CARRA2¶

Copernicus pan-Arctic Regional Reanalysis (second-generation CARRA)

Available

At a glance¶

Note on period: The CDS dataset titles say "from 1986 to present", but the sub-daily data start on 1985-09-01 and the daily and monthly statistics (reanalysis-pan-carra-means) on 1985-10-01 (per the CDS catalogue; the request form also offers 1985). As of 2026-10-12 both datasets end on 2025-12-31. Once monthly updates (2–3 months behind real time) begin, the end date will move forward.

Overview¶

The second-generation C3S Arctic regional reanalysis. It replaces CARRA1's two European-sector domains (West and East) with a single 2.5 km polar stereographic grid covering the whole pan-Arctic. Based on an adapted HARMONIE-AROME cy46h1 system (hydrostatic) with 3D-Var upper-air assimilation in 3-hourly cycles, SEKF surface assimilation and ERA5 lateral boundary conditions, it provides 3-hourly analyses, hourly forecasts and daily and monthly statistics from September 1985. Surface processes use the multi-layer SURFEX scheme with a 12-layer explicit snow scheme and multi-energy balance (MEB) for the forest canopy. Single, pressure, height and model levels are all in one CDS entry, with spatial subsetting.

Available variables¶

Key parameters from the parameter tables in the CARRA2 Data User Guide. Names in backticks are GRIB short names; CDS requests use long names such as 2m_temperature and total_precipitation. The CDS sub-daily dataset lists 75 variable names.

instantaneous mean accumulated maximum minimum static — instantaneous: value at the time stamp · mean/accumulated/maximum/minimum: over the indicated interval · static: does not change with time

Surface and single levels (analysis and forecast, instantaneous)¶

Surface and single levels (forecast-only accumulations and extremes)¶

Pressure levels (20 levels, 1000–50 hPa, analysis and forecast)¶

Height levels (18 levels, 15–3000 m) and model levels (65 levels, analysis only)¶

Invariant fields¶

The full variable list is in the official documentation.

How to get the data¶

• CDS — pan-Arctic regional reanalysis sub-daily data (reanalysis-pan-carra) — Web download / CDS API (cdsapi)
  • Requires an ECMWF account and acceptance of the dataset licence (CC-BY). Choose the level type (level_type: single_levels / pressure_levels / height_levels / model_levels), the levels (level_location) and analysis or forecast; use area to subset a region. GRIB2 is the native format; NetCDF is an experimental conversion.
• CDS — daily and monthly statistics (reanalysis-pan-carra-means) — Web download / CDS API
  • Pre-computed daily and monthly means, extremes and totals. Choose time_aggregation (daily/monthly) and product_type (analysis_based/forecast_based). Per the CDS warning of 2026-05-15, gusts are not yet available, surface runoff and fog are missing for many periods, and maximum/minimum temperature and sea ice surface temperature also have gaps, which are being back-filled.
• MET Norway example notebooks (metno/carra2_cds) — Python examples (cdsapi, Jupyter)
  • Examples from the CARRA2 producer for downloading, regional subsetting, time series and sea ice maps.

Example code¶

Put the CDS API url and your key in ~/.cdsapirc before running. Request keys and values were taken from the CDS request form (API definition) and the MET Norway examples.

# pip install --upgrade cdsapi
import cdsapi

client = cdsapi.Client()

# 1) Forecast: total precipitation (accumulated since forecast start) — +18h minus +6h gives 12-hour precipitation
client.retrieve(
    "reanalysis-pan-carra",
    {
        "level_type": "single_levels",
        "product_type": "forecast",
        "variable": ["total_precipitation"],
        "year": ["2020"], "month": ["01"], "day": ["01"],
        "time": ["00:00", "12:00"],
        "leadtime_hour": ["6", "18"],
        "area": [80, -30, 60, 30],        # north, west, south, east — only the region you need
        "data_format": "grib",
    },
    "carra2_tp_fc_20200101.grib",
)

