Flood damage estimation beyond stage^damage functions: an Australian example

Transcription

1 Flood damage estimation beyond stage^damage functions: an Australian example M.H. Middelmann-Fernandes Geoscience Australia, Canberra, Australia Correspondence: M. H. Middelmann-Fernandes, Geoscience Australia, GPO Box 378, Canberra ACT 2601, Australia. DOI: /j X x Key words Damage; damage limitation; flood; loss; stage; velocity. Abstract A number of factors influence the direct consequence of flooding. The most important are depth of inundation, velocity, duration of inundation and water quality and the interaction of these factors with human society. Although computer modelling techniques exist that can provide an estimate of these variables, this information is seldom used to estimate the impact of flooding on a community. This work describes the first step to improve this situation using data collected for the Swan River system in Perth, Western Australia. Here, it is shown that residential losses are underestimated when stage damage functions or the velocity stage damage functions are used in isolation. This is because the functions are either limited to assessing partial damage or structural failure resulting from the movement of a house from its foundations. This demonstrates the need to use a combination of techniques to assess the direct cost of flooding. Introduction Floods frequently have a direct impact on Australian communities; therefore, estimation of flood damage, both past and future, is important. Accurate loss estimation provides useful input into the management of flood risk. In conjunction with other considerations such as community safety, cost information influences the formulation of flood policy including the allocation of funding, regulations and plans aimed at reducing flood risk. A comparison of the cost of rapid-onset natural disasters in Australia over the period has confirmed that floods are the most costly natural disaster (BITRE, 2008). This and earlier comparisons (BTE, 2001) have been used to influence natural disaster expenditure in favour of flood risk management projects. The limitations of cost estimates in Australia are, however, well recognized (SCARM, 2000; BTE, 2001; Middelmann, 2007). Furthermore, the development of guidelines for properly costing flood damages has been identified as part of the future work programme of the National Flood Risk Advisory Group, an Australian body of government and nongovernment flood management professionals. Accurate loss data play an integral role in assessing the cost benefit of any proposed mitigation strategy. Some examples for Australia are highlighted in BTRE (2002). Damage estimates influence decisions on disaster relief assistance, land-use planning and research priorities. However, obtaining accurate loss data is a challenge shared globally. This is demonstrated for Germany by Merz et al. (2004) and for the United States by Downton and Pielke (2005). Numerous parameters contribute to flood damages including water depth (McBean et al., 1986; Smith, 1994; USACE, 1996; Green, 2003), flow velocity (Sangrey et al., 1975; USBR, 1988; Clausen, 1989), duration of inundation (Parker et al., 1987; Lekuthai and Vongvisessomjai, 2001; FEMA, 2005) and contamination, sediment or debris load (Haehnel and Daly, 2002; Penning-Rowsell et al., 2003; Thieken et al., 2005). Building construction, age and materials (Dale et al., 2004; Zhai et al., 2005), and warning time and previous experience with flooding also influence flood damages (Smith, 1986; Smith et al., 1995). Further examples from the literature are provided by Kelman and Spence (2004) in their overview of flood actions on buildings and in Proverbs and Soetanto (2004) in their book on flooddamaged property. In summary, the consequences of flood damages are the result of the interaction of a number of flood characteristics with human behaviour. Of all these flood characteristics, only one, depth of inundation, has received widespread usage in the assessment of residential flood damages. This paper contributes to attempts to improve the assessment of flood damages. Velocity and water depth combinations can be used to provide an indicator of the impact

2 Flood damage estimation beyond stage damage functions 89 against an object such as a person or a building. Therefore, velocity is an important parameter in assessing flood damages and the potential for loss of life or injuries resulting from flooding. There is only one published example of applying velocity and water depth to estimate damage to buildings in Australia. In this mitigation study, losses were estimated using damage functions developed for the United States (Black, 1975) to buildings in Queanbeyan and Canberra as a result of dam failure (Smith, 1990). Subsequent work on the development of similar loss functions, however, demonstrated the inappropriateness of the United States curves in Australia (Dale et al., 2004). This paper considers only the direct tangible