Contrasting seismic risk for Santiago, Chile, from near-field and distant earthquake sources

Abstract

More than half of all the people in the world now live in dense urban centres. The rapid expansion of cities, particularly in low-income nations, has enabled the economic and social development of millions of people. However, many of these cities are located near active tectonic faults that have not produced an earthquake in recent memory, raising the risk of losing hard-earned progress through a devastating earthquake. In this paper we explore the possible impact that earthquakes can have on the city of Santiago in Chile from various potential near-field and distant earthquake sources. We use high-resolution stereo satellite imagery and imagery-derived digital elevation models to accurately map the trace of the San Ramón Fault, a recently recognised active fault located along the eastern margins of the city. We use scenario-based seismic-risk analysis to compare and contrast the estimated damage and losses to the city from several potential earthquake sources and one past event, comprising (i) rupture of the San Ramón Fault, (ii) a hypothesised buried shallow fault beneath the centre of the city, (iii) a deep intra-slab fault, and (iv) the 2010 Mw 8.8 Maule earthquake. We find that there is a strong magnitude–distance trade-off in terms of damage and losses to the city, with smaller magnitude earthquakes in the magnitude range of 6–7.5 on more local faults producing 9 to 17 times more damage to the city and estimated fatalities compared to the great magnitude 8+ earthquakes located offshore in the subduction zone. Our calculations for this part of Chile show that unreinforced-masonry structures are the most vulnerable to these types of earthquake shaking. We identify particularly vulnerable districts, such as Ñuñoa, Santiago, and Providencia, where targeted retrofitting campaigns would be most effective at reducing potential economic and human losses. Due to the potency of near-field earthquake sources demonstrated here, our work highlights the importance of also identifying and considering proximal minor active faults for cities in seismic zones globally in addition to the more major and distant large fault zones that are typically focussed on in the assessment of hazard.

Full text

Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020
https://doi.org/10.5194/nhess-20-1533-2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.
Contrasting seismic risk for Santiago, Chile, from near-field
and distant earthquake sources
Ekbal Hussain1,2, John R. Elliott1, Vitor Silva3, Mabé Vilar-Vega3, and Deborah Kane4
1COMET, School of Earth and Environment, University of Leeds, Leeds, LS2 9JT, UK
2British Geological Survey, Natural Environment Research Council, Environmental Science Centre,
Keyworth, Nottingham, NG12 5GG, UK
3GEM Foundation, Via Ferrata 1, 27100 Pavia, Italy
4Risk Management Solutions, Inc., Newark, CA, USA
Correspondence: Ekbal Hussain (ekhuss@bgs.ac.uk)
Received: 1 February 2019 – Discussion started: 11 March 2019
Revised: 20 March 2020 – Accepted: 29 March 2020 – Published: 29 May 2020
Abstract. More than half of all the people in the world now
live in dense urban centres. The rapid expansion of cities,
particularly in low-income nations, has enabled the eco-
nomic and social development of millions of people. How-
ever, many of these cities are located near active tectonic
faults that have not produced an earthquake in recent mem-
ory, raising the risk of losing hard-earned progress through a
devastating earthquake. In this paper we explore the possible
impact that earthquakes can have on the city of Santiago in
Chile from various potential near-field and distant earthquake
sources. We use high-resolution stereo satellite imagery and
imagery-derived digital elevation models to accurately map
the trace of the San Ramón Fault, a recently recognised ac-
tive fault located along the eastern margins of the city. We
use scenario-based seismic-risk analysis to compare and con-
trast the estimated damage and losses to the city from several
potential earthquake sources and one past event, comprising
(i) rupture of the San Ramón Fault, (ii) a hypothesised buried
shallow fault beneath the centre of the city, (iii) a deep intra-
slab fault, and (iv) the 2010 Mw8.8 Maule earthquake. We
find that there is a strong magnitude–distance trade-off in
terms of damage and losses to the city, with smaller magni-
tude earthquakes in the magnitude range of 6–7.5 on more
local faults producing 9 to 17 times more damage to the
city and estimated fatalities compared to the great magnitude
8+ earthquakes located offshore in the subduction zone. Our
calculations for this part of Chile show that unreinforced-
masonry structures are the most vulnerable to these types
of earthquake shaking. We identify particularly vulnerable
districts, such as Ñuñoa, Santiago, and Providencia, where
targeted retrofitting campaigns would be most effective at
reducing potential economic and human losses. Due to the
potency of near-field earthquake sources demonstrated here,
our work highlights the importance of also identifying and
considering proximal minor active faults for cities in seismic
zones globally in addition to the more major and distant large
fault zones that are typically focussed on in the assessment
of hazard.
1 Introduction
Earthquakes are caused by the sudden release of accumu-
lated tectonic strain that increases in the crust over decades
to millennia. Many faults are often not recognised as danger-
ous because they have not recorded an earthquake in living
and written memory (e.g. England and Jackson, 2011). Since
probabilistic seismic-hazard assessments (PSHAs) rely on
knowledge of past seismicity to determine hazard levels, the
regions around these faults are often deemed to be low haz-
ard in seismic-risk assessments until an earthquake strikes
and the assessment is revised (Stein et al., 2012). The 2010
Mw7.0 Haiti earthquake, with its close proximity to an urban
centre, was a stark reminder of how ruptures on these faults
can be so deadly, especially when they are located near ma-
jor population centres in poorly prepared low-income nations
(Bilham, 2010).
Published by Copernicus Publications on behalf of the European Geosciences Union.

1534 E. Hussain et al.: Seismic risk in Santiago
The South American country of Chile is one of the most
seismically active countries in the world. Since 1900 there
have been 11 great earthquakes in the country with magni-
tudes 8 or larger (USGS, 2019). All of these were located
on or near the subduction interface where the Nazca plate is
subducting beneath the South American plate at 7.5 cm yr−1
(DeMets et al., 2010), giving rise to the Andean mountain
range (Armijo et al., 2015; Oncken et al., 2006). It is there-
fore unsurprising that shaking from offshore subduction zone
events dominates the seismic hazard – thus the building de-
sign – criteria in Chile (e.g. Fischer et al., 2002; Pina et al.,
2012; Santos et al., 2012). The most recent great event was
the Mw8.8 Maule earthquake, which struck southern Chile in
2010, generating a tsunami and causing 521 fatalities. How-
ever, large, shallow-crustal (<15 km depth) earthquakes are
not uncommon in Chile. Since 1900 there have been nine
magnitude 7+shallow-crustal earthquakes located inland
and therefore not directly associated with slip on the sub-
duction megathrust (USGS, 2019). But most of these faults
accumulate strain at slower rates compared to the subducting
plate boundary and thus rupture infrequently.
The San Ramón Fault is one such fault. It runs along
the foothills of the San Ramón mountains and bounds the
eastern margin of the capital city Santiago, a conurbation
which hosts 40 % of the country’s population within the
city’s metropolitan region (7 million according to 2017 es-
timates). Due to the rapid expansion of the city in the 20th
and 21st centuries (Ramón, 1992), parts of the fault now lie
beneath the eastern communes (districts) of the city (Fig. 1),
in particular Puente Alto, La Florida, Peñalolén, La Reina,
and Las Condes. Yet it was only as recent as the past decade
that Armijo et al. (2010) recognised that the San Ramón
Fault is an active Quaternary thrust fault and poses a sig-
nificant hazard to the city. Using field mapping and satel-
lite imagery they estimated a slip rate of ∼0.5 mm yr−1for
the fault, a much slower loading rate compared to the overall
7.5 cm yr−1plate convergence rate in the subduction zone.
Palaeo-seismic trench studies across the San Ramón Fault
scarp revealed records of two historical ∼5m slip events –
approximately equivalent to a pair of Mw7.5 earthquakes –
17–19 and ∼8 kyr ago (Vargas et al., 2014). However, based
on geophysical investigations of the fault region, Estay et al.
(2016) concluded that the San Ramón Fault is segmented into
four sub-faults that are most likely activated independently in
earthquakes with moment magnitude in the range of 6.2 and
6.7. While Estay et al. (2016) do not discount the possibil-
ity of a larger rupture linking across all four segments, evi-
dence from the trench studies (Vargas et al., 2014) suggests
that larger-magnitude earthquakes are possible on the fault.
Therefore for the seismic hazard and risk analysis in this pa-
per we take the worst-case scenario of a complete rupture
along the fault as our scenario case study.
Riesner et al. (2017) used balanced kinematic reconstruc-
tions of the geology across the region to deduce a long-term
average shortening rate of between 0.3 and 0.5 mm yr−1,
Figure 1. Declassified corona satellite image (∼2 m resolution)
from 1970 (left) and the same region in the SPOT imagery (1.5m
resolution; right) used in this study showing the eastward expan-
sion of the city over the San Ramón Fault (red lines). Four notable
regions are highlighted in blue: (a) an alluvial fan that is clearly
visible in the older imagery but completely covered with buildings
in the recent image, (b) expansion into the foothills of the moun-
tain and onto the hanging wall of the San Ramón thrust fault (these
are often more affluent neighbourhoods with better views across the
city), (c) urban densification in the central regions, and (d) land use
change from farmland to dense urban neighbourhood masking the
fault trace.
which is compatible with the earlier estimates by Armijo
et al. (2010) and the recurrence slip rates deduced from
the palaeo-earthquakes in the trench study by Vargas et al.
(2014). This suggests that most of the active deformation
across the West Andean fold-and-thrust belt is accommo-
dated on the San Ramón Fault. Therefore, despite the very
low slip rates, the long time interval since the last earthquake
means that significant strain has now accumulated on the
fault, and if it were to rupture completely, it could produce
earthquakes of an equivalent magnitude as those recorded in
the trench.
Pérez et al. (2014) performed a detailed analysis of local
seismicity for the region and showed that the microseismic-
ity at depth (∼10 km) can be associated with the San Ramón
Fault, implying that the fault is indeed active and accumulat-
ing strain. Vaziri et al. (2012) used Risk Management Solu-
tions’ (RMS) commercial catastrophe risk modelling frame-
work to estimate the losses from future earthquakes on the
San Ramón Fault for Santiago. They estimate that a Mw6.8
earthquake on the fault could result in 14 000 fatalities with a
building loss ratio of 6.5 %. However, the spatial distribution
of these losses remains unclear.
Building on this previous work of identifying the hazard
and losses, we aim to contrast the risk posed by the San
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1535
Ramón Fault and place it in the context of other potential
earthquake sources and a previous far-field subduction earth-
quake (Maule 2010). As we are examining the losses due to
a very-near-field source with exposed elements immediately
adjacent to the potential rupture, we seek to delineate the lo-
cation, extent, and segmentation of the San Ramón Fault to
improve the accuracy of the ground motion. Stereo satellite
optical imagery is often used to derive high-resolution digital
elevation models (DEMs) over relatively large areas, which
can be useful in identifying subtle active tectonic geomorphic
markers of faulting as well as in examining fault segmenta-
tion (Elliott et al., 2016). In this paper we use DEMs created
from high-resolution satellite imagery from the SPOT and
Pléiades satellites (1.5 and 0.5 m resolution respectively) to
better characterise the surface expression of the San Ramón
Fault and to also look for other potential fault splays within
the city limits. Following a similar method as Chaulagain
et al. (2016) and Villar-Vega and Silva (2017) and using
the Global Earthquake Model’s (GEM) OpenQuake Engine
(Silva et al., 2014), we explore the contrasting losses to the
residential-building stock in the capital through scenario cal-
culations for (a) future earthquakes on the San Ramón Fault,
(b) earthquakes on a hypothesised shallow splay buried be-
neath the centre of the city, (c) deep intra-slab events, and
(d) the 2010 Mw8.8 Maule earthquake. Our models help us
to identify particularly vulnerable parts of the city and en-
able us to make targeted geographical recommendations to
improve the seismic resilience of these communities.
