Integrating long term seismic risk chang

Nat Hazards DOI 10.1007/s11069-013-0714-y ORIGINAL PAPER Integrating long-term seismic risk changes into improving emergency response and land-use planning: a case study for the Hsinchu City, Taiwan Hung-Chih Hung • Ming-Chin Ho • Yi-Jie Chen • Chang-Yi Chian Su-Ying Chen • Received: February 2013 / Accepted: May 2013 Ó Springer Science+Business Media Dordrecht 2013 Abstract An increasing number of rapidly growing urban areas in Asia are becoming more vulnerable to seismic hazards in their development process However, local authorities rarely integrate seismic risk into the procedure of emergency and land-use planning This article explores the question of whether seismic risks for urban areas are increasing or diminishing over time, while trends such as population growth and land development in hazard-prone areas increase the potential for loss in disasters The net effects of such urbanization factors are examined through the use of simulation models that estimate building inventory and seismic loss changes Seismic losses are modeled for a comparative analysis under the same hypothetical earthquake events hitting at different points in a city area’s long-term development A case study of seismic risk assessments is illustrated by the Hsinchu City, Taiwan Results of a prospective analysis indicate that, for the same seismic events, overall risk is expected to increase due to a forecast 2.9 % growth H.-C Hung (&) Center for Land and Environmental Planning, National Taipei University, 151, University Road, San Shia, New Taipei City 23741, Taiwan e-mail: hung@mail.ntpu.edu.tw M.-C Ho Architecture and Building Research Institute, 13th Floor, 200, Section Beisin Road, Sindian, New Taipei City 231, Taiwan e-mail: ho@abri.gov.tw Y.-J Chen Á C.-Y Chian Department of Real Estate and Built Environment, National Taipei University, 151, University Road, San Shia, New Taipei City 23741, Taiwan e-mail: ejay0701@yahoo.com.tw C.-Y Chian e-mail: changyi@mail.ntpu.edu.tw S.-Y Chen National Science and Technology Center for Disaster Reduction, 9F, 200, Section Beisin Road, Sindian, New Taipei City 23143, Taiwan e-mail: suin@ncdr.nat.gov.tw 123 Nat Hazards in building inventory This increment in loss is largely attributed to a large amount of initial buildings predicted to be developed into commercial and industrial uses However, the spatial pattern of risk would change slightly; particularly, the southeastern, eastern, and some older core areas would be the most vulnerable and risky both at current and future time periods The approach here enables city planners to incorporate seismic risk analysis into predisaster emergency and land-use planning to encourage risk-reduction strategies Keywords Seismic risk Á Land-use planning Á Vulnerability Á Emergency management Á Urban dynamics Introduction Several studies recognize the importance of reducing seismic risk through predisaster preparedness and mitigation measures rather than postdisaster relief when formulating general city development policies (Burby et al 2000; Comerio 2004; Tate et al 2010) Predisaster risk mitigation measures can range from building code standards to emergency response and land-use planning (Nelson and French 2002; Lindell et al 2006; Schwab et al 2007) In preparing such measures, an impending need for decision-makers is to evaluate the possible consequences and losses caused by earthquakes for current land uses and their changes over time Traditionally, local authorities in Taiwan and other developing countries seldom took into account seismic risks and costs involved in the procedure of local or city development decisions (Sengezer and Koc¸ 2005; Hung and Chen 2007; Erdik and Durukal 2008; Hung 2009), because they lacked instruments and financial resources to implement risk management policy to tackle seismic impacts (Sengezer and Koc¸ 2005; Hung and Chen 2007) This leads to urban areas becoming more vulnerable and riskier to disasters due to development continuing to encroach upon hazard-prone areas such as floodplains, coastal regions, and seismic fault zones (NRC 2006; Nirupama and Simonovic 2007; Berke et al 2009) Advances in technology have increased the applicability of geographic information system (GIS)-based methodology to identify seismic hazards and create vulnerability and risk maps that can enhance policy-makers’ understanding of the factors resulting in losses and their spatial variability (Olshansky and Wu 2001; Sarris et al 2010; Liu et al 2013) However, few studies have applied GIS-based approach to explore how the dynamics of numerous factors of urban growth affect overall trends in urban risks and how to use them in planning process A typical output of risk assessment initiatives is an atlas depicting the distribution of seismic hazard or vulnerability based on present-day or ad hoc land-use scenarios This approach fails to characterize risk in several critical ways First, the geographic distribution and extent of a given hazard are portrayed through maps, but the spatial variation in loss or risk changes over time within hazard areas generally is not Second, most of the impacts focus on vulnerability of current built environment rather than on its changes over time This skews mitigation measures in the direction of current conditions A more inclusive analysis might suggest approaches for the long-term conditions such as considering resilient types of infrastructure and improving emergency response capacities, which could connect emergency management with land-use planning in a more effective way 123 Nat Hazards Third, the current process of risk assessment often overlooks the links between urban development factors and risk (or vulnerability) changes This limits the application of risk assessment for disaster management activities A screening tool should be useful for the assessment of which sectors, regions, or communities are most vulnerable or risky, and suggest the factors that cause or exacerbate vulnerability and risk over time This article proposes a novel methodology to help overcome the above-mentioned shortcomings in the existing seismic risk assessments and applications To illustrate the methodology, a