cost for residential structure and contents using synthetic damage functions and compares two methods for analysing these. The first method uses information on water depth only. The second method estimates damage using water depth and velocity to estimate which houses are likely to move off their foundations. Stage--damage functions Direct flood damages to residential buildings are typically assessed worldwide using stage damage functions (McBean et al., 1986; Smith, 1994; Dutta et al., 2001; Meyer and Messner, 2005; Messner et al., 2007). Damage is based on stage height (i.e. water depth) either as a percentage damage or loss to building structure and/or to building contents. Considerable uncertainty is inherent in stage damage functions. For example, data for both actual direct structural and content damages for residential buildings from the 1986 flood in Sydney, Australia, show considerable scatter (Smith, 1994). In the same paper, the considerable smoothing needed to develop stage damage curves from raw data is acknowledged. To the author s knowledge, no stage damage functions facilitate estimation of total loss; therefore, in flood events where buildings are destroyed or considered not worth repairing, the use of such functions in isolation underestimates damage. Stage damage functions are produced in a number of different ways. They can be empirical curves based on damages from a historical flood or flood events in a specific location and therefore represent actual damages from that event. Alternatively, synthetic stage damage functions can be developed. Unlike synthetic functions that are based on one or two parameters, for example, water depth, duration and/or warning time (Penning-Rowsell and Chatterton, 1977; Parker et al., 1987), empirical stage damage functions will include the influence of many physical factors on buildings (e.g. velocity, water depth, sediment, contamination, debris load, duration of inundation and warning time). The damage functions produced from historical data may then be used to estimate damage for a subsequent flood event at the same location, although if the specific circumstances between the floods are different, then estimates will not be realistic. Examples of empirical curves include the work by Nascimento et al. (2006) for Itajuba, Brazil; Dutta and Tingsanchali (2003) for Bangkok, Thailand, and the Snowy Mountain Engineering Corporation (SMEC, 1975) for Brisbane, Australia. Synthetic functions are hypothetical curves developed independently from historical flood data for a specific area; therefore, they do not rely on the time-consuming and often difficult collection of damage data. Unlike empirical curves, they can also be used in different areas, enabling unqualified comparisons between these areas. Synthetic functions may be developed using data from surveys, insurance companies, loss adjusters or quantity surveyors, enabling extrapolation to other areas (Greenaway and Smith, 1983). Examples of synthetic functions for residential buildings include ANU- FLOOD for Australia (Smith and Greenaway, 1988; Greenaway and Smith, 1993), the Blue Manual for the United Kingdom (Penning-Rowsell and Chatterton, 1977; Penning- Rowsell et al., 2003) and HOWAD (Flood Damage Simulation Model, German: Hochwasser-Schadens-Simulations-Model) for Germany (Neubert et al., 2009). With some recent exceptions (Parker et al., 2008), synthetic functions may overestimate the actual damages experienced by a community because the effects of flood loss-reducing measures, such as lifting contents above the flood level, are not taken into account. However, as synthetic functions represent the full extent of damage or loss that may be sustained, their use is recommended by Smith (1994) in his review of urban stage damage curves. The use of both synthetically and empirically derived damage curves is recommended by McBean et al. (1986), who found that studies that used synthetic damage curves calibrated against observed flood damage were the most accurate in assessing damages. In Australia, the ANUFLOOD suite of stage damage functions is the most commonly used tool to assess flood damage (Middelmann et al., 2005a). Their limitations are, however, highlighted in the ANUFLOOD handbook (Greenaway and Smith, 1983), which emphasizes that the use of stage damage functions is only appropriate in gently flowing waters (velocity o 1 m/s) because of the increased likelihood of buildings in faster flowing floodwaters suffering structural failure. A report by the New Zealand Institute of Economic Research (NZIER, 2004) further suggests that the use of synthetic stage damage functions is really only applicable when the rate of rise of a flood is slow, velocity is low and there is a low silt content in the water. This study used synthetic stage damage functions developed using data provided by a quantity surveyor on the cost to repair single-storey residential structures in the city of Perth, Western Australia (K. Dale, unpublished data).