We also explore losses to the non-residential-building
stock using the Risk Management Solutions (RMS) commer-
cial risk model. The RMS model provides a different view of
the commercial risk, where the model is well calibrated due
to the availability of losses from previous events and covers
the insured assets, which is often one of the main mecha-
nisms to recover from disasters.
2 Fault geomorphology from satellite imagery
Freely available global elevation data from the Shuttle Radar
Topography Mission (SRTM) (Farr et al., 2007) have a spa-
tial resolution of 30 m, which is insufficient to accurately
map the San Ramón Fault scarp or look for other potential
fault splays expressed in the geomorphology. To overcome
the low-resolution issue, we analysed SPOT-6 stereo satel-
lite imagery over a 35km×36km region covering Santiago
city and the San Ramón mountains. The SPOT-6 panchro-
matic imagery (acquired in 2014) has a spatial resolution of
1.5 m. We also requested the acquisition of very high reso-
lution (0.5 m panchromatic, acquired in 2016) Pléiades tri-
stereo imagery over a smaller region (5km ×36km) cover-
ing just the San Ramón Fault (Fig. 2). We performed pho-
togrammetry analysis using commercial software (ERDAS
IMAGINE 2015) to produce topographic point clouds from
the SPOT and Pléiades stereo imagery. We removed exces-
sive low noise from the point clouds by initially doing a
ground classification with only the highest points in a 3.5 m
by 3.5 m grid and removing points that were highly isolated
in wide and flat neighbourhoods before redoing the ground
classification with the filtered points. We then created raster-
gridded digital elevation models with 10 and 2m ground res-
olutions with the de-noised SPOT and Pléiades point clouds
(∼83 and ∼60 million points respectively). We did this by
first triangulating the point cloud into a temporary triangular
irregular network (TIN) and then rasterising the TIN into a
digital elevation model.
The San Ramón Fault is not immediately obvious in the
Pléiades elevation map (Fig. 3) beyond the overall morphol-
ogy of the uplifted San Ramón mountains with a relief of
2.5 km above the Santiago basin. However the fault scarp is
clear in the hillshaded DEM, slope and, and terrain rugged-
ness index (TRI) maps (Fig. 3a, iii–v) as a north–south trend-
ing lineament. The terrain ruggedness index is a measure of
the local variation in elevation about a central pixel (Riley
et al., 1999; Wilson et al., 2007). A TRI of 0 indicates flat
terrain while a value of 1 indicates extremely rugged terrain.
Such analysis can highlight the change in elevation and slope
at a fault scarp. We use these datasets to map the surface ex-
pression of the San Ramón Fault, confirming and building
on previous work by Armijo et al. (2010), who used a 10m
DEM.
The identification of the active fault trace at the surface
provides some evidence for the length, location, and segmen-
tation of the fault at depth, and that information is used as
a constraint in the subsequent risk analysis for the range of
earthquake sources that we seek to test. Measuring the verti-
cal offset across the scarp can give an idea of the past activ-
ity along the fault with the caveat that due to natural erosion
processes, scarps tend to degrade with time. To do this we
plotted a series of west–east profiles across the foothills of
the San Ramón mountains to identify and measure the scarp
height along the fault by determining the vertical offset be-
tween the best fit lines through the point cloud either side
of the fault scarp (green and red lines in Figs. 4 and 5). The
variable topographic slope along the fault means it is difficult
to fit lines of equal length for each profile. In most cases we
have tried to ensure a fit through at least 500m, but where
possible 1 km, of points on either side of the fault scarp.
In the northern section we found scarp heights to vary be-
tween ∼5 and ∼119 m along the fault trace (Fig. 4). Profile
dshows no clear evidence of a fault scarp, but since this pro-
file is near a stream channel the scarp is moderated by fluvial
erosion. However, it contains a clear break in slope, which is
indicative of active faulting. Profiles c and k cross anticlines
(∼145 and ∼44 m high respectively) that have likely grown
as a result of long-term movements in the hanging wall of the
fault. The anticline shown in profile kcuts across the north-
ern section of an alluvial fan, implying its growth post-dates
the age of the fan deposit.
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1536 E. Hussain et al.: Seismic risk in Santiago
Figure 2. (a) Map of central Chile showing the location of great earthquakes for the past century in the subduction zone where the Nazca
plate is converging beneath the South American plate at a rate of 43mm yr−1(Zheng et al., 2014). The Santiago metropolitan region is
shown in the dark grey outline, subdivided by commune (names of all the communes are given in Fig. S1). (b) A 90 m SRTM shaded terrain
map of the region around Santiago city (light grey). The SPOT and Pléiades satellite data used in this study cover the region shown by the
maroon and purple polygons respectively. The San Ramón Fault is shown in red, while the dotted blue line is the location of our inferred
buried fault within the city (see text for details). The dashed black lines are the mountain front faults mapped by Armijo et al. (2010).
In the southern section the scarp heights vary between ∼2
and ∼30 m (Fig. 5). Profiles land mshow two folds with
heights of ∼23 and ∼102 m respectively. The growth of the
fold in profile p(∼68 m high) shows evidence that it blocked
and diverted the Maipo river further south to its current posi-
tion.
Our estimate of ∼33 m for profile jand ∼39 m for pro-
file gare equivalent to the 31 and 40m estimated by Armijo
et al. (2010). However, our estimate of ∼36 m for profile his
significantly less than their 54–60 m. The range is probably
due to our interpretation of the upper slope, which varies due
to scarp degradation. This may be because of the relatively
lower resolution DEM of 10 m used by Armijo et al. (2010)
compared to our 2 m DEM.
Our geomorphic analysis of the surface trace of the San
Ramón Fault confirms the findings of Armijo et al. (2010) for
the central and northern part of the fault (red line in Fig. 3)
and extends the trace of the fault further to the south (blue
line in Fig. 3).
In their trench study Vargas et al. (2014) measured fault
displacements of the order of ∼5 m. Projected to the ver-
tical for a 45◦dipping fault, this gives a vertical offset of
about 3.5 m. Given that our average measured scarp height is
∼32 m, it is likely that this represents cumulative displace-
ment over numerous earthquakes. However there is signifi-
cant variation along the fault, with the smallest scarp of 2m
and the largest at 119 m. This variation is probably due to
erosion along the mountain front leading to variable degra-
dation of the fault scarps. The smaller scarps represent the
cumulative displacements of fewer earthquakes. While these
observations enable us to determine the active fault segments
that comprise the San Ramón Fault, the variations in scarp
height mean it is difficult to trace specific historical ruptures
along the fault.
Our observations of the fault traces show a network of fault
segments that are ∼0.5 to ∼8 km long. Our assumption is
that at depth these segments represent the same fault. This is
motivated by field observation that large earthquake ruptures
often consist of multiple rupture segments (e.g. Barka et al.,
2002; Civico et al., 2018). It is possible that the two main
strands of the San Ramón Fault (Figs. 4 and 5) could rup-
ture independently. However, these individual fault segments
are not long enough to produce a magnitude 7.5 earthquake
from a 5 m slip event, which justifies exploring earthquake
scenarios that rupture across both strands of the fault.
The West Andean frontal faults drawn by Armijo et al.
(2010) appear to terminate at the northern and southern mar-
gins of the city. Although it is possible that the San Ramón
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1537
Figure 3. (a) (i) The Pléiades satellite optical multispectral image, (ii) the elevation map created using photogrammetry analysis of the
panchromatic optical image, (iii) the hillshaded digital elevation model (DEM), (iv) slope map, and (v) – the terrain ruggedness index (TRI).
Data gaps are on steep slopes in shadow, resulting in low contrast and inability to derive heights from stereo image matching. (b) The SPOT
satellite multispectral image (left) and the resulting hillshaded DEM (right) derived from stereo panchromatic pairs. (Pléiades © CNES,
2016; distribution by Airbus DS/Spot Image.)
Fault accommodates the full shortening across the region, it
is also possible that the frontal faults extend further west be-
neath the city (Fig. 2b), hidden by the sediments of the cen-
tral depression. Our investigations using the SPOT satellite
DEM and point cloud data do not show any clear evidence
of a fault scarp within the central regions of the city. How-
ever, this could be masked by urban development, or the fault
could be buried, as in the case of the Pardisan thrust fault be-
neath the city of Tehran in Iran, where no primary fault is vis-
ible at the surface (Talebian et al., 2016). Similarly, the 2011
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1538 E. Hussain et al.: Seismic risk in Santiago
Figure 4. The northern section of the Pléiades-derived DEM (2m resolution) indicated in Fig. 3. The black points on the profiles are ground
pixels within a 30 m swath of the profile line from the Pléiades-imagery-derived point cloud, while the grey are from the SPOT point cloud.
The red and green lines are best fit lines through the point clouds either side of the fault scarps. The scarp height is estimated from the vertical
offset between these two lines. The blue lines are the height of anticlines measured from the downslope side.
Mw6.3 Christchurch (New Zealand) earthquake occurred on
a previously unrecognised fault buried right beneath the cen-
tre of the city (Elliott et al., 2012), but its impact was much
greater than the larger earthquake (Mw7.1) that struck the
year before outside of the city.
Riesner et al. (2017) proposed a first-order model for the
deeper structure of the San Ramón Fault and found that it
constitutes the frontal expression of a major west-vergent
fold-and-thrust belt that extends laterally for thousands of
kilometres along the western flank of the Andes (see also
Armijo et al., 2015). Since it is well known that frontal faults
of fold-and-thrust belts tend to migrate out of the central
highlands through progressive growth of new faulting (e.g.
Davis et al., 1983; Dahlen, 1990; Reynolds et al., 2015), it is
not unreasonable to assume that younger faults would extend
further west from the San Ramón Fault. We project the loca-
tion of this inferred fault along strike from the West Andean
cordillera frontal fault (Fig. 2b) and assume the dip is the
same as for the San Ramón Fault. In Sect. 3 we will explore
the losses from moderate-magnitude earthquakes (Mw6 and
Mw6.5) on this hypothesised buried fault within the city with
larger magnitude events on the San Ramón Fault (Mw7 and
Mw7.5), consistent with the palaeo-seismic trench work of
Vargas et al. (2014). The magnitudes for the central Santiago
splay scenario were determined using standard fault scaling
relationships (Wells and Coppersmith, 1994), where a rup-
ture with a length of ∼25 km, a width of ∼12 km, and a
co-seismic slip of 1 m would result in an earthquake with
moment magnitude in the range of 6–6.5.
In 1647 a large earthquake destroyed Santiago, which at
the time was a 100-year-old Spanish town, and killed an es-
timated one-fifth of its inhabitants (de Ballore, 1913; Udías
et al., 2012). Details of this earthquake remain poorly un-
derstood, and there is much debate on the epicentral loca-
tion (e.g. Lomnitz, 1983; Comte et al., 1986; Lomnitz, 2004).
