ORI GIN AL PA PER

Multi-level geospatial modeling of human exposurepatterns and vulnerability indicators

Christoph Aubrecht • Dilek Ozceylan • Klaus Steinnocher •

Sergio Freire

Received: 13 April 2012 / Accepted: 9 September 2012 / Published online: 18 September 2012� Springer Science+Business Media B.V. 2012

Abstract In the context of disaster risk management and in particular for improving

preparedness and mitigation of potential impacts, information on socioeconomic charac-

teristics including aspects of situation-specific human exposure and vulnerability is con-

sidered vital. This paper provides an overview on available multi-level geospatial

information and modeling approaches from global to local scales that could serve as

inventory for people involved in disaster-related areas. Concepts and applications related to

the human exposure and social vulnerability domains are addressed by illustrating the

varying dimensions and contextual implications. Datasets and methods are highlighted that

can be applied to assess earthquake-related population exposure, ranging from global and

continental-scale population grids (with a focus on recent developments for Europe) to

high-resolution functional urban system models and space–time variation aspects. In a

further step, the paper elaborates on the integration of social structure on regional scale and

the development of aggregative social and economic vulnerability indicators which would

eventually enable the differentiation of situation-specific risk patterns. The presented

studies cover social vulnerability mapping for selected US federal states in the New

Madrid seismic zone as well as the advancement of social vulnerability analysis through

integration of additional economic features in the index construction by means of a case

study for Turkey’s provinces.

Keywords Multi-level geospatial information � Spatial disaggregation � Population grid �Human exposure � Socio-economic vulnerability index � Risk reduction

C. Aubrecht (&) � K. SteinnocherAIT Austrian Institute of Technology, Vienna, Austriae-mail: christoph.aubrecht@ait.ac.at

D. OzceylanSakarya University, Sakarya, Turkey

S. FreireFCSH, e-GEO, Universidade Nova de Lisboa, Lisbon, Portugal

123

Nat Hazards (2013) 68:147–163DOI 10.1007/s11069-012-0389-9

1 Introduction

Integrated disaster risk management starts much earlier and extends beyond the immediate

emergency response actions following a catastrophic incident (UNDP 2004). In many

projects dealing with crisis management, the main focus has been on the phase starting at

the point when an unwanted event happens and lasting until the activities return to normal

routines (i.e., ad hoc reaction rather than proactive mitigation). There has been less

emphasis on the other phases of the disaster management cycle including risk analysis as

well as prevention and preparedness, even though related aspects have a strong influence

on the general status of a society and its citizens. Especially, the potential of a crisis to

escalate into a large-scale disaster is heavily dependent on the overall level of preparedness

as well as on the planning of mitigation and response actions and their timely execution.

Response and mitigation preparations considering multi-dimensional influencing factors

as well as space–time dependencies lessen impacts to social systems and in addition, reduce

the overall time for recovery. In the context of large-scale disasters, there is a clear indi-

cation that proper planning and preventive actions could have at least decreased the pos-

sibility of the crisis to escalate (OECD 2004). Furthermore, there are also indications to

obvious general failures in implementing proper emergency response systems and applying

lessons learnt from major disasters such as Hurricane Katrina or the recurrent Mediterranean

forest fires (Tsubokawa 2006). In modern societies, it is considered rational to expect that

decision makers do everything to avoid disasters that affect the societies’ individual

members and even their function and prosperity as a whole. However, as the course of

events of many natural and technological disasters has recently demonstrated, these efforts

often fail to go beyond short-term actions and be sustainable (Ingram et al. 2006; UNISDR

2004). Besides a set of other influencing factors, the lack of a preventive attitude at many

levels (e.g., regional/international, personal/political) has been identified as essential in

contributing to this failure (Xanthopoulos 2007). Therefore, among the most important

elements in mitigating crisis impacts and reducing societal vulnerability are proper planning

and related multi-sectoral cooperation. There is a need for multi-level geospatial infor-

mation to support improved decision making and to enable modeling of potential impacts on

social systems that vary according to their inherent vulnerabilities (Aubrecht et al. 2012a).

In the United Nations International Strategy for Disaster Reduction’s (UNISDR) ‘‘Ter-

minology on Disaster Risk Reduction’’ vulnerability is described as ‘‘the characteristics and

circumstances of a community, system, or asset that make it susceptible to the damaging

effects of a hazard.’’ Examples for the many aspects of vulnerability are provided, including

‘‘poor design and construction of buildings, inadequate protection of assets, lack of public

information and awareness, limited official recognition of risks and preparedness measures,

and disregard for wise environmental management.’’ It is also referred to the significant

variation in vulnerability within a community and over time which is an issue of high

concern in the scientific literature (Cutter 2003; Aubrecht et al. 2012b). The UNISDR

definition identifies vulnerability as being ‘‘independent of its exposure.’’ However, it is

explicitly stated that the term is commonly used ‘‘more broadly to include the element’s

exposure.’’ (UNISDR 2009) Building upon that, a central objective of vulnerability

assessment is, in our understanding, to provide indications where, when, and how people

and associated assets might be affected by a certain impact, including its aftermath (Birk-

mann 2007). Variable patterns of vulnerability in that context determine where and when a

mere natural event potentially turns into a disaster (EEA 2010). Results of vulnerability and

eventually risk assessments should provide decision- and policymakers with supporting

information to target response and mitigation actions adequately.

148 Nat Hazards (2013) 68:147–163

123

For assessment of the social dimension of vulnerability, mapping and analysis of

population exposure forms an essential part and is usually considered the starting point.

Integration of social structure and varying aspects of resilience would then further dif-

ferentiate situation-specific vulnerability patterns on a local scale. Resilience in that

context has been defined by UNISDR (2009) as ‘‘the ability of a system, community or

society exposed to hazards to resist, absorb, accommodate to and recover from the effects

of a hazard in a timely and efficient manner.’’ In a disaster risk management context,

assessment of human vulnerability has generally been disregarded in favor of hazard

analysis studies (Pelling 2004). Accurately estimating population exposure including

spatio-temporal aspects is a key component of catastrophe loss modeling, one element of

effective integrated risk analysis and emergency management (Aubrecht et al. 2010a). This

paper will provide an overview of concepts and applications in the human exposure and

social vulnerability domains, highlighting the varying scales and dimensions and the

related implications considered in an earthquake context.

