.3, first-order finite-volume solver (full shallow-water equations)

#1= under 0.01 % off

Run time, finest grid: 1.1 min · AMD EPYC 7763 64-Core Processor, 2 cores, 15.6 GB, Linux 6.17.0-1022-azure (GitHub Actions (github-hosted, ubuntu24))

Settings: lisflood (fv1 solver, defaults; CPU, double precision)

Water conserved

• < 0.0001 % off

Scored with method 3.0 · Score card: LISFLOOD-FP 8.3 FV1

Run it yourself

LISFLOOD-FP 8.3 FV1 · runs on Linux · produced by scripts/run_lisflood.py in open-solvers.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
# LISFLOOD-FP 8.3 source (Zenodo, MD5 abaa2e41b58f6c7df4443274bf130d4a), built CPU-only in double precision:
sudo apt-get install -y libnetcdf-dev libnuma-dev
cmake -S <lisflood>/source -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build --target lisflood
python -m pip install -e ".[dev]"
python scripts/run_lisflood.py --exe build/lisflood --solver fv1 A2C-EA2

The model files the suite wrote for this run (.zip, 32 KB): the exact inputs the solver read (one folder per grid, where the test has several), with their SHA-256.

LISFLOOD-FP ACC

LISFLOOD-FP 8.3, the local-inertia solver it is best known for; it leaves out the advection term

#1= under 0.01 % off

Run time, finest grid: 32.2 s · AMD EPYC 7763 64-Core Processor, 2 cores, 15.6 GB, Linux 6.17.0-1022-azure (GitHub Actions (github-hosted, ubuntu24))

Settings: lisflood (acceleration solver, defaults; CPU, double precision)

Water conserved

• < 0.0001 % off

Scored with method 3.0 · Score card: LISFLOOD-FP 8.3 ACC

Run it yourself

LISFLOOD-FP 8.3 ACC · runs on Linux · produced by scripts/run_lisflood.py in open-solvers.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
# LISFLOOD-FP 8.3 source (Zenodo, MD5 abaa2e41b58f6c7df4443274bf130d4a), built CPU-only in double precision:
sudo apt-get install -y libnetcdf-dev libnuma-dev
cmake -S <lisflood>/source -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build --target lisflood
python -m pip install -e ".[dev]"
python scripts/run_lisflood.py --exe build/lisflood --solver acc A2C-EA2

The model files the suite wrote for this run (.zip, 32 KB): the exact inputs the solver read (one folder per grid, where the test has several), with their SHA-256.

LISFLOOD-FP DG2

LISFLOOD-FP 8.3, second-order discontinuous Galerkin solver

#1= under 0.01 % off

Settings: lisflood (dg2 solver, defaults; CPU, double precision)

Water conserved

• < 0.0001 % off

Scored with method 3.0 · Score card: LISFLOOD-FP 8.3 DG2

Run it yourself

LISFLOOD-FP 8.3 DG2 · runs on Linux · produced by scripts/run_lisflood.py in open-solvers.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
# LISFLOOD-FP 8.3 source (Zenodo, MD5 abaa2e41b58f6c7df4443274bf130d4a), built CPU-only in double precision:
sudo apt-get install -y libnetcdf-dev libnuma-dev
cmake -S <lisflood>/source -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build --target lisflood
python -m pip install -e ".[dev]"
python scripts/run_lisflood.py --exe build/lisflood --solver dg2 A2C-EA2

The model files the suite wrote for this run (.zip, 33 KB): the exact inputs the solver read (one folder per grid, where the test has several), with their SHA-256.

