Meteorologist, oceanographer, Python solvem probler, experimenter, Raspberry Pi tinkerer, geek dad, manager, space enthusiast, ham radio lover, science columnist, ADHDer, synthetic intellect curious, linux admin cosplayer.

NCEP/NCAR Reanalysis I — Technical Research (Post 46)

Scope: Technical architecture of the NCEP/NCAR Reanalysis I (Kalnay et al. 1996, BAMS 77:437-471).

Key paper

Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha, G. White, J. Woollen, Y. Zhu, M. Chelliah, W. Ebisuzaki, W. Higgins, J. Janowiak, K. C. Mo, C. Ropelewski, J. Wang, A. Leetmaa, R. Reynolds, R. Jenne, and D. Joseph, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437-471. DOI: 10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2

Extension / corrections paper

Kistler, R., E. Kalnay, W. Collins, S. Saha, G. White, J. Woollen, M. Chelliah, W. Ebisuzaki, M. Kanamitsu, V. Kousky, H. van den Dool, R. Jenne, and M. Fiorino, 2001: The NCEP-NCAR 50-Year Reanalysis: Monthly Means CD-ROM and Documentation. Bull. Amer. Meteor. Soc., 82(2), 247-267. Extends coverage back to 1948; identifies three uncorrected bugs.

Citation table

Claim Primary URL Wayback URL
Paper published BAMS 77:437-471, March 1996 https://journals.ametsoc.org/view/journals/bams/77/3/1520-0477_1996_077_0437_tnyrp_2_0_co_2.xml https://web.archive.org/web/2024/https://journals.ametsoc.org/view/journals/bams/77/3/1520-0477_1996_077_0437_tnyrp_2_0_co_2.xml
~29,937 citations (SciSpace, 2024-2025); “one of the most cited papers in the geosciences” https://scispace.com/papers/the-ncep-ncar-40-year-reanalysis-project-yeggi2oyjq https://web.archive.org/web/2024/https://scispace.com/papers/the-ncep-ncar-40-year-reanalysis-project-yeggi2oyjq
Model: T62 spectral, ~210 km equivalent grid, 28 sigma levels (5 in PBL, ~7 above 100 hPa); lowest level ~5 hPa from surface, top ~3 hPa https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
System frozen as of 11 January 1995 (date NCEP implemented this configuration operationally) https://climatedataguide.ucar.edu/climate-data/ncep-ncar-r1-overview https://web.archive.org/web/2024/https://climatedataguide.ucar.edu/climate-data/ncep-ncar-r1-overview
Assimilation: Spectral Statistical Interpolation (SSI) = 3D-Var; Parrish-Derber 1992 https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
Parrish & Derber 1992 SSI paper: Mon. Wea. Rev. 120:1747-1763 https://journals.ametsoc.org/view/journals/mwre/120/8/1520-0493_1992_120_1747_tnmcss_2_0_co_2.xml https://web.archive.org/web/2024/https://journals.ametsoc.org/view/journals/mwre/120/8/1520-0493_1992_120_1747_tnmcss_2_0_co_2.xml
Hardware: Cray YMP/8 (128 MW, Unicos 7); ~24 h YMP wall-clock (2-7 processors) per month of reanalysis https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
Cray C90 acquired by NCEP early 1994; operational systems migrated off YMP April 1994 https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
Output classes A/B/C/D: A = strongly observation-influenced; B = obs + model; C = model-derived; D = climatology https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
SST boundary condition: Reynolds OISST (post-1982, AVHRR era) + UKMO GISST earlier https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
Frozen-system principle: “frozen state-of-the-art analysis/forecast system” so trends reflect data, not algorithm changes https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
Roy Jenne (NCAR) led US-Russia bilateral data exchange (20 tapes, 57 USSR stations, 1961-1978) with Shumbera (NCDC) https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
NCAR held GTS upper-air tapes from March 1962; earlier data required dedicated rescue https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC https://web.archive.org/web/2024/https://www.cpc.ncep.noaa.gov/products/wesley/cdrom/bams96/BAMSPAPR.ASC
Output formats: 6-hourly, daily, monthly; Gaussian grid 192×94 (T62) or 2.5° (144×72); now in NetCDF4 https://www.cen.uni-hamburg.de/en/icdc/data/atmosphere/reanalysis-atmosphere/ncep.html https://web.archive.org/web/2024/https://www.cen.uni-hamburg.de/en/icdc/data/atmosphere/reanalysis-atmosphere/ncep.html
Reanalysis 1 still continuously updated; coverage 1948-01-01 to present https://climatedataguide.ucar.edu/climate-data/ncep-ncar-r1-overview https://web.archive.org/web/2024/https://climatedataguide.ucar.edu/climate-data/ncep-ncar-r1-overview
Kistler 2001 errors found: (1) snow cover 1974-94 (1973 binary repeated); (2) PAOBS 180° longitude shift 1979-92 SH; (3) spectral snow precipitation bias https://ftp.cpc.ncep.noaa.gov/wd51we/reanal/bams_paper.2001/reanl2.htm https://web.archive.org/web/2024/https://ftp.cpc.ncep.noaa.gov/wd51we/reanal/bams_paper.2001/reanl2.htm
NCEP system change Feb 1997: MVS mainframe → networked UNIX https://ftp.cpc.ncep.noaa.gov/wd51we/reanal/bams_paper.2001/reanl2.htm https://web.archive.org/web/2024/https://ftp.cpc.ncep.noaa.gov/wd51we/reanal/bams_paper.2001/reanl2.htm
Kalnay was Director of NCEP/EMC 1987-1997; PhD MIT 1971 under Charney (first woman MIT meteorology PhD); died 13 August 2024 https://en.wikipedia.org/wiki/Eugenia_Kalnay https://web.archive.org/web/2024/https://en.wikipedia.org/wiki/Eugenia_Kalnay
AMS in memoriam confirms EMC tenure and 40-year reanalysis as Kalnay’s landmark legacy https://headlines.ametsoc.org/2026/02/17/in-memoriam-eugenia-kalnay/ https://web.archive.org/web/2026/https://headlines.ametsoc.org/2026/02/17/in-memoriam-eugenia-kalnay/

