You are on page 1of 9

2D Hydrology:- Using the Shallow water wave equation to model hydrologic response

Mr. Rudy Van Drie Engineer, Burrapine, NSW, Australia E-mail: rudy@balancernd.com.au Dr. Petar Milevski Senior Flood modeller, Wollongong City Council, Australia E-mail: PMilevski@wollongong.nsw.gov.au Mr. Michael Simon Director Hydro-Sim, Essen Germany E-mail: simon.m@hydro-sim.de Abstract: The shallow water wave (SWW) equation is well known for its applications in shallow water hydraulics such as resolving flow behaviour in river and floodplain systems. In addition various solution approaches exist that may utilise finite difference, finite element, finite volume and volume of fluid approaches. However up until recently, the ability of the SWW equation to properly resolve catchment hydrology has been questioned. It is known that some approaches are more unstable than others. This paper puts forward findings using a Finite Volume approach that has shown itself to remarkably stable and accurate. Comparison with a well known and accepted hydrologic model provide for very similar results. In fact it is likely that the SWW equation approach may be more realistic than the lumped hydrologic modelling approach. A number of catchments will be shown over a range of size and complexity all showing very promising results.

1. INTRODUCTION
The need to develop methods to assess hazard due to flooding is well known and documented. The January 2011 flooding of south east Queensland and the subsequent Queensland Flood Commission have shed considerable light on the fact that current approaches to flood mapping in Australia have been less than adequate in terms of both accuracy and time taken to produce results. Part of the problem relates to the excessive time taken to prepare formal flood studies. Often years are taken and $100,000s per catchment to provide flood maps that continue to present relatively low resolution results. At times in some places the results are simply incorrect possibly missing vital diversions, or other effects. The current approaches generally use fixed grid model methods with resolutions of 10x10m or 5x5m of only a portion of the catchment. In order to account for hazard in highly urban areas the cell size needs to be considerably lower. Some models have been produced at 2x2m or 1x1m of only a very small portion of the entire catchment, as either a nested or linked sub model. However this approach is not capable of fully representing correctly all of the dynamic behaviour up and down the catchment. Further some 2D flow models struggle to replicate hydrologic response correctly (Clark, Ball, Babister, 2008). This type of modelling approach is also mostly, reliant on hydrologic models to provide the details of flow (quantity) into the 2D hydraulic models, as such there are many opportunities for misrepresentation of flow behaviour to creep into this modelling approach. Of the available methods of solving the 2D flow equations (Shallow Water Wave Equations) researchers have generally found the Finite Volume approach the most robust compared to Finite Difference or Finite Element, particularly as related to both, shock capturing and wetting and drying. (Jakeman 2006)

2D Hydrology:- Using the SWW equation to model hydrologic response In order to develop these hydraulic flow solving models into hydrologic models they essentially need to represent the following essential aspects: Include the response of the entire catchment (dynamically) , with Rainfall over the catchment (Spatially variable) Surface Roughness (Spatially variable) Boundary conditions (Tide etc. Temporally varying as required) Optionally infiltration, although this can be applied via excess rainfall concepts

VanDrie

In order to provide adequate resolution of flow behaviour the cell size also needs to be variable. This is particularly true in urbanised catchments, or to include specific detail such as bridge piers. However only finite volume approaches generally use unstructured variable sized grids. Hence in order to model entire catchments providing robust hydrologic response, inclusive of detailed hydraulic behaviour, plausibly between individual buildings, it would appear a candidate is an approach using an unstructured grid finite volume shallow water wave equation solver, with a rainfall forcing function. This paper identifies such a model and provides evidence of robust performance over a wide range of scales and applications. Further the code described is Free and Open Source Software (FOSS) and hence is capable of relatively rapid development. As interest in the code increases it is likely that the number of developers will increase, resulting in further code development.

2. ROLE OF THE HYDROLOGIC MODEL


Hydrologic models are currently still used in the majority of applications to provide input into 2D hydraulic models. This two step process is fraught with danger as it becomes very unclear as to sources of errors. Errors are often overcome by manipulation of parameters such as roughness in the hydraulic model, even though the problem may stem from poor understanding of the hydrologic model, leading to erroneous determined flow predictions. This issue was raised with an example by researchers (Van Drie, R., Milevski, P., Simon, M., 2011) that showed poor use of a hydrologic model leading to flows being used in the 2D model that were underestimated by 44%. The resulting flood level and behaviour obviously affected. Hydrologic models will remain to serve this role and the role of a sanity check for 2D hydraulic models, to ensure that as the use of 2D models increases in hydrologic applications, they are replication hydrologic response correctly.

