TUFLOW Benchmarking

From Tuflow
Jump to: navigation, search

Recently Uploaded

Welcome to the TUFLOW Benchmarking page of the TUFLOW Wiki. This page lists various documents that demonstrate the accuracy of TUFLOW, either via comparisons to real world flood event recordings, scale model flume datasets or analytical theory.

Recently added documents are listed in the thumbnails below.

Click the text link in the lower half of the image for a redirect to the document

University Thesis Studies

Boyte (2014)

This thesis investigated incorporating hydrology into direct rainfall models, with consideration given to hydraulic resistance mechanisms at shallow flow. The direct rainfall methodology was implemented into a two dimensional shallow water model, TUFLOW GPU; which was compared against an industry standard hydrologic model, XP RAFTS.

The primary objectives were to determine whether TUFLOW GPU was a suitable software package to use in industry applications, whether the direct rainfall model was able to reproduce the hydrology of a real storm event in a gauged catchment more accurately than the hydrologic model; and to understand hydraulic resistance mechanisms at shallow flow and at different roughness scales. These objectives were met through numerical modelling with real data produced from experiments, stream gauges, or analytical solutions. Dressler’s sloping dam break analytical model was used to validate TUFLOW GPU, a gauged catchment in New South Wales was used to compare hydrology representation in the direct rainfall model and hydrologic model, and experimental data from an open channel at shallow flow was analysed to analyse hydraulic resistance mechanisms. Monte Carlo testing by simulating non uniformity in bed roughness was undertaken on an ungauged catchment in New South Wales to determine the practical impacts of secondary flows, which arose after analysis of the experimental data.

Boyte 2014 Dressler.PNG Boyte 2014 Catchment.PNG

Leister (2010)

The author undertook research to ascertain the accuracy of TUFLOW in calculating the energy losses associated with the contraction and expansion of flow through a constriction and to ascertain the most appropriate method/s for reliably modelling the energy losses associated with bridge piers. To undertake the research 2D model results were compared to physical flume test undertaken by Liu et al (1957).

The research involved the development of a series of flumes within TUFLOW that were used to simulate a number of scenarios that were modelled in a physical flume by Liu et al (1957). These scenarios included constriction widths varying between 2 and 6 feet, as well as a number of pier combinations involving square shaft, single shaft, double shaft and round-ended narrow pier types. The TUFLOW flumes were of varying grid sizes to test the model’s ability to replicate the physical models results at varying grid resolutions. The afflux predicted by each of these scenarios within TUFLOW was compared to the results obtained from the physical flume tests.

The results from the analysis undertaken have shown that TUFLOW can, within reasonable bounds, reproduce the results of the physical model. Recommendations regarding the modeling of constrictions and piers within a 2D hydraulic model are made.

Caddis (2010)

The methods used to estimate soils infiltration in TUFLOW have been based on the findings of this thesis. The research undertaken looks at the influence of losses on 2D hydrodynamic flood modelling applications. The main objectives were to:

  • Investigate depression storage inherent in digital topographic data and its influence on modelled losses;
  • Assess the influence that application of loss models to the ground surface rather than the rainfall hyetograph can have on surface runoff;
  • Assess the influence that varying soil types have on surface runoff; and
  • Assess the influence that ponding depth has on infiltration, and the subsequent affect on surface runoff.
  • Click Here To Open Thesis (Caddis, 2010)

Huxley (2004)

This thesis validates TUFLOW against independent analytical calculations. The study used over 300 benchmark models to verify the accuracy of TUFLOW for a range of flow conditions (super critical, critical and subcritical). The specific test cases included:

  • 1D culvert flow;
  • 1D weir flow;
  • 2D weir flow;
  • 2D channel flow;
  • 2D floodplain flow; and
  • 2D channel/floodplain flow.

The TUFLOW results were found to be within a 2% accuracy of the analytical estimates in 97% of the benchmark models.

