Nat Hazards (2010) 53:43–61 DOI 10.1007/s11069-009-9409-9 ORIGINAL PAPER
Estimating probable maximum loss from a Cascadia tsunami Dale Dominey-Howes Æ Paula Dunbar Æ Jesse Varner Æ Maria Papathoma-Ko¨hle
Received: 5 May 2008 / Accepted: 25 May 2009 / Published online: 17 June 2009 Springer Science+Business Media B.V. 2009
Abstract The Cascadia margin is capable of generating large magnitude seismic-tsunami. We use a 1:500 year tsunami hazard flood layer produced during a probabilistic tsunami hazard assessment as the input to a pilot study of the vulnerability of residential and commercial buildings in Seaside, OR, USA. We map building exposure, apply the Papathoma Tsunami Vulnerability Assessment Model to calculate building vulnerability and estimate probable maximum loss (PML) associated with a 1:500 year tsunami flood. Almost US$0.5 billion worth of buildings would be inundated, 95% of single story residential and 23% of commercial buildings would be destroyed with PML’s exceeding US$116 million. These figures only represent a tiny fraction of the total values of exposed assets and loss that would be associated with a Cascadia tsunami impacting the NW Pacific coast. Not withstanding the various issues associated with our approach, this study represents the first time that PML’s have ever been calculated for a Cascadia type tsunami, and these results have serious implications for tsunami disaster risk management in the region. This method has the potential to be rolled out across the United States and elsewhere for estimating building vulnerability and loss to tsunami. Keywords Tsunami Cascadia Building vulnerability assessment Loss PTVA model D. Dominey-Howes (&) Australian Tsunami Research Center, School of Biological, Earth and Environmental Science, University of New South Wales, Sydney, NSW 2052, Australia e-mail:
[email protected] P. Dunbar National Geophysical Data Center, National Oceanic and Atmospheric Administration (NOAA), Boulder, CO 80303, USA J. Varner Cooperative Institute for Research in Environmental Sciences (CIRES), University of Colorado, Boulder, CO 80309, USA M. Papathoma-Ko¨hle Department of Geography and Regional Research, University of Vienna, 1010 Wien, Austria
123
44
Nat Hazards (2010) 53:43–61
1 Introduction and aims The Cascadia Subduction Zone (CSZ) (Fig. 1a) generates large (moment magnitude [ 8) ‘mega-thrust’ earthquakes similar to the 2004 Indian Ocean event (Satake and Atwater 2007). Coastal sediments in Oregon, Washington and British Columbia attest to the occurrence of at least eight Holocene earthquake-tsunami (Kelsey et al. 2005; Losey 2005; Nelson et al. 2006; Peters et al. 2007). Recurrence intervals along the CSZ range from 300 to 1,000 years with an average of 500–600 years (Peters et al. 2007). Having recognized the hazard, efforts have begun to estimate future tsunami occurrence using a probabilistic tsunami hazard assessment (PTHA) framework (The Tsunami Pilot Working Group 2006). Once estimates are available, it is desirable to move to the next stage of the risk management process by calculating likely probable maximum losses (PMLs) for particular events. This is important as PMLs are used to develop disaster preparedness and response plans, to establish appropriate mitigation efforts such as landuse zoning policies, to develop and apply building codes and regulations and to determine where appropriate, insurance premiums. To estimate PMLs, information is needed about the extent and severity of the hazard (in this case inundation distance and flow depth), asset exposure (e.g., buildings located within the expected flood zone), the vulnerability of those buildings and their market value (or replacement cost).
(a)
(b) NORTH AMERICAN PLATE
52
123.96 W 123.94 W
123.90 W
46.04 N
46.04 N
British Columbia
Gearhart
46.02 N
Va n
123.92 W
46.02 N
PACIFIC OCEAN
co
uv er I
Necanicum River
s Vancouver
46.00 N
46.00 N
Seaside
Seattle
Washington
45.98 N
45.98 N
ION
de F
UCT
uca
UBD
IA S
Rid
ge
CAD
CAS
48
E
Jua n
ZON
123.96 W
123.94 W
123.92 W
123.90 W
Portland
JUAN DE FUCA PLATE
44
Oregon PACIFIC PLATE
0
200
California
km
40 N
130
126
122 W
Fig. 1 a Map of the Cascadia Subduction Zone (including crustal plates) and NW USA (together with places referred to in the text), b Seaside study location, c PTHA 1:500 tsunami map
123
Nat Hazards (2010) 53:43–61
45
(c)
Fig. 1 continued
Wood (2002), Wood et al. (2002a, b) and Wood and Good (2004) developed a framework for identifying ‘relative vulnerability hotspots’—places that are exposed to hazard processes. As valuable as the framework is, they acknowledge that their approach is an ‘issues identification tool’ that does not provide a quantification of PML (Wood and
123
46
Nat Hazards (2010) 53:43–61
