In today’s fast-changing world, gaining an adequate understanding of the built environment is a critical prerequisite for many decision-making processes across industries. Specifically, an understanding of the location and dimensions of buildings (the “Building Footprints”) serves as foundational information for processes including insurance risk assessment, telecommunications network planning and deployment, and municipal urban planning.
In this blog post, we discuss challenges and common pitfalls we have observed over the past decade in pursuit of offering best-in-class Building Footprints at a country-scale, and highlight how Ecopia AI overcame these challenges to create and maintain the most comprehensive, accurate, and up-to-date Building Footprints map of the United States.
Challenges of accurate mapping at a country-scale
Gaining an adequate understanding of Building Footprints at a country-scale is a non-trivial task. Challenges encountered when sourcing Building Footprints can be summarized as follows:
1. Lack of coverage/comprehensiveness. Most high-value enterprise decision-making applications require a complete understanding of buildings across a given area of interest. Unfortunately, most commercially available Building Footprints databases are incomplete. This is primarily because they were generated through practices such as:
- Scraping fragmented Building Footprints open data: Attempting to connect with thousands of municipal sources — many of which do not produce or release Building Footprints data — to piece together a partial database of varying accuracy and recency; or
- Mining open-source imagery: Mapping Building Footprints based on freely available geospatial imagery that is often out-of-date (commonly by several years) and low-resolution (offering a sub-par level of detail) — resulting in an out-of-date and incomplete representation of reality.
2. Low accuracy. Enterprise decisions are often only as accurate as the data they are based on. However, generating and maintaining highly accurate Building Footprints at-scale requires robust Artificial Intelligence (AI) technology paired with scalable quality assurance (QA) practices, in addition to high-quality image sources. Many vendors do not have the required capabilities and image partnerships and, as a result, cannot offer clear specifications of accuracy metrics, or obscure the concepts by making broad claims of accuracy — which, as highlighted in Figures 2 and 3, are often unsubstantiated. When assessing accuracy, the following metrics should be carefully considered:
- False Negatives: The count of Building Footprints that were missed, represented as a percentage of the total Building Footprints that were supposed to be captured.
- False Positives: The count of the excess Building Footprints that were wrongfully captured, represented as a percentage of the total Building Footprints that were supposed to be captured.
- Interpretation Accuracy: The count of Building Footprints deemed to be a good representation of the underlying features as visible from the source geospatial imagery, represented as a percentage of the total Building Footprints that were captured in the delivered results.
3. Out-of-date. There is often a lack of transparency surrounding the updating schedules of third-party Building Footprints sources. This is because most providers receive their data from inconsistent sources (i.e. scraping fragmented Building Footprints data from municipal sources, or mining open source geospatial imagery), and consequently have no control over the frequency or quality of the updates. For instance, scraping data from municipal sources relies on the budgeting, image capture, and production cycles of each municipality (many of which offer updates infrequently, if at all). Comparably, open source imagery relies on the goodwill of imagery providers to continue to donate updated imagery to the open-source community (from our experience, due to revenue cannibalization risk, image providers predominantly offer outdated or reduced-quality imagery of minimal commercial value).
Ecopia’s solution: the most comprehensive, accurate, and up-to-date Building Footprints
Ecopia has generated hundreds of millions of Building Footprints across over 100 countries around the world. Example clients leveraging this data include the National Oceanic & Atmospheric Association (NOAA), Geoscape Australia (formerly PSMA), Land Information New Zealand, Ordnance Survey, the World Bank, the Bill & Melinda Gates Foundation, Munich Re, and CoreLogic.
Ecopia has generated Building Footprints for the entirety of:
- Australia (16M Building Footprints)
- United States (173M Building Footprints)
- sub-Saharan Africa (416M Building Footprints)
- all with industry-leading specifications surrounding comprehensiveness, accuracy, and recency — specifications that have consistently been independently validated by our customers.
Specific to the United States, Ecopia created and continues to maintain the preeminent database of Building Footprints with specifications as follows:
1. Comprehensive. To ensure the most comprehensive coverage, Ecopia first sourced the best commercially available high-resolution geospatial imagery across the contiguous United States (fresh <50cm aerial and satellite imagery). This imagery was then mined using Ecopia’s AI-based system to create an entirely proprietary database of every Building Footprints larger than 100 square feet — including all sheds, townhouses, barns, and other permanent structures. The resulting database currently contains over 173M Building Footprints across the United States.
2. Accurate. Ecopia’s system iteratively combines AI and quality assurance (QA) processes to mine high-quality image sources — technology that has been refined over the past decade, and results in the ability to maintain industry-leading accuracy standards at-scale. Ecopia contractually guarantees the following accuracy specifications for Building Footprints anywhere in the world:
- False Positives: <5%
- False Negatives: <5%
- Interpretation Accuracy: >95%
Within the United States, extensive independent customer evaluation has validated the accuracy of Ecopia’s Building Footprints to exceed these specifications as follows:
- False Positives: 0.91%
- False Negatives: 1.05%
- Interpretation Accuracy: 98.64%
3. Up-to-date. Ecopia has built a proprietary network of leading geospatial imagery providers that capture high-resolution (<50cm) aerial and satellite imagery of the United States every year — enabling Ecopia to offer annual updates generated with the best commercially available geospatial data on the market. As this blog post is being written, Ecopia has just completed a second annual update of the United States. In addition to updating the map annually, this capability enables Ecopia to offer a unique perspective on all new building construction, demolition, and modification across the country.
In summary, enterprise decision-making requires an adequate understanding of the built environment — of which Building Footprints form a foundational element. When investing in mapping data, we suggest carefully considering the above components — ensuring that decisions are being made on the most comprehensive, accurate, and up-to-date information.
Contact us today to learn about how Ecopia’s Building Footprints can help empower better data-driven decisions within your organization. Visit us at www.ecopia.ai to learn more about our commercial offering, and follow along on LinkedIn & Twitter @EcopiaAI.