Why GIS Professionals will not be replaced by AI

How is geospatial being impacted by AI?

I often have conversations with students, graduates and emerging spatial professionals and one concern that they have is whether artificial intelligence will reduce the employment opportunities in geospatial. On a different, but related tangent, sometimes IT professionals ask me what makes GIS distinctive from any other branch of technology.

There is no doubt that technology advances have already had a significant impact on the geospatial profession and adjacent professions. For example the discipline of photogrammetry where LiDAR and one type of AI, edge matching and automated feature extraction and classification from imagery, has had a major impact on the data capture branch of geospatial professions. The surveying profession is also being impacted by LiDAR, drones and GPS.

Certain geospatial tasks such as adding metadata or writing documentation are being aided by Large Language Models so there is no doubt that the geospatial industry will be aided impacted by advances in AI.

There are early attempts to create maps from natural language instructions, which can be fine for certain use cases where the data is easily accessible, simple and has been structured and where the use case is not critical (so errors are unlikely to cause major issues). For example "create a map of the coffee shops in my city that have opened in the last year":

• This is simple data (point data for coffee shops)

• Coffee shop data is widely available and doesn't require much processing (data format changes, reprojections etc) to be used

• For the use cases that this request is probably being used for if the results are not accurate it is probably not going to matter

What is so special about spatial?

Accuracy Considerations

However, there are many aspects of the use of geospatial technology where the analysis is not as simple as the example given above. This is where the role of geospatial professionals remains critical and is why the geospatial profession is not going to be replaced with AI any time soon. My perspective on AI is:

Artificial intelligence is a great tool, provided you already know the answer to the question that you are asking it.

AI can be used to speed up certain tasks such as image analysis, but understanding when to trust the results and when to take a better look at the output data remains a skill that will continue to be needed.

Decision-making in complex situations

Another critical skill of GIS professionals is making the right choices from many options to meet the requirements of a specific use-case. Take for example if there was a request from a user for road network analysis. An early decision that a GIS professional will need to make is which road centerline dataset is going to be fit for purpose to meet the requirements. There are many potential road centerline datasets, with different strengths and weaknesses and different costs.

1. Open StreetMap might be one contender, where the cost factor is a significant advantage. However if there are complex requirements, such as lane specific functions then OSM might not currently have the data to support that.

2. Another option might be a government supplied road network generated from decades of data capture but where perhaps the spatial accuracy in rural areas is not great.

3. Another option might be a commercial road-centerline dataset, with high accuracy and lane specific data, but with high costs.

4. If super-high sub-meter accuracy is required then perhaps a new data capture activity might be necessary.

Evaluating these options to make the best decision according to many factors is something that AI cannot currently achieve, and it will take a long time for AI to be capable of making these types of subjective decisions.

Special characteristics

There are some aspects of GIS that are different to most other branches of IT. If the 'general IT' governance has representatives that have sufficient understanding of these differences then GIS sitting under general IT governance can work OK. But if the general IT governance does not understand specialist aspects of GIS then that is when there can be problems, and perhaps separate GIS Governance may be required, or at a minimum better representation of GIS perspectives.

 In many respects GIS shares many characteristics that are common to all other branches of technology: therefore, there is sometimes an assumption is made that geospatial  governance can be covered under general IT governance (or data governance etc). This can be true provided that geospatial characteristics are understood and adequately taken into consideration. The follow are just some of the many characteristics and examples of geospatial technology that are not common with other types of systems:

Spatial query functions

GIS applications have additional querying capabilities in addition to standard SQL. Spatial query functions retrieve and analyze spatial data based on their spatial relationships. These functions enable users to perform operations such as finding nearby features, identifying intersecting geometries, and determining containment or proximity between spatial objects. 

Topological relationships and accuracy

In GIS, topological relationships refer to the spatial connections and configurations between geographic features, such as adjacency, containment, intersection, and connectivity. Topological accuracy describes the precision and reliability of these relationships, ensuring that spatial data is correctly represented and analyzed according to real-world geographic phenomena.         

Geodesy

Geodesy is vital for GIS as it establishes datum and coordinate systems, guides projection selection to minimize distortions, supports geodetic control and surveying for precise spatial reference points, and underpins spatial analysis and modeling for informed decision-making. Understanding the implications of reprojecting geospatial data between projections can be important if the loss of accuracy is important for a specific use-case.               

Geospatial anomalies

Irregularities or unexpected patterns in geospatial data that can impact analysis and interpretation. Examples include doughnut polygons (inner holes in polygons), dis-contiguous polygons (disconnected shapes), and the Modifiable Areal Unit Problem (MAUP), which highlights how different spatial aggregations can lead to varying analytical results.