Good 2004; p. 265). They note communities interested in identifying and quantifying vulnerability to [tsunami] damage and loss must apply an objective scientific weighting scheme to the rankings of specific vulnerability attributes at the local level. Objective analysis of, for instance, residential building vulnerability, should be carried out by technical experts/engineers at high-resolution scales as assessment tools and data become available. Recent reports state that there is still a need for credible fragility models and laboratory data to understand the interaction of tsunami with the built environment (Bernard et al. 2007; Grundy et al. 2005). The Papathoma Tsunami Vulnerability Assessment (PTVA) model is a first generation loss estimation tool designed to achieve the quantification of vulnerability advocated by Wood and Good (2004) and Bernard et al. (2007). The PTVA model was developed using information from historic tsunami, postevent surveys and damage assessments (Papathoma 2003; Papathoma and Dominey-Howes 2003; Papathoma et al. 2003). Papathoma (2003) identified and ranked using expert judgement, a series of attributes (engineering, environmental, social, etc.) responsible for controlling the type and severity of tsunami damage to buildings. The 2004 Indian Ocean event enabled the model to be tested, validated and evaluated and provides a robust framework to estimate tsunami vulnerability of buildings (Dominey-Howes and Papathoma 2007). The PTVA model is dynamic. The attribute data contained within the primary (GIS) database may be continuously and quickly updated, allowing investigation of spatial and temporal vulnerability. Table 1 outlines selected classes/attributes within the model. In the absence of fully-developed and tested tsunami building fragility-damage assessment tools, the PTVA model provides a framework capable of generating high-resolution first order assessments of vulnerability and PML. We apply the PTVA model to Seaside (46000 0000 N, 123930 0000 W on the northwest coast of Oregon, 110 km northwest of Portland), OR (USA) (Fig. 1b), because (1) a 1:500 year PTHA tsunami flood map has recently been established (Fig. 1c) (The Tsunami Pilot Working Group 2006), (2) it can provide the detail advocated by Wood and Good (2004). The site for this study includes the city boundaries of Seaside and Gearhart (hereafter referred to as ‘Seaside’) (Fig. 1b). The aims of this study are to: 1. Map and quantify the ‘exposure’ of one-story residential and commercial buildings within the 1:500 year tsunami flood hazard zone in Seaside 2. Use the PTVA model to quantify the vulnerability of these structures; and 3. To provide a preliminary estimate of PMLs in 2006 US$ for the buildings in the 1:500 year tsunami flood zone
2 Methods In order to quantify the vulnerability of buildings in Seaside to the 1:500 year tsunami flood, we undertook the following: 2.1 Step 1: Identification of tsunami inundation zone based on a PTHA We selected the 1:500 year tsunami flood layer (Fig. 1c) generated by the PTHA assessment described in the Tsunami Pilot Working Group (2006). We use the 1:500 year layer
123
Nat Hazards (2010) 53:43–61
47
Table 1 Selected class/attribute fields within the PTVA model relevant to this study (after Dominey-Howes and Papathoma, 2007) Major class
Attribute
Attribute description
The built environment
Number of floors in each building
Only one floor
High
More than one floor
Low
Description of ground floor
Open plan with movable objects (e.g., tables and chairs)
High
Open plan or with big glass windows without movable objects
Medium
None of the above
Low
Building surroundings
Building material, age, design
Shape and orientation of building
The environment
Natural environment
Land cover (vegetation)
Vulnerability descriptor
No barrier
Very high
Low/narrow earth embankment
High
Low/narrow concrete wall
Medium
High concrete wall
Low
Buildings of field stone, unreinforced, crumbling and/or deserted, wood frame and wood construction
High
Ordinary masonry brick buildings, cement mortar, no reinforcement
Medium
Pre-cast concrete skeleton, reinforced concrete
Low
Square or oblong shaped structure
High
Non cubic shaped building (e.g., has hexagonal or circular shaped floor plan)
Low
Walls parallel to shoreline
High
Corners of building facing the shoreline
Low
Narrow intertidal zone
High
Intermediate intertidal zone
Medium
Wide intertidal zone
Low
No vegetation cover
High
Scrub and low vegetation, small trees
Medium
Trees and dense scrub
Low
The ‘‘vulnerability descriptors’’ in the last column are defined as ‘relative’ to one another and are based on previous field surveys
rather than the 1:100 year layer they developed because no buildings are located within the 1:100 year flood layer (thus exposure, vulnerability and PML are negligible) and because no other flood return periods (e.g., 1:1,000 year) were available. It should be noted that since the 1:500 year tsunami flood layer is based upon a probabilistic calculation, it does not exactly correlate with a specific event of a given magnitude. The Tsunami Pilot Working Group (2006) was unable to determine a ‘credible worst case scenario’. However, since the 1:500 year layer is dominated by a local CSZ tsunami (the Tsunami Pilot Study Working Group 2006), we examine vulnerability deterministically based upon the probabilistic 1:500 year flood layer.
