Methodology#

The Global Risk Model methodology integrates hazard, exposure, vulnerability, and risk assessment approaches to provide comprehensive disaster risk evaluation across multiple hazard types.

Hazard Methodology#

The hazard component characterizes the frequency and intensity of natural hazards across different geographic regions. Our approach integrates multiple hazard models to provide comprehensive coverage.

Key Features#

  • Multi-hazard assessment framework

  • Probabilistic hazard modeling

  • Regional hazard characterization

  • Standardized intensity measures

#

Exposure Methodology#

What is an exposure model? An exposure model is essentially an inventory of all the buildings, people, and economic assets that could be affected by natural hazards. Think of it as a comprehensive database that tells us “what’s at risk” before a disaster happens.

The Global Exposure Model provides a comprehensive framework for assessing the distribution and characteristics of built infrastructure that could be affected by natural hazards worldwide. This model serves as a foundational component of multi-hazard risk assessment, offering standardized methodologies for quantifying and characterizing exposure elements across different geographic scales and disaster scenarios including earthquakes, floods, hurricanes, extreme weather events, and pandemics.

Key Features

The model captures essential information for each building type including:

  • Building counts: How many buildings of each type exist in each area

  • Occupancy data: Number of people who live or work in these buildings

  • Physical characteristics: Construction materials, structural systems, and building height

  • Economic values: Replacement costs and built-up areas

  • Vulnerability attributes: Factors that influence how buildings respond to earthquake shaking and other natural hazards

Geographic Coverage and Regional Organization

The global exposure model covers 215 countries, encompassing all continental territories reported by the United Nations. The model excludes only small territories (primarily islands in the Pacific and Indian Oceans) located in regions of low seismic hazard with populations below 1,600 inhabitants.

Data Quality and Resolution

While the methodology is consistent across regions, the models vary in terms of:

  • Geographic resolution: From individual buildings to administrative divisions

  • Data quality: Depending on availability of national statistics and local expertise

  • Temporal coverage: Based on census vintages and data collection periods

  • Building stock completeness: Residential, commercial, and industrial coverage varies by country

Regional studies (listed in the References section) provide detailed methodologies for each area, with improvements including incorporation of land use data, seismic code information, construction cost analyses, and extensive review by local experts.

Methodology#

Unlike previous global models that relied on broad assumptions applied uniformly worldwide, this model adopts a bottom-up approach wherever possible. This means we gather detailed local data country-by-country rather than making top-down estimates. This approach allows us to:

  • Identify the most common construction types in each region

  • Locate areas with high concentrations of vulnerable buildings (such as unreinforced masonry or informal construction)

  • Understand regional variations in construction practices and building quality

  • Provide more accurate risk assessments for local decision-making across multiple hazards

The methodology presented here builds upon the comprehensive framework described in yepes2023, which details the complete development process of this global building exposure model.

The Challenge of a Global Building Inventory#

Creating a comprehensive global building inventory for earthquakes and other natural hazard risk assessment presents unique challenges. Ideally, we would have detailed information for every building including:

  • Physical characteristics: Age, height, materials, structural system, irregularities

  • Design standards: Building code compliance and construction quality

  • Occupancy details: Building use and number of occupants

  • Economic data: Replacement costs and built-up areas

However, most countries lack centralized databases with this level of detail, necessitating innovative approaches to combine multiple data sources.

Bottom-Up vs. Top-Down Approaches#

Traditional Top-Down Approach:

  • Uses global datasets (population, GDP, socio-economic indicators)

  • Applies uniform assumptions and mapping schemes nationally

  • Advantages: Consistent methodology, globally applicable

  • Limitations: High uncertainty, neglects regional construction variations, less suitable for local analyses

Our Bottom-Up Approach:

  • Collects national statistics country-by-country at the finest administrative level

  • Uses housing censuses, commercial surveys, and local datasets

  • Incorporates local construction practices and regional variations

  • Provides more accurate representation of actual building stock

  • Enables better sub-national risk assessments

This approach allows for more rigorous modeling of asset numbers, local construction practices, and realistic construction costs, as demonstrated in regional applications (e.g., Argentina - yepes2017).

Modelling Framework#

An ideal building inventory for natural hazard risk assessment should include information at the building level regarding the physical characteristics (e.g., age of construction, height, built-up area, predominant materials, lateral load resisting system, irregularities, design code provisions), occupancy type, number of occupants and replacement cost. However, the majority of the countries do not have centralized databases with such level of detail. Therefore, it is necessary to develop exposure models by collecting and combining multiple sources of data.