2 Multi-level population mapping for human exposure assessment

In the context of disaster risk management and particularly for exposure and impact

assessments, the quality of available input data in terms of both spatial and thematic

accuracy and reliability is one of the most important factors. Census data available in

inhomogeneous spatial reference units are considered the standard information input for

assessing potentially affected people, for example in case of an emergency. However, there

is a strong demand on population data that are independent from enumeration and

administrative areas. Raster representations meet this demand but are not yet available

globally in both spatial and thematic consistency. Reallocating aggregated population

counts from administrative areas to a regular grid requires areal interpolation methods such

as dasymetric mapping. This technique utilizes ancillary data to disaggregate coarse

population data to areas where it is effectively present, at a finer resolution. Land use/land

cover (LULC) maps are often used as a basis for the disaggregation process in that regard

(Eicher and Brewer 2001; Mennis and Hultgren 2006; Langford 2007).

For applications on a supra-regional level, rather coarse-scale raster data on population

patterns are mostly well suited, but for sub-regional analyses, representations on higher

spatial resolution are required, that is, fine-scale population grids which eventually go down

to basic building level. Information on functional relationships in urban and suburban

environments as well as high-level population distribution information is often not quickly

available in case of emergency and cannot be immediately produced with high accuracy.

Rapid mapping concepts are mostly applied in structural post-disaster damage assessment

and mainly rely on Earth Observation data from different sensors as, for example, provided

by the International Charter Space and Major Disasters (www.disasterscharter.org, last

accessed on July 12, 2012). Improving the usability of satellite-derived products and

improving the benefits of satellite data for disaster management support in general (e.g., in

the pre-event stage) is still subject to ongoing research (Buehler et al. 2006).

2.1 Supra-regional to global scale

Figure 1 shows a comparison of selected global-scale datasets on population distribution

and earthquake hazard patterns. For an overview of broad-scale approaches to population

distribution modeling, Balk et al. (2006) presented a list of datasets all focusing on

Nat Hazards (2013) 68:147–163 149

123

representing resident population based on the highest-resolution input data available. The

first and least complex dataset is the Gridded Population of the World (GPW) (CIESIN and

CIAT 2004) representing the residence-based spatial distribution of human populations

across the globe. ‘‘The Global Demography Project’’ (Tobler 1979) can be considered the

first major effort in that regard. Produced at the National Center for Geographic Infor-

mation and Analysis (NCGIA) in 1995, this first version of GPW was sort of a precursor

for the subsequent series of approaches to consistently map population on a global scale.

The current third edition of this broad-scale data product (GPWv3) aims at providing a

spatially disaggregated population layer that is compatible with datasets from social,

economic, and earth science fields. The output shows the distribution of human population

converted from national or sub-national spatial units (usually administrative units) of

varying resolutions to a series of geo-referenced grids at a resolution of 2.5 arc min

(Deichmann et al. 2001). Population data estimates are available for the period 1990–2015

by quinquennial years with estimates for 2005, 2010, and 2015 produced in collaboration

with the United Nations Food and Agriculture Programme (FAO) as ‘‘GPW: Future

Estimates’’ (Balk et al. 2005a).

Fig. 1 Compilation of global-scale datasets on population density (GPWv3) and EQ hazard distribution(peak ground acceleration and frequency) for Japan (Source: NASA Socioeconomic Data and ApplicationsCenter, CIESIN/Columbia University 2012). OpenStreetMap is provided on the top left for orientation

150 Nat Hazards (2013) 68:147–163

123

The Global Rural–Urban Mapping Project (GRUMP) provides a new suite of data

products that add rural–urban specification to GPWv3, hence featuring an increased level

of complexity (Balk et al. 2005b). This project emerged out of a need for researchers to be

able to distinguish population spatially by urban and rural areas (Montgomery et al. 2003).

There is, however, no single definition of what makes an area ‘‘urban’’. Balk (2009) points

out the United Nations World Urbanization Prospects (UN 2006) which identifies each

country’s definition of the term ‘‘urban’’ where criteria include a variety of population size

or density thresholds associated with administrative areas, capital cities, and combinations

thereof. The central data product resulting from GRUMP (current status: Version 1,

GRUMPv1) is a ‘‘Gridded Population of the World with Urban Reallocation’’ in which

spatial and population data of both administrative units and urban extents are gridded at a

resolution of 30 arc-s (i.e., 25-fold higher resolution compared to GPW).

Considering a set of additional assumptions about real-world population distribution,

another group of modeled datasets is based on the accessibility concept. Basic motive for

that kind of methodology is that people tend to live in or close to cities and tend to move

toward areas that are well connected with urban centers (Balk et al. 2006). This premise

basically holds true even for rural regions for which it is expected that areas of higher

population density are located preferentially close to transport links and bigger cities. The

concept of accessibility and related indicators has been in use for a long time in particular

in transportation research (Koenig 1980; Halden 2003; Gutierrez et al. 2010). Applying

accessibility indicators to population distribution and reallocation models has, however,

not been that popular until first implementations for producing continental-scale databases

for Africa, Asia, and Latin America in the mid 1990s (Deichmann 1996, 1998). Just

recently Langford et al. (2008) examined how alternative population distribution models

(i.e., approaches using even-distribution within census tracts vs. dasymetric mapping)

influence GIS-based accessibility analyses, thus approaching this topic from the opposite

side. Deichmann (1997) provides a comprehensive overview on the use of various

accessibility indicators in GIS. He emphasizes that there are several ways to define

accessibility and presents different concepts. Accessibility can be defined as the ability for

interaction or contact with sites of economic or social opportunity. This definition is

supported by Goodall (1987) who states ‘‘The concept expresses the ease with which a

location may be reached from other locations. [It] summarizes relative opportunities for

contact and interaction.’’ Aubrecht et al. (2010b) presented a case study where an acces-

sibility surface was produced for the territory of Austria, subsequently used for population

disaggregation. Referring to available high-resolution raster census data, that study illus-

trated first validation results providing an idea of what can be expected from supra-

regionally modeled population datasets in terms of statistical accuracy and deviations.