Demo 2D

Worked example that ships with the suite: first-order finite volume

#1= under 0.01 % off

Run time, finest grid: 47.1 s · AMD EPYC 9V74 80-Core Processor, 2 cores, 15.6 GB, Linux 6.17.0-1022-azure (GitHub Actions (github-hosted, ubuntu24))

Water conserved

• < 0.0001 % off

Scored with method 3.0 · Score card: demo-swe2d (first-order FV, HLL + hydrostatic reconstruction)

Run it yourself

demo-swe2d (first-order FV, HLL + hydrostatic reconstruction) · version 0.1.0 · runs on any · produced by scripts/run_demo_2d.py in track-a.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
python -m pip install -e ".[dev]"
python scripts/run_demo_2d.py

The model files for this run will be published with its next run.

TELEMAC-2D

openTELEMAC-2D 9.1.1, run with its default finite-volume settings

#1= under 0.01 % off

Settings: telemac2d (Selafin, nodes at the grid's cell centres)

Water conserved

• 0.0086 % off

Run with a 1 s largest step, as for the other codes: with TELEMAC's step left to the 5-minute output interval, the dry start took long first steps that undersampled the inflow and lost 6 % of the water.

Scored with method 3.0 · Score card: openTELEMAC-2D v9.1.1

Run it yourself

openTELEMAC-2D v9.1.1 · runs on Linux · produced by scripts/run_telemac.py in track-a.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
python -m pip install -e ".[dev]"
conda create -n telemac -c conda-forge opentelemac=v9.1.1 python=3.12
python scripts/run_telemac.py --telemac-prefix "$(conda info --base)/envs/telemac" A2C-EA2

The model files the suite wrote for this run (.zip, 191 KB): the exact inputs the solver read (one folder per grid, where the test has several), with their SHA-256.

HEC-RAS 6.6

HEC-RAS 6.6 2D with its Eulerian shallow-water option (SWE-EM), the release before 7.0

#7= 0.72 % off

Run time, finest grid: 3.8 min · AMD EPYC 7763 64-Core Processor, 2 cores, 16 GB, Windows 2025Server (GitHub Actions (github-hosted, win25-vs2026))

Settings: hecras 2D (headless terrain + seeded mesh + boundary lines, SWE-EM; depth = cell volume / area)

Water conserved

• 0.72 % off

As SWE-ELM: HEC-RAS's own accounting shows a 1.4 % volume error.

Scored with method 3.0 · Score card: HEC-RAS 6.6 2D SWE-EM

Run it yourself

HEC-RAS 6.6 2D SWE-EM · runs on Windows · produced by scripts/run_hecras_2d_cases.py in hecras.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
# Windows: HEC-RAS 6.6 from HEC (HEC-RAS_66_Setup.exe, SHA-256 42A370B1…4715EB), silent install /S /v/qn
python -m pip install -e ".[dev]" "ras-commander==0.105.0" h5py psutil pywin32 pillow pythonnet tifffile
set HECRAS_TAG=66
python scripts/run_hecras_2d_cases.py A2C-EA2 --accept-hec-terms --equations em

The model files the suite wrote for this run (.zip, 5,219 KB): the exact inputs the solver read (one folder per grid, where the test has several), with their SHA-256.

HEC-RAS 7.0

HEC-RAS 7.0.1 2D with its Eulerian shallow-water option (SWE-EM), which HEC recommends for dam breaks

#7= 0.73 % off

Run time, finest grid: 4.0 min · AMD EPYC 7763 64-Core Processor, 2 cores, 16 GB, Windows 2025Server (GitHub Actions (github-hosted, win25-vs2026))

Settings: hecras 2D (headless terrain + seeded mesh + boundary lines, SWE-EM; depth = cell volume / area)

Water conserved

• 0.73 % off

Scored with method 3.0 · Score card: HEC-RAS 7.0.1 2D SWE-EM

Run it yourself

HEC-RAS 7.0.1 2D SWE-EM · runs on Windows · produced by scripts/run_hecras_2d_cases.py in hecras.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
# Windows: HEC-RAS 7.0.1 from HEC (HEC-RAS_701_Setup.exe, SHA-256 1FE76297…3CEE73), silent install /S /v/qn
python -m pip install -e ".[dev]" "ras-commander==0.106.1" h5py psutil pywin32 pillow pythonnet tifffile
python scripts/run_hecras_2d_cases.py A2C-EA2 --accept-hec-terms --equations em

The model files the suite wrote for this run (.zip, 5,671 KB): the exact inputs the solver read (one folder per grid, where the test has several), with their SHA-256.