case study is demonstrated by the assessments of seismic risk for the Hsinchu City, Taiwan The ways in which seismic risks are measured, incorporated a building inventory dynamic modeling into a GIS-based earthquake-loss estimation framework, provide a new approach to conceptualize and map seismic risks and their changes over time Finally, we discuss how the findings provide direction for more effective approaches to risk mitigation, emergency response, and land-use planning Urban development and seismic risk Researchers generally agree that high-quality land-use planning is a useful tool for reducing risk and improving emergency response in natural disasters (Burby et al 2000; Olshansky 2001; Amini Hosseini et al 2009) In preparing such plans, in Taiwan, for example, the Disaster Prevention and Protection Act (DPPA) and the Urban Planning Act attempt to develop technical tools to assess hazard and risk for local communities, based on creation of hazard maps, with clear implications for prioritizing mitigation measures for the areas of greatest hazard or risk However, local authorities acquire little guides on how to create such maps and how to use them in the planning process A number of GIS-based approaches related to seismic risk have been developed, some of which are dealing with the identification and mapping of hazard or vulnerability (Tyagunov et al 2006; Nath et al 2008; Sarris et al 2010), the application to emergency support system, urban planning, or construction of seismic risk mitigation database (Martelli et al 2007; Inel et al 2008) There are also several GIS-based tools, such as the US Federal Emergency Management Agency’s (FEMA) HAZUS-MH, United Nation’s Risk Assessment Tools for Diagnosis of Urban Area against Seismic Disasters, and Taiwan’s TELES (Taiwan Earthquake Loss Estimation System), which were designed to estimate the hazards and losses to the built environment from earthquake and intended for application by local planners and risk managers (Al-Momani and Harrald 2003; Schneider and Schauer 2006; Yeh et al 2006) The research efforts regarding seismic risk analysis have been altered from comparison conditions between countries toward more focused on the region- or city-scale analyses (Yong et al 2002; Barbat et al 2010; Sarris et al 2010; Tate et al 2010;) Particularly, the efforts to incorporate land-use, demographic, socioeconomic, and resilient factors into seismic hazard analysis have been increasingly emphasized by disaster risk managers (Cutter et al 2003; Carren˜o et al 2007; Chang et al 2012) This thread implies that seismic risk assessment should inform decision-makers not only about the spatial distribution of vulnerable or risky areas, but also about the possible changes of their distribution, and the major factors contributing to these changes and impacts on seismic risks A few studies have focused on the discussion of built environment changes and urbanization over time, as well as their impacts on disaster risks For example, Davidson and Rivera (2003) and Jain et al (2005) modeled the changes of building inventory over time and forecasted how those changes affected hurricane risk at the county level 123 Nat Hazards Olshansky and Wu (2001) and Hung and Chen (2007) used a comparative static approach to model seismic risk for various land-use plans Chang et al (2012) developed simulation models to explore and compare casualties and transportation risk of earthquake at different points in time Generally, previous seismic risk analyses included the application of hazard, vulnerability or risk map for hazard identification, compilation of monetary losses as a proxy of built environment impacts, and examination at the national or regional levels However, existing approaches rarely communicate vulnerability and risk changes associated with urban growth to city planners The methodology applied here is built on the existing efforts through the combination of a multidisciplinary seismic risk model with a GIS framework The overall goal is to generate descriptive maps that display the long-term spatial variation in seismic risk associated with building stock changes over time, as well as to examine the urbanization factors that vary the distribution of seismic risk Methods 3.1 Seismic risk model Seismic risk consists of the interaction between hazard and vulnerability (UNISDR 2005; NRC 2006; Hung and Chen 2007) Hazard is expressed as the likelihood of an event with a specific intensity, place, and period of time that could cause losses or casualties Vulnerability, indicating propensity to experience loss should the hazard event occur, includes exposure and sensitivity components (Chang et al 2012) The vulnerability of an area reflects land uses, and physical and socioeconomic factors that influence loss in a disaster and ability to cope Hence, the simplest expression of seismic risk is: Risk ¼ Hazard  Vulnerability ð1Þ In this study, we simply emphasized vulnerability changes rather than hazard or physical damage, concern of current as well as future trends, and application to a rapidly growing, medium-sized, city area The focus is especially on building disruption to illustrate different facets of land-use and community loss in earthquake, which is important to consider when planning for earthquakes A seismic risk model was developed to project buildings disruption and loss for the purpose of an intertemporal analysis This model addresses how building damages and losses would be influenced by urban growth and building stock changes The model consists of three sequential submodels which estimate building inventory changes, building damages, and direct economic losses, respectively (Fig 1) This framework is very similar to that of well-established models such as the loss estimation model for earthquake in HASUS-MH (Kircher et al 2006) 3.2 Building inventory change model A prospective analysis was concentrated on modeling building disruption-specifically, building damage and loss in earthquakes The likelihood of seismic events is expected to remain constant over the study period, while the changes in seismic vulnerability and risk are anticipated due to continued urbanization The requirement is to develop