nasa-ibm-ai4science/Sombench-pretraining-data

SomBench Pre-training Corpus: Multimodal Lunar Tiles

Dataset Summary

This includes a small sample from SomBench: a corpus of co-registered, multimodal lunar image tiles built for large-scale self-supervised (foundation-model) pre-training. It contains a subset of modalities from the low-resolution (WAC-anchored) and high-resolution (NAC-anchored) tracks specifically used in pretraining.

Tiles are anchored to individual LROC Experiment Data Record (EDR) image canvases rather than to a fixed map grid. A sliding window enumerates fixed-shape 512 × 512-pixel patches over each EDR, and every co-registered modality is snapped to that same per-tile boundary. Each tile therefore bundles many physically distinct measurements (optical, topographic, spectral, radar, thermal, gravity) over the same ground patch, covering the same spatial bounds. Anchoring to image canvases also preserves the optical observation context, and overlapping EDR coverage exposes models to the same terrain under different illumination conditions.

The corpus is organized into two parallel tracks that share construction logic but differ in their optical anchor and ground scale:

• WAC_LowRes/: anchored to LROC Wide-Angle Camera (WAC) visible imagery at 100 m/pixel; each tile covers ≈ 51.2 × 51.2 km.
• NAC_HighRes/: anchored to LROC Narrow-Angle Camera (NAC) imagery at 1 m/pixel; each tile covers ≈ 512 × 512 m.

Both tracks use a 512 × 512-pixel tile size for the anchor modality (WAC or NAC). Coarser modalities are resampled, while preserving native resolution as much as possible, to align to the exact bounds of the anchor modality.

Note: The full dataset is available on AWS at s3://nasa-lunar-fm-bench/ and can be accessed using the AWS CLI:

aws s3 ls s3://nasa-lunar-fm-bench/ --no-sign-request

Supported Tasks and Applications

• Self-supervised / foundation-model pre-training: masked image modeling and multimodal SSL over co-registered layers.
• Multimodal representation learning and cross-modal fusion: joint embeddings across optical, topographic, spectral, radar, thermal, and gravity modalities.
• Multi-scale learning: pairing meter-scale NAC context with hundreds-of-meters-to-kilometer static layers within one framework.
• Downstream fine-tuning: the test split is reserved for evaluation and fine-tuning on the SomBench application benchmarks (crater detection, IMP segmentation, ice prospectivity).

Tracks

Property	WAC_LowRes/	NAC_HighRes/
Optical anchor	WAC VIS (100 m/px)	NAC (1 m/px)
Tile ground extent	51.2 km × 51.2 km	512 m × 512 m
Tile size (pixels)	512 × 512	512 × 512
Modality directories	~29	~31
Source EDR products	54,080 WAC EDRs	1,095 NAC EDRs
Zones covered (LTM_CODE)	92 (90 LTM + LPS_N/S)	80
Tiles	963,609	1,000,113
Dataset size	38 TB	1.4 TB

Directory Layout

Sombench-pretraining-data/
├── WAC_LowRes/            # low-resolution (WAC-anchored)
    ├── WAC_LowRes.parquet
    ├── aspect/
    ├── dtm/
    ├── metadata/
    ├── slope/
    ├── uv/
    └── vis/
├── NAC_HighRes/           # high-resolution (NAC-anchored)
    ├── NAC_HighRes.parquet
    ├── aspect_3m/
    ├── dtm_3m/
    ├── metadata/
    ├── nac/
    └── slope_3m/

The .parquet file in each track has a row for each tile and includes the paths to the paired modalities (e.g. DTM_TILE), metadata associated with the EDR image (e.g. INCIDENCE_ANGLE), and DATASET assignment for training.

Within each modality sub-directory, tiles are stored as one netCDF file per patch, named by the source EDR product id and the sliding-window row/column:

{product_id}_r{row}_c{col}.nc

Modalities

Each modality lives in its own sub-directory (the lowercase "dataset key"). The table lists source, native resolution, and band content. Availability differs by track and by region: some layers exist only at the poles, others only outside the poles. Band counts are structural (from the SomBench paper); native resolutions are the upstream product resolutions before per-tile snapping.

NOTE: This table includes all modalities that are included in SomBench; however, the HuggingFace sample inlcudes only modalities used in pre-training, including 6 WAC_LowRes modalities (vis, uv, dtm, aspect, slope, metadata) and 5 NAC_HighRes modalities (nac, dtm, aspect, slope, metadata).

