AGIT2026: AGIT 2026 – CONFERENCE FOR GEOINFORMATICS
PROGRAM FOR WEDNESDAY, JULY 8TH
Days:
previous day
next day
all days

View: session overviewtalk overview

09:00-10:15 Session 1A: WS: Geo-Knowledge Graphs and Geospatial Artificial Intelligence
Location: GI Studio
09:00
Workshop on Geo-Knowledge Graphs and Geospatial Artificial Intelligence

ABSTRACT. Hands-on workshop on Geo-Knowledge Graphs (GeoKGs) and Geospatial Artificial Intelligence (GeoAI). The workshop introduces the fundamentals of knowledge graphs, explains their differences from relational databases, and guides participants step by step in building a GeoKG from geospatial data. It further demonstrates how GeoKGs can support semantic integration, querying, reasoning, and GeoAI workflows, providing practical skills and reusable methods for geospatial research and applications.

09:00-10:15 Session 1B: WS: Mapping Hope
Location: Dekanatssaal
09:00
Workshop: Mapping Hope - Part 2

ABSTRACT. A collaborative space connecting geospatial researchers, peacebuilding practitioners, policy actors, and the Peace Tech industry with a focus on conflict prevention, digital inclusion, and locally grounded innovation across Sub-Saharan Africa.

09:00-10:15 Session 1C: WS: Multi-facetted Threats in the Modern World
09:00
Workshop: Multi-facetted Threats in the Modern World

ABSTRACT. We propose a workshop at AGIT 2026 titled Multi-facetted Threats in the Modern World, focusing on the various origins of threats and resulting risks to society. The workshop targets experts in disaster research and related fields and - in particular - aims to bring together attendees from the co-located DOD2026-conference with AGIT 2026.

09:00-10:15 Session 1D: WS: Vom Luftbild zum GeoAI-Orthophoto

Workshop

Location: HS 413
09:00
Workshop: Vom Luftbild zum GeoAI-Orthophoto: Integration von UAV-, Satelliten- und KI-Daten für die Erdbeobachtung der Zukunft
PRESENTER: Matthias Möller

ABSTRACT. Die hochauflösende Erfassung der Erdoberfläche befindet sich im Wandel: Klassische Luftbildbefliegungen werden zunehmend durch hochauflösende Satellitendaten, UAV-Systeme und KI-gestützte Auswertungsverfahren ergänzt. Der Workshop zeigt, wie sich diese Datenquellen systematisch vergleichen und künftig integrieren lassen. Im Fokus stehen Genauigkeit, Informationsgehalt, Kosten sowie neue Potenziale durch GeoAI. Ziel ist es, aktuelle Entwicklungen einzuordnen und Strategien für eine hybride, zukunftsfähige Orthophoto-Produktion zu diskutieren.

10:30-12:00 Session 2: KEYNOTE: Trisalyn Nelson & AGIT Opening
Location: Audimax
10:30
GIScience Provides a Path to Better Bicycling

ABSTRACT. Lack of available data can be a barrier to bicycling and pedestrian research and planning. Much of my team's research focuses on building novel data sources to map bicycling infrastructure, safety, and volume. In this presentation I will outline how we are using OpenStreetMaps to map bicycle infrastructure, crowdsource webmaps (BikeMaps.org and WalkRollMap.org) to map safety and access concerns, and modelling ridership using Strava data. I will discuss how crowdsource data provide unique opportunities to enhance active transportation data, but also challenges particularly around representativeness and equity.

13:00-14:15 Session 4A: GIP Forum 2026
Location: Blauer Hörsaal
13:00
GIP Forum 2026

ABSTRACT. ITS Vienna Region ist GIP-Österreich Betreiberin im Auftrag des ÖVDAT und launcht 2026 die neue Generation der GIP "GIP2.0". Wir möchten das zum Anlass nehmen, um die neue GIP2.0 im Detail vorzustellen und relevante F&E&I Projekte im Umfeld zu präsentieren bzw. mit den Expert:innen vor Ort zu diskutieren.

