Computational mapping of historic built environments in American cities from Sanborn maps: A case from the Chicago Urban Heritage Project

Sanborn Fire Insurance maps are some of the most complete records of the historical built environment that researchers have access to, with building-level data for over 12,000 cities contained in hundred-year-old atlases. A challenge facing the historical urban planning research community is ensuring the preservation of these atlases’ information through scanning, but further digitization through vectorization has been largely contained to time-consuming hand-tracing methodologies that are prone to human error. While there is existing research to extract Sanborn footprints through machine learning, the methodology is costly in time and processing power. This paper contributes a scalable, open-sourced computational workflow for extracting building footprints from Sanborn maps to create vector polygons. We demonstrate this model’s scalability, speed, and accuracy by using a case from our Chicago Urban Heritage Project to map Chicago’s Hyde Park neighborhood in the mid-1920s. This case provides an example for how these automatically extracted footprints can open the door for spatial analysis of new eras of the historical built environment when combined with traditional datamining techniques.