3 90 Middelmann-Fernandes Cellars were not considered as these are rare in Australian houses. The contents curve was developed using data on insured replacement cost and assumes that no measures are adopted to reduce the potential loss (K. Dale, unpublished data). Both the structure and the contents curves are shown in Figure 1. Velocity-- stage-- damage functions While current stage damage functions only assess partial damage, records of historical floods demonstrate that complete failure of buildings has occurred, for example, during the 1893 and 1974 floods in South East Queensland, Australia (Middelmann et al., 2001). Moreover, with Figure 1 Stage damage structure and contents functions for singlestorey residential buildings (K. Dale, unpublished data). considerable uncertainty existing in stage damage functions, the development and use of synthetic damage functions incorporating additional factors is important. Where high velocities exist, buildings can be moved off their foundations even in relatively shallow waters as a result of buoyant and horizontal forces as demonstrated by Black (1975) for houses in the United States. Black s work has been cited in the global literature to demonstrate the importance of considering velocity (Clausen and Clark, 1990; Smith, 1994; Roos, 2003; Kelman and Spence, 2004), although Sangrey et al. (1975) questioned the accuracy of the curves. More recent work by Dale et al. (2004) developed velocity stage damage functions for select Australian residential buildings, modifying Black s (1975) methodology to correctly apply the effects of drag. The size and shape of a typical single-storey house in Perth, Australia, was adopted, which was found to be 2.7 times larger than the houses used in Black s (1975) North American study. The Australian study concluded that the water depth and velocity combinations needed to move Australian buildings off their foundations are higher because the greater weight of the houses generally makes them more stable at low velocities (Dale et al., 2004). The same authors, however, found that the lightest buildings modelled (fibro and timber houses with steel or metal roofs) could still fail at velocities as low as 0.4 m/s where water depths in the vicinity of 2.4 m were present. Figure 2 shows the six Australian velocity stage damage curves developed by assessing gravity, buoyancy and dynamic forces (due to flowing water) for combinations of wall type (brick veneer, fibro or timber) and roof type (tile or steel), for various combinations of water depth and flow velocity. Building type plays a significant role in the Figure 2 Velocity stage damage functions for single-storey detached houses (Dale et al., 2004 and K. Dale, unpublished data).

4 Flood damage estimation beyond stage damage functions 91 resiliency of buildings, with heavier buildings being less buoyant and therefore less likely to be displaced from their foundations due to the greater frictional force present (Dale et al., 2004). The velocity stage damage functions in Figure 2 assess the overall instability. Where the combination of velocity and water depth at a particular structure falls to the right of the curve, the structure is considered to have become unstable and moved off its foundations. Structures are considered stable where the combination of velocity and water depth at the structure falls to the left of the curve (Dale et al., 2004). Study area Located on the Indian Ocean coastline, Perth is the capital city of Western Australia (Figure 3). With a population in excess of 1.5 million, Perth is Australia s fourth most populous city (ABS, 2008). The results presented in this study were for a single urban region, which was selected as information on velocity and water depth was available. The Swan River is the most significant river flowing through Perth, dividing the metropolitan area into two. Six major tributaries contribute flow to the Swan River, of which the Canning River contributes the most. Both urban and rural areas were modelled, the latter crucial in identifying potentially flood-prone areas where future development may need to be restricted or regulated. Method The one-dimensional hydrodynamic model HEC-RAS (US Army Corps of Engineers, 2003a, b) was used to model flooding in the Swan River and major tributaries over a combined total distance of 188 km. Eight scenarios were modeled, ranging from the 10-year average recurrence interval (ARI) to the 2000-year ARI. A detailed description of the hydrologic and hydraulic modelling for Perth is documented by Middelmann et al. (2005b) and summarized in Middelmann (2009). The stage damage functions shown in Figure 1 were applied to a subset of the building exposure dataset described in Middelmann (2009). In that paper, residential losses to Perth were estimated solely using stage damage functions. The synthetic stage damage functions developed for brick veneer buildings were considered similar enough to those that would be estimated for fibro and timber-walled houses because of anticipated similar wetting and drying performances. Therefore, specific functions for these building types were not developed, particularly as several assumptions were made using statistical data in estimating wall