13:00-14:15 Session 4B: Scalable Data Infrastructures
13:00
Getting from Data to Training: Reproducible Cloud Pipelines for Earth Observation (and beyond)

ABSTRACT. Deep learning on large EO archives is often I/O- and data-engineering bound rather than model-code bound. I present a reproducible two-stage Python workflow that materialises versioned STAC-derived chip datasets on object storage (Icechunk/Zarr), then fine-tunes models with a high-throughput GPU ingestion pipeline (NVIDIA DALI). In a controlled benchmark, accelerated ingestion reduced end-to-end fine-tuning time by ~1.85x.

13:07
Habitat / Land Cover Change Detection in Alpine Protected Areas: A Comparison of AI Architectures

ABSTRACT. Rapid climate change demands frequent habitat monitoring in alpine ecosystems. We compare post-classification vs. direct change detection using geospatial AI (U-Net, Prithvi, Clay, ChangeViT) on high-res multimodal data from Gesäuse National Park, Austria. Post-classification with foundation models achieves superior multi-class accuracy. LiDAR integration boosts segmentation from 30% to 50% accuracy. Results show promise for operational monitoring despite class imbalance challenges.

13:22
Integrating Local Large Language Models into WebGIS: An Open-source Framework for Natural Language Interaction

ABSTRACT. WebGIS applications are well-known for interacting with the geospatial information, where integration of large language models (LLM) can be promising to offer chat-based natural language interaction (NLI) features to explore on location-related questions. We implemented an open-framework that enables chat-based user interaction into geospatial actions by respecting data privacy. A python native prototype application is developed, tested using Ollama with controlled WebGIS interaction.

13:37
Towards the unified provision of country wide high resolution geospatial products using STAC

ABSTRACT. Countrywide high-resolution datasets in Austria are scattered over various platforms preventing uniform access. This leads to redundant work for various users to integrate them into their workflows. Inspired by the provision of satellite based EO data, we used Spatio temporal asset catalogs (STAC) to establish uniform access demonstrated through the prototype for selected datasets. This represents a first towards the uniform provision of Austrian geospatial datasets.

13:52
Semantic Earth observation (EO) data cubes to analyse the data availability of Sentinel-2 and Landsat

ABSTRACT. Earth observation (EO) based analyses assume the availability of cloud-free remote sensing data. To evaluate these prerequisites – especially to gain location-specific information about the annual number of cloud-free observations – we leverage a semantic data cube approach. The applied EO cube framework (i) fetches cloud-optimized data on demand via STAC, (ii) contains pre-defined options for semantic and knowledge-based modelling, and (iii) scales through chunking and parallel execution.

13:00-14:15 Session 4C: YR: Meet the Keynote Speaker
Location: GI Studio
13:00
Meet the Keynote Speaker: Trisalyn Nelson
13:00-14:15 Session 4E: WS: Change, Displacement and Migration - Narratives, Evidence, Models and local Perspectives, Big (EO) Data and Machine Learning
Location: HS 413
13:00
Climate Change, Displacement and Migration - Narratives, evidence, models and local perspectives, Big (EO) Data and Machine Learning

ABSTRACT. The Groundswell Report projected up to 84 mio climate migrants within Sub-Saharan Africa by 2050 under a pessimistic scenario, yet observational evidence that links climate with migration is sparse. We shed light on this nexus from different angles. What signals of a changing climate do we observe in Earth observation (EO)? What do Call Detail Records tell us? How do mathematical climate-mobility models conceptualize the nexus? How do quantitative methods relate to qualitative research?

14:45-16:00 Session 5A: Stadtplanung und Planungsprozesse (DE)
Location: Blauer Hörsaal
14:45
Entwicklung einer grenzübergreifenden Datenbasis für die Raumbeobachtung in der EUREGIO Salzburg - Berchtesgadener Land - Traunstein

ABSTRACT. Das Ziel des Projekts RABE ist die Entwicklung einer Datengrundlage für grenzübergreifende Raumbeobachtung in der EUREGIO Salzburg - Berchtesgadener Land - Traunstein. Um Hürden in der grenzüberschreitenden Zusammenarbeit abzubauen, werden nationale Daten harmonisiert und langfristig aktualisierbare Indikatoren errechnet, die in einer Online-Plattform zugänglich gemacht werden. Zur Vereinfachung grenzübergreifender Prozesse ist die Etablierung von Schnittstellen für Datenaustausch vorgesehen.