3. ANUGA AS A HYDROLOGIC MODEL


ANUGA is a hydrodynamic modelling tool that allows users to model realistic flow problems in complex two-dimensional geometries, such as dam breaks, flooding, storm surges and tsunami. Most ANUGA components are written in the object-oriented programming language Python and most users will interact with ANUGA by writing small Python programs based on the ANUGA library functions. Computationally intensive components are written for efficiency in C routines working directly with Python numpy structures. The user specifies a study area represented by an irregular triangular mesh and comprises the topography and bathymetry, manning's roughness, boundary conditions, rainfall and/or stream flows, wind stress or pressure gradients. An example of application of rainfall to a catchment polygon is shown in figure 1. Note the refined mesh in the same model (figure 2.)

HWRS 2012 Van Drie, Milevski Simon

2 of 9

2D Hydrology:- Using the SWW equation to model hydrologic response

VanDrie

Figure 1. Catchment Polygon in Domain

Figure 2. Refine mesh over area of interest.

The rainfall is applied to the cells covered by the polygon (or polygons) and the shallow water wave (SWW) equation solver resolves all aspects of routing. Water flows over the terrain hillsides toward the streams where flow concentration is also modeled with the same SWW solution. ANUGA tracks the evolution of water depth and horizontal momentum within each triangle over time by solving the shallow water wave equation governing equation using a finite-volume method (Figure 3.). Stage and momentum time series can also be plotted directly via the viewer (Figure 4.).

Figure 3. Momentum on fine grid

Figure 4. Time series of momentum from viewer

It should also be noted that the results file (extension .SWW) is also compatible with tools such as waterRIDE (Worley Parsons) and Mirone (J.F. Luis 2007). Further, built-in scripts allow for easy export of any results to any GIS platform, or the creation of section-profile plots and time series plots of conserved and derived quantities. Further details of built in functions and detailed background can be found in the ANUGA User Manual.

4. APPLICATION / VALIDATION
In order to validate the use of ANUGA as a sound hydrologic model, it is necessary to apply and compare it to known results for a variety of applications over a range of scales. ANUGA has been validated for a range of hydraulic applications including tsunami (Nielsen, et al 2005), wave tank experiments (Barnes, Baldock, 2007) and predicting possible impacts of climate

HWRS 2012 Van Drie, Milevski Simon

3 of 9

2D Hydrology:- Using the SWW equation to model hydrologic response change (Baldock et al 2007). It has also been trialed on small (12.5 hectare) to medium catchments (VanDrie, Milevski 2009) where not only did it adequately reproduce hydrographs compared to a well validated hydrologic model, it also identified a valid range of Mannings roughness beyond which users should question its appropriateness. Questions were raised regarding the use of high Roughness values in other models, as there appears to be no valid descriptions of Mannings N values of 10 or 20? Further validation was undertaken on a 720 hectare catchment by VanDrie, Milevski Simon, (2011) (Figure 6.). This work was recently revised (2012) and confirmed that the rapid rise of flood water immediately downstream of a railway culvert could only have occurred due to a major blockage and effective dam Break of the debris raft. This validation work is still ongoing as a dam break facility is added to the ANUGA model. Figure 5 shows the current comparisons. Note the very good validation against the recorded level generally, and the near matching peak accounting for the dam break.

VanDrie

Figure 5. Validation Stage Plot

Figure 6. 720 hectare Catchment in the viewer

It should be noted that after the 1998 flood event, statements in property owners insurance claims for the properties immediately down stream identify an extremely rapid rise in flood water that is not explained by the event flood model without a debris raft dam break.

5. VALIDATING LARGER CATCHMENTS


Considering that ANUGA was able to replicate hydrograph shape and peaks well for 12 ha and 720 ha catchments, it was considered of value to investigate the range of scale to which this may apply. Considering the availability of a well validated larger model, the 87km long river reach down stream of the St. Francis Dam (Begnudelli, Sanders, 2007) was used as an initial trial as there good validation data. The extreme dynamic behaviour, such as valley sloshing, hydraulic jumps and multiple shocks below the dam provide a good test for ANUGAs ability to model such dynamic behaviour. Although the area modeled is around 350,000 hectares, being a dam break model no rainfall was used (Figure 7.). The aim initially was to ensure valid reproduction of results along such a long river reach and to capture the extreme behaviour. The previous 2007 work used the BREZO model. In order to fully investigate the extreme sloshing a second refined ANUGA model was also set up (Figure 8.).