123
48
Nat Hazards (2010) 53:43–61
2.2 Step 2: Identification (and manipulation) of data sets for use in the tsunami vulnerability assessment To quantify building vulnerability, we needed data about Table 1 attributes at a high resolution. We consulted the US multi-hazard loss estimation software (HAZUS-MH) database for Oregon. HAZUS-MH includes attributes potentially useful for vulnerability assessment and loss estimation and presents data in aggregate form at census block level. We identified HAZUS-MH attributes similar to the PTVA Model attributes including construction material, building condition, number of floors and building row number from the coast. Although HAZUS-MH contains some data for these attributes, their coarse aggregate nature meant that they are at a scale at odds with the very high resolution spatial analysis we wished to undertake. We therefore obtained the Clatsop County Tax Assessor GIS taxlot database, which includes Seaside (hereafter referred to as the ‘taxlot database’). The taxlot database contains multiple attributes for every taxlot (corresponding to individual buildings or land parcels). From this, attributes relevant to the PTVA model included the number of floors, type and use of structure, year built and market value of the improvement (building). The taxlot database contains 185 numerical classes (referred to as ‘‘Stat_Class’’) to classify building types and land use (Table 2). For example, there were 46 separate classes that identified 1- and 1.5-story residential structures (Table 2). In order to simplify the taxlot database, we reclassified the building and land parcel classes (Table 2, column 3). 2.3 Step 3: Combining the taxlot database with the 1:500 year inundation depth zones and generation of water depth above ground surface We obtained probabilistic tsunami water height raster layers produced by the Tsunami Pilot Study Working Group (2006) from the USGS. The set of rasters contained annual tsunami probability of exceedance for wave heights from 0.5 to 10.5 m above mean high water (MHW) (see Tsunami Pilot Study Working Group (2006) for a description of how this was achieved). The individual probabilistic wave height rasters were combined into one, which represented the maximum tsunami water height with a 500-year recurrence interval. This was achieved by summing the wave height for each grid where the annual probability of exceedance was 0.002 or greater. We then determined the maximum 1:500 water depth above the land surface. To do this, we subtracted the DEM developed in the Seaside Tsunami Pilot Study and calculated the water depth raster from the 500-year wave height one. The resulting raster layer provided 1:500 water depths for each pixel (with 0.000558 degree cell size or about 60 m). A water depth was then assigned to each taxlot in the taxlot database layer using the ArcGIS zonal statistics tool to compute the mean value of the water depth raster contained within each taxlot polygon. 2.4 Step 4: Collecting high resolution data relating to the PTVA model attributes and ground-truthing the HAZUS-MH and taxlot databases Two key PTVA model attributes (building material, condition) were not available in the taxlot database. However, the HAZUS-MH database provides summaries of the proportion of buildings constructed from different materials. For example, 99% of residential structures in Oregon are constructed of wood. Although we used satellite imagery, we were unable to determine the spatial distribution of these structures so we visited Seaside to
123
Nat Hazards (2010) 53:43–61
49
Table 2 Clatsop County Tax Assessor taxlot database classes, class descriptions and our simplified (revised) Stat_Class number Original Clatsop County Tax Assessor taxlot Stat_Class
Description
Our revised Stat_Class
110, 111, 112, 120, 121, 122, 123, 124, 125, 130, 131, 132, 133, 134, 135, 140, 141, 142, 143, 144, 145, 150, 151, 152, 153, 154, 155, 160, 161, 162, 163, 164, 165, 170, 171, 172, 173, 174, 175, 180, 181, 182, 183, 184, 185, 630
Residential 1 story*
100
116, 126, 127, 128, 129, 136, 137, 138, 139, 146, 147, 148, 149, 156, 157, 158, 159, 166, 167, 168, 169, 177, 178, 186, 188, 189, 656, 657, 646, 176, 179, 187, 637
Residential 2 ? story
120
420, 421
Residential apartments (low rise)
130
422, 423, 900
Residential apartments (medium and high rise, e.g., condos)
140
190, 181, 182, 183, 194, 199, 200, 300, 500
Residential single-wide manufactured homes (MFD), vacant MFD and miscellaneous outbuildings
150
232, 233, 234, 242, 243, 244, 252, 253, 254
Residential all plexes (duplex, triplex, etc.)
155
410, 411, 412, 413, 414
Commercial—motels, hotels B&B
160
440, 441, 442, 443, 444, 446, 447, 448, 449, 475, 490, 430, 431, 432, 433, 434, 435, 540
Commercial—retail, department stores, etc.
170
480, 481, 482, 483, 484, 485, 698
Commercial—automotive, service stations, etc.
180
400, 651, 460, 461, 462, 463, 464, 499
Commercial—other, misc., communications—all types, financial—banks, ATMs, etc.
190
470, 471, 472, 473, 474, 476, 477
Industrial—warehouse, storage, etc.
200
700, 710, 720, 730, 740, 750, 760, 770, 780, 790
Industrial—all others (e.g., manufacturing)
210
450, 452, 453, 454, 455, 496, 520, 530, 566, 570
Medical—hospitals, veterinary, day care, nursing home, etc.
230
000, xxx
Land—vacant land/ contiguous land holding
0, 240
564
Emergency service—fire, police, ambulance
250
493, 494, 495, 497, 492, 511
Recreation—sports, art venues, RV parks, etc.
