Catalog > Google WeatherNext 3 forecast, 48 hour, 0.05 degree, virtual > Validation report

Google WeatherNext 3 forecast, 48 hour, 0.05 degree, virtual validation report

This dataset validation report plots a sample of values from the Google WeatherNext 3 forecast, 48 hour, 0.05 degree, virtual dataset over time and across space, comparing where possible to a previously validated reference dataset. It also reports the quantity of unavailable values and their associated timestamps. These analyses are one layer of a multi-layered dataset validation process we perform at dynamical.org and also provide users a preview of the dataset contents.

Report generation start time: 2026-10-31T19:58:32 UTC

Summary

This dataset has been reviewed and is ready for use.

What looks good

Both variables were available at all 48 scanned initializations. The sampled decode check covered 20 of 48 initialization regions and five leads, reporting 200 successfully decoded references.
The plotted fields show consistent geographic orientation and plausible units and ranges.
Trained on point-location station data, this distinct high-resolution model head predicts 0.05° data covering the globe, including over ocean. It is not 0.1° data interpolated to 0.05°.

Review notes

Longitudes use the 0–360 convention, and the horizontal dimensions are named y and x. y holds latitude ascending from south to north and x holds longitude. For example, 100°W is x=260 in this dataset.

Datasets

Role	Name	ID	Version	URL
Validation	Google WeatherNext 3 forecast, 48 hour, 0.05 degree, virtual	`google-weathernext3-forecast-48-hour-0-05-degree-virtual`	`v0.1.0`	`https://google-weathernext3.r2.dynamical.org/google-weathernext3-forecast-48-hour-0-05-degree-virtual/v0.1.0.icechunk`
Reference	NOAA GFS analysis	`noaa-gfs-analysis`	`0.1.0`	`s3://dynamical-noaa-gfs/noaa-gfs-analysis/v0.1.0.icechunk`

Run parameters

Parameter	Value
Validation dataset type	forecast
Validation time range	init_time 2026-06-30T00:00 → 2026-07-01T23:00
Reference time range	time 2021-05-01T00:00 → 2026-10-31T17:00
Point 1	lat=39.7500, lon=255.0000
Point 2	lat=25.0000, lon=300.0000
Spatial comparison time	init=2026-07-01T00:00, lead=24h (reference at 2026-07-02T00:00)
Timeseries period	init=2026-07-01T00:00
Statistic	mean when declared, otherwise standard_deviation
Store type	virtual — availability manifest-probed; value series sampled

Availability

The plot below shows the availability of each variable over time — light red marks unavailable data, dark green marks present data.

availability heatmap

Every variable is complete at every scanned position.

Decode health

200 references decoded successfully, sampled across 20 of 48 append-dim regions, 5 leads and up to 6 present levels per group variable.

Per-variable details

`dew_point_temperature_2m` [statistic=mean]

Metadata

Field	Value
units	`degree_Celsius`
long_name	2 metre dewpoint temperature
short_name	2d
standard_name	dew_point_temperature
step_type	instant
sampled statistic	statistic=mean
comment	WeatherNext 3's station-trained head, defined everywhere.

Point time series — 2026-06-30T00:00 - 2026-07-01T23:00, sampled lead=2h

Point	min	mean	max
P1	-2.64	3.929	9.305
P2	22.37	23.28	23.72

Spatial — init=2026-07-01T00:00, lead=24h (reference not available)

Source	min	mean	max
Validation	-70.11	4.337	29.89

Temporal — init=2026-07-01T00:00

Source	min	mean	max
P1 Validation	1.76	6.78	10.74
P2 Validation	22.37	22.96	23.38

Availability — 48 of 48 positions complete

`temperature_2m` [statistic=mean]