HEC-RAS 6.6

HEC-RAS 6.6, the release before 7.0, run the same way (1D, and 2D with its default SWE-ELM): kept so a change between versions shows

#7= 0.77 % off

Run time, finest grid: 3.9 min · AMD EPYC 7763 64-Core Processor, 2 cores, 16 GB, Windows 2025Server (GitHub Actions (github-hosted, win25-vs2026))

Settings: hecras 2D (headless terrain + seeded mesh + boundary lines, SWE-ELM; depth = cell volume / area)

Water conserved

• 0.77 % off

HEC-RAS's own volume accounting reports 1.4 % more water stored than flowed in (98,576 m³ from 97,194 m³). Integrating its water levels over the 2 m DEM gives +0.76 %. Run on the EA's own 2 m DEM as sub-grid terrain.

Scored with method 3.0 · Score card: HEC-RAS 6.6 2D SWE-ELM

Run it yourself

HEC-RAS 6.6 2D SWE-ELM · runs on Windows · produced by scripts/run_hecras_2d_cases.py in hecras.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
# Windows: HEC-RAS 6.6 from HEC (HEC-RAS_66_Setup.exe, SHA-256 42A370B1…4715EB), silent install /S /v/qn
python -m pip install -e ".[dev]" "ras-commander==0.105.0" h5py psutil pywin32 pillow pythonnet tifffile
set HECRAS_TAG=66
python scripts/run_hecras_2d_cases.py A2C-EA2 --accept-hec-terms --equations elm

The model files the suite wrote for this run (.zip, 5,219 KB): the exact inputs the solver read (one folder per grid, where the test has several), with their SHA-256.

HEC-RAS 7.0

HEC-RAS 7.0.1 (US Army Corps of Engineers), built and run headless with its default settings; in 2D, its default full shallow-water option (SWE-ELM)

#7= 0.79 % off

Run time, finest grid: 3.3 min · AMD EPYC 9V74 80-Core Processor, 2 cores, 16 GB, Windows 2025Server (GitHub Actions (github-hosted, win25-vs2026))

Settings: hecras 2D (headless terrain + seeded mesh + boundary lines, SWE-ELM; depth = cell volume / area)

Water conserved

• 0.79 % off

Scored with method 3.0 · Score card: HEC-RAS 7.0.1 2D SWE-ELM

Run it yourself

HEC-RAS 7.0.1 2D SWE-ELM · runs on Windows · produced by scripts/run_hecras_2d_cases.py in hecras.yml.

git clone https://github.com/scott-jeffers/benchmark1d2d && cd benchmark1d2d
git checkout 71dc10fc07a9
# Windows: HEC-RAS 7.0.1 from HEC (HEC-RAS_701_Setup.exe, SHA-256 1FE76297…3CEE73), silent install /S /v/qn
python -m pip install -e ".[dev]" "ras-commander==0.106.1" h5py psutil pywin32 pillow pythonnet tifffile
python scripts/run_hecras_2d_cases.py A2C-EA2 --accept-hec-terms --equations elm

The model files the suite wrote for this run (.zip, 5,671 KB): the exact inputs the solver read (one folder per grid, where the test has several), with their SHA-256.

Not run yet: GeoClaw, TUFLOW, MIKE 21, Iber, BASEMENT, Delft3D FM

Run it and add yours →

Status of this test

Candidate. To become frozen it needs: maintainer review recorded.