Besides the abovementioned global- and continental-scale population databases, there is

another dataset available on that spatial level—LandScanTM—comprising a worldwide

population database compiled on a 30 arc-s latitude/longitude grid. In addition to trans-

portation networks and populated places, LandScan includes likelihood coefficients based

on parameters such as elevation, slope, nighttime lights and land cover for apportioning

census counts and UN estimates to each grid cell, while less effort is spent on using the

highest possible resolution population input information. This dataset has a categorically

different focus compared with the previously described population distribution models as it

aims at measuring ambient population instead of attempting to represent nighttime census

resident population (Dobson et al. 2000; Bhaduri et al. 2002). These datasets have different

objectives, make different assumptions, use different methodologies, and are designed to

measure two different indicators. As stated on the iSciences ‘‘Global Data Hound’’ blog in

Nat Hazards (2013) 68:147–163 151

123

April 2009 (http://www.terraviva.net/data/blogs.htm), ‘‘it would not be fair to say that one

or the other is ‘better’, it is more a question of what tool(s) are best for the job at hand.’’

The level of aggregation of all the above described databases (30 arc-s resolution at

best, approximately 1 km2) is still too coarse to adequately support risk analysis at a

detailed local level. However, other information sources (e.g., remote sensing imagery)

may allow disaggregation at a finer resolution. The development of the LandScan Global

Population Database represented a great improvement over pure residence-based popula-

tion datasets. However, its representation of ‘‘ambient population’’ corresponds to a

temporal averaging (over 24 h) that is not ideal for use in time-specific hazards such as an

earthquake or tsunami event. At a regional level, it has been attempted to overcome these

limitations by developing population distribution databases featuring higher temporal and

spatial detail (McPherson et al. 2006; Bhaduri et al. 2007; Freire and Aubrecht 2010),

including variation in the daily cycle, but such efforts are still to be applied on global-scale

datasets.

2.2 The European situation

For the last couple of years, the development of high-resolution population grids has been

a priority topic of the national statistical offices and Eurostat, the statistical office of the

European Commission. A number of European countries have recently changed their

population census to register-based statistics, with population numbers linked to address

points. Based on these point datasets, population grids of any size can easily be generated,

in due consideration of data privacy. These activities led to a European initiative called

Geostat (http://www.essnet-portal.eu/geostat-0, last accessed on July 12, 2012) coordinated

by Eurostat with the objective to establish a 1-km population grid for the member states of

the European Union. Those countries that already have a register-based census would

deliver aggregated grid data. For the remaining countries, a disaggregation method should

provide the required information.

As stated above, disaggregation methods such as dasymetric mapping rely on ancillary

data for spatially disaggregating available coarse-scale population data to effective pop-

ulated areas on a finer resolution. Mostly land use/land cover (LULC) maps are used in that

regard, on a European scale, for example, CORINE land cover data (CLC) were applied as

basis for spatial disaggregation of residential population data, by either (1) estimating

population density weights for the CLC classes (Gallego 2010) or (2) strictly considering

mere residential CLC classes (Steinnocher et al. 2006). The drawback of these approaches

is the limited spatial resolution of the CLC dataset that leads to over- or underestimation of

sparsely populated areas, respectively.

With the recently published EEA Fast Track Service Precursor on Land Monitoring

(EEA 2008), a new dataset is now available that provides the degree of soil sealing for

EU27? countries. By applying this dataset as a proxy for population density, the spatial

disaggregation can be improved significantly. The EEA Fast Track Service Precursor on

Land Monitoring is a raster dataset for built-up areas including continuous degree of

imperviousness ranging from 0 to 100 % in spatial resolution of 1 ha. The dataset is based

on orthorectified high-resolution satellite imagery (Image2006 European Coverages,

Muller et al. 2009), acquired primarily in the reference year 2006 (±1 year). Supervised

classification techniques were used to automatically map built-up areas, followed by visual

improvement in the classification results. The degree of soil sealing for the classified built-

up zones was derived from calibrated normalized difference vegetation index (NDVI). The

dataset covers EU27 and neighboring countries, 38 countries in total, and has just recently

152 Nat Hazards (2013) 68:147–163

123

been updated for the reference year 2009 (EEA 2008). We will refer to this dataset as

H(igh) R(esolution) soil sealing layer.

Applying that HR soil sealing layer as a proxy for population (or rather housing)

density, the accuracy of disaggregation results can be improved significantly. The method

is based on the assumption that the population density is proportional to housing density,

no population resides outside housing areas, and the ratio between population and housing

density is constant within a region. The housing density is derived from the HR soil sealing

layer, assuming that the degree of soil sealing is proportional to housing density. Since this

assumption does not hold for all cases, the soil sealing layer requires further processing. In

order to get a representation of housing densities, it is necessary to mask out all sealed

surface areas with non-residential function. These include road and rail networks, as well

as industrial and commercial areas. Masking the transport network is based on linear road

and rail data derived from OpenStreetMap, which are rasterized and expanded up to 50 m

depending on the road type in order to cover associated areas as well. Non-residential built-

up areas, such as industrial and commercial land use, are derived from respective CLC

classes. Due to the large minimum mapping unit of CLC, masking is limited to areas larger

than 25 ha. The remaining areas covered by the adapted HR soil sealing layer are assumed

to represent residential building densities and are used as input to the population disag-

gregation approach (Steinnocher et al. 2011).

