Buildings Characterization at Global Scale: Challenges of Geospatial Semantics

Keywords: geospatial semantics, built environments, volunteered geographic information
Session Type: Virtual Paper Abstract
Day: Monday
Session Start / End Time: 2/28/2022 11:20 AM (Eastern Time (US & Canada)) - 2/28/2022 12:40 PM (Eastern Time (US & Canada))
Room: Virtual 59

Abstract

The knowledge of building characteristics across the world is a valuable source of information for many research areas such as population mapping, mitigation of hazards, energy analysis, and many others. In recent years wider availability of the building and architectural data on the Internet allowed researchers to combine diverse datasets to facilitate more detailed analysis of the global building stock, its vulnerability, and preparedness for various types of hazards. However, putting together sources of the data that come from different governments, commercial vendors, non-governmental organizations, and volunteers brings a number of challenges. In the building and architectural domain the biggest challenge is the semantic compatibility of the data sources: data may conform to a variety of building codes, architectural traditions, or may be expressed in several languages. In addition, the resulting combined datasets typically have a very uneven quality in terms of data completeness and often exhibit significant gaps in coverage and attributes. Here we present an approach towards development of a unifying framework for integration of global building geoinformation at various scales and levels of generalization. The framework builds upon existing interoperability standards such as CityGML, Global Earthquake Model (GEM) Taxonomy, and others. We discuss the problems of semantic interoperability of disparate building data and potential solutions based on Bayesian modeling, GeoAI, and data science approaches.