Download inputs (.zip)Cite this test View on GitHub

Technical details for reviewers Checks, reference, inputs, sources

Checks and why each limit was chosen

Check	Difference measured as a percentage of	Latest	Difference	Why this check
mass_conservation expected_state_error of volume · in the score	already a percentage	–		The only exact statement the test admits: nothing leaves a closed domain, so any difference is mass created or lost by the scheme (suite target < 0.1 %).

Reference

No exact or measured answer exists. The test is scored on states that follow from the inputs alone; published package results are context only.

Expected state	Value	Why it must hold
final_stored_volume volume	97,200 m3	Closed boundaries and a dry initial bed: stored volume at the end of the run equals the linearly interpolated inflow hydrograph integrated to t_end.

More figures

Cumulative inflow (exact)
Stored volume, demo solver

In a closed domain the two must coincide at every instant; the gating check is the final value.

Gaps are times when the cell at that point is dry.

Promotion checklist

Only frozen tests have fixed inputs and answers. This list is computed from the manifest and the published score cards; the rules are in the method.

✓Data licence verifiedfor candidateOGL-UK-3.0
✓All inputs availablefor candidateall committed or fetched by SHA-256
✓Reference (truth) availablefor candidateobjective expected states
✓Every tolerance justifiedfor candidateeach gating criterion has a rationale
✓A published run passesfor frozen6 passing score card(s)
✓Order of accuracy shownfor frozennot a verification case
✓Two independent codes passfor frozenpassing codes: ANUGA 4.0.1, LISFLOOD-FP 8.3 ACC, LISFLOOD-FP 8.3 DG2, LISFLOOD-FP 8.3 FV1, demo-swe2d, openTELEMAC-2D v9.1.1
✗Maintainer review recordedfor frozenno maintainer review recorded

Inputs

File	Role	Format	SHA-256	Get it
`inputs/terrain.asc`	terrain	esri_ascii_grid	`92d85335a20c…`	Fetched by hash from source
`inputs/active_area.shp`	active_area	shp	`427f2f26033a…`	Download
`inputs/inflow.csv`	boundary_condition	csv	`533033b11da4…`	Download
`inputs/output_points.csv`	probes	csv	`8ead735333d9…`	Download
`inputs/case.json`	case_parameters	json	`f8c74353d715…`	Download

About this test

Parameter	Value
Manning's n	0.03
Grid	20 m
Run time	48 h
Inflow volume to t_end	97,200 m³

Probe levels and velocities are compared with the published package spread for context only.

Data and licence

Inputs come from the GOV.UK 'Benchmarking model data' archive, published under the Open Government Licence v3.0 (the page licence; the archive carries no other terms). Attribution: Environment Agency, SC120002 (2013), test specifications May 2010.

Archive: https://assets.publishing.service.gov.uk/media/6033a9bbd3bf7f72182e99ca/Benchmarking_Model_Data.zip (SHA-256 388c2789570e5975fe909033e34bedbe5d7bde5e6d7f9b7d7c1eaf04e58bc732)
Landing page: https://www.gov.uk/flood-and-coastal-erosion-risk-management-research-reports/2d-benchmarking-evaluating-the-latest-generation-of-the-hydraulic-models-for-fcrm
Report: https://assets.publishing.service.gov.uk/media/6033a943d3bf7f721f4b0d49/_SC120002_Benchmarking_2D_hydraulic_models_Report.pdf

Fetch the non-committed inputs with python scripts/fetch_inputs.py A2C-EA2.

Sources

Néelz, S., Pender, G. (2013). Benchmarking the Latest Generation of 2D Hydraulic Modelling Packages. SC120002. www.gov.uk/flood-and-coastal-erosion-risk-management-researc
Néelz, S., Pender, G. (2010). Benchmarking of 2D Hydraulic Modelling Packages. SC080035/SR2. assets.publishing.service.gov.uk/media/5a75765c40f0b6360e474

Test A2C-EA2_floodplain_depressions · version 1.0.0 · data licence OGL-UK-3.0