You are on page 1of 13

Science of the Total Environment 538 (2015) 445–457

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage:

Flood risk and adaptation strategies under climate change and urban
expansion: A probabilistic analysis using global data
Sanne Muis a,⁎, Burak Güneralp b, Brenden Jongman a, Jeroen C.J.H. Aerts a, Philip J. Ward a
Institute for Environmental Studies (IVM), VU University Amsterdam, The Netherlands
Department of Geography, Texas A&M University, College Station, USA


• We develop a probabilistic method to

assess future flood risk on a country-
• Urban expansion will drive large in-
creases in flood risk in Indonesia.
• Climate change amplifies coastal risk by
19–37%, its impact on river risk is un-
• Adaptation is effective and increasingly
urgent, regardless of climate uncer-
• Global data is used successfully for a
probabilistic risk analysis, including ad-

a r t i c l e i n f o a b s t r a c t

Article history: An accurate understanding of flood risk and its drivers is crucial for effective risk management. Detailed risk pro-
Received 15 May 2015 jections, including uncertainties, are however rarely available, particularly in developing countries. This paper
Received in revised form 12 August 2015 presents a method that integrates recent advances in global-scale modeling of flood hazard and land change,
Accepted 12 August 2015
which enables the probabilistic analysis of future trends in national-scale flood risk. We demonstrate its applica-
Available online 28 August 2015
tion to Indonesia. We develop 1000 spatially-explicit projections of urban expansion from 2000 to 2030 that
Editor: D. Barcelo account for uncertainty associated with population and economic growth projections, as well as uncertainty in
where urban land change may occur. The projections show that the urban extent increases by 215%–357%
Keywords: (5th and 95th percentiles). Urban expansion is particularly rapid on Java, which accounts for 79% of the national
Flood risk increase. From 2000 to 2030, increases in exposure will elevate flood risk by, on average, 76% and 120% for river
Adaptation and coastal floods. While sea level rise will further increase the exposure-induced trend by 19%–37%, the re-
Global modeling sponse of river floods to climate change is highly uncertain. However, as urban expansion is the main driver of
Urbanization future risk, the implementation of adaptation measures is increasingly urgent, regardless of the wide uncertainty
Climate change impacts
in climate projections. Using probabilistic urban projections, we show that spatial planning can be a very effective
adaptation strategy. Our study emphasizes that global data can be used successfully for probabilistic risk assess-
ment in data-scarce countries.
© 2015 Elsevier B.V. All rights reserved.

⁎ Corresponding author at: Institute for Environmental Studies — Instituut voor Milieuvraagstukken (IVM), VU University Amsterdam, De Boelelaan 1087, 1081 HV Amsterdam, The
E-mail address: (S. Muis).
0048-9697/© 2015 Elsevier B.V. All rights reserved.
446 S. Muis et al. / Science of the Total Environment 538 (2015) 445–457

1. Introduction addition, the increasing availability of land-use and land-cover data at

continental- to global-scales (Klein Goldewijk et al., 2011; Schneider
Over the past decades, urban areas around the world have expanded et al., 2009) has opened the way for large-scale forecasting of land-
rapidly (Seto et al., 2011; Angel et al., 2011). According to the latest change dynamics (Güneralp et al., 2015; Nelson et al., 2010), that in-
global projections, the urban population will increase from 2.8 billion clude the uncertainties of the underlying drivers (Güneralp and Seto,
in 2000 to 5 billion in 2030 (UN, 2013); and urban land cover will 2013).
triple from 2000 to 2030 (Seto et al., 2012). Nearly half of the urban ex- These developments allow for a large-scale assessment of flood risk
pansion is projected to take place in Asia, where some of the world's that integrates future dynamics of hazard and exposure, and includes a
highest concentrations of people and assets are located in low-lying quantification of uncertainties. Here, we present a first of its kind
areas and are highly prone to flooding from rivers and coastal surge country-scale flood risk assessment that integrates these methodologi-
(Hallegatte et al., 2013; Jongman et al., 2012; Nicholls and Cazenave, cal advances, using Indonesia as a case-study. We use probabilistic and
2010). Major floods can be a key threat to these cities and can cause cat- spatially-explicit projections of new urban land to 2030, which include
astrophic losses with large impacts on national economies. the uncertainties in the changes in the population and economy, the
Global flood risk has been increasing during the last few decades two major drivers of urban land change, as well as the uncertainty in
(Bouwer et al., 2007). Hereby, flood risk is defined as the combination where urban land change may occur. These projections are combined
of hazard (i.e. the frequency and intensity of floods), exposure (i.e. the with future climate projections of both river and coastal flood hazards.
population and assets located in flood-prone areas) and vulnerability We evaluate the effect of two adaptation strategies: spatial planning
(i.e. the susceptibility of the exposed elements to the hazard) (IPCC, in flood-prone areas and enhanced flood protection. This is the first
2012; UNISDR, 2011). Several studies show that increasing exposure is large-scale flood risk analysis that assesses the effectiveness of spatial
the main driver of increasing risk (e.g. Bouwer, 2011; Visser et al., planning based on probabilistic urban projections. The main aims of
2014); although recent studies show that reductions in vulnerability this paper are to:
in the past decades may have reduced the overall risk, which could
mask potential signals of climate change (Jongman et al., 2015; • Develop a methodology based on global data to assess national-scale
Mechler and Bouwer, 2014). Over the coming decades, flood risk is ex- flood risk under current and future scenarios;
pected to continue increasing due to socio-economic development and • Develop spatially-explicit and probabilistic projections of urban
climate change (Visser et al., 2014; IPCC, 2012; UNISDR, 2011). By 2030, expansion and river and coastal flood hazard projections from
40% (195,000 km2) of the global urban land is projected to be located in 2000–2030;
high-frequency flood zones, compared to 30% in 2000 (Güneralp et al., • Combine the exposure and hazard projections to assess how projected
2015). According to IPCC (2013), global sea levels are likely to rise changes in drivers contribute to future changes in flood risk;
between 0.26 m and 0.82 m by 2100 relative to 1986–2005, resulting • Assess to what extent spatial planning of new urban areas and en-
in more severe coastal floods. Studies of global-scale river flood risk hanced flood protection can offset projected increases in flood risk.
project large climate-induced increases over the coming century
(Arnell and Lloyd-Hughes, 2014; Hirabayashi et al., 2013).
Increasing global losses demand effective and efficient risk-reducing Our case-study is Indonesia, a country that already faces high flood
strategies (Aerts et al., 2014; Brown et al., 2014), whose implementation risk, and is undergoing rapid urbanization. While in 2000 about one
should be based on an accurate understanding of the drivers of risk (e.g. third of the total population lived in urban areas, by 2010 this had in-
Jha et al., 2012; Jongman et al., 2014a; Mechler and Bouwer, 2014). This creased to one half (UN, 2014). In addition, catastrophic events occur
means it is of key importance to integrate both climate and anthropo- frequently, for example, wide-spread floods in 2014 resulted in estimat-
genic drivers (Kim and Chung, 2014). Future risk projections are charac- ed direct damages of ca. USD 400 million (EM-DAT, 2013).
terized by high uncertainty, which is most frequently addressed by
using different scenarios. While a scenario-based approach is useful to 2. Materials and methods
explore the potential impact of climate change, unlike a probabilistic ap-
proach, they provide any quantitative information on the uncertainty to Here, we provide a brief overview of the applied modeling frame-
assess the relative risk of particular adaptation actions. In response to work (Fig. 1). Details of each step are given in the following paragraphs.
the debate on the need for probabilistic climate change scenarios In short, a damage model combines the three risk elements (i.e. hazard,
(Pittock et al., 2001; Schneider, 2001, 2002), various studies have exposure, and vulnerability) to compute the Expected Annual Damage
focused on quantifying the probabilities of (the impacts of) climate (EAD) expressed in million USD/year (e.g. Ward et al., 2011). Unless
change (e.g. Ward et al., 2013a; Webster et al., 2012; Purvis et al., stated otherwise, all values in this paper are expressed in values of the
2008; New et al., 2007). The uncertainties related to projected year 2000 using GDP deflators of the World Bank (2014). By simulating
socio-economic development have received less attention in the flood the flood damages associated with certain exceedance probabilities, we
risk community, and the few studies that integrate future climate can calculate the EAD by estimating the area under an exceedance prob-
and socio-economic projections are based on discrete scenarios (e.g. ability loss curve (i.e. risk curve). The EADs are aggregated on a provin-
Hinkel et al., 2014; Hallegatte et al., 2013; Penning-Rowsell et al., cial level.
2013; te Linde et al., 2011). We use two model cascades to simulate flood hazard; river floods
Probabilistic estimates of current and future risk that include the are based on GLOFRIS (Winsemius et al., 2013; Ward et al., 2013b),
dynamics of hazard, exposure, and vulnerability are therefore rare and coastal floods are based on the DIVA database (Vafeidis et al.,
(e.g. Aerts et al., 2014). This is especially true for developing countries, 2008). Probabilistic urban projections based on Güneralp and Seto
where the data needed to conduct such analyses are scarce. With the in- (2013) and Seto et al. (2012) are combined with estimates of economic
creasing availability of hydrological datasets with continental to global- value per surface area to express economic exposure. Vulnerability is
coverage and new generations of flood models, several models for represented by a depth-damage curve taken from Budiyono et al.
assessing flood hazard at the global-scale have been developed in recent (2014). We do not include flood protection. There are few empirical
years (e.g. Hinkel et al., 2014; Hirabayashi et al., 2013; Winsemius et al., data available to estimate protection standards in Indonesia, but gener-
2013). Although there are constraints due to the relatively low spatial ally protection standards can be considered to be low (e.g. Hallegatte
resolution of the data, particularly related to the global elevation et al., 2013; Ward et al., 2010) since damaging floods occur frequently
datasets available, a major benefit of this development is that it enables (EM-DAT, 2013; Marfai and King, 2007). Although absolute flood risk
large-scale analysis of floods, even in countries where data are scarce. In estimates are associated with large uncertainties (Apel et al., 2008;
S. Muis et al. / Science of the Total Environment 538 (2015) 445–457 447

Fig. 1. Flowchart showing the general methodological approach where the three risk components (hazard, exposure, and vulnerability) are combined in a damage model.