280
445, 550, 560, 561, 562, 563, 564, 565, 567, 491
Public services—schools, post offices, churches, libraries, regional centers
290
* 1.5-story structures were houses with 1-room attics Note: In this table, we have listed every building class (e.g., 1 story, industrial, recreation etc) we examined in our study. However, in this paper we only report on the vulnerability and likely PMLs of 1- and 1.5* story residential (hereafter referred to as ‘1 story’) and commercial buildings
123
50
Nat Hazards (2010) 53:43–61
Table 3 Attributes from the PTVA model and their weighting within the study Attribute
Weighted value
Water depth above ground surface
8
Building row number (from the sea)
7
Building material
6
Number of floors
5
Orientation of building
4
Condition of building
3
Building surroundings
2
Land cover
1
These specific attributes were selected for this study because they are (1) known to be important in predicting/controlling vulnerability and (2) because data related to these attributes was readily available to us. The weighting of the attributes was determined by Papathoma (2003) via a process of ‘expert judgement’
ground truth the taxlot database. Selected ‘representative’ residential and commercial blocks located within the peninsula area of Seaside (the area of Fig. 1b located between the Pacific Ocean and the Necanicum River) were chosen and information related to eight key PTVA model attributes (Table 3) was recorded for every structure. The peninsula region was selected because it is the area affected by the deepest tsunami inundation (Fig. 1c) and is therefore, extraordinarily vulnerable. We used the ground-truthed data from 12 residential and 12 commercial blocks together with a rapid visual assessment of the remaining building stock in the peninsula area to generate a high resolution dataset of building attributes. 2.5 Step 5: Calculation of PTVA vulnerability scores for ground-truthed blocks We restrict our calculation of building vulnerability to the ground-truthed one-story residential and commercial buildings on the peninsula region of Seaside. The percentage distribution of these vulnerability scores was then scaled up to all one-story residential and commercial buildings within the peninsula area. For each attribute, descriptions of possible alternatives were provided, and these alternatives were given a raw score. The list of attributes, and thus raw scores, varied depending upon the particular locality being evaluated. These are shown in Table 4. To calculate the overall ‘‘vulnerability score’’ for each building, we: • Weighted the attributes (Table 3) from highest (those most significant in controlling vulnerability) to lowest (those least significant in controlling vulnerability). The weighting of the attributes was determined by expert judgement by Papathoma (2003) from an analysis of numerous tsunami damage assessment surveys and engineering reports • Transformed (or standardized) the raw scores for each attribute to a ratio scale to permit interattribute manipulation. Thus, we take the raw score and divide that number by the highest possible value for that attribute. For example, for water depth, the raw score may be 4 (indicating a water depth of between 2.0 and 2.99 m) and 4 is divided by 6 (which is the maximum possible score for this attribute ; see Table 4). So, 4/6 = 0.66. We then complete this calculation for each attribute
123
4
3
2
1#
2.0–2.99
1.0–1.99
0.5–0.9
0.01–0.49
1#
2
3
2*
2 or 1# more
1 story
Cubic/ parallel
Cubic with corner to sea
Irregular
1#
2
Excellent
Fair
Poor
1#
2
3*
High concrete wall
Medium brick/ concrete wall
Low brick wall
None
3*
Building surroundings (g)
4*
Condition of building (f) Raw Condition Raw Surroundings score score
Orientation of building (e)
Raw Floor Raw Orientation score number score
Number of floors (d)
1#
2
3
4*
2
3*
Raw score
Large trees 1#
Bushes and low trees
None
Raw Cover score
Land cover (h)
Note: Each of the attributes investigated within the PTVA model contributes to the ‘overall’ vulnerability of an individual building structure. Since the attributes are all very different, they have to be reduced to a common numeric value to enable the calculation of an overall vulnerability score. The overall vulnerability score of each building structure is a summation of the relative numeric raw score for each attribute. The raw score of each attribute will vary, but in every case ranges from a ‘high vulnerability’ (which in Table 4 is indicated by the *) and a ‘low vulnerability’ (which is indicated by the #). See text for an explanation of the calculation of the final ‘vulnerability’ score
Steel and concrete
Concrete
37#
1#
Stone and concrete; concrete and brick
Down to…
Wood only; wood on concrete base; steel frame with wood walls
5
37*
3.0–3.99
1
6*
C4.0
Building material (c)
Raw Row Raw Material score number score
Building row number (from the sea) (b)
Depth
Water depth above ground surface (m) (a)
Table 4 Attributes, their descriptions and the ‘raw score’ for each description
Nat Hazards (2010) 53:43–61 51
123
52
Nat Hazards (2010) 53:43–61
Table 5 Range of vulnerability classes and their respective numeric distributions
Numeric values
Vulnerability class
30.53–36
High vulnerability
25.031–30.529
Medium-high vulnerability
19.56–25.03
Medium vulnerability
14.08–19.55
Medium-low vulnerability
8.609–14.079
Low vulnerability
To calculate the vulnerability score (V) of each individual building, we then sum as follows: V ¼ ð8 aÞ þ ð7 bÞ þ ð6 cÞ þ ð5 d Þ þ ð4 eÞ þ ð3 f Þ þ ð2 gÞ þ ð1 hÞ where a, b, c, d, e, f, g and h are the standardized scores for water depth, row number, material, floors, orientation, condition, surroundings and land cover, respectively (Table 4, row 1). The lowest possible vulnerability score is: Rð8 0:16Þ þ ð7 0:027Þ þ ð6 0:25Þ þ ð5 0:5Þ þ ð4 0:33Þ þ ð3 0:33Þ þ ð2 0:25Þ þ ð1 0:33Þ ¼ 8:609 The highest possible vulnerability score is: Rð8 1Þ þ ð7 1Þ þ ð6 1Þ þ ð5 1Þ þ ð4 1Þ þ ð3 1Þ þ ð2 1Þ þ ð1 1Þ ¼ 36 The ‘vulnerability score’ of each building will lie between these values. Since we now know that the building vulnerability score will lie between 8.609 and 36, we divide this range into ‘equal intervals’ (Table 5). 2.6 Step 6: Extrapolation from ground-truthed blocks to whole of Seaside peninsula and calculation of PML’s Once we had estimated the vulnerability of buildings in our ground-truthed blocks, we took the relative distribution of ‘high vulnerability’ and ‘medium vulnerability’ (etc.) structures and scaled these numbers up to the total number of buildings present on the whole peninsula. Finally, we calculated the PMLs (both absolute number of buildings and their $ value) for one-story residential and commercial buildings.
