goosedb

DuckDB · Parquet
DashboardThe files

9.2 GB · 1750–2026

Climate observations

Live over NOAA's Global Historical Climatology Network — a 9.20 GB public Parquet dataset on S3, read in place with HTTP range requests. Nothing is downloaded or copied.

Public role · 9 columns visible · 1 withheld
14,031 files · 1750–2026

Selects a partition

Selects partitions

Grain

~14 MB
per request

Observations

3.9M

2025–2025, TMAX

Mean

16.92 °C

across every matching reading

Range

-75.5 – 51.6

°C, lowest to highest recorded

95th percentile

34.4 °C

across every matching reading

Maximum temperature over time

Grouped on a substring of the raw YYYYMMDD column rather than a parsed timestamp — the same answer, measured 3.7× faster over two million rows.

Seasonality

Month of the year across the whole range, ignoring which year each reading fell in — another substring of the same column.

Scanning Parquet

By country

Grouped on a code taken from the station ID — GHCN's own FIPS encoding — so this is one scan with nothing joined to it. The name is filled in from the catalog.

Scanning Parquet

Scanned Parquet
2.40 s
scan 761.3 ms
bytes resident
cold start
12 rows
policy: public

Read 13.4 MB from 8 of 14,031 files · 9.20 GB dataset · fetched once, then read from memory

0.1456% of the dataset