Title:City definition affects long-term urban scaling analyses in the United States (1900-2015)

Abstract:The scaling relations between city attributes and population are emergent and ubiquitous aspects of urban growth. Quantifying these relations and understanding their theoretical foundation, however, is difficult due to the challenge of defining city boundaries and a lack of historical data to study city dynamics over time and space. To address this issue, we analyze scaling between city infrastructure and population across 857 United States metropolitan areas over an unprecedented 115 years using dasymetrically refined historical population estimates, historical urban road network data, and multi-temporal settlement data to define dynamic city boundaries based on settlement density. We demonstrate the clearest evidence that urban scaling exponents can closely match theoretical models over a century if cities are defined as dense settlement patches. In contrast, these results strongly disagree with theoretical models and vary in time if we aggregate cities into administrative-defined metropolitan areas, a method used in previous analyses. Despite the close quantitative agreement with theory, the empirical scaling relations unexpectedly vary across space, a finding that is supported by global scaling analysis. Finally, our analysis of scaling coefficients reveal that a city in 2015 uses more developed land and kilometers of road than a city in 1900 with a similar population, which may imply cities have become increasingly harmful to the local environment. Our results show the strengths and limitations of current theoretical models, and offer a new way to study urban systems based on historical data.