The Potential Pitfalls of "Geo" and "Spatial" Prefixes in Technology Innovations

Chris North, M.Sc.

In the rapidly evolving landscape of technology, the prefixes "Geo" and "Spatial" (and often “Location”) are frequently attached to various innovations to imply an enhancement or specialization of a geographic nature. However, this practice often leads to confusion, as the distinction between the geo-enhanced versions and their standard counterparts can be ambiguous or even non-existent. This article explores the pitfalls of this nomenclature, illustrating both genuine differentiations and those that are less convincing.

Full disclosure, I have long lamented the debates over whether it should be “Geo-this”, “Location-that” or “Spatial-these” or “GIS” or “GeoSpatial” blah, blah, blah.? I see this as largely “To-may-to To-mah-to”. It’s no secret I like “GIS” (more specifically, I like “Geography”), but I digress.

The Concept of Geo and Spatial

The prefixes "Geo" and "Spatial" are derived from geography and space, respectively. Historically, these prefixes are used pretty much interchangeably, and the choice of one vs the other is usually a matter of preference or originality (who said it first).? Case in point, generally we accept that the process of converting mailing addresses to coordinates is called “geocoding”, but there is no reason it couldn’t be called “spatialcoding”. The latter sounds silly only because we’ve always used the former.

For this article, I will be using these prefixes interchangeably, generally using the vernacular. The context in which I am referring to these prefixes is when they are intended to denote technologies or methods that incorporate geographic or spatial data and considerations. I’m taking issue with the notion that just by adding these prefixes to some existing technology, the resulting technologies will offer unique functionalities tailored to spatial data and its analysis. In some cases, this is true. In others, not so much.

Genuine Differentiation: Spatial Statistics

I do believe that spatial statistics is an area where the prefix "Spatial" signifies a true and necessary differentiation from traditional statistics. Standard statistical methods assume that data points are independent of one another and identically distributed. However, spatial data often violates these assumptions due to spatial autocorrelation – the concept that objects closer to each other are more likely to be similar than those further apart.

Spatial statistics incorporates methods to account for this spatial dependency. Techniques such as kriging, spatial autocorrelation measures (e.g., Moran's I, Geary's C), and spatial regression models are specifically designed to handle spatial data’s unique characteristics. These methods enable more accurate analysis and interpretation of geographic patterns and relationships, which would be missed or misinterpreted using conventional statistical methods.

Ambiguous Differentiation: Geodatabases

I think talking about a geodatabase, while not wrong, is confusing. A geodatabase, ostensibly, is a database that is optimized for storing, querying, and manipulating spatial data. On the surface, this seems like a necessary distinction, given the specialized requirements of spatial data storage and retrieval, such as spatial indexing and efficient querying of spatial relationships.

However, upon closer inspection, many of the capabilities attributed to geodatabases are not fundamentally different from those of traditional databases. Modern relational databases like PostgreSQL with the PostGIS extension, or even NoSQL databases such as MongoDB, can be equipped to handle spatial data effectively. They provide spatial indexing, querying, and analysis capabilities that work in a similar way to to those of “proprietary” (I dislike that term, but that’s for another article) geodatabases like Esri’s ArcGIS geodatabase.

To be clear I think Esri’s geodatabase should be more accurately considered a data model rather than a database. It provides a framework for organizing and managing spatial data, incorporating various data types, relationships, and rules within a geographic context. This model supports the storage of complex spatial information, including vector and raster data, topological relationships, and metadata, within standard database management systems like SQL Server, Oracle, or PostgreSQL. I personally think we should talk about a “geodatamodel” that serves as a structured schema to enhances the capabilities of conventional databases to handle geographic data efficiently. This would avoid the ambiguity that stems from the fact that there’s a conflation of the storage of spatial data in a way that is responsive and the organization of spatial data in a way that supports analysis and decision making.