Lomnitz (1970) notes that historical descriptions of the dam-
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1539
Figure 5. The southern section of the Pléiades DEM indicated in Fig. 3. The black points on the profiles are ground pixels from the Pléiades-
imagery-derived point cloud, while the grey are from the SPOT point cloud.
age indicate an epicentre within 50 miles of Santiago at most,
while Poirier (2006) mentions that the earthquake did not
produce any devastating tsunamis, both pointing to a source
on a fault near the city. As there is no evidence of this earth-
quake in the trench studies along the San Ramón Fault (Var-
gas et al., 2014), it is unlikely that the earthquake originated
there as suggested by Rauld (2002). While it is possible that
the earthquake occurred on one of the faults in the princi-
pal cordillera (Farías et al., 2010), our assumption was that
the main activity in the region is on the frontal portions of
the fold-and-thrust belt (e.g. Dahlen, 1990; Reynolds et al.,
2015). It is also possible that the 1647 earthquake occurred
on the West Andean faults to the north and south of the city
(Fig. 2); however, due to the lack of any evidence for either
case we feel it is reasonable to explore the worst-case sce-
nario of an earthquake occurring on the extension of these
faults through the city.
Another possible candidate for the 1647 earthquake is an
earthquake in the subducting slab beneath the city. Therefore,
we also examine intra-slab faulting scenarios. This is moti-
vated by the most damaging earthquake in terms of fatali-
ties in south central Chile in the previous century. The 1939
earthquake (Ms∼7.8) caused ∼28000 deaths (many times
more than the great 1960 subduction earthquake) and pro-
duced extensive damage to the city of Chillán (Saita, 1940;
Frohlich, 2006), about 200 km south of Santiago. Beck et al.
(1998) modelled the first P-wave motions for this earthquake
and concluded that it was a normal-faulting event within the
down-going slab at a depth of 80–100 km. Since the subduct-
ing slab beneath Santiago is also about 80 km beneath the
city (Hayes et al., 2012), we explore the losses from similar
normal-faulting events in the slab beneath Santiago.
3 Earthquake scenarios for the residential-building
stock
The development and implementation of measures to min-
imise the physical impact due to earthquakes require a com-
prehensive understanding of the potential for human and eco-
nomic losses, which is usually achieved through earthquake
risk assessment studies (e.g. Silva et al., 2015b; Chaulagain
et al., 2016). For risk management purposes, risk is the po-
tential economic, social, and environmental consequences of
hazardous events that may occur in a specified period of time
(see Grossi and Kunreuther, 2005, for details).
We use the GEM OpenQuake Engine v3.3.2 (Silva et al.,
2014; GEM, 2019) to calculate the damage and losses to
residential buildings from earthquake scenarios on predeter-
mined faults for all 52 communes that make up the Santiago
metropolitan region (∼1.1 million buildings). In the sections
below we briefly describe the key components of the dam-
age and risk calculation: exposure, hazard, and vulnerability
(Fig. 6).
3.1 The residential-building exposure model
In order to describe the residential-building stock of the San-
tiago metropolitan region we used the exposure model es-
tablished by Santa-María et al. (2017). The exposure model
was built using data from the national population and hous-
ing census surveys (2002 and 2012) and information from
the 2002–2014 Formulario Único de Estadísticas de Edifi-
cación (Unique Edification Statistic Form; UESF). The ex-
posure model describes the number and distribution of resi-
dential buildings at the census block resolution and contains
information on the main material of construction, number
of storeys (Fig. S2), age of construction, expected ductil-
ity, the number of people living in each building, and the
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1540 E. Hussain et al.: Seismic risk in Santiago
Figure 6. A graphical representation of the damage and loss calculation workflow using the Global Earthquake Model’s OpenQuake Engine
(Silva et al., 2014). Black boxes represent model calculators, while white boxes are data inputs/outputs.
replacement cost per unit area. The replacement cost in-
cludes an estimate of the structural, non-structural, and con-
tent costs of each building. We assume the earthquake sce-
narios occur at night, and therefore the residential fatality es-
timates represent the night-time losses. Table 1 summarises
the most important information in the residential-building
exposure model. The most commonly used building mate-
rial for residential buildings is masonry (79 % of all build-
ings), with confined masonry the dominant building typology
(39 % of total buildings), followed by reinforced-masonry
(26 %) and unreinforced-masonry structures (14 %). To im-
prove computing efficiency we resample the Santa-María
et al. (2017) exposure model from the census-block resolu-
tion to a 1km ×1km grid (Fig. S1). Our exposure model re-
veals that Puente Alto, Maipú, and La Florida are the most
populated communes, together accounting for about 26 % of
all residential homes in the Santiago metropolitan region.
Puente Alto and La Florida are centred on the San Ramón
Fault.
Figure 7 shows the fraction of the total building stock
in each commune categorised into the five building classes
against the percentage of people living below the poverty
line, defined as USD 400 per month for a family of four
(Ministerio de Desarrollo Social, 2016). The coefficients of
determination (R2) for the fit through each building class
are 0.63, 0.01, 0.45, 0.20, and 0.73 for reinforced-concrete
(RC), confined-masonry (MCF), reinforced-masonry (MR),
unreinforced-masonry (MUR), and wooden (W) buildings
respectively. It is clear that the scatter in the data for
the masonry buildings is large and reflected in the low
R2values, particularly for MCF buildings, implying lit-
tle correlation between levels of poverty and the fraction
of confined-masonry buildings. However, we find that the
fraction of reinforced-concrete buildings decreases signifi-
cantly with the proportion of people living below the poverty
line. This trend is balanced by an increase in the fraction
of wooden, reinforced-masonry, and unreinforced-masonry
buildings with level of poverty.
3.2 Definition of the earthquake scenarios
Unlike probabilistic seismic-hazard analysis where the risk
calculation is initiated with a stochastic event dataset, in
this study we calculate the damage and losses for specific
earthquake scenarios on predetermined faults. We chose a
scenario-based approach because it provides clear commu-
nication of the relative scale of potential damage and losses
from the recently recognised proximal San Ramón Fault
versus that from the better-characterised offshore subduc-
tion faulting, which is important for emergency management
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1541
Table 1. Summary of the building classes and typologies in our Santiago residential-building exposure model, based on
Yepes-Estrada et al. (2017).
Class Building typologies (GEM taxonomy) Count % of total Storeys
RC Non-ductile reinforced-concrete walls (CR_LWAL-DNO) 79 198 5.65 1–7
Ductile reinforced-concrete walls (CR_LWAL-DUH) 72 882 5.20 1–8+
MCF Non-ductile confined-masonry walls (MCF_LWAL-DNO) 395 349 28.19 1–3
Ductile confined masonry (MCF_LWAL-DUH) 149 714 10.68 1–5
MR Ductile reinforced-masonry walls (MR_LWAL-DNO) 259 223 18.48 1–3
Non-ductile reinforced masonry (MR_LWAL-DUH) 109 430 7.80 1–5
MUR Non-ductile unreinforced-masonry walls (MUR_LWAL-DNO) 161779 11.54 1–2
Non-ductile unreinforced-adobe walls (MUR-ADO_LWAL-DNO) 32 160 2.29 1–2
W Non-ductile light-wood walls (W-WLI_LWAL-DNO) 12796 0.91 1–2
Ductile light-wood walls (W-WLI_LWAL-DUM) 129 372 9.23 1–3
UNK Unknown or insufficient information available (UNK) 319 0.02
Total 1 402 222
Figure 7. The fraction of residential buildings by building class –
RC is reinforced concrete, MCF is confined masonry, MR is rein-
forced masonry, MUR is unreinforced masonry, and W is wooden
construction – against the proportion of people living below the
poverty line in the communes of the Santiago metropolitan region.
Solid lines represent best-fit trends through the data (linear for all
cases except for reinforced concrete, which is exponential), with
coefficients of determination of 0.63, 0.01, 0.45, 0.20, and 0.73 for
RC, MCF, MR, MUR, and W respectively. The poverty line is de-
fined as USD 400 per month for a family of four (Ministerio de
Desarrollo Social, 2016).
planning and for raising societal awareness of risk (Silva
et al., 2014).
We modelled the San Ramón Fault as a set of four rect-
angular slip planes (total length of 35 km) to account for the
changes in geometry along strike due to fault segmentation
from our DEM analysis (Fig. S3). The prescribed fault planes
dip at 45◦to the east and extend from the surface down to
12 km depth based on the structural cross sections drawn by
Armijo et al. (2010) and the depth of microseismicity deter-
mined by Pérez et al. (2014) as indicators of the down-dip
width that is locked and accumulating strain. The location
of the hypothesised splay fault is in line with the West An-
dean Front and 12 km west of the San Ramón Fault, consis-
tent with the approximate 10 km spacing inferred in the ma-
jor thrust faults beneath the San Ramón–Farellones Plateau
(Pérez et al., 2014). We represent the splay fault using a sin-
gle 45◦eastward-dipping rectangular plane extending from
0.5 km below the surface down to 12 km depth, with a north–
south strike running 25 km along longitude 70.65◦W. The
deep intra-slab fault scenario is modelled using a single
westward-dipping rectangular plane with a length of 35 km
in the subducting slab beneath the city. We used a 70◦dip
for the intra-slab fault to represent a similar earthquake to
the 1939 Chillán earthquake, for which Beck et al. (1998) es-
timated a 60–80◦dipping fault plane. We used the Slab1.0
model (Hayes et al., 2012) to set the top depth and bottom
depth of the fault plane at 85 and 98 km respectively for this
locality. Earthquakes on the San Ramón and Santiago splay
faults are prescribed as having a pure thrust mechanism (rake
+90◦), while the intra-slabs are normal (rake −90◦). The
fault characteristics are summarised in Table S1.
The hazard component of the calculation concerns de-
termining the spatial pattern of the key shaking parameters
from each scenario event by employing a ground motion pre-
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1542 E. Hussain et al.: Seismic risk in Santiago
diction equation (GMPE). The hazard parameters used here
are peak ground acceleration (PGA) and spectral accelera-
tion (SA). There are many GMPEs available in the litera-
ture (see Douglas and Edwards, 2016, for a review and http:
//www.gmpe.org.uk, last access: 1 January 2019, for an up-
dated compendium). In our analysis we use three equations
for shallow-crustal earthquakes (Akkar et al., 2014; Bindi
et al., 2014; Boore et al., 2014) and two for the intra-slab sce-
nario calculations (Abrahamson et al., 2016; Montalva et al.,
2017). These were selected for the OpenQuake Engine ac-
cording to expert opinion during the Global GMPEs project
(Stewart et al., 2012, 2015) and have been updated since.
Averaging several selected GMPEs helps to partially prop-
agate the epistemic uncertainty of the distribution of shaking
that arises from a non-perfect knowledge of ground motion.
There are several methods to calculate the distance from each
exposure point to the rupture. To remain consistent across
the GMPEs we implement the form of the equations that use
the Joyner–Boore distance, defined as the shortest horizontal
distance from each exposure element to the surface projec-
tion of the rupture area. However, we present the damage
and loss results at the district level by calculating the sum of
the losses of all points within each district.
For each scenario we produce 1000 realisations of the
ground motion in the region to account for the aleatory vari-
ability in the ground motion and assume the entire fault rup-
tures in the earthquake. We account for the spatial correlation
of the intra-event variability during the generation of each
ground motion field according to the methods described by
Jayaram and Baker (2009) to ensure assets located close to
each other will have similar ground motion levels.
For the 2010 Mw8.8 Maule earthquake, we directly used
the USGS ShakeMap as the input ground shaking for the
damage and risk assessment calculations (see Villar-Vega
and Silva, 2017, for details of this procedure).
3.3 Site effects
The Santiago metropolitan region is located in a narrow basin
between the Andes and coastal mountains filled with Qua-
ternary fluvial and alluvial sediments (Armijo et al., 2010).