The methodology was applied to a European-wide dataset, covering EU27 and EFTA

(European Free Trade Association) countries. Population input data dated 2006 were

provided by Eurostat on municipality level, thus temporally corresponding with the HR

soil sealing layer. Disaggregation of the population counts was performed for each input

region (basic reference unit: LAU 2; ‘‘local administrative unit’’ referring to municipalities

or equivalent units in the 27 EU member states), with an output grid featuring a defined

spatial resolution of 1 km2 (Fig. 2). Comparison of the resulting population grid against

local population numbers provided by Statistics Austria showed significant improvement in

this disaggregation method compared with previous CLC-based approaches (Steinnocher

et al. 2011). The resulting European population grid was validated against the aggregated

grids from Austria, Denmark, and the Netherlands. All tested sites show a high correlation

between the disaggregated grid and the references (R2 of 0.90, 0.86, and 0.92,

respectively).

In a final processing step, the available bottom-up grids were integrated by Eurostat in

order to obtain the best available information on population in Europe. The resulting

population grid is therefore a hybrid information layer, providing aggregated population

grids from register-based census for those countries where this source is available and

disaggregated population grids for the other countries. The hybrid population grid is

available at the webpage of the European Forum for Geostatistics http://www.efgs.info/

data/GEOSTAT_Grid_POP_2006_1K (last accessed on July 12, 2012).

2.3 Local scale

Information on functional patterns in urban environments as well as high-level population

distribution information is often not quickly available in case of emergency which is why

rapid mapping concepts often rely on various Earth Observation data sources. For building

comprehensive urban data management systems and mapping the complex urban envi-

ronment at a high level of detail, diverse input data are required, with functional infor-

mation such as socioeconomic and explicit demographic data on the one hand and ‘‘real

world’’ physical properties as derived from remote sensing on the other hand (Steinnocher

Nat Hazards (2013) 68:147–163 153

123

and Kostl 2007). Earth Observation data classification is limited to physical characteristics

of the analyzed objects but does not include process-related information. With respect to

man-made features, this means that buildings can be detected as such, while building use

and related socioeconomic activities usually cannot be derived that way. Urban system

modeling based on remote sensing and geoinformation technology often does not go

beyond a certain spatial and thematic level regarding the corresponding reference objects.

Only the integration of population and socio-economic features and thus moving from land

cover detection to land use assessment enables modeling of societal vulnerability as well as

damage potential and impact patterns. Integrative approaches considering remote sensing

and ancillary information are therefore expected to further increase in importance in the

future (Aubrecht et al. 2009).

Aubrecht et al. (2010a) presented a model approach for a selected study area in Austria

with the objective of identifying functional and socioeconomic relationships in a suburban

environment on a very high level of both spatial and thematic detail. Very high-resolution

(1-m pan-sharpened) satellite imagery (IKONOS) and airborne laser scanning data were

considered in a joint classification using object-based image analysis which resulted in

derivation of detailed structural land cover information. Georeferenced postal address data

then enabled linking this geometric base data framework with ancillary spatial and space-

related information such as yellow pages company data and population data from the

census. Integrating all available datasets resulted in a 3D population model on sub-building

level (see Fig. 3) which was finally consulted for exposure and impact assessments (e.g.,

earthquake, noise propagation).

Fig. 2 1-km2 population grid of EU27 and EFTA countries based on disaggregation of LAU 2 populationto the HR soil sealing layer (subset)

154 Nat Hazards (2013) 68:147–163

123

The spatial distribution of population in general, and hence its exposure to hazards, is

time-dependent, especially in metropolitan areas. Due to human activities and mobility, the

distribution and density of population varies greatly in the daily cycle (Freire 2010).

Therefore, a more accurate assessment of population exposure and risk analysis requires

going beyond residence-based census maps and figures (representing a nighttime situation)

in order to be prepared for events that can occur any time and day (e.g., 1755 Lisbon

tsunami around 10 a.m., 2010 Haiti earthquake at 4:53 p.m., 2011 Japan tsunami at 2:46

p.m.).

The recently developed LandScan USA is an expansion to the basic LandScan Global

product that features ambient population distribution on a 1-km raster (Bhaduri et al.

2007). A multi-dimensional dasymetric modeling approach allowed the creation of a high-

resolution spatio-temporal population distribution dataset. At a 90-m resolution (3 arc-s ),

LandScan USA contains both nighttime residential and daytime population distribution

information incorporating movement of workers and students (Fig. 4).

Freire and Aubrecht (2010, 2011) illustrated the modeling and analysis of the spatio-

temporal distribution of population in the daily cycle in the Lisbon Metropolitan Area to

reassess potential human exposure to earthquake and tsunami hazard. Total population

Fig. 3 Population distribution information, disaggregated to sub-building level based on functional urbansystem modeling (Aubrecht et al. 2010a)

Nat Hazards (2013) 68:147–163 155

123

numbers and spatial distribution in the municipalities vary significantly between day and

night (i.e., population increases by up to 60 % during daytime). In order to approximate the

pattern variation between daytime and nighttime, data on workforce and commuting sta-

tistics as well as land use and infrastructure information were integrated with the basic

census counts in the spatial population disaggregation and reallocation process. Results

show increase in potential population exposure to tsunami inundation by 100 % during the

day.

In ongoing research, this approach has been extended to include evacuation modeling

considering tsunami hazard for part of the same study area whereby the initial 50-m grid

output was advanced to a high-level 3D building model (Freire et al. 2011, 2012). Figure 5

shows the evacuation model (illustrating evacuation travel time in minutes) for population

in buildings potentially exposed to tsunami flood waters. The model considers both hori-

zontal exits and vertical shelters for evacuation, whereby specific flood depth and building

height are identified in 3D preprocessing.

3 From human exposure patterns to social and economic vulnerability indicators

Mapping and analysis of human exposure forms an essential step for the assessment of the

social dimension of vulnerability and is usually considered the starting point. Measures of

exposure can thereby include a quantitative assessment of types of assets or the number of

people in an area as elaborated in the previous chapter. As UNISDR (2009) illustrates,

these can then be combined ‘‘with the specific vulnerability of the exposed elements to any

particular hazard, [in order] to estimate the quantitative risks associated with that hazard in

the area of interest.’’ Integration of social structure and varying aspects of resilience

enables development of social vulnerability indicators and eventually differentiation of

situation-specific risk patterns on regional scales.

