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.

William (Bill) Bourke

Australian meteorologist; principal author of the foundational papers (Bourke 1972, 1974) that made the spectral transform method workable for the primitive equations on a sphere; long-serving modeller at the Commonwealth Meteorology Research Centre (CMRC), Australian Numerical Meteorology Research Centre (ANMRC), and Bureau of Meteorology Research Centre (BMRC) in Melbourne.

IMPORTANT NAMING NOTE — DISAMBIGUATION

There are several William Bourkes in the Australian record. The William Meskill (“Bill”) Bourke in the Australian Dictionary of Biography (1913–1981, ADB vol. 17, 2007) is a lawyer and Labor / DLP politician for the seat of Fawkner, not the meteorologist. Online searches confound the two. The meteorologist is a different man, publishing as W. Bourke or W. P. Bourke (the middle initial appears on the 1977 Methods in Computational Physics chapter and other Bureau / BMRC documentation). I have not been able to verify the meteorologist’s full middle name, birth date, or birth place from open sources; the ADB entry that comes up first is the wrong person. Any biographical write-up must avoid conflating them. (Sources: Bennett, S., “Bourke, William Meskill (Bill) (1913–1981)”, Australian Dictionary of Biography, ANU, https://adb.anu.edu.au/biography/bourke-william-meskill-bill-12238 — wrong Bourke; BMRC reference list (Phillips, LLNL/PCMDI, 1996) shows “W.P. Bourke” as the author of Bourke et al. 1977.)

1. Biographical arc — what is known and what is not

Status as of 2026-05-12 — almost certainly still alive. Bourke published a substantial historical paper as recently as 2021 in the Proceedings of the Royal Society of Victoria: Bourke, W., 2021, “Pioneering of numerical weather prediction in Australia: Dick Jenssen, Uwe Radok and CSIRAC,” Proc. R. Soc. Victoria 133(2): 67–81. (https://www.publish.csiro.au/RS/RS21010, https://connectsci.au/rs/article/133/2/67/33879/Pioneering-of-numerical-weather-prediction-in.) For a person who was already publishing as an established scientist in 1972, this implies he was probably born around the mid-1930s to early 1940s; he would now be roughly in his eighties or close to ninety. I could find no published obituary in 2010–2025 obituary databases (Legacy.com.au, Mytributes, Echovita searches were negative; Obituaries Australia / National Centre of Biography returns only the unrelated politician). Flag: I have no confirmed birth date, birth place, parents, schooling, or undergraduate university for the meteorologist. The natural assumption — University of Melbourne or Sydney, given his subsequent career in Melbourne — is plausible but not documented in the open sources I could locate.

Career, what is documentable.

  • By the late 1960s Bourke was attached to the Commonwealth Meteorology Research Centre (CMRC), the joint CSIRO / Bureau of Meteorology unit established in 1969 in Melbourne to do numerical weather prediction research. CMRC was a creature of Bill Gibbs (Director of Meteorology 1962–78) and C. H. B. Priestley (Chief, CSIRO Division of Meteorological Physics, Aspendale), explicitly modelled on the von Neumann / Charney / Smagorinsky tradition. (Sources: AMOS Inspiration for Awards pages on Gibbs and Priestley, https://www.amos.org.au/activities/awards/inspiration-for-awards/; Gibbs & Priestley, “Origins of CMRC”, in CMRC/ANMRC Valedictory Report, ANMRC, Melbourne, 1984; Encyclopedia of Australian Science, ANMRC entry, https://www.eoas.info/biogs/A000912b.htm.)
  • 1969–1974: CMRC. The two foundational spectral papers (1972, 1974) carry “Commonwealth Meteorology Research Centre, Melbourne” as Bourke’s affiliation. The 1972 paper was received by MWR during 1972; the multi-level 1974 