This course introduces core concepts, methods, and tools in Geospatial Artificial Intelligence (GeoAI) for students on the MSc Geospatial Data Science and AI. Students will learn how modern AI models (e.g., convolutional neural networks, Transformers, graph neural networks, and geospatial foundation models or large language models) can be applied to Earth observation, environmental monitoring, smart cities, and social sensing problems. Through weekly reading discussions and hands-on labs, students will design and implement end-to-end GeoAI workflows on real geospatial datasets, culminating in a course project that tackles a real-world application. The course emphasises spatially explicit thinking, reproducible coding practices, and critical reflection on ethics, privacy and reproducibility in GeoAI.

Weeks 1 to 7: 2-hour lecture (weekly) plus 2-hour computer lab class (weekly)

Week 8: 4-hour oral presentations and peer feedback

Weeks 9-10: 4-hour computer lab class (weekly)

None.

GEOG5019 Principles of GIS or equivalent GIS knowledge

1. An essay (40%)

2. Short oral presentation. (10%)

3. A project report 50%)

This course aims to equip students with the conceptual understanding and practical skills needed to design and implement Geospatial Artificial Intelligence (GeoAI) workflows for real-world geospatial problems across environmental, urban, and societal domains.

This course will introduce students to the historical roots and core methodological foundations of GeoAI within GIScience, covering spatially explicit deep learning, graph-based models, representation learning, and geospatial foundation models.

Through student-led reading, hands-on labs, and a final project, this course aims to develop students' ability to critically select, implement, and evaluate GeoAI methods using open-source tools and reproducible coding practices.

A further aim is to foster critical awareness of ethical, legal, and societal implications of GeoAI (e.g., bias, privacy, and sustainability) and to encourage students to articulate responsible and inclusive approaches to GeoAI practice and research.

By the end of this course students will be able to:

1. Explain the conceptual foundations, historical development, and current research agenda of GeoAI within GIScience, identifying its distinctive contributions and limitations for geospatial problem-solving.

2. Justify appropriate GeoAI methods and data representations for a given application scenario, taking into account spatial autocorrelation, scale, uncertainty, and constraints in the geospatial domain.

3. Evaluate the performance of GeoAI models by experimenting on spatially-aware designs and metrics (e.g. spatial cross-validation).

4. Communicate results effectively to both technical and non-technical audiences through visual, oral, and written formats.

5. Develop end-to-end GeoAI workflows that ingest, preprocess, and model different types of geospatial data (e.g. Earth observation imagery, vector features, networks, and location-based data) by implementing state-of-the-art, open-source toolkits and algorithms.

6. Assess ethical, legal, and societal implications of GeoAI (e.g., issues of bias, fairness, privacy, energy cost, and reproducibility) and practical mitigation strategies.

7. Build a feasible and reproducible GeoAI project, including a presentation of project proposal, shared code repositories, and a detailed project report.