a building inventory change model (BICM) to quantify how and why the number, location, and use type of building change in a city over time However, existing studies on urban dynamic 123 Nat Hazards Population by building occupancy types Building Inventory Change Model Land uses Demography Socioeconomics Location Population by building structural types Damage Model Hypothetical earthquake scenarios Population by building damage states Direct Economic Loss Model Loss estimates Ground shaking intensity measures Legend Analysis Data Scenario Fig Conceptual diagram of the seismic risk model models are rarely focused on predicting and characterizing such change process in the building inventory (Wegener 1994; Verburg et al 2004) The quantity, density, and spatial distribution of building stocks and their changes are accompanied by several urbanization characteristics, such as demographic growth, socioeconomic changes, location, expansion of built-up areas, and land-use types These urbanization characteristics have a major effect upon the growth and changes in building inventory, while also influencing the distribution of seismic risk (Chang et al 2012) A new BICM was built to project the building inventory growth The probability of building use in area i changing from type j to k (Pri[(.)jk]) can be assumed as a probabilistic function of the urbanization characteristics: Pri ẵbuilding inventory changeịjk ẳ f land usesi ; demographyi ; socioeconomic factorsi ; locationi Þ ð2Þ The BICM was further specified as a generalized form of multinomial logit model in order to incorporate GIS into the analysis and to fit the format needs of TELES input: expbl0 ỵ bl1 xi1 ỵ bl2 xi2 ỵ ỵ blm xim ị Pri ẵjjk ẳ PN l l l l iẳ1 expb0 ỵ b1 xi1 ỵ b2 xi2 ỵ ỵ bm xim ị 3ị where Pri[j|k] is the probability that area i is developed from building use type j to k; xim is explanatory or independent variable; bm is the logit coefficient, and N is the full set of building inventory changes This model is analogous to the land-use change model developed by Landis and Zhang (1998a, b), which has been widely applied in forecasting urban growth and land-use changes 3.2.1 Land uses The explanatory variables involved in the BICM were identified from literature and the characteristics of urban growth that would influence building inventory growth and changes (Landis and Zhang 1998b; Verburg et al 2004) The first set of explanatory variables is land-use characteristics, which refer to initial land-use types: residential, 123 Nat Hazards commercial (e.g., retail, service and office), industrial, and other uses (i.e., public and educational uses) Urban economists suggest that ‘higher-order’ land uses (e.g., commercial and residential uses) are capable of generating higher land rents (Mills and Hamilton 1997) This implies that the previously developed ‘higher-order’ uses should, all else being equal, be less likely to be redeveloped into ‘lower-order’ uses (e.g., industrial use) By the same logic, the ‘lower-order’ uses should be more likely to be redeveloped into ‘higher-order’ uses The land-use data (1995 and 2007) used in the model were collected from the database of the Taiwan Land Use Investigation 3.2.2 Demographic and socioeconomic factors Demographic and socioeconomic factors would determine the levels of demand for various types of building development The variables include (1) initial number of households and employment (1995), (2) rate of household growth during 1995–2007, (3) initial number of jobs (1995), (4) rate of job growth during 1995–2007, and (5) ratio of jobs to households (1995), assuming that population (or household) and employment growth would cause higher probability for the conversion of ‘lower-order’ uses to ‘higher-order’ uses Additionally, the agglomeration economics and market demand would cause ‘size effect’ of population and employment The larger scales of the initial population and employment would increase the probability of building redeveloping into residential and commercial uses; thus, a positive effect is expected However, the effect upon conversion to industrial and other uses may be either positive or negative depending on the local context of land development (Landis and Zhang 1998b) The data of these variables were obtained from the database of the Taiwan Population and Housing Census 3.2.3 Location The locational characteristics were measured in three ways: (1) accessibility (1995), (2) policy constrains on land uses (1995), and (3) neighboring or adjacent uses (1995) Accessibility was scaled by using GIS to measure the Euclidean distance from each area to downtown core (Hung and Wang 2011) The building uses proximity to city center are encouraged to change to ‘higher-order’ uses for reasons of minimized travel costs (Mills and Hamilton 1997) The expectation is the nearer the building uses to city center, the greater the probability of change to residential or commercial uses but the less the likelihood of change to industrial and other uses Local governments in Taiwan utilize zoning to stipulate which uses and densities are permitted where Local municipalities prepare their General City Plan (GCP) as the ultimate ‘build-out’ and land-use regulation boundaries New development is supposed to be channeled to areas within the GCP boundaries and steered away from unincorporated areas outside plan boundaries Thus, whether an area is located within the GCP boundary could serve as a simple proxy for policy constrains on land uses The areas within the GCP boundary are more likely to be developed or changed Generally, building use types are strongly influenced by the pattern of neighboring or adjacent uses For example, an area surrounded by commercial uses would be more likely to be developed into commercial uses than other uses GIS was utilized to compute the initial percentages (1995) of surrounding areas in commercial, residential, industrial, and other uses 123 Nat Hazards 3.3 Damage and direct economic loss model To construct damage and direct economic loss models, TELES was used, which was developed by the Taiwan National Science Council (TNSC) and Taiwan National Center for Research on Earthquake Engineering.1 TELES is composed of