Merz et al., 2010), our main interest is in relative changes, which are Ward et al., 2013b). This results in hazard maps showing inundation
more robust than absolute estimates (e.g. Bubeck et al., 2011). extent and depth (30″ × 30″) for nine return periods: 2, 5, 10, 25,
50, 100, 250, 500, and 1000 years. Flood volumes are simulated
2.1. Hazard with PCR-GLOBWB-dynRout, a global distributed hydrological
model with a resolution of 0.5° × 0.5° (approximately
2.1.1. Coastal floods 50 km × 50 km at the equator) (Bierkens and van Beek, 2009; van
We calculate coastal inundation using a GIS-based planar approach, Beek and Bierkens, 2008). To simulate floods under historical climate
which uses extreme sea levels and a Digital Elevation Model (DEM) as conditions (1960–2000), we force the PCR-GLOBWB model with the
inputs. We define inundated areas as areas that have an elevation climate reanalysis dataset EU-WATCH (Weedon et al., 2011). For
lower than the extreme sea level and a direct connection to the sea, each grid cell, the different return periods of flood volumes are ex-
following Hanson et al. (2010). We derive extreme water levels tracted based on a Gumbel extreme value distribution, as described
corresponding to return periods of 10, 100 and 1000 years from the in Ward et al. (2013b). Subsequently, these flood volumes and the
DIVA (Dynamic Interactive Vulnerability Assessment) tool (Hinkel and HydroSHEDS DEM are used as input to the GLOFRIS downscaling rou-
Klein, 2009; Vafeidis et al., 2008). The water levels vary spatially across tine (Winsemius et al., 2013). River flooding is considered to play a
the 426 coastal segments in Indonesia, and are linked to the occurrence role only in streams characterized by a Strahler order of 5 or larger
of high tides and storm surges. Coastal inundation due to tropical and only flood levels corresponding to a return period of two years
cyclones, tsunamis and waves are not included. The DIVA segments or higher are assumed to lead to overbank flooding (e.g. Dunne and
are linked to the DEM using the Euclidean Allocation tool from ArcGIS. Leopold, 1978).
The DEM is derived from the HydroSHEDS database and is based on To simulate the impact of climate change on river floods, we use the
high-resolution elevation data (30″ × 30″, approximately 1 km × 1 km bias-corrected future climate projections from the Inter-Sectoral Impact
at the equator) obtained during NASA's Shuttle Radar Topography Mis- Model Intercomparison Project (ISI-MIP; Hempel et al., 2013) which
sion and corrected for hydrological applications (Lehner et al., 2008). contains projections from five global climate models (GCMs), each
Since regional sea level rise (SLR) projections are not available for driven by the four different RCPs from the IPCC. Next, we force
Indonesia, we use the most recent global projections from IPCC (2013) PCR-GLOBWB with each of these 20 different simulations, providing
for four different Relative Concentration Pathways (RCP2.6, RCP4.5, time-series of flood volumes for the period 2010–2049 (referring to
RCP6.0, and RCP8.5; see Moss et al., 2010). By 2100, these pathways cor- the climate in 2030). We then use the same model cascade as for the
respond to increases of global mean sea level with 0.44 m, 0.53 m, historical projections, resulting in inundation maps for each return peri-
0.55 m, and 0.74 m, respectively. For 2030, the SLR projections do not od and GCM-RCP combination. Because the bias-correction method
deviate much for the different RCPs, we therefore apply the median used in ISI-MIP cannot remove bias in the distribution of rainfall days
and likely ranges of SLR, which correspond to low, mean, and high SLR across the month, as well as interannual variability, we carried out a fur-
estimates of 0.09 m, 0.12 m, and 0.17 m respectively (Table AII7.7 in ther bias correction on the EAD values, using the delta-method:
IPCC, 2013). Following Menéndez and Woodworth (2010), we assume
no changes in storminess and the frequency of extreme events and EAD ¼ a EADEU−WATCH
therefore SLR is added to the extreme water levels under current cli-
mate (Nicholls et al., 2011).

2.1.2. River floods where EAD⁎ is the bias corrected risk and a the delta change factor a,
We generate inundation maps for river flooding using the hydrolog- which is equal to EADggcm,rcp/EADgcm,hist. In Supplementary Material A
ical and hydraulic components of GLOFRIS (Winsemius et al., 2013; we show the effect of applying a bias correction on our results.
448 S. Muis et al. / Science of the Total Environment 538 (2015) 445–457

2.2. Exposure 2.2.2. Economic exposure

To represent economic exposure for one uniform urban land-use
We represent flood exposure by a map of economic exposure per class, we use the (spatially) weighted average of the exposure estimates
urban grid cell (30″ × 30″). Such a map requires two inputs: (1) a for 12 different urban land-use classes in Jakarta reported by Budiyono
map indicating urban extent, i.e. a map of urban grid-cells; and (2) the et al. (2014). The economic exposure in Jakarta is 6.7 million USD/
economic value of these urban grid-cells. These two inputs are de- km2. For the other Indonesian provinces, we assume that regional dif-
scribed below. ferences in economic exposure are proportional to regional differences
in GDP per capita in 2010 (BPS, 2013), similar to Jongman et al.
2.2.1. Urban extent (2012). To estimate future economic exposure, we assume that these
The initial land cover map for ca. 2000, is derived from the Moderate will increase proportionally to the national projections of growth in
Resolution Imaging Spectroradiometer (MODIS) v5 land cover product, GDP per capita. Using the mean of the probabilistic projections of GDP
which has a horizontal resolution of 500 m × 500 m (Friedl et al., 2010), and population derived for the urban simulations, we calculate the
resampled to 1 km × 1 km resolution. mean increase in GDP per capita relative to 2000, resulting in a factor in-
We develop probabilistic and spatially-explicit projections of crease of 2.2, 2.9, 3.6 and 4.3 for the years 2015, 2020, 2025 and 2030,
urban expansion in Indonesia for the years 2015, 2020, 2025, and respectively. Since there are no regionally disaggregated GDP projec-
2030. The two major drivers of urban expansion are population and tions for Indonesia, we assume that the relative distribution of GDP
economic growth (Angel et al., 2011). Individual projections of per capita across regions remains constant during the study period.
both drivers are associated with large uncertainties and are prone
to error (Alho, 1997; Cohen, 2004; Pritchett, 2000), which are 2.3. Vulnerability
reflected in urban forecasts (Güneralp and Seto, 2013; Seto et al.,
2012). To account for the uncertainties in drivers, we generate Vulnerability is represented by a depth-damage function (or vulner-
1000 simulations of new urban land area, based on probabilistic pro- ability curve) that estimates the damage for a given inundation depth.
jections of urban population and gross domestic product (GDP). To Based on expert meetings and a workshop, Budiyono et al. (2014)
obtain probabilistic projections of total population, we combine the developed curves for 12 different land-use classes in Jakarta. To convert
UN World Population Projections 2012 Revision (UN, 2013) with un- these into one curve for the urban land-use class, we take the area-
certainty estimates published by U.S. National Research Council weighted average based on the spatial distribution of the different
(NRC, 2000) and use this information to fit a probability density land-use classes within Jakarta (Fig. 2), similar to Huizinga (2007). By
function (PDF) (Supplementary Material B). To get the urban popu- using this area-weighted curve, we assume that the land-use distribu-
lation, we then draw 1000 values from the PDF using Latin Hyper- tion in Jakarta is representative of the distribution in other urban areas
cube (LHC) sampling and multiply them by 0.63, which is the in Indonesia, and that these urban areas have an equal vulnerability to
projected urban population proportion (UN, 2012). For GDP, we as- floods. While vulnerability may change in time (Bouwer, 2013; Visser
sume a uniform distribution using the minimum and maximum esti- et al., 2014), there is no quantitative information available on how this
mate from country-level GDP projections from the four IPCC SRES may affect urban damages. Hence, we assume the vulnerability curve
scenarios (Nakicenovic et al., 2000). The SRES scenarios have explic- does not change in time. Other flood characteristics than inundation
itly not been appointed any likelihood; however, they cover a wide depth, such as flow velocity, duration, water quality, and precautionary
range of plausible economic futures and hence we argue that they measures play on important role in the amount of flood damage (e.g.
can be used as a basis for exploring a realistic set of future projec- Merz et al., 2010). Although with large uncertainties (e.g. De Moel
tions. The minimum and maximum estimates range from ca. 600 bil- et al., 2014; Merz and Thieken, 2009), research shows that flood damage
lion USD for the A2 scenario and 1.4 trillion USD for the A1 scenario to residential buildings is strongly dependent on the water depth of a
(1990 values). We again employ LHC sampling to draw 1000 values flood, and depth-damage curves are the standard approach to assess
from the resulting uniform GDP distribution for 2030 and divide urban flood losses (Smith, 1994). The curves developed by Budiyono
them by the urban population projections for the same year, which et al. (2014) based on river floods, are the only curves developed for
provides us with 1000 projections of GDP per capita. To calculate Indonesia. Damages for river and coastal flood may differ; however, as
urban land expansion due to the urban population increase only, the curves are based on water depth only, we assume the same vulner-
we first overlay the urban extent map and the population density ability is applicable to coastal floods as well.
map and calculate the country's average urban population density
ca. 2000, which was 272.9 m2/person. We divide each value in the 2.4. Adaptation strategies
set of 1000 urban population estimates by the urban population den-
sity. Then, we derive 1000 projections of per capita GDP in 2030 We evaluate two adaptation strategies: (1) spatial planning of new
using the probabilistic projections of the country's GDP and popula- urban areas; and (2) enhancing flood protection. The first adaptation
tion. We incorporate the increase in per capita urban land due to the strategy aims to limit the amount of new urban land in flood-prone
increase in per capita GDP using the observed relationship between areas by enforcing building restrictions. We carry out urban simulations
the two at country level (Supplementary Material C). This enables in which urban expansion into flood-prone areas is either restricted or
us to estimate the total area of new urban land. limited. In these simulations, we define flood-prone areas as areas ex-
The new urban land is spatially allocated using a grid-based land posed to either coastal or river flood with a return period of 100 years.
change model based on GEOMOD (Pontius et al., 2008), and described For grid-cells in these areas, we reduce the suitability value (thus lower-
in Seto et al. (2012) and Güneralp and Seto (2013). The required inputs ing the chance of these locations to urbanize) to 0, 1/2, and 1/3 of their
are summarized in Supplementary Material D. Through statistical baseline suitability values to reflect full, medium, and weak enforce-
analysis of the empirical patterns created by the overlay of the initial ment of the building restrictions, respectively. Subsequently, we calcu-
land cover map with the driver maps (roads, slopes etc.), the model late the EAD for these three levels of enforcement. The second
estimates overall suitability of each grid cell for change. Based on adaptation strategy is the enhancement of flood protection. Flood de-
these inputs, the model spatially allocates new urban areas. For each fences are designed to resist water levels that correspond to a given re-
new urban cell to be allocated, there may be more than one equally suit- turn period. We assess the effectiveness of three different protection
able locations and the model randomly picks one of these. Thus, the spa- standards, corresponding to floods with return periods of 10, 50 and
tial configuration of urban expansion may be different between any two 100 years. We incorporate flood protection in the model framework
realizations with the same amount of projected increase in urban land. by truncating the risk curve at the exceedance probability assumed by
S. Muis et al. / Science of the Total Environment 538 (2015) 445–457 449

