Wind Atlas quickstart — windowed read (live)¶

Download the mean wind speed at 100 m for a small bounding box from the Global Wind Atlas and read it back. The layer is a 13.9 GB global Cloud-Optimized GeoTIFF, but the backend reads only the bbox window over /vsicurl/, so a small area returns in seconds — no whole-file download.

Runs live against public figshare (no credentials).

In [ ]:

from earthlens.core import EarthLens

from earthlens.core import EarthLens

Download one windowed subset¶

A ~0.5° box near DTU (Denmark). download() returns the written GeoTIFF path(s).

In [ ]:

paths = EarthLens(
    data_source="solar-wind-atlas",
    variables=["wind_100m"],
    lat_lim=[55.0, 55.5],
    lon_lim=[12.0, 12.5],
    path="swa_out",
).download(progress_bar=False)
paths

paths = EarthLens( data_source="solar-wind-atlas", variables=["wind_100m"], lat_lim=[55.0, 55.5], lon_lim=[12.0, 12.5], path="swa_out", ).download(progress_bar=False) paths

Read the result back with pyramids¶

The output is an EPSG:4326 GeoTIFF of mean wind speed in m/s.

In [ ]:

import numpy as np
from pyramids.dataset import Dataset

ds = Dataset.read_file(str(paths[0]))
arr = np.asarray(ds.read_array(), dtype='float64')
finite = arr[np.isfinite(arr)]
print('grid     :', ds.columns, 'x', ds.rows, 'EPSG', ds.epsg)
print(
    f'wind m/s : min {finite.min():.2f}  mean {finite.mean():.2f}  max {finite.max():.2f}'
)

import numpy as np from pyramids.dataset import Dataset ds = Dataset.read_file(str(paths[0])) arr = np.asarray(ds.read_array(), dtype='float64') finite = arr[np.isfinite(arr)] print('grid :', ds.columns, 'x', ds.rows, 'EPSG', ds.epsg) print( f'wind m/s : min {finite.min():.2f} mean {finite.mean():.2f} max {finite.max():.2f}' )

Why this is fast¶

download() opens the remote COG over /vsicurl/ and reads only the tiles overlapping the bbox via pyramids Dataset.crop(bbox=) — a few hundred KB, not the 13.9 GB global file. The Global Solar Atlas layers (ghi, dni, …) instead download their full ~2.7 GB ZIP once and crop locally, since they are not served as range-accessible COGs.