type (Middelmann, 2009). As the velocity stage damage functions (Figure 2) are applicable only to brick veneer, fibro and timber detached Figure 3 The Perth study area, Western Australia. houses, only these structures are used in the comparison of loss in this study. The combination of water depth and velocity at each house was used to determine at which buildings the thresholds shown in Figure 2 were exceeded, and therefore the building could be assumed to have failed. Where a building was modelled to move off its foundations, total structure and contents loss was assumed. No damage to structure or contents was estimated if a building was considered stable. Comparison of loss estimates Table 1 shows the number of flood-affected single-storey detached houses by scenario estimated using the velocity stage damage functions shown in Figure 2 to have failed by moving off their foundations. Also shown is the number of

5 92 Middelmann-Fernandes houses affected by overfloor flooding estimated using the stage damage functions in Figure 1. The range in maximum stage height and velocity at flood-affected houses is shown in Table 1 to provide an indication of the variation in flood hazard between the scenarios. In this case study, residential buildings are only moved off their foundations for the more severe modelled events (4200-year ARI), where the houses are completely submerged and velocities are in the order of 1 m/s. The lighter timber and fibro houses are the most affected because lower velocities are required to move the buildings off their foundations. Three percent of buildings affected by overfloor flooding are estimated by the velocity stage damage functions to become buoyant in the 500-year ARI scenario, increasing to 16% during the 2000-year ARI scenario. In order to enable a direct comparison between losses estimated using the velocity stage damage functions and those estimated using stage damage functions, the stage damage functions were applied only to those houses identified using the former method as having moved off their foundations. Figure 4 shows the estimated replacement cost for the buildings based on the velocity stage damage functions. Also shown is the estimated value to rebuild or repair the same buildings based on loss estimates derived using the stage damage functions. Both contents and structure losses are included. Figure 4 shows that where buildings fail, the use of stage damage functions underestimates loss. The disparity in loss estimated between the two types of damage functions is greater when only structure is considered. For example, structural losses are 45% larger when estimated using the velocity stage damage functions over the stage damage functions during the 1000-year ARI scenario. This compares with only 31% higher for combined structural and contents losses in the same scenario. The incorporation of contents decreases the difference in losses estimated because the Table 1 Estimated number of houses affected by overfloor flooding or estimated to have failed by moving off their foundations Average recurrence interval (years) Maximum stage height (m) Flood-affected brick veneer, timber or fibro detached houses Maximum velocity (m/s) Number moved off foundations Number with overfloor flooding The maximum stage height and the maximum velocity of flood-affected houses are also shown. Figure 4 Comparison of the repair or replacement cost for detached brick veneer, fibro and timber flood-damaged houses using information on velocity and water depth, and water depth only. Losses are presented in Australian dollars at 31 September 2008 values.

6 Flood damage estimation beyond stage damage functions 93 distinction in maximum losses between the two types of damage functions is much less for contents than for structure. Published Australian stage damage damage functions based on Australian empirical or overseas data are compared in a paper by Blong (2002). The paper showed that the National Hazard Research Centre s potential loss curves (Blong, 2002) based on insurance data from the United Kingdom yielded the highest estimates of combined damage functions developed in Australia at the time of comparison. The Centre s functions were compared with empirical stage damage functions developed following several Australian flood events. The functions were also compared with ANUFLOOD, synthetic functions developed during the 1980s for residential building styles prevalent at that time. Maximum loss estimated by the Natural Hazard Research Centre s synthetic functions was 43% of the sum insured for structure and 89% of the sum insured for contents (Blong, 2002). The stage damage functions used in this study (Figure 1), developed using synthetic data from the casestudy area, estimated maximum losses that are notably higher than these earlier curves, being 26% higher for structure and 10% higher for contents. The difference in loss estimated between the two types of damage functions presented in this paper is therefore even greater when the older Australian synthetic and empirical stage damage functions are used. Figure 5 shows the structure and contents loss estimates for all detached brick veneer, fibro and timber houses affected by overfloor flooding. As in Figure 