You are on page 1of 6

30th Hydrology and Water Resources Symposium 4 - 7 December 2006 Launceston, TAS

Design flood estimation in small catchments using twodimensional hydraulic modelling A case study
Steve Muncaster Associate Water Technology 15 Business Park Drive, Notting Hill VIC 3168 Email: steve.muncaster@watech.com.au Warwick Bishop, Associate Water Technology 15 Business Park Drive, Notting Hill VIC 3168, Email: warwick.bishop@watech.com.au Andrew McCowan Director Water Technology 15 Business Park Drive, Notting Hill VIC 3168, Email: Andrew.mccowan@watech.com.au
Abstract: Design flood estimates are often required for small catchments as part of development planning and/or infrastructure design. Traditionally, the probabilistic Rational Method has been the principal approach to estimate design peak flow with simple hydraulic calculations employed to size culverts and bridge waterways. The application of the Rational Method, while well ingrained in engineering practice, relies on significant simplifications of the catchment runoff process. These simplifications can lead to uncertainty surrounding design flood estimates. More sophisticated analysis is possible through the use of runoff routing models such as RORB, Rafts or Urbs. Such models allow greater detail to be incorporated into the analysis at a subcatchment level, although some simplifications of the runoff process are still necessary. Recent developments in twodimensional hydraulic models enable the direct application of rainfall excess onto the computational grid. Increasingly high resolution topographic data is becoming available, often associated with proposed development of a particular area. These two developments facilitate the application of twodimensional hydraulic models as a runoff routing model in small catchments. This paper discusses the key aspects of this application including appropriate topographic data sources, computational grid resolution and effective roughness values. Results from a preliminary application to small rural catchments in the Geelong region are presented and used to illustrate key aspects. In particular, the selection of appropriate hydraulic roughness is critical. Better representation of catchment storage is provided through the use of topographic data in the computational grid. The use of a two-dimensional hydraulic model integrates the hydrologic and hydraulic aspects into a single model. Further investigation is required to assess the role of hydraulic roughness in determining surface runoff rates. Keywords: 1 runoff routing, hydraulic roughness, catchment storage, flood estimation catchment losses) and the available catchment storage. The Rational Method lumps the loss characteristic into the runoff co-efficient and the storage catchment is reflected in the time of concentration. Runoff routing models (RORB, (Laurenson and Mein 1997), WBNM (Boyd et al 1994), & URBS (Carroll 2003), etc) split these two characteristics with the runoff production dealt with by the loss model. The catchment storage is reflected by the use of a power relationship between catchment storage and m outlet. The familiar S = KQ is found in various forms in RORB, WBNM and URBS. More rigorous analysis of catchment response, particularly catchment storage, have been in the past limited by the lack of available data required to construct such models.

INTRODUCTION

The flood behaviour analysis of small catchments is commonly undertaken as part of infrastructure and development conceptual design. This flood analysis typically consists of two components: hydrologic analysis (determination of peak flows and flood hydrographs) and a hydraulic analysis (determination of flood depths, extents and conceptual design of hydraulic structures). Historically, the common approach employed the Rational Method and simple culvert routines as the two principal analysis tools. A more refined approach may employ a runoff routing model (e.g. RORB (Laurenson and Mein 1997)). Key catchment characteristics for determining flood response are runoff production (i.e.