Vertex density

Understanding vertex density is crucial in GIS as it directly impacts the accuracy of spatial representations and the performance of geospatial analyses. In the context of the Coastline Paradox, where the length of a coastline increases as the measurement scale becomes finer due to increased vertex density, knowing how vertex density affects accuracy helps in selecting appropriate levels of detail for mapping. Moreover, managing vertex density efficiently is essential for optimizing computational resources and enhancing the performance of GIS operations, ensuring smooth data processing and analysis.            

Cartographic considerations

Scale considerations are crucial as they influence the accuracy and interpretation of spatial data. Understanding scale-dependent display ensures that features are represented appropriately at different zoom levels, avoiding generalization errors or data loss. Cartographic license, is a related concept of map design, emphasizing the importance of accurately representing data without misleading interpretations, especially when communicating information to the public or decision-makers.

Layer interactions

In the domain of GIS, layer interactions are crucial as they can lead to aggregate compounding inaccuracy when two or more layers with spatial inaccuracies are combined. Understanding how different layers interact spatially is essential to identify and mitigate errors that may arise from overlapping or compounding inaccuracies, ensuring the overall integrity and reliability of geospatial analyses and decision-making processes. 

Foreign key relationships are unnecessary

One key strength and distinctive capability of GIS is that foreign key relationships are unnecessary to relate different datasets together because spatial data in GIS is inherently linked by their geographic coordinates. Unlike traditional databases where foreign keys establish relationships between tables, GIS data can be spatially joined based on their spatial proximity or shared geometry without the need for explicit foreign key constraints. This spatial relationship allows for seamless integration and analysis of diverse datasets, making foreign key relationships redundant in GIS data management.

Prevalence of external data

In the domain of GIS, the prevalence of external data sources is crucial due to the spatial nature of geographic information. Unlike typical IT systems that primarily deal with internally created data, GIS relies heavily on diverse external geospatial datasets to enrich analyses, enhance decision-making, and provide a comprehensive understanding of spatial relationships and patterns.  

Spatial Accuracy

All types of information systems (including GIS) have various dimensions of accuracy such as completeness, coverage, currency, consistency, validity and attribution accuracy. However GIS has added accuracy considerations around the proximity of representative spatial features in the virtual world compared to the actual location of the actual objects they represent in the real world.               

Visualisation

Most IT systems such as financial, payroll, or HR systems primarily visualize tabular data, graphs and metrics, highlighting numerical trends and performance indicators without the spatial context inherent in GIS visualizations. GIS visualizations focus on mapping geographic information, such as terrain, land use, or infrastructure, to provide spatial context and insights.  This can include 3D and time-comparison visualisations.               

Spatial Concepts

Specialist spatial concepts like Tobler’s Law, spatial gravity models, kriging, Voronoi Diagrams, and Fractal Geometry are crucial in GIS for advanced spatial analysis and modeling. These concepts help in understanding spatial relationships, predicting spatial patterns, interpolating data, partitioning spaces based on proximity, and representing complex geometric structures, enhancing the depth and accuracy of geospatial analyses and decision-making processes.

Specialised Data Structures and Formats

Specialized data structures and formats like 3D formats, LiDAR formats, and the OpenStreetMap (OSM) road segment data model differ from typical tabular relational data structures in other IT systems due to their focus on spatial representation. These GIS data structures are designed to store and manage geographic information, including elevation data, point clouds, and detailed spatial attributes, enabling complex spatial analysis and visualization that go beyond traditional tabular data.  

Spatial Network Processing

Spatial Network Processing, including algorithms like the Travelling Salesman Problem and Dijkstra's Algorithm, is crucial in GIS for optimizing routing, logistics, and network analysis. These tools help in finding the most efficient paths, calculating distances, and determining connectivity within spatial networks, enabling better decision-making in transportation planning, emergency response, and infrastructure management by identifying optimal routes and analyzing spatial relationships in complex networks.         

Performance

GIS performance characteristics often differ from other IT systems due to the large volumes of spatial data involved and the complexity of computations required for processes like spatial analysis, network routing, geoprocessing and cartographic display. The handling of big data sets, intricate spatial algorithms, and the need for real-time processing in GIS applications can lead to unique performance challenges, necessitating specialized hardware, optimization techniques, and scalable architectures to ensure efficient and timely geospatial data processing and analysis.          

Spatial temporal changes

Managing spatial-temporal changes like parcel splits or merges involves specialized processes to track and update the evolving geographic features over time. Techniques such as versioning, temporal databases, and spatial-temporal data models are used to capture, store, and analyze changes in spatial data, enabling the representation of dynamic phenomena like land parcel modifications accurately. These unique GIS processes facilitate the maintenance of historical records, tracking of spatial changes, and analysis of temporal patterns, essential for effective land management, urban planning, and environmental monitoring.

There are many other examples of aspects of the geospatial profession that are not easily replaced by AI, and that make GIS distinctive from other forms of IT. Perhaps you might like to add some suggestions into the comments below.

All opinions expressed above are personal perspectives and do not necessarily reflect the views of my employer nor any other party.