Skip to content

NEXRAD radar — usage#

Install#

pip install earthlens[radar]      # boto3 (unsigned S3); no credentials needed

pyart (for reading/gridding the assembled .ar2v volumes) is not required to fetch — it's a downstream follow-on.

Request#

variables maps WSR-88D site id → an (advisory) moment list — a Level-II volume carries every moment, so the whole volume is fetched and the list value is informational:

from earthlens.earthlens import EarthLens

lens = EarthLens(
    data_source="radar",                    # alias: "nexrad"
    variables={"KTLX": ["reflectivity"]},   # Oklahoma City
    start="2024-06-01T18:00:00",            # ISO datetime (sub-hourly feed)
    end="2024-06-01T19:00:00",
    lat_lim=[33, 37],
    lon_lim=[-100, -95],
    path="out/radar",
)
inventory = lens.download()   # GeoDataFrame of assembled volumes
  • Stations — keys of variables. Any valid four-letter id works; the bundled station catalog provides geometry + bbox discovery for the full ~210-site WSR-88D network (a site absent from the catalog still fetches, just with no point geometry). StationCatalog().in_bbox(w, s, e, n) lists the catalogued radars in a box.
  • Time window — start / end filter volumes by scan-start time. Because the feed is a rolling buffer, the window must be recent (~last hour); older windows return an empty inventory.
  • fmt — defaults to "%Y-%m-%dT%H:%M:%S" (the feed is sub-hourly).

Output#

download() returns a GeoDataFrame with one row per assembled volume:

column meaning
station_id WSR-88D site id
volume rolling volume number
scan_time volume scan-start (UTC)
n_chunks chunks concatenated
path local .ar2v file
geometry station point (EPSG:4326), or None if uncatalogued

The assembled .ar2v files are written to path; read them with pyart.io.read_nexrad_archive(...).

aggregate= is not supported (raw polar volumes aren't griddable by the pyramids reducer — the facade rejects it for this vector backend). A failed volume is skipped with a warning rather than aborting the batch.