Historical Gridded Snow Products: A Field Guide

Where to get it

• NCAR GDEX d526000 (NetCDF, gridded + station reports)
• Rutgers Global Snow Lab v4 copy (tar, 2015 build)

What's valuable

The only continuous daily snowfall grid built from station data that covers the whole 20th century. Its companion station-report files let you trace which stations fed each cell on each day, and for historical work that traceability is worth as much as the grid.

How to use it vs. the others

Use this as the observed baseline and the benchmark any regional product must beat. Prefer the GDEX copy: it is NetCDF, split by year, and citable. At 1° it cannot resolve ridge–valley or lake-effect gradients, so treat its values as area averages, not point estimates. It is built from COOP, so any 1960s grid is off-limits as an input to a 1960s blind test.

Watch out: Coarse. Inherits COOP observer-practice inhomogeneities. Nothing after 2009.

Where to get it

• NSIDC G10021 landing page (HTTPS, no login)

What's valuable

Snow, temperature and precipitation on one shared grid from the same COOP and MSC station base. That makes it the easiest observed product for checking snowfall, temperature and precipitation against each other, for example testing phase and ratio assumptions against observations.

How to use it vs. the others

Best for 1959–2009 diagnostics that need snowfall and temperature on the same grid. It starts in 1959, so in a September 1959 frozen experiment it is a holdout benchmark only.

Watch out: Same 1° limits and COOP dependence as Kluver. Nothing after 2009.

Where to get it

• NCEI metadata page (returns a metadata error as of Sept 2026)
• Request by name from NCEI: [email protected]

What's valuable

Twice Kluver's resolution, with observation-time metadata attached to each value, which is rare and directly useful for aligning observation windows.

How to use it vs. the others

Worth requesting from NCEI. If you get it, it is the sharpest observed benchmark for 1948–1999, and its time-of-observation data can be used to check how reanalysis snowfall is aggregated into daily windows.

Watch out: Not downloadable right now: no NCEI directory and no known mirror.

Where to get it

• NCEI RSI data directory
• RSI overview
• NCEI snowstorm database (station storm totals)

What's valuable

High-resolution storm-total analyses for major snowstorms back to 1900, and the station-level storm totals behind them.

How to use it vs. the others

Use for event validation, e.g. checking whether a reconstructed daily grid reproduces the footprint and gradient of named storms. Not a climatology: quiet winters and small events are missing.

Watch out: Covers major storms only. Station-based, so it overlaps the COOP truth set.

Where to get it

• NSIDC G02158

What's valuable

1 km snow depth, SWE and daily solid precipitation, with assimilated ground, airborne and satellite observations.

How to use it vs. the others

Use it to learn fine-scale terrain and elevation structure in the modern era, which can then be applied back in time as climatological downscaling patterns. Not a historical product.

Watch out: Starts late 2003. Its snowfall field is modeled, then nudged toward observations.

Where to get it

• NOHRSC snowfall analysis
• Version 2 archive

What's valuable

The closest thing to a modern observed high-resolution snowfall grid: it merges station reports with model background fields.

How to use it vs. the others

Modern-era target for training or calibrating the fine-scale spatial structure of snowfall, and a validation reference for any model applied to recent years.

Watch out: 2008 onward only; background-field dependence varies by version.

Where to get it

• NOAA PSL 20CR
• NCAR GDEX d131003

What's valuable

The only reanalysis reaching back to the 1800s, with an ensemble that provides a spread-based uncertainty.

How to use it vs. the others

The pre-1940 driver. Take circulation, thickness and 850 hPa temperature from ensemble members, not the ensemble mean, so the spread carries through into snowfall uncertainty. Treat its snowfall and precipitation as weakly constrained.

Watch out: Only surface pressure is assimilated; temperature and precipitation are model-generated and spread is large before about 1920.

Where to get it

• NCAR GDEX d626000
• ECMWF dataset page (may block scripts)

What's valuable

An independent 20th-century reanalysis with its own snowfall field, useful as a structural cross-check on 20CRv3.

How to use it vs. the others

Use for comparison: where 20CRv3 and ERA-20C disagree on thermodynamic fields, flag those days as low confidence.

Watch out: Coarse. Single deterministic member.

Where to get it

• ECMWF dataset page (may block scripts)

What's valuable

Coupled ocean–atmosphere version of ERA-20C, with an ensemble.

How to use it vs. the others

Secondary cross-check. Its main extra value is coastal and sea-surface temperature consistency for nor'easter-season thermodynamics.

Watch out: Coarse; access via ECMWF.

Where to get it

• Copernicus CDS single levels
• ARCO-ERA5 cloud Zarr (no login)
• NCAR GDEX d633000

What's valuable

Snowfall (large-scale and convective), precipitation type, freezing level, 850 hPa temperature, thickness, and snow depth, hourly from 1940.

