Thoughts on Geocomputing

To see where we are going as a geospatial community, it’s my belief that we must often look back from whence we came. As a Geospatial developer who is developing my own geocomputation course, I often reflect upon geocomputing’s origins. This interdisciplinary specialization combines geospatial science and computer science and the benefits to improve efficiencies and tackle complex problems in geospatial topics in academy and industry are numerous. While geography as a discipline has a history spanning over 2,000 years going back to ancient civilizations, most notably Eratosthenes in ancient Greece, GIS has its roots in the 1960s when Roger Tomlinson created the first Geographic Information System for the Canadian government. This marked what we now recognize as the start of the computer revolution, integrating computers into the discipline of geography and mapping our physical world. GIS has since been embraced as a standalone discipline and has also been adopted by several other disciplines and industries to solve spatial problems and build more sophisticated geographical analyses and models.

The underlying tool for these sophisticated workflows has been the computer, hence the emergence of geocomputation. While a relatively young term that emerged in 1996, it clearly outlined a scope outside of ‘quantitative geography,’ with emphasis on “creative and experimental” applications and the development of new tools and methods (Longley et al. 1998). Simply put, it allows practitioners to develop outside the restrictions of propriety software. Using programmatic languages such as Python, JavaScript, or R, these tools enable geocomputational experts to address highly complex, often non-deterministic problems in fields such as urban planning, natural resource management, and other areas studying spatial phenomena.

Now we sit at the forefront of another change within the discipline: Large Language Models in Artificial Intelligence (LLM AI). The most significant impact that LLMs are having within the geospatial sphere is the application of algorithms to identify real-world objects from raster imagery. These models, coupled with geocomputing, are analyzing vast amounts of spatial data far more quickly and accurately than humans, enabling geospatial professionals to extract valuable insights from complex datasets. This doesn’t necessarily equate to a loss of jobs or abandonment of specializations. LLM AI models serve as an enhanced assistant—in the common parlance of Silicon Valley, “AI makes us 10x engineers, bro.” However, without understanding the fundamentals, the tool isn’t as helpful; thus, we still require expertise in the subject matter.

Looking ahead, I’m optimistic that AI will continue to help practitioners instead of hindering them, but only if we truly work on becoming great developers. This requires staying sharp on common computer science topics and algorithms, bringing them into the geospatial discipline, and weaving these ideas into our development workflows to solve problems - no matter the industry.

References:

Longley, Paul, Sue M. Brooks, Rachael McDonnell, and Bill MacMillan, eds. 1998. Geocomputation: A Primer. Chichester, England; New York: Wiley.