Socioeconomic development is considered to affect the distribution of future patterns of

human exposure. For modeling future socioeconomic patterns and monitoring spatio-tem-

poral changes, we take official Eurostat population prospects into consideration. Predicted

changes in population counts available on NUTS 2 level (i.e., corresponding to provinces)

from the ‘‘EUROPOP2008—Convergence scenario, regional level’’ are disaggregated based

Fig. 4 Daytime versus nighttime population distribution information for lower Manhattan in New YorkCity, disaggregated to a 90-m grid (LandScan USA dataset)

156 Nat Hazards (2013) 68:147–163

123

on the assumption that the relative distribution patterns remain constant, that is, population is

reallocated to those grid cells already populated in the reference year depicting the ‘‘current’’

situation (2006). Future population densities (2030) are, thus, again calculated based on the

2006 version of the HR soil sealing layer described above, for lack of available future

prospects in that regard. However, in the frame of the geoland2 project (www.gmes-geoland.

info, last accessed on July 12, 2012), the HR soil sealing layer has just recently been updated

for the reference year 2009. This will allow extrapolations to the short to medium time scale

in the near future. Structural population information inherent in the Eurostat population

convergence scenario is used for calculating the change of the proportion of elderly people

(60? years of age) whereby these changes are again proportionally applied to the populated

grid cells (Aubrecht et al. 2011a).

Including that structural population information in human exposure mapping is con-

sidered an important step toward a potential assessment of social vulnerability where the

inherent structural patterns of a socioeconomic system play an important role (e.g., health

vulnerability, etc.). Social structure and economic strength are two important factors that

affect a region’s vulnerability. An established infrastructure and strong social and eco-

nomic structure can make a community resilient and speed up response and recovery

processes, following disastrous events (Ozceylan and Coskun 2012).

In the following paragraphs, we will illustrate social vulnerability assessment by using

aggregative indicators. The application of complex multi-level indicators is strongly

related to input data availability. Many information sources on social characteristics are

available on regional scales (e.g., federal states, provinces, etc.), while on a local level

information access is often constraint due to data privacy issues. As illustrated above, data

disaggregation using weighting factors based on certain kind of land cover and land use

functions has proved reasonable for basic locational aspects such as population distribu-

tion. Certain structural population characteristics such as information on age or gender can

be attached to that. However, this becomes more complex when input information dis-

tribution is not proportionally related to land use patterns, and therefore, other disaggre-

gation parameters have to be applied. This would be the case, for example, for poverty

levels or language proficiency.

The studies that are presented in the following are therefore carried out on regional

level. First, an application will be shown covering essential social characteristics on a

regional scale for selected US federal states. In a further step, we highlight by means of a

Fig. 5 Evacuation time model (right) for population in buildings potentially exposed to tsunami floodwaters (left) (Freire et al. 2012)

Nat Hazards (2013) 68:147–163 157

123

case study for Turkey’s provinces how social vulnerability analysis can be further

advanced through additional integration of economic features in the index construction.

Optional availability of input data for local-scale analysis as can be the case for research

purposes but as well in emergency situations would enable indicator downscaling, thus

allowing socioeconomic vulnerability mapping on high resolution.

3.1 Application of social vulnerability indicators

A conceptual framework developed for the New Madrid Seismic Zone Catastrophic Event

Planning Project (Elnashai et al. 2009) initiated by the Federal Emergency Management

Agency (FEMA) integrates the use of social vulnerability factors and the selection of

preparedness actions. The objective is to provide decision support for regional planners to

identify appropriate preparedness actions for catastrophic earthquake scenarios.

Vulnerability indicators have been selected for eight federal states in the New Madrid

seismic zone through consensus building processes with involved stakeholders. This

subset of social factors that eventually influence the selection of preparedness actions is

derived from census data and includes (1) poverty level (measured by percent of popu-

lation living in poverty), (2) lack of proficiency in English (measured by percent of

population not proficient in English), (3) vulnerability of age groups (measured as percent

of population under 5 and above 65 years of age), and (4) disabled population (measured

by the number of disabled residents). Figure 6 shows a map of the social vulnerability

index aggregated from the four social factors listed above (equally weighted) for the State

of Illinois.

3.2 Progressing toward socioeconomic vulnerability

Increasing extreme events and related catastrophic consequences to societal systems have

resulted in some changes in the perspective on disasters. The extremely destructive nature

of recent disasters such as the 2010 Haiti Earthquake or the 2011 Tohoku Earthquake and

related tsunami has drawn the attention to the importance of ‘‘management of disaster

risks’’ and fundamental issues like vulnerability and resilience assessment (Vogel et al.

2007).

In an attempt to advance social vulnerability analysis to a higher level of thematic

detail, Ozceylan (2011) integrated economic parameters into the index development for

Turkey’s provinces. Factors in that index were determined after extensive literature review

and chosen based on their applicability to Turkey. Then, selected groups of experts

weighted these factors according to their expected impacts on social and economic vul-

nerability. Finally, sub-index and index values have been calculated for each province in

Turkey using these weights.

The final aggregative socioeconomic vulnerability index as defined by Ozceylan and

Coskun (2012) consists of two sub-indexes in the social and economic domain, respec-

tively. The social vulnerability sub-index builds upon a set of interrelated factor groups

such as (1) population characteristics (i.e., population density, annual population growth

rate, birth rate), (2) vulnerable groups (i.e., vulnerable age group, people with disabilities,

female ratio, average household size), (3) preparedness level (i.e., education level, rate of

mandatory earthquake insurance coverage, number of volunteers in disaster-related

NGOs), (4) health service capacity (i.e., number of hospital beds, number of doctors), and

(5) shelter capacity (i.e., sheltering possibilities). The economic vulnerability sub-index on

the other hand builds upon factor groups including (1) employment (i.e., unemployment

158 Nat Hazards (2013) 68:147–163

123

rate, employment of women) and (2) other welfare-related indicators (i.e., GDP age

dependency ratio, rate of socially dependent population, renters, number of industrial

establishments, number of bank branches). Figure 7 shows a map of the socioeconomic

vulnerability index aggregated from the social and economic factors listed above for the

provinces of Turkey.