Using numerical simulations of the Santiago basin taking ac-
count of the superficial geology, Pilz et al. (2011) showed
that there is a strong and sometimes complex basin amplifi-
cation effect on the peak ground velocity from hypothetical
earthquakes on the San Ramón Fault. While in this study we
are unable to account for the complexities of basin resonance
and topography, we attempt to take into account the basin
amplification effect in our ground motion calculations by us-
ing the Vs30 values, the shear-wave velocities in the top 30 m
of soil.
In this study two datasets were used to obtain the Vs30
information for Santiago. The first consists of local micro-
zonation studies, which contain seismic-zonation maps (Pas-
ten, 2007; Leyton et al., 2011) and proposed Vs30 values for
Figure 8. A map of the Vs30: shear-wave velocity in the top 30 m
of soil in metres per second at the exposure locations. The circled
data are from microzonation studies (Leyton et al., 2011; Humire-
Guarachi, 2013), and the remaining are estimates derived from the
topographic slope (Allen and Wald, 2007). The green lines indicate
the surface trace of the San Ramón Fault used in the seismic-risk-
scenario calculations. Red colours indicate a relatively high Vs30
value and are generally in regions with exposed or shallow bedrock.
Relatively slow Vs30 values are associated with sedimentary basins.
soil types in each zone, taking into account additional infor-
mation from soil penetration tests (Humire-Guarachi, 2013).
However, given the cost and time demand of such studies,
microzonation maps are usually focussed on limited areas.
Therefore, for the remaining zones we supplemented the mi-
crozonation data with velocities from the USGS Global Vs30
Map Server (Allen and Wald, 2007). This method derives
maps of seismic site conditions using topographic slope as
a proxy, assuming that stiffer materials (i.e. higher Vs30 val-
ues) are more likely to maintain a steep slope, while deep
basin sediments are deposited mainly in environments char-
acterised by a lower velocity.
Figure 8 shows the Vs30 values used in this study at the
building exposure locations, with the values from the micro-
zonation studies indicated in circles. Note that this will prob-
ably be an underestimate of the full basin effects (Joyner,
2000). For example, not accommodating for basin resonance
will mean that our models do not take into account the par-
ticular vulnerability of buildings of certain heights that are
prone to resonance, which was an important factor for ex-
ample in the Kathmandu rupture and basin amplification in
Nepal, with 4–5 s of resonance (Galetzka et al., 2015).
3.4 Building fragility and vulnerability models
The physical, or structural, vulnerability for a built system is
defined as its susceptibility to losses when subjected to earth-
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1543
quake shaking. In our scenario calculations we use two main
forms of vulnerability models: fragility functions, which are
used to relate earthquake shaking to certain levels of physi-
cal damage to a building (e.g. extensive damage, collapse),
and vulnerability functions (structural and occupants), which
relate the earthquake shaking of a structure to the economic
and human losses.
Villar-Vega et al. (2017) analytically derived fragility
functions for the 57 building classes in the exposure dataset
developed for the South America Risk Assessment (SARA)
project (Yepes-Estrada et al., 2017). For our analysis we use
the subset of these equations that represents the building
exposure in the Santiago metropolitan region (Table 1 and
Fig. S1). To derive the fragility functions, Villar-Vega et al.
(2017) represented the structural capacity of each building
class by a set of single-degree-of-freedom (SDOF) oscilla-
tors. Each oscillator was subjected to a suite of ground mo-
tion records representative of the South American tectonic
environment and seismicity using GEM’s Risk Modellers’
Toolkit (Silva et al., 2015). From each analysis, the max-
imum spectral displacement of each SDOF oscillator was
used to allocate it into a damage state (e.g. collapse). In this
paper, we focus our scenario analysis on the spatial distribu-
tion of collapsed buildings, which comprises not only phys-
ically collapsed buildings but also partially collapsed struc-
tures (Villar-Vega et al., 2017).
The simplification of each building typology to a single-
degree-of-freedom oscillator means that the calculated
fragility functions only approximate the building response
to ground shaking. Therefore these would not be sufficient
to investigate building-by-building scale losses from earth-
quake shaking. However, we believe it is sufficient to ex-
plore aggregated district level losses. And so, while the sce-
nario calculations are done on a 1km ×1 km gridded expo-
sure model, the losses presented in the following sections ag-
gregate these to the district level.
A vulnerability curve establishes the probability distribu-
tion of a loss ratio (e.g. fatalities / total number of occupants)
given a shaking intensity measure level (Figs. 9 and S5).
Vulnerability curves are generally empirically derived using
loss data, usually collected through insurance claims or gov-
ernmental reports. A database of fragility and vulnerability
functions can be found in the OpenQuake platform (Yepes-
Estrada et al., 2016; Martins and Silva, 2018). We used these
vulnerability functions to directly model fatalities and repair
costs, where the loss ratio for the former would be the ratio
of fatalities to exposed population, and for the latter the ratio
would be that of repair cost to cost of replacement for a given
building typology.
3.5 Residential-building collapse and loss results
The median predicted ground motion for the larger earth-
quake scenario considered for each fault is given in Fig. 9. It
shows the relatively simple ground motion patterns from the
single rectangular Santiago splay fault and a more complex
pattern from the San Ramón Fault. For the San Ramón case
most of the high ground shaking is around the communes
to the east of the city, while for the splay fault case there is
a more even distribution of shaking across the central com-
munes.
Our damage and loss results, averaged over the GMPEs
used in each scenario calculation, reveal that the collapsed-
building estimates for each scenario are distributed unevenly
across the city (Fig. 10). Figure 11 shows a summary of the
damage and loss results for all scenarios. It is clear that the
damage and losses are greater for the larger-magnitude earth-
quake considered in each case as one would expect since
larger earthquakes, at a given depth, produce higher-intensity
ground shaking.
For the San Ramón scenarios the losses are mostly con-
centrated in the communes around the fault. Most collapsed
buildings are located in Puente Alto (23 100–28 800; 16%–
20 %), La Florida (12 400–15700; 15 %–19%), and Las Con-
des (10 500–12 800; 20 %–25 %), where the first numbers in
the brackets are the building collapse counts for the two San
Ramón Fault scenarios and the second two numbers the per-
centage of collapse of the total number of exposed buildings
in the commune. We calculate fewer residential-building col-
lapses in Peñalolén (5900–7400; 14 %–17 %) and La Reina
(5000–6200; 18 %–23 %) despite these communes also be-
ing located on the fault. This discrepancy could be explained
through a combination of greater exposed population – and
thus more residential buildings (Fig. S6 shows the percentage
of collapsed buildings) – and the level of poverty.
Puente Alto has the largest population of these communes
(622 356) and also the greatest percentage living below the
poverty line (Table 2). Puente Alto and La Florida also gen-
erally contain a greater proportion of masonry constructions
(93 % and 86 % compared to 79 % and 71 % for Peñalolén
and La Reina respectively), which perform poorly in the
San Ramón earthquake scenarios. While Peñalolén has a low
fraction of RC residential buildings (5 %), which generally
perform well in our calculation, it is compensated by a large
proportion of wooden structures (17 %), which perform the
best when subject to seismic shaking.
In general the greatest percentage of collapsed buildings
(Fig. S6) occurs, as expected, in the communes directly on
the fault, e.g. Vitacura (3000–3600; 21%–25 %) and Las
Condes (10 500–12 800; 20 %–25 %). However there are sev-
eral communes with high collapse fractions that are not lo-
cated on the fault, notably Ñuñoa (8400–10 600; 20%–25 %)
and Macul (4200–5300; 18 %–23%). These both have a high
fraction of unreinforced-masonry buildings, 27 % and 13 %
respectively compared to an average of 9% for the com-
munes on the fault (Table 2). Unreinforced-masonry build-
ings are the most likely building class to collapse in all the
scenarios considered in this study (Table 4).
In terms of anticipated fatalities for the larger San Ramón
scenario (Fig. 12), the communes of Ñuñoa and Providen-
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1544 E. Hussain et al.: Seismic risk in Santiago
Figure 9. The estimated median peak ground acceleration (PGA) as a fraction of gravity gfor the largest earthquake scenario considered
for each fault. For the San Ramón (red line) and Santiago splay fault (green line) cases, these estimates were obtained using the Akkar et al.
(2014) ground motion prediction equation, while those for the intra-slab fault are estimates from the Abrahamson et al. (2016) equation.
The USGS peak ground accelerations for the Mw8.8 Maule earthquake are shown at the bottom. The ground motions for the full set of
earthquake scenarios are given in Fig. S4.
Table 2. Exposed populations and buildings for the 10 most affected communes in terms of average modelled building collapse across the
six earthquake scenarios (full list in Supplement; Tables S2 and S3). The communes are ranked in order of average collapse count.
Communes Area Population Popndensity Popnbelow Buildings in exposure model (% of total)
km2(per km2) poverty line∗(%) RC MCF MR MUR W Total
Puente Alto 88 622 356 7072 8.0 3 61 29 3 4 143 463
La Florida 71 356 925 5027 3.1 6 51 26 9 8 81 493
Santiago 22 371 250 16 875 5.9 38 19 15 27 1 57 341
Ñuñoa 17 273 354 16 080 2.4 36 32 18 13 1 42 598
Las Condes 99 296 251 2992 0.6 38 31 21 8 1 51 646
Maipú 133 608 094 4572 5.2 4 44 35 12 6 142 828
Peñalolén 54 197 909 3665 4.8 5 39 30 9 17 42 562
La Pintana 31 191 306 6171 13.9 2 39 36 10 13 40 847
Providencia 14 88 928 6352 0.7 51 25 18 6 0 22 080
Macul 13 116 694 8976 5.3 16 32 17 27 7 23 528
∗defined as USD 400 monthly income (in 2015 US dollars) for a family of four (Ministerio de Desarrollo Social, 2016).
cia (10 km west of the San Ramón fault trace) are modelled
as experiencing the highest fatality rates of 4–5 per 1000
per 1000 people (Fig. S7). In terms of absolute numbers,
the largest number of fatalities (Fig. 12) occurs in Ñuñoa
and Las Condes (1120–1420 and 1080–1330 respectively).
Overall the fatalities across the region are estimated in the
range of 9700–12 700 (Fig. 11), resulting in a fatality rate of
0.15 %–0.19 % (Table S3). The residential losses in terms of
replacement costs average USD8–10 billion (5 %–7 % mean
loss ratio). The greatest replacement costs are for Santiago
(USD 1.3 billion), but they are also high (USD 0.5+bil-
lion) for the communes of Puente Alto, Las Condes, and La
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1545
Figure 10. The distribution of collapsed buildings for the earthquakes considered in each set of magnitude pair scenarios for the San Ramón
Fault (green line; a–b), the Santiago splay fault (dashed cyan line; c–d), and a deep intra-slab fault (e–f). The collapse counts are the average
for the GMPEs used in each calculation and include the total count of both complete and partial collapses. Note that the range of the colour
scale changes between the upper four and lower two panels. The collapse fraction for each commune is given in the Supplement (Fig. S6).
Names of all the communes are given in Fig. S1.
Florida on top of the San Ramón Fault, as well as for Ñuñoa
further west (Fig. S8).
For the earthquake scenarios on a buried fault splay be-
neath the centre of the city, the distribution of collapsed res-
idential buildings is similar to the San Ramón scenario with
damage concentrated towards the eastern communes of the
city on the hanging wall of the fault. Most collapses oc-
cur in Puente Alto (13 600–20 900; 9 %–15 %) and Santi-
ago (10 000–15 400; 17 %–27 %). As in the case of the San
Ramón Fault the high collapse count in Puente Alto probably
reflects the large number of residential buildings in that com-
mune. Of the communes directly next to the fault splay, San-
tiago has the largest number of residential buildings (57 341).
