[Paper Review] MTBF-33: A multi-temporal building footprint dataset for 33 counties in the United States (1900 – 2015)

Johannes Uhl, Stefan Leyk|arXiv (Cornell University)|Mar 21, 2022

Land Use and Ecosystem Services17 references20 citations

TL;DR

MTBF-33 is a unique, open-access dataset of 6.2 million building footprints across 33 U.S. counties (1900–2015), derived from harmonized cadastral, tax assessment, and building footprint data. It enables high-resolution, multi-temporal analysis of urban growth and built environment evolution, supporting validation of historical land use models and training of deep learning systems for remote sensing-based urban change detection.

ABSTRACT

Despite abundant data on the spatial distribution of contemporary human settlements, historical datasets on the long-term evolution of human settlements at fine spatial and temporal granularity are scarce, limiting our quantitative understanding of long-term changes of built-up areas. This is because commonly used large-scale mapping methods (e.g., computer vision) and suitable data sources (i.e., aerial imagery, remote sensing data, LiDAR data) have only been available in recent decades. However, there are alternative data sources such as cadastral records that are digitally available, containing relevant information such as building construction dates, allowing for an approximate, digital reconstruction of past building distributions. We conducted a non-exhaustive search of open and publicly available data resources from administrative institutions in the United States and gathered, integrated, and harmonized cadastral parcel data, tax assessment data, and building footprint data for 33 counties, wherever building footprint geometries and building construction year information was available. The result of this effort is a unique dataset that we call the Multi-Temporal Building Footprint Dataset for 33 U.S. Counties (MTBF-33). MTBF-33 contains over 6.2 million building footprints including their construction year, and can be used to derive retrospective depictions of built-up areas from 1900 to 2015, at fine spatial and temporal grain. Moreover, MTBF-33 can be employed for data validation purposes, or to train statistical learning models aiming to extract historical information on human settlements from remote sensing data, historical maps, or similar data sources.

Motivation & Objective

• To address the scarcity of fine-grained, long-term historical building footprint data in the U.S. for urban change research.
• To integrate and harmonize open, publicly available cadastral and property data from 33 U.S. counties with building construction year attributes.
• To create a multi-temporal geospatial dataset that supports retrospective mapping of built-up areas from 1900 to 2015.
• To provide a benchmark dataset for validating historical land use change models and training deep learning models on urban change signals.
• To enable longitudinal analysis of building stock dynamics across urban areas of varying age and development patterns.

Proposed method

• Collected open data from county and state government websites, including parcel records, tax assessments, and building footprint GIS layers.
• Identified 33 counties where both building footprints and construction year ("year built") data were publicly available.
• Integrated and harmonized heterogeneous data sources into a consistent geospatial vector format using Albers equal-area conic projection (SR-ORG:74801).
• Retained all plausible construction year values from source data but constrained temporal scope to 1900–2015 to reduce survivorship bias.
• Produced 33 ESRI Shapefiles, one per county, with polygon geometries and associated construction year attributes.
• Applied data quality checks and reported completeness rates and temporal statistics per county to support user awareness of data limitations.

Experimental results

Research questions

• RQ1 How can historical building footprint data be reconstructed at high spatial and temporal resolution using non-remote sensing data sources?
• RQ2 To what extent can cadastral and tax assessment data provide reliable, multi-temporal building footprint information for urban change analysis?
• RQ3 What are the temporal and spatial patterns of built-up area expansion across diverse U.S. urban and rural counties from 1900 to 2015?
• RQ4 How does survivorship bias in construction year data affect the accuracy of historical urban extent reconstructions?
• RQ5 In what ways can this dataset improve the training and validation of deep learning models for historical urban change detection from remote sensing or maps?

Key findings

• MTBF-33 contains over 6.2 million building footprints with construction year attributes across 33 U.S. counties, spanning from 1810 to 2015.
• The dataset exhibits high completeness in year built data, with 97.7% completeness in Sarasota County (FL) and 95.9% in Boulder County (CO).
• The mean construction year across all counties is 1968, with a median of 1971, indicating a concentration of building activity in the mid-20th century.
• Survivorship bias increases significantly before 1900, with fewer records preserved for earlier periods, limiting reliability for pre-1900 reconstructions.
• The dataset enables retrospective mapping of built-up areas with fine temporal granularity, supporting detailed analysis of urbanization patterns.
• MTBF-33 is publicly available at https://doi.org/10.17632/w33vbvjtdy, facilitating reuse in urban modeling, population downscaling, and AI training.