This section presents the general modeling framework undertaken for the development of the global building inventory for multi-hazard risk assessment. The studies indicated #in regional references describe in more detail the methodologies for the respective study area, which can vary depending on the data and resources available. Previous global exposure modeling initiatives have favored a top-down methodology. In this approach, global datasets (e.g., population counts, national GDP, or other socio-economic data) are used to infer building counts, and the definition of building classes is obtained through national mapping-schemes. This approach has the advantage of using a uniform methodology, and it can be globally adjusted based on the available data. On the other hand, this approach is affected by large uncertainties and bias due to the employment of the same factors and mapping schemes nationally, neglecting the potential changes in regional construction practices. Such limitations may render these models less suitable for sub-national risk analyses. For the global model presented herein, we aimed at employing a bottom-up approach whenever the necessary data was available. In this modelling option, national statistics (e.g., housing census, surveys of commercial and industrial facilities) are collected (country-by-country) are used at the smallest available administrative level. These datasets typically include information about the type of buildings, main construction materials, number of storeys, and epoch of construction, allowing a more rigorous modeling of the number of assets, local construction practices, and construction costs (see the example for Argentina described by yepes2017). We note that despite the national coverage of the datasets, the data is usually collected at the local level (e.g., building-by-building), and then aggregated at a given administrative division.

Building Classification System#

Understanding Building Types#

Why classify buildings? Different building types respond differently to earthquake or other natural hazards. A modern steel-frame office building will behave very differently from an old adobe house or a reinforced concrete apartment building when subjected to earthquake shaking, hurricane winds, or flood waters. By grouping buildings with similar hazard response characteristics, we can better predict potential damage and losses across multiple hazard types.

To identify the most predominant construction typologies in each country and group buildings with similar characteristics, comprehensive building classes were developed for each region. An extensive review of local, national, and regional technical reports and scientific literature was conducted to analyze construction practices, which are frequently associated with cultural heritage, meteorological conditions, and #local availability of construction materials. #### Data Sources for Building Classification

Examples of detailed reports include:

  • National inventories: Catalogue of Building Typologies in India (NDMA, 2013)

  • Regional systems: HAZUS building classification for the United States (fema2017), European building classification (NERA project, Crowley et al., 2012)

  • Local studies: Dictionary of building typologies in Manila, Philippines (Bautista et al., 2014)

  • Global databases: World Housing Encyclopedia (WHE) with over 150 detailed reports on construction techniques, UN-HABITAT and PAGER building inventory databases (jaiswal2010)

  • Post-earthquake surveys: Damage assessments providing real-world vulnerability evidence

  • Virtual reconnaissance: Google Street View, Google Earth, Mapillary, and Tencent Maps/Baidu Total View surveys

  • Expert collaboration: Local contributors reviewing and improving building classes

A complete list of contributors to the global exposure model can be found at https://globalquakemodel.org/risk-model-contributors.

The GEM Building Taxonomy#

Building classes are defined using the GEM building taxonomy v3.3. The GEM taxonomy uses key structural attributes that are relevant for multiple hazard types:

  • Predominant construction material: steel (S), wood (W), reinforced concrete (RC), unreinforced masonry (MUR), reinforced masonry (MR), confined masonry (MCF), adobe (ADO), rammed earth (ER), bamboo (WBB), and mixed-types (MIX). Additional details include masonry unit types (stone, clay bricks, concrete blocks) and concrete technology (cast-in-place or precast).

  • Lateral load resisting system: moment frames (bare - LFM, infilled - LFINF, and braced - LFBR), walls (LWAL), dual wall-framed systems (LDUAL), and post and beam (LPB).

  • Number of stories: depending on the information and needs, specific heights or a range of the number of stories was indicated (e.g., H:1, H:1-3).

  • Ductility level: four levels of ductility were defined (DNO - nonductile, DUL - low ductility, DUM - moderate ductility, and DUH - high ductility). When available, the code level was also included in the details (CDN - no code, CDL - low code, CDM - moderate code, and CDH - high code).

  • Occupancy class: residential (RES), commercial (COM), and industrial (IND). For some countries, the type of activity was also included in commercial buildings.

Multi-Hazard Applications#

The global building inventory includes over 1,000 building classes, reflecting the tremendous diversity in construction practices worldwide. Construction materials serve as primary indicators of building vulnerability and enable rapid identification of vulnerable construction concentrations in hazard-prone areas.

While originally developed with earthquake risk assessment as the primary focus, the classification system captures information relevant for multiple natural hazards:

  • Earthquakes: Structural system, materials, and ductility levels determine seismic performance

  • Hurricanes/Cyclones: Building height, structural system, and construction quality affect wind resistance

  • Floods: Building elevation, materials (water resistance), and structural integrity influence flood vulnerability

  • Volcanic hazards: Roof characteristics and structural systems affect ash load capacity

  • Pandemics: Occupancy densities and building ventilation characteristics (captured in occupancy classifications) influence disease transmission risk

  • Extreme temperatures: Construction materials and building envelope characteristics affect thermal performance

Examples of multi-hazard applications include the Middle East study (dabbeek2020) which addresses earthquakes, strong winds, and floods simultaneously.

Data Sources and Spatial Distribution#

Census Data as Foundation#

The primary data sources for the global model are national censuses, which provide the most comprehensive information about living conditions, building characteristics, and economic activities within countries. These include:

Population and Housing Censuses:

  • Conducted every 10 years through full field enumeration

  • Cover all persons and dwellings in urban and rural areas

  • Include building materials, construction types, and occupancy information

  • Follow international standards enabling data harmonization

Economic and Establishment Censuses:

  • Conducted every 5 years, often using sample surveys

  • Focus on commercial and industrial buildings

  • Provide information on business activities and building characteristics

  • Examples include surveys in Bangladesh, Ethiopia, and Mexico

While census questionnaires are not specifically designed for natural hazard risk assessment, many attributes can be effectively used to infer building structural characteristics and vulnerability levels across different hazard types.