14:52
9 zu 1: Österreichweite Vergleichbarkeit der Flächenwidmung durch Datenharmonisierung

ABSTRACT. Die Präsentation zeigt einen Geodaten‑Explorer, der unter anderem eine von DataScience-Service erstellte generalisierte und homogenisierte Flächenwidmungsansicht für ganz Österreich bereitstellt. Diese Ansicht ermöglicht bundesweite Vergleiche sowie präzise standortbezogene Abfragen, während die offiziellen Informationen der Bundesländer gewahrt bleiben. Grundlage bilden regelmäßig bezogene Datensätze, die harmonisiert, vereinheitlicht und in einem konsolidierten Schema zusammengeführt werden.

14:59
Datengestützte Bottom-up Mobilitätsanalysen für integrierte Raum- und Verkehrsplanung

ABSTRACT. Die Publikation enthält Inhalte des Forschungsprojekts LösungsWege (gefördert vom BMIMI). Darin wurde u. a. ein datenbasierter Ansatz entwickelt, um kleinräumige Bewertungsgrundlagen zur Analyse von Mobilitätsmustern und deren Umweltwirkungen aufzuzeigen. Ziel ist es, Gemeinden und Regionen evidenzbasierte Entscheidungsgrundlagen für die Entwicklung integrierter, bedarfsgerechter Mobilitätsangebote bereitzustellen und damit nachhaltige Verkehrs- und Raumentwicklung gezielt zu unterstützen.

15:14
GIS als Beitrag zur systematischen Ausstattung von Quartieren mit Sharing Hubs

ABSTRACT. Sharing Hubs kombinieren – angepasst an die örtlichen Gegebenheiten und Nutzendenprofile – unterschiedliche Angebote von Fahrzeug- und Parkplatz-Sharing und bündeln und reduzieren damit Verkehrsflächen. GIS-bezogene Schwerpunkte des Forschungsprojekts Sharing Hubs sind die Identifizierung und Charakterisierung von geeigneten Fokusräumen für geteilte Mobilität, Visualisierungen erfolgsversprechender Szenarien sowie die Unterstützung der Konzeption von Betreibermodellen.

15:21
Urbane Digitale Zwillinge für die Klimawandelanpassung in Gemeinden: Der grenzübergreifende Ansatz des Projekts Twin4Clim

ABSTRACT. Der Klimawandel trifft insbesondere alpine Regionen stark und auch im Grenzraum Salzburg–Bayern sind die Auswirkungen bereits deutlich spürbar. Das Projekt Twin4Clim entwickelt grenzübergreifende Urbane Digitale Zwillinge zur Unterstützung kommunaler Klimaanpassungsmaßnahmen. Drei Use Cases zu baulichen Entwicklungsszenarien, grüner Infrastruktur sowie zu Starkregen zeigen konkrete Ansätze. Ergebnisse werden in einem offenen SDDI-Katalog und als übertragbare Handlungsanleitungen bereitgestellt.

15:28
Plattform Räumliche Energieplanung: Harmonisierte Geodaten als Grundlage der Wärmewende
PRESENTER: Anna Gabbert

