Datasets:
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
HLS v2 training cube
One NetCDF file, train.nc (about 233 GB). It concatenates 400 HLS S2-opt v4 tiles along pixels, and concatenates years 2013–2025 along time.
| Axis | Size | Meaning |
|---|---|---|
| pixels | 13,395,600 | 400 tiles in ingest order, raster order inside each tile |
| time | 1,586 | 122 steps per year, 2013 through 2025 |
| bands | 13 | B01–B07, B08, B8A, B09–B12 |
| granules | 499,675 | source scenes for the picked observations |
| tiles | 400 | MGRS tile names |
Time labels run from 2013001 to 2025364. Each year is a 3-day pick-one series (001, 004, …, 364), not a median composite. Leap day falls in the last bin.
Values
data is int16 digital number. Reflectance is DN * 1e-4. Fill is -9999. Chunks are (1024, 1586, 13), zlib level 1.
fmask is the original HLS Fmask byte. Fill is 255. Cloudy observations are kept. sensor is 1 for S30 and 2 for L30. Landsat-30 has no red-edge or B08/B8A/B12 slot, so those positions stay -9999.
source_idx is an int32 index into this file's granule table. obs_time is the picked observation as YYYYDDD. lon and lat are the 30 m pixel centers.
This repository is the single concatenated cube. It does not contain the 5,600 yearly and merged tile files that were used to build it.
- Downloads last month
- 49