Directory (low / high)	Modality	Source / instrument	Latitude	Native res	Bands	Pixel value units
nac (high only)	NAC panchromatic imagery; high-res anchor	LRO LROC NAC	90°S–90°N (sparse, non-uniform)	1 m	1	I/F: unitless
vis (low only)	WAC visible reflectance (415, 566, 604, 643, 689 nm); low-res anchor	LRO LROC WAC	90°S–90°N	100 m	5	I/F: unitless
uv (low only)	WAC ultraviolet (321, 360 nm)	LRO LROC WAC	90°S–90°N	500 m	2	I/F: unitless
wac_mosaic	WAC global morphologic mosaic (643 nm)	LRO LROC WAC	90°S–90°N	100 m	1	I/F: unitless
wac_norm_ref	WAC normalized multi-band reflectance	LRO LROC WAC	60°S–60°N	500 m	7	I/F: unitless
wac_nr_643_hr	WAC 643 nm high-resolution normalized reflectance	LRO LROC WAC	90°S–90°N	100 m	1	I/F: unitless
tio2	WAC-derived TiO₂ abundance	LRO LROC WAC	70°S–70°N	400 m	1	wt%
dtm / dtm_60m	SLDEM2015 elevation (topography)	LOLA + Kaguya TC	90°S–90°N	60 m	1	meters
slope / slope_60m	SLDEM2015 slope	LOLA + Kaguya TC	90°S–90°N	60 m	1	degrees
aspect / aspect_60m	SLDEM2015 aspect (downslope geographic azimuth, 0/360°: north-facing, 90°: east-facing; sine/cosine encoded)	LOLA + Kaguya TC	90°S–90°N	60 m	2	aspect: degrees, sine/cosine: unitless
dtm_3m (high only)	NAC-stereo DTM elevation	USGS NAC DTM	90°S–90°N (sparse, non-uniform)	3 m	1	meters
slope_3m (high only)	NAC-stereo DTM slope	USGS NAC DTM	90°S–90°N (sparse, non-uniform)	3 m	1	degrees
aspect_3m (high only)	NAC-stereo DTM aspect (downslope geometric azimuth, 0/360°: faces map-up, 90°: faces map-right; sine/cosine encoded)	USGS NAC DTM	90°S–90°N (sparse, non-uniform)	3 m	2	aspect: degrees, sine/cosine: unitless
roughness	LOLA roughness (50 m baseline)	LRO LOLA	90°S–90°N	1 km	1	meters
geomap	USGS Unified Geologic Map (categorical, 43 units)	USGS	90°S–90°N	60 m	1	unitless
gravity	GRAIL free-air gravity disturbance	GRAIL	90°S–90°N	20 km	1	milligal
mi_norm_ref	Kaguya MI normalized reflectance (VIS–NIR)	SELENE/Kaguya MI	55°S–55°N	60 m	8	I/F: unitless
mi_mineralogy	Kaguya MI mineralogy (olivine, OPX, CPX, plagioclase, FeO, plag. grain size, OMAT)	SELENE/Kaguya MI	55°S–55°N	60 m	7	OMAT: unitless, grain size: microns, FeO: wt%, all others: wt%/100
sw_fe	Kaguya MI space-weathering Fe (smFe, mpFe, npFe)	SELENE/Kaguya MI	55°S–55°N	1 km	3	wt%
sp_mineralogy	Kaguya SP polar mineralogy (olivine, plag., HCP, LCP, FeO, npFe, OMAT)	SELENE/Kaguya SP	80–90°N; 80–90°S	1 km	7	OMAT: unitless, FeO: wt%, all others: wt%/100
minirf_s1	Mini-RF radar reflectivity (S-band)	LRO Mini-RF	90°S–90°N	90 m	1	decibels
minirf_cpr	Mini-RF circular polarization ratio (CPR)	LRO Mini-RF	90°S–90°N	90 m	1	unitless
treg	Diviner nighttime regolith temperature anomaly	LRO Diviner	70°S–70°N	240 m	1	Kelvin
hpar	Diviner H-parameter (regolith density scale height)	LRO Diviner	70°S–70°N	240 m	1	meters
rock_abundance	Diviner rock abundance (areal fraction of m-scale rocks)	LRO Diviner	70°S–70°N	240 m	1	unitless
tbol	Diviner bolometric temperature; 24 sub-solar-longitude phases (15° steps) + closest	LRO Diviner	90°S–90°N	15 km	25	Kelvin
tbol_poles	Diviner polar bolometric temperature; 24 phases × summer/winter + closest	LRO Diviner	80–90°N; 80–90°S	240 m	49	Kelvin
dice	Diviner ice stability depth	LRO Diviner	80–90°N; 80–90°S	240 m	1	centimeters
psr	LOLA permanently shadowed regions (PSR) mask	LRO LOLA	80–90°N; 80–90°S	20 m	1	unitless
avg_illum	LOLA average Sun illumination (18.6-yr precession cycle)	LRO LOLA	75–90°N; 75–90°S	120 m	1	unitless
albedo	LOLA 1064 nm normal albedo	LRO LOLA	50–90°N; 50–90°S	1 km	1	I/F: unitless
hydrogen	Lunar Prospector hydrogen abundance	LP Neutron Spectrometer	60–90°N; 60–90°S	15 km	1	wt%
metadata	Associated per-tile metadata (illumination geometry, location, …)	LRO LROC NAC/WAC	90°S–90°N	-	-	coordinates & angles: degrees
craters (low only)	Robbins (2019) crater labels, rasterized to the tile grid	Robbins crater catalog	90°S–90°N	-	-	unitless