30th Hydrology and Water Resources Symposium 4 - 7 December 2006 Launceston, TAS

Assessment of the variation of available catchment storage within small catchments can be aided by available topographic data. Twodimensional hydraulic models can utilise available topographic data in the analysis of available catchment storage and its role in flood response. Recent developments in the cost effective collection of topographic data, such as Airborne Laser Scanning (ALS), has lead to increased availability of detailed topographic information and the potential use of two-dimensional hydraulic analysis. Further, two-dimensional hydraulic models have evolved significantly over recent years (McCowan et al 2002). Generally the use of two-dimensional (2D) hydraulic models has been for the determination of flood extents and levels over large floodplains. With this approach, a hydrologic model/analysis has provided a flood hydrograph as a boundary condition to the hydraulic model. However, many 2D hydraulic models are now capable of modelling the effect of net rainfall on the computational grid. This development has enabled the integration of the hydrologic and hydraulic components of flood behaviour analysis for small catchments. This paper discusses the critical aspects of the application of 2D hydraulic models as a tool for runoff routing in small catchments. Particular attention is paid to the following: o Analysis of catchment storage o Role of hydraulic roughness (Mannings n) in determining catchment response o Conservation of mass within two dimensional models when applied with direct net rainfall. A case study for several small catchments to the south of Geelong is provided to illustrate this approach. This paper aims to provide an initial insight to the application of 2D hydraulic models to runoff routing and is not intended as a rigorous review of this application. 2 APPLICATION OF HYDRAULIC MODELLING TO RUNOFF ROUTING AT CATCHMENT SCALE

overland flow paths. These models include Liu et al (2003), Jain and Singh (2005), and Fortin et al. (2001). A review of similar applications is provided in Singh and Woolhiser (2002). Typically these models have employed the kinematic wave and/or diffusive wave approach to the solution of the Saint Venant equations. These approaches overcome the computational demands of the full dynamic wave hydraulic model (Singh 1996). The kinematic wave and/or diffusive wave approach typically uses a lower spatial resolution (larger computational grid size). This lower spatial resolution facilitates the use of more coarse topographic data as the basis for the computational grid. Further these models typically simulate the runoff generation component as well as the runoff routing component of surface flows. Fully-dynamic wave 2D hydraulic models have generally been employed for floodplain hydraulic analysis. Recent developments have seen dynamic wave hydraulic models, as such MIKEFlood (DHI 2005) and TUFLOW (WBM 2006), able to have direct net rainfall on the computational grid. The fully-dynamic wave models are generally employed with a high spatial resolution (smaller computational grid size) than the diffusive wave models discussed above. This high spatial resolution allows for small features which influence catchment storage characteristics to be adequately schematised and resolved (Horritt and Bates 2001). The increased availability of high resolution topographic data has the potential to aid the application of fully-dynamic wave models to runoff routing. Various formulations of hydraulic roughness have been employed in the 2D hydraulic models. Determination of hydraulic roughness parameters is well developed for floodplain applications of 2D hydraulic models. For these floodplain applications, hydraulic roughness values are ideally determined through model calibration against observed flood levels and extents. However, in the absence of observed flood data, many references (e.g. Chow 1959) provide guidance. In typical floodplain applications of 2D hydraulic models, flow depths are generally in order of tens of centimetres to several metres. When applied for the purposes of runoff routing, flow depths away from the main waterways are small (< 0.1 m). At such shallow flow depths, the effective roughness may increase due to the effect and type of ground cover. As a result, the roughness value applied to the same type of ground cover may vary as the flow depth

A number of hydraulic based models have been developed for routing of surface runoff at the catchment scale. These models use topographic information at various resolutions, and hydraulic solution techniques as the basis to evaluate the connectivity and conveyance of

30th Hydrology and Water Resources Symposium 4 - 7 December 2006 Launceston, TAS

increases. Further investigation of appropriate roughness values away the watercourses is required to refine the application of dynamic wave 2D hydraulic models to runoff routing. 3 CASE STUDY ARMSTRONG CREEK, GEELONG 3.1 Study area

External inflow catchments:- MIKE Flood model was applied as per the enclosed catchments for the portion of the catchment within the study area. Runoff contributions from the catchment external to the study area were then determined by a RORB model.