However, the greatest impact in terms of collapse fraction
is in the communes in the central districts near the fault,
with the highest fraction of collapse occurring in Santi-
ago (10 000–15 400; 17 %–27 %), Providencia (3400–5500;
15 %–25 %), Independencia (2500–3700; 16 %–24 %), and
Ñuñoa (6600–10 300; 15 %–24 %). The estimated fatalities
for the buried splay scenarios are similar to or slightly above
those for the San Ramón cases despite being a magnitude
lower in scale, in the range of 6500–11 500 (0.10 %–0.17%
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1546 E. Hussain et al.: Seismic risk in Santiago
Figure 11. A summary of the total number of building collapses
and fatalities as well as the total replacement costs for each sce-
nario calculation. The solid circles are the average across the GM-
PEs considered in each scenario, which are indicated by the smaller
polygons. The spread in the estimates from the GMPEs indicates
the epistemic uncertainty in our calculations. Error bars represent
1 standard deviation determined from the 1000 Monte Carlo simu-
lations. The yellow star denotes the actual number of building col-
lapses (4306) in Santiago in the 2010 Maule earthquake (Elnashai
et al., 2010).
loss ratio). The most affected communes are also similar, in-
cluding Santiago, San Miguel, Providencia, and Ñuñoa, with
fatality fractions of 4–5 per 1000 for the larger Mw6.5 sce-
nario (Fig. S7). The greatest number of fatalities for both
magnitudes (Fig. 12) is also in Santiago and Ñuñoa (870–
1600 and 710–1310 respectively; Table S3). The residential
replacement costs are USD 6.1–9.6 billion (4 %–6 % loss ra-
tio) for the two magnitude scenarios (Table S3). The greatest
losses are in Santiago, Ñuñoa, and Puente Alto (Fig. S8).
The overall collapse count for the magnitude 7 deep intra-
slab scenario is small, but the magnitude 7.5 scenario re-
sults in a substantial number of collapsed buildings (about
60 000), with most collapsed homes and fatalities (Fig. 12)
generally located in the more populous communes. The ex-
tent of collapse across the city is more diffusive due to the
buried nature of the intra-slab source, with building collapse
up to 8% in the centre (Lo Espejo, San Joaquin, and Inde-
pendencia). The estimated total number of fatalities (Fig. 11)
for the larger event is 3180 (0.05%), with the largest number
of fatalities (200–300; Table S3) each in Santiago, Ñuñoa,
Maipú, and Puente Alto (Fig. 12). The estimated replacement
cost is USD 3 billion (2 % loss ratio), with the greatest costs
distributed in the same commune as fatalities (Fig. S8).
Central government statistics estimate a total collapse
count of 81 444 residential buildings throughout Chile in the
2010 Mw8.8 earthquake, with the most damage occurring
in the Maule, Biobío, O’Higgins, and Santiago metropolitan
regions (Elnashai et al., 2010; de la Llera et al., 2017) and
4306 of these collapses occurring in the Santiago metropoli-
tan region (yellow star in Fig. 11). While the collapse count
is smaller than our modelled estimate of 9800 ±8000, it is
within the error margin. The discrepancy could have arisen
due to a slightly different exposure model. The actual expo-
sure in 2010 would have been different than our exposure
model estimates, which use data from 2014. Moreover, there
is often ambiguity regarding the classification of actual struc-
tural collapse and damage beyond repair (and thus in need of
demolition). See Villar-Vega et al. (2017) for a discussion on
this topic.
While we estimate building collapse fractions up to 21 %
(Providencia) for the Mw7 San Ramón scenario, the aver-
age collapse ratio across all the communes is ∼8 %. This is
larger than the 6.5 % estimated by Vaziri et al. (2012) for a
magnitude 6.8 earthquake on the fault. However, their esti-
mate of 14 000 fatalities is larger than the 9700 we estimate
for a Mw7 earthquake. This difference is most likely due to
variations in the exposure model and calculation procedure
(i.e. choice of ground motion models). But since Vaziri et al.
(2012) used an industry exposure model we are not able to
determine the exact cause behind the difference in our esti-
mates.
In general, across all the scenarios considered in this study,
the largest number of collapsed buildings occurs in the highly
populous communes (which therefore have more buildings)
close to the fault. However, the collapse and fatality frac-
tions (the number of collapses or fatalities over the exposure;
Figs. S6 and S7) reveal particularly vulnerable areas. Sev-
eral communes experience relatively large damage and loss
fractions, which is an indication of the vulnerability of the
communities in these communes. Of particular importance
are Ñuñoa, Providencia, and Santiago, which generally have
high loss fractions with 3, 3, and 2 fatalities per 1000 people
and damage fractions of 15 %, 15 %, and 15 % respectively,
averaged over the six earthquake scenarios. In comparison
the average loss fraction across all communes and all scenar-
ios is 0.8 fatalities per 1000 people and 7 % building collapse.
Therefore targeted measures to retrofit particularly vulner-
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1547
Figure 12. The estimated fatalities in each commune for the earthquakes considered in each scenario for the San Ramón Fault (red line), the
Santiago splay fault (dashed cyan line), and a deep intra-slab fault. Note that the range of the colour scale changes between the upper four
and lower two panels.
able residential buildings (unreinforced masonry) could re-
duce the seismic risk faced by communities living in these
communes.
4 Non-residential insured losses
We also used Risk Management Solutions’ (RMS) commer-
cial Chile Earthquake Model, developed in 2011, with the
most recent industry exposure database (IED) to derive in-
dustry loss estimates for specific earthquake scenarios. The
exposure model contains only non-residential-building in-
formation and does not include public infrastructure such
as roads or bridges. The exposed values (or “total insured
value”, TIV) in this dataset include commercial buildings,
contents, and business interruption and are aggregated at the
commune level.
Table 3 and Fig. 13a give a summary of the average gross
loss ratios (calculated loss over the total insured value) for
the maximum-magnitude scenarios on the San Ramón and
Santiago splay faults. The gross losses are the full replace-
ment costs for the property after accounting for insurance
penetration and after the application of deductibles, limits,
and co-insurance. This is often referred to as the insured
loss. It is worth noting that these losses are a subset of
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1548 E. Hussain et al.: Seismic risk in Santiago
Table 3. Gross loss (GR) ratios for the Santiago non-residential ex-
posure.
Event Average GR loss ratio
Fault Magnitude (Loss/total insured value)
San Ramón 7.0 5 %
San Ramón 7.5 10 %
Santiago splay 6.0 0 %
Santiago splay 6.5 5 %
the full economic losses in an earthquake since the gross
losses account for insurance penetration, which is always
less than 100 %. The average insurance penetration (residen-
tial and non-residential) in the Santiago metropolitan region
was 31.5 % in 2011. However, only an estimated 30 % of
small commercial business owners had earthquake insurance,
which rises to greater than 75 % for large commercial and in-
dustrial facilities (Muir-Wood, 2011).
5 Discussion
Whilst a detailed past record of earthquakes and variability of
recurrence on the San Ramón Fault is not precisely known,
the palaeo-seismic work of Vargas et al. (2014) tentatively
points towards a recurrence interval of the order of ∼8 kyr;
determined from records of two past earthquakes 17–19 and
∼8 kyr ago. Given that the last event was ∼8 kyr ago it is
prudent to consider a San Ramón rupture scenario of Mw7.5
as a real possibility to plan for. Vargas et al. (2018) find that
present urbanisation of eastern Santiago reached 55 % of the
San Ramón fault trace, evidencing that this active geological
structure has not been considered in urban regulations devel-
oped for the metropolitan region.
5.1 Residential collapse by building class
Buildings of differing construction material type are known
to perform markedly differently under seismic shaking (Park
and Hamza, 2016). We aim to identify the expected better-
and less-well-performing building classes exposed in Santi-
ago under our varying earthquake shaking scenarios. In or-
der to compare the collapse ratios for the different building
classes defined in Table 1 (RC, MCF, MR, MUR, and W),
we calculate the normalised collapse fraction, NCFs
t, for each
building class according to
NCFs
t=cs
t/et
(Cs/E),(1)
where etis the number of exposed residential buildings of
building class t, and cs
tis the number of collapsed buildings
of the same class in earthquake scenario s.Csis the total
number of collapsed buildings in scenario s, and Eis the
total number of exposed buildings in the city.
Therefore, if NCFt>1, then typology tis more likely
to collapse than the average. The results of this normal-
isation are shown in Table 4. Across the six earthquake
scenarios, we find that reinforced-concrete (RC) residen-
tial buildings perform best with a normalised collapse frac-
tion NCFRC =0.5. It is also clear that unreinforced-masonry
(MUR) structures collapse the most across all scenarios, with
an average NCFMUR =2.6, implying that MUR structures
are over 2.5 times more likely to collapse than the aver-
age. Typically, 3 %–13 % of buildings in the most affected
communes (Table 2) are unreinforced-masonry constructions
except for Santiago and Macul, where 27 % of the build-
ings are MUR. This is why we typically observe relatively
large collapse fractions in Santiago (1700–15 400; 3 %–
27 %) and to a lesser extent Macul (600–5300; 3 %–23 %)
over the six earthquake scenarios considered in this study.
Wooden residential homes (W) perform very well with an
average NCFW=0.2. Confined-masonry (NCFMCF =0.9)
and reinforced-masonry (NCFMR =0.8) perform better than
unreinforced-masonry buildings and slightly better than the
average. It is clear that masonry construction in general per-
forms worse during earthquakes.
5.2 Magnitude–distance trade-off
The calculated collapses and losses for each of the sce-
narios using the residential-building exposure show the ex-
pected pattern of greater losses for a larger earthquake on
a given fault (Fig. 11 and Table 5). Our calculations show
that the San Ramón earthquake scenarios are the most dam-
aging to the city, with 136400 residential-building collapses
from a magnitude 7 and 181 300 collapses from a magni-
tude 7.5 earthquake on the fault, a difference of 25%. The
estimated damage from the buried Santiago splay fault com-
prises 107 300 collapses for a magnitude 6 and 172 200 col-
lapses for a magnitude 6.5 earthquake, a difference of 38 %.
We note a similar pattern in the losses, with an increase of
24 % and 20% in fatalities and replacement cost respectively
between a magnitude 7 and 7.5 earthquake on the San Ramón
Fault and an increase of 38 % and 43 % on the splay fault
earthquakes. It is clear from our calculations that a half mag-
nitude increase in earthquake size results in more damages
and losses from the Santiago splay fault than the San Ramón
Fault. This is probably because shaking from the splay fault
exposes more of the densely populated communes in the cen-
tre of the city, where a small increase in shaking can have
more impact.
In all cases, the magnitude 8.8 Maule earthquake pro-
duces fewer losses to the city than the smaller local earth-
quakes despite having 100–10 000 times the moment re-
lease of the other scenarios considered here. According to
our models, it is clear that there is a trade-off between
earthquake magnitude and distance in terms of residential-
building collapses and fatalities. Chatelain et al. (1999) also
found a similar trade-off in their earthquake risk assessment
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1549
Figure 13. The distribution of (a) non-residential and (b) residential replacement costs for maximum-magnitude scenarios considered for the
San Ramón Fault and the Santiago splay fault (moment magnitude 7.5 and 6.5 respectively). The gross loss ratios represent the calculated
losses in each scenario over the total insured value after the application of policy conditions and deductibles. The residential replacement
costs are the costs to repair or replace buildings and their contents damaged in each scenario. The residential replacement cost maps for all
scenarios are given in Fig. S8.
