Introduction: The Dissolution of GIS and the Rise of the Workflow

This manuscript is titled SemanticGIS: Designing and Documenting Geospatial Workflows. This title is crafted to emphasise an important distinction: SemanticGIS names the methodology and pedagogy this book teaches; Geospatial Workflow names what a practitioner produces by applying it. One is the discipline; the other is its output.

We are witnessing an accelerating diversification of both what spatial analysis represents and how it is deployed. We must therefore decouple the intellectual core of the discipline—the design of the workflow—from its operational execution.

The ambition to steer this discipline away from its purely operational, software-bound origins is not novel; it has been part of an ongoing academic turf war since the 1990s. Early definitions, such as Roger Tomlinson’s (1968) characterisation of Geographic Information Systems (GIS) as a “data bank” and a “set of procedures”, firmly entrenched the field as a mechanical container for data. The 1990s marked an intellectual rebellion against this view, most notably when Michael Goodchild (1992) famously introduced Geographic Information Science in an attempt to reclaim the meaning of the abbreviation.

This manuscript does not attempt to adjudicate that debate. Instead, for the sake of pedagogical precision, it fixes the two terms to two distinct referents. We reserve GIS strictly for the operational technology stack — the software and code through which spatial logic is executed, whether QGIS, ArcGIS, or a Python ecosystem. We use Geospatial Workflow for the process this manuscript actually teaches: the disciplined, documented reasoning by which a spatial question is translated into an executable analysis. This is a terminological choice for clarity within these pages, not a claim about how the wider field should resolve its own vocabulary.

From a pedagogical perspective, it is time to refocus. What we once called GIS has fundamentally dissolved. It is no longer a standalone scientific discipline; it has become infrastructure. This is not a speculative claim about a distant future in which autonomous agents execute entire analyses unsupervised; it is a verifiable reality today. A student can ask any conversational AI exactly how to perform a spatial join in QGIS and receive a correct, step-by-step answer in seconds. But the correctness of that answer is conditional in a way that is easy to overlook: the AI will faithfully answer the question it is asked, not the question that should have been asked. Consider a student working with a dataset in a geographic coordinate system who asks how to create a buffer around a set of points. The AI will readily explain how to invoke the buffer tool—yet the buffer operation is not meaningfully defined on unprojected coordinates in the first place. (As Chapter [X] will detail, distances and areas computed in a geographic coordinate system do not correspond to real-world distances or areas at all.) The real, prior question—how to reproject the data so that a buffer becomes meaningful—was never asked, because the student did not yet know it needed to be. This is a version of what is sometimes called the X-Y problem: help sought for a proposed solution, X, when the underlying need was actually Y, and no one—including the AI—surfaces the mismatch, because the AI answers the question actually posed. The comprehension gap here does not belong to the artificial intelligence; it belongs to the human framing the prompt, in a process whose underlying conditions have not yet been established.

This reframes what has changed. The procedural knowledge that once justified an entire university module—which button to press, which menu to open—genuinely is no longer scarce. What remains scarce, and what remains teachable, is the conceptual substrate that allows a student to pose the right question in the first place: an understanding of data models, projections, attribute typing, and topology sound enough to recognise when a workflow is about to go wrong before an AI—or a GIS—executes it uncritically. Judgement, in this sense, is not merely deciding which operation to run; it is possessing the conceptual grounding to interrogate the fitness of an operation before delegating its execution. By the time current students complete their degrees, they will very likely be directing agents through geospatial analyses rather than performing every operation by hand. This is already how a growing share of professional spatial data science is conducted—inside AI-supported coding environments rather than traditional desktop interfaces. Preparing students for that reality, however, does not mean handing them an autonomous agent on day one. In a learning environment, an agent that performs the operation on request short-circuits the very judgement this manuscript aims to build. It is akin to handing a calculator to a student who has not yet learned to reason about numbers: it produces answers without producing understanding.

This manuscript therefore insists that practitioners perform their own operations throughout their foundational training. We mandate this not because the mechanical execution itself carries intellectual weight, but because doing the work yourself is what exposes the gaps in your workflow design. Delegation is a privilege earned by—not a substitute for—having mastered the epistemology first.

This manuscript builds upon Nicholas Chrisman’s (1997) definition of GIS as an “organised activity” by which people measure and operate upon geographical phenomena. We share Chrisman’s view of the “organised activity”, but we no longer call this organised activity “GIS.” Instead, we use the term Geospatial Workflow. The focus of SemanticGIS is to teach the rigorous design and documentation of these workflows.

