GeoAI integrates AI, geospatial big data, and high-performance computing for solving data- and computation-intensive geospatial problems. This field has gained continuous momentum, driven by strong demands in geography and the rapid advancement of AI.
Wenwen Li´s paper “GeoAI for Science and the Science of GeoAI”, published in the latest edition of the Journal of Spatial Information Sciences, review trends in GeoAI research and discuss cutting-edge advances in GeoAI and its roles in accelerating environmental and social sciences.
The paper summarizes the milestone development of GeoAI methods and emphasizes the importance of integrating key spatial principles into AI model design to enhance the methodological foundation of GeoAI, which includes spatially explicit modeling, multi-source and multimodal GeoAI modeling, spatiotemporal and multi-scale joint learning, and geography-informed model training and validation.
The paper further describes four interconnected research pillars that shape the science landscape of GeoAI: predictability, interpretability, reproducibility, and social responsibility.
The author emphasizes the critical need to advance GeoAI methods and techniques, enhancing their predictive capabilities with greater reliability and transparency. To further advance GeoAI, spatial thinking and convergent thinking will need to be consistently embed into the design of GeoAI models while also maintaining a commitment to social responsibility.
Reference
Li, W., Arundel, S., Gao, S., Goodchild, M., Hu, Y., Wang, S., & Zipf, A. (2024). GeoAI for Science and the Science of GeoAI. Journal of Spatial Information Science, (29), 1-17. https://josis.org/index.php/josis/article/view/349/193