UCGIS I-GUIDE Community Champions expand the community reach of I-GUIDE. This year, the Community Champions will have an opportunity to develop their own instructional materials (such as a single lab exercise) that involves the use of spatial AI while it leverages the I-GUIDE Platform. The materials will involve AI-ready spatial data and related machine learning models and applications while promoting FAIR (Findable, Accessible, Interoperable, and Reusable) data principles, and responsible and open science practices. Ultimately the materials will be hosted on the I-GUIDE Platform for wider access by the geospatial community.

Yao-Yi Chang, University of Minnesota

This hands-on lab introduces students to SpaBERT, a spatially informed language model that enhances transformer architectures with geospatial pretraining objectives. The module will guide learners through training and applying SpaBERT to geographic datasets, helping them understand how spatial proximity and semantic context combine to improve the representation of geographic entities. Students work with global and regional datasets (e.g., OpenStreetMap features or gazetteers) through workflows to preprocess geographic data, integrate spatial embeddings, and generate region-level embeddings that capture semantic, spatial, and topological patterns. Emphasis is placed on constructing and visualizing contextual region representations to support downstream tasks like population prediction or movement analysis. The lab will have conceptual and practical components to prepare students for applying spatial language models in domains like urban analytics, transportation, and geographic information science. It equips students with skills in spatial AI, natural language processing, and data-driven geographic reasoning.

Alexander Michels, University of Texas at Dallas

Computer vision is an increasingly useful tool for geospatial problem-solving due to the proliferation of street-view imagery (SVI). SVI is well-known for its utility for self-driving cars as well as problems in safety, mobility, planning, economics and more. My spatial Al lab will cover the fundamental concepts in computer vision, open data sources for SVI data, and example analyses with an Al-ready dataset. Aimed at upper-level undergraduates and graduate students, the lab begins with foundational computer vision concepts, followed by an overview of SVI applications in safety, planning, and economics. Students will use an AI-ready dataset and apply pre-trained machine learning models (e.g., from Hugging Face) to perform spatial AI tasks. The lab concludes with links to additional learning resources from the I-GUIDE Platform and UCGIS GIS&T Body of Knowledge. It will be piloted in three courses between 2026 and 2027 and will include slides and a lesson plan for instructional use across disciplines.

Yi Qi, University of Southern California

Urban trees provide critical ecosystem services that enhance environmental quality, community resilience, and human well-being. Traditional field-based tree inventories are costly and time-consuming, and recent advances in remote sensing and deep learning have opened new opportunities for mapping urban tree canopy at scale. This project will develop an open and transferable deep learning workflow for mapping urban tree canopy using freely available National Agriculture Imagery Program data. Building on prior work supported by the Bezos Earth Fund and piloted in Los Angeles, I will disseminate datasets, models, and training materials through the I-GUIDE Platform. By providing transparent workflows, instructional materials, and accessible tools, this project will not only deliver valuable resources for educators and students but also help democratize spatial Al for urban forestry, enabling local governments, researchers, and community organizations to monitor, plan, and sustain their tree canopy resources more effectively.

Siqin (Sisi) Wang, University of Southern California

This project brings spatially explicit species distribution modeling (SOM) using Maxent (Maximum Entropy Modeling) into the classroom. Designed for advanced graduate-level spatial analysis courses, it integrates machine learning principles, spatial data manipulation, and model interpretation to teach students how environmental variables influence spatial distributions of particular species. Using real-world crop data from Hawaii, students combine georeferenced occurrence points with environmental and soil variables to predict suitable cultivation areas. The curriculum covers data preparation, model configuration, execution, and validation, emphasizing machine learning concepts and spatial interpretability. Students learn to adjust Maxent parameters, interpret ROC curves and AUC scores, and use jackknife tests and response curves to assess variable importance. The outputs can readily be visualized with GIS to enhance communication and spatial storytelling.

Qiusheng Wu, University of Tennessee at Knoxville

Spatial Artificial Intelligence (GeoAI) is transforming Earth system analysis, yet integrating these powerful tools into geospatial education is often hindered by accessibility and infrastructure challenges. This project addresses these barriers by developing open-source GeoAI teaching modules that combine cloud-based data access, interactive visualization, and hands-on deep learning for spatial analysis. Using Python tools like SAMGeo and GeoAI, learners will be guided through end-to-end workflows—including data acquisition from platforms like the Microsoft Planetary Computer and AWS Open Data, model training, and geospatial image segmentation. Designed for reproducibility and scalability, the modules will be hosted on I-GUIDE’s JupyterHub as interactive notebooks. To broaden impact, the modules will be piloted through live training sessions and YouTube tutorials, and the I-GUIDE learning platform.