Refining Property Risk Assessment Through Geospatial Precision

By Dan Pribe

Actuary of the Future, July 2025

Geospatial imagery and high-resolution image technology are transforming industries, including property insurance. The Society of Actuaries Research Institute recently published a research paper demonstrating the use of geospatial data and high-resolution imagery alongside artificial intelligence in the P&C industry.

The paper, titled “Advanced Analytics in Insurance: Utilizing Building Footprints Derived from Machine Learning and High-Resolution Imagery” and authored by Greg Dietzen, Garrett Bradford, Kailey Adams and Claire Palmer of Milliman, studies how leveraging precise building footprint data — along with structure characteristics and other property features — can improve risk assessment and underwriting accuracy for insurers.

Background

Accurate location data attached to property information is necessary to assess location-based risk. This data can take several forms such as coordinate pairs (i.e., latitude and longitude), delivery points (e.g., mailbox) or the centroid of a parcel that a building sits on.

Geocoding is the process of linking location data to address information. Geocoding, however, may not provide the most accurate information depending on the underlying quality of its data.

In contrast to geocoding, building footprints can provide information on size, shape and location of structures using field surveys, mapping, aerial photography, and/or remote sensing. Field surveys and mapping are highly accurate but are time consuming and labor intensive. Remote sensing and high-resolution imagery refers to imagery obtained from airborne crafts or satellites that can efficiently delineate buildings over large areas with comparable accuracy.

The creation of building footprints from high-resolution imagery can be developed through simple rules-based methods (i.e., using pre-defined rules guided by the user) to sophisticated deep machine learning approaches. The common goal of these methods is to segment an image into categories corresponding to different parts of a landscape.

Building footprints are a necessary component of flood risk assessment, as they provide information about the location, size, and layout of structures relative to flooding sources.

Building footprints also provide the location and spatial arrangement of homes within a community and the distance between homes and wildland vegetation. Thus these footprints are essential when assessing risks due to wildfire and the pathways through which they spread.

Flood Risk Case Study

A case study conducted in Hillsborough County, Florida illustrates the importance of location accuracy. The study assessed expected flood losses and National Flood Insurance Program (NFIP) estimated premiums for single-family homes using two distinct location methods:

1. Parcel centroid data.
2. Building footprint data.

The results demonstrated that location discrepancies had a tangible effect on risk calculations. For 68% of homes analyzed, location estimates differed by more than ten meters, significantly influencing expected flood losses and insurance premiums. Elevation played a critical role, as homes with footprint-derived locations positioned three feet lower than their parcel-derived counterparts faced storm surge losses over three times higher. Similarly, when elevation differences were reversed, the risk varied just as dramatically.

Beyond horizontal inaccuracies, the study highlighted elevation differences of up to ten feet across short distances of five meters, with some cases showing differences of up to 30 feet when parcel-derived locations were significantly further from structures. These discrepancies directly impacted flood risk, with elevation proving to be the most decisive factor.

In some cases, parcel-based locations differed from footprint-based locations by over 300 meters, leading to significant miscalculations in elevation and proximity to flooding sources such as coastlines and rivers. The substantial variance in estimated flood losses and insurance premiums based on the method used underscores the necessity of footprint-based geospatial data for flood risk determination. The continued reliance on only parcel-based geocoding could result in unfairly high or low premiums, potentially exposing insurers to unexpected risk.

Wildfire Risk Case Study

The second case study examined the impact of building footprint and vegetation cover in calculating underwriting wildfire risk for three communities in California: Scripps Ranch (San Diego County), Grizzly Flats (El Dorado County), and Fountaingrove (Sonoma County).

Three critical factors were analyzed:

1. Vegetation Cover: Data on surrounding vegetation was collected for each home, categorized into wildfire preparedness zones (immediate: 0–5 ft, intermediate: 5–30 ft, and extended: 30–100 ft). Grizzly Flats appeared to have a higher wildfire risk based on vegetation coverage.
2. Structure Separation: The distance between buildings was evaluated using building footprint data. Greater separation limits fire spread, whereas closer structures increase the likelihood of widespread damage. Scripps Ranch exhibited higher risk due to smaller structure separation.
3. Distance to Wildland Vegetation: Buildings closer to wildland areas face greater wildfire risk. Geographic data was used to assess the exposure of each community to wildfires progressing through the Wildland Urban Interface (WUI). Under the study’s parameters, Grizzly Flats appeared most vulnerable.

The wildfire case study underscored the utility of footprint data in quantifying structure separation — a key driver in wildfire conflagration—and wildland vegetation proximity. Additionally, high-resolution imagery enabled the measurement of vegetation coverage near homes, providing crucial data for assessing both property-level and community-wide wildfire risk.

Additional Applications and Future Research

While the studies focused on flood and wildfire risk, high-resolution imagery has broader applications in property insurance.

1. Building Footprint Analysis:
• Essential for commercial properties where policies may cover multiple structures.
• Used in square footage calculations for replacement cost assessments.
• Improves validation of prefilled insurance data for more accurate underwriting.
2. Ground Feature Identification:
• Enhances risk assessment by detecting property characteristics remotely.
• Could reduce the need for in-person property inspections, minimizing costs for insurers and policyholders.
• Machine learning models can analyze aerial imagery to classify roof shape, material, and condition, tree overhangs, and other property or liability exposures.

Future Case Study Considerations

Further research is needed to incorporate estimated wildfire loss data, enabling a better understanding of premium differences between homes and communities. Additional studies could also expand ground feature detection beyond building footprints and vegetation cover to include roof type and attached fences, further refining wildfire risk models.

Moreover, the study relied on publicly available footprint data and did not compare different high-resolution sources or machine learning methods. Future research should assess the relative accuracy of various data sources, ensuring that insurers use the most effective techniques for improving risk modeling and insurance losses.

Conclusion

By adopting advanced geospatial technology and machine learning-driven analytics, insurers can reduce costs, improve risk differentiation, and create fairer premium structures. As these techniques become more commercially available, the property insurance industry should leverage their full potential, ultimately leading to more accurate and equitable insurance practices in an increasingly uncertain climate.

This article is provided for informational and educational purposes only. Neither the Society of Actuaries nor the respective authors’ employers make any endorsement, representation or guarantee with regard to any content, and disclaim any liability in connection with the use or misuse of any information provided herein. This article should not be construed as professional or financial advice. Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries or the respective authors’ employers.

Daniel S. Pribe, FSA, MAAA, is owner of Iris-Edge. Dan can be contacted at dpribe@iris-edge.com.