Feature extraction and machine learning techniques for identifying historic urban environmental hazards: New methods to locate lost fossil fuel infrastructure in US cities

Abstract

U.S. cities contain unknown numbers of undocumented “manufactured gas” sites, legacies of an industry that dominated energy production during the late-19th and early-20th centuries. While many of these unidentified sites likely contain significant levels of highly toxic and biologically persistent contamination, locating them remains a significant challenge. We propose a new method to identify manufactured gas production, storage, and distribution infrastructure in bulk by applying feature extraction and machine learning techniques to digitized historic Sanborn fire insurance maps. Our approach, which relies on a two-part neural network to classify candidate map regions, increases the rate of site identification 20-fold compared to unaided visual coding.

When Environmental Inequality Racialized: Historical Evidence from Providence, Rhode Island

“…Methodologically, the study demonstrates the value of researchers’ growing capacity to harness newly georeferenced historical population data (Logan et al 2011) and new computational approaches for recovering detailed spatial information locked away in historical maps, organizational directories, and other primary source material (Bell et al 2020; Berenbaum et al 2019; Tollefson et al 2021; Trepal et al 2020). With such tools now becoming available, researchers can peer further into the past and train their focus more efficiently and more effectively at smaller scale units of analysis.…”

“…To overcome this problem, we developed a novel computational pipeline on the basis of an ensemble of feature extraction and machine learning tools. As described in Tollefson, Frickel, and Restrepo (2021), the publicly available pipeline (available at https://github.com/ TollefsonJ/Detect_MGP) identifies circular features of Sanborn maps that correspond to so-called gas holders: the large cylindrical tanks used to store manufactured gas at central production sites and to maintain gas line pressure at district substations (see Figure 2). Using this tool, we generated a cumulative list of relevant sites from the map corpus.…”

Combining Remote-Sensing-Derived Data and Historical Maps for Long-Term Back-Casting of Urban Extents

“…While most historical topographic maps only contain limited information on land use and building functions, some map collections such as the Sanborn Fire Insurance maps [ 20 ] contain such information. However, the extraction of such information requires more complex information extraction methods, such as deep learning [ 98 ]. Future work could also focus on the extraction of functional properties from such historical maps, facilitated by contemporary remote sensing data.…”

Creating building-level, three-dimensional digital models of historic urban neighborhoods from Sanborn Fire Insurance maps using machine learning

“…These methods may not be directly applicable to Sanborn maps. Although there is literature that uses machine learning to extract information from Sanborn maps [ 37 ], these methods are limited to specific types of buildings, such as manufactured gas production and storage sites, and are difficult to generalize to other information on Sanborn maps (e.g., building footprints of dwellings and stores). It is still difficult to create efficient workflows for extracting building-level information from Sanborn maps.…”

“…Over the past decade, the rapid advancement of machine learning techniques has facilitated the development of automated and semi-automated workflows for extracting geographic information from historical maps [10,[23][24][25] (see a more detailed discussion in the Background section). Methods have been developed to efficiently detect textual labels textual labels [26][27][28], land use [29,30], building footprints [31,32], road networks [33][34][35][36], and landmarks [37,38]. Existing methods, however, are mostly focused on maps other than Sanborn maps, such as the topographic maps from the United States Geological Survey (https://www.usgs.gov/programs/national-geospatial-program/historical-topographic-mapspreserving-past), which contain different types of geographic information and use different symbol and color systems than the Sanborn maps.…”