Table 4. The fraction of collapsed residential buildings by building class (Table 1) normalised to the total collapse fraction in each earthquake
scenario. SR: San Ramón; SS: Santiago splay; IS: intra-slab. The number indicates the moment magnitude of the earthquake source. Values
greater than 1 indicate that the building typology is more likely to collapse than the average.
Building Total Normalised collapsed fraction
typology exposed SR7 SR7.5 SS6 SS6.5 IS7.0 IS7.5
RC 152 080 0.7 0.7 0.6 0.7 0.1 0.2
MCF 545 063 0.9 0.9 0.8 0.9 0.8 0.9
MR 368 653 0.8 0.8 0.8 0.9 0.7 0.8
MUR 193 939 2.0 1.9 3.0 2.6 3.4 2.8
W 142 168 0.2 0.2 0.2 0.3 0.1 0.2
Total 1 401 903
for Quito, Ecuador. Therefore, simply focussing on large
offshore megathrust earthquakes would mask the significant
risks posed from moderate-size earthquakes on smaller but
more local active faults.
5.3 Residential and non-residential insured losses
Figure 13b shows the loss distribution of the residential-
building replacement costs due to a magnitude 7.5 earth-
quake on the San Ramón Fault and a magnitude 6.5 earth-
quake on a buried splay fault beneath the centre of the
city. While these are not directly comparable with the non-
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1550 E. Hussain et al.: Seismic risk in Santiago
Table 5. Summary of losses from each fault and earthquake scenario.
Fault Earthquake Building Fatalities Replacement cost
magnitude collapse (billions USD)
San Ramón 7 136 400 9670 8.3
San Ramón 7.5 181 300 12 650 10.4
Santiago splay 6 107 300 6470 6.1
Santiago splay 6.5 172 200 11 430 9.6
Intra-slab 7 23 500 930 1.2
Intra-slab 7.5 63 500 3180 3.1
Maule 8.8 9800 510 3.1
residential insured losses (Fig. 13a), partly due to the fact
that the insured losses are impacted by insurance penetra-
tion and also include business interruption and policy condi-
tions, there are some clear differences between the two that
are worth noting.
For the San Ramón scenario the highest residential-
building replacement costs are generally concentrated in the
communes with large populations (hence more residential
buildings) close to the fault and those with a large frac-
tion of reinforced-concrete buildings, which are more ex-
pensive than masonry structures. The greatest losses occur
in Puente Alto, La Florida, Las Condes, and Santiago, while
the insured losses were equally high in all communes along
the fault, including high loss ratios in Lo Barnechea and
Huechuraba, two communes not directly above the fault.
This difference reflects the concentration of high-value
commercial properties in the more affluent eastern com-
munes, where businesses are more likely to be insured (Muir-
Wood, 2011), particularly in Las Condes and La Reina. The
residential losses reflect the damages in highly populous
communes, as evidenced by the losses in Santiago, Ñuñoa,
and La Florida.
Similarly the losses in residential buildings for a magni-
tude 6.5 earthquake on the Santiago splay are concentrated
in the eastern communes on the hanging wall of the fault
(Santiago, Ñuñoa, and Puente Alto), while the insured losses
concentrate around the central business districts of Santiago
and Huechuraba.
5.4 Caveats and limitations
There are a number of caveats and sources of uncertainty
in estimating damage and losses from past and hypothetical
earthquake scenarios using the method described in this pa-
per. One of the main sources of uncertainty is the exposure
model, which does not exactly correspond to the exposed
population and building portfolio that were affected by a past
earthquake. For this study the exposure model was developed
by Santa-María et al. (2017) using information from census
surveys and housing information from 2014. Therefore, it is
important to note that the exposure model used in investigat-
ing the damage and losses for the hypothetical earthquake
scenarios on the San Ramón Fault, Santiago splay, and deep
intra-slab faults does not represent the exact current expo-
sure of the Santiago metropolitan region. This is particularly
important for Santiago, where rapid eastward expansion of
the city into the foothills of the San Ramón mountains puts
an increasingly greater population closer to the San Ramón
Fault (Fig. 1) and onto its hanging wall, where ground accel-
erations are typically higher. Therefore, methods that allow
for a near-continuous update of building inventories and lo-
cations are needed to maintain the veracity of exposure.
One of the largest sources of uncertainty in the calcula-
tions is in the GMPEs. In order to capture the epistemic
uncertainty in both median ground motion predictions and
their associated aleatory variability, we used several equally
weighted GMPEs (e.g. Bommer et al., 2005; Bommer et al.,
2010). The differences in the datasets used to derive each
GMPE and the way each GMPE calculates the ground mo-
tions are partly why the uncertainty range in our estimates of
the number of collapsed buildings is large (Fig. 11).
The development of fragility models also involves large
uncertainties. The fragility functions used in this study
were developed using a probabilistic approach where a set
of structures are tested against a suite of ground motion
records (Villar-Vega et al., 2017). Since it is time- and cost-
prohibitive to develop a fragility function for every build-
ing in the city, each of the buildings is allocated to a build-
ing typology within a general building class in the exposure
model. Each typology is then represented by a single-degree-
of-freedom block to calculate its response to the ground mo-
tion records. It is important to note that this level of simplifi-
cation adds uncertainty in the final response of any particular
typology in the event of an earthquake (see Villar-Vega and
Silva, 2017, for a discussion). Also, it is not possible to say
how any individual building will respond to an earthquake
using this approach as the fragility functions do not include
the unique complexities in the design and construction of ev-
ery building.
The results shown in this paper are from scenario-based
calculations on predetermined faults, and thus we cannot pro-
vide the relative likelihood of a shaking event as in a prob-
abilistic seismic-hazard analysis (PSHA) framework. How-
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1551
ever, this is an important start to motivate continued work
on the recurrence interval of faults in the region, begun by
Armijo et al. (2010) and Vargas et al. (2014) on the San
Ramón Fault. The SPOT DEM is not good enough to con-
clusively identify the presence of any geomorphic marker
that may result from a splay fault within the city. Its very
existence, let alone relative level of activity, remains an open
question. Higher-resolution DEMs or detailed field surveys
might be able to resolve this issue.
It is important to note that although the focus of this pa-
per has been to explore the direct damage and losses due
to earthquakes, in the case of an actual event there are of-
ten cascading hazards in the form of liquefaction, tsunamis,
landslides, and blocked waterways leading to floods, fires,
etc. that can lead to loss of lives and livelihoods (Gill and
Malamud, 2014). Nevertheless, historically most fatalities in
earthquakes were due to direct building collapse apart from
the large tsunami death tolls from a few 21st-century megath-
rust earthquakes (Ambraseys and Bilham, 2011).
6 Conclusions
In this study we use high-resolution DEMs of the metropoli-
tan area of Santiago city and the foothills of the San Ramón
mountains created using SPOT and Pléiades satellite imagery
to accurately map the surface expression and segmentation
of the San Ramón Fault. We recognise that estimates of the
impact from specific earthquakes (historical or hypothetical)
can support decision makers in the development of risk re-
duction strategies. We therefore use the OpenQuake Engine
to calculate damage and losses for realistic earthquake sce-
narios on the mapped San Ramón Fault as well as a potential
unknown fault, buried directly beneath the centre of the city,
and a deep intra-slab fault. We compare these losses with
those for the 2010 magnitude 8.8 offshore Maule earthquake
as a reference level of a recent event. Our calculations show
that there is a strong magnitude–distance trade-off in terms
of direct damage to the exposed building portfolio and fa-
talities, with smaller, more local shallow earthquakes caus-
ing greater losses to the city than a larger offshore megath-
rust earthquake. It is clear that the eastward expansion of the
city into the foothills of the San Ramón mountains has ex-
posed a large number of (predominantly affluent) people to
a future earthquake on the San Ramón Fault. While the re-
currence interval of large earthquakes on the fault are long
(∼8 kyr) and only very loosely constrained, the last Mw7.5
event was ∼8kyr ago (Vargas et al., 2014). So it is pru-
dent to consider the potential impacts of a San Ramón rup-
ture scenario of similar magnitude. We calculate losses using
scenario-based models of San Ramón earthquake ruptures in
the magnitude range of 7–7.5 under the current residential
exposure. Our models estimate 181000±80 000 partial to to-
tal building collapses, 12700 ±4500 fatalities, and replace-
ment costs of USD 10 ±3 billion for the larger-magnitude
earthquakes the San Ramón Fault can accommodate, assum-
ing all fault segments fail at once. While these numbers are
subject to considerable uncertainty arising from a changing
exposure as well as from uncertain ground motion predic-
tion equations and fragility functions, they provide an in-
formative guide to the potential scale of losses from a large
earthquake on the San Ramón Fault. For all modelled scenar-
ios, the most vulnerable building class is unreinforced ma-
sonry, while wooden structures and reinforced concrete are
the most resilient to earthquake shaking. Therefore, effective
near-term risk reduction measures could target unreinforced-
masonry homes for retrofitting campaigns, particularly in
Ñuñoa, Providencia, and Santiago, while in the mid to long
term a drive towards reinforced-concrete homes would sig-
nificantly reduce the risks to future earthquakes from both
near- and far-field sources. This work also reinforces the need
to identify active faults adjacent to or beneath cities in ac-
tively deforming zones and the need to update the exposure
models as such cities encroach onto these faults. We have
highlighted that local crustal earthquakes in the magnitude
range of 6–7.5 can have a much greater impact than larger
distant earthquakes. Therefore the frequency of major distal
earthquakes has to be balanced by the potential for infrequent
but much more potent, smaller local earthquakes on less ac-
tive faults.
Code and data availability. The latest version of the Open-
Quake Engine can be downloaded from the Global Earth-
quake Model GitHub repository: https://github.com/gem/oq-engine
(GEM, 2019). The point clouds created from both the SPOT
and Pléiades stereo satellite imagery have been uploaded to the
OpenTopography platform (http://opentopography.org, last access:
May 2020) and are available to download for free.
Supplement. The supplement related to this article is available on-
line at: https://doi.org/10.5194/nhess-20-1533-2020-supplement.
Author contributions. EH conducted the research and wrote the
manuscript. JRE devised and led the project and assisted with the
satellite data acquisition, analysis, and manuscript writing. VS and
MVV assisted with the residential seismic risk calculations. DK
provided the non-residential risk calculation results. All authors
contributed to discussions and manuscript preparation.
Disclaimer. Any opinions, findings, conclusions, or recommenda-
tions expressed in this paper are those of the authors and do not
necessarily reflect those of Risk Management Solutions, Inc.
Acknowledgements. This work has been supported by the
NERC/AHRC/ESRC Global Challenges Research Fund
(GCRF) awarded to the Seismic Cities project (grant number
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1552 E. Hussain et al.: Seismic risk in Santiago
NE/P015964/1), by the Royal Society GCRF Challenge grant
(CHG\R1\170038), and in part by the British Geological
Survey (BGS) Overseas Development Assistance programme
(NEE6214NOD). This paper is published with the approval of
the BGS Executive Director. We would like to thank the Seismic
Cities team for helpful discussions, especially Paula Repetto for
help in accessing the Chilean datasets. The lead author would like
to thank Catalina Yepes-Estrada, Anirudh Rao, and Marco Pagani,
as well as all other members of the GEM Hazard and Risk team
for their patience in explaining the details and methodologies of
the OpenQuake Engine. The lead author would also like to thank
the Research Center for Integrated Disaster Risk Management
(CIGIDEN) in Chile for hosting the January 2017 OpenQuake
Engine training workshop. We also thank Tim Craig for suggesting
the need to also consider intra-slab earthquakes as potential major
shaking sources and Laura Gregory for help with interpreting fault
profiles. We thank colleagues at RMS, including Justin Moresco,
Chesley Williams, and Robert Muir-Wood. We gratefully acknowl-
edge the CEOS Seismic Pilot for providing the Pléiades stereo
imagery over the San Ramón Fault (© CNES, 2016; distribution
Airbus DS/Spot Image). Pléiades images made available by CNES
in the framework of the CEOS WG Disasters. JRE is supported by
a Royal Society University Research fellowship (UF150282).