Regional availability. Most modalities are present in both polar and non-polar tiles wherever data exists. The exceptions: sp_mineralogy, tbol_poles, dice, psr, avg_illum, albedo, and hydrogen are polar-only, while tio2, hpar, and rock_abundance are non-polar-only. A sub-directory may still appear in both tracks even where its tiles are populated for one region only.

Tile File Format

Every tile is a self-describing netCDF4 file (written via h5netcdf with Bitshuffle + LZ4 chunk compression; one chunk per variable):

• Coordinate vectors: 1-D x and y pixel-center coordinates in the tile's CRS.
• Global attributes: full CRS as a WKT string (crs), per-axis pixel resolution (pix_res_x, pix_res_y), and a comma-separated list of band names (band_names).
• Value-range clipping applied at write time from a per-modality registry (e.g. slope ∈ [0°, 90°], normalized reflectance ∈ [0, 1], albedo ∈ [0, 1], TiO₂ ∈ [0, 100] %). NaN encodes a genuine data gap and is preserved through clipping; the categorical geomap layer is exempt so class indices are unchanged.
• Snapped resolution: because coarse modalities are resampled to align with the anchor grid, the stored pixel size can differ slightly from native (e.g. 60 m → ≈ 56.9 m in the NAC track; 100 m → ≈ 102.4 m). The per-tile CRS and resolution are always recorded in the file's attributes.

Multi-band / multi-phase notes

• tbol: all 24 sub-solar-longitude snapshots (15° steps) plus a duplicated "closest" band → 25 bands.
• tbol_poles: 24 phases × summer/winter (48) plus a "closest" band → 49 bands (polar tiles only).
• Multi-band spectral/mineral layers (mi_norm_ref, mi_mineralogy, wac_norm_ref, sp_mineralogy, sw_fe) are stacked along a band dimension; if any band is missing, the whole layer is recorded as missing for that tile.

Catalogs and Metadata

Each track has a Parquet catalog (WAC_LowRes.parquet and NAC_HighRes.parquet) with one row per tile and these columns:

• Identification: PRODUCT_ID (source EDR id), ROW, COL (sliding-window offset), LTM_CODE (LTM zone or LPS_N/LPS_S).
• Per-modality: {MODALITY}_TILE, relative path to the tile's netCDF file (e.g. WAC_VIS_TILE, NAC_TILE, DTM_60M_TILE, TBOL_TILE, …) and {MODALITY}_FRACTION_NULL, the tile's NaN fraction for that layer. (METADATA_TILE, and low-res CRATERS_TILE, have no null-fraction column.)
• Geometry: CENTER_LATITUDE/CENTER_LONGITUDE, corner UPPER_LEFT_* / LOWER_RIGHT_* lon/lat, and projected bounds BOUNDS_XMIN/XMAX/YMIN/YMAX.
• Viewing geometry: INCIDENCE_ANGLE, EMISSION_ANGLE, PHASE_ANGLE, SUB_SOLAR_GROUND_AZIMUTH, SUB_SOLAR_LATITUDE, SUB_SOLAR_LONGITUDE.
• Split & flags: DATASET (train/val/test) and ALLOW_NANS_OPTICAL_DTM_SLOPE_ASPECT, a boolean marking tiles permitted to contain NaN in the optical / DTM / slope / aspect layers (primarily polar tiles admitted with partial data; 4,588 in the low-res track, 3,888 in the high-res track).