How to use it vs. the others

The post-1940 driver and the main source of SLR and phase predictors. Aggregate to each station's observation window, not to UTC days.

Watch out: Its snow depth assimilates station reports, so it leaks into any COOP holdout. Weaker before about 1958 because upper-air observations were sparse. Snowfall is model output.

Where to get it

• Copernicus CDS
• Google Earth Engine hourly

What's valuable

Snow depth (physical and water equivalent), snow cover fraction, melt and density at 9 km with elevation-corrected temperature.

How to use it vs. the others

Use for snowpack questions only. Its snowfall is ERA5's interpolated, adding no new information about how much snow fell.

Watch out: No snow data assimilation, so its snowpack can drift. Accumulations run from 00 UTC through the day, unlike ERA5's per-hour values.

Where to get it

• NCAR GDEX d640000

What's valuable

An independent full reanalysis starting in 1947 — the only one besides ERA5 covering the late 1940s and 1950s.

How to use it vs. the others

Use as an independent check on ERA5 thermodynamics for 1947–1958, where ERA5 is weakest.

Watch out: Access is simplest through GDEX.

Where to get it

• NOAA PSL

What's valuable

Long, stable, very widely used. Most classic teleconnection and circulation climatologies were built on it.

How to use it vs. the others

Use for circulation indices and pattern typing, where matching the published literature matters more than resolution.

Watch out: Too coarse for snowfall itself.

Where to get it

• NASA GLDAS

What's valuable

Long offline land-model record with snowfall and snowpack terms.

How to use it vs. the others

Minor role: a cross-check on snowpack evolution.

Watch out: Forcing-dependent. Ends 2014.

Where to get it

• NCAR GDEX d628000

What's valuable

Mature reanalysis, now superseded by JRA-3Q.

How to use it vs. the others

Prefer JRA-3Q unless you are matching published JRA-55 results.

Watch out: Ended in early 2024.

Where to get it

• NOAA PSL

What's valuable

Assimilates precipitation, so its precipitation is more realistic than in global reanalyses. Includes snow fields.

How to use it vs. the others

Modern-era mesoscale snow environment and precipitation-type climatology for the eastern US.

Watch out: 1979 onward only; check the current update status.

Where to get it

• CFSR NCAR GDEX d093000
• CFSv2 NCAR GDEX d094000

What's valuable

Coupled system with snowfall and snow fields.

How to use it vs. the others

Secondary modern cross-check.

Watch out: Discontinuity at the CFSR→CFSv2 join in 2011.

Where to get it

• NASA NLDAS-2 forcing

What's valuable

Hourly gauge-based precipitation forcing, with a frozen fraction, at 1/8°.

How to use it vs. the others

Modern hourly precipitation for building event-scale training data.

Watch out: The rain/snow split is a model partition, not an observation.

Where to get it

• NASA GMAO

What's valuable

Snowfall (PRECSNO), and a precipitation-corrected land surface.

How to use it vs. the others

Modern-era intercomparison member.

Watch out: 1980 onward only.

Where to get it

• NOAA PSL

What's valuable

Long, fine-resolution modeled SWE driven by gridded station precipitation and temperature from 1915.

How to use it vs. the others

Pre-1950 snowpack context at high resolution. Its gridded precipitation and temperature forcing is also a candidate liquid-precipitation layer.

Watch out: SWE is modeled, not observed. There is a 2019 version extending to 2018.

Where to get it

• ORNL Daymet

What's valuable

1 km temperature, precipitation and SWE with terrain-aware interpolation.

How to use it vs. the others

Modern fine-scale terrain patterns for downscaling templates.

Watch out: SWE comes from a simple model.

Where to get it

• NSIDC-0719

What's valuable

4 km SWE and snow depth constrained by SNOTEL and COOP observations — one of the best modern snowpack grids.

How to use it vs. the others

Modern snowpack validation.

Watch out: Uses COOP observations, so the same overlap concerns apply.

Where to get it

• Rutgers Global Snow Lab
• NCEI Snow Cover CDR

What's valuable

The longest satellite snow-cover record.

How to use it vs. the others

Large-scale snow-cover context from 1966 onward.

Watch out: Coarse, and records cover only, not snowfall.

Where to get it

• NSIDC G02156

What's valuable

Daily snow/no-snow maps drawn by analysts from satellite imagery.

How to use it vs. the others

Checking modern snow-cover persistence.

Watch out: Resolution changes in 2004 and 2014.

Where to get it

• NSIDC MOD10A1 v61

What's valuable

Fine-resolution snow-cover fraction.

How to use it vs. the others

Fine-scale modern snow-cover patterns.

Watch out: Gaps under cloud; cover only.

Where to get it

• NASA GPM data directory

What's valuable

Radar precipitation retrievals from orbit, including snowfall.

How to use it vs. the others

Research into mid-latitude snowfall microphysics; little use for a historical grid.

Watch out: Swath sampling; short record.