Financial support. This work has been supported by the
NERC/AHRC/ESRC Global Challenges Research Fund (GCRF)
awarded to the Seismic Cities project (grant no. NE/P015964/1), by
the Royal Society GCRF Challenge (grant no. CHG\R1\170038),
and in part by the British Geological Survey (BGS) Official
Development Assistance programme (grant no. NEE6214NOD).
Review statement. This paper was edited by Uwe Ulbrich and re-
viewed by Robin Lacassin and one anonymous referee.
References
Abrahamson, N., Gregor, N., and Addo, K.: BC Hydro ground
motion prediction equations for subduction earthquakes, Earthq.
Spectra, 32, 23–44, 2016.
Akkar, S., Sandıkkaya, M., and Bommer, J.: Empirical ground-
motion models for point-and extended-source crustal earthquake
scenarios in Europe and the Middle East, B. Earthq. Eng., 12,
359–387, 2014.
Allen, T. I. and Wald, D. J.: Topographic slope as a proxy for seis-
mic site-conditions (Vs30) and amplification around the globe,
Tech. rep., United States Geological Survey (USGS), available
at: https://pubs.usgs.gov/of/2007/1357/pdf/OF07-1357_508.pdf
(last access: 1 March 2018), 2007.
Ambraseys, N. and Bilham, R.: Corruption kills, Nature, 469, 153–
155, , 2011.
Armijo, R., Rauld, R., Thiele, R., Vargas, G., Campos, J., Lacassin,
R., and Kausel, E.: The West Andean thrust, the San Ramon
fault, and the seismic hazard for Santiago, Chile, Tectonics, 29,
TC2007, https://doi.org/10.1029/2008TC002427, 2010.
Armijo, R., Lacassin, R., Coudurier-Curveur, A., and Carrizo, D.:
Coupled tectonic evolution of Andean orogeny and global cli-
mate, Earth-Sci. Rev., 143, 1–35, 2015.
Barka, A., Akyuz, H. S., Altunel, E., Sunal, G., Cakir, Z., Dikbas,
A., Yerli, B., Armijo, R., Meyer, B., de Chabalier, J. B., Rock-
well, T., Dolan, J. R., Hartleb, R., Dawson, T., Christofferson,
S., Tucker, A., Fumal, T., Langridge, R., Stenner, H., Lettis, W.,
Bachhuber, J., and Page, W.: The surface rupture and slip distri-
bution of the 17 August 1999 Izmit earthquake (M 7.4), North
Anatolian fault, B. Seismol. Soc. Am., 92, 43–60, 2002.
Beck, S., Barrientos, S., Kausel, E., and Reyes, M.: Source charac-
teristics of historic earthquakes along the central Chile subduc-
tion Askew et Alzone, J. S. Am. Earth Sci., 11, 115–129, 1998.
Bilham, R.: Lessons from the Haiti earthquake, Nature, 463, 878–
879, https://doi.org/10.1038/463878a, 2010.
Bindi, D., Massa, M., Luzi, L., Ameri, G., Pacor, F., Puglia, R., and
Augliera, P.: Pan-European ground-motion prediction equations
for the average horizontal component of PGA, PGV, and 5 %-
damped PSA at spectral periods up to 3.0 s using the RESORCE
dataset, B. Earthq. Eng., 12, 391–430, 2014.
Bommer, J. J., Scherbaum, F., Bungum, H., Cotton, F., Sabetta, F.,
and Abrahamson, N. A.: On the use of logic trees for ground-
motion prediction equations in seismic-hazard analysis, B. Seis-
mol. Soc. Am., 95, 377–389, 2005.
Bommer, J. J., Douglas, J., Scherbaum, F., Cotton, F., Bungum, H.,
and Fäh, D.: On the selection of ground-motion prediction equa-
tions for seismic hazard analysis, Seismol. Res. Lett., 81, 783–
793, 2010.
Boore, D. M., Stewart, J. P., Seyhan, E., and Atkinson, G. M.: NGA-
West2 equations for predicting PGA, PGV, and 5% damped PSA
for shallow crustal earthquakes, Earthq. Spectra, 30, 1057–1085,
2014.
Chatelain, J.-L., Tucker, B., Guillier, B., Kaneko, F., Yepes, H., Fer-
nandez, J., Valverde, J., Hoefer, G., Souris, M., Dupérier, E., Ya-
mada, T., Bustamante, G., and Villacis, C: Earthquake risk man-
agement pilot project in Quito, Ecuador, GeoJournal, 49, 185–
196, 1999.
Chaulagain, H., Rodrigues, H., Silva, V., Spacone, E., and Varum,
H.: Earthquake loss estimation for the Kathmandu Valley, B.
Earthq. Eng., 14, 59–88, 2016.
Civico, R., Pucci, S., Villani, F., Pizzimenti, L., De Martini, P. M.,
Nappi, R., and Group, O. E. W.: Surface ruptures following the
30 October 2016 Mw 6.5 Norcia earthquake, central Italy, J.
Maps, 14, 151–160, 2018.
Comte, D., Eisenberg, A., Lorca, E., Pardo, M., Ponce, L., Saragoni,
R., Singh, S., and Suárez, G.: The 1985 central Chile earthquake:
A repeat of previous great earthquakes in the region?, Science,
233, 449–453, 1986.
Dahlen, F.: Critical taper model of fold-and-thrust belts and accre-
tionary wedges, Annu. Rev. Earth Pl. Sc., 18, 55–99, 1990.
Davis, D., Suppe, J., and Dahlen, F.: Mechanics of fold-and-thrust
belts and accretionary wedges, J. Geophys. Res.-Sol. Ea., 88,
1153–1172, 1983.
de Ballore, F. D. M.: Historia sísmica de los Andes Meridionales al
sur del paralelo XVI, Anales de la Universidad de Chile, 9–63,
1913 (in Spanish).
de la Llera, J. C., Rivera, F., Mitrani-Reiser, J., Jünemann, R., For-
tuño, C., Ríos, M., Hube, M., Santa María, H., and Cienfuegos,
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1553
R.: Data collection after the 2010 Maule earthquake in Chile, B.
Earthq. Eng., 15, 555–588, 2017.
DeMets, C., Gordon, R. G., and Argus, D. F.: Geologically current
plate motions, Geophys. J. Int., 181, 1–80, 2010.
Douglas, J. and Edwards, B.: Recent and future developments in
earthquake ground motion estimation, Earth-Sci. Rev., 160, 203–
219, 2016.
Elliott, J., Nissen, E., England, P., Jackson, J. A., Lamb, S., Li,
Z., Oehlers, M., and Parsons, B.: Slip in the 2010–2011 Canter-
bury earthquakes, New Zealand, J. Geophys. Res..-Sol. Ea., 117,
B03401, https://doi.org/10.1029/2011JB008868, 2012.
Elliott, J., Walters, R., and Wright, T.: The role of space-
based observation in understanding and responding to ac-
tive tectonics and earthquakes, Nat. Commun., 7, 13844,
https://doi.org/10.1038/ncomms13844, 2016.
Elnashai, A. S., Gencturk, B., Kwon, O.-S., Al-Qadi, I. L., Hashash,
Y., Roesler, J. R., Kim, S. J., Jeong, S.-H., Dukes, J., and
Valdivia, A.: The Maule (Chile) earthquake of February 27,
2010: Consequence assessment and case studies, Tech. rep.,
Mid-American Earthquake Center, Department of Civil and
Environmental Engineering, University of Illinois at Urbana-
Champaign, available at: http://hdl.handle.net/2142/18212 (last
access, 1 June 2018), 2010.
England, P. and Jackson, J.: Uncharted seismic risk, Nat. Geosci.,
4, 348–349, https://doi.org/10.1038/ngeo1168, 2011.
Estay, N. P., Yáñez, G., Carretier, S., Lira, E., and Maringue,
J.: Seismic hazard in low slip rate crustal faults, estimat-
ing the characteristic event and the most hazardous zone:
study case San Ramón Fault, in southern Andes, Nat. Hazards
Earth Syst. Sci., 16, 2511–2528, https://doi.org/10.5194/nhess-
16-2511-2016, 2016.
Farías, M., Comte, D., Charrier, R., Martinod, J., David, C., Tassara,
A., Tapia, F., and Fock, A.: Crustal-scale structural architecture
in central Chile based on seismicity and surface geology: Impli-
cations for Andean mountain building, Tectonics, 29, TC3006,
https://doi.org/10.1029/2009TC002480, 2010.
Farr, T. G., Rosen, P. A., Caro, E., Crippen, R., Duren, R.,
Hensley, S., Kobrick, M., Paller, M., Rodriguez, E., Roth,
L., Seal, D., Shaffer, S., Shimada, J., Umland, J., Werner,
M., Oskin, M., Burbank, D., and Alsdorf, D.: The Shut-
tle Radar Topography Mission, Rev. Geophys., 45, RG2004,
https://doi.org/10.1029/2005RG000183, 2007.
Fischer, T., Alvarez, M., De la Llera, J., and Riddell, R.: An inte-
grated model for earthquake risk assessment of buildings, Eng.
Struct., 24, 979–998, 2002.
Frohlich, C.: Deep earthquakes, Cambridge University Press, Cam-
bridge, UK, 2006.
Galetzka, J., Melgar, D., Genrich, J. F., Geng, J., Owen, S., Lindsey,
E. O., Xu, X., Bock, Y., Avouac, J.-P., Adhikari, L. B., Upreti, B.
N., Pratt-Sitaula, B., Bhattarai, T. N., Sitaula, B. P., Moore, A.,
Hudnut, K. W., Szeliga, W., Normandeau, J., Fend, M., Flouzat,
M., Bollinger, L., Shrestha, P., Koirala, B., Gautam, U., Bhat-
terai, M., Gupta, R., Kandel, T., Timsina, C., Sapkota, S. N.,
Rajaure, S., and Maharjan, N.: Slip pulse and resonance of the
Kathmandu basin during the 2015 Gorkha earthquake, Nepal,
Science, 349, 1091–1095, 2015.
GEM: The OpenQuake-engine User Manual, Global Earthquake
Model (GEM) OpenQuake Manual for Engine version 3.3.2,
Tech. rep., Global Earthquake Model (GEM), Pavia, Italy,
GitHub, available at: https://github.com/gem/oq-engine, last ac-
cess: 1 August 2019.
Gill, J. C. and Malamud, B. D.: Reviewing and visualizing the in-
teractions of natural hazards, Rev. Geophys., 52, 680–722, 2014.
Grossi, P. and Kunreuther, H.: Catastrophe modeling: A new ap-
proach to managing risk, vol. 25, Springer Science & Business
Media, 2005.
Hayes, G. P., Wald, D. J., and Johnson, R. L.: Slab1.0:
A three-dimensional model of global subduction zone
geometries, J. Geophys. Res.-Sol. Ea., 117, B01302,
https://doi.org/10.1029/2011JB008524, 2012.