Questionable Differentiation: Geo AI

Geo AI is a term that has gained traction as artificial intelligence and machine learning are applied to spatial data. The idea is that Artificial Intelligence (AI) algorithms can be enhanced to specifically address the complexities and nuances of spatial data, such as spatial autocorrelation and spatial heterogeneity.

In practice, however, the term Geo AI for me falls into a grey area. Many AI techniques used in Geo AI applications, such as convolutional neural networks (CNNs) for image analysis or recurrent neural networks (RNNs) for time-series predictions, are standard AI methods applied to the attributes of spatial datasets. The key difference lies in the preprocessing and integration of spatial data, rather than in the AI techniques themselves.

To add to the complexity, AI is really an umbrella term that covers both Machine Learning (ML) and Deep Learning (DL), and these subsets of AI are focused on algorithms that enable computers to learn from and make predictions or decisions based on data. This conflation of AI and ML also leads to confusion, as it implies that all AI technologies involve learning from data, overshadowing other important AI methodologies and creating unrealistic expectations about the capabilities and scope of AI systems.

Alas, “GeoAIMLDL” does not roll off the tongue quite as nicely.

I’d be remiss if I did not mention the amazing potential AI can play in automating the processing of very large data sets or assisting in the creation and execution of complex workflows for GIS applications. True, AI can automatically classify, cluster, and detect patterns within vast geospatial datasets which would be prohibitively time-consuming for humans to process manually. But I fail to see what is inherently spatial about the AI in these applications - the AI is in the automation and in the volume, not in the clustering. Are these useful? Absolutely! Are they “geographic” in-and-of-themselves? I’m not convinced.

There are some specialized methods in Geo AI, such as spatially-explicit flood prediction tools that show promise. However, from where I sit, currently the majority of "Geo AI" applications could (and should) be described as AI applied to spatial data without significant modifications to the underlying algorithms. Again, there is nothing inherently wrong with this, but it does raise the question of whether the term Geo AI genuinely signifies a new domain of AI, or simply a marketing tactic to highlight the spatial nature of the data involved.?

Consequences of Indiscriminate Differentiation

So, why do I care so much about this?

The indiscriminate application of these prefixes can have several negative consequences. First, it can dilute the perceived value of genuinely specialized technologies. When every new product is labeled with "Geo" or "Spatial," it becomes harder to distinguish which innovations are truly groundbreaking.

Second, it can mislead users into believing they need specialized tools for tasks that could be accomplished with standard or existing technologies. This can result in unnecessary expenditure and complexity, as users investing in "geo-enhanced" products that offer no real advantage over their traditional counterparts. This can then become a barrier to to the broader adoption of GIS within organizations.

Third, it can stifle innovation by creating an environment where superficial branding is prioritized over substantive technological advancements. Companies and organizations might focus more on rebranding existing technologies and systems with geo-centric labels rather than investing in developing truly novel solutions that leverage the unique value of spatial data.

Lastly, it is likewise dangerous to assume that a system is inherently superior simply because it stores location data, such as the latitude and longitude of fire hydrants in an asset management system. The mere presence of spatial data does not guarantee that the system effectively leverages this information for enhanced functionality or decision-making. Without proper tools and methodologies to analyze and interpret spatial data, the system is not utilizing the full potential of location, and thus fails to enable more effective management and planning of assets within a geographic context.

Conclusion

The prefixes "Geo" and "Spatial" can signify important technological advancements and specializations, as seen in fields like spatial statistics. However, their overuse and simplistic application to existing technologies often obscures the true nature of these innovations, leading to confusion and potentially inflated expectations.

I’m not suggesting these technologies do not have a role to play in geographic problem solving. I’m suggesting that we play a risky game if we are merely labeling something that is inherently non-spatial as something that is.

Rather, for the continued advancement of geospatial technologies, I am suggesting it is crucial to apply these prefixes judiciously, ensuring they reflect genuine differentiation. By doing so, we can better recognize and appreciate the true innovations in our field, while avoiding the pitfalls of superficial branding. The goal should be to foster clarity and innovation, providing users with the tools they truly need to harness the power of geography in salving real world challenges.