How it works

Climate risk is uneven across space and time and is generally represented as a grid, like the pixels of an image. Over a region the size of a county, there are actually many different values of risk:

• High average temperatures in one part of the county but not others.

• Extreme rainfall on one side of a mountain range but drought potential on the other.

The work of regional processing is to aggregate values derived from high-resolution climate models into a meaningful aggregate risk exposure score across an area of interest, typically as a mean or max statistic of intersecting data. While the process of extracting risk exposure metrics is seemingly simple, it has historically had a very high barrier to entry.

To access and work with climate risk exposure, data has typically required dedicated spatial data expertise, knowledge of the current state of climate research, and effort to test and validate the solution. From the beginning, one of Sust Global’s primary goals has been to lower this barrier, and to make climate data more accessible and actionable.

Example of gridded climate data and unevenness in risk across a geography.

Sust Global’s regional processing is an advancement over previous approaches, better capturing high resolution detail while being less sensitive to outliers. Previous approaches at regional risk processing suffered from several shortcomings. They used point-based processing at the center of a region of interest, which returned risk scores sensitive to the exact point selected, and sensitive to outliers in the area of interest. Regional processing solves this issue, giving a much more accurate description of risk across an entire region.

Regional processing gives a more accurate picture of risk across a user defined region (zip code, county boundary, municipality, etc), compared to centroid based processing on that region. Differences can be substantial, potentially several times the annual likelihood of an event occurring. Shown here, the difference in risk between Regional Processing (y-axis) and centroid-based processing (x-axis), for counties of the lower 48 states.

At its core regional processing is a straightforward capability but with huge implications for the accuracy of metrics served. Regional processing – in conjunction with near real-time observation ingestion with artificial intelligence driven resolution enhancement, served using a scalable cloud-native data platform – allows users to investigate risk exposure across a spatial region more effectively.

Beyond direct regional analysis, users can now perform analyses such as putting an asset point’s risk exposure in context of the surrounding region. Such analyses can show how a particular asset’s location compares to the rest of the zipcode, county or state average, and allow for corresponding decisions to minimize risk exposure.

In addition, regional processing is an important precursor to a deeper understanding of climate risk and continually improving on the picture of risk we serve. Embedding regional understanding of climate into our workflows allows us to build better tools to understand changing risk and resilience around the world, such as population exposure to chronic hazards, the potential for carbon sequestration as part of a regenerative agriculture project, and optimal site selection for wildfire risk reduction projects.

Regional processing of administrative units to understand water stress across Australia. A score of 1.0 indicates regions that consume available water faster than it is recharged by hydrological processes.

Understanding the future likelihood of a cyclone hit at counties along the US Eastern Seaboard.

Regional processing also opens up the potential to analyze an area in greater detail. The accuracy gains from using regional processing allows for understanding risks at a gridded level, across an entire region. The wildfire maps below show one example of this, clearly denoting areas of high (and low) wildfire risk, based on the modeled unified wildfire likelihood.

A view into increasing wildfire risk in the US Pacific Northwest, powered by regional processing. High-resolution future wildfire data over Oregon and Washington, powered by Super Resolution AI Downscaling (patented). Left: Annual Wildfire Risk, 1980-2010. Right: Annual Wildfire Risk, 2022-2052

Conclusion

Regional processing provides a powerful way for users to understand risk across an entire region more accurately, opening up opportunities for more specific location based analytics and meaningful conversations about risk mitigation measures.

At Sust Global, we are building the software of climate adaptation. We are actively serving teams working on climate risk assessments, reporting and risk management with effective tools to integrate climate analytics into their workflows. If you are looking into site-specific climate risk assessments, site selection for climate aware investments or climate informed financial instruments, we would love to hear more from you.