ABSTRACT. Die Wärmewende erfordert konsistente und vergleichbare Geodaten als Grundlage räumlicher Planungsprozesse, insbesondere auf kommunaler Ebene. In Österreich erschweren föderal organisierte, heterogene Datenhaltungen jedoch integrierte räumliche Analysen sowie die Umsetzung nationaler und europäischer Vorgaben. Die Plattform Räumliche Energieplanung (PREP) begegnet dieser Herausforderung durch die Harmonisierung von Datenmodellen und Governance-Strukturen. Ziel ist ein bundesländerübergreifend koordinierter und vergleichbarer Betrieb von Energieinformationssystemen. Aufbauend auf methodischen Vorarbeiten wird ein einheitlicher Daten- und Integrationsrahmen geschaffen, der Energiedaten interoperabel nutzbar macht. Technischer Kern ist ein modulares, georeferenziertes Gebäudemodell als zentraler räumlicher Basislayer. Darauf aufbauend werden thematische Layer zu Wärme/Kälte, Mobilität, erneuerbaren Potenzialen und relevanter Infrastruktur integriert, gebäudescharf verknüpft und bundesweit vergleichbar aufbereitet. Die Anbindung heterogener Datenquellen erfolgt über standardisierte Modelle, definierte Schnittstellen und qualitätsgesicherte Transformationsprozesse. Dadurch werden Interoperabilität, Datenschutz und die Datenhoheit der Länder bei dezentraler Umsetzung in den bestehenden Landes GIS Systemen gewahrt. Die entstehende Infrastruktur unterstützt zentrale Aufgaben der räumlichen Energieplanung und erleichtert die Erfüllung europäischer Berichtspflichten. Die Umsetzung erfolgt schrittweise mit Pilotregionen, begleitendem Wissenstransfer und sukzessiver Skalierung. Als bundesländerübergreifende öffentlich öffentliche Partnerschaft etabliert PREP eine föderal getragene Geodateninfrastruktur und stärkt die Anwendung datenbasierter, GIS-gestützter Planungsansätze.

14:45-16:00 Session 5B: Climate Change & Urban Heat
14:45
Spatial Patterns of Urban Heat and Population Exposure in Zagreb (2001–2024)
PRESENTER: Dino Bečić

ABSTRACT. This study examines spatiotemporal changes in land surface temperature (LST) and near-surface air temperature (Tair) in Zagreb from 2001–2024 using Landsat imagery, CERRA reanalysis, and population data. Results show a mean LST increase of 4.44 °C and significant heat clustering (Moran’s I = 0.767). Fifty-four high-heat, high-density neighbourhoods were identified as priority zones for climate adaptation.

15:00
Towards Real Urban Comfort: Bridging the Gap Between Subjectively Perceived and Objectively Measured Comfort Parameters Using Low-Cost Sensors

ABSTRACT. This paper introduces a mobile, low-cost multi-sensor prototype to bridge the gap between objective and subjective urban comfort data. Leveraging "Urban Comfort Walks," the platform captures 18 features and real-time human feedback. We propose a three-step synthesis: (I) EW-AHP triangulation, (II) GraphSAGE for spatial interpolation, and (III) XAI for model transparency. This prototype serves as a scalable, human-centric diagnostic tool for urban planning.

15:15
Spatio-Temporal Trend and Hotspot Persistence Modelling of Urban Land Surface Temperature Using Multi-Year Landsat Time Series
PRESENTER: Mahbubul Alam

ABSTRACT. Rapid urbanisation intensifies surface heating in Dhaka. This study analyses long‑term Land Surface Temperature trends using Landsat 8–9 data (2014–2024) processed in Google Earth Engine. Mann–Kendall and Theil–Sen methods detect warming patterns, while a percentile‑based hotspot model identifies persistent heat clusters. Results show strong warming in dense built‑up areas and stable hotspots linked to urban morphology.

15:22
Measurable Targets and Response Indıcators for Clımate Adaptatıon in Germany's Buıldıng Sector

ABSTRACT. Currently, some European countries are seeking to regularly review the implementation of their climate adaptation efforts by setting measurable targets and monitoring progress. In Germany, the German Strategy for Adaptation to Climate Change (DAS) and the Federal Climate Adaptation Act (KAnG) provide the legal framework. Regular monitoring reports are published on developments. The DAS 'Buildings' field of action covers residential and non residential buildings, including land and real estate.

15:29
City-Scale Rooftop and Solar Panel Mapping from Very-High-Resolution Orthophotos Using Deep Learning

ABSTRACT. The rapid growth of rooftop photovoltaic (PV) systems requires scalable geospatial methods for automated rooftop and solar panel mapping. This study compares Mask R-CNN and YOLOv8 using 6 cm orthophotos of Salzburg. YOLOv8 achieved higher segmentation accuracy and substantially faster inference. Object-level validation confirmed reliable rooftop and solar panel detection, supporting scalable city-wide PV monitoring and urban energy planning applications.

