Object Preview

USDA NASS Cropland Data Layer (CDL)

Annual crop-specific land cover classification of the conterminous United States, produced by the USDA National Agricultural Statistics Service (NASS), reformatted into a cloud-optimized, version-controlled Icechunk Zarr store.

One Icechunk repository at v0.1.0.icechunk/ with two groups:

group	variable	years	resolution	grid (y × x)
`30m`	`crop_type (year, y, x) uint8`	2008–2025	30 m	105,432 × 160,171
`10m`	`crop_type (year, y, x) uint8`	2024–2025	10 m	316,295 × 480,509

CRS: EPSG:5070 (USA Contiguous Albers Equal Area Conic, NAD83); coordinates are pixel centres in metres. A CF spatial_ref variable carries the WKT and GeoTransform.
Pixel values are categorical class codes. The full code → class-name and code → color mappings are embedded in the crop_type attributes (class_names, class_colors, CF flag_values/flag_meanings).
0 = Background (also the fill value: outside a given year's classified extent). 81 = Clouds/No Data is a real class within the classified extent.
From product year 2024 the native CDL resolution is 10 m; the 30 m product for 2024+ is NASS's resampling of the 10 m CDL, included for continuity with 2008–2023.
Per-year provenance (source zip URL, SHA-256, placement) is stored in group attributes and icechunk commit metadata; each year is tagged (e.g. 30m-2025).
Storage layout: zarr v3 sharded arrays with (1, 512, 512) inner chunks — a field-scale or point query fetches only ~tens of KB per year via range requests.
Multiscale overviews (zarr multiscales convention): each group carries a 2×-per-level pyramid in child groups (10m/2x … 10m/512x, 30m/2x … 30m/256x), block-majority (mode) resampled from the native array. For a quick national map, read a coarse level (e.g. group="30m/32x" ≈ 960 m pixels) instead of the full-resolution array.

Reading the data

Select an area of interest in projected coordinates (or transform lon/lat with pyproj), then plot it with the official class colors embedded in the attributes:

License

US Public Domain. The USDA NASS Cropland Data Layer has no copyright restrictions, is considered public domain, and is free to redistribute. USDA NASS asks for acknowledgement when the data are used. This README and the store metadata are provided under the same terms.

Citation

United States Department of Agriculture (USDA) National Agricultural Statistics Service (NASS), 20260227, Cropland Data Layer: USDA NASS, USDA NASS Marketing and Information Services Office, Washington, D.C. Online Links: https://croplandcros.scinet.usda.gov/

Provenance & processing

Source files are the official national GeoTIFF zips from NASS Research and Science. Pixel values are bit-identical to the source rasters (validated by sampled comparison at ingest).

Processing code (MIT licensed): https://github.com/colinahill/usda_cropland_data

Contact (data producer): USDA NASS Spatial Analysis Research Section, SM.NASS.RDD.GIB@usda.gov · CDL FAQ

USDA NASS Cropland Data Layer (CDL)

One Icechunk repository at v0.1.0.icechunk/ with two groups:

group	variable	years	resolution	grid (y × x)
`30m`	`crop_type (year, y, x) uint8`	2008–2025	30 m	105,432 × 160,171
`10m`	`crop_type (year, y, x) uint8`	2024–2025	10 m	316,295 × 480,509

CRS: EPSG:5070 (USA Contiguous Albers Equal Area Conic, NAD83); coordinates are pixel centres in metres. A CF spatial_ref variable carries the WKT and GeoTransform.
Pixel values are categorical class codes. The full code → class-name and code → color mappings are embedded in the crop_type attributes (class_names, class_colors, CF flag_values/flag_meanings).
0 = Background (also the fill value: outside a given year's classified extent). 81 = Clouds/No Data is a real class within the classified extent.
From product year 2024 the native CDL resolution is 10 m; the 30 m product for 2024+ is NASS's resampling of the 10 m CDL, included for continuity with 2008–2023.
Per-year provenance (source zip URL, SHA-256, placement) is stored in group attributes and icechunk commit metadata; each year is tagged (e.g. 30m-2025).
Storage layout: zarr v3 sharded arrays with (1, 512, 512) inner chunks — a field-scale or point query fetches only ~tens of KB per year via range requests.
Multiscale overviews (zarr multiscales convention): each group carries a 2×-per-level pyramid in child groups (10m/2x … 10m/512x, 30m/2x … 30m/256x), block-majority (mode) resampled from the native array. For a quick national map, read a coarse level (e.g. group="30m/32x" ≈ 960 m pixels) instead of the full-resolution array.

