Urban Environmental Insights from Citizen Science Data Using AI
Federico Stachurski (University of Glasgow)
Wednesday 4th November 12:00-13:00
Maths 311B
Abstract
This work, developed within the GALLANT WS3 project (Glasgow as a Living Lab Accelerating Novel Transformation, Work Stream 3, WS3), explores how qualitative, unstructured citizen science data from CommuniMap can be turned into meaningful urban environmental insights. CommuniMap, developed by Spotteron, collects geolocated images and short textual descriptions across different community categories (“CoLabs”).
First, we introduce Ask the Map, an interactive multimodal retrieval framework that allows users to query the dataset using natural language and retrieve relevant geospatial entries. The system combines text and image embeddings to capture both semantic meaning and visual context. We find that performance depends strongly on the type of query: visually grounded queries (e.g., “flooding in the streets”) benefit from image dominant retrieval, while more descriptive queries rely more on text.
Second, we show how these citizen observations can be pushed further into quantitative analysis. Focusing on the tree CoLab, we estimate tree characteristics directly from images and apply allometric models to approximate carbon stock and sequestration, providing a scalable way to assess urban greenery. Overall, this work highlights how AI and multimodal machine learning within GALLANT WS3 can bridge citizen science and environmental analytics, supporting use cases such as urban greening assessment and flood risk identification.
Add to your calendar
Download event information as iCalendar file (only this event)