Detailed notes

1. The model (T62L28)

  • NCEP global spectral model, frozen at its mid-1990s operational configuration (the MRF / GFS lineage). The configuration was the one that went operational on 11 January 1995.
  • Triangular truncation T62: total wavenumber 62; equivalent to about 210 km horizontal grid spacing.
  • 28 sigma vertical levels: 5 in the boundary layer, ~7 above 100 hPa, lowest level ~5 hPa from the surface, top level near 3 hPa.
  • Output Gaussian grid: 192 × 94 ≈ 1.875° × 1.875°. Some derived products also on 2.5° × 2.5° (144 × 72).

2. Data assimilation: SSI (3D-Var)

  • Spectral Statistical Interpolation = 3-dimensional variational analysis in spectral space (Parrish and Derber 1992, Mon. Wea. Rev. 120:1747-1763).
  • Analyzes spectral coefficients of NMC model directly; minimizes objective function that is formally the same as OI but no longer requires the locally-isotropic data selection that OI demanded.
  • Background error covariance from “NMC method”: average difference between 12- and 24-h forecasts valid at the same time, used as a proxy for forecast error covariance.
  • No nonlinear normal mode initialization needed (the variational balance constraint on time derivative of the divergence equation handles initialization implicitly).
  • Irony of co-authorship: Lev Gandin — the leading theorist of optimal interpolation (Gandin 1963) — is a co-author of the Kalnay paper that consolidated OI’s legacy in the very system that retired it. The reanalysis was OI’s monument and its operational obituary in one stroke.

3. The frozen-system principle

  • Core conceptual breakthrough: lock the model and assimilation system across the entire reanalysis period so apparent climate trends reflect data changes, not algorithm drift.
  • Operational analysis archives are heterogeneous (continuous model and scheme upgrades cause artificial trends); the reanalysis substitutes a single fixed system across all years.
  • The remaining inhomogeneity is the observing network itself — most famously the pre-/post-satellite (TIROS-N, late-1978) discontinuity in tropical and Southern Hemisphere fields, and (separately) the 1957 IGY radiosonde expansion.

4. Observation rescue

  • Input streams: land surface synoptic, ship and buoy, radiosonde, pibal, aircraft (AIREP/ACARS), satellite radiances and retrievals (TOVS from 1979), plus PAOBS (Australian Bureau bogus surface pressure analyses for the data-sparse Southern Ocean).
  • NCAR held GTS tapes back to March 1962; pre-1962 data required dedicated rescue.
  • Roy Jenne (NCAR) coordinated the historical archive work — leading the US-Russia bilateral exchange (with Shumbera at NCDC) that yielded 20 magnetic tapes of upper-air data for 57 USSR stations, 1961-1978.
  • Punch cards, paper tapes, ship logs, and microfilm radiosonde records were all in scope. Jenne’s group at NCAR is the institutional memory the project relied on.