3 Results There are a total of 4,910 buildings (of different classes) and land plots within Seaside (Table 6, column 2). Here, we focus on the vulnerability of one-story residential and commercial buildings only since collectively they represent over 50% of the building stock, and we only wish to ‘demonstrate’ the capability of our approach (rather than provide a comprehensive assessment of building vulnerability and PML). 3.1 Summary of building exposure within the 1:500 year tsunami flood layer, spatial distribution and replacement costs A total of 3,032 buildings (62% of the total) are located within the 1:500 year tsunami flood hazard layer. Their spatial distribution is shown in Fig. 2a, and the total number of
123
Nat Hazards (2010) 53:43–61
53
Table 6 Number of buildings and land plots (by class), exposure and replacement cost in the tsunami inundation zone Building class Total number Total number of buildings of buildings and plots in and plots 1:500 year tsunami flood
Total $ value of buildings in the 1:500 year tsunami flood
Total number of buildings in tsunami water [4 m deep
Percentage of building stock in tsunami water [4 m deep
Replacement costs ($) of buildings destroyed by tsunami water [4 m deep
1 story residential
2,949
1,874
189,954,374
258
13.76
31,159,699
2 ? story residential
1,221
577
105,849,803
173
29.98
36,495,884
181 Plexes (duplex, triplex, etc)
133
22,697,035
24
18.04
4,712,938
Commercial
417
355
85,845,670
38
10.70
14,606,689
Industrial
55
29
4,683,521
6
20.69
162,084
55 Recreation, public service and regional centers
42
59,005,132
3
7.14
283,389
Medical and emergency services
32
22
6,530,069
3
13.64
392,364
Vacant land
439a
Total
4,910
3,032
$474,565,604
505
–
$87,813,047
a
These vacant land parcels have no structures built on them. However, the Clatsop County Assessor Database includes these land plots as ‘rateable’ areas. Thus the total number of ‘building plots’ equals 4,910
buildings (by class) is summarized in Table 6 (column 3). Figure 2b and c shows the spatial distribution of the one-story residential and commercial structures (by water depth), respectively. The total market value of all buildings exposed to inundation within the 1:500 year tsunami flood hazard layer is almost US$0.5 billion (Table 6, column 4). 3.2 Summary of exposure of buildings under tsunami flood water [4 m deep, their spatial distribution and replacement costs Figures 1c and 2 indicate that some of the peninsula would be under tsunami flood water [4 m deep. Table 6 (columns 5 and 6) summarizes the total number of buildings (by class) and the percentage of the total building stock that would be submerged under tsunami water [4 m deep. The replacement costs for those buildings are summarized in Table 6 (column 7). Severe structural damage is known to have occurred to buildings in flood waters of significantly lesser depth [e.g., 2004 Indian Ocean Tsunami (Dominey-Howes and Papathoma 2007) and 1996 Irian Jaya Tsunami (Matsutomi et al. 2001)]. However, we assume that any structure submerged in tsunami water[4 m deep will be damaged beyond repair. The total replacement cost in this case would exceed US$87 million (or 18.5% of the value of all building stock) (Table 6, column 7).