Fig. 2. Vulnerability curves for various urban land classes in Jakarta from a series of expert meetings and a workshop described in Budiyono et al. (2014) and the weighted average curve
(black line) used in this study.

the protection standard (Feyen et al., 2012), and estimating the EAD as with a return period of 100 years or higher. We further discuss im-
the area of the remaining part of the risk curve. We do not include dike proving the global data needed to assess coastal flood hazard in
failures at return periods shorter than the design standard. Section 3.6.

3. Results and discussion

3.2. Current flood risk
3.1. Validation of discharge and surge levels
For 2000, the simulated EAD is 31 million USD for coastal floods,
3.1.1. River floods and 75 million USD for river floods; this corresponds to 0.01% and
The (extreme) discharges from the PCR-GLOBWB/GLOFRIS 0.02% of Indonesia's GDP, respectively. Reported data on historical
modeling cascade have been validated in previous studies (Van flood losses are scarce in Indonesia, particularly for coastal floods.
Beek et al., 2009; Ward et al., 2013a, 2013b). Globally there is a rea- While this is the reason for conducting this analysis, it makes verifi-
sonable agreement with observed discharge (within 25% relative cation of our results difficult. For river floods, the disaster database
error for the majority of locations), although there are large regional EM-DAT (2013) reports total damages of 5.5 billion USD from 1990
differences. Validating the model in Indonesia is complicated, due to to 2013. The majority of this total results from a single 3 billion
the existence of many small rivers combined with the relatively USD event in January 2013 in Jakarta. The huge damages are mainly
coarse resolution of the discharge data (0.5°). Moreover, observed due to the collapse of a 30 m long section of the Western Flood
daily discharge data are not readily available. We verify whether Canal, which led to major flooding in Jakarta's main business district.
the modeling cascade is applicable in Indonesia by comparing the Such dike collapses are not included in our modeling framework. The
observed mean maximum discharge with the simulated mean annu- return period of this flood event is estimated to be much lower than
al maxima discharge (1960–1999) of the nearest PCR-GLOBWB cell several previous events that caused much less damage (Sagala et al.,
(Supplementary Material E1). This data is only available near Jakarta, 2013). In addition, our estimates are based on average urban dam-
where there are two main rivers: the Ciliwung River, which has a ages, which will underestimate damage in the capital's main busi-
specific discharge of around 100 m 3/s/100 km 2 (Tachikawa et al., ness district. If we exclude this event, the annual average of
2004) and the Citarum, which has as specific discharge in the order reported losses over 1990–2013 are 121 million USD. This is higher
of 3 m3/s/100 km2 (Takeuchi et al., 1995). This totals to a specific dis- than our EAD estimate of 75 million USD, but is of the same order
charge of approximately 250 m3/s/0.5°, which is in the same order as of magnitude. The lower estimate could be because we only model
the PCRGLOW-WB cell that has a value of 263 m3/s/0.5°. urban damages, while reported damages from EM-DAT also include
rural losses. Furthermore, most floods in Indonesia's lowlands are as-
3.1.2. Coastal floods sociated with rivers that do not freely drain to the sea because of high
To assess whether the extreme surge levels in DIVA are applicable coastal water levels. These interactions between river runoff and sea
in Indonesia, we compare the dataset against sea level observations levels are not included in our model framework.
from the University of Hawaii ( For While river flood risk is almost 2.5 times higher than coastal flood
all 13 stations that have more than 5 years of observations, we risk, the EAD for river floods is more equally divided between the differ-
estimate the extreme values based on the annual maxima method ent regions (Fig. 3A–C). The largest share of the national risk is located
(Gumbel, 1941). Supplementary Material E2 provides detailed on Java, which accounts for 40% of river flood risk and 56% of coastal
information per station. On average, DIVA overestimates the 10-yr flood risk. With a share of 42%, Sumatra has the second largest share
return period with 0.8 m. Over all stations the differences range of coastal flood risk. Although there are no land-locked provinces,
from − 0.4 to 2.0 m. Unfortunately none of the stations has an obser- there are a number of provinces for which our simulations result in no
vation record long enough for validating the more extreme sea levels coastal flood risk.
450 S. Muis et al. / Science of the Total Environment 538 (2015) 445–457

Fig. 3. EAD for the different Indonesian provinces and the regional share of the national risk for 2000 and 2030. For river floods, we show risk for the baseline scenario (A) and the risk
projection under the mean ensemble of all GCMs and RCP6.0 (B). For coastal floods, we show risk for the baseline projection (C) and the risk projections for mean SLR (D). The colors
of the seven main regions in Indonesia correspond to the colors in the pie charts.

