Tracking Between Cameras: Flow Maps and the Paths Animals Actually Take

Opening Scene

Trail camera seven catches a collared elk on Tuesday morning. Trail camera twelve, four miles northeast, catches the same collar’s signal Thursday evening. A point map would show two dots, unconnected, no different from two separate elk seen in two separate places. But a ranger tracking migration knows those two dots are really one story: an animal moved from here to there, likely along a specific corridor, likely for a specific reason tied to season or forage. Drawing that connection — an arrow from camera seven to camera twelve, weighted by how many collared animals made that same trip — turns two isolated facts into a picture of movement across the whole range.

A point map answers “where.” A flow map answers “from where, to where, and how much.”

In Plain English

A flow map visualizes movement or connection between locations, typically as lines or arcs between origin and destination points, weighted by volume, frequency, or magnitude. It’s a fundamentally different question from either a point map or a choropleth: those describe a static snapshot, while a flow map describes a transition. Flow maps are powerful and easy to make unreadable at the same time — origin-destination data grows combinatorially, and a flow map with too many overlapping lines becomes exactly the kind of unreadable tangle that a heatmap was built to avoid on the point-density side. Getting a flow map right means being deliberate about which flows matter enough to show, and how to encode their magnitude (line width, color, animation) without the map collapsing into visual spaghetti.

The Old Way

Before flow maps are designed deliberately, common defaults include:

Drawing every single origin-destination pair with equal visual weight — producing a tangle of crisscrossing lines where the busiest corridor looks no different from a one-time outlier trip.
Straight lines with no regard for actual geography — connecting two points with a direct line that ignores terrain, roads, or corridors the movement actually followed, misrepresenting the real path.
No aggregation threshold — plotting every recorded movement individually rather than aggregating repeated trips along the same corridor into one weighted line, guaranteeing clutter as data volume grows.

None of these are deliberate distortion. They’re what happens when movement data gets mapped the same way static point data would be, without accounting for how differently flow data behaves visually.

What’s Changing (and Why AI Is the Reason)

AI-assisted trajectory inference can now reconstruct likely movement paths between sparse observation points, filling in a plausible route between camera seven and camera twelve rather than a naive straight line, based on terrain and historical corridor data.
This raises the importance of labeling inferred paths as inferred. A model can produce a highly plausible route between two sightings; it cannot guarantee the animal actually took that exact path, and a flow map that doesn’t distinguish “observed” from “inferred” movement risks presenting a guess as a fact.
Automated flow aggregation and thresholding can now dynamically simplify a flow map based on zoom level or volume, showing only the most significant corridors at a glance while still preserving the full detail available on drill-down — a capability that used to require extensive manual curation.

The Metaphor, Fully Extended

Ranger Station Element	Geospatial Concept
The same collared elk seen at camera seven, then camera twelve days later	Two point observations that represent a single movement, not two unrelated events
An arrow drawn between the two cameras, weighted by how many animals made that trip	A flow map line, encoded by volume or frequency of movement
A tangle of arrows connecting every camera pair with equal weight	An unreadable flow map with no aggregation or magnitude encoding
A likely migration corridor inferred from terrain, drawn between two sparse sightings	An AI-inferred trajectory connecting sparse observation points, distinct from a directly tracked path
A simplified flow overlay showing only the busiest corridors at a glance	Automated flow aggregation and thresholding adapting detail to zoom level

For Beginners: What to Actually Do

Before drawing a flow map, aggregate repeated origin-destination trips into weighted lines rather than plotting every individual movement separately.
Encode magnitude clearly — line width, color intensity, or animation speed should make the busiest corridors visually obvious at a glance.
Distinguish directly observed movement from inferred or interpolated paths, both in the visual encoding and in the map’s documentation.
Limit the number of flows shown at once; a threshold that hides low-volume connections usually makes the map more readable and more honest about what actually matters.

For Practitioners and Leaders: The Deeper Layer

Require that flow maps built on AI-inferred trajectories visually and textually distinguish inference from direct observation, given how easily a plausible path can be mistaken for a confirmed one.
Invest in dynamic, zoom-aware flow simplification for any dashboard with growing origin-destination data, rather than letting flow maps degrade into unreadable tangles as volume increases.
Treat flow map design as a distinct discipline from point and choropleth mapping in your team’s visualization guidelines, since the failure modes and clutter thresholds are genuinely different.
Periodically audit which corridors a flow map foregrounds — an aggregation or thresholding choice that quietly buries a real but low-volume pattern can matter as much as one that fabricates a false one.

Quick Recap

Flow maps visualize movement or connection between locations, answering “from where to where” rather than a point map’s simple “where.”
Unaggregated flow data becomes an unreadable tangle quickly; deliberate magnitude encoding and thresholding are essential, not optional polish.
AI-assisted trajectory inference can reconstruct plausible movement paths between sparse points, but inferred routes must be visually distinguished from directly observed ones.
Automated, zoom-aware flow simplification now lets a single flow map stay readable across both a high-level overview and a detailed drill-down.

Where This Fits in the Series

Having covered static density in Article 5, this article shifts to movement itself — the second core technique in the series. Article 7 addresses a related but distinct question: once you know where something is, which region does it actually fall inside?