The case study focuses on 6 small catchments located to the south west of Geelong. The study area has been identified as a possible future urban growth area for Geelong. Figure 1 (provided at the end of the paper) shows the general location of the study area and catchment delineation. The naming convention for the catchments was taken as the City of Greater Geelong specification. The catchment areas are: C253&C254 442 ha, C255C 324 ha , C257 2833 ha, C267A - 74 ha, C267C 45 ha and C267D -32 ha. The study area consists mainly of rural land. The associated drainage is via a number of drainage depressions and larger watercourses. Generally the watercourses in the study area are not well defined with extensive shallow flooding occurring frequently. The seven catchments can be divided into the following two groups o o Enclosed catchments catchments entirely contained within the study area. External inflow catchments Catchments not fully enclosed in the study area and have catchment inflows from outside the study area. General model application

The MIKE Flood models employed a 5 m computational grid for all the catchments in this case study. Both the MIKE Flood and RORB models were run for a range of design storm durations in order to assess the critical durations. 3.3 Available input model data

There is no streamflow data available for the catchments in the study area. As such, formal calibration of the hydrologic and hydraulic analyses has not been possible. A photogrammetric survey covered the entire study area consisting of a regular grid of spot elevation, contours and breaklines to define linear features. Further, the details of culverts, bridges and retarding basins were incorporated into the model. 3.4 Design rainfall and losses

3.2

Design rainfall depths and temporal patterns were obtained from Australian Rainfall and Runoff (IEAust 1999). A uniform design rainfall spatial pattern was applied. To determine net 100 year design rainfalls for input into the MIKE Flood, an initial loss of 10 mm and continuing loss of 2 mm/hour were adopted for the rural areas. 3.5 External catchment design inflow hydrographs

The 2D hydraulic modelling package employed for this study was MIKE Flood. The MIKE Flood package is a state of the art tool for floodplain modelling that has been formed by the dynamic coupling of Danish Hydraulics Institutes well proven MIKE 11 river modelling and MIKE 21 fully two-dimensional modelling systems. (DHI 2005) Two approaches have been used for the application of MIKE Flood. These two approaches were developed to reflect the two catchment types as follows: o Enclosed catchments:- MIKE Flood was applied as an integrated hydrologic and hydraulic model to the entire catchment. Direct net rainfall was applied to the computational grid.

For the external inflow catchments, the RORB models were developed for the entire catchments. The RORB parameter kc was determined using a regional prediction formula (Pearse et al 2002). The RORB models incorporated the existing retarding basins. The design losses, for the RORB models, were adopted as an initial loss of 10 mm and continuing loss of 2 mm/hour. In urban areas, a fraction impervious of 0.45 was adopted.

30th Hydrology and Water Resources Symposium 4 - 7 December 2006 Launceston, TAS

3.6

Hydraulic roughness

4.2

Flood hydrograph shape

As discussed, the estimation of hydraulic roughness is complicated by the shallow flow depths occurring in catchments away from the formal watercourses. To assess the impact of roughness on computed peak flow at the catchment outlet, the following two hydraulic roughness scenarios were tested: 1. Uniform roughness across the entire catchment (Mannings n = 0.04) 2. Varied roughness with higher roughness in areas away from the watercourses. (Mannings n = 0.04 for watercourses and n = 0.1 for remainder of catchment) 4 RESULTS 4.1 100 year peak flow

As RORB models were developed for the two external inflow catchments, design 100 year flood hydrographs were available for comparison to the MIKEFlood hydrographs. Figures 2 and 3 shows the RORB model and MIKEFlood 100 year design flood hydrographs for the C253&C254, and C257 catchments for the uniform roughness scenario.
16 14 12

RORB model MIKEFlood

Flow (m3/s)

10 8 6 4 2 0 0:00

For the enclosed catchments, the subcatchment peak flows obtained from the MIKE Flood models were compared to peak flows obtained from the Rational Method. The Rational Method as outlined by VicRoads (1999) was applied with a 10 year runoff coefficient (C10) of 0.1. The 10 year runoff coefficient was obtained from Australian Rainfall and Runoff (IEAust 1999). Table 1 Design 100 year peak flow estimates: Enclosed catchments
Catchment Rational Method peak flow (m3/s) 8.6 2.9 2.0 1.6 MIKEFlood peak flow (m3/s) Uniform Varied roughness roughness (floodplain 0.1) 19.6 4.4 4.6 4.8 13.3 2.6 3.4 3.8

