Conference Agenda

Sustainable management of building stocks requires precise and comprehensive mapping of buildings' locations, shapes, and usages. However, detailed datasets with sufficient spatial precision and essential attributes are rarely available beyond national scales. Existing continental or global building inventories often lack the necessary spatial detail and attribute completeness required for many policy-relevant applications, limiting comparative research and large-scale analyses.

Here, we present EUBUCCO v1.0, Europe's most precise, attribute-rich, and comprehensively validated 3D database of ~320 million individual buildings, characterized by complete coverage of building height and usage-type information. EUBUCCO v1.0 achieves high spatial precision and completeness by integrating governmental, volunteered, and satellite-derived data sources using a novel, machine learning–based conflation framework. First, this framework spatially aligns and matches building footprints using 87 shape characteristics and contextual indicators, achieving an unprecedented matching accuracy at the continental scale with an F1 score of 99.5%. Then, it systematically merges and validates building footprints and attributes across data sources. Lastly, missing height and usage type attributes are inferred through machine learning models, achieving state-of-the-art generalization performance, with a mean absolute error (MAE) of 1.3 meters for building height and an F1 score of 0.72 for usage type classification.

Our analysis further reveals that relying solely on governmental or OpenStreetMap (OSM) data leads to spatially incomplete and geographically biased inventories, emphasizing the need for integrating satellite-derived data alongside robust conflation methodologies. Additionally, we demonstrate that commonly used intersection-based matching approaches, without prior spatial alignment, omit approximately one-quarter of true building matches, causing duplicates, limiting attribute enrichment, and erroneously excluding numerous buildings. Compared to the earlier version (EUBUCCO v0.1), these methodological improvements yield a 59% increase in the number of buildings and a 6-33% increase in ground-truth building attributes. By enabling comparisons across multiple sources, we provide transparent assessments of data completeness and ML-based attribute inference, establishing EUBUCCO v1.0 as Europe's most reliable and comprehensively validated building stock dataset. It provides a robust foundation for low-carbon, climate-resilient urban planning and evidence-based management of Europe's building stock.

Energy prices have become a central concern for European households, impacting living costs and public welfare. Ongoing global conflicts have contributed to the volatility of fuel markets, making energy unaffordable for thousands of families. In Scotland, for instance, fuel poverty has been on the rise since 2017, with current estimates indicating that approximately 34% of the population is living in poverty or extreme fuel poverty (Scottish Housing Condition Survey, 2022).

Qualitative research has introduced the concept of double energy vulnerability to describe the difficult and often enduring trade-offs many households face, such as choosing between heating their homes, preparing hot meals, or affording transportation (Martiskainen et al., 2021). Simultaneously, the UK’s NetZero Agenda, a national strategy aiming to reduce greenhouse gas emissions to net zero by 2050, is placing a growing pressure on the domestic sector to transition towards cleaner energy sources through household-level retrofitting and energy efficiency improvements.

These retrofitting schemes report an under-delivery rate according to the Energy Demand Centre, partially due to the lack of information on demand and socioeconomic circumstances at the local level that will enable local councils to offer relevant programs for different types of households according to their context (Kasprowicz, V. et al, 2024). A refined representation of how energy use is spatially distributed in Scotland and its relationship with socioeconomic characteristics at the household level is needed to approach these upcoming challenges. Unfortunately, the granularity required to understand the energy distribution at this scale is usually non-existent or unavailable due to privacy issues.

The goal of this study is to develop a realistic spatial microsimulation of Scotland’s population and its current energy consumption patterns using synthetic population to overcome the privacy restrictions, drawing on data from the Scottish House Condition Survey (2022), the Census (2022), and the Consumer Data Research Centre (2015). The study faces some key challenges, including: (1) integrating multi-scale, temporal data sources while addressing privacy and granularity constraints; (2) evaluating and selecting the most suitable method for synthetic population generation; and (3) identifying and recommending optimal validation methodology to ensure policy relevance of results.

Following the development of an energy consumption spatial model for Scotland, the generated synthetic population enables the creation of exploratory policy simulations to assess potential impacts on local policy frameworks. In this session, we will present the synthetic population tailored for energy use studies and preliminary insights on spatial disparities of energy consumption differences.