123
Fig. 2 a Exposure of all (total = 3,032) buildings and land plots (all classes) by water depth in the 1:500 year tsunami flood hazard layer. Total market value exceeds $474 million; b distribution (exposure) of all (total = 1,874) one-story residential buildings in the 1:500 year tsunami flood hazard layer. Total market value exceeds $189 million; c distribution (exposure) of all (total = 355) commercial buildings in the 1:500 year tsunami flood hazard layer. Total market value exceeds $85 million
54
123
Nat Hazards (2010) 53:43–61
Nat Hazards (2010) 53:43–61
55
3.3 Results of building tsunami vulnerability assessment Figure 3 shows the spatial distribution of the 1,017 exposed one-story residential and 199 commercial buildings in the peninsula region. We ground-truthed 131 (or 12.88%) of the one-story residential and 98 (or 49.25%) of the commercial structures. 3.3.1 Residential building tsunami vulnerability assessment Of the 131 ground-truthed one-story residential structures (Fig. 4a, b), 14.5% have a vulnerability classification of ‘high’ and 81.7% are ‘medium-high’. The remaining classifications are shown in Table 7 (columns 1 and 3). These relative percentages were then scaled up for the entire peninsula region. We assume that a total loss of structure occurs for buildings classified ‘medium-high’ and ‘high’ (which in the vast majority of cases correlated with water depth [4 m). Therefore, the PML is the aggregate replacement costs of buildings that fall within these classifications. This replacement cost, calculated from the 2006 market value of the residential buildings, amounts to more than US$103 million (Table 7, column 5). 3.3.2 Commercial building tsunami vulnerability assessment Of the 98 ground-truthed commercial structures (Fig. 4c), none have a vulnerability classification of ‘high’, and only 23.5% have a classification of ‘medium-high’. The remaining classifications are shown in Table 7 (columns 1 and 3). These percentages were also scaled up for the entire peninsula region, and it was again assumed that a total loss occurs for buildings classified ‘medium-high’ and ‘high’. Consequently, the PML for these structures exceeds US$13 million (Table 7, column 5).
4 Discussion Wood and Good (2004) provide a framework to begin to examine tsunami vulnerability at the local level. We go well beyond Wood and Good (2004) using a specific technical tool to identify and quantify one-story residential and commercial building vulnerability and to estimate PMLs—something that to our knowledge, has never been done for tsunami in the United States. We attempted to use data from the HAZUS-MH database to estimate PMLs from the 1:500 year tsunami. However, HAZUS-MH data are at present, too coarse to be used for this type of detailed tsunami vulnerability assessment. The County Tax Assessor Taxlot Database on the other hand, is useful, although we are uncertain if other similar databases are as accurate. This issue would need to be investigated for each community subjected to similar detailed vulnerability assessment. There are a large number of buildings (3,032) with a high market value within the 1:500 year tsunami flood layer, and they are densely concentrated on the low-lying peninsula (Fig. 2a). This is a significant issue during tsunami inundation since the peninsula is connected to the mainland by six small bridges with one lane of traffic flow in each direction. In the event of an evacuation, significant bottlenecks would be expected to develop at these critical points. Some 1,874 single-story residential structures with a combined value of more than US$189 million are located within the 1:500 year tsunami flood layer. However, exposure
123
Fig. 3 a Distribution (exposure) of all (total = 1,017) one-story residential building structures in the 1:500 year tsunami flood hazard layer in the peninsula region only. PML of building structures in water depth [4 m exceeds $18 million; b distribution (exposure) of all (total = 199) commercial buildings in the 1:500 year tsunami flood hazard layer in the peninsula region only. PML of buildings in water depth [4 m exceeds $7 million
56
123
Nat Hazards (2010) 53:43–61
Fig. 4 a (Upper) Distribution (exposure) of ground-truthed one-story residential buildings in the 1:500 year tsunami flood hazard layer in the northern peninsula region only. (Lower) Actual ‘vulnerability’ of those structures according to the PTVA Model; b (Upper) Distribution (exposure) of ground-truthed one-story residential buildings in the 1:500 year tsunami flood hazard layer in the southern peninsula region only. (Lower) Actual ‘vulnerability’ of those structures according to the PTVA model; c (Upper) Distribution (exposure) of ground-truthed commercial buildings in the 1:500 year tsunami flood hazard layer in the peninsula region only. (Lower) Actual ‘vulnerability’ of those structures according to the PTVA model
Nat Hazards (2010) 53:43–61 57
123
58
Nat Hazards (2010) 53:43–61
Table 7 Building vulnerability and PML of one-story residential and commercial buildings located on the peninsula PTVA vulnerability class
Number of ground-truthed structures
Percentage of building stock
Scaled-up market value $
Probable maximum loss $
Residential buildings High Medium-high Medium
19
14.5
15,568,512
107
81.7
87,675,303
3
2.3
2,458,186
Medium-low
0
Low
2
Total
131
– 1.5 100
– 1,638,791 107,340,792
15,568,512 87,675,303 –a –a –a 103,243,815
Commercial buildings High
–
–
–
Medium-high
23
23.5
13,431,515
13,431,515
Medium
51
52.0
29,782,925
–a
5
5.1
2,919,895
–a
Medium-low
0
Low
19
Total
98
19.4 100
11,095,600
–a
57,229,935
13,431,515
Total PML for one-story residential and commercial buildings in the 1:500 year tsunami
116,675,330
a
While buildings with a vulnerability classification of ‘Medium’, ‘Medium-low’ and ‘Low’ will be damaged by the 1:500 year tsunami flood, we assume that they will not be damaged beyond repair. Therefore, we do not include the likely replacement costs for these structures since we are only concerned with the replacement costs of structures that are completely destroyed by the tsunami