4, structural losses are estimated using information on water depth only and using the velocity stage damage functions. Losses are shown for the eight scenarios modelled. While Figure 4 highlighted that losses are underestimated by not considering the failure of buildings, Figure 5 highlights that losses are also underestimated if the buildings that suffer only partial damage are excluded in the loss estimates. For example, while a A$10 million loss to structure is estimated using the stage damage functions for the 200- year ARI scenario, no loss is estimated using the velocity stage damage functions as the occurrence of structural failure is first modelled in the 500-year ARI scenario. The average loss per building is substantially higher where failure occurs and when only the affected buildings are considered. For example, using data on velocity and water depth, over Australian dollar:1 million of damage is estimated in the 500-year ARI event, based on six buildings suffering structural failure. This equates to an average loss per building of A$ In the same scenario, up to A$17 million damage is estimated using information on depth of inundation only for a total of 192 buildings. This equates to an average loss per building of A$ Figure 6 shows the estimated loss for house structure and contents using both methods. Losses were calculated by first estimating which buildings move off their foundations based on the velocity stage damage functions. The stage damage functions were then applied to assess partial damage to the remaining flood-affected buildings. The importance of considering the impact of velocity and water depth increases during the more severe flood events as shown in Figure 6. While buildings that suffer failure contribute only 7% to the total losses in the 500-year ARI Figure 5 Comparison of the repair or replacement cost for brick veneer, fibro and timber detached houses using information on velocity and water depth, and water depth only. Losses are presented in Australian dollars at 31 September 2008 values.

7 94 Middelmann-Fernandes Figure 6 Estimated losses for brick veneer, fibro and timber house structure and contents derived using information on velocity and water depth, and water depth only. Losses are presented in Australian dollars at 31 September 2008 values. scenario, this increases to 27% in the 2000-year ARI scenario. Given the low flow velocities estimated around buildings in Perth, the proportion to which losses are underestimated by using information on depth of inundation only to assess damage will be significantly greater in areas with higher velocities and similar water depths. While the lower probability events may be of less importance in a cost benefit analysis as their losses are multiplied by the probability of occurrence, the recovery time (financial, social, etc.) following an extreme event is likely to be much larger and requires greater resources. Conclusions Estimation of flood loss is complex but an accurate estimation is important for all areas of flood risk management. This paper highlights the importance of going beyond the traditional use of stage damage functions to assess direct tangible costs. Loss was estimated using both stage damage functions and velocity stage damage functions. The results for Perth, Australia, demonstrate that both types of damage functions presented in this paper have their limitations. The velocity stage damage functions identify only those buildings that fail by moving off their foundations; therefore, buildings that may experience only partial damage are not accounted for in the estimation of loss. In comparison, the stage damage functions do not consider that buildings may fail and therefore may underestimate damage with a maximum loss of 69% estimated using these functions. Used in combination, the two methods provide a potentially more accurate reflection of the total direct cost to residential structures than the use of any particular loss function in isolation. The results presented in this paper reinforce the importance of considering multiple methods in assessing damage. Acknowledgements The author would like to thank and acknowledge the engineering and spatial analysis expertise provided by Ken Dale and Lisa Cornish. The valuable review comments received by colleagues and through the journal review process are much appreciated. This paper has been published with the permission of the Chief Executive Officer, Geoscience Australia. References ABS. Population projections, Australia, 2006 to Publication , released September Canberra: Australian Bureau of Statistics, BITRE. About Australia s regions. Canberra: Bureau of Infrastructure, Transport and Regional Economics, Black R.D. Flood proofing rural residences. A project Agnes report. New York: Department of Agricultural Engineering, Cornell University, Blong R. Estimating residential flood damage. In: D.I. Smith & J. Handmer, eds. Residential flood insurance. Canberra: Water Research Foundation of Australia, 2002, BTE. Economic costs of natural disasters in Australia. Report 103. Canberra: Bureau of Transport Economics, BTRE. Benefits of flood mitigation in Australia. Report 106. Canberra: Bureau of Transport and Regional Economics, 2002.