14:45-16:00 Session 5D: WS: Fully Automated Flood Detection: Harnessing the Power of Deep Learning and Sentinel for processing large amounts of Data
Location: HS 413
14:45
Fully Automated Flood Detection: Harnessing the Power of Deep Learning and Sentinel for Processing Large Amounts of Data

ABSTRACT. Streamlining workflows for flood mapping is of utmost importance to quickly deliver information about the situation of affected areas. Therefore, this workshop will be a 75-minutes-long hands-on session with discussions about - how to procedurally Earth observation obtain data using STAC, - apply deep learning models, and - generate a final map ready to use. All workshop material will be provided inside a web browser using Google Colab. Attendees should bring their own laptop.

14:45-16:00 Session 5E: WS: Teaching Modern GIS - Exploring Smart Techniques in ArcGIS
Location: HS A1.052
14:45
Teaching Modern GIS - Exploring Smart Techniques in ArcGIS

ABSTRACT. The evolving topic of Modern GIS technologies has transformed the way we represent and utilize geographic data. This workshop aims to address these changes by providing lecturers with suggestions and inspiration for using Modern GIS technologies. Thereby forming coherent workflow in ArcGIS on urban sustainability and resilience in Berlin. In this context it presents an iterative didactic concept for working with spatial data.

16:30-17:30 Session 6B: Poster Lightning Talks
16:30
Poster Lightning Talks
17:30
Constrained Photogrammetry for High-Precision Metrology in refractive Media

ABSTRACT. Underwater 3D reconstruction is challenged by refractive distortion and unstable low-cost optics. This study presents a new constrained photogrammetry technique using a GoPro HERO4. By pre-calibrating intrinsic parameters via OpenCV and fixing them during bundle adjustment, lens distortion is decoupled from refraction. Compared to unconstrained workflows, this method reduces scaling RMSE by 35.6% (22.2mm to 14.3mm), eliminates "doming," and achieves sub-millimeter linear accuracy.

16:30-17:30 Session 6C: PANEL: The future of GI-Education in the light of AI
Location: GI Studio
16:30
UNIGIS International
PRESENTER: Martin Loidl

ABSTRACT. MODERATOR: Martin Loidl

Generative AI marks a disruptive change in many fields, not least also in academia and academic education. In general terms we are confronted with the questions about which intellectual competences that we want to retain as humans and for what we are happy to rely on machines. This general question is even more pressing in IT-related fields like Geoinformatics, where AI not only supports workflows, but for themselves become key methodological toolsets. We can view this from two perspectives: AI4Geo - how AI technologies can address challenges in spatial domains, and Geo4AI - how spatial thinking can improve AI's performance and thus contribute back to the broader AI community. Will we need GI-experts in 5 to 10 years from now? Which competences will they need to have? Can we take a high-level role as spatial thinkers without hands-on experience in applying analytical methods?

 

PANEL: Johanne Scholz | Céline Rozenblat | Mike Gould | Josef Strobl

16:30-17:30 Session 6D: WS: Unified access to high resolution raster EO datasets in Austria through STAC
Location: Dekanatssaal
16:30
Unified access to high resolution raster EO datasets in Austria through STAC

ABSTRACT. We developed a prototype for the unified access to high resolution EO raster products in Austria using SpatioTemporal Asset Catalogs (STAC). Within this workshop, we first outline the required steps to represent EO datasets as STAC collections, followed by a hands-on tutorial demonstrating the established services. The first part of the workshop will concentrate on automatic access through Python and the second part focuses on manual access through QGIS.

16:30-17:30 Session 6E: WS: Harnessing GeoAI tools for Urban Citizen Science
Location: HS 413
16:30
Harnessing GeoAI tools for Urban Citizen Science

ABSTRACT. GeoAI is transforming the way geospatial professionals and users interact with maps and spatial data, as well as how they can understand their surroundings. The rise of this tool also helps support planning professionals, citizen scientists, and decision makers with limited GIS knowledge to utilize GIS applications in their work. This workshop will provide a framework for participants to utilize GeoAI tools through different ArcGIS Online applications via a sustainable planning lens.