air_temp_c dict | precip_1h_mm dict | hn24_cm dict |
|---|---|---|
{
"variable": "air_temp_c",
"models": [
"hrrr",
"gefs",
"ecmwf_ens"
],
"rows": 380579,
"scheme": "unseen stations, from 2025-10-01",
"stations": 98,
"crps_network": 1.2993698621198109,
"levels": [
0.05,
0.25,
0.5,
0.75,
0.95
],
"crps_hrrr": 2.138313626305047,
"crpss... | {
"variable": "precip_1h_mm",
"models": [
"hrrr",
"gefs",
"ecmwf_ens"
],
"rows": 373680,
"scheme": "unseen stations, from 2025-10-01",
"stations": 97,
"crps_network": 0.12412213458724505,
"levels": [
0.05,
0.25,
0.5,
0.75,
0.95
],
"crps_hrrr": 0.20593761604754213,
"... | {
"variable": "hn24_cm",
"models": [
"hrrr",
"gefs",
"ecmwf_ens"
],
"rows": 171165,
"scheme": "unseen stations, from 2025-10-01",
"stations": 98,
"crps_network": 0.5107894634924134,
"levels": [
0.05,
0.25,
0.5,
0.75,
0.95
],
"crps_hrrr": 2.060570827872952,
"crpss_vs... |
{
"variable": "air_temp_c",
"models": [
"hrrr"
],
"rows": 1431748,
"scheme": "unseen stations, from 2025-10-01",
"stations": 98,
"crps_network": 1.3658678572989464,
"levels": [
0.05,
0.25,
0.5,
0.75,
0.95
],
"crps_hrrr": 2.1288727156411453,
"crpss_vs_hrrr": 0.35840792769632... | {
"variable": "precip_1h_mm",
"models": [
"hrrr"
],
"rows": 1407168,
"scheme": "unseen stations, from 2025-10-01",
"stations": 97,
"crps_network": 0.1366380620949586,
"levels": [
0.05,
0.25,
0.5,
0.75,
0.95
],
"crps_hrrr": 0.2149628749542891,
"crpss_vs_hrrr": 0.364364371643... | {
"variable": "hn24_cm",
"models": [
"hrrr"
],
"rows": 588385,
"scheme": "unseen stations, from 2025-10-01",
"stations": 98,
"crps_network": 0.5576407569297743,
"levels": [
0.05,
0.25,
0.5,
0.75,
0.95
],
"crps_hrrr": 2.0569934692562426,
"crpss_vs_hrrr": 0.7289049453660136,
... |
wxfuser station archive
Paired material for calibrating weather forecasts at mountain stations: station observations, and archived model forecasts extracted at the same points. It is built by wxStationFuser for Tree60 Weather.
snotel/
The hourly record of NRCS SNOTEL stations (currently Montana), from each station's first hourly report.
stations.parquet: triplet, name, latitude, longitude, elevation (m), state.obs/<triplet>.parquet:valid_time(UTC),air_temp_c,precip_1h_mm,snow_depth_cm,swe_mm, and the other columns of the wxfuser observation schema.
Processing, relative to the raw AWDB service:
- AWDB stamps hourly data in station standard time (PST for western SNOTEL). It is converted to UTC here.
precip_1h_mmis the increment of the cumulative gauge after removing sensor jitter (a future-minimum filter). Plain positive differences overcount summer precipitation roughly tenfold.- Values are range- and step-checked, and stuck sensors are nulled.
Source: USDA NRCS National Water and Climate Center, AWDB REST API (public domain).
points/<model>/<YYYY-MM>.parquet
Every archived run of a model initialised in that month, at every station point
(nearest grid cell): station_id, model, init_time, lead_h, valid_time, and
the forecast columns fc_air_temp_c, fc_rh_pct, fc_wind_speed_ms,
fc_wind_gust_ms, fc_precip_1h_mm (mean hourly rate over the step, mm).
Ensemble models are the member mean.
| model | runs | source store |
|---|---|---|
| hrrr | 00z and 12z, 2018-07 on, 48 h | NOAA HRRR via dynamical.org |
| gefs | 00z, 2020-10 on, to 168 h | NOAA GEFS via dynamical.org |
| ecmwf_ens | 00z, 2024-04 on, to 168 h | ECMWF IFS ENS via dynamical.org |
Forecast data comes from dynamical.org archives of NOAA and ECMWF open data. See dynamical.org and each originating centre for licensing and attribution terms. ECMWF open data is CC BY 4.0.
- Downloads last month
- 988