apac-nwp-forecast-raw: Dataset Card

APAC NWP Forecast — raw per-run parquet (Bronze)

Immutable per-run parquet capture of Asia-Pacific NWP forecasts (JMA MSM, DWD ICON), one file per model run. This is the Bronze layer of the dataset family — the raw export kept exactly as fetched, before it is reshaped into the analysis-ready cube.

Most users want the analysis-ready cube, not this: jimtseng/apac-nwp-forecast (Silver, Zarr Mode A).

Dataset family

Time coverage: this Bronze layer accumulates from the go-forward pipeline onward. Historical runs predate it (the old pipeline discarded the parquet after conversion), so the historical original lives in apac-nwp-forecast-zip. Silver spans both.

Layout

data/model=<model>/year=YYYY/month=MM/<run>.parquet
# e.g. data/model=dwd_icon/year=2026/month=07/20260701T00Z.parquet

Hive-style partitioning (model, year, month). Filenames are the model run initialisation time (UTC), e.g. 20260701T00Z = the 00z run of 2026-07-01.

Schema

Long table: one row per (location_id, time), with grid columns latitude, longitude, elevation, provenance run_init / scraped_at, and the weather variables (irradiance, temperature, humidity, wind, pressure, precipitation, cloud cover, and — dwd_icon only — snow). Integer variables (radiation uint16, cloud/humidity uint8) reflect the source quantization; zstd-compressed. The points form a full regular lat/lon grid.

import pandas as pd
from huggingface_hub import hf_hub_download
p = hf_hub_download("jimtseng/apac-nwp-forecast-raw",
                    "data/model=dwd_icon/year=2026/month=07/20260701T00Z.parquet",
                    repo_type="dataset")
df = pd.read_parquet(p)

License & attribution

Data: CC BY 4.0 — Weather data by Open-Meteo.com, based on open data by JMA (MSM) and DWD (ICON). Please retain this attribution when redistributing or displaying the data.