data/schema.json

{
  "description": "Raw little-endian tensor members inside each WebDataset .tar shard. Each sample is identified by a zero-padded decimal key. Reshape bytes using numpy: np.frombuffer(data, dtype=dtype).reshape(shape).",
  "members": {
    "s2_spring.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128,
        10
      ],
      "band_names": [
        "B2",
        "B3",
        "B4",
        "B8",
        "B5",
        "B6",
        "B7",
        "B8A",
        "B11",
        "B12"
      ],
      "description": "Sentinel-2 L2A 10-band seasonal median composite (spring 2025)",
      "units": "DN (divide by 10000 to get reflectance [0,1])"
    },
    "s2_summer.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128,
        10
      ],
      "band_names": [
        "B2",
        "B3",
        "B4",
        "B8",
        "B5",
        "B6",
        "B7",
        "B8A",
        "B11",
        "B12"
      ],
      "description": "Sentinel-2 L2A 10-band seasonal median composite (summer 2025)",
      "units": "DN (divide by 10000 to get reflectance [0,1])"
    },
    "s2_autumn.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128,
        10
      ],
      "band_names": [
        "B2",
        "B3",
        "B4",
        "B8",
        "B5",
        "B6",
        "B7",
        "B8A",
        "B11",
        "B12"
      ],
      "description": "Sentinel-2 L2A 10-band seasonal median composite (autumn 2025)",
      "units": "DN (divide by 10000 to get reflectance [0,1])"
    },
    "s2_winter.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128,
        10
      ],
      "band_names": [
        "B2",
        "B3",
        "B4",
        "B8",
        "B5",
        "B6",
        "B7",
        "B8A",
        "B11",
        "B12"
      ],
      "description": "Sentinel-2 L2A 10-band seasonal median composite (winter 2025)",
      "units": "DN (divide by 10000 to get reflectance [0,1])"
    },
    "s1_spring.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128,
        2
      ],
      "band_names": [
        "VV",
        "VH"
      ],
      "description": "Sentinel-1 GRD linear gamma-0 (spring 2025)",
      "units": "linear gamma-0 (not dB); typical range 0-1"
    },
    "s1_summer.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128,
        2
      ],
      "band_names": [
        "VV",
        "VH"
      ],
      "description": "Sentinel-1 GRD linear gamma-0 (summer 2025)",
      "units": "linear gamma-0 (not dB); typical range 0-1"
    },
    "s1_autumn.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128,
        2
      ],
      "band_names": [
        "VV",
        "VH"
      ],
      "description": "Sentinel-1 GRD linear gamma-0 (autumn 2025)",
      "units": "linear gamma-0 (not dB); typical range 0-1"
    },
    "s1_winter.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128,
        2
      ],
      "band_names": [
        "VV",
        "VH"
      ],
      "description": "Sentinel-1 GRD linear gamma-0 (winter 2025)",
      "units": "linear gamma-0 (not dB); typical range 0-1"
    },
    "tree_species.u8": {
      "dtype": "uint8",
      "shape": [
        128,
        128
      ],
      "description": "Tree species classification class IDs (0 = background/no-data)"
    },
    "mean_height.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128
      ],
      "description": "Mean tree height per 10 m cell (m); NaN = no inventory trees"
    },
    "median_height.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128
      ],
      "description": "Median tree height per 10 m cell (m)"
    },
    "height_variance.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128
      ],
      "description": "Variance of tree height per 10 m cell"
    },
    "tree_count.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128
      ],
      "description": "Number of inventory trees per 10 m cell"
    },
    "tree_density.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128
      ],
      "description": "Sum of tree_count in 3x3 neighbourhood"
    },
    "tree_count_variance.f32": {
      "dtype": "float32",
      "shape": [
        128,
        128
      ],
      "description": "Spatial variance of tree count per cell"
    },
    "dop20_rgb.u8": {
      "dtype": "uint8",
      "shape": [
        6400,
        6400,
        3
      ],
      "band_names": [
        "R",
        "G",
        "B"
      ],
      "description": "DOP20 true-colour aerial orthophoto at 20 cm GSD. Pixel extent is 6400\u00d76400 = 1280 m \u00d7 1280 m (co-registered with the 128\u00d7128 10 m Sentinel grid). Only present when INCLUDE_DOP20=1 and dop20_available=True for the patch.",
      "units": "uint8 [0, 255]",
      "optional": true
    }
  },
  "meta_member": ".json",
  "crs": "EPSG:25832",
  "patch_size_px": 128,
  "resolution_m": 10,
  "dop20_patch_size_px": 6400,
  "dop20_resolution_m": 0.2
}