several modules that measure hazards and losses in various ways (Yeh et al 2006) For this illustrative case study, only parts of the modules were used Building damages were estimated using direct physical damage module that was built specifically to Taiwan building types and construction practices Buildings were grouped in terms of the model building types (MBT), seismic design levels, and occupancy classes TELES assesses the damage probability matrices for the MBT, which are specific to structural types and building height, and translates ground-shaking intensity into the probability of experiencing different damage states.2 The methodology allows the building damage results link to land-use types by combining damage into occupancy classes The results of occupancy class probability of damage were multiplied by replacement costs of structures and nonstructures for each MBT and total floor area to estimate direct economic loss (Yeh et al 2006) Direct economic loss results were aggregated by structural type, damage class, and occupancy class for each area of analysis The estimation of earthquake hazard and losses requires three types of input for using TELES First, the data included the scenario basis, attenuation relationship, and soil map Second, aggregated census tract data were used to estimate direct physical damages It included building types and aggregated data on the general building stock, which were obtained from the Taiwan Taxation Agency Third, the replacement costs of structures and nonstructures were taken from a survey of building contractors This input is needed for calculating direct economic losses Results and discussion 4.1 Background and earthquake scenarios Hsinchu City is located in the northwestern coast of Taiwan (Fig 2) It has a population of 0.41 million and an area of 104.15 km2 The city is a high-density, rapidly growing, and a major high-tech industry assemblage area in Taiwan, which would increase its vulnerability and exposure to seismic hazards Hsinchu is situated in one of the moderate-tohighest activity seismic zones in Taiwan Crustal- or subcrustal-type earthquakes occur more frequently and can reach magnitude 7.0 or greater in this area (Campbell et al 2002) For example, a major earthquake event, the Mw = 7.1 Hsinchu-Taichung earthquake, occurred in 1935 Furthermore, the 1845 Mw = 6.0 Taichung earthquake and the 1999 Mw = 7.6 Chi–Chi earthquake which struck the areas of neighboring were also crustal or subcrustal events Our research, thus, focused on these types of earthquakes A singlescenario event was used for comparative static modeling of seismic risks at different points In 1998, the TNSC started the HAZ-Taiwan project to promote research on seismic damage and economic loss estimation HAZ-Taiwan closely resembles the approach employed in HAZUS TELES is a successive software version of HAZ-Taiwan, in which the analysis modules and parameters have been revised to accommodate the local environment and engineering practices in Taiwan (Yeh et al 2006; Hung and Chen 2007) There are four damage states in TELES for building damage estimation These states include ‘slight’, ‘moderate’, ‘extensive’, and ‘complete’ damage 123 Nat Hazards Fig Earthquake epicenters and location of the Hsinchu City in time This approach allows controlling for the hazard, so that any changes in risk can be directly resulted from changes in vulnerability factors.3 Seismic risks were evaluated for two hypothetically probabilistic earthquake scenario events—each with a characteristic earthquake magnitude, location, and probability (Table 1) In comparison with the notably historical damaging earthquake in this region (the 1935 Hsinchu-Taichung earthquake), the Shitan scenario event has the same magnitude but an epicentral location assumed in the midpoint of Shitan fault (Fig 2) The Hsincheng scenario, developed in consultation with the Hsinchu City Disaster Mitigation and Prevention Plan (HCDMPP), is also assumed an epicenter located in the midpoint of Hsincheng fault, which is one of the most active faults affecting Hsinchu These two hypothetical scenarios represent strong but realistic events that are very likely to occur in this area The scale of analysis in TELES is based on census tract for each Li (village), which is the basic unit of city administration in Taiwan Census tracts were grouped into 115 Lis for the BICM, and the outputs could be directly input into TELES Lis were defined according to municipal boundaries, as well as homogeneous soil, demography, and built environment 4.2 Earthquake damages and losses in hypothetical earthquakes By running TELES for each of the two hypothetical earthquake scenario events, the estimated peak ground acceleration (PGA) is within 0.27 and 0.53 g, and shows that the A full range of seismic hazard assessment is beyond the scope of this article But our approach can readily be extended to multiple-scenario events for probabilistic analysis (Shaw et al 2007) Scenario analysis used here aims to quantify risk This approach is increasingly used in urban planning and disaster risk management in various ways (Chang et al 2012) 123 Nat Hazards Table Two hypothetical earthquake scenario events Earthquake event Magnitude Fault type Location (two degree belts) Depth (km) Probability Hsincheng fault 7.1 Reverse 269,000, 2,742,000 10 0.0025 Shitan fault 7.1 Reverse 245,917, 2,721,782 10 0.0025 earthquake event on the Hsincheng fault would be more hazardous than on the Shitan fault The areas estimated to have the highest PGA are concentrated in the east and southeast portions of Hsinchu City Table shows the variability in total and expected annual direct economic losses, as well as percentage in significantly damaged buildings Predictions of losses and damages to buildings caused by each hypothetical earthquake event vary considerably due to earthquake damage and epicenter The larger damaging event is the earthquake on Hsincheng fault, with the expected percentages in significantly damaged buildings approximating % A validation exercise found that the results are reasonable when viewed against the most resent comparable disaster—the 1999 Mw = 7.6 Chi–Chi earthquake The Chi–Chi earthquake caused 