CrocoLake

CrocLakeSource reads a local CrocoLake parquet database — ocean in-situ observations from ARGO, GLODAP and SprayGliders — and writes the observations in each window as a DART obs_seq file.

Usage

import datetime
from dartobsgen import ObsGenConfig, CrocLakeSource, generate_obs_sequences

config = ObsGenConfig(
    start=datetime.datetime(2010, 5, 1),
    end=datetime.datetime(2010, 5, 3),
    lat_min=5,   lat_max=60,
    lon_min=-100, lon_max=-30,
    obs_types=["ARGO_TEMPERATURE", "ARGO_SALINITY"],
    assimilation_frequency=datetime.timedelta(hours=6),
    output_dir="./obs_output",
)

source = CrocLakeSource(
    crocolake_path="/path/to/crocolake/",
    dart_path="/path/to/DART/",
)

if __name__ == "__main__":
    written = generate_obs_sequences(config, source)
    print(written)

dart_path is the root of a DART clone; it is used to resolve the DART obs type definitions. crocolake_path is the directory holding the parquet database.

Because observations exist independently of any model run here, start is the natural way to specify the time range — pick the time you initialize the model and the first analysis lands one assimilation_frequency later. See Time windows.

Observation types

obs_types accepts DART compound names, DART variable names, or CrocoLake variable names, freely mixed. The full table of supported types and the obs_type_map override are in Observation types.

Parallelism

Each worker process opens the parquet database independently and writes its own output file, so windows parallelize cleanly — see Parallel generation.