Building Class Distribution and Mapping Schemes#

Mapping schemes that establish the relationship between the attributes provided in the different census databases and the identified set of building classes are available for each country in the country-specific sections. The development of the mapping schemes follows the methodology described in yepes2017. For each country, three different schemes were derived:

  1. urban areas,

  2. rural areas, and

  3. cities with large and dense urban agglomerations (e.g., capital cities).

For countries with large populations (e.g., India, China, or Mexico) additional mapping schemes were proposed for the major cities based on the level of urbanization (see for example rao2020 for India). Mapping schemes are a major source of epistemic uncertainty, and they depend on the available information in the original data, the expertise of the modeler, and the judgement of the local experts.

The ductility level of the buildings was inferred through the year of construction, evaluation of the ratio of informal construction, and the evolution of building codes and standards in the respective country. When the year of construction was not available in the housing variables, past censuses were used to infer the ratio of new construction, as well as national statistics. Regarding informal construction rates, few countries specifically reported the number and type of slums (e.g., India, Bangladesh). However, informal construction is not limited to low-income houses, but instead, it refers to structures built without engineering design and construction control, usually present in developing countries. Fractions of informal construction were estimated based on the construction permit statistics, households with unsatisfied basic needs, and other reports at the national or regional scale (e.g., Fernandes, 2011; Guevara and Arce, 2016). Regarding the evolution of building codes, Crowley et al. (2021) present a detailed example of the analysis carried out in Europe. This study shows the seismic design coefficients for each region, as well as the evolution of the design regulations in different countries. Similar information was considered for the assignment of the ductility and code level in the Middle East, Latin America and the Caribbean, Africa, South and Southeast Asia.

Average Built-Up Areas and Replacement Cost#

The replacement cost was estimated based on the expected reconstruction strategy of each country. For most countries, it was assumed that damaged or collapsed buildings would be repaired or reconstructed following the ‘‘building back better’’ concept, and not according to the previous conditions. For example, if an adobe house collapses, the replacement cost will be the value corresponding to the minimum affordable housing cost established in the country (e.g., a confined or reinforced masonry structure). On the other hand, it is unrealistic to assume that such strategy would be rigorously followed is less developed countries, as indicated by the local experts (e.g., Afghanistan, India, Malawi, Myanmar, Mongolia, Tanzania). The replacement cost of a building includes the structural components (i.e., foundations, frames, floors, walls, and staircases), non-structural components (i.e., partition walls, facades, veneers and finishings, plumbing, electric wiring, HVAC, lifts, lighting, ceilings, and heating), and its contents (i.e., shelves, machinery, office and communication equipment, manufacturing equipment). It was assumed that the replacement cost of each one of these components depends on the main material of construction, occupancy class, and region.

The average built-up area and the replacement cost depend on the building class, the type of occupancy, and the settlement type (urban or rural). Reference values are commonly reported per square meter, and in some cases, national statistical offices include average dwelling areas and construction costs considering the structural and nonstructural components (e.g., Australia, Indonesia, Japan, Thailand, United States and most of Europe). When national information was not available, regional and global valuation surveys.

To homogenize and compare values between countries, the replacement cost reported in the model is in US dollars and has been adjusted to 2021 values. The cost of the land is not included, even though it could represent a significant fraction of the real estate value.

Regional Studies and References#

The following regional studies contributed to the global exposure mosaic:

  • Africa: paul2022

  • Australia: dunford2014

  • Canada: journeay2022

  • Central America: calderon2022

  • Central Asia: pittore2020

  • China: ma2021

  • Costa Rica: calderon2019

  • Europe: crowley2020

  • GED4GEM: gamba2012

  • India: rao2020

  • Iran: motamed2019

  • Middle East: dabbeek2020

  • New Zealand: abbott2020

  • Pacific Island Countries: pcrafi

  • South America: yepes2017

  • Turkey: rao2021

  • United States: fema2017

#

Vulnerability Methodology#

The vulnerability component defines the susceptibility of exposed elements to damage from hazard events. This includes fragility functions, damage-to-loss relationships, and consequence models.

Key Components#

  • Fragility function development

  • Damage state definitions

  • Loss estimation models

  • Building taxonomy integration

#

Risk Methodology#

Risk assessment combines hazard, exposure, and vulnerability components to quantify potential impacts including:

Loss Types#

  • Economic losses - Direct and indirect financial impacts

  • Human casualties - Fatalities and injuries

  • Building damage - Structural and non-structural damage

  • Population impacts - Displaced and affected populations

Assessment Scales#

  • Global - Worldwide risk patterns

  • Regional - Continental and sub-continental analysis

  • National - Country-level assessments

  • Subnational - Provincial and local analysis