8,773 buildings severe damage or collapse—yielded a damage rate of averaged 2–18 % in various areas (Chen et al 2002) While the rates in some areas are higher than in our case, it is within one order of magnitude Figure 3a, b shows the spatial distribution of percentage in significantly damaged buildings superimposed on Li boundaries throughout Hsinchu Both the earthquake scenarios would cause significantly damaged buildings over the city spatially very uneven, and the most damaged areas are concentrated in the eastern, southeastern, and some parts of the downtown core The results represent the overall disruptiveness of building damages, and suggest the potential of casualties as well as where heavy seismic impacts might occur and emergency management is required Expected annual losses for each hypothetical earthquake event were calculated by multiplying the total expected loss by the probability of each earthquake event occurrence This value can be considered as a long-term and rough measure of seismic risk The spatial distribution of estimated seismic risk resulting from two hypothetical earthquake events is highly consistent with the spatial patterns of significantly damaged buildings, although seismic risk distributes highly heterogeneous (Fig 3c, d) The aggregate loss results mask substantial intracity geography of risk and provide the decision-makers an overall picture for prioritizing risk-reduction measures and allocating emergent aid resources Table Estimated earthquake losses for buildings with current land uses Earthquake event Expected loss (NT$ billion)a Probability Expected annual lossb (NT$ million) Percentage in significantly damaged buildingsc Hsincheng fault 11.26 0.0025 28.15 6.94 2.92 0.0025 7.30 2.31 Shitan fault a New Taiwan Dollar (NT$) is convertible with US Dollar at an exchange rate in April 2013 of NT$1 = US$0.03 b Expected annual loss = expected loss probability of each earthquake event occurrence c Significantly damaged = ‘moderate’, ‘extensive’, or ‘complete’ damage states 123 Nat Hazards Fig a Distribution of percentage in significantly damaged buildings resulting from the earthquake on Hsincheng fault; b distribution of percentage in significantly damaged buildings resulting from the earthquake on Shitan fault; c distribution of expected annual direct economic losses to buildings resulting from the earthquake on Hsincheng fault; d distribution of expected annual direct economic losses to buildings resulting from the earthquake on Shitan fault 4.3 The determinants of building inventory changes Four separate logit models were applied to predict redevelopment activities in Hsinchu These logit models were specified as conversion probability from various building uses to residential, industrial, commercial, as well as public and educational uses, respectively (Table 3) As expected, initial land uses played a role in explaining residential building conversions The more the initial residential uses, the higher the probability of an area being converted to residential uses The areas with higher numbers of households, jobs, jobs– households ratios, or higher household growth rates were more likely to be converted to residential uses Distance from the city center within km served to encourage residential development It is worthy to note that areas surrounded by higher percentages of various land developments were all more likely to be developed into residential uses The spillover effects from residential, commercial, and industrial uses were significantly positive to 123 Variables Conversion to residential uses Conversion to industrial uses Conversion to commercial uses Conversion to public and educational uses Constant -293.5** (-2.62)a 10,547.7** (14.3) 3,267.1** (11.71) 3,620.8** (14.47) Initial residential use 0.02** (5.91) 0.08** (3.12) 0.06** (12.06) 0.02 (1.25) Initial industrial use -0.02** (-6.92) 0.13** (7.91) 0.05** (17.42) 0.2** (21.17) Initial commercial use -0.03** (-5.82) 0.87** (12.39) 0.23** (30.11) 0.6** (17.38) Initial public and educational uses -0.01** (-47.79) -0.01** (-17.11) -0.001** (-28.81) 0.002** (5.84) 0.007** (19.34) Households 0.00012** (4.70) -0.004** (-12.40) -0.002** (-32.89) Employment 0.0002** (15.98) 0.001** (6.36) 0.0002** (15.92) -0.004** (-18.72) Jobs/household ratio 1.497** (10.26) -16.71** (-6.91) 2.10** (8.68) 3.07** (5.73) Employment growth (%) 0.003 (0.18) -1.89** (-8.08) 0.29** (9.36) -1.46** (-21.45) Household growth (%) 0.078** (11.45) -1.11** (-11.79) 0.68** (37.38) -0.48** (-26.67) Within city plan area 0.001 (0.04) 2.95** (20.69) 1.10** (36.81) 3.58** (41.74) Distance to CBD within km 1.327** (36.93) -2.90** (-6.99) 1.57** (18.63) -4.71** (-20.25) Distance to CBD within 1–2 km 0.800** (26.43) -1.30** (-6.76) 2.96** (43.68) -5.07** (-23.63) Distance to CBD within 2–3 km 0.624** (22.05) -0.80** (-4.90) 1.99** (28.47) -2.99** (-16.16) -4.86** (-24.78) Distance to CBD within 3–4 km 0.418** (12.97) -1.39** (-9.00) 1.24** (17.40) Distance to CBD within 4–5 km 0.648** (19.04) -7.20** (-13.13) 0.82** (10.89) -5.10** (-17.79) Percentage of adjacent areas in residential use 289.89** (2.59) -1,055.0** (-14.3) -3,280.6** (-11.7) -3,625.3** (-14.5) Percentage of adjacent areas in commercial use 283.05* (2.52) -1,055.8** (-14.4) -3,276.8** (-11.8) -3,625.9** (-14.5) Percentage of adjacent areas in industrial use 288.73** (2.58) -1,055.0** (-14.4) -3,271.9** (-11.7) -3,624.1** (-14.8) 123 Percentage of adjacent areas in public and educational uses 288.67* (2.57) -1,054.7** (-14.4) -3,271.4** (-11.7) -3,632.1** (-14.5) Log likelihood function -169,379.5 -66,976.7 -167,647.3 -174,542.4 Pesudo-R2b 0.80 0.72 0.67 0.57 a t value in parentheses; * significant at the 0.05 level; ** significant at the 0.01 level b Pesudo-R2 = - (M/MC), where M is the log likelihood function of the estimated model; MC is for the constants-only model Nat Hazards Table Results of logit analysis for building inventory changes Nat Hazards residential use conversion A high mixture of land uses is, thus, expected as a major pattern of future land development in Hsinshu Results also show that initial residential, commercial, and industrial uses had positive effect upon industrial building development In