The Earthquake Risk Model for the Pacific Islands has compiled by the GEM Foundation using publicly accessible sources of information. Even though the exposure model was developed for 19 countries, given the disparity in the hazard levels that can be found in the region, the risk metrics were calculated exclusively for the 9 countries that present non-negligible levels of seismic hazard.
1. Exposure Modelling
In the framework of the Global Seismic Risk Model, an open exposure model describing the residential, commercial and industrial building stock for 19 countries has been developed, using only publicly available sources of information. The 19 countries were divided in three groups. For the first group the information used to build the exposure models was retrieved from the Pacific Catastrophe Risk Assessment and Financing Initiative (PCRAFI), a joint initiative from different organizations to provide disaster risk modelling tools for the Pacific Islands. For the second group of countries, which includes the three unincorporated United States territories located in the Pacific, the information was collected from HAZUS. For the remaining countries, the primary source of information was the latest national and sub-national population and housing census databases of each country. The table below presents the population and number of residential, commercial and industrial buildings for the 19 countries, as well the primary source of information.
Cook Islands (COK)
Marshall Islands (MHL)
Solomon Islands (SLB)
Northern Mariana Islands (MNP)
American Samoa (ASM)
New Caledonia (NCL)
Papua New Guinea (PNG)
Timor Leste (TLS)
Pacific Islands Exposure Map
Although 15 countries are included in the PCRAFI initiative, three of them (Fiji, Papua New Guinea, and Timor-Leste) had insufficient information and therefore the data could not be considered. PCRAFI data describes the occupancy type, the number and type of buildings, the number of stories, the floor area and the economic value of each asset. A building class was then assigned to each occupancy and construction type, according to different publications that describe the building practices of the region. After analyzing the housing census data for most of the countries, it was noticed that the number of households reported was systematically lower than the total number of residential buildings described in PCRAFI. The difference in these two metrics can be related to the definition of household: while in some countries it is defined as “a group of persons who sleep in the same housing unit and have a common arrangement in the preparation and consumption of food”, in other cases it is stated that “the definition is based on eating together rather than on living or sleeping in the same building or dwelling”, indicating that a household may include more than one building. Moreover, the PCRAFI methodology uses satellite imagery to obtain the building footprints, which can lead to an overestimations of the number of buildings, since this automatic process may, in certain cases, identify other surfaces as buildings. In such cases, the total number of households in each administrative region was distributed according to the number of buildings in the same area and, consequently, the number of dwellings per building is below one.
Since the PCRAFI data was collected in 2010 and 2011, the economic values had to be updated, considering the inflation rates of each country. The structural, non-structural, and contents values were estimated as a fraction of the total value provided by PCRAFI, using different ratios according to the general occupancy of each asset (residential, commercial and industrial).
For the unincorporated United States territories located in the Pacific (American Samoa, Guam and Northern Mariana Islands), the HAZUS building stock inventory was used. This methodology uses a dasymetric mapping to distribute buildings inside each census block. Then, a general and specific occupancy is assigned to each location, based on land use maps and on more specific information, when available. To assign building classes to each building, the Earthquake Model State Occupancy to Building Type Mapping Scheme for Hawaii is used, since these do not exist for the Pacific territories. Replacement costs are estimated based on RSMeans values, which are then adjusted considering local costs of materials and services.
For the remaining countries in which the census databases were used (Fiji, New Caledonia, Papua New Guinea and Timor Leste), the development of the exposure model followed these main steps:
1 - Mapping census data to building classes:
The housing census databases typically provide information about the number of households and specific attributes (e.g. exterior walls material, roof material, or number of rooms) but do not describe the number of households in specific building classes. For example, the census data for Timor-Leste provide different tables that detail the number of households by construction material of the external walls. However, it is not possible to know the percentage of concrete/brick buildings that have corrugated iron/zinc or concrete roofs. Therefore, it was necessary to establish a relationship between the attributes used in the census data and a list of building classes, which is herein called a mapping scheme. The number of buildings per building class was calculated by multiplying the quantity defined in the census by the associated building fraction, at each geographical scale. Since the building classes are expected to have distinct distributions depending on the settlement type (for example metropolitan, urban and rural areas), when the geographical regions of the country were categorized according to these classes, different mapping schemes were developed.
2 - Converting the number of dwellings/households to number of buildings:
In most of the countries analyzed, the census data reports the total number of households by geographical region. Whilst this information is useful to estimate the total area or replacement cost of a given building class, it does not give information about the number of buildings in each location. Given the construction practices of these countries, the vast majority of the population live in detached buildings. Therefore, in rural areas the number of households was considered equivalent to the number of buildings, as well as for the single storey buildings in urban areas. In urban or metropolitan areas, the number of buildings with 2 or more floors was obtained by dividing the total number of households by the average number of households per building, for each building class.
3 - Estimation of areas, replacement costs and occupants per building class:
The final step to complete the exposure model is the estimation of the replacement cost. In this context, the replacement cost includes the cost of structural and non-structural components and contents, but not the cost of the land. Since construction costs are commonly found per square meter of dwelling, the average floor area per dwelling was used. Given the difficulty in finding reliable estimation for the contents cost, this value was estimated as a percentage of the structural value of the building. Depending on the information available for each country, the cost was differentiated across metropolitan, urban, and rural areas. Finally, the population in each region was distributed across the different building classes proportionally to the number of dwellings.
