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.

Verified Facts - Post 10: The Swedes Got There First (Swedish NWP / BESK)

Post: /_posts/2026-04-03-The-swedes-got-there-first.md Verified: 2026-05-17

Primary sources consulted

  • Persson, A. (2014). “Fifty Years Later,” History of Meteorology 6 (open access PDF, journal.meteohistory.org) - confirms 45,000 troops 1954, ~0.85/0.82/0.70 correlations, Bolin extensions.
  • Yang, C. (2025/2026). “The First Compute Arms Race: the Early History of Numerical Weather Prediction.” arxiv.org/html/2506.21816v1 - Yang explicitly cites Persson 2005 for Sweden; confirms September 1954 / 45,000 troops / nuclear protection equipment / “tremendously successful” / Herlin/MVC narrative.
  • History of Information, “The Royal Swedish Air Force Weather Service Computes the First Routine Real-Time Numerical Weather Forecasting” (id=2073) - December 1954, North Atlantic, three times a week, Tellus 6 (1954) 139-149 citation.
  • Wikipedia: BESK, BARK, Conny Palm, Bert Bolin, Paul Crutzen, Hans Riesel, Arne Beurling, Aksel C. Wiin-Nielsen, Karlqvist gap.
  • EMC/NOAA EMC history of NWP.
  • MIT News obit for Norman Phillips (2019).

Verified – correct in post

BESK and BARK

  • BESK = Binär Elektronisk SekvensKalkylator. Word length 40 bits, memory 512 words, addition 56 us, multiplication 350 us, 2,400 vacuum tubes (Wikipedia BESK).
  • BESK held title of fastest computer in the world for a short time (Wikipedia, History of Information, Yang).
  • BESK installed/operational by December 1953; inaugurated 1 April 1954; decommissioned 1966 (Wikipedia BESK; Tekniska museet via Drottninggatan95a.se).
  • BESK located at Drottninggatan 95A, Stockholm (KTH/MMN premises). Confirmed.
  • BARK used 8,000 telephone relays and 80 km of cable; completed February 1950, operational 28 April 1950 (Wikipedia BARK).
  • Aiken quote “first computer I have seen outside Harvard that actually works” – present in Wikipedia BARK article (marked needing citation; comes from Lundin 2006 oral history).
  • Conny Palm: full name Conrad “Conny” Rudolf Agaton Palm, statistician at L.M. Ericsson, queueing theory pioneer, died 27 December 1951 at age 44, led BARK project. Confirmed.
  • Karlqvist gap: discovered 1953 by Olle Karlqvist while designing ferromagnetic surface layer for BESK magnetic drum memory. Confirmed.

Swedish NWP operational story

  • Sweden first operationalized real-time NWP in 1954; forecasts began routinely in December 1954, three times a week, for the North Atlantic region, using barotropic model developed by Stockholm University’s Institute of Meteorology (History of Information; Yang). Confirmed.
  • The primary citation is “Staff Members, Institute of Meteorology, University of Stockholm (1954). Results of forecasting with the barotropic model on an electronic computer (BESK). Tellus 6, 139-149.” Confirmed.
  • Bolin (1955) “Numerical forecasting with the barotropic model,” Tellus 7, 27-49 – extends to 72 hours, reports correlations ~0.85 (24h), 0.82 (48h), 0.70 (72h). Confirmed against post-cited primary.
  • September 1954 military exercise: 45,000 troops in central Sweden, including testing nuclear protection equipment; real-time upper-air forecasts critical for simulated fallout. Yang/Persson 2014. Confirmed.
  • “The operationalized NWP forecasts turned out to be tremendously successful compared to subjective forecasts.” Direct quote from Yang (citing Persson 2005); Persson 2005 is the post’s footnote [^1]. Confirmed.
  • Oskar Herlin = leader of Military Weather Central (MVC) under Swedish Air Force; “practical man” who supported NWP. Confirmed (Yang).
  • SMHI was initially skeptical; SMHI took greater involvement after success, did not operationalize NWP until 1961; fully responsible for NWP only by 1973. Confirmed (Yang).

Personnel

  • Carl-Gustaf Rossby: 1939 Assistant Chief of US Weather Bureau (confirmed Wikipedia). Returned to Sweden 1947, founded Department of Meteorology at Stockholm University (“Rossby’s Institute”) and International Meteorological Institute (IMI). Confirmed: he was “first head of the Department of Meteorology at MIT” per Yang and also at Stockholm; founded Tellus (per Yang: “He also launched a new journal, the Tellus, to serve as a counterweight to the 2 other European theoretical meteorology journals”). Confirmed.
  • Rossby died 19 August 1957 at age 58 in Stockholm. Confirmed (Wikipedia).
  • Rossby divided his time between Stockholm, Chicago, and Woods Hole. Confirmed (Wikipedia).
  • Bert Bolin: born 15 May 1925 in Nyköping, Sweden; spent year 1950 at IAS Princeton with Charney and von Neumann; first chairman IPCC 1988-1997; IPCC shared 2007 Nobel Peace Prize with Al Gore; died 30 December 2007. Confirmed (Wikipedia).
  • Aksel Wiin-Nielsen: Danish (born 17 Dec 1924, died 26 April 2010); first Director of ECMWF 1 January 1974 - 31 December 1979; trained at “Rossby’s Institute” Stockholm. Confirmed.
  • Bo Döös: Swedish meteorologist, director of GARP (Global Atmospheric Research Programme) 1971-1982. Confirmed.
  • Norman Phillips: visited Stockholm to help Rossby set up weather model in 1953 (MIT obit calls it “brief stint”; Lewis 1998 BAMS article cited via search said “spent 6 months in Stockholm writing a numerical forecast program for the Swedish Air Force”). The specific “autumn 1953” is plausible but only loosely supported by available open sources; Persson 2005 paywalled.
  • Paul Crutzen: arrived Stockholm University Department of Meteorology in July 1959 (Wikipedia says July 1959, not just “1959”); employed as programmer before PhD; PhD awarded 1968; supervisors Bolin and Witt; 1995 Nobel Prize in Chemistry; ozone research. Confirmed.
  • Hans Riesel: 1957, used BESK nights/weekends, found 18th Mersenne prime 2^3217-1, 969 digits; held record 1957-1961. Confirmed.
  • Arne Beurling: 1940 deciphered Geheimschreiber (Siemens & Halske T52) single-handedly in two weeks using pen and paper; from 1954 professor at IAS Princeton; took over Einstein’s office. Confirmed (Wikipedia).
  • “A magician does not reveal his secrets” quote – attributed to Beurling and cited via Beckman 2002 (footnote 4 of post); not found in Wikipedia but Beckman is the standard source.