1:00

2:00

3:00

4:00

5:00

6:00

7:00

8:00

9:00

10:00 11:00 12:00

Time (h)

Figure 2 C253&C254 100 year design flood hydrographs


70

60

50

Flow (m3/s)

40

RORB Mike Flood

30

20

10

C255 C C267 A C267 C C267 D

0 0:00

2:00

4:00

6:00

8:00

10:00 12:00 14:00 16:00 18:00 20:00 22:00

0:00

Time (h)

Figure 3 C257 100 year design flood hydrographs 4.3 Mass conservation

For the external inflow catchments, Table 2 shows the MIKEFlood 100 year peak flows compared to the RORB model peaks flows. Table 2 Design 100 year peak flow estimates: External inflow catchments
Catchment RORB model (m3/s) 12.7 44 MIKEFlood peak flow (m3/s) Uniform Varied roughness roughness (floodplain 0.1) 14.8 66 13.6 48

To ensure the MIKEFlood model was adequately conserving mass, a comparison of the net rainfall volume to the computed runoff volume was made with the results shown in Table 3. Table 3 Mass conservation comparison
Catchment C255 C C267 A C267 C C267 D Mass error (% of total inflow) 2.7 1.8 1.0 1.0

C253 &C254 C257

30th Hydrology and Water Resources Symposium 4 - 7 December 2006 Launceston, TAS

4.4

Catchment storage

For enclosed catchments over the duration of the 100 year design flood, the flood storage within the catchment and flow rate at the outlet was determined at 5 minute intervals during the hydrograph recession. Figure 4 shows the storage-outflow relationships for the uniform roughness scenario configuration discussed in Section 3.6.
90000 80000 70000 60000

1 in 5 to 10. It is considered that this higher slope results in the MIKE Flood model producing larger peak flow estimates. The routing of overland flow in MIKE Flood accounts for the slope of the terrain. The Rational Method has no allowance for terrain slope in its application. Similarly, for the C253&254, and C257 catchments, the MIKE Flood peak estimates are larger than the RORB model estimates. For both sets of catchments, the use of the varied roughness scenario has lead to a decrease in the MIKEflood peak flow estimates. These reductions range from 8% (C253/254) to 41 % (C267A). As expected, the increase in Mannings n for overland flow areas, leads to a decrease in flow velocity, increased attenuation of the runoff and in turn lower peak flows.
25

Storage (m3)

C255C C267A C267C C267D

50000 40000 30000 20000 10000 0 0 5 10 15

20

Flow (m3/s)

Figure 4 Catchment storage relationships 4.5 Overland path delineation

As the 2D hydraulic model calculates flow depths across the entire catchment, the flood extent and overland flow path can be determined. Figure 1 displays the 100 year flood extents. This has particular benefit in flat terrain where the overland flow paths are illdefined. 5 DISCUSSION For the enclosed catchments, 100 year peak flows from the MIKE Flood models exceed the Rational Method, except for C267A with the varied roughness scenario. As there is no calibration data available, an assessment of the reliability of the peak estimates is not possible. However, MIKE Flood peak flows appear to be of the same order as the RORB and Rational method. The MIKE flood model peak flow estimate for C255C is significantly higher than the Rational Method estimate. The catchment for C255C is relatively linear in shape. It is considered that the shape of the catchment results in a shorter overland flow paths to the main watercourse. These shorter overland flow paths then result in a higher peak flow. The MIKE Flood models yield significantly higher peak flows for C267C and C267D than the Rational Method. These sub-catchments are located in the upper part of the C267. As a result, the slope of the terrain is of the order of