Industrial and commercial exposure model
For the development of the industrial and commercial exposure models the data available in the census was also the primary source of information used. In most cases, the number of workers in each region or administrative division is available, disaggregated by type of industry or sector of work. An average number of workers per business was defined to estimate the number of buildings in each sector, based on the available literature or reports for each country. The total number of buildings was also adjusted to match the ratio of population per building (industrial and commercial) verified in the neighboring countries with similar economic characteristics. Subsequently, mapping schemes were defined for each sector to define the number of buildings per building class. Based on the building class and on the average number of workers in each sector, the average building area and cost per square meter were defined for each building class.
The figures below present the distribution of buildings and the aggregated replacement cost for the 9 countries with significant seismic hazard (American Samoa, Tonga, Vanuatu, Solomon Islands, Timor-Leste, Samoa, Fiji, New Caledonia and Papua New Guinea). The exposure (in terms of the number of buildings) is also presented on an evenly spaced grid in the interactive map below.
The vulnerability component characterizes the likelihood to suffer damage or loss given a hazard intensity. The relation between probability of loss and hazard intensity is expressed by a vulnerability function, whilst the relation between probability of damage and hazard intensity is represented by fragility functions. For the Pacific Islands region, the general set of global vulnerability functions from GEM were employed. The methodology followed for the development of these functions relied on the following steps:
Identification of the most common building classes in the region, using peer-reviewed literature, web surveys (https://platform.openquake.org/building-class/), and World Housing Encyclopedia reports.
Development of simplified numerical models for each building class, using data from the literature and results from experimental campaigns (e.g. yield and ultimate global drift, elastic and yield period of the first mode of vibration, participation factor of the first mode of vibration, common failure mechanisms). Some of the building classes had to be explicitly modelled using complex 3D models due to the lack of information in the literature.
Selection of ground motion records using local strong motion databases, and considering the local seismicity and tectonic environment. To this end, seismic hazard disaggregation at the location of the most urbanized centers supported the identification of the combinations of magnitude and distance, which contribute the most to the seismic hazard. The use of a large set of actual time histories aims at propagating the record-to-record variability to the vulnerability assessment.
Performing nonlinear time history analysis to evaluate the structural response (i.e. engineering demand parameter (EDP) – maximum displacement and acceleration) of the simplified numerical model against the selected ground motion records. This step uses the open-source package for structural analysis OpenSees, and the Risk Modelers Toolkit supported by GEM.
Evaluation of the structural responses of the numerical models in order to evaluate the evolution of damage with increasing hazard intensities. In this process, the probability of exceeding a number of damage states for a set of intensity measure levels is defined (i.e. fragility functions).
The fragility functions can be converted into vulnerability functions (i.e. probability of loss ratio conditional on ground shaking) using a damage-to-loss model. Such functions can be used directly in the assessment of economic and human losses due to earthquakes.
A similar approach has been followed for the development of fragility and vulnerability functions within the scope of regional programs in Central America and the Caribbean, South America, Sub-Saharan Africa and South-East Asia. This framework aims at being a dynamic tool which can be improved upon the release of new models and datasets. As an example, fragility models for four building classes in the Pacific Islands are illustrated below.
3. Seismic Hazard
The main components concerning the probabilistic seismic hazard model for the region can be found in the associated technical documentation. The seismic hazard in terms of the peak ground acceleration (PGA) for a probability of exceedance of 10% in 50 years (equivalent to approximately 475 years return period) is presented in the interactive map below.
Pacific Islands Hazard Map
4. Seismic Risk Results
As previously explained, the risk metrics were calculated for the 9 countries in the region that have non-negligible levels of seismic hazard. The risk results suggest that the countries with greatest earthquake risk in the Pacific are Vanuatu, Tonga, Solomon Islands and Papua New Guinea. The region is part of the circum-Pacific belt (also called “Ring of Fire”), one of the most seismically active regions of the world. Throughout the history, high magnitude earthquakes have caused human and economic losses in these countries, due to the convergence between the Pacific and Indo-Australian plate.
In addition to being in one of the most seismically active regions in the world, Papua New Guinea shows also a high concentration of population which increases the risk of economic and human losses. In fact, according to the Global Seismic Risk Model estimations, this country presents not only a high loss ratio but also high absolute average annual losses. The capital stock of the other countries in the region with high seismic hazard (Tonga, Vanuatu and Solomon Islands) are relatively low, leading to low absolute average annual losses, but high loss ratios.
More detailed risk results by country are presented in specific country risk profiles [link]. In each profile it is possible to find seismic hazard, exposure and risk maps, sets of exceedance probability curves and the list of the top regions at risk at a subnational level. The seismic risk (in terms of normalized average economic losses) is also presented in the interactive map below.
Pacific Islands Risk Map
5. Partners and Contributors
The Pacific Islands seismic risk model extensively relies on the enthusiasm and commitment of various organizations that openly collaborated with GEM and its partners. The creation of this model would not have been possible without the support provided by many experts. A list of the individuals that contributed to the development of the Pacific Islands seismic risk model is provided below.