Fig. 6 Distribution of social vulnerability index patterns for the State of Illinois (at county level) and thepercentile equivalent (within state) (Yeletaysi et al. 2009; adapted)

Fig. 7 Distribution of socioeconomic vulnerability index patterns for Turkey (on province level) (Ozceylan2011)

Nat Hazards (2013) 68:147–163 159

123

4 Conclusions and outlook

This paper provides an overview of available multi-level geospatial information and

modeling approaches from global to local scales. Illustrating the varying dimensions and

related implications considered in an integrated disaster risk management context, concepts

and applications related to the human exposure and social vulnerability domains were

addressed. Datasets and methods were highlighted that can be applied to assess earthquake-

related population exposure, ranging from global- and continental-scale population grids

(with a focus on novel developments for Europe) to local-scale functional urban system

models and integration of space–time variation aspects. In a further step, the paper elab-

orates on the integration of social structure and the development of aggregative social and

economic vulnerability indicators which would eventually enable the differentiation of

situation-specific risk patterns on regional scales. The presented studies cover social vul-

nerability mapping for selected US federal states in the New Madrid seismic zone as well

as the advancement of social vulnerability analysis through additional integration of

economic features in the index construction by means of a case study for Turkey’s

provinces.

Identifying distinct daytime and nighttime population distribution characteristics on

local scale is a major improvement for human exposure mapping compared with standard

census-based models, but does, however, only display part of reality. Looking at future

developments in that field, new technology-driven advancements including improved data

storage and processing capabilities allow moving toward real-time representation of human

movement (Aubrecht et al. 2011b, 2012a). One way to record time-specific population

distribution information in case of emergency situations is mapping cell phone user

activity. Distinct time and location information of user actions allow mapping distribution

patterns of cell phone subscribers and applying that as proxy for spatio-temporal popu-

lation distribution (Loibl and Peters-Anders 2012). Thus, the population number at a

certain time in a certain area can be examined and in a further step, the potential time-

specific human exposure to a dangerous situation or hazard evaluated. Another option for

analyzing population dynamics and deriving spatio-temporal patterns of human exposure is

opening up with the availability of location-specific volunteered geographic information

(VGI) drawing upon the increasing number of persons who are equipped with ‘‘location

sensors’’ in the form of GPS-enabled mobile devices (Goodchild 2007). The willingness to

share situational experiences with others is generally increasing rapidly and is facilitated

by rising new technologies supporting the spatial component of social networks. As a

special type of user-generated content, VGI offers great potential to produce near real-time

location-specific information related to human activities, even though overall accuracy

remains an issue of debate (Cheng et al. 2011).

Data quality and availability as well as privacy concerns are highlighted and considered

as important issues. Furthermore, objectiveness and lack of conceptual consensus for

model development have to be discussed. While assessment of human exposure relies on

quantitative measures and spatial statistics, indicator development including expert-based

factor selection for social and economic vulnerability analyses introduce a certain source of

uncertainty and subjectivity. Nonetheless, aggregative indexes on regional level can offer

consistent comparability and broad applicability over wide areas and provide valuable

support for decision makers and emergency managers on various levels.

With the recent focus on data sharing and integration initiatives such as INSPIRE,

GEOSS, GMES, and others (Havlik et al. 2011), particularly in emergency situations,

reliable and consistent input data for local-scale analysis will likely become available for

160 Nat Hazards (2013) 68:147–163

123

authorized users as it is already the case for certain purposes on higher level (e.g., Disaster

Charter, SAFER). In line with near real-time coverage of spatial and temporal charac-

teristics of human activities including population movements and functional socioeco-

nomic aspects, this would enable indicator downscaling and thus allow vulnerability and

risk mapping at high spatio-temporal resolution.

References

Aubrecht C, Steinnocher K, Hollaus M, Wagner W (2009) Integrating earth observation and GIScience forhigh resolution spatial and functional modeling of urban land use. Comput Environ Urban Syst33:15–25

Aubrecht C, Kostl M, Steinnocher K (2010a) Population exposure and impact assessment: benefits ofmodeling urban land use in very high spatial and thematic detail. In: Tavares JMRS, Natal Jorge RM(eds) Computational vision and medical image processing: recent trends. computational methods inapplied sciences. Springer, Berlin, vol 19, pp 75–89

Aubrecht C, Yetman G, Balk D, Steinnocher K (2010b) What is to be expected from broad-scale populationdata?—showcase accessibility model validation using high-resolution census information. In: PainhoM, Santos MY, Pundt H (eds) Proceedings of AGILE 2010—The 13th AGILE international conferenceon geographic information science

Aubrecht C, Steinnocher K, Kostl M, Zuger J, Loibl W (2011a) Spatially modeled high detail population andclimate prospects for a European transect—an outlook to future patterns of vulnerability. In: GuarascioM, Reniers G, Brebbia CA, Garzia F (eds) Safety and security engineering IV. WIT Transactions onthe Built Environment. WIT Press, Ashurst, vol 117, pp 219–232

Aubrecht C, Ungar J, Freire S (2011b) Exploring the potential of volunteered geographic information formodeling spatio-temporal characteristics of urban population: a case study for Lisbon Metro usingfoursquare check-in data. In: Proceedings of the 7th international conference on virtual cities andterritories, Lisbon, pp 57–60

Aubrecht C, Freire S, Loibl W, Steinnocher K, Ungar J (2012a) The contribution of multi-level geospatialinformation to assessing urban social vulnerability to earthquakes. In: Joint conference proceedings 9thinternational conference on urban earthquake engineering (9CUEE) & 4th Asia conference on earth-quake engineering (4ACEE). Center for Urban Earthquake Engineering, Tokyo Institute of Technol-ogy, Japan, 1867–1875

Aubrecht C, Freire S, Neuhold C, Curtis A, Steinnocher K (2012b) Introducing a temporal component inspatial vulnerability analysis. Disaster Adv 5(2):48–53

