Using a real-world case study, Jacobs have benchmarked Flood Modeller’s 1D urban and 2D ADI solvers (linked) against two other widely available software products.
Our benchmarking tests have shown that Flood Modeller:
Was up to 49% faster.
Results were not sensitive to the timestep selection.
Results were in close comparison and reflect, to a high level of acceptability, the flood
mechanisms at study location.
Was numerically stable throughout the simulation.
The differences in flood level results, between software products, is within currently accepted modelling tolerances (Environment Agency Minimum Technical Requirements, and CIWEM’s Urban Drainage Group code of practice for the hydraulic modelling of urban drainage systems 2017).
Calibration data is not available for the case study. Through model calibration and/or adjustments to run parameters, for each software product, it may be possible to produce results that are closer in comparison
Dataset
The dataset is for a small heavily urbanised catchment in England. The upper and middle
parts of the catchment are steep and culverted. The lower part of the catchment is culverted
and includes a short section of open channel. The catchment discharges, through an outfall,
into a watercourse.
The same input data was made available for constructing the model within each software
product:
A model schematic and node naming convention is illustrated in Exbibit A.
To enable accurate benchmarking of results the hydraulic model schematisation and the
model components, available within each software product, were selected to ensure a close
match between the three software products.
Default/recommended model parameters have been used for each software product.
Each benchmark test was run on the same computer and only used the CPU solvers that are
available within each software product. The computer used was an Intel® Xeon® CPU E5-
1650 v3 @ 3.50GHz (RAM 32.0 GB).
Benchmarking results
The results of each of the software simulations have been extracted, analysed, and compared
to highlight similarities and differences in model results and numerical performance between
the software products.
Simulation times and effect of timestep selection
When setting the model timestep at 1s Flood Modeller completed the simulation in 266s.
49% faster than the next fastest software product.
A comparison of the percentage difference in maximum flood extent between the 0.5s and
0.2s showed that for Flood Modeller the difference in maximum flood extent was 0.2%.
Flood extent and level
The percentage of flood extent overlap between Flood Modeller and the other software
products is up to 79.2%, this was higher than the overlap between the other two software
products. The flood extent produced by Flood Modeller is shown in exhibit B.