3.3. Future trends in risk drivers scenarios lead to an increase in inundated area by 2%, 4%, and 7%, re-
spectively. Under the high SLR scenario of + 0.19 m, the majority of
3.3.1. Urban expansion by 2030 the increase in inundation extent takes place in four provinces: East
In 2000, the extent of urban area in Indonesia was about 11,000 km2 Java, Banten, and West and East Kalimantan, which account for respec-
(0.6% of total land), of which 67% was located on Java (5.3% of total land tively 17%, 20%, 28% and 13% of the increase.
area of the island). Our urban projections suggest a rapid increase in
urban extent by 2030 (Fig. 4) that ranges from 215% to 357% (5th and 3.4. Future changes in flood risk
95th percentiles) with a mean of 277%. Java accounts on average for
79% of the urban expansion; implying that by 2030 approximately 3.4.1. River floods
75% of total urban land in the country could be located on that island River flood risk is projected to increase rapidly from 2000 to 2030
alone. Other parts of Indonesia are also expected to experience rapid (Fig. 6A), although there is wide uncertainty in the results. Assuming
urban expansion, albeit with lower rates than Java (Supplementary changes in exposure only (red line in Fig. 6A), the mean increase in
Material F). In Sumatra, Kalimantan, and Sulawesi, urban extent in- EAD due to urban expansion is 638% (523% and 772% for the 5th and
creases on average by 195%, 144% and 147%, respectively. Maluku and 95th percentiles, respectively).1 When separating the drivers responsi-
Western New Guinea are expected to experience relatively low rates ble for this increase, on average 165% is due to urban expansion and
of urban expansion. 179% is due to an increase in economic exposure in existing urban
areas. The increase in economic exposure in future urban areas accounts
3.3.2. Changing flood hazard due to climate change for the remaining 293% of the increase. We project the EAD to increase
most rapidly on Java, whose share of national risk increases from 40% Impact of climate change on river floods. There are large spatial to 46% from 2000 to 2030 (Fig. 3A–B).
differences in the projected changes in river flood hazard. Fig. 5A When comparing national flood losses over time it is also useful to
shows the mean magnitude of changes in flood volumes with a normalize risk for future changes in GDP, which is shown in Fig. 6B. In
100-year return period across the 20 different GCM-RCP combinations. this case, the EAD increases from 0.02% GDP in 2000 to 0.03% GDP in
Floods are projected to become more severe in parts of Sumatra, 2030 (0.03% and 0.04% for the 5th and 95th percentiles). If we include
Kalimantan, Sulawesi, Maluku Island, and Papua, while for Java the climate change projections, the normalized risk projections for 2030
mean signal shows a decrease in flood volume. However, there is wide range from −6% to +238% compared to 2000 (−6% is the 5th percen-
uncertainty across the projections. This is illustrated by the number of tile of the climate projection resulting in the lowest EAD, and +238% is
projections that agree on the sign of change (Fig. 5B). the 95th percentile of the climate projection resulting in the highest
EAD). Assuming a change in hazard only (i.e. keeping future exposure
constant), risk changes between − 37% and + 91% over the period Effect of sea level rise on coastal floods. The effect of SLR will be an
increase in the inundation extent and depth for each return period. For a 1
The reported results in terms of percentiles are calculated based on the set of 1000
flood with a 100-year return period, the low, mean, and high SLR projections for urban expansion combined with the different climate projections.
S. Muis et al. / Science of the Total Environment 538 (2015) 445–457 451

Fig. 4. The probabilities of urban expansion by 2030 derived by stacking all 1000 urban expansion projections.

2000–2030 (Table 1). The mean ensemble of all climate projections re- positive, which implies that flood risk will become increasingly costly
sults in a change of 1%. However, the rapid increases in exposure mean relative to GDP. Assuming changes in exposure only (red line in
that there is a large certainty that future risk will increase, irrespective Fig. 6D), we project an increase of between 80% and 183% (5th and
of climate change. Only 3 out of the 20 climate projections indicate a 95th percentiles). The mean increase in coastal EAD is equal to 120%.
small decrease in normalized risk from 2000 to 2030, and only when This is 1.6 times higher than the increase in river EAD, indicating that
combined with the urban projections in the lowest 20th percentile. urban expansion preferentially takes place along the coastline. This is
This highlights that the implementation of risk-reducing measures consistent with earlier findings (e.g. Jha et al., 2012; Jongman et al.,
remains urgent, regardless of the uncertainty in climate change 2012). From 2000 to 2030, SLR amplifies the exposure-induced risk in-
projections. crease (normalized) by, on average, 19%, 25%, and 37% based on the low,
mean, and high SLR projections, respectively. The impact of SLR be-
3.4.2. Coastal floods comes more apparent over time, as demonstrated by the divergence
For coastal floods, all risk projections show strong increases of projections with and without SLR (Fig. 6D). Hence, beyond 2030,
(Fig. 6C&D). Considering changes in exposure only (red line in SLR may become a more dominant driver. Our results are consistent
Fig. 6C), absolute risk more than doubles from 2000 to 2015 with a with other studies (e.g. Thieken et al., 2014; Storch and Downes,
mean increase of 238%. By 2030, the mean increase is equal to 841% 2011) that also conclude that the impact of climate change in the near
(673% and 1029% are the 5th and 95th percentiles, respectively). future will most likely be relatively minor compared to that of rapid ur-
Assuming changes in economic exposure and flood hazard, risk banization and economic growth.
increases between 751 and 1269% (5th and 95th percentiles). On a
national scale, urban expansion accounts for almost half of the total 3.5. Effectiveness of adaptation strategies
risk increase (Table 1), while increasing exposure accounts for approx-
imately one-fifth. SLR has a minor contribution (ranging from 0.8% to We evaluate the potential of two adaptation strategies: spatial plan-
1.8%), however, the three risk drivers reinforce each other, for example ning of new urban land; and enhanced flood protection. The effective-
when urban expansion takes place in areas that becomes prone to ness of the adaptation measures is defined as the reduction in EAD
flooding with SLR. The EAD increases most rapidly on Java, whose re- compared to projections without adaptation.
gional share of national risk increases from 56% in 2000 to 72% in
2030 (Fig. 3C–D). Coastal risk is also projected to increase in provinces 3.5.1. Spatial planning
for which we simulate no coastal flood risk under current climate, The increase in EAD can be offset by limiting the amount of new
such as South Sulawesi and West Papua. urban land in flood-prone areas, which can be achieved by enforcing
Once risk is normalized for GDP (to account for economic growth), building restrictions. For full enforcement of the restrictions, there is a
the increasing trend over 2000–2030 is less steep, though still strongly risk reduction of 50%–57% (5th and 95th percentiles) for river floods
452 S. Muis et al. / Science of the Total Environment 538 (2015) 445–457

Fig. 5. Changes in flood volume with a 100-year return period under climate change based on 20 different combinations of 5 GCMs and 4 RCPs. We show the mean magnitude of change
over the simulation ensemble (A) and the number of projections (out of 20) that agree on the sign of change (B).

(mean climate ensemble for RCP6.0; Table 1; Fig. 7A). For weak enforce- standards also greatly reduces coastal risk. A protection standard of
ment, the risk reduction ranges from 42% to 58% (5th and 95th percen- 10-year return period reduces flood risk, on average, by 63%, while a
tiles). For coastal risk, full enforcement result in a reduction of 76% to protection standard of 100-year reduces risk by 95%. Compared to spa-
84% (5th and 95th percentiles; Fig. 7B). Even for weak enforcement, tial planning, the effectiveness of flood protection has less regional var-
the reduction is still 73% (95th percentile). Spatial planning is most ef- iation (Supplementary Material G; Figs. E&D). This is because whereas
fective in provinces where urban expansion is the prime driver of in- spatial planning reduces the projected increase in flood exposure
creases in flood risk (Supplementary Material G; Figs A&C). (with no impact on current risk), flood protection reduces current risk.
The uncertainty in the EAD projections increases with lower levels of With this analysis, we show that enhanced flood protection can be a
enforcement (Fig. 7). This is because, with lower levels of enforcement, highly effective strategy. This is in agreement with local studies in Ho
urban expansion is still allowed to take place in flood prone areas, and Chi Minh City (Lasage et al., 2014) and Rotterdam (De Moel et al.,
thus the risk reduction depends on the particular urban expansion pat- 2014) that also demonstrate that the implementation of flood protec-
terns that emerge in each individual urban projection. Spatial planning tion may offset the increase in risk. More research is however required
is more effective for coastal flood risk than river flood risk due to more to assess the costs of the construction and maintenance of such mea-
rapid urban expansion in coastal areas. While we demonstrate that sures (e.g. Aerts et al., 2014; Jongman et al., 2014a). It is likely that
spatial planning can potentially be very effective, we did not assess flood protection is financially feasible only in high-risk areas; and
the feasibility of its actual implementation. A full restriction on urban even then, the implementation may exceed the institutional and finan-
development may not be feasible; we however argue that the land cial capabilities of many countries that face high flood risk. Furthermore,
may be utilized in a way that is compatible with flooding. This means re- here we analyze the two adaptation strategies separately, while the
ducing the vulnerability of new urban land in flood-prone areas, which most effective way is likely a combination of different strategies.
could be achieved by enforcing building codes, such as wet- or dry
proofing of building (Kreibich et al., 2009; Poussin et al., 2012), zoning 3.6. Implications for flood risk assessment and management
of land-use functions (Koks et al., 2013), or elevating an area of individ-
ual buildings (e.g. De Moel et al., 2013; Lasage et al., 2014). This is the first flood risk analysis that assesses future trends in risk
drivers and explicitly quantifies the uncertainties in socio-economic
3.5.2. Enhanced flood protection projections, other than using discrete scenarios (e.g. Hinkel et al.,
If we assume a relatively low flood protection standard of a 10-year 2014; Hallegatte et al., 2013; Penning-Rowsell et al., 2013; te Linde
return period uniformly throughout all urban areas in Indonesia, river et al., 2011). Probabilistic urban projections quantify the uncertainty as-
flood risk is reduced by 53% on average (Table 1). This emphasizes sociated with future projection, and as such they can provide more de-
that floods with a relatively low return period are responsible for the tailed information. This may result in a better understanding of the
majority of the EAD. A protection level corresponding to a 100-year dynamics of future risk, and thus better-informed decision-making.
return period could reduce the EAD by 93%. Raising the protection For example, in this case-study for Indonesia, we show that regardless
S. Muis et al. / Science of the Total Environment 538 (2015) 445–457 453