to tsunami flooding does not automatically mean that these structures will be damaged beyond repair. If we assume that only those residential structures affected by tsunami water depths [4 m will be completely destroyed, then the replacement cost of these 258 homes (14% of the single story residential building stock) alone exceeds US$31 million (Table 6, column 7). Three hundred and fifty-five commercial buildings (out of 417) with a combined value of US$85 million are located within the 1:500 year tsunami flood layer. However, only 38 of these (10.7% of the stock) are located in areas where tsunami water depth will be[4 m. Assuming that all these buildings would be destroyed, their replacement costs would exceed US$14 million (Table 6, column 7). Our assumption about water depth and total building loss needs to be explored in greater detail however. The vulnerability/resilience of an individual structure is a function of a number of parameters that include structural engineering, age, condition, material, distance from the coast, engineering standards, orientation and shape. The PTVA model attempts to account for these factors and ascribes a vulnerability score to each structure. Thus, two adjacent building structures both in 4 m of tsunami flood water, may sustain varying degrees of damage according to these factors. Therefore, when the PTVA model is applied to buildings in Seaside, the patterns shown in Figs. 2 and 3 vary, and the PMLs change dramatically. In our analyses, it is assumed that buildings categorized as ‘high’ or ‘medium-high’ vulnerability will be completely destroyed. Consequently, over 95% (Table 7, column 3) of one-story residential buildings in Seaside would be destroyed with a PML in excess of
123
Nat Hazards (2010) 53:43–61
59
US$103 million. When the same analysis is repeated for commercial buildings, some 23% (Table 7, column 3) would be destroyed with a PML of approximately US$13 million. Therefore, total PML for these two building classes would exceed US$116 million (Table 7, column 5) out of a total exposure worth close to US$0.5 billion. The PML of US$116 million is the minimum that would be expected at Seaside for a 1:500 year tsunami. The real figure would be much higher as other classes of buildings would also be destroyed. Also, our calculations have only considered the costs associated with the full replacement of destroyed buildings. It does not include the costs associated with the repair of partly-damaged buildings. The PTVA model is a useful tool for quantifying the vulnerability of buildings enabling estimates of PML and could easily be ‘rolled-out’ across the United States to help gain a high-resolution first-order assessment of exposure, vulnerability and PML. This loss estimation tool should be useful to emergency management and local government officials in prioritizing disaster mitigation efforts. As fragility curves become available, individual fragility functions could be added as an attribute field within the PTVA model. This pilot study was designed to test the capability of the PTVA model rather than provide a comprehensive assessment of building vulnerability and PML for Seaside (and the Pacific NW coast). There are several limitations to this work: • We have deterministically quantified exposure, vulnerability and PML based upon a probabilistic map that does not directly equate to an actual event. Use of a ‘credible worst case scenario’ would increase our confidence in the estimates of exposure and PML. When a ‘credible worst case scenario’ is developed, this study should be repeated to gain a more comprehensive estimate of PMLs • PMLs associated with the 1:500 year tsunami do not (in our study) take account of earthquake-related damage to structures prior to the arrival of the tsunami. Given that the 1:500 year tsunami flood layer is dominated by a Cascadia type earthquake, such an event will likely have caused significant structural damage to buildings that would then be affected by a tsunami. As such, PMLs are likely to be higher than reported here. Further work should try to incorporate this issue • Ground truthing was limited to a few blocks on the peninsula. Increased coverage of blocks on both the peninsula and mainland regions of Seaside would increase our confidence in the representativeness of the building stock surveyed (even if it did not change the distribution of building types/vulnerabilities) • We have only included the peninsula region in our analysis. Further analysis should focus on the mainland. Additionally, we simplified tsunami inundation to a single wave running across the region parallel with the shoreline. In reality, the wave would funnel up the Necanicum River estuary (Fig. 1b, c). Inundation of buildings adjacent to the estuary could occur before water from the coast reaches them. This might alter the relative vulnerability (and PMLs) at these locations • The County Tax Assessor Taxlot Database is useful, although we are uncertain if other databases are as accurate. This would need checking ‘on the ground’ for each community and could potentially limit the usefulness of the PTVA model; and • We have made no attempt to estimate human vulnerability, which is important for a complete tsunami vulnerability assessment. Databases containing business information such as number of hotel beds, employees, etc, could also be used to determine the variability in human exposure and vulnerability
123
60
Nat Hazards (2010) 53:43–61
5 Conclusion It is now known that the Cascadia margin is capable of generating large magnitude seismic-tsunami. Consequently, there is an urgent need to identify and apply quantitative tools to assess the relative vulnerability of buildings located within expected tsunami inundation zones and to use the assessments of vulnerability to estimate PML for future possible events. The results of such analyses may then be used for developing appropriate mitigation strategies. We have used a 1:500 year tsunami hazard flood layer produced during a PTHA as the input to a study of the vulnerability of one-story residential and commercial buildings in the low-lying peninsula region of Seaside, OR, USA. We have mapped building exposure, applied the ‘‘PTVA model’’ to calculate their vulnerability and estimate PMLs associated with a 1:500 year tsunami flood. Almost US$0.5 billion worth of buildings would be inundated, 95% of single-story residential and 23% of commercial buildings would be destroyed with PMLs exceeding US$116 million. This study represents the first time that PMLs have ever been calculated for a Cascadia type tsunami, and the results have serious implications for risk management. It is clear from this pilot study that the PTVA model may be a useful tool for investigating high resolution vulnerability of structures and for estimating PMLs within United States coastal tsunami inundation zones and could be applied elsewhere. Acknowledgments NOAA/NGDC is thanked for providing the resources to enable Dale Dominey-Howes and Papathoma-Ko¨hle to participate in this study. Two anonymous referees and Filippo Dall’Osso are thanked for making numerous helpful comments on an earlier version of this manuscript all of which significantly improved the final version.
