Quantifying the geographic distribution of building coverage across the US for urban sustainability studies. (September 2018)

• Record Type: Journal Article
• Authors: Soliman, A.; Mackay, A.; Schmidt, A.; Allan, B.; Wang, S.
• Is Part Of: Computers, environment and urban systems. Volume 71(2018)
• Page Start: 199
• Page End: 208
• Publication Date: 2018-09
• Subjects: Building coverage ratio -- Green infrastructure -- Urban intensification -- NLME -- Urban sustainability; City planning -- Data processing -- Periodicals; Regional planning -- Data processing -- Periodicals; 303.4834
• DOI: 10.1016/j.compenvurbsys.2018.05.010
• Languages: English
• ISSNs: 0198-9715

Abstract

Abstract: The geographic distribution of building coverage is an important factor when evaluating the potential for green infrastructure and other sustainable development practices. However, comprehensive building coverage datasets are not available for many urban communities. We predicted the distribution of building coverage ratio across the contiguous United States (CONUS) at the scale of census block group using a bootstrap strategy applied to a nonlinear mixed effects model. Building footprints from 12 metropolitan areas were examined for completeness and aggregated at the level of U.S. census block groups ( n = 19, 109) and used to train the model. The best predictive model included both the percentage of the impervious surface area and housing unit density as nonlinear fixed effects in the form of a multivariate power function. Urban-type class, defined by median construction year of housing units and land use composition, were included in the model as a random effect. Cross-validation indicated that the selected model has a mean error of 0.049% (95% CI 0.047 and 0.051) of the estimated proportion of block group land area represented by building coverage. Adopting a bootstrapping strategy allowed selection of a subset with a minimum distance between samples to avoid spatial autocorrelation of residuals, while repeating the sampling for 1000 iterations to estimate the confidence intervals of the model parameters. Results are provided as open access data which quantify the geographical distribution of building coverage in terms of the ensemble predictions, mean and standard deviation. In addition, we describe trends in the model parameters across Urban-type classes. Although they were included as random effects in our predictive model, they will provide the urban research community with quantitative information about the effect of neighborhood age and type on the degree of urban intensification at a national scale. Highlights: A new open access dataset that quantifies the building coverage area of urban census block groups across the contiguous U.S. The dataset is based on the prediction of an ensemble of mixed effect models that account for urban spatial heterogeneities. Percentage of impervious land and housing density have fixed non-linear associations with the estimated area of buildings. Cross-validation showed that model estimated area of census blocks represented by buildings with a mean error of 0.049 %.