Fig. 6. Projections of changes in flood risk (EAD; expected annual damage) between 2015 and 2030. River flood risk is shown under 20 different projections of climate change (5 GCMS and
4 RCPs) and for all projections of urban expansion. Absolute values are shown in (A), while values normalized to GDP are shown in (B). The light red shaded band shows the 5th–95th
percentiles for the projections with no climate change (i.e. urban expansion only). The light orange shaded band shows the 5th–95th percentiles over the lowest and highest risk projection
when the urban projections are combined with the 20 climate change projections. Coastal flood risk is shown under different scenarios of sea level rise (SLR) and for all projections of urban
expansion using absolute values (C) and normalized to GDP (D). The red shaded band shows the 5th–95th percentiles for the projections with no SLR (i.e. urban expansion only). The blue
shaded band shows the 95th percentile and 5th percentile when the urban projections are combined with high SLR.

of the wide uncertainty in river hazard projections, it is very likely that management, both in Indonesia and more widely. For example, due to
flood risk will increase by 2030 as urban expansion is the main driver of more rapid urban expansion in the coastal zone, coastal flood risk in-
future risk. Assessing the likelihood of various levels of risk would have creases more rapidly than river flood risk. SLR aggravates this trend,
been impossible to assess without the use probabilistic scenarios. whereas the climate change trends for river floods are highly uncertain,
Jointly considering the future trends in river and coastal flood risk on and although unlikely, also a decrease in risk is possible. While climate
a country-scale demonstrates some key implications for flood risk change may imply that a certain increase in flood hazard is inevitable
454 S. Muis et al. / Science of the Total Environment 538 (2015) 445–457

Table 1 sources of uncertainty (Güneralp and Seto, 2013; Hall, 2007). The
The effect of drivers on flood risk from 2000 to 2030 and the effectiveness of adaptation assigned probabilities are subjective as it is impossible to estimate
measures based on changes in EAD in absolute values. If based on probabilistic projections,
mean values are shown, with the 5th and 95th percentile values for the urban projections
how much uncertainty remains unquantified (e.g. Schneider, 2002;
shown in brackets. For river floods, the climate change projection shows the mean ensem- Dessai and Hulme, 2004). Nevertheless, future projections involve
ble of the 20 different combinations of GCMs and RCPs, while the minimum and maximum large uncertainties and although discrete scenarios can account for un-
projections are shown in brackets. For coastal floods, we show the mean SLR projection certainty (for example by using a low, mean and high estimate) they
with the values for the low and high SLR projections in brackets. The combined impact
provide a limited representation of the underlying sources of uncertain-
of the drivers shows the 5th percentile of the lowest projection and the 95th percentile
of the highest projection. For the evaluation of adaptation strategies, the values are shown ty. Thus, the use of probabilistic projections in flood risk analysis pro-
for the mean GCM ensemble for RCP6.0 for river floods, and for mean SLR for coastal vides additional, quantitative information that is complementary to
floods. the information provided by discrete scenarios.
River Coast

Baseline risk projection EAD (million USD) 3.7. Limitations and future research directions
75.0 30.9
There are several limitations intrinsic to various parts of the analysis.
Change in risk due to single drivers Change in EAD (%) First, we emphasize that while the global modeling of river floods has
and their combined impact
advanced rapidly in recent years (e.g. Arnell and Lloyd-Hughes, 2014;
Economic growth 122 182 Ward et al., 2014; Hirabayashi et al., 2013), global-scale modeling of
Urban expansion 166 (125-215) 445 (348-554)
coastal flood hazard has received less attention. To the best of our
Climate change 1 (-37-91) 11 (8-17)
All 651 (305-1395) 971 (751-1269) knowledge, the DIVA database (e.g. Hinkel et al., 2014) is the only avail-
able global dataset of extreme water levels. While the DIVA database
Effectiveness of adaptation measures Reduction in EAD relative to no
adaptation (%)
contains extreme water levels, and has been instrumental in global-
scale coastal flood risk studies: the validation of the model has received
Spatial planning
Low enforcement 33 (22-43) 65 (56-73)
little attention. In addition, the limited number of return periods used to
Medium enforcement 50 (42-58) 76 (70-81) construct the risk curve may cause overestimation of risk (Ward et al.,
High enforcement 57 (50-64) 80 (76-84) 2011). Hence, there is a need for improved global datasets of coastal
Flood protection water-level extremes. This could be achieved by the application of glob-
Low protection level 53 63
al hydrodynamic models with high resolution meteorological data and
Medium protection level 86 91
High protection level 93 95 bathymetry in order to produce time-series and extract return periods
of storm surges worldwide (Lowe et al., 2010). While this has been car-
ried out regionally – for example by Haigh et al. (2013) for the
Australian coastline – an analysis on a global-scale has not yet been con-
over the coming decades, large increases in flood exposure can be con- ducted. This would be particularly beneficial for risk assessments in de-
trolled by effective local governance. This means that there is a clear role veloping countries, where most of the urbanization is expected to take
for local authorities to enforce land-use regulations and prevent urban place, yet, for which national data and observations of flood levels are
development in flood-prone areas. We demonstrate how probabilistic often non-existent.
urban projection can be used to explore the effectiveness of spatial plan- Second, our analysis applies a global DEM derived from the SRTM
ning for different levels of enforcement. Results show that spatial plan- mission. Over the entire world, the SRTM has a vertical error in the
ning can be a highly effective strategy, particularly in rapidly urbanizing range of ∼ 7 m (Rodríguez et al., 2006). Since the STRM is surface
areas. Furthermore, higher levels of enforcement reduce the uncertainty model, the elevation are often underestimation due to land cover
of where urban expansion is to take place by confining urban expansion (e.g., forest or built environment), which will lead to an underestima-
in a relatively smaller area. tion of flood extent (Baugh et al., 2013), particularly in a tropical region
We emphasize that while the probabilistic urban projections quanti- like Indonesia. Also the horizontal resolution of the used DEM
fy the uncertainty associated with economic and population growth, as (1 × 1 km2) is hampering an accurate modeling of floods as it is not of
well where urban expansion is to take place, we cannot account for all sufficient detail to resolve all terrain features that control flooding

Fig. 7. Cumulative probability density functions showing the effect of spatial planning for three different levels of enforcement for the year 2030. For the sake of clarity, we have chosen not
to show the results for all RCPs, but only for RCP6.0. (A) River flood risk for the mean ensemble of all GCMs for RCP6.0. Dashed lines show the individual curves for individual GCMs;
(B) Coastal floods under mean SLR.
S. Muis et al. / Science of the Total Environment 538 (2015) 445–457 455

