Origin–destination flows

A flat table of origins and destinations becomes an animated flow map automatically. kepler does not auto-detect flows the way it does points and geometry, so the panel emits the layer itself.

SELECT o.lat AS origin_lat, o.lon AS origin_lon,
       d.lat AS dest_lat,   d.lon AS dest_lon,
       COUNT(*) AS trips
FROM trips t
JOIN zones o ON o.id = t.origin_zone
JOIN zones d ON d.id = t.dest_zone
GROUP BY 1, 2, 3, 4;

One row is one flow. The trips column becomes its magnitude.

The two ways to say where

A flow layer takes either a pair of coordinate pairs or a pair of H3 indices. When a query carries both, H3 wins — a hexagon flow is the more specific description of the same movement.

Coordinate pairs

Role	Detected from
Origin lat	origin_lat, origin_latitude, from_lat, start_lat, source_lat, pickup_lat, lat0
Origin lng	origin_lon, origin_lng, origin_long, origin_longitude, from_lon, from_lng, start_lon, start_lng, source_lon, source_lng, pickup_lon, pickup_lng, lng0, lon0
Dest lat	dest_lat, dest_latitude, destination_lat, to_lat, end_lat, target_lat, dropoff_lat, lat1
Dest lng	dest_lon, dest_lng, dest_long, dest_longitude, destination_lon, destination_lng, to_lon, to_lng, end_lon, end_lng, target_lon, target_lng, dropoff_lon, dropoff_lng, lng1, lon1

The lists are long because the naming conventions in the wild are: pickup/dropoff from taxi data, from/to from network tables, source/target from graph exports, lat0/lat1 from kepler's own convention. All four roles must resolve — three of four is not a flow.

H3 pairs

Role	Detected from
Origin H3	origin_h3, source_h3, from_h3, h3_0
Destination H3	dest_h3, target_h3, to_h3, h3_1

SELECT h3_from AS origin_h3,
       h3_to   AS dest_h3,
       COUNT(*) AS trips
FROM journeys
GROUP BY 1, 2;

These four roles are autodetect-only

originH3, destH3 and the two velocity pairs are not exposed in the Field mapping editor, which offers twelve of the twenty-one roles. If your H3 columns are called something the list above does not cover, alias them in the query — SELECT h3_pickup AS origin_h3 — rather than looking for a dropdown. See Field roles.

Magnitude

Role	Detected from
Count	count, trips, magnitude, weight, flow, total, volume

Optional. Without it every flow is drawn with the same weight, which is occasionally what you want — a network diagram rather than a volume map — but usually is not.

Aggregate before the map. A million individual journeys between two hundred zones is at most forty thousand flows after a GROUP BY, and the map cannot show you the difference anyway.

Line style

Flow line style in the panel options picks how the lines are drawn:

Value	Reads as
Straight (default)	a clean network; easiest to read when flows are dense
Curved	separates the two directions of a pair, so A→B and B→A do not overlap
Animated	direction is unmistakable; costs a little rendering, and moves in a dashboard

Curved is worth reaching for the moment your data is bidirectional. Straight lines between the same two points sit exactly on top of each other and a strong A→B flow becomes indistinguishable from a balanced pair.

Reading a flow map

Colour, width scaling and the magnitude thresholds all live in kepler's layer panel, as they would for any layer.

One flow-specific setting is worth knowing before you go hunting for a bug: Max Top Flows defaults to 5,000. Hand the layer more flows than that and the smallest ones are simply not drawn. If a flow you expected is missing from a dense map, raise that number — or, better, aggregate harder in SQL — before suspecting detection.

Why this needs a pre-release kepler

The Flow layer arrived in kepler.gl's 3.3.0 line and exists nowhere in the 3.2 stable release. That is the reason the plugin pins 3.3.0-alpha.11 rather than the stable version, and the reason the layer is enabled explicitly at startup rather than relying on the upstream default. See Differences from stock kepler.gl.