The spatial analysis of survey data combines social and spatial sciences in an interdisciplinary way, which entails a number of technical and legal challenges. The original survey data with the residential addresses of the people surveyed are subject to strict data protection and are made available to researchers in specially secured so-called ‘secure rooms’ in research data centres (FDZ). Thus, a data protection-compliant process is required that enables analyses to be carried out on the basis of aggregated data (e.g. to be able to investigate correlations), but does not permit re-identification. Researchers' access to the secure rooms on site is limited in terms of space and time. Comprehensive GIS skills are often required for the analyses, and the preparation of geodata sets and linking logics can be time-consuming.

The ‘Geolinking Service SoRa’ was implemented to close this gap and support for instance interdisciplinary research on housing equity. Linking socio-demographic and socio-economic data from surveys with spatial characteristics, such as population density and accessibility can give highly valuable insights and help answering questions like: How liveable is the neighbourhood for people who are particularly old, single parents or low-income earners compared to other social groups? How many parks and green spaces are there in the neighbourhood? How dense is the neighbourhood? Do residents living in areas with good access to daily amenities report higher satisfaction with their living conditions?

As a new research data infrastructure, SoRa Geolinking service intends to operationalise and simplify the linking of survey data with geodata. A distributed and decentralised infrastructure with several components enables the linking of different research data centres. Users can integrate SoRa into their scripts via an R package and access ready-made linking methods and geodata sets. The on-site use of SoRa within a secure room allows the original addresses to be used, while outside in public mode a synthetic structure dataset can be used to test links in advance. SoRa follows the principles of open source and FAIR, and can be expanded to include additional FDZs with new surveys.

The presentation will analyse exploratory spatial patterns of housing equity in Germany. The analyses are primarily based on survey data from the DIW Socio-Economic Panel (SOEP) as well as geodata from the IOER Monitor and the Federal Agency of Cartography (BKG), and can also be used for different time periods. The functionalities of SoRa will be demonstrated and a live demo (using JupyterLab) will show the user perspective. Finally, a discussion of the results and the potential of SoRa for socio-spatial research will conclude the presentation.

Urban mobility hubs are growing in complexity. Adaptive operational design for train stations, where the system depend on coordination across spatial, digital, and human layers, has a great importance for future-ready urban transport. This research introduces a hybrid Digital Twin for Harburg Train station, integrating real-time sensor data through innovative 5G networks, UWB sensors and radio cameras to build an AI predictor connected to the stations Building Information Model (BIM) using indoor navigation system.

The project includes for the simulations multimodal data, from radar, cameras, environmental sensors, and crowd flow tracking and real data to build a responsive model. A key innovation is the use of “data walks”: structured on-site observations of passenger behavior, used to annotate and contextualize sensor data for training transformer-based AI models. These models predict congestion points, operational inefficiencies, and route conflicts in real time. Indoor navigation improves user experience and generates feedback data, continuously updates the information flow. Scenario simulations are conducted at macro (timetable), meso (platform flow), and micro (individual movement) levels, enabling testing under different urban and operational conditions such as delays, events, or diverse weather.

To operationalize real-world testing, the system is implemented in the Oberhafen Gleishalle, an empty old train station in Hamburg that provides a spatial base to Harburg Station. Here, students from the university, designed and conducted scenario-based experiments addressing urban challenges such as governance, sustainability, mobility management, privacy, and public engagement. These student scenarios act as experimental probes, exploring how Digital Twins and data analytics can enhance decision-making and participatory urban development. Students engaged directly with the system, learning to interact with and behave inside the twin, using real Harburg data to guide actions and insights. On the other hand, these scenarios were evaluated through architectural design parameters, emphasizing human-centered spatial experience, accessibility, and usability, bridging digital systems with spatial design thinking.

This hybrid approach aligns operational station design and management with integrated urban mobility and land use strategies. The approach also serves as a calibration for the simulations, and it demonstrates how Digital Twins can become adaptive planning and decision support tools to build AI-powered urban systems.