5. Temporal coverage

  • Originally 1957-1995 (the “40-year reanalysis” of the paper’s title — note the period is centred on the IGY 1957-58 starting point because radiosonde coverage exploded then).
  • Extended back to 1948 in Kistler et al. 2001 (the “50-year reanalysis”).
  • 2026: continuously updated daily; coverage 1948-01-01 → present. Reanalysis 1 is the longest-running continuous reanalysis product in the world.

6. Reynolds SST boundary

  • Sea-surface temperature lower boundary condition: Reynolds OISST reanalysis from 1982 onward (AVHRR era), UKMO GISST for earlier periods.
  • The Reynolds SST analysis is itself a separate OI-based product — another data layer that exists thanks to OI infrastructure even as the atmospheric analysis moved to 3D-Var.

7. Output products & quality classes

  • 6-hourly assimilation cycles; daily and monthly means.
  • Pressure-level files (standard isobaric surfaces 1000-10 hPa); sigma-level files; and surface flux files with the famous A/B/C/D quality flags:
    • A = analysis variable strongly influenced by observations (upper-air temperature, geopotential, wind in well-sampled regions).
    • B = both observations and model contribute strongly (lower-tropospheric humidity).
    • C = no direct observational influence; field comes entirely from the model (precipitation, surface heat fluxes, soil moisture, cloud).
    • D = field taken from climatology (some surface boundary fields).
  • The A/B/C convention has shaped every reanalysis user’s caveat list since 1996: Class C variables are model output that happens to have been produced inside a data-assimilation envelope, not observations.

8. Computational platform

  • Production ran on the NCEP Cray YMP/8 (128 MW memory, Unicos 7).
  • Throughput: roughly 24 hours of YMP wall-clock (using 2-7 processors) produced one month of reanalysis + forecasts per day, i.e. a 30:1 wall-clock-to-simulated-time ratio.
  • The Cray C90 was newly acquired by NCEP in early 1994; operational suites migrated off the YMP in April 1994. The reanalysis ran in the YMP/C90 transition era at NCEP/NMC (the centre was renamed from NMC to NCEP on 5 October 1995).
  • Production wall-clock for the full 40 years: rough estimate ≈ 40 × 12 production-days ≈ 16 months at one month/day throughput, with parallel production streams shortening calendar duration in practice. Exact total wall-clock: UNVERIFIED.

9. Citation impact

  • The Kalnay 1996 paper is the most-cited paper in atmospheric science and one of the most-cited in all geoscience.
  • SciSpace reports ~29,937 citations (a 2024-2025-era figure).
  • Trajectory (approximate, from secondary references): >10,000 well before 2010; ~20,000 by ~2015; ~30,000 by mid-2020s. Exact Google Scholar count at 2010 / 2020 / 2025 milestones: UNVERIFIED (Scholar pages blocked).

Disambiguations (do-not-confuse list)

  • Reanalysis I (R1) = Kalnay et al. 1996 (this post).
  • Reanalysis 2 (R2) = Kanamitsu et al. 2002 (BAMS 83:1631-1643), bug-fixed R1 with same T62L28 but improved parameterizations and the corrections Kistler 2001 documented.
  • CFSR = Saha et al. 2010 (BAMS 91:1015-1057), fully coupled high-resolution successor (T382, coupled ocean).
  • ERA-15 / ERA-40 / ERA-Interim / ERA5 = ECMWF separate research stream (Agent #3 scope).
  • MERRA / MERRA-2 = NASA GMAO, separate again.
  • Eugenia Kalnay (1942-2024): PhD MIT 1971 under Charney (first woman MIT meteorology PhD); Director of NCEP/EMC 1987-1997; Lowry Chair Oklahoma 1997, then Maryland. Co-developer with Zoltan Toth of the breeding method (1992) — links to Post 45 EPS history.
  • David Parrish & John Derber (NCEP/EMC): authors of the SSI scheme that became the assimilation engine of R1.
  • Lev Gandin (1921-1997): OI pioneer; on the R1 author list — bridges Posts 25-26 (OI) to Post 46.
  • Roy Jenne (NCAR): the institutional memory who made historical reanalysis feasible.
  • Richard Reynolds (NCEP CPC): SST analysis that fed R1’s lower boundary.

Unverified items needing human follow-up

  • Exact wall-clock for the full 40-year production run.
  • Citation count at specific anniversaries (2010, 2020, 2025) via Google Scholar.
  • Exact split of YMP vs C90 production hours during reanalysis.
  • Whether the BAMS paper’s lead text quotes match the canonical PDF (the CPC ASCII version appears to be a near-final preprint).