3D Lunar Surface Modeling from Chandrayaan-2 TMC DEM Data

Overview

This project turns Chandrayaan-2 Terrain Mapping Camera (TMC) digital elevation models into explorable, true-scale lunar terrain in Unreal Engine 5. Each DEM tile is converted into a 16-bit heightmap, the Moon's curvature is added relative to the tile centre, and a null-value mask hides gaps in the data. The terrain is then lit with an adjustable sun (azimuth, elevation and intensity) and rendered with Lumen, Nanite and World Partition, so the same pipeline serves visual inspection, mission planning and training scenarios.

Grey lunar terrain rendered in Unreal Engine 5, showing craters and ridges lit at a low sun angle — **Figure 1:** UE5 Recreation of Chandrayaan-2 TMC Tile `ch2_tmc_ndn_20220614T1810117259_d_dtm_d32`

Keywords: Chandrayaan-2, TMC, Digital Elevation Model, Lunar Terrain, Heightmap, Unreal Engine 5, Lumen, Nanite, World Partition, QGIS

Highlights

< 10 sper DEM tile, from import to heightmap, with the automation plugin

16-bitheightmaps: 65,536 height levels instead of 256 at 8-bit

10 mtypical TMC spatial resolution, used to scale the terrain in UE5

FinalistISRO Immersion Startup Challenge, AI for Space and Geospatial Innovation

Method

Data acquisition: TMC DEM tiles (.tif heightmaps) for a lunar patch are downloaded from the ISSDC Chandrayaan-2 map browser.
Preparation in QGIS: multiple tiles are stitched by latitude and longitude, kept to a single projection (polar-stereographic and equirectangular tiles cannot be mixed), cropped and exported as one .tif.
Heightmap conversion: in Python, the DEM is read with GDAL, converted to 16-bit integers and saved as a single-channel .png heightmap.
Curvature: each pixel's latitude and longitude is converted to x, y, z, its height is measured relative to the normal at the central pixel, shifted so the borders sit at the base, and added to the original heightmap.
Null mask: an 8-bit binary mask marks valid DEM values as 255 and null values as 0.
Unreal Engine 5: the curved heightmap is imported as a landscape, scaled with x = y = spatial resolution and z = (max height − min height) / 512 (1 UE unit = 1 cm), and the mask is applied as a material to hide null areas.
Automation: a custom UE plugin takes the heightmap and mask as inputs and builds the terrain, processing each tile in under 10 s.

Flat-base heightmap plus lunar curvature equals curved heightmap, imported into UE5, then a mask removes null areas — **Figure 2:** Heightmap with Flat Base + Lunar Curvature = Curved Heightmap, Imported into UE5; the Null-Value Mask Then Removes the Invalid Boundary (Red)

Rendering

The sun is a directional light whose azimuth, elevation and intensity can be changed live, with a Milky Way sky sphere for background illumination. Low sun elevations bring out crater rims and ridges, which matters for landing-site and traverse analysis.

The same lunar tile rendered under three sun settings — **Figure 3:** One Tile Under Three Sun Settings. (Left to Right) Azimuth 270°, Elevation 10°, 2.5 lux; Azimuth 270°, Elevation 15°, 2.5 lux; Azimuth 180°, Elevation 15°, 1 lux

Lumen: dynamic global illumination with diffuse bounce lighting and soft shadows, using software ray tracing so it runs without ray-tracing hardware.
Nanite: virtualized geometry that switches to coarser triangles when zoomed out and keeps pixel-scale detail up close, with no manual LOD setup.
World Partition: streams only the visible cells of large terrains, with hierarchical LOD for distant areas.
Virtual Shadow Maps and Large World Coordinates: long shadow distances and 64-bit positions for accurate large-scale scenes.

Lunar tile rendered with Lumen disabled (top) and enabled (bottom) — **Figure 4:** Lumen Disabled (Top) and Enabled (Bottom)

Two wireframe views of the lunar mesh, zoomed out and zoomed in, comparing Nanite (blue) and traditional (orange) meshes — **Figure 5:** Lunar Surface Wireframes, Zoomed Out (Left) and Zoomed In (Right). Nanite (Blue) Coarsens Distant Geometry; the Traditional Mesh (Orange) Keeps Full Detail

Top-down lunar terrain divided into a grid of World Partition cells — **Figure 6:** World Partition Cells (Orange) Streamed over the Lunar Terrain

Results

Verification

The UE5 terrain was checked against the ISSDC preview and against hillshade and contour renderings of the same DEM in QGIS. Contours, peaks and craters line up across all four views. The approach follows the one NASA used to recreate the Apollo 17 landing site in Unreal Engine from Lunar Reconnaissance Orbiter data.

Four vertical strips of the same lunar tile: ISSDC snapshot, QGIS hillshade, QGIS contours and the UE5 recreation — **Figure 7:** (Left to Right) ISSDC Snapshot, QGIS Hillshade, QGIS Contours and Our UE5 Recreation of the Same Tile

Apollo 17 lunar module photograph beside NASA's Unreal Engine recreation — **Figure 8:** Reference: Apollo 17 Photograph (Left) and NASA's UE5 Recreation from LRO Data (Right). Images: NASA

Different Lunar Patches

Lunar terrain at true scale with a human mannequin for size reference — **Figure 9:** Lunar Surface at True Scale with a Mannequin for Reference

Three further lunar tiles rendered at 1:100 scale, with zoomed-in views of highlighted regions — **Figure 10:** Three More TMC Tiles at 1:100 Scale, with Zoomed-in Views of the Regions in Red. (Left to Right) `…20200121T0825168134_d_dtm_g26`, `…20211221T1917095490_d_dtm_d32`, `…20210513T1401073872_d_dtm_hw1`

Applications

Mission planning: terrain and lighting context for landing sites and surface operations.
Rover navigation: traverse planning that accounts for energy, lighting and communication line of sight.
Vehicle design: flying new vehicle concepts in simulation to evaluate safety and operability.
Training and outreach: immersive, flight-like environments for astronauts and education.

Tools: Unreal Engine 5, QGIS, Python (GDAL, NumPy, PyPDS, Pillow)