Assessment of methods for creating a national building statistics database for atmospheric dispersion modeling

Mesoscale meteorological codes and transport and dispersion models are increasingly being applied in urban areas. Representing urban terrain characteristics in these models is critical for accurate predictions of air flow, heating and cooling, and airborne contaminant concentrations in cities. A key component of urban terrain characterization is the description of building morphology (e.g., height, plan area, frontal area) and derived properties (e.g., roughness length). Methods to determine building morphological statistics range from manual field surveys to automated processing of digital building databases.

In order to improve the quality and consistency of mesoscale meteorological and atmospheric dispersion modeling, a national dataset of building morphological statistics is needed. Currently, due to the expense and logistics of conducting detailed field surveys, building statistics have been derived for only small sections of a few cities. In most other cities, modeling projects rely on building statistics estimated using intuition and best guesses. There has been increasing emphasis in recent years to derive building statistics using digital building data or other data sources as a proxy for those data. Although there is a current expansion in public and private sector development of digital building data, at present there is insufficient data to derive a national building statistics database using automated analysis tools. Too many cities lack digital data on building footprints and heights and many of the cities having such data do so for only small areas.

Due to the lack of sufficient digital building data, other datasets are used to estimate building statistics. Land use often serves as means to provide building statistics for a model domain, but the strength and consistency of the relationship between land use and building morphology is largely uncertain. In this paper, we investigate whether building statistics can be correlated to the underlying land use. If a reasonable correlation exists, then a national building statistics database could be created since land use is available for the entire U.S. Digital datasets of building footprint and height information have been obtained, validated and analyzed for eight western U.S. cities covering areas ranging from 6 km2 to 1653 km2. Building morphological statistics (including mean and standard deviation of building height, plan area fraction and density, rooftop area density, frontal area index and density, building-to-plan area ratio, complete aspect ratio, height-to-width ratio, roughness length, displacement height, and sky view factor) have been computed for each city at 250-m resolution and are being correlated to underlying land use type. This paper will summarize the building statistics from the eight cites focusing on the variability within each city and between cities as a function of land use.