Huxley 2004 Variance.PNG

Barton (2001)

This thesis study investigated the ability of 2D hydrodynamic models to adequately predict energy losses through an abrupt constriction. In particular, the investigation focuses on the impact that model spatial resolution has on the ability of the model to predict expansion and contraction losses due to the abrupt constriction. Principal outcomes of the study were:

  • An improved understanding of different numerical solution schemes;
  • An improved understanding of the nature of contracting and expanding flow;
  • The confirmation that the spatial resolution of 2D models does have an impact on the ability of these models to predict energy losses due to turbulent effects;
  • An understanding of the importance of the eddy viscosity formulation technique on the predictive ability of 2D models; and
  • A preliminary assessment of the impact of varying the eddy viscosity formulation technique.

Barton 2009.PNG

Syme (1991)

TUFLOW was first developed as a result of this Masters study. This thesis summarises the mathematical theory that underpins TUFLOW. It presents a discussion on the process of selecting TUFLOW's 1D and 2D schemes. It also details:

  • The methodology used for coding the 2D scheme;
  • The wetting and drying method; the dynamic 2D/1D link; and
  • The stabilisation of oblique water level boundaries.
  • Click Here To Open Thesis (Syme, 1991)

Other Benchmark Studies

TUFLOW has been benchmarked against other software, physical models and analytical equations on many occasions:

  1. Daniel Sheehy, Dr Sharmil Markar, Dr David Newton from WRM Water & Environment Pty Ltd assessed the capability of the TUFLOW HPC to accurately simulate an idealised dam break scenario by comparing the model results to analytical solutions. The results were also compared the model results for coastal inundation by a tsunami to real-world data from the 2004 Banda Ache (Indonesia) tsunami. The findings demonstrate that TUFLOW HPC correctly captures the dam break flood fronts and the flood wave propagation. As such TUFLOW HPC is well suited for dam break flood modelling:
  2. The Environment Agency in the United Kingdom have documented independent testing of most 2D modelling packages. Testing was undertaken in both 2010 and 2013 and all three TUFLOW engines were submitted for testing in 2013. This is a good resource for comparing TUFLOW to other available software. The tests have subsequently been rerun and documented using the latest version of TUFLOW in 2017:
    The above benchmark documents have been superseded by the document below. It is a revision of the benchmark modelling using the 2017-09-AC version of TUFLOW. It includes results for TUFLOW Classic, FV, GPU and the new TUFLOW HPC solver.
  3. Chris Huxley compared the speed of TUFLOW Classic against the TUFLOW HPC for high resolution integrated urban drainage applications. The modelling included the underground stormwater network and 1D elements and 2D for the representation of above ground flooding. Models ranging in size from 7,500 2D cells to 12,500,000 2D cells were assessed.
  4. Wood Rodgers compared TUFLOW results against recorded flume data and standard engineering equations. The assessment results were presented at the 2015 Floodplain Management Association Conference in Rancho Mirage, California. The TUFLOW results are summarized in the following link: Chris Huxley built on the dataset for the 2015 TUFLOW User Conference
  5. Bill Syme presented model validation/benchmark results for bends, structures and obstructions at the 2011 Australian TUFLOW Workshop and 2011 Association of State Floodplain Managers (ASFPM) Conference
  6. Bill Syme presented model validation results for a case study in the Eudlo Creek Catchment, Australia at the 2006 Association of State Floodplain Managers (ASFPM) Conference. TUFLOW was benchmarked against a physical model and FESWMS.
  7. Additional TUFLOW benchmark articles are available from the "Library" section of the TUFLOW website:

Real World Calibration/Validation

TUFLOW has been used on thousands of flood studies worldwide. Model calibration to historic events is a critical element of any flood study. Here are some links to public domain documents which have included TUFLOW model calibration/validation to historic flood events.

Hardware Benchmark Testing

The link below summarises computer specifications and model simulation times for a benchmark model that is run using TUFLOW Classic (CPU) and TUFLOW GPU. The results are a useful resource for people who are upgrading hardware for the specific purpose of modelling.