Humire-Guarachi, F. A. H.: Aplicación de métodos geofísicos basa-
dos en ondas superficiales para la caracterización sísmicade
suelos. Aplicación a la microzonificación sísmica del norte y
poniente de Santiago, Tech. rep., Pontifica Universidad Catolica
de Chile, available at: https://tinyurl.com/vqfkauq (last access:
1 January 2019), 2013.
Jayaram, N. and Baker, J. W.: Correlation model for spatially dis-
tributed ground-motion intensities, Earthquake Engineering &
Structural Dynamics, 38, 1687–1708, 2009.
Joyner, W. B.: Strong motion from surface waves in deep sedimen-
tary basins, B. Seismol. Soc. Am., 90, S95–S112, 2000.
Leyton, F., Sepúlveda, S. A., Astroza, M., Rebolledo, S., Acevedo,
P., Ruiz, S., Gonzalez, L., and Foncea, C.: Seismic zonation of
the Santiago basin, Chile, in: 5th International Conference on
Earthquake Geotechnical Engineering, 2011.
Lomnitz, C.: Major earthquakes and tsunamis in Chile during the
period 1535 to 1955, Geol. Rundsch., 59, 938–960, 1970.
Lomnitz, C.: On the epicenter of the great Santiago earthquake of
1647, B. Seismol. Soc. Am., 73, 885–886, 1983.
Lomnitz, C.: Major earthquakes of Chile: A historical survey, 1535–
1960, Seismol. Res. Lett., 75, 368–378, 2004.
Martins, L. and Silva, V.: A Global Database of Vulnerability Mod-
els for Seismic Risk Assessment, in: Proceeding of the 16th Eu-
ropean Conference on Earthquake Engineering, 2018.
Ministerio de Desarrollo Social: Estimaciones de la pobreza
por ingresos y multidimensional en comunas con rep-
resentatividad, Tech. rep., Gobierno de Chile, available
at: http://observatorio.ministeriodesarrollosocial.gob.cl/
casen-multidimensional/casen/docs/estimacion_pobreza_
ingreso_multidimensional_comunal.pdf (last access: 1 Novem-
ber 2018), 2016 (in Spanish).
Montalva, G. A., Bastías, N., and Rodriguez-Marek, A.:
Ground-Motion Prediction Equation for the Chilean Sub-
duction Zone, B. Seismol. Soc. Am., 107, 901–911,
https://doi.org/10.1785/0120160221, 2017.
Muir-Wood, R.: Designing optimal risk mitigation and risk transfer
mechanisms to improve the management of earthquake risk in
Chile, OECD Working Papers on Finance, Insurance and Private
Pensions, p. 1, 2011.
Oncken, O., Hindle, D., Kley, J., Elger, K., Victor, P., and Schem-
mann, K.: Deformation of the central Andean upper plate system:
Facts, fiction, and constraints for plateau models, in: The An-
des, 3–27, Springer, https://doi.org/10.1007/978-3-540-48684-
8_1, 2006.
Park, R. and Hamza, M.: Earthquake Resistant Structures, in: Ref-
erence Module in Materials Science and Materials Engineering,
Elsevier, https://doi.org/10.1016/B978-0-12-803581-8.00856-0,
2016.
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020

1554 E. Hussain et al.: Seismic risk in Santiago
Pasten, C.: Respuesta sísmica de la cuenca de Santiago, Tech. rep.,
Facultad de Ciencias Físicas y Matemáticas. Santiago, Chile,
available at: http://repositorio.uchile.cl/handle/2250/102937 (last
access: 1 April 2020), 2007.
Pérez, A., Ruiz, J., Vargas, G., Rauld, R., Rebolledo, S., and Cam-
pos, J.: Improving seismotectonics and seismic hazard assess-
ment along the San Ramon Fault at the eastern border of Santiago
city, Chile, Nat. Hazards, 71, 243–274, 2014.
Pilz, M., Parolai, S., Stupazzini, M., Paolucci, R., and Zschau, J.:
Modelling basin effects on earthquake ground motion in the San-
tiago de Chile basin by a spectral element code, Geophys. J. Int.,
187, 929–945, 2011.
Pina, F., Ventura, C., and Nunez, T.: Seismic Risk Assessment in
Northern Chile – Lessons from a reconnaissance visit before the
next mega-earthquake in Peru-Chile, in: Proceedings of the 15th
World Conference on Earthquake Engineering, Lisbon, Portugal,
2012.
Poirier, J.-P.: The 1755 Lisbon disaster, the earthquake that shook
Europe, Eur. Rev., 14, 169–180, 2006.
Ramón, A. d.: Santiago de Chile [1541–1991]: Historia de una so-
ciedad urbana, MAPFRE, 1992 (in Spanish).
Rauld, R.: Análisis morfoestructural del frente cordillerano de San-
tiago oriente, entre el Río Mapocho y la Quebrada de Macul (in
Spanish), PhD thesis, Memoria de Título (Inédito), Universidad
de Chile, Departamento de Geología, 2002.
Reynolds, K., Copley, A., and Hussain, E.: Evolution and dynamics
of a fold-thrust belt: The Sulaiman Range of Pakistan, Geophys.
J. Int., 201, 683–710, 2015.
Riesner, M., Lacassin, R., Simoes, M., Armijo, R., Rauld, R., and
Vargas, G.: Kinematics of the active West Andean fold-and-thrust
belt (central Chile): Structure and long-term shortening rate, Tec-
tonics, 36, 287–303, 2017.
Riley, S. J., DeGlioria, S. D., and Elliot, R.: Index that quantifies
topographic heterogeneity, Intermountain Journal of Sciences, 5,
23–27, 1999.
Saita, T.: The great Chilean earthquake of January 24, 1939, B.
Earthq. Res. I. Tokyo, 18, 446–459, 1940.
Santa-María, H., Hube, M. A., Rivera, F., Yepes-Estrada, C., and
Valcárcel, J. A.: Development of national and local exposure
models of residential structures in Chile, Nat. Hazards, 86, 55–
79, 2017.
Santos, S., Lima, S., and Arai, A.: Comparative study of codes for
the seismic design of structures, Revista IBRACON de Estruturas
e Materiais, 5, 812–819, 2012.
Silva, V., Crowley, H., Pagani, M., Monelli, D., and Pinho, R.:
Development of the OpenQuake engine, the Global Earthquake
Model’s open-source software for seismic risk assessment, Nat.
Hazards, 72, 1409–1427, 2014.
Silva, V., Casotto, C., Rao, A., Villar, M., Crowley, H., and Vam-
vatsikos, D.: OpenQuake Risk Modeller’s Toolkit – User Guide,
Tech. rep., Global Eearthquake Model Technical Report 2015–
09, 2015.
Silva, V., Crowley, H., Varum, H., and Pinho, R.: Seismic risk as-
sessment for mainland Portugal, B. Earthq. Eng., 13, 429–457,
2015b.
Stein, S., Geller, R. J., and Liu, M.: Why earthquake hazard maps
often fail and what to do about it, Tectonophysics, 562, 1–25,
2012.
Stewart, J., Douglas, J., Di Alessandro, C., Bozorgnia, Y., Abra-
hamson, N. A., Boore, D. M., Campbell, K., Delavaud, E., Erdik,
M., and Stafford, P. J: Selection of a Global Set of GMPEs for
the GEM-PEER Global GMPEs Project, UCLA: Civil and Envi-
ronmental Engineering, available at: https://escholarship.org/uc/
item/6pn9s2hg (last access: 1 April 2020), 2012.
Stewart, J. P., Douglas, J., Javanbarg, M., Bozorgnia, Y., Abraham-
son, N. A., Boore, D. M., Campbell, K. W., Delavaud, E., Erdik,
M., and Stafford, P. J.: Selection of Ground Motion Prediction
Equations for the Global Earthquake Model, Earthq. Spectra, 31,
19–45, https://doi.org/10.1193/013013EQS017M, 2015.
Talebian, M., Copley, A. C., Fattahi, M., Ghorashi, M., Jackson,
J. A., Nazari, H., Sloan, R., and Walker, R.: Active faulting
within a megacity: The geometry and slip rate of the Pardisan
thrust in central Tehran, Iran, Geophysical Supplements to the
Monthly Notices of the Royal Astronomical Society, 207, 1688–
1699, 2016.
Udías, A., Madariaga, R., Buforn, E., Muñoz, D., and Ros, M.: The
large Chilean historical earthquakes of 1647, 1657, 1730, and
1751 from contemporary documents, B. Seismol. Soc. Am., 102,
1639–1653, 2012.
USGS: USGS earthquake catalogue, available at: https://tinyurl.
com/yxl6nh9d, last access: 1 January 2019.
Vargas, G., Klinger, Y., Rockwell, T., Forman, S., Rebolledo, S.,
Baize, S., Lacassin, R., and Armijo, R.: Probing large intraplate
earthquakes at the west flank of the Andes, Geology, 42, 1083–
1086, 2014.
Vargas, G. E., Contardo, J. I., Tello, S. P., Cáceres, D. E., and Yáñez,
C. J.: Misleading urbanization? The San Ramón fault as a new
seismic risk scenario and the sustainability of Santiago, Chile,
Revista de Urbanismo, 1–20, 2018 (in Spanish).
Vaziri, P., Zoback, M., Tabucchi, T., and Cabrera, C.: Compara-
tive analysis of economic and human casualty seismic risk for
South American Andean capital cities, in: Proceedings of the
15th World Conference on Earthquake Engineering, Lisbon, Por-
tugal, 2012.
Villar-Vega, M. and Silva, V.: Assessment of earthquake damage
considering the characteristics of past events in South America,
Soil Dyn. Earthq. Eng., 99, 86–96, 2017.
Villar-Vega, M., Silva, V., Crowley, H., Yepes, C., Tarque, N.,
Acevedo, A. B., Hube, M. A., Gustavo Coronel, D., and María,
H. S.: Development of a fragility model for the residential
building stock in South America, Earthq. Spectra, 33, 681–604,
https://doi.org/10.1193/010716EQS005M, 2017a.
Wells, D. L. and Coppersmith, K. J.: New empirical relationships
among magnitude, rupture length, rupture width, rupture area,
and surface displacement, B. Seismol. Soc. Am., 84, 974–1002,
1994.
Wilson, M. F., O’Connell, B., Brown, C., Guinan, J. C., and Grehan,
A. J.: Multiscale terrain analysis of multibeam bathymetry data
for habitat mapping on the continental slope, Mar. Geod., 30, 3–
35, 2007.
Yepes-Estrada, C., Silva, V., Rossetto, T., D’Ayala, D., Ioannou, I.,
Meslem, A., and Crowley, H.: The Global Earthquake Model
Physical Vulnerability Database, Earthq. Spectra, 32, 2567–
2585, 2016.
Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020 https://doi.org/10.5194/nhess-20-1533-2020

E. Hussain et al.: Seismic risk in Santiago 1555
Yepes-Estrada, C., Silva, V., Valcárcel, J., Acevedo, A. B., Tarque,
N., Hube, M. A., Coronel, G., and María, H. S.: Modeling the
Residential Building Inventory in South America for Seismic
Risk Assessment, Earthq. Spectra, 33, 299–322, 2017.
Zheng, L., Gordon, R. G., and Kreemer, C.: Absolute plate veloci-
ties from seismic anisotropy: Importance of correlated errors, J.
Geophys. Res.-Sol. Ea., 119, 7336–7352, 2014.
https://doi.org/10.5194/nhess-20-1533-2020 Nat. Hazards Earth Syst. Sci., 20, 1533–1555, 2020