Reading the data

Select an area of interest in projected coordinates (or transform lon/lat with pyproj), then plot it with the official class colors embedded in the attributes:

License

Citation

Provenance & processing

Source files are the official national GeoTIFF zips from NASS Research and Science. Pixel values are bit-identical to the source rasters (validated by sampled comparison at ingest).

Processing code (MIT licensed): https://github.com/colinahill/usda_cropland_data

Contact (data producer): USDA NASS Spatial Analysis Research Section, SM.NASS.RDD.GIB@usda.gov · CDL FAQ

1import numpy as np
2import matplotlib.pyplot as plt
3from matplotlib.colors import BoundaryNorm, ListedColormap
4from matplotlib.patches import Patch
5from pyproj import Transformer
6
7# 10 km x 10 km window around a point in Iowa (lon/lat -> Albers EPSG:5070)
8x, y = Transformer.from_crs("EPSG:4326", "EPSG:5070", always_xy=True).transform(-93.45, 42.15)
9aoi = ds["crop_type"].sel(year=2025, x=slice(x - 5000, x + 5000), y=slice(y + 5000, y - 5000))
10
11# class_names / class_colors attrs map each integer code to its name and hex color
12names, colors = aoi.attrs["class_names"], aoi.attrs["class_colors"]
13
14# build the colormap from just the classes present in the AOI (skip Background = 0)
15present = np.unique(aoi)
16cmap = ListedColormap([colors[str(c)] for c in present])
17cmap.set_under(colors["0"])  # Background
18norm = BoundaryNorm([present[0] - 0.5, *(c + 0.5 for c in present)], cmap.N)
19
20fig, ax = plt.subplots(figsize=(9, 7))
21aoi.plot(ax=ax, cmap=cmap, norm=norm, add_colorbar=False)
22ax.legend(
23    handles=[Patch(color=colors[str(c)], label=names[str(c)]) for c in present],
24    loc="center left", bbox_to_anchor=(1.02, 0.5), fontsize=8,
25)
26ax.set_title("USDA Cropland Data Layer 2025")
27fig.tight_layout()

1import numpy as np
2import matplotlib.pyplot as plt
3from matplotlib.colors import BoundaryNorm, ListedColormap
4from matplotlib.patches import Patch
5from pyproj import Transformer
6
7# 10 km x 10 km window around a point in Iowa (lon/lat -> Albers EPSG:5070)
8x, y = Transformer.from_crs("EPSG:4326", "EPSG:5070", always_xy=True).transform(-93.45, 42.15)
9aoi = ds["crop_type"].sel(year=2025, x=slice(x - 5000, x + 5000), y=slice(y + 5000, y - 5000))
10
11# class_names / class_colors attrs map each integer code to its name and hex color
12names, colors = aoi.attrs["class_names"], aoi.attrs["class_colors"]
13
14# build the colormap from just the classes present in the AOI (skip Background = 0)
15present = np.unique(aoi)
16cmap = ListedColormap([colors[str(c)] for c in present])
17cmap.set_under(colors["0"])  # Background
18norm = BoundaryNorm([present[0] - 0.5, *(c + 0.5 for c in present)], cmap.N)
19
20fig, ax = plt.subplots(figsize=(9, 7))
21aoi.plot(ax=ax, cmap=cmap, norm=norm, add_colorbar=False)
22ax.legend(
23    handles=[Patch(color=colors[str(c)], label=names[str(c)]) for c in present],
24    loc="center left", bbox_to_anchor=(1.02, 0.5), fontsize=8,
25)
26ax.set_title("USDA Cropland Data Layer 2025")
27fig.tight_layout()

USDA NASS Cropland Data Layer (CDL)

USDA NASS Cropland Data Layer (CDL)

USDA NASS Cropland Data Layer (CDL)

Contents

Reading the data

License

Citation

Provenance & processing

USDA NASS Cropland Data Layer (CDL)

Contents

Reading the data

License

Citation

Provenance & processing