Balk D (2009) More than a name: why is global urban population mapping a GRUMPy proposition? In:Gamba P, Herold M (eds) Global mapping of human settlement: experiences, data sets, and prospects.Taylor & Francis series in remote sensing applications. CRC Press, Boca Raton

Balk D, Brickmann M, Anderson B, Pozzi F, Yetman G (2005a). Mapping global urban and rural populationdistributions. Annex estimates of future global population distribution to 2015. Environmental andnatural resources working paper 24, FAO, Rome, Italy

Balk D, Pozzi F, Yetman G, Deichmann U, Nelson A (2005b) The distribution of people and the dimensionof place: methodologies to improve the global estimation of urban extents. In: Proceedings of the 5thinternational symposium remote sensing of urban areas (URS 2005), Tempe, AZ, USA

Balk DL, Deichmann U, Yetman G, Pozzi F, Hay SI, Nelson A (2006) Determining global populationdistribution: methods, applications and data. In: Hay SI, Graham AJ, Rogers DJ (eds) Global mappingof infectious diseases: methods, examples and emerging applications. Advances in parasitology.Academic Press, London, vol 62, pp 119–156

Bhaduri B, Bright E, Coleman P, Dobson J (2002) LandScan: locating people is what matters. Geoinfor-matics 5(2):34–37

Bhaduri B, Bright E, Coleman P, Urban M (2007) LandScan USA: a high-resolution geospatial and temporalmodeling approach for population distribution and dynamics. GeoJournal 69(1):103–117

Birkmann J (2007) Risk and vulnerability indicators at different scales: applicability, usefulness and policyimplications. Environ Hazards 7:20–31

Buehler YA, Kellenberger TW, Small D, Itten KI (2006) Rapid mapping with remote sensing data duringflooding 2005 in Switzerland by object-based methods: a case study. In: Martin-Duque JF, Brebbia CA,Emmanouloudis DE, Mander U (eds) Geo-environment and landscape evolution II. WIT Press, Ashurst

Nat Hazards (2013) 68:147–163 161

123

Center for International Earth Science Information Network (CIESIN), and Centro Internacional de Agri-cultura Tropical (CIAT) (2004) Gridded population of the world, version 3 (GPWv3) and the globalrural-urban mapping project, version 1 (GRUMPv1). Columbia University, Palisades, NY. Available athttp://sedac.ciesin.columbia.edu/gpw. Accessed 12 July 2012

Cheng Z, Caverlee J, Lee K, Sui DZ (2011) Exploring millions of footprints in location sharing services. In:Fifth international AAAI conference on weblogs and social media (ICWSM-11), Barcelona

Cutter S (2003) GI science, disasters, and emergency management. Trans GIS 7(4):439–445Deichmann U (1996) Asia medium resolution population database documentation. NCGIA, UC Santa

Barbara and UNEP/GRID GenevaDeichmann U (1997) Accessibility indicators in GIS. United Nations Statistics Division, Department for

Economic and Policy Analysis, New York, NY, USADeichmann U (1998) Africa medium resolution population database documentation. NCGIA, UC Santa

Barbara, UNEP/GRID Sioux Falls and World Resources InstituteDeichmann U, Balk D, Yetman G (2001) Transforming population data for interdisciplinary usages: from

census to grid. GPWv2 working paper. CIESIN, Columbia University, Palisades, NY, USADobson JE, Bright EA, Coleman PR, Durfee RC, Worley BA (2000) LandScan: a global population

database for estimating populations at risk. Photogramm Eng Remote Sens 66(7):849–857EEA European Environment Agency (2008) LMCS high resolution soil sealing layer. EIONET European

topic centre on spatial information and analysis. Available at http://sia.eionet.europa.eu/CLC2006/FTSP/built-up_areas. Accessed 12 July 2012

EEA European Environment Agency (2010) Mapping the impacts of natural hazards and technologicalaccidents in Europe. Technical report 13/2010, Copenhagen, Denmark

Eicher CL, Brewer CA (2001) Dasymetric mapping and areal interpolation: implementation and evaluation.Cartogr Geogr Inf Sci 28(2):125–138

Elnashai AS, Cleveland LJ, Jefferson T, Harrald JR et al (2009) Impact of new Madrid seismic zoneearthquakes on the central USA, vol 2. Mid-America Earthquake Center Report 09-03, Urbana-Champaign, Illinois, USA

Freire S (2010) Modeling of spatio-temporal distribution of urban population at high resolution—value forrisk assessment and emergency management. In: Konecny M, Brandova TL, Zlatanova S (eds) Geo-graphic information and cartography for risk and crisis management, Lecture notes in geoinformationand cartography. Springer, Berlin, vol 1, pp 53–67

Freire S, Aubrecht C (2010) Towards improved risk assessment—mapping spatio-temporal distribution ofhuman exposure to earthquake hazard in the Lisbon Metropolitan Area. In: Proceedings of Gi4DM2010, international symposium on geoinformation for disaster management, Torino

Freire S, Aubrecht C (2011) Assessing spatio-temporal population exposure to tsunami hazard in the lisbonmetropolitan area. In: Proceedings of ISCRAM 2011, 8th international conference on informationsystems for crisis response and management, Lisbon, pp 5

Freire S, Aubrecht C, Wegscheider S (2011) Spatio-temporal population distribution and evacuationmodeling for improving tsunami risk assessment in the Lisbon metropolitan area. In: Proceedings ofGi4DM 2011, international symposium on geoinformation for disaster management, Antalya, pp 6

Freire S, Aubrecht C, Wegscheider S (2012) When the tsunami comes to town—improving evacuationmodeling by integrating high-resolution population exposure. In: Rothkrantz, Ristvej, Franco (eds)Proceedings of ISCRAM 2012, 9th international conference on information systems for crisis responseand management. Vancouver, Canada, April 22–25, 2012 (in press)