8 Flood damage estimation beyond stage damage functions 95 Clausen L.K. Potential dam failure: estimation of consequences and implications for planning. Enfield, UK: Middlesex Polytechnic, Clausen L. & Clark P.B. The development of criteria for predicting dambreak flood damages using modeling of historical dam failures. In: W.R. White, ed. International conference on river flood hydraulics. Hydraulics Research Ltd., September. Chichester, UK: John Wiley and Sons Ltd, 1990, Dale K., Edwards M., Middelmann M. & Zoppou C. Structural flood vulnerability and the Australianisation of Black s curves. Risk 2004 conference proceedings. Risk Engineering Society, 8 10 November 2004, Melbourne, Downton M.W. & Pielke R.A. How accurate are disaster loss data? The case of U.S. flood damage. Nat. Hazards 2005, 35, (2), Dutta D., Herath S. & Musiake K. Direct flood damage modelling towards urban flood risk management. Proceedings of the Joint Workshop on Urban Safety Engineering, International Centre for Urban Safety Engineering, Report 1, September 2001, Thailand, 2001, Dutta D. & Tingsanchali T. Development of loss functions for urban flood risk analysis in Bangkok. Proceedings of the 2nd International Symposium on New Technologies for Urban Safety of Mega Cities in Asia, International Centre for Urban Safety Engineering, October 2003, Tokyo, Japan, 2003, FEMA. Effects of long and short duration flooding on building materials. Orlando: DFO, Federal Emergency Management Agency, DR-1539/1545/1551/1565, Green C. Handbook of water economics: principals and practice. Chichester: John Wiley, Greenaway M.A. & Smith D.I. ANUFLOOD field guide. Canberra: Centre of Resource and Environmental Studies, Australian National University, Greenaway M.A. & Smith D.I. ANUFLOOD programmer s guide and user s manual. Canberra: Centre for Resource and Environmental Studies, Australian National University, Haehnel R.B. & Daly S.F. Maximum impact force of woody debris on floodplain structures. US Army Corps of Engineers Engineer Research and Development Center, Technical report ERDC/CRREL TR-02-2, Kelman I. & Spence R. An overview of flood actions on buildings. Eng. Geol. 2004, 73, Lekuthai A. & Vongvisessomjai S. Intangible flood damage quantification. Water Resour. Manag. 2001, 15, (5), McBean E., Fortin M. & Gorrie J. A critical analysis of residential flood damage estimation curves. Can. J. Civ. Eng. 1986, 13, Merz B., Kreibich H., Thieken A. & Schmidtke R. Estimation uncertainty of direct monetary flood damage to buildings. NHESS 2004, 4, (1), Messner F., Penning-Rowsell E., Green C., Meyer V., Tunstall S. & van der Veen A. Evaluating flood damages: guidance and recommendations on principles and methods. FLOODsite report number T , FLOODsite consortium, Meyer V. & Messner F. National flood damage evaluation methods a review of applied methods in England, the Netherlands, the Czech Republic and Germany. FLOODsite Project Report Contract No. GOCE-CT , Middelmann M., Harper B. & Lacey R. Flood risks. In: K. Granger & M. Hayne, eds. Natural hazards and the risks they pose to South East Queensland. Canberra: Geoscience Australia, 2001, Middelmann M., Sheehan D., Jordon P., Zoppou C. & Druery C. National catalogue of flood studies. New South Wales Floodplain Management Conference, February 2005, Narooma, 2005a. Middelmann M., White J., Rodgers S. & Zoppou C. Riverine flood hazard. In: T. Jones, M. Middelmann & N. Corby, eds. Natural hazard risk in Perth, Western Australia. Canberra: Geoscience Australia, 2005b, Middelmann M.H. ed. Natural hazards in Australia: identifying risk analysis requirements. Canberra: Geoscience Australia, Middelmann M.H. Residential flood losses in Perth, Western Australia. In: P. Samuels, S. Huntington, W. Allsop & J. Harrop, eds. Flood risk management: research and practice. London: Taylor and Francis Group, 2009, Nascimento N., Baptista M., Silva A., Léa Machado M., Costa de Lima J., Gonçalves M., Silva A., Dias R. & Machado E. Flood-damage curves: methodological development for the Brazilian context. Water Pract. Technol. 