HWRS 2012 Van Drie, Milevski Simon

4 of 9

2D Hydrology:- Using the SWW equation to model hydrologic response

VanDrie

Figure 7. St Francis Dam Break Model (350,000ha)

Figure 8. Detail of Sloshing below dam

In comparing the results to the previous work by Sanders (2007) both the timing of the arrival of peaks and the estimates of peak flows are all very similar. Indeed the hydrograph shapes (Figure 9.) also appear quite similar except for some oscillation in the Brezo results which are more than likely induced by instability in the model. ANUGA shows lower oscillation which is more than likely sloshing rather than instability.

St Francis Dam Break: Estimated Time of Arrival of Peaks Comparison XS A-A B-B C-C D-D E-E F-F G-G Time (sec) Brezo Front Brezo Peak 50 360 1560 2220 3720 4620 7500 7980 10800 11520 13980 14460 18360 19200 ANUGA entire 180 1800 3960 8100 11700 14940 19800

150 2400 4680 8400 11100 14700 19500

ESTIMATED PEAK DISCHARGES COMPARED (m3/s) Brezo ANUGA Historic Brezo ANUGA Est. Peak (Sanders) XS XS Location 48145 43500 45694 1 DS Dam Wall 37748 2 Midway to PS2 36817 30500 32783 A-A 3 Power house 2 22500 23940 B-B 4 Saugus substation 17000 15920 C-C 5 Edison Camp at Kemp 11500 10732 D-D 6 Bardsdale Bridge 7600 7063 E-E 7 Santa Paula Bridge 6000 5973 F-F 8 Saticoy Bridge 4000 4440 G-G 9 Ventura Ocean

HWRS 2012 Van Drie, Milevski Simon

5 of 9

2D Hydrology:- Using the SWW equation to model hydrologic response

VanDrie

Figure 9. Comparison of Hydrograph Shapes BREZO (Sanders, 2007) top, ANUGA bottom Given the positive results of the Dam Break exercise, ANUGA has also been applied to: - Wupper River in Germany 83,000 hectares (Figure 10.) - Condamine/Balonne River Queensland Australia, 8,500,000 hectares (Figure 11.) Both these have rainfall being applied directly to the entire catchment. Since ANUGA is capable of using an unstructured (variable size) mesh, the process is for an initial run to provide an outline of wet extent which will be used to seed the refined mesh. It has been found that several (~2) iteration of this process quickly provides an optimized mesh layout of highly refined in areas to capture detail but also minimizing the total number of cells to control memory requirements and speed. Note that these issues are likely to be addressed in the future by utilising adaptive meshing and GPU processing rather than CPU processing. Note ANUGA is currently capable of running in parallel on multiple cores, multiple cpus and multiple machines (Cluster). However the GPU approach is likely to be an even greater speed-up. The interim results to date are very encouraging. Comparison with hydrographs generated by the widely accepted hydrologic model WBNM indicate ANUGA is performing very well as a hydrologic model. The immediate response of WBNM is not reflected, yet peak flows and timing are generally

HWRS 2012 Van Drie, Milevski Simon

6 of 9

2D Hydrology:- Using the SWW equation to model hydrologic response consistent the reason for this is that WBNM has no data regarding the terrain or surface roughness.

VanDrie

Figure 10. 83,000Ha Wupper ANUGA model

Figure 11. 8,500,000Ha Condamine/Balonne

Figure 12 shows the depth resulting from a 1:500 year critical storm duration for the catchment.

Figure 12. First refinement results 1:500 year flood depth from rainfall on 8,500,000 hectare catchment

6. EVEN LARGER (A SINGLE NATIONAL MODEL?):


It is understood that a model of the entire nation of France (54,563,000 hectares) was developed using JFLOW by JBA consulting. Very little is known of this application, or whether it uses flow or rainfall as input, it is however a fixed grid model rather than unstructured grid. However, validating rainfall distribution would become a more vital issue on very large and multiple catchments such as this. The ability of JFLOW to provide this extreme capability is apparently reliant on access the GPU rather than CPU which is reported to provide a speed up of 100X.
http://www.gccapitalideas.com/2010/06/17/a-national-flood-modeling-solution-for-mainland-france/