The Boundary: Projects vs. Workflows

To navigate this modern landscape, we must strictly define our boundaries. Practitioners rarely perform spatial analysis for its own sake. The overarching endeavour—whether it is planning urban infrastructure, simulating traffic architectures, or managing environmental risk—is the Domain Project. The Project belongs to the wider world; it sets the brief and defines the human intent.

Within that overarching Project lies the Geospatial Workflow. The Workflow is the specific, documented sequence of semantic logic designed to answer the spatial questions demanded by the Domain Project.

The Value of Formalism: Generating Semantic Assets

When introduced to the SemanticGIS workflow, one might ask: Why all this formalism? Why not just jump straight to executing the algorithmic schema?

In extremely simple scenarios, a routine professional might indeed skip to execution. However, in the vast majority of modern spatial problems, the value of the formalistic approach profoundly outweighs the initial overhead. The advantage of the SemanticGIS methodology is that you are not merely generating a simple map or report; you are generating a suite of Semantic Assets:

1. The Design Rationale: A permanent record of every “Wicked Problem” encountered and the logic used to resolve it.

2. An Explicit Data Model: A text-indexed bridge connecting the raw, systemic data to the human intent of the workflow.

3. The Analytical Recipe: A decoupled, reproducible script (a Directed Acyclic Graph) that allows the project to be picked up by a new practitioner—or an autonomous computational agent—without you in the room.

These assets provide distinct value across the profession:

• For the Student: As spatial operations rapidly diversify, learning to click buttons in a single software environment offers fleeting career value. The enduring value lies in mastering the abstract, interoperable concepts these assets represent.

• For the Consultant: The value lies in deconstructing the brief into operational elements up front. Through the Design Rationale, this deconstruction is communicated to the client, ensuring compliance with explicit and implicit requirements early in the process. Delivering these Semantic Assets alongside the final map provides immense added value.

• For the Scientist: The ability to share and debate a workflow with peers without getting bogged down in software execution details or violating sensitive data protocols is paramount. Executional minutiae typically distract from the intellectual science.

Ultimately, these well-documented assets become indispensable. In an era where computational execution is increasingly “black-boxed”—often performed by AI agents—the original human intent of a workflow can easily become obscured. The ability to return to the Semantic Assets to audit whether the final mathematical output genuinely matches the initial human question is of profound scientific value.

The structure of the manuscript — iterative deepening of complexity

The five activities of the SemanticGIS workflow, coupled with continuous workflow stewardship, form the architectural scaffolding of this manuscript. However, mastering this workflow is like learning a new language: it demands iterative deepening of complexity rather than linear memorisation.

It is pedagogically fruitless to attempt to learn all the spatial “nouns” (the concrete concepts and data structures), followed by all the “verbs” (the abstract operations that act upon them), and finally the “grammar” (how nouns and verbs combine into a valid analytical recopy). In isolation, these elements hold no scientific meaning. They must be applied simultaneously to answer a spatial question.

Therefore, the teaching within this manuscript is structured as an iterative spiral of increasing realism:

• The Foundation (Part 2 The Hermetic Example Workflow): We begin with artificially simplified, closed-system examples. These initial iterations demonstrate the entirety of the pipeline—from defining the human intent to operationalising the syntax—allowing the practitioner to grasp the grammatical structure of the workflow without the friction of messy, real-world data.
• The Escalation of Realism (Part 3 Getting your feet wet): Once the foundational grammar is understood, we iteratively introduce noise. We expose systemic blind spots, conflicting cultural schemas, and the necessity of negotiating the sensoric extent.
• The Heterogeneous Convergence (Part 4 The deep dive): Finally, we abandon simple datasets entirely. The workflows will escalate to bridge chaotic sensoric manifolds (e.g., raw drone imagery and point clouds) with programmatic algorithmic logic and 3D urban environments, preparing the practitioner for the true complexity of the modern Domain Project.
• On your own: Supporting concepts and the technology stack. These parts are designed to provide help and guideance not co much relatet to the workflow but many of the concepts and tecnologies involved in the workflow.
• Finding inspiration. In this final section, we present inspirational workflows from different domains.

Through this iterative expansion, the practitioner learns to govern the spatial logic with absolute confidence, regardless of which operational tools are eventually deployed to execute it.