References Bernard B, Dengler L, Yim S (2007) National Tsunami Research Plan: report of a workshop sponsored by NSF/NOAA. NOAA Technical Memorandum OAR PMEL-133, 135 p. http://nctr.pmel.noaa.gov/ pubs.html Dominey-Howes D, Papathoma M (2007) Validating a tsunami vulnerability assessment model (the PTVA Model) using field data from the 2004 Indian Ocean tsunami. Nat Hazards 40:113–136. doi: 10.1007/s11069-006-0007-9 Grundy P, Thurairaja A, Walker G (2005) Some reflections on the structural engineering aspects of tsunami. In: Lam NTK, Wilson JL, Butler B (eds) Proceedings, Australian Earthquake Engineering Society, Albury, New South Wales, p 7-1-7-4 Kelsey H, Nelson A, Hemphill-Haley E, Witter R (2005) Tsunami history of an Oregon coastal lake reveals a 4600 year record of great earthquakes on the Cascadia subduction zone. Bull Geol Soc Am 117 (7–8):1009–1032. doi:10.1130/B25452.1 Losey R (2005) Earthquakes and tsunami as elements of environmental disturbance on the Northwest coast of North America. J Anthropol Archaeol 24:101–116. doi:10.1016/j.jaa.2005.02.001 Matsutomi H, Shuto N, Imamura F, Takahashi T (2001) Field survey of the 1996 Irian Jaya earthquake tsunami in Biak island. Nat Hazards 24:199–212. doi:10.1023/A:1012042222880 Nelson A, Kelsey H, Witter R (2006) Great earthquakes of variable magnitude at the Cascadia subduction zone. Quat Res 65:354–365. doi:10.1016/j.yqres.2006.02.009 Papathoma M (2003) Tsunami vulnerability assessment using a geographic information system with special reference to Greece. Unpublished PhD Thesis, Coventry University, UK, 289 p Papathoma M, Dominey-Howes D (2003) Tsunami vulnerability assessment and its implications for coastal hazard analysis and disaster management planning, Gulf of Corinth, Greece. Nat Hazards Earth Syst Sci 3:733–744 Papathoma M, Dominey-Howes D, Zong Y, Smith D (2003) Assessing tsunami vulnerability, an example from Herakleio, Crete. Nat Hazards Earth Syst Sci 3:377–389
123
Nat Hazards (2010) 53:43–61
61
Peters R, Jaffe B, Gelfenbaum G (2007) Distribution and sedimentary characteristics of tsunami deposits along the Cascadia margin of western North America. Sediment Geol 200:372–386. doi:10.1016/ j.sedgeo.2007.01.015 Satake K, Atwater B (2007) Long-term perspectives on giant earthquakes and tsunamis at subduction zones. Annu Rev Earth Planet Sci 35:349–374. doi:10.1146/annurev.earth.35.031306.140302 The Tsunami Pilot Study Working Group (2006) Seaside, oregon tsunami pilot study—modernization of FEMA Flood Hazard Maps, Joint NOAA/USGS/FEMA special report, 94 p, 7 appendices. http://pubs. usgs.gov/of/2006/1234/ Wood N (2002) Vulnerability of port and harbor communities to earthquake and tsunami hazards in the Pacific Northwest. Unpublished PhD dissertation, Department of Geosciences, Oregon State University Wood N, Good J (2004) Vulnerability of port and harbor communities to earthquake and tsunami hazards: the use of GIS in community hazard planning. Coast Manage 32:243–269. doi:10.1080/0892075 0490448622 Wood N, Good J, Goodwin R (2002a) Reducing vulnerability of ports and harbors to earthquake and tsunami hazards. In: Proceedings, solutions to coastal disasters, American Society of Civil Engineers, Reston, Virginia, pp 949–963 Wood N, Good J, Goodwin R (2002b) Vulnerability assessment of a port and harbor community to earthquake and tsunami hazards: integrating technical and expert and stakeholder input. Nat Hazards Rev 3(4):148–157. doi:10.1061/(ASCE)1527-6988(2002)3:4(148)
123