contrast, the areas in which distance from the city center within km and those which were located outside of the GCP boundary were less likely to be developed into industrial uses These findings show that industrial uses were expected to be developed as an assemblage of mixed-use associated with commercial and residential uses in the suburbs of the GCP areas An area was more likely to be changed into industrial uses if it had a higher number of jobs, but a lower number of households and population growth The areas surrounded by the higher percentage of residential, commercial, and industrial uses were less likely to be converted to industrial uses The effects of initial residential, commercial, and industrial uses upon commercial uses conversion were significantly positive Surprisingly, the areas surrounded by residential, commercial, and industrial uses were less likely to be developed into commercial uses This follows the fact that commercial uses were concentrated at the city core, but relatively rare in both suburbs and the areas outside the GCP boundary Moreover, demand factors also significantly affected commercial use changes The areas with richer jobs, higher jobs–household ratios, household, and job growth were more likely to be changed to commercial uses Public and educational uses were less likely to be developed in the areas surrounding the city core, although initial residential, commercial, public, and educational uses had positive effects upon the uses conversion to public and educational uses The areas located within the GCP boundary were more likely to be converted to public and educational uses if they were surrounded by areas of less development Public and educational land development favored the areas with more households, less jobs, higher jobs–household ratios, and slower growth of households and jobs 4.4 Earthquake losses of building inventory changes 1000m2 Parameters estimated by the logit models were employed to predict the probabilities, spatial distribution, and magnitudes of long-term changes in various building uses, as well as to examine the factors that influence seismic risk Figure compares the building floor areas of initial uses with the predicted conversions to residential, commercial, industrial, public, and educational uses In Hsinchu, as currently developed, the total building floor areas are predicted to have an increase of 719,800 m2 in long term, an appreciation of 2.9 % compared to initial building inventory The most massive changes involve conversion to commercial and industrial uses, with an increase of total 669,000 m2 building floor areas, accounting of 10.8 and 2.8 % of initial commercial and industrial uses, respectively a 18,000 15,000 12,000 9,000 6,000 3,000 Initial building floors Predicted building floors 15,947 15,978 (0.2%)a 5,910 1,137 Residential uses 6,547 (10.8%) 1,169 (2.8%) Industrial uses Commercial uses : percentage of floors increase in parentheses Fig Estimates of the building floor area changes in the Hsinchu City 123 1,607 1,627 (1.2%) Public and educational uses Nat Hazards Table Estimated earthquake loss in changed building inventory Earthquake event Total (NT$ billion) Mean of Lis (NT$ billion) SD (NT$ billion) Expected annual loss (NT$ million) Increase in expected annual lossa (%) Hsincheng fault 12.36 0.11 0.07 31.0 2.85 (10.1 %) 2.93 0.03 0.02 Shitan fault 7.32 0.02 (0.3 %) a Increase in expected annual loss = expected annual loss under changed building inventory - expected annual loss under initial building inventory The net implication of these changes for expected losses is, according to the modeled results, slight to moderate: the hypothetical Hsincheng fault event would cause higher levels of risk than Shitan event in the long term (Table 4) In fact, while results indicate insignificant increases in the expected earthquake losses, these differences did not fully involve residential, commercial, and industrial property losses Figure 5a, b plots the distribution of estimated building earthquake economic losses for the changed building inventory throughout Hsinchu Comparing Figs 5a with 3c, and 5b with 3d, the changes of spatial distribution patterns are slight All these figures show that the most hazardous and risky areas are majorly concentrated in the eastern and southeastern portions, as well as in some downtown cores (Fig 5c, d) These areas are clustered by high-tech industries (i.e., Hsinchu Science Park), high density of mixture uses, and early urbanized land-use activities Numerous limitations of the model and results should be noted First, the results are specific to the particular earthquake scenarios The values of loss and their spatial distribution should change with different earthquake events, but it is unlikely that overall intertemporal trends would be significantly different Second, the model is subject to several computational assumptions (e.g., the negligence of building code and structure practice upgrades) and data assumptions These limitations imply that the actual loss or risk values should be explained as order-of-magnitude estimates, rather than accurate predictions However, these model and data assumptions were made consistently for both initial and long-term building inventory changes, and greater confidence can be placed in the conclusions related to risk trends over time 4.5 Application to emergency response and land-use planning Seismic risk analysis can help city planners in the selection of safer sites for development, regulation of land uses, and improvement of building codes, construction practices, emergency services, as well as promotion of risk communication Especially, in the Hsinchu City, the highly developed downtown cores rapidly growing suburb areas, city planners need to assess and monitor city growth and land-use development over time for preparing emergency infrastructure, as well as for supporting land-use and predisaster hazard-mitigation planning to reduce property loss and save the lives (Lindell et al 2006; Schwab et al 2007; Amini Hosseini et al 2009) According to DPPA, the Hsinchu City government attempts to implement the HCDMPP, which is a predisaster emergency plan for identifying the facilities and