JNWPU / IBM comparisons

  • JNWPU established 1954 by NWS + military weather services. IBM 701 installed March 1955. First operational forecasts later 1955 were “very disappointing and not usable by forecasters” (EMC NOAA history – exact wording matches post). Confirmed.
  • IBM 701 used Williams tube memory.
  • JNWPU achieved a useful operational model in 1958 with one-level improved barotropic model running on IBM 704 with core memory. Confirmed (EMC NOAA history; Wikipedia History of NWP).
  • IBM 701 mean time between failure ~30 minutes – cited via Persson 2005 (post’s footnote 1). Not separately verified open-access.

Other claims

  • Saab Lansen wing profile computation on BESK; SARA computer built by Saab as BESK copy “twice as fast” – supported by Yang and Tekniska museet sources.
  • BESK Boys / Facit EDB / Stemme leaving 1956 – supported by Petersson 2005 IEEE Annals (cited in post).

NOT VERIFIED / WEAK POINTS

  1. BESK operational date in spec table. Post lists “Operational 1953” – BESK first ran in late 1953 (Wikipedia, Yang) but formal inauguration was 1 April 1954. The “fastest computer for a few weeks in late 1953” is supported, so 1953 is defensible. Minor framing point only.

  2. BARK completion date. Post says “completed in April 1950.” Wikipedia BARK says “completed in February 1950” and “became operational April 28, 1950.” So “completed April 1950” is loose – “operational April 1950” would be more accurate. Wording change only.

  3. BARK expansion of acronym. Post says “Binär Automatisk Relä-Kalkylator” (Binary Automatic Relay Calculator). Wikipedia BARK gives “Binär Aritmetisk Relä-Kalkylator” (Binary Arithmetic Relay Calculator). Sources vary; the original development project name uses “Aritmetisk.” Recommended fix.

  4. Norman Phillips visit “autumn 1953.” Open sources say “brief stint” (MIT) or “spent 6 months” (Lewis 1998 via search) writing numerical forecast program for Swedish Air Force. The specific season “autumn 1953” is consistent with the Persson 2005 narrative (which the post cites) but not independently verifiable from open access. Leave wording but note we relied on Persson 2005.

  5. Bolin “1950-51” at Princeton. Wikipedia says “spent a year in 1950 at the Institute for Advanced Study.” Persson 2005 likely supports “1950-51.” Defensible.

  6. Paul Crutzen “1959” at MISU. Wikipedia says specifically July 1959. Post says “In 1959.” OK.

  7. Rossby’s heart attack details (rheumatic fever as a boy, weak heart). Persson 2005 source; not separately verified open access.

  8. Saab SARA “twice as fast” claim. Supported by general Swedish computing history sources but not crisply verified in this fact-check pass.

  9. Stig Comet head of BESK AND FRA bureau chief. Plausible; cited via Lundin 2006 (footnote 3). Not independently verified open access.

  10. “Hemmafruar med erfarenhet av att sticka” knitting housewives quote. Cited Lundin 2006 (footnote 3). Not independently verified open access – but Lundin is the cited primary.

Wayback Machine archives requested

Issued web.archive.org/save/* requests for:

  • en.wikipedia.org/wiki/BESK
  • arxiv.org/html/2506.21816v1
  • en.wikipedia.org/wiki/Bert_Bolin
  • en.wikipedia.org/wiki/Hans_Riesel
  • en.wikipedia.org/wiki/Aksel_C._Wiin-Nielsen
  • en.wikipedia.org/wiki/Paul_Crutzen
  • historyofinformation.com/detail.php?id=2073
  • emc.ncep.noaa.gov/emc/pages/ourhistory.php

All returned 200/HTML responses indicating successful capture (or already cached).

Summary

Post 10’s core thesis – Sweden produced the first routine real-time NWP forecasts beginning December 1954, a full year before the JNWPU’s disappointing May 1955 IBM 701 forecasts – is well-supported across multiple independent sources (Yang 2025 arXiv, History of Information, NOAA EMC, Persson 2014). The September 1954 / 45,000 troops / nuclear protection equipment story is directly confirmed in Yang 2025 (citing Persson 2005). BESK specifications match Wikipedia. The Bolin Tellus 7 (1955) correlations 0.85/0.82/0.70 are sourced to the cited primary.

Minor corrections recommended for the BARK acronym expansion and BARK completion date. The “Operational 1953” entry in the BESK spec table is defensible but the formal inauguration was April 1954.