(Schumann et al., 2014). To conclude, the uncertainty of global elevation which we used as our case study. Moreover, by using probabilistic pro-
data is large, which will have a large influence on both our river flood in- jections, we are able quantify the uncertainties associated with the
undation modeling using GLOFRIS and particularly on the coastal flood drivers of future risk and adaptation. We find that, by 2030, urban ex-
inundation modeling applying a planar approach. pansion is the main driver of increasing flood risk (changes range
Third, the analysis could be improved by the use of regional projec- from +125 to +215% for river floods and from +348 and +554% for
tions of climate change. Climate projections based on GCMs, as used in coastal floods), particularly on Java and along the coasts. Whereas SLR
this paper, involve large uncertainties. Developing regional climate pro- will exacerbate this upward risk trend (changes from SLR alone range
jections could lead to a better understanding of future changes in cli- from +8 to +17%), the impact of climate change on river floods is high-
mate, and hence runoff. The same applies for SLR projections: since ly uncertain (changes from climate change alone range from − 37 to
regional projections of SLR are not available for Indonesia, we applied +91%). However, the increases in exposure are so large that even if cli-
the global mean SLR projection from the IPCC (2013). However, SLR is mate change were to reduce the severity of river floods, the overall im-
not uniform across the globe and according to Han et al. (2010) pact will still be an increase in flood risk. Hence, the implementation of
Indonesia may experience significantly more SLR than the global mean. effective risk-reducing measures remains urgent, regardless of the un-
Fourth, while land subsidence is a major driver of flood risk in many certainty in climate change projections. Under the scenarios used the
parts of the world (Syvitski et al., 2009), including highly populated potential risk reduction ranges from 22–85% and 53–95%, which
coastal areas in Indonesia (Abidin et al., 2011; Chaussard et al., 2013), shows that spatial planning and enhancing flood protection can be
we do not consider this process. Estimates for subsidence rates are avail- highly effective strategies. Spatial planning is more effective for river
able only for a few locations in Indonesia; moreover, the estimation of floods than for coastal floods (because urban expansion is more rapid
future projections of subsidence on a national-scale is complex and in- in the coastal zone) and uncertainties in flood risk reduce with stronger
volves large uncertainties. We stress that global- to national-scale levels of enforcement.
flood risk assessments would benefit from the quantification of subsi-
dence rates for large regions, for example, based on satellite radar inter- Acknowledgments
ferometry (e.g. Erban et al., 2014).
Fifth, there are limitations in the urban damage assessment. The as- The research leading to these results has received funding from the
sumption that differences in economic exposure are proportional to dif- European research project RISES-AM (grant agreement no. 603396).
ferences in GDP per capita does not fully capture regional differences in This research has also been partly funded by the Dutch research pro-
damages (Jongman et al., 2012). It accounts for the fact that higher gram Knowledge for Climate and Delta Alliance research project
values for economic exposure are more likely to be encountered in HSINT02a (Jakarta Climate Adaptation Tools). P.J.W. received
areas with a higher GDP per capita; however, this results in high eco- additional funding from the Netherlands Organisation for Scientific
nomic exposure for provinces that have a low population density and Research (NWO) in the form of a VENI grant (grant no. 863-11-011).
a high regional GDP due to oil production, such as East Kalimantan We thank Gil Pontius for providing the source code for the land
(Fig. 3). In addition, the density of buildings within areas classified as change model, GEOMOD, and the Texas A&M Supercomputing Facility
urban may differ greatly (Jongman et al., 2014b). The urban projections for providing computational resources that have been used for the
are based on the two prime drivers, population and economic growth. urban simulation.
Note that the informal sector is not included in the official GDP esti-
mates, but could constitute a considerable share of the overall economic
Appendix A. Supplementary data
growth in Indonesia. In addition, the forecasts of GDP growth differenti-
ated by economic sector may provide more accurate forecasts of urban
Supplementary data to this article can be found online at http://dx.
expansion. There are also several other factors that could significantly
influence urban expansion, such as agricultural productivity, land-use
policies, and changes in transportation infrastructure (Güneralp and
Seto, 2013; Seto et al., 2012). References
Finally, despite a growing understanding that all risk components Abidin, H.Z., Andreas, H., Gumilar, I., Fukuda, Y., Pohan, Y.E., Deguchi, T., 2011. Land sub-
dynamically interact through time (e.g. Meyer et al., 2013; Visser sidence of Jakarta (Indonesia) and its relation with urban development. Nat. Hazards
59 (3), 1753–1771.
et al., 2014), there is still limited knowledge on how to incorporate
Aerts, J.C.J.H., Botzen, W., Emanuel, K., Lin, N., de Moel, H., 2014. Evaluating flood resil-
this in risk modeling. The link between the different risk components ience strategies for coastal megacities. Science 344, 473–475.
could be improved in several ways. For example, although vulnerability 1126/science.1248222.
may change over time (Jongman et al., 2015) we do not consider such Alho, J.M., 1997. Scenarios, uncertainty and conditional forecasts of the world population.
J. R. Stat. Soc. A Stat. Soc. 160 (1), 71–85 (Retrieved from http://www.ncbi.nlm.nih.
changes as no future projections of vulnerability exist. Also, the dynamic gov/pubmed/12292503).
link between land-use change and hydrological change should be in- Angel, S., Parent, J., Civco, D.L., Blei, A., Potere, D., 2011. The dimensions of global urban
cluded as, for instance, the extension of urban areas may result in expansion: estimates and projections for all countries, 2000–2050. Prog. Plan. 75
(2), 53–107.
more frequent river flooding due to higher runoff coefficients. In addi- Apel, H., Merz, B., Thieken, A.H., 2008. Quantification of uncertainties in flood risk assess-
tion, there is rapid deforestation in Indonesia, which may amplify future ments. Int. J. River Basin Manag. 6 (2), 149–162.
river flood risk (Bradshaw et al., 2007). Lastly, we consider coastal and 2008.9635344.
Arnell, N.W., Lloyd-Hughes, B., 2014. The global-scale impacts of climate change on water
river floods independent of each other. Several recent studies have resources and flooding under new climate and socio-economic scenarios. Clim.
pointed to the importance of including interactions between the two Chang. 122 (1–2), 127–140.
(e.g. Klerk et al., 2015; Svensson and Jones, 2004). Baugh, C.A., Bates, P.D., Schumann, G., Trigg, M.A., 2013. SRTM Vegetation Removal and
Hydrodynamic Modeling Accuracy. Water Resour. Res. 49 (9), 5276–5289. http://
4. Conclusions Bierkens, M.F.P., van Beek, R.L.P., 2009. Seasonal predictability of European discharge:
NAO and hydrological response time. J. Hydrometeorol. 10 (4), 953–968. http://dx.
We present an approach that utilizes the recent advances in the
Bouwer, L.M., 2011. Have disaster losses increased due to anthropogenic climate change?
global modeling of flood hazards and land change. The approach is Bull. Am. Meteorol. Soc. 92 (1), 39–46.
based on global data, combines climate change projections with proba- Bouwer, L.M., 2013. Projections of future extreme weather losses under changes in cli-
bilistic projections of urban expansion, and includes adaptation strate- mate and exposure. Risk Anal. 33 (5), 915–930.
gies. This makes it possible to assess future trends in national-scale Bouwer, L.M., Crompton, R.P., Faust, E., Höppe, P., Pielke Jr., R.A., 2007. Confronting disas-
flood risk in data-scarce developing countries such as Indonesia, ter losses. Science 318 (5851), 753.
456 S. Muis et al. / Science of the Total Environment 538 (2015) 445–457