Ideally, through model calibration against observed streamflow data, appropriate values and spatial variation of Mannings n could be quantified. The variation of Mannings n from the watercourse to overland flow areas is likely related to ground cover/vegetation and flow depth. Further, a parallel could be drawn in hydrological runoff routing, as such URBS (Carroll 2002), with the division of channel and sub-catchment routing parameters. The mass balance errors of 1 - 2.7 % are considered reasonable for this type of application and reflect the stability of the computational scheme. The catchment discharge storage relationships shown in Figure 4 display a high degree of linearity across a range of flows with an apparent non-linear response in the lower flow range. Further investigation is required to assess whether the non-linear response is related to in-channel flow, while the linear response is related to extensive overland flow. 6 CONCLUSIONS This paper outlines an application of a 2D hydraulic model to routing of surface runoff within a small catchment. The functionality of direct net rainfall of the computational grid has enabled such application. The increased availability of high resolution topographic data has further added to the development of such models. Particularly, the spatial variation of hydraulic roughness is seen as an important model parameter. The suitability of standard hydraulic roughness (Mannings n) values in shallow

30th Hydrology and Water Resources Symposium 4 - 7 December 2006 Launceston, TAS

overland flow investigation.

areas

requires

further Jain, M.K., and V.P. Singh, DEM-based modelling of surface runoff using diffusion wave equation, Journal of Hydrology, 302, 107-126, 2005. Laurenson, E.M. and R.G. Mein, RORB Version 4, Runoff Routing Program User Manual, Dept. of Civil Engineering, Monash University 1997. Liu, Y.B., S. Gebremeskel, F. De Smedt, L. Hoffmann, and L. Pfister, A diffusive transport approach for flow routing in GIS-based flood modelling, Jour. of Hydrology, 283, 91-106, 2003. McCowan, A.D., E.B. Ramussen and P.E. Berg, Improving the performance of a two dimensional hydraulic model for floodplain application, Proc. Conf. on Hydraulics in civil Engineering, Hobart, 2002. Pearse M., P. Jordan and Y. Collins, A simple method for estimating RORB model parameters for ungauged rural catchments. 27th Hydrology and Water Resources Symposium, Melbourne. Institution of Engineers, Australia. 2002 Singh, V.P. and D.A. Woolhiser, Mathematical modelling of watershed hydrology. J. Hydrol. Eng., ASCE 7 (4), 270-292. 2002. VicRoads, Road Design Guidelines Part 7 Drainage. VicRoads. Melbourne. 1999. WBM, TUFLOW User Manual, June 2006.

The application allows the evaluation of catchment discharge storage relationships. Further investigation of these relationships may yield refinements of the relationships currently in the use by hydrological runoff routing models. REFERENCES Boyd, M.J., E.H. Rigby, M.G. Sharpin and R. VanDrie, Enhanced runoff routing model WBNM. Instn. Engineers Australia, Water Down Under Conference, pp.445-448. 1994. Carroll D, URBS a catchment management and flood forecasting rainfall runoff routing model. Version 3.94 User manual. Brisbane 2002 Chow, V.T., Open channel hydraulics. McGraw Hill, New York. 1959. DHI, MIKEFlood, Technical Reference and Model Documentation. 2005. Fortin, J.P., R. Turcotte, S. Massicote, R. Moussa and J. Fitzback, A distributed watershed model compatible with remote sensing and GIS data. Part -1. Description of the model, J. Hydrol. Eng., ASCE 6 (2), 91-99. 2001. Horritt, M.S and P.D. Bates, Effect of spatial resolution on raster based model of flood flow. Journal of Hydrology, 253, 239-249, 2001. Institution of Engineers Australia, Australian Rainfall and Runoff, Vols 1&2. (Ed: Pilgrim D.H.) 1999.

Figure 1 Catchments and flood extents

You might also like