guidelines to assist emergency response activities, such as the assignment of emergency operation centers, medical care, shelters, emergency roads, fire, logistics, and police systems This plan focuses more on the allocation of emergency infrastructure, but it lacks incorporation of seismic risk analysis into the planning process 123 Nat Hazards Fig a Distribution of estimated annual economic losses for the changed building inventory resulting from the earthquake on Hsincheng fault; b distribution of estimated annual economic losses for the changed building inventory resulting from the earthquake on Shitan fault; c increase in annual economic losses for building inventory changes resulting from the earthquake on Hsincheng fault; d increase in annual economic losses for building inventory changes resulting from the earthquake on Shitan fault Using estimators of seismic risk model, Fig delineates the map of HCDMPP in downtown core overlaid on the distribution of predicted percentage of changed building inventory in significantly damaged buildings resulting from the earthquake on Hsincheng fault It indicates that the eastern and southeastern parts of the downtown core will have a long-term rise in vulnerability and risks, because significant increases in residential and commercial uses are expected in these areas This would enhance the demand for emergency response infrastructure In case of strong earthquakes, some of emergency roads and facilities would be blocked from the higher percentage of collapsed buildings Through the dynamics of damage and risk analyses, we demonstrate some of the types of insights needed by the collaboration of emergency managers with city planners to improve emergency preparedness and land-use planning In the highly vulnerable and risky areas, pre-earthquake emergency planning, such as allocation of emergency infrastructure, is necessary to be incorporated into land-use planning to ensure the capability of systems to absorb the impacts of strong earthquake and to reduce vulnerability 123 Nat Hazards Fig Distribution of emergency response facilities superimposed on the predicted distribution of percentage of changed building inventory in significantly damaged buildings resulting from the earthquake on the Hsincheng fault Conclusions This study proposes a novel methodology to investigate the dynamics of urban earthquake risk through modeling and comparing risk over time Findings provide substantial information to inform city planners for future infrastructure design and seismic risk reduction, as well as to link the process of emergency preparedness with land-use planning While most of the research focused on the current seismic risk, especially at the global and 123 Nat Hazards national scales, few studies have examined seismic risk over time empirically at the city levels Generally, the study predicted a long-term trend of seismic risk changes and found that the net effect of the multiplicity factors that influence risk—from population growth to land-use, socioeconomics, location, and their interaction—is rare to consider in the preparedness of emergency responses and land-use plans In the case of Hsinchu City, there are two principle findings First, overall risk and damage would be increasing The simulation analyses found a projected 0.3–10.1 % growth of expected annual loss while total building inventory was predicted to increase by approximately 2.9 % This overall increase in risk is attributable mostly to a large amount of initial buildings predicted to be developed into commercial and industrial uses Second, the risk is spatially diversified, although the pattern of risk distribution is predicted to change slightly As the seismic risk at different points in time is characterized by spatial disparities, the southeastern, eastern, and some older core areas are expected to be the most vulnerable and risky both at current and future time periods These findings collectively indentify and indicate ‘hot spots’ of particularly risky areas for decision-makers to prepare for and reduce risk The methodology presented here enables maximally using existing data on representing urban change, which is available for conducting risk assessment and supporting city (or regional) planning However, it is worthy to note the application of this methodology for other cities in Taiwan as well as other local administrative units elsewhere in the world For example, the projection of future characteristics of city areas entails substantial uncertainty because assumptions and omissions were consistently made for the analysis Our modeled results were confirmed in the seismic loss and damage analyses It shows that the findings are reasonable, at least an order-of-magnitude, when compared with historical earthquake events Thus, findings on intertemporal trends are more reliable than loss estimates for a particular point in time In order to improve the model robustness, future research should incorporate more earthquake scenarios into the seismic risk analyses This work is critical for conducting a more complete seismic risk assessment to support policy-making Moreover, additional city case studies in methodology applications remain to be investigated Other cities such as metropolitan areas would have more sizeable stocks of highly vulnerable buildings Similarly, many cities in developing countries also experienced rapid growth and continual change, and major portions of developments or buildings may be comparatively vulnerable due to lack of building code provisions, land-use regulation, and enforcement Understanding the risk characteristics of confluence of these developments and changes is crucial for reducing the risk of disasters that will likely strike at some future points in time Acknowledgments This research was supported by grants from the National Science Council, Taiwan (NSC93-2415-H-305-012), and the Architecture and Building Research Institute, Taiwan References Al-Momani NM, Harrald JR (2003) Sensitivity of earthquake loss estimation model: how useful are the prediction Int J Risk Assess Manag 4(1):1–19 Amini Hosseini K, Jafari MK, Hosseini M, Mansouri B, Hosseinioon S (2009) Development of urban planning guidelines