Gallego FJ (2010) A population density grid of the European Union. Popul Environ 31(6):460–473Goodall B (1987) Dictionary of human geography. Penguin, HarmondsworthGoodchild MF (2007) Citizens as sensors: the world of volunteered geography. GeoJournal 69(4):211–221Gutierrez J, Condeco-Melhorado A, Martın JC (2010) Using accessibility indicators and GIS to assess

spatial spillovers of transport infrastructure investment. J Transp Geogr 18(1):141–152Halden D (2003) Accessibility analysis concepts and their application to transport policy, programme and

project evaluation. In: Pearman A, Mackie P, Nellthorp J (eds) Transport projects, programmes andpolicies: evaluation needs and capabilities. Ashgate Publishing Ltd, Hampshire, pp 227–242

Havlik D, Zeug G, Aubrecht C, Steinnocher K (2011) Data sharing and integration initiatives for crisismanagement—what are the benefits of INSPIRE, GEOSS, SEIS and Co.?: In Knezic, Poudyal Chhetri,Ozunu (eds) Proceedings of the international emergency management society (TIEMS) 18th annualconference, Bucharest, Romania, pp 407–418

Ingram JC, Franco G, Rio CR, Khazai B (2006) Post-disaster recovery dilemmas: challenges in balancingshort-term and long-term needs for vulnerability reduction. Environ Sci & Policy 9(7–8):607–613

Koenig JG (1980) Indicators of urban accessibility: theory and application. Transportation 9(2):145–172

162 Nat Hazards (2013) 68:147–163

123

Langford M (2007) Rapid facilitation of dasymetric-based population interpolation by means of raster pixelmaps. Comput Environ Urban Syst 31(1):19–32

Langford M, Higgs G, Radcliffe J, White S (2008) Urban population distribution models and serviceaccessibility estimation. Comput Environ Urban Syst 32(1):66–80

Loibl W, Peters-Anders J (2012) Mobile phone data as source to discover spatial activity and motionpatterns. In: Jekel T, Car A, Strobl J, Griesebner G (eds) G1_Forum 2012: geovizualisation, societyand learning. Herbert Wichmann, VDE, Berlin/Offenbach, pp 524–533

McPherson TN, Rush JF, Khalsa H, Ivey A, Brown MJ (2006) A day-night population exchange model forbetter exposure and consequence management assessments. In: Proceedings of the 86th AMS annualmeeting, sixth symposium on the urban environment, Atlanta, GA

Mennis J, Hultgren T (2006) Intelligent dasymetric mapping and its application to areal interpolation.Cartogr Geogr Inf Sc 33(3):179–194

Montgomery MR, Stern R, Cohen B, Reed H (2003) Cities transformed: demographic change and itsimplications in the developing world. NRC, Washington, DC

Muller R, Krauß T, Lehner M, Reinartz P, Forsgren J, Ronnback G, Karlsson A (2009) IMAGE2006European coverage—methodology and results. GMES fast track land service 2006–2008, GermanAerospace Center (DLR). (further information on ESA Earthnet Online, https://earth.esa.int/web/guest/content?p_r_p_564233524_assetIdentifier=esa-datasets-6287, Accessed 12 July 2012)

OECD Organisation for Economic Co-Operation and Development (2004). Large-scale disasters: lessonslearned. Paris, France

Ozceylan D (2011) Social and economic vulnerability index development for disasters: an application oncities in Turkey. PhD Thesis, Social Science Institute, Sakarya University (in Turkish)

Ozceylan D, Coskun E (2012) The relationship between Turkey’s Provinces’ development levels and socialand economic vulnerability to disasters. J Homel Secur Emerg Manag 9(1), Article 11

Pelling M (2004) Visions of risk: a review of international indicators of disaster risk and its management.UNDP Bureau for Crisis Prevention and Recovery, Geneva

Steinnocher K, Kostl M (2007) Mapping urban functions by integrating socioeconomic information, addressdata, and remote sensing. 2007 urban remote sensing joint event. Paris, France, April 11–13, 2007

Steinnocher K, Weichselbaum J, Kostl M (2006) Linking remote sensing and demographic analysis inurbanised areas. In: Hostert P, Damm A, Schiefer S (eds) 1st EARSeL workshop of the SIG urbanremote sensing, Berlin

Steinnocher K, Kostl M, Weichselbaum J (2011) Grid-based population and land take trend indicators—newapproaches introduced by the geoland2 core information service for spatial planning. New Techniquesand Technologies for Statistics, NTTS 2011, Brussels

Tobler W (1979) Smooth pycnophylactic interpolation for geographical regions. J Am Stat As 74(367):519–530

Tsubokawa H (2006) Insurance issues of catastrophic disasters in Japan: lessons from the 2005 HurricaneKatrina disaster. In Ikeda S, Fukuzono T, Sato T (eds) A better integrated management of disasterrisks: toward resilient society to emerging disaster risks in mega-cities. TERRAPUB Terra ScientificPublishing Company, Tokyo, Japan, pp 193–198

UNDP United Nations Development Program (2004) Reducing disaster risk: a challenge for development.UNDP Bureau for Crisis Prevention and Recovery, New York

UNISDR United Nations International Strategy for Disaster Reduction (2004) Living with risk: a globalreview of disaster reduction initiatives, Geneva, Switzerland

UNISDR United Nations International Strategy for Disaster Reduction (2009) Terminology on disaster riskreduction, Geneva, Switzerland

United Nations (2006) World population prospects: the 2006 revision. Newest releases (currently 2010revision). Available at http://esa.un.org/unpd/wpp and http://unstats.un.org. Accessed 12 July 2012

Vogel C, Moser SC, Kasperson RE, Dabelko GD (2007) Linking vulnerability, adaptation, and resiliencescience to practice: pathways, players, and partnerships. Glob Environ Change 17:349–364

Xanthopoulos G (2007) Forest fire policy scenarios as a key element affecting the occurrence and char-acteristics of fire disasters. In: Proceedings of the 4th international wildland fire conference, Sevilla

Yeletaysi S, Ozceylan D, Fiedrich F, Harrald JR, Jefferson T (2009) A framework to integrate socialvulnerability into catastrophic natural disaster preparedness planning. In: Proceedings of TIEMS 2009annual conference, Istanbul, Turkey, pp 380–389

Nat Hazards (2013) 68:147–163 163

123