Interface I3sPointLayerOptions

Request authorizer for secured ArcGIS services. Pass an ESRI authorizer (e.g. getEsriAuthorizer() from @bitruvius/esri-auth, or set one globally via configureEsriAuth() and omit this). Public services need neither this nor requestInit.

Auto symbology*: when no explicit model is set, parse the service's drawingInfo and, if it symbolizes with a 3D PointSymbol3D Object (a glTF "asset" — the Esri norm for trees / turbines), resolve + load that real .glb model and render it instanced per feature, auto-sized by the sizeInfo visual variable. Falls back to GPU dots when there is no resolvable model.

Resident memory budget.

Scene-anchor center [lng,lat] — pin for a wide-extent root whose OBB center is degenerate.

Initial color mode (default 'height' — feature positions carry no per-point color until symbology lands; height-ramps the dots so they're visible).

Vertical datum: drop on the map ('basemap'), or place at true absolute height. Default: 'basemap' for a ground-relative service (elevationInfo.mode relativeToGround/ onTheGround), else inherited from the declared vertical CRS (vcsWkid) when a geoid is available, else 'basemap'.

Advanced: a custom point decoder (overrides the built-in Draco decoder + the worker pool).

Constant on-screen point size (CSS px). Default 3 — feature points are sparse, so a fixed pixel size reads better than physical-radius attenuation.

Optional — you should not normally need this.* The geoid undulation N (m) at the dataset, used to place absolutely-georeferenced data on mean-sea-level terrain. The SDK resolves it automatically (it fetches the EGM96 grid on demand and samples it at the dataset), so pass this only to override that — e.g. for a dataset referenced to a local vertical datum EGM96 does not model.

An override shifts placement by the DIFFERENCE between your value and the EGM96 undulation at the dataset — so passing the conventional EGM96 value for your area is a no-op rather than a ~30 m jump.

Requantization grid (m) for the i32 position buffer. Default 0.001 (1 mm).

Auto-ground from the first decoded tile's lowest point (default true).

MapLibre layer id (must be unique within the map).

Label text COLOUR (any CSS colour, default white). Convenience — overridden by labelStyle.color.

Text labels — field.* Label each feature with the value of this attribute (by name), rasterised into a sprite atlas and drawn as screen-facing billboards over the base render (dots / models / icons). Overrides the service's labelingInfo field. Categorical / small label sets (species, type, class) are ideal; per-feature UNIQUE text at scale (every id) wants an SDF atlas — a follow-up that shares this renderer.

Cap on DISTINCT label strings rasterised into the atlas (default 1024). Beyond it, further unique values render with no label (logged once) — the categorical-atlas limit; unbounded unique text is the SDF follow-up.

Extra WORLD-space lift (metres) of labels above their feature, ON TOP OF the per-feature model height (model mode raises each label by its own model height so it clears the model). Default 0.

Screen-space lift (CSS px) of labels above their anchor point — the "slightly above" float that keeps the text clear of the feature. Default 8.

Text labels — master toggle.* false forces labels OFF even when the service declares them; true forces them ON, choosing a field automatically (the service's labelingInfo, else the renderer's own category field — e.g. tree species / furniture type). Omit for the default: labels show when the service declares labelingInfo or a labelField is given, otherwise off.

Label text SIZE in CSS px (default 14). Convenience for the common case — overridden by an explicit labelStyle.font.

On-screen size multiplier for label billboards (default 1).

Full text style for labels (font / color / halo / haloWidth / …). Takes precedence over labelSize / labelColor; merged over the service's labelingInfo symbol style, which is merged over the renderer defaults.

LOD quality bias on maxScreenThreshold (>1 coarser/lighter, <1 sharper). Default 1 in dots mode; 2.6 in model mode (heavy instanced models refine coarser).

Model mode*: render each feature as an instanced 3D MODEL (the Esri ObjectSymbol3DLayer / "models on points" path) instead of GPU dots — a glTF/GLB URL, raw bytes, or a built-in NormalizedMesh. v1 uses a uniform modelScale; per-feature scale/rotation from attributes is a follow-up.

Model mode DISTANCE-WEIGHTED LOD: the distance (m) over which the LOD coarsens ~2× — so mid/far features thin out progressively (Esri-style dense-near / sparse-far) instead of loading full density across the whole view. Default 400. Smaller = thins harder + closer (lighter / faster); larger = denser far. 0 disables (uniform LOD).

Model mode: max concurrent in-flight node loads (fetch → decode → GPU upload). Default: 4 with a main-thread decoder, else max(8, 2 × the worker pool's concurrency) — lower = smoother navigation (less per-frame main-thread work) but slower fill-in; higher = faster fill, more pan jank. (A worker-pooled decoder will lift this constraint.)

DISTANCE-LOD cusp.* Max distance (m) at which features load + render; beyond it the engine culls those nodes (so a whole national/global point-feature service doesn't stream + draw out to the frustum far plane / past the horizon — the "everything loads" perf hit). Default: ADAPTIVE — the distance at which a nominal feature would be ~modelMinScreenPx CSS px tall (so it tracks viewport + fov + the asset's height), clamped to a 4 km ceiling. Applies in BOTH model and dots mode. Set a fixed metre value to override.

Cull features once they would be smaller than this many CSS px tall on screen (drives the adaptive modelMaxDistance). Default 5. Larger → cull closer (lighter / faster); smaller → persist farther. Ignored when modelMaxDistance is set.

Model mode: per-feature HEADING attribute (by name, e.g. 'Rotatie', degrees CW from north).

Uniform model scale (metres) in model mode. Default 1 — multiplies any per-feature scale.

Model mode: per-feature SCALE attribute (by name, e.g. 'Boomhoogte') → model size = value·modelScaleFactor + modelScaleBase. Sizes each model by a real attribute.

Fired with the working-set load fraction (0–1) as it streams.

Portal base URL for resolving WELL-KNOWN NAMED web styles — a styleSymbolReference whose styleName (e.g. 'EsriThematicTreesStyle') has no per-layer styleUrl, so its 3D model assets live in an ArcGIS portal item. Default 'https://www.arcgis.com'. Set for an enterprise portal that hosts the style.

Static auth/headers escape hatch (e.g. a Bearer token) for tile + metadata fetches. Superseded by auth when both are set.

Geometry codec seams (Draco/meshopt/ktx2) if the model GLB is compressed.

Start with cast shadows on (model mode only — the tree models cast into the shared cascade atlas alongside any mesh/building datasets). Needs a sun via I3sPointLayer.setSunState; toggle live with I3sPointLayer.setShadows. Default false.

Recompute CREASE-AWARE smooth per-vertex normals on the auto-resolved model(s) — average the face normals at each shared vertex, but only across faces within a ~60° crease so genuine hard edges (box/furniture corners, cylinder rims) survive. Esri's schematic web-scene models (e.g. the LOD3 thematic trees) ship FLAT per-face normals, so a faithful renderer draws them faceted ("geometry lines" on the low-poly canopy); ArcGIS smooths them at load. Default true (matches Esri; the crease test makes it safe for boxy assets too). Set false to render the authored per-face normals verbatim.

SceneServer or layer URL (…/SceneServer or …/SceneServer/layers/{id}).

Decode tiles off the main thread in a Web Worker pool (default true) — keeps navigation smooth while tiles stream. Set false to decode on the main thread (e.g. a CSP that blocks workers).