18.4 Visual Workflows — ModelBuilder, QGIS Graphical Modeler, Atlas

Key takeaways

• Visual workflow tools let you build analysis pipelines without code.
• Great for reproducibility, teaching, and collaboration with non-programmers.
• Export to scripts when pipelines grow or need version control.

Introduction

Not every GIS analyst is a programmer. Visual workflow tools let you chain operations graphically — connect boxes representing tools, see the data flow, and run the pipeline with a click. This lesson covers the main options and when they're the right choice.

ArcGIS ModelBuilder

Esri's flagship visual pipeline editor. Drag tools onto a canvas, connect inputs and outputs, add parameters. Run the model as a tool that colleagues can use.

Strengths

• Comprehensive: any ArcGIS geoprocessing tool is a model node.
• Exports to Python (arcpy script).
• Fits neatly with ArcGIS Pro's project structure.

Weaknesses

• Esri-only.
• Commercial licence.
• Models can get visually unwieldy beyond ~20 tools.

QGIS Graphical Modeler

QGIS's equivalent. Free, open source, with access to all of QGIS's processing tools plus SAGA, GRASS, and Orfeo.

Strengths

• Free.
• Works with open source algorithms.
• Exports to Python.
• Can chain QGIS / GRASS / SAGA tools seamlessly.

Weaknesses

• Model file format (.model3) is QGIS-specific.
• Less polished UX than ModelBuilder.

FME (Feature Manipulation Engine)

Commercial ETL tool focused on data transformation. Visual pipelines with 500+ transformers for format conversion, attribute handling, spatial operations.

Strengths

• Best-in-class for data conversion across obscure formats.
• Enterprise features (scheduling, notifications).
• Powerful attribute handling.

Weaknesses

• Commercial.
• Steeper learning curve.
• Overkill for simple pipelines.

Atlas visual workflows

Atlas.co's browser-based workflow builder. Drag-and-drop with live map preview.

Strengths

• Browser-based — no install.
• Modern UX.
• Cloud-native data sources.
• Collaborative.

When to use visual tools

• You're learning GIS and want to see the data flow.
• Your team has mixed technical backgrounds.
• The workflow is one-off and disposable.
• You need to hand off to a non-programmer.
• Quick prototyping before coding.

When to skip them for code

• The workflow will be reused many times — scripts version-control better.
• Complex logic (loops, error handling).
• Integration with other systems (APIs, databases).
• Testing — code is testable; visual models aren't.
• Transparency for code review.

Hybrid approach

• Prototype in the visual modeller.
• Export to Python.
• Refactor, parameterise, add tests.
• Version-control the script.
• Optionally keep the visual model for documentation.

Both QGIS and ArcGIS ModelBuilder export to Python — use this bridge routinely.

A worked example — land-use change pipeline

Visual model of a deforestation workflow:

1. Input 1: landcover_2010.tif.
2. Input 2: landcover_2020.tif.
3. Tool A: Align rasters.
4. Tool B: Reclassify both to binary (forest / non-forest).
5. Tool C: Raster calculator: year2020 - year2010.
6. Tool D: Polygonize loss patches.
7. Tool E: Filter patches ≥ 0.5 ha.
8. Tool F: Export to GeoPackage.
9. Output: deforestation.gpkg.

Seven tools, one page — a reviewer instantly understands.

Keeping visual models reproducible

• Parameterise — don't hard-code file paths; expose them as inputs.
• Document — write text annotations directly on the canvas.
• Version — commit the model file to git with each revision.
• Validate outputs — include QA tools that check sanity.
• Package — with sample inputs so others can run.

Self-check exercises

1. When is ModelBuilder preferable to a Python script?

For one-off analyses shared with non-programmer colleagues, for teaching GIS visually, and for rapid prototyping before moving to code. ModelBuilder makes the workflow self-documenting in a way scripts don't — the visual layout shows data flow immediately. For anything that becomes production or reused many times, a script is superior.

2. Why export a visual model to Python?

Version control (Python scripts diff cleanly; model files are opaque), testing, integration with other code, parameterisation beyond what the GUI exposes, and the ability to run from a command line or CI. Most visual tools are fine for creation but scripts are better for lifetime maintenance.

3. What's a sign that a visual model has outgrown the tool?

When you're zooming out to see the whole canvas, when you need branching logic (if/else, try/except), when the same sub-pipeline is repeated many times, or when you need iterating over collections. All of these are straightforward in code and awkward in visual tools. Refactor to a script.

Summary

• Visual workflow tools are great for clarity, teaching, and one-off pipelines.
• ArcGIS ModelBuilder, QGIS Graphical Modeler, FME, and Atlas cover the spectrum.
• Export to code when pipelines grow or need CI / version control.
• Use both — visual for design, code for production.

Further reading

• Esri — ModelBuilder Best Practices.
• QGIS Documentation — Graphical Modeler.
• FME documentation.
• Python Scripting for ArcGIS — comprehensive scripting reference.