HWRS 2012 Van Drie, Milevski Simon

7 of 9

2D Hydrology:- Using the SWW equation to model hydrologic response

VanDrie

Figure 13. Flood model of ALL of France

7. CONCLUSIONS:
ANUGA is not the only two dimensional model capable of modelling large catchments and nor is it the only model capable of modelling rainfall on the domain. What is unique about ANUGA is that mesh resolution can be changed. This allows the user to create models which can be very detailed is area of interest while keeping mesh size coarse in other areas which helps reduce model run times. ANUGA is also free and open source which allows users to contribute to its development and add features. As discussed in this paper, ANUGA can be used on a macro scale to simulate riverine flooding on a catchment scale, with the ability to apply rainfall directly on the domain or using the traditional method of applying hydrographs on the domain (or a combination of the two). Application of rainfall on the domain allows the user to not only determine mainstream flooding but also determine overland flow paths at the same time. ANUGA is also a very capable model on a micro scale. Figure 14 shows an example of flow through a very complex curved labyrinth weir located in Wollongong NSW. This structure exhibits extreme dynamic behaviour resulting in multiple hydraulic jumps and reflective moving shock waves through out the structure and energy dissipater area. This type of application can certainly assist immensely in the analysis and design of such complex structures, in providing the designer with specific insight into detailed momentum plumes within such a structure (Figure 15.).

Figure 14. Brokers Rd basin

Figure 15. Momentum plume within the basin

HWRS 2012 Van Drie, Milevski Simon

8 of 9

2D Hydrology:- Using the SWW equation to model hydrologic response

VanDrie

8. REFERENCES
ANUGA User Manual Release 1.2.1, Geoscience Australia and the Australian National University Baldock, T., Barnes, M., Guard, P. Thomas Hie , D. Hanslow, D., Ranasinghe, R., Gray, D., Nielsen, O., 2007, Modelling tsunami inundation on coastlines with characteristic form, 16th Australasian Fluid Mechanics Conference, Gold Coast, Australia BARNES, M.P. AND BALDOCK, T.E., 2007. Direct Bed Shear Stress Measurements in Laboratory Swash. Journal of Coastal Research, SI 50 (Proceedings of the 9th International Coastal Symposium), 641 645. Gold Coast, Australia, ISSN 0749.0208 Begnudelli,L., Sanders, B.F., 2007, Simulation of the St. Francis Dam-Break Flood, ASCE; JOURNAL OF ENGINEERING MECHANICS, NOVEMBER 2007 Clark, K. Ball, J. Babister, M. (2008), Can Fixed Grid 2D Hydraulic Models be Used as Hydrologic Models?. Water Down Under 2008, 31st Hydrology and Water Resources Symposium and the 4th International Conference on Water Resources and Environment Research Conference, Adelaide, April 4-17, 2008. Jakeman, J. 2006, ON NUMERICAL SOLUTIONS OF THE SHALLOW WATER WAVE EQUATIONS, ANU Thesis Luis, J.F.2007, Mirone: A multi-purpose tool for exploring grid data. Computers & Geosciences, 33, 31-41 Milevski, P. 2009, 2012, Calibration of Towradgi Creek Catchment ANUGA Flood Model on the 17 August 1998 Storm Event, Wollongong City Council Internal Report Nielsen, O. and Gray, D. 2005. Hydrodynamic inundation modelling for disaster risk management. Extended Abstracts of the 17th BMRC Modelling Workshop, BMRC Research Report no.111, pp65-67 (this volume). Nielsen, O., Roberts, S., Gray, d., McPherson, A.,Hitchman, A., 2005, Hydrodynamic modelling of coastal inundation, MODSIM2005 VanDrie, R., Milevski, P., 2009, ANUGA Free & Open Source 2-D Hydrodynamic Model, with Hydrologic Capability, 32nd Hydrology and Water Resources Symposium - Newcastle Van Drie, R., Milevski, P., Simon, M., 2011, Validation of a 2-D Hydraulic Model - ANUGA, to undertake Hydrologic Analysis, IAHR, BRISBANE. Zoppou, C., Roberts, S. 1999, Catastrophic Collapse of Water Supply Reservoirs in Urban Areas, ASCE Journal of Hydraulic Engineering 1999, Pgs. 686 695

HWRS 2012 Van Drie, Milevski Simon

9 of 9

You might also like