Derived layer

A derived layer is an item whose features are computed from other layers by a pipeline of analysis steps. The output rows are materialized into a PostGIS table and behave like any other layer: you can put them on a map, style them, query them, or feed them into another derived layer.

Pipeline model

A derived layer is an ordered list of steps, each producing a new intermediate result from its inputs. Available steps:

• Filter. Keep rows matching an expression.
• Calculate. Add a computed column.
• Buffer. Expand each input by a fixed distance.
• Spatial join. Attach attributes from one layer to features of another based on spatial relationship.
• Group by. Roll up rows into groups with aggregates.
• Dissolve. Merge adjacent polygons that share an attribute.
• Centroid. Replace each feature's geometry with its centroid.
• Fishnet. Generate a regular grid over an extent.
• Contour. Generate iso-lines from a value column.
• Clip. Trim features to fit inside a polygon.

Steps reference earlier steps' outputs by name. The last step's output is the derived layer's features.

Inputs

Each step takes one or more layer references as input. A layer reference is any of:

• A data layer sublayer (this org's, or another org's shared to you).
• Another derived layer's output.
• An ArcGIS service sublayer (streaming).

The pipeline runs server-side; the upstream layer doesn't have to be loaded in a browser for the derived layer to compute.

Materialization

A derived layer is materialized: when you save it, the portal runs the pipeline and stores the result rows in a real PostGIS table. Downstream queries hit the table directly, not the pipeline.

This means:

• Fast at query time. No per-request recomputation.
• Stale when inputs change. The pipeline doesn't re-run automatically on input edits. The detail page surfaces a Refresh button to recompute.

A scheduled-refresh option (every N minutes, hourly, nightly) is available for derived layers whose inputs change predictably.

Cost and limits

Each step has a documented cost profile. The detail page shows the last run's wall time and row count. Pipelines whose final output exceeds 5M rows require an explicit acknowledgement to save; very large outputs can saturate the org's storage budget quickly.

Sharing

Standard three-tier. A user who can read the derived layer but not its inputs still sees the materialized output rows (because those live in this item's table). This is the right call for most cases; if you need to lock derived output to upstream readers, share at the same tier as the most-restrictive input.

See also

• Buffer. The simplest one-step derived layer recipe.
• Spatial join. The most common multi-step recipe (filter, then join).
• Group by. Roll-up rollups.