dapper.sample_era5_land

dapper.sample_era5_land(domain, start_date, end_date='latest', *, backend='auto', sampling_method='auto', variables='elm', output_dir=None, overwrite=False, keep_downloads=False, arco_max_grid_cells=25, arco_max_estimated_seconds=300.0, cds_client=None, gdrive_folder='dapper_era5_land', job_name=None, gee_scale='native', gee_years_per_task=5, skip_tasks=False)

Sample ERA5-Land through ECMWF ARCO or Google Earth Engine.

Dates use an inclusive start and exclusive output end. ARCO downloads are synchronous and written as GEE-compatible CSV files under output_dir. GEE exports remain asynchronous Drive tasks. backend='auto' chooses ARCO only when every feature fits its service limits and the estimated request is below the configured cell/time thresholds.

ECMWF omits the incomplete 1950-01-01 source day from its ERA5-Land ARCO archive. When ARCO is selected for a request beginning on that date, Dapper warns and begins sampling on 1950-01-02 instead of silently switching the entire request to GEE. Sampling metadata records both dates. Use export_met(..., clip_to_full_years=True) to omit partial boundary years, or select backend='gee' explicitly when 1950-01-01 is required.

For polygon supports, sampling_method='auto' uses a local area-weighted mean of intersecting ERA5-Land cells with ARCO and GEE’s spatial mean with GEE. Points use nearest-cell sampling. Use sampling_method='nearest' to sample polygon representative points instead.

Sampling provenance is attached without changing the input Domain’s model coordinates or weights. GEE geometry-reference coordinates are recorded as sampling_reference_lon and sampling_reference_lat when available.

Return type:

Domain

Parameters:

• domain (Domain)

• backend (str)

• sampling_method (str)

• output_dir (str | Path | None)

• overwrite (bool)

• keep_downloads (bool)

• arco_max_grid_cells (int)

• arco_max_estimated_seconds (float)

• gdrive_folder (str)

• job_name (str | None)

• gee_years_per_task (int)

• skip_tasks (bool)