2006, 1, (1), doi: /WPT Neubert M., Naumann T. & Deilmann C. Synthetic water level building damage relationships for GIS-supported flood vulnerability modelling of residential properties. In: P. Samuels, S. Huntington, W. Allsop & J. Harrop, eds. Flood risk management: research and practice. London: Taylor and Francis Group, 2009, NZIER. Economic impacts on New Zealand of climate changerelated extreme events. Focus on freshwater floods. Report to the New Zealand Climate Change Office. Wellington, New Zealand: New Zealand Institute of Economic Research, Parker D.J., Green C.H. & Thompson P.M. Urban flood protection benefits: a project appraisal guide (The red manual). Aldershot: Gower Technical Press, Parker D.J., Priest S., Schildt A. & Handmer J. Modelling the damage reducing effects of flood warnings, FLOODsite Final Report T , HR Wallingford, UK, Penning-Rowsell E. & Chatterton J.B. The benefits of flood alleviation: a manual of assessment techniques (The blue manual). Aldershot, UK: Gower Technical Press, Penning-Rowsell E., Johnson C., Tunstall S., Tapsell S., Morris J., Chatterton J., Coker A. & Green C. The benefits of flood and coastal defence: techniques and data for London: Enfield: Flood Hazard Research Centre, Middlesex University, Proverbs D.G. & Soetanto R. Flood damaged property: a guide to repair. Oxford, UK: Blackwell Publishing, 2004.

9 96 Middelmann-Fernandes Roos I.W. Damage to buildings. Delft Cluster-publication DC the Netherlands, Delft: DUP Standard, Sangrey D.A., Murphy P.J. & Nieber J.L. Evaluating the impact of structurally interrupted flood plain flows. Technical Report 98. Ithaca, NY, USA: Cornell University Water Resources and Marine Sciences Centre, SCARM. Floodplain management in Australia. Best practice principles and guidelines. Collingwood, Victoria: CSIRO Publishing, Agriculture and Resource Management Council of Australia and New Zealand, Standing Committee on Agriculture and Resource Management, Report No. 73. SMEC. Brisbane River flood investigations final report. Canberra: Snowy Mountain Engineering Corporation for Cities Commission, Smith D.I. Cost effectiveness of flood warnings. In: D.R. Smith & J.W. Handmer, eds. Flood warning in Australia: policies, institutions and technology. Canberra: Centre for Resource and Environmental Studies, 1986, Smith D.I. The worthwhileness of dam failure mitigation: an Australian example. Appl. Geogr. 1990, 10, Smith D.I. Flood damage estimation A review of urban stage-damage curves and loss functions. Water SA 1994, 20, (3), Smith D.I. & Greenaway M.A. The computer assessment of flood damages: ANUFLOOD. In: P.W. Newton, R. Sharpe & M.A.P. Taylor, eds. Desktop planning, advanced microcomputer applications for physical and social infrastructure planning. Melbourne: Hargreen, 1988, Smith D.I., Handmer J.W., McKay J.M., Switzer M.A.D. & Williams B.J. Non-structural measures for flood mitigation: current adoption in urban areas. Vol. 1, Report to the National Landcare Program, Department of Primary Industries and Energy by the Centre for Resource and Environmental Studies. Canberra: Australian National University, Thieken A.H., Muller M., Kreibich H. & Merz B. Flood damage and influencing factors: new insights from the August 2002 flood in Germany. Water Resour. Res. 2005, 41, USACE. Risk based analysis for flood damage reduction studies. Manual no Washington, DC, USA: U.S. Army Corps of Engineers, USACE. HEC-RAS river analysis system, hydraulic reference manual, Version 3.1. California, USA: Hydrologic Engineering Centre, United States Army Corps of Engineers, 2003a. USACE. HEC-RAS river analysis system, user manual, Version 3.1. California, USA: Hydrologic Engineering Centre, United States Army Corps of Engineers, 2003b. USBR. Downstream hazard classification guidelines, U.S Department of the Interior Bureau of Reclamation, ACER technical memorandum No. 11, Colorado, Denver, Zhai G., Fukuzono T. & Ikeda S. Modelling flood damage: case of Tokai flood J. Am. Water Resour. As. 2005, 41, (1),