BPS, 2013. Trends of selected socio-economic indicators of Indonesia. Baruna Surindo
Perkasa — National Statistical Agency, Jakarta (Retrieved from http://www.bps.go. IPCC, 2013. Climate change 2013: the physical science basis. In: Stocker, T.F., Qin, D.,
id/aboutus.php?booklet=1). Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., …, Midgley, P.M. (Eds.), Contribu-
Bradshaw, C.J.A., Sodhi, N.S., Peh, K.S.-H., Brook, B.W., 2007. Global evidence that defores- tion of Working Group I to the Fifth Assessment Report of the Intergovernmental
tation amplifies flood risk and severity in the developing world. Glob. Chang. Biol. 13 Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom
(11), 2379–2395. and New York, NY, USA.
Brown, S., Nicholls, R.J., Hanson, S., Brundrit, G., Dearing, J.A., Dickson, M.E., ... Woodroffe, Jha, A.K., Bloch, R., Lamond, J., 2012. Cities and flooding: a guide to integrated urban flood
C.D., 2014. Shifting perspectives on coastal impacts and adaptation. Nat. Clim. Chang. risk management for the 21st century. World Bank (URL: https://openknowledge.
4 (9), 752–755.
Bubeck, P., De Moel, H., Bouwer, L.M., Aerts, J.C.J.H., 2011. How reliable are projections of Jongman, B., Ward, P.J., Aerts, J.C.J.H., 2012. Global exposure to river and coastal flooding:
future flood damage? Nat. Hazards Earth Syst. Sci. 11 (12), 3293–3306. http://dx.doi. long term trends and changes. Glob. Environ. Chang. 22 (4), 823–835. http://dx.doi.
org/10.5194/nhess-11-3293-2011. org/10.1016/j.gloenvcha.2012.07.004.
Budiyono, Y., Aerts, J.C.J.H., Brinkman, J.J., Marfai, M.A., Philip, W., Ward, P.J., 2014. Flood Jongman, B., Hochrainer-Stigler, S., Feyen, L., Aerts, J.C.J.H., Mechler, R., Botzen, W.,
risk assessment for delta mega-cities: a case study of Jakarta. Nat. Hazards http:// Bouwer, L.M., Pflug, G., Rojas, R., Ward, P.J., 2014a. Increasing stress on disaster-risk finance due to large floods. Nat. Clim. Chang. 1–5
Chaussard, E., Amelung, F., Abidin, H.Z., Hong, S.-H., 2013. Sinking cities in Indonesia: NCLIMATE2124.
ALOS PALSAR detects rapid subsidence due to groundwater and gas extraction. Re- Jongman, B., Koks, E.E., Husby, T.G., Ward, P.J., 2014b. Financing increasing flood risk: ev-
mote Sens. Environ. 128 (2013), 150–161. idence from millions of buildings. Nat. Hazards Earth Syst. Sci. 2 (1), 137–165. http://
Cohen, B., 2004. Urban growth in developing countries: a review of current trends and a Jongman, B., Winsemius, H.C., Aerts, J.C.J.H., Coughlan de Perez, E., van Aalst, M., Kron, W.,
caution regarding existing forecasts. World Dev. 32 (1), 23–51. Ward, P.J., 2015. Declining vulnerability to river floods and the global benefits of ad-
1016/j.worlddev.2003.04.008. aptation (PhD thesis), Jongman B. (2015). Unravelling the drivers of flood risk across
De Moel, H., Vliet, M., Aerts, J.C.J.H., 2013. Evaluating the effect of flood damage-reducing spatial scales. VU University Amsterdam (Chapter 6, 194 pp.).
measures: a case study of the unembanked area of Rotterdam, the Netherlands. Reg. Kreibich, H., Piroth, K., Seifert, I., Maiwald, H., Kunert, U., Schwarz, J., Merz, B.,
Environ. Chang. Thieken, a.H., 2009. Is flow velocity a significant parameter in flood damage
de Moel, H., Bouwer, L.M., Aerts, J.C.J.H., 2014. Uncertainty and sensitivity of flood risk cal- modelling? Nat. Hazards Earth Syst. Sci. 9, 1679–1692.
culations for a dike ring in the south of the Netherlands. Sci. Total Environ. 473–474, 5194/nhess-9-1679-2009.
224–234. Kim, Y., Chung, E.S., 2014. An index-based robust decision making framework for water-
Dessai, S., Hulme, M., 2004. Does climate adaptation policy need probabilities? Clim. Pol. shed management in a changing climate. Sci. Total Environ. 473–474, 88–102. http://
4, 107–128.
Dunne, T., Leopold, L.B., 1978. Water in environmental planning. W. H. Freeman, San Klein Goldewijk, K., Beusen, A., Van Drecht, G., De Vos, M., 2011. The HYDE 3.1 spatially
Fransico, CA, p. 818. explicit database of human-induced global land-use change over the past
EM-DAT, 2013. The OFDA/CRED International Disaster Database Retrieved from, http:// 12,000 years. Glob. Ecol. Biogeogr. 20 (1), 73–86. (on 20 Augustus 2014). 8238.2010.00587.x.
Erban, L.E., Gorelick, S.M., Zebker, H.A., 2014. Groundwater extraction, land subsidence, Klerk, W.J., Winsemius, H.C., van Verseveld, W.J., Bakker, A.M.R., Diermanse, F.L.M., 2015.
and sea-level rise in the Mekong Delta, Vietnam. Environ. Res. Lett. 9 (8), 084010. The co-incidence of storm surges and extreme discharges within the Rhine–Meuse Delta. Environ. Res. Lett. 10 (3).
Feyen, L., Dankers, R., Bódis, K., Salamon, P., Barredo, J.I., 2012. Fluvial flood risk in Europe (IOP Publishing: 035005).
in present and future climates. Clim. Chang. 112, 47–62. Koks, E.E., De Moel, H., Aerts, J.C.J.H., Bouwer, L.M., 2013. Effect of spatial adaptation mea-
s10584-011-0339-7. sures on flood risk: study of coastal floods in Belgium. Reg. Environ. Chang. http://dx.
Friedl, M.A., Sulla-Menashe, D., Tan, B., Schneider, A., Ramankutty, N., Sibley, A., Huang, X.,
2010. MODIS collection 5 global land cover: algorithm refinements and characteriza- Lasage, R., Veldkamp, T.I.E., De Moel, H., Van, T.C., Phi, H.L., Vellinga, P., Aerts, J.C.J.H.,
tion of new datasets. Remote Sens. Environ. 114 (1), 168–182. 2014. Assessment of the effectiveness of participatory developed adaptation
1016/j.rse.2009.08.016. strategies for HCMC. Nat. Hazards Earth Syst. Sci. 2 (1), 385–425. http://dx.doi.
Gumbel, E.J., 1941. The Return Period of Flood Flows. The Annals of Mathematical Statis- org/10.5194/nhessd-2-385-2014.
tics 12 (2), 163–190. Lehner, B., Verdin, K., Jarvis, A., 2008. New global hydrography derived from spaceborne
Güneralp, B., Seto, K.C., 2013. Futures of global urban expansion: uncertainties and impli- elevation data. Eos 89 (10), 93–104.
cations for biodiversity conservation. Environ. Res. Lett. 8 (1), 014025. http://dx.doi. Lowe, J.A., Woodworth, P.L., Knutson, T., Mcdonald, R.E., McInnes, K.L., Von Storch, H., ...
org/10.1088/1748-9326/8/1/014025. Weisse, R., 2010. Past and future changes and waves. In: Church, J.A., Woodworth,
Güneralp, B., Güneralp, İ., Liu, Y., 2015. Changing global patterns of urban exposure to P.L., Aarup, T., Wilson, W.S. (Eds.), Understanding Sea-Level Rise and Variability, 1st
flood and drought hazards. Glob. Environ. Chang. 31, 217–225. edition Blackwell Publishing Ltd., pp. 326–375.
1016/j.gloenvcha.2015.01.002. Marfai, M.A., King, L., 2007. Potential vulnerability implications of coastal inundation due
Haigh, I.D., MacPherson, L.R., Mason, M.S., Wijeratne, E.M.S., Pattiaratchi, C.B., Crompton, to sea level rise for the coastal zone of Semarang city, Indonesia. Environ. Geol. 54 (6),
R.P., George, S., 2013. Estimating present day extreme water level exceedance prob- 1235–1245.
abilities around the coastline of Australia: tropical cyclone-induced storm surges. Mechler, R., Bouwer, L.M., 2014. Understanding trends and projections of disaster losses
Clim. Dyn. 42 (1–2), 139–157. and climate change: is vulnerability the missing link? Clim. Chang. http://dx.doi.
Hall, J.W., 2007. Probabilistic climate scenarios may misrepresent uncertainty and lead to org/10.1007/s10584-014-1141-0.
bad adaptation decisions. Hydrol. Process. 21, 1127–1129. Menéndez, M., Woodworth, P.L., 2010. Changes in extreme high water levels based on a
hyp.6573. quasi-global tide-gauge data set. J. Geophys. Res. 115 (C10011).
Hallegatte, S., Green, C., Nicholls, R.J., Corfee-Morlot, J., 2013. Future flood losses in 1029/2009JC005997.
major coastal cities. Nat. Clim. Chang. 3 (9), 802–806. Merz, B., Thieken, A.H., 2009. Flood risk curves and uncertainty bounds. Nat. Hazards 51,
1038/nclimate1979. 437–458.
Han, W., Meehl, G.A., Rajagopalan, B., Fasullo, J.T., Hu, A., Lin, J., ... Yeager, S., 2010. Patterns Merz, B., Kreibich, H., Schwarze, R., Thieken, A.H., 2010. Review article “Assessment of
of Indian Ocean sea-level change in a warming climate. Nat. Geosci. 3 (8), 546–550. economic flood damage.”. Nat. Hazards Earth Syst. Sci. 10 (8), 1697–1724. http://
Hanson, S., Nicholls, R.J., Ranger, N., Hallegatte, S., Corfee-Morlot, J., Herweijer, C., Chateau, Meyer, V., Becker, N., Markantonis, V., Schwarze, R., van den Bergh, J.C.J.M., Bouwer, L.M.,
J., 2010. A global ranking of port cities with high exposure to climate extremes. Clim. ... Viavattene, C., 2013. Review article: assessing the costs of natural hazards — state of
Chang. 104 (1), 89–111. the art and knowledge gaps. Nat. Hazards Earth Syst. Sci. 13 (5), 1351–1373. http://
Hempel, S., Frieler, K., Warszawski, L., Schewe, J., Piontek, F., 2013. A trend-preserving bias
correction — the ISI-MIP approach. Earth Syst. Dyn. 4 (2), 219–236. Moss, R.H., Edmonds, J.A., Hibbard, K.A., Manning, M.R., Rose, S.K., van Vuuren, D.P., ...
10.5194/esd-4-219-2013. Wilbanks, T.J., 2010. The next generation of scenarios for climate change research
Hinkel, J., Klein, R.J.T., 2009. Integrating knowledge to assess coastal vulnerability to sea- and assessment. Nature 463 (7282), 747–756.
level rise: the development of the DIVA tool. Glob. Environ. Chang. 19 (3), 384–395. nature08823. Nakicenovic, N., Alcamo, J., Davis, G., de Vries, B., Fenhann, J., Gaffin, S., Dadi, Z., 2000. spe-
Hinkel, J., Lincke, D., Vafeidis, A.T., Perrette, M., Nicholls, R.J., Tol, R.S.J., Marzeiong, B., cial report on emissions scenarios: a special report of the Working Group III of the In-
Fettweish, X., Ionescus, C., Levermann, A., 2014. Coastal flood damage and adaptation tergovernmental Panel on climate change. Cambirdge University Press, Cambridge,
costs under 21st century sea-level rise. Proc. Natl. Acad. Sci. U. S. A. p. 599 (Retrieved from on 20
10.1073/pnas.1222469111. November 2013).
Hirabayashi, Y., Mahendran, R., Koirala, S., Konoshima, L., Yamazaki, D., Watanabe, S., ... Nelson, E., Sander, H., Hawthorne, P., Conte, M., Ennaanay, D., Wolny, S., Manson, S.,
Kanae, S., 2013. Global flood risk under climate change. Nat. Clim. Chang. 3 (9), Polasky, S., 2010. Projecting global land-use change and its effect on ecosystem ser-
816–821. vice provision and biodiversity with simple models. PLoS One 5 (12), e14327.
Huizinga, H.J., 2007. Flood damage functions for EU member states. Technical report by
HKV Consultants and Joint Research Centre of the European Commission. New, M., Lopez, A., Dessai, S., Wilby, R.L., 2007. Challenges in using probabilistic climate
IPCC, 2012. In: Field, C.B., Barros, V., Stocker, T.F., Dahe, Q., Qin, D., Dokken, D.J., …, change information for impact assessments: an example from the water sector.
Midgley, P.M. (Eds.), Managing the Risks of Extreme Events and Disasters to Advance Phil. Trans. R. Soc. A 365 (1857), 2117–2131.
Climate Change Adaptation. Cambridge University Press, Cambridge, p. 582 http://dx. 2080.
S. Muis et al. / Science of the Total Environment 538 (2015) 445–457 457