# 2) Monthly statistics: monthly mean 2 m temperature (analysis based)
client.retrieve(
    "reanalysis-pan-carra-means",
    {
        "time_aggregation": "monthly",
        "level_type": "single_levels",
        "variable": ["2m_temperature"],
        "product_type": "analysis_based",
        "year": ["1986"],
        "month": [f"{m:02d}" for m in range(1, 13)],
        "data_format": "netcdf",
    },
    "carra2_t2m_monthly_1986.nc",
)

Practical tips¶

• 10u and 10v are components relative to the polar stereographic grid, not eastward and northward (Data User Guide). For geographic winds, rotate them yourself or use the true-north wind direction 10wdir with the wind speed 10si.
• Some surface variables (users report SST and sea ice fraction among them) come on 2880 x 2880 points. As the guide says, drop the extra 11 rows and columns (all NaN) to match the 2869 x 2869 grid.
• 2 m temperature is too warm over forests (e.g. Siberia and the Sahtu region of Canada, sometimes more than 10 K warmer than ERA5). This is an official known issue; 2 m temperature per forest tile is to be provided later.
• Within about 100 km of the domain edge, precipitation is generally too low because of the ERA5 lateral boundaries; beyond about 200 km the effect is negligible. Exclude the edge band from analyses.
• Near some Greenland ice-sheet stations (NASA-SE 2012–2019, Petermann ELA 2011–2014 and others) 2 m temperature is badly biased, and the guidance is that these data should not be trusted. For ice-sheet work, check the list of sites and periods on the Known issues page first.
• Melt on glaciers is not represented, so glacier surface runoff is unrealistically low, and glacier extent is fixed at mid-period values. Do not use these fields as they are for glacier mass-balance work.
• Do not join CARRA2 to CARRA1 as one time series: domain, grid, model (non-hydrostatic Cy40h1 → hydrostatic Cy46h1), surface scheme and forecast length (30 → 18 h) all differ. If you downloaded roughness length from the CDS before 2026-09-01, you may actually have received surface runoff; download it again.

Notices and change history¶

References for your paper¶

Dataset citation¶

Sub-daily data (reanalysis-pan-carra) — DOI 10.24381/f5effe24

@misc{carra2_subdaily,
  author    = {{Copernicus Climate Change Service}},
  title     = {pan-Arctic regional reanalysis sub-daily data from 1986 to present},
  year      = {2025},
  publisher = {ECMWF},
  doi       = {10.24381/f5effe24},
  note      = {Accessed on DD-MMM-YYYY}
}

Daily and monthly data (reanalysis-pan-carra-means) — DOI 10.24381/b19380d8

@misc{carra2_means,
  author    = {{Copernicus Climate Change Service}},
  title     = {pan-Arctic regional reanalysis daily and monthly data from 1986 to present},
  year      = {2025},
  publisher = {ECMWF},
  doi       = {10.24381/b19380d8},
  note      = {Accessed on DD-MMM-YYYY}
}

Acknowledgements¶

Acknowledge the use of C3S data and give the DOI of the CDS dataset you actually used (sub-daily, or daily and monthly) with your access date. As of 2026-10-12 we found no peer-reviewed paper describing CARRA2; the system is documented in the Data User Guide, the Full system documentation and the test evaluation report (Yang et al., 2025) in the Copernicus Knowledge Base.

Information on the datasets used in this study was compiled with the help of the Korean Climate Data Navigator (KCDN) (Sim, 2026; https://doi.org/10.5281/zenodo.23300923).

Links¶

• CDS — CARRA2 sub-daily dataset
• CDS — CARRA2 daily and monthly dataset
• CARRA2 Data User Guide (Copernicus Knowledge Base)
• CARRA2 Known issues (Copernicus Knowledge Base)
• C3S — Copernicus Arctic Regional Reanalysis service (CARRA1 vs CARRA2)
• CARRA2 user workshop (June 2026) — system, observations and known issues (Schyberg)
• Previous generation CARRA (CARRA1)