← Digital Twin
Digital Twin · Data

Data compilation

Five distinct data streams are integrated into the BEAM Digital Twin — from physical sensor readings to 3D spatial models, geospatial layers, and machine learning outputs. Each stream is processed differently before being served to the platform.

Stream 01
3D building models

The DT's 3D environment combines two complementary building model sets — one providing high geometric detail, the other reflecting more recent construction. Both cover NUS Kent Ridge Campus including UTown, excluding Kent Vale.

3D building model overview
3D model overview
replace with your photo
2019–2020 model · LOD 3.2
2023 LiDAR scan model
2023 LiDAR scan model
replace with your photo
2023 model · LOD 2.1
Attribute2019–2020 model2023 model
SourceVirtual Campus projectGPS Lands LiDAR scan
File format.STL.OBJ
File size431 MB~5 MB
Level of detailLOD 3.2 — high detailLOD 2.1 — simplified
CoverageKent Ridge + UTownKent Ridge + UTown
LimitationsMissing newer buildings (Techno Edge, E7, COM3, S9)Missing latest buildings (Valour House, retrofitted YIH)

Ref: BEAM Phase 1 Report §4.3.1, Table 4

Stream 02
Greenery inventory

A comprehensive vegetation model built from three separate data sources — each capturing a different layer of the campus's green cover, from individual roadside trees down to forest canopies and ground-level shrubs.

01
GIS shapefile → 3D trees

Tree inventory shapefile from UCI — containing species, height, canopy spread, and trunk girth — imported into Grasshopper/Rhino via the Urbano plugin. Each tree point is matched to a species in the BIM vegetation library and generated as a full 3D geometry.

Roadside trees
02
LiDAR point cloud → understory

Campus-wide laser-scan point cloud clipped beneath tree canopies to isolate shrub and turf layers. Outlines are extracted and species assigned parametrically — capturing vegetation that satellite imagery or manual surveys cannot detect beneath the canopy.

Shrubs + turf
03
Satellite DEM → forests

Maxar satellite DEM at 0.5 m resolution used to model dense forest areas and groves. Greenery blocks follow forest boundary and height profiles; trees inside are generated randomly but matched to surrounding species for ecological accuracy.

Forest areas
3D Rhino greenery model
3D Rhino model
replace with your photo
3D Rhino model
GIS greenery layer
GIS layer view
replace with your photo
GIS model
ENVI-Met model
ENVI-Met conversion
replace with your photo
ENVI-Met model
GIS synchronisation
Rhino ↔ ArcGIS Pro

All Rhino-generated models — trees, shrubs, turf, buildings, and roads — are imported into ArcGIS Pro maintaining accurate geospatial coordinates. The 3D and GIS models remain synchronised, enabling diverse spatial analyses and correlations with weather station data.

Simulation conversion
ENVI-Met ready

The 3D greenery model can be converted to ENVI-Met for microclimate simulations. Each tree is represented as pixel cubes incorporating Leaf Area Index (LAI), albedo, emissivity, and transmittance. Terrain data ensures buildings and greenery follow actual ground contours.

Ref: Lu et al. (2025) Journal of Digital Landscape Architecture · BEAM Phase 1 Report §4.4

Stream 04
GIS layers

Nine independently toggleable geospatial layers form the 2D environment of the DT — sourced from the greenery database, 2019–2020 Virtual Campus GIS, and satellite-derived data. Each layer is served as a GeoJSON file from S3 and rendered by deck.gl.

GIS layers in Digital Twin
GIS layers screenshot from DT
replace with your photo
GIS layers view in the Digital Twin
LayerGeometry typeSource
BuildingsPolygon3D model (derived)
Big trees · Small treesPointGreenery database
Tree canopyPolygonGreenery database
ShrubsPolygonGreenery database
TurfPolygonGreenery database
ForestPolygonVirtual Campus 2019–2020
Roads · Pavement · Sports fieldsPolygonVirtual Campus 2019–2020
Administrative zones (1, 2, 3, 4, 7)PolygonVirtual Campus 2019–2020
NUS boundaryPolygonVirtual Campus 2019–2020

Ref: BEAM Phase 1 Report §4.3.2, Table 5