Nicholls, R.J., Cazenave, A., 2010. Sea-level rise and its impact on coastal zones. Science Te Linde, A.H., Bubeck, P., Dekkers, J.E.C., de Moel, H., Aerts, J.C.J.H., 2011. Future flood risk
328 (5985), 1517–1520. estimates along the river Rhine. Nat. Hazards Earth Syst. Sci. 11 (2), 459–473. http://
Nicholls, R.J., Marinova, N., Lowe, J.A., Brown, S., Vellinga, P., de Gusmão, D., ... Tol, R.S.J.,
2011. Sea-level rise and its possible impacts given a “beyond 4 °C world” in the Thieken, A.H., Cammerer, H., Dobler, C., Lammel, J., Schöberl, F., 2014. Estimating changes
twenty-first century. Phil. Trans. R. Soc. A 369 (1934), 161–181. in flood risks and benefits of non-structural adaptation strategies — a case study from
10.1098/rsta.2010.0291. Tyrol, Austria. Mitig. Adapt. Strateg. Glob. Chang.
NRC, 2000. Beyond six billion: forecasting the world's population. National Research 014-9602-3.
Council, Washington D.C., United Sates of America, p. 236. UN, 2012. World urbanization prospects: the 2011 revision. Population Division of the
Penning-Rowsell, E.C., Yanyan, W., Watkinson, A.R., Jiang, J., Thorne, C.R., 2013. Socioeco- Department of Economic and Social Affairs of the United Nations Secretariat, New
nomic scenarios and flood damage assessment methodologies for the Taihu Basin, York.
China. J. Flood Risk Manag. 6 (1), 23–32. UN, 2013. World population prospects: the 2012 revision. Population Division of the De-
2012.01168.x. partment of Economic and Social Affairs of the United Nations Secretariat, New York.
Pittock, A.B., Jones, R.N., Mitchell, C.D., 2001. Probabilities will help us plan for climate UN, 2014. World urbanization prospects: the 2013 revision. Population Division of the
change. Nature 413, 249. Department of Economic and Social Affairs of the United Nations Secretariat, New
Pontius, R.G., Boersma, W., Castella, J.-C., Clarke, K., Nijs, T., Dietzel, C., ... Verburg, P.H., York.
2008. Comparing the input, output, and validation maps for several models of land UNISDR, 2011. Global assessment report on disaster risk reduction. Revealing Risk,
change. Ann. Reg. Sci. 42 (1), 11–37. Redefining Development (Geneva).
Poussin, J.K., Bubeck, P., Aerts, J.C.J.H., Ward, P.J., 2012. Potential of semi-structural and non- Vafeidis, A.T., Nicholls, R.J., McFadden, L., Tol, R.S.J., Hinkel, J., Spencer, T., ... Klein, R.J.T.,
structural adaptation strategies to reduce future flood risk: case study for the Meuse. 2008. A new global coastal database for impact and vulnerability analysis to sea-
Nat. Hazards Earth Syst. Sci. 12 (11), 3455–3471. level rise. J. Coast. Res. 244, 917–924.
3455-2012. Van Beek, R.L.P., Bierkens, M.F.P., 2008. The global hydrological model PCR-GLOBWB: con-
Pritchett, L., 2000. Understanding patterns of economic growth: searching for hills among ceptualization, parameterization and verification. Department of Physical Geography,
plateaus, mountains, and plains. World Bank Econ. Rev. 14 (2), 221–250. http://dx. Faculty of Earth Sciences, Utrecht University, p. 53 (Utrecht, The Netherlands. Re- trieved from on 13 No-
Purvis, M.J., Bates, P.D., Hayes, C.M., 2008. A probabilistic methodology to estimate future vember 2013).
coastal flood risk due to sea level rise. Coast. Eng. 55 (12), 1062–1073. http://dx.doi. Van Beek, L.P.H., Bierkens, M.F.P., 2009. The Global Hydrological Model PCR-GLOBWB:
org/10.1016/j.coastaleng.2008.04.008. Conceptualization. Parameterization and Verification (Utrecht: Utrecht University)
Rodríguez, E., Morris, C.S., Belz, J.E., 2006. A Global Assessment of the SRTM Performance. available online at
Photogramm. Eng. Remote Sens. 3 (12), 249–260. Visser, H., Petersen, A.C., Ligtvoet, W., 2014. On the relation between weather-related di-
72.3.249. saster impacts, vulnerability and climate change. Clim. Chang. 125, 461–477. http://
Sagala, S., Lassa, J., Yasaditama, H., Hudalah, D., 2013. The evolution of risk and vulnerabil-
ity in Greater Jakarta: contesting government policy in dealing with a megacity's ex- Ward, P.J., Marfai, M.A., Yulianto, E., Hizbaron, D.R., Aerts, J.C.J.H., 2010. Coastal inundation
posure to flooding. An academic response to Jakarta Floods in January 2013 (No. 2). and damage exposure estimation: a case study for Jakarta. Nat. Hazards 56 (3),
IRGSC Working Paper No. 2, Kupang, Indonesia, p. 18 (Retrieved from http://irgsc. 899–916.
org/pubs/wp/IRGSCWP002jakartaflood.pdf on 10 September 2014). Ward, P.J., De Moel, H., Aerts, J.C.J.H., 2011. How are flood risk estimates affected by the
Schneider, S.H., 2001. What is “dangerous” climate change? Nature 411, 17–19. http://dx. choice of return-periods? Nat. Hazards Earth Syst. Sci. 11 (12), 3181–3195. http://
Schneider, S.H., 2002. Can we estimate the likelihood of climatic changes at 2100? An Ed- Ward, P.J., van Pelt, S.C., de Keizer, O., Aerts, J.C.J.H., Beersma, J.J., van den Hurk, B.J.J.M., te
itorial Comment. Clim. Chang. 52, 423–425. Linde, A.H., 2013a. Including climate change projections in probabilistic flood risk as-
Schneider, A., Friedl, M.A., Potere, D., 2009. A new map of global urban extent from MODIS sessment. J. Flood Risk Manag. (2006, n/a–n/a).
satellite data. Environ. Res. Lett. 4 (4). Ward, P.J., Jongman, B., Sperna Weiland, F.C., Bouwman, A., van Beek, R.L.P., Bierkens,
044003. M.F.P., ... Winsemius, H.C., 2013b. Assessing flood risk at the global scale: model
Schumann, G.J.P., Bates, P.D., Neal, J., Andreadis, K.M., 2014. Fight Floods on a Global Scale. setup, results, and sensitivity. Environ. Res. Lett. 8 (4), 044019.
Nature 507 (7491), 169. 1088/1748-9326/8/4/044019.
Seto, K.C., Fragkias, M., Güneralp, B., Reilly, M.K., 2011. A meta-analysis of global urban Ward, P.J., Jongman, B., Kummu, M., Dettinger, M.D., Sperna Weiland, F.C., Winsemius,
land expansion. PLoS One 6 (8), e23777. H.C., 2014. Strong influence of El Niño Southern Oscillation on flood risk around the
0023777. world. Proc. Natl. Acad. Sci. U. S. A. Early Ed.
Seto, K.C., Güneralp, B., Hutyra, L.R., 2012. Global forecasts of urban expansion to 2030 and 1409822111.
direct impacts on biodiversity and carbon pools. Proc. Natl. Acad. Sci. U. S. A. 109 (40), Webster, M., Sokolov, A.P., Reilly, J.M., Forest, C.E., Paltsev, S., Schlosser, A., Wang, C.,
16083–16088. Kickligher, D., Sarofim, M., Melillo, J., Prinn, R.G., Jacoby, H.D., 2012. Analysis of cli-
Smith, D.I., 1994. Flood damage estimation — a review of urban stage-damage curves and mate policy targets under uncertainty. Clim. Chang. 112, 569–583. http://dx.doi.
loss functions. Water SA 20, 2310238. org/10.1007/s10584-011-0260-0.
Storch, H., Downes, N.K., 2011. A scenario-based approach to assess Ho Chi Minh City's Weedon, G.P., Gomes, S., Viterbo, P., Shuttleworth, W.J., Blyth, E., Österle, H., ... Best, M.,
urban development strategies against the impact of climate change. Cities 28 (6), 2011. Creation of the WATCH forcing data and its use to assess global and regional
517–526. reference crop evaporation over land during the twentieth century.
Svensson, C., Jones, D.A., 2004. Dependence between sea surge, river flow and precipita- J. Hydrometeorol. 12 (5), 823–848.
tion in south and west Britain. Hydrol. Earth Syst. Sci. 8 (5), 973–992. Winsemius, H.C., van Beek, R.L.P., Jongman, B., Ward, P.J., Bouwman, A., 2013. A frame-
Syvitski, J.P.M., Kettner, A.J., Overeem, I., Hutton, E.W.H., Hannon, M.T., Brakenridge, G.R., work for global river flood risk assessments. Hydrol. Earth Syst. Sci. 17 (5),
... Nicholls, R.J., 2009. Sinking deltas due to human activities. Nat. Geosci. 2 (10), 1871–1892.
681–686. World Bank, 2014. World development indicators Retrieved January 15, 2014, from
Tachikawa, Y., James, R., Abdullah, K., Mohd., N., Desa, bin Mohd, 2004. Catalogue for riv-
ers in South East Asia and the Pacific – Volume 1. Acquired via http:// (Accessed on July, 24 2015).
Takeuchi, K., Jayawardena, A.W., Takahasi, W., 1995. Catalogue for rivers in South East
Asia and the Pacific – Volume 1. Acquired via
rsc/riverCatalogue/Vol_01/ (Accessed on July, 24 2015).