Loading on the Hub. load_dataset(...) and the dataset viewer return this catalog (per-tile paths and metadata), not the netCDF imagery. Fetch each tile from the path in its {MODALITY}_TILE column (e.g. via huggingface_hub).

The METADATA_TILE column (and the metadata modality directory) carries associated per-tile metadata (illumination geometry, location, etc.).

Splits

Splits are assigned at the grid-cell (LTM zone) level to prevent spatial leakage:

• Zone partition: unique LTM_CODE grid cells are randomly partitioned into train / validation / test at a target of 75 % / 15 % / 10 % of cells. Whole zones go to a single split, so nearby tiles never straddle splits.
• Buffer exclusion: tiles that straddle two zones are dropped to guarantee strict spatial separation.
• Zone/partition assignments are applied consistently across both tracks.

Released split sizes (DATASET column of the catalogs):

Split	WAC_LowRes	NAC_HighRes
Train	754,529	750,557
Val	115,548	157,153
Test	93,532	92,403
Total	963,609	1,000,113

Train + validation are intended for pre-training; test is reserved for downstream evaluation and fine-tuning. Because zones vary in tile count, the realized tile ratio differs from the 75/15/10 cell ratio.

Coverage and Projections

Tiles span from the equator to ±82° latitude in 90 Lunar Transverse Mercator (LTM) zones plus two Lunar Polar Stereographic (LPS) caps (|φ| ≥ 82°), using WKT definitions from McClernan et al. (2025). Non-polar windows are rejected if the required anchor modality contains any NaN; polar windows are admitted with any valid data to maximize coverage, so polar tiles more often contain partial NaN. Tiles that straddle two zones are dropped, keeping split boundaries clean.

Known Limitations

Heterogeneous spatial resolution, illumination-driven appearance changes, label scarcity/class imbalance, and tens-of-meters absolute geolocation uncertainty all apply. In practice: expect resampling artifacts where coarse modalities are snapped to the anchor bounds; treat NaN as a genuine data gap (especially in polar tiles, which are admitted with partial coverage); and use the ALLOW_NANS_OPTICAL_DTM_SLOPE_ASPECT flag when building NaN-free training sets. See the SomBench paper for the full discussion and recommended practices.

Citations

@article{fraccaro2026lfm,
  title  = {Multimodal-Multiresolution Foundation Model for Lunar Remote Sensing},
  author = {Fraccaro, Paolo and Nyirjesy, Gabby and Szwarcman, Daniela and Patil, Himanshu
            and Gaur, Vishal and Lal, Rohit and Slank, Rachel A. and Dawson, Geoffrey
            and Debary, Hiyam and Dionelis, Nikolaos and Barker, Michael K. and Annex, Andrew
            and Viswanathan, Vishnu and Morse, Zachary and Schaefer, Ethan I. and Kumar, Ankur
            and Watson, Campbell D. and Dawson-Rigas, Rebekah I. and Maskey, Manil
            and Roy, Sujit and Ramachandran, Rahul and Bernab\'e-Moreno, Juan},
  year   = {2026}
  howpublished = {\url{https://huggingface.co/collections/nasa-ibm-ai4science/nasa-ibm-lunar-fm-and-downstream-models}}
}

@misc{sombench2026collection,
  author = {Patil, Himanshu and Nyirjesy, Gabby and Slank, Rachel A. and Gaur, Vishal
          and Szwarcman, Daniela and Fraccaro, Paolo and Dionelis, Nikolaos and Barker, Michael K.
          and Annex, Andrew and Viswanathan, Vishnu and Morse, Zachary and Schaefer, Ethan I.
          and Debary, Hiyam and Kumar, Ankur and Lal, Rohit and Dawson, Geoffrey
          and Watson, Campbell and Dawson-Rigas, Rebekah I. and Maskey, Manil
          and Bernab\'e-Moreno, Juan and Ramachandran, Rahul and Roy, Sujit},
  title        = {{SomBench}: Benchmark Dataset for Advancing Machine Learning in Lunar Science},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/collections/nasa-ibm-ai4science/lunar-fm-ml-ready-benchmark-dataset-sombench}}
}

License

Released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.