for improving emergency response capacities in seismic areas of Iran Disasters 33(4):645–664 123 Nat Hazards Barbat AH, Carren˜o M-L, Pujades LG, Lantada N, Cardona OD, Marulanda MC (2010) Seismic vulnerability and risk evaluation methods for urban areas A review with application to a pilot area Struct Infrastruct Eng 6(1–2):17–38 Berke PR, Song Y, Stevens M (2009) Integrating hazard mitigation into new urban and conventional developments J Plan Educ Res 28:441–455 Burby RJ, French SP, Nelson A (1998) Plans, code enforcement, and damage reduction: evidence from Northridge earthquake Earthq Spectra 14(1):59–74 Burby RJ, Deyle RE, Godschalk DR, Olshansky RB (2000) Creating hazard resilient communities through land-use planning Nat Hazard Rev 1(2):99–106 Campbell KW, Thenhaus PC, Barnhard TP, Hampson DB (2002) Seismic hazard model for loss estimation and risk management in Taiwan Soil Dyn Earthq Eng 22:743–754 Carren˜o M-L, Cardona OD, Barbat AH (2007) Urban seismic risk evaluation: a holistic approach Nat Hazards 40:137–172 Chang SE, Gregorian M, Pathman K, Yumagulova L, Tse W (2012) Urban growth and long-term changes in hazard risk Environ Plan A 44:989–1008 Chen LC, Hung HC, Chen SY, Lai MJ (2002) The socioeconomic impacts and post-disaster recovery of Chi–Chi earthquake: a case study of Taichung and Nantou County, Taiwan Civ Hydraul Eng 29(2):8–20 Comerio MC (2004) Public policy for reducing earthquake risks: a US perspective Build Res Inf 32(5):403–413 Cutter SL, Boruff BJ, Shirley WL (2003) Social vulnerability to environmental hazards Soc Sci Q 84:242–261 Davidson R, Rivera M (2003) Projecting building inventory changes and the effect on hurricane risk J Urban Plan Dev 129(4):211–230 Erdik M, Durukal E (2008) Earthquake risk and its mitigation in Istanbul Nat Hazards 44:181–197 Hung HC (2009) The attitude towards flood insurance purchase when respondents’ preferences are uncertain: a fuzzy approach J Risk Res 12(1–2):239–258 Hung HC, Chen LC (2007) The application of seismic risk-benefit analysis to land-use planning in Taipei City Disasters 31(3):256–276 Hung HC, Wang TW (2011) Determinants and mapping of collective perceptions of technological risk: the case of the Second Nuclear Power Plant in Taiwan Risk Anal 31(4):668–683 Inel M, Senel SM, Toprak S, Manav Y (2008) Seismic risk assessment of buildings in urban areas: a case study for Denizli, Turkey Nat Hazards 46(3):265–285 Jain VM, Davidson R, Rosowsky D (2005) Modeling changes in hurricane risk over time Nat Hazards Rev 6(2):88–96 Kircher CA, Whitman RV, Holmes WT (2006) HAZUS earthquake loss estimation methods Nat Hazards Rev 7(2):45–59 Landis J, Zhang M (1998a) The second generation of the California urban futures model Part 1: model logic and theory Environ Plan B 25(5):657–666 Landis J, Zhang M (1998b) The second generation of the California urban futures model Part 2: specification and calibration results of the land-use change submodel Environ Plan B 25(6):795–824 Lindell MK, Prater CS, Perry RW (2006) Fundamentals of emergency management Federal Emergency Management Agency, Emergency Management Institute, Emmitsburg, MD Liu YC, Chien SY, Hung HC, Chen LC (2013) Exploring the spatial patterns of disaster impacts caused by Typhoon Morakot J Chin Inst Eng doi:10.1080/02533839.2012.727594 Martelli L, Filippini M, Bagli S, Severi P, Tomasetti F (2007) Seismic risk reduction: a local effects map for territorial and urban planning The example of the Rimini Territorial plan Boll Geofis Teor Appl 48(2):151–162 Mills ES, Hamilton BM (1997) Urban economics, 5th edn Scott, Foresman and Company, London Nath SK, Thingbaijam KKS, Raj A (2008) Earthquake hazard in Northeast India—a seismic microzonation approach with typical case studies from Sikkim Himalaya and Guwahati city J Earth Syst Sci 117:809–831 Nelson A, French SP (2002) Plan quality and mitigating damage from natural disasters: a case study of the Northridge earthquake with planning policy considerations J Am Plan As 68(2):194–207 Nirupama N, Simonovic SP (2007) Increase of flood risk due to urbanisation: a Canadian example Nat Hazards 40(1):25–41 NRC (2006) Facing hazards and disasters: understanding human dimensions National Research Council, Committee on Disaster Research in the Social Science, The National Academics Press, Washington, DC 123 Nat Hazards Olshansky RB (2001) Land use planning for seismic safety: the Los Angeles County experience, 1971–1994 J Am Plan As 67(2):173–185 Olshansky RB, Wu Y (2001) Earthquake risk analysis for Los Angeles County under present and planned land uses Environ Plan B 28(3):419–432 Sarris A, Loupasakis C, Soupios P, Trigkas V, Vallianatos F (2010) Earthquake vulnerability and seismic risk assessment of urban areas in high seismic regions: application to Chania City, Crete Island, Greece Nat Hazards 54:395–412 Schneider PJ, Schauer BA (2006) HAZUS-its development and its future Nat Hazards Rev 7(2):40–44 Schwab AK, Eschelbach K, Brower DJ (2007) Hazard mitigation and preparedness Wiley, Hoboken, NJ Sengezer B, Koc¸ E (2005) A critical analysis of earthquakes and urban planning in Turkey Disasters 29(2):171–194 Shaw D, Yeh CH, Jean WU, Loh CH, Kuo YL (2007) A probabilistic seismic risk analysis of building losses in Taipei: an application of HAZ-Taiwan with its pre-processor and post-processor J Chin Inst Eng 30(2):289–297 Tate E, Cutter S, Berry M (2010) Integrated multihazard mapping Environ Plan B 37(4):646663 Tyagunov S, Gruănthal G, Wahlstroăm R, Stempniewski L, Zschau J (2006) Seismic risk mapping for Germany Nat Hazards Earth Syst Sci 6:573–586 UNISDR (2005) Hyogo framework for action 2005–2015: building the resilience of nations and communities to disasters United Nations International Strategy for Disaster Reduction, Geneva Verburg PH, Schot PP, Dijst MJ, Veldkamp A (2004) Land use change modelling: current practice and research priorities GeoJournal 61(4):309–324 Wegener M (1994) Operational urban models J Am Plan As 60(1):17–29 Yeh CH, Loh CH, Tsai KC (2006) Overview of Taiwan earthquake loss estimation system Nat Hazards 37(1–2):23–37 Yong C, Ling C, Guăendel F, Kulhanek O (2002) Seismic hazard and loss estimation for Central America Nat Hazards 25(2):161–175 123