How 1.6 billion buildings power ICE’s global climate risk analytics

How 1.6 billion buildings power ICE's global climate risk analytics

Authors: ICE Climate Research and Content

• ICE Climate's exposure datasets enable climate risk assessments at the building footprint level globally
• These risks can be aggregated to municipalities, countries, corporations, pools of mortgages, and real estate portfolios to understand exposure across asset classes

Modeled flood depths for a 1-in-100-year rain-driven flood event in 2020

Figure 1.A residential neighborhood in Nevada with building footprints shaded by rain-related flood risk (modeled flood depths during a 1-in-100-year rainfall event) in 2020. Source: ICE as of 11/09/2024.

When it comes to flood risk, location matters. Take a map of flood risk in a residential neighborhood near Reno, Nevada (Figure 1). Some areas of this neighborhood are projected to experience more than 15 cm of flooding during a 1-in-100-year rainfall event, while other areas just down the block have negligible risk.

On the other side of the United States, in the port city of Norfolk, Virginia, many neighborhoods face similar risks due to coastal flooding (Figure 2). Still farther away, rain-related flood risks are significant for many residential and commercial buildings in Hanover, Germany (Figure 3), and coastal flood risks are significant in the greater Bangkok area (Figure 4). By 2050, the current location, shape, size or orientation of a particular home, warehouse, mall, or museum could mean the difference between a usable building and an unusable one.

2020 - Modeled flood depths for a 1-in-100-year coastal flood event

2050 - Modeled flood depths for a 1-in-100-year coastal event (SSP5-8.5)

Figure 2.A neighborhood in Norfolk, Virginia, with building footprints shaded by coastal flood risk (modeled flood depths during a 1-in-100-year coastal flood event) in 2020 and under a Shared Socioeconomic Pathway 5-8.5 climate scenario1in 2050. Source: ICE as of 11/09/2024.

Projected rainfall-driven flood depths for a 1-in-100-year event in 2050 (SSP5-8.5)

Figure 3.A view of a part of the greater Hanover, Germany area, with building footprints shaded by projected rain-related flood risk (modeled flood depths during a 1-in-100-year flood event) under Shared Socioeconomic Pathway 5-8.5 in 2050. Source: ICE as of 11/09/2024. Contains information from GlobalMLBuildingFootprints,2which is made available here under the Open Database License (ODbL).

It can be challenging to understand and map these sorts of climate risks. On top of the uncertainties inherent in climate models themselves,3many exposure models approximate buildings as points, when in fact large structures like distribution centers, convention centers, stadiums, airports, and malls often have spatial footprints measured in thousands of square meters. Given the difference that 100 meters can make for perils like flooding - inundation versus no flooding at all - approximating buildings as point locations can make it difficult to assess which structures might become impaired under different climate risk scenarios (e.g., Figures 2 and 3).

Projected coastal flood depths for a 1-in-100-year event in 2050 (SSP5-8.5)

Figure 4.A view of an area along the Chao Phraya River south of central Bangkok, with building footprints shaded by projected coastal flood risk (modeled flood depths during a 1-in-100-year coastal flood event) under Shared Socioeconomic Pathway 5-8.5 in 2050. Source: ICE as of 11/09/2024. Contains information from GlobalMLBuildingFootprints,4which is made available here under the Open Database License (ODbL).

To meet this challenge, ICE Climate is constructing next-generation global exposure datasets that incorporate information derived from building footprints. The new global exposure layers5include data from several propriety and open data sources; in total, these datasets incorporate about 1.6 billion building footprints worldwide. Even though individual building-level risk estimates have their limitations, this level of granularity is powerful: it allows ICE Climate to aggregate and assess risks consistently, anywhere in the world, whether the risks are associated with global corporations and their assets, the homes in pools of mortgages and real estate portfolios, or the buildings located within municipalities and sovereign nations.

After integrating building footprint data from various sources into these global exposure layers, there are areas of the world with missing building footprint and rooftop coverage. These areas include China, central Africa, North and South Korea, Taiwan, New Zealand, parts of Spain, and several countries in the former Soviet Union (Figure 5). In these areas, ICE Climate uses information from satellite-derived human settlement data (the Global Human Settlement Layer, or GHSL) produced by the European Commission in 2018.6

The GHSL is a dataset of 6.49 trillion pixels (10-meter by 10-meter global resolution) that indicates where human structures exist. In this case, ICE Climate groups these pixels into 40 square meter "structure clusters." ICE Climate then uses the structure clusters in areas that lack coverage in other datasets. At the country level, about 80% of countries and territories have greater than 50% building footprint data coverage - with remaining areas filled in with structure clusters (Figure 5).

Figure 5.Countries shaded by rooftop and building footprint coverage in the ICE Climate global exposure layers. In areas with low coverage, ICE Climate uses information from the Global Human Settlement Layer. Source: ICE as of 11/09/2024.

These kinds of unified maps of global built structures enable ICE Climate to assess climate risks at the individual tax-parcel level within the United States and any given area of land globally.

The reasoning for incorporating the ability to interrogate climate risks for any given area globally is simple: where structures exist and are at risk today is key information. However, where structures may not be able to exist tomorrow due to the developable land carrying too much risk is just as critical.

In the coming years, these kinds of climate-related risks will affect individuals, communities, and countries across the world, as well as the international financial markets that tie us all together. Our core mission at ICE Climate is to provide data and insights to help build resilience at every level. The building footprint and exposure datasets discussed in this article are a foundational component of this effort, enabling us to map the exposure of countries, corporations, and communities around the world to projected wildfire, inland and coastal flooding, and hurricane risks at the asset level.

Upcoming articles will explain how these exposure datasets are combined with ICE Climate's global hazard projections to estimate expected property and economic losses across the world - and how these loss estimates translate into material considerations for investors, corporations, and local and sovereign governments.

1O'Neill et al. (2015) The roads ahead: Narratives for shared socioeconomic pathways describing world futures in the 21st century. Global Environmental Change, JGEC-1388. Available at: https://korbel.du.edu/sites/default/files/2021-12/The%20Roads%20Ahead.pdf

2Global ML Building Footprints, Microsoft. Available at: https://github.com/microsoft/GlobalMLBuildingFootprints/

3Roston, E, Karra, E, Kaufman, K & A Rangarajan (9 Aug 2024). The Risky Business of Predicting Where Climate Disaster Will Hit. Bloomberg. Available at: https://www.bloomberg.com/graphics/2024-flood-fire-climate-risk-analytics/

4Microsoft's GlobalMLBuildingFootprints.

5These layers are trivial transformations that contains only extractions and no Open Street Map data.

6Pesaresi, Martino; Politis, Panagiotis (2022): GHS-BUILT-S R2022A - GHS built-up surface grid, derived from Sentinel-2 composite and Landsat, multitemporal (1975-2030). European Commission, Joint Research Centre (JRC) [Dataset] DOI:10.2905/D07D81B4-7680-4D28-B896-583745C27085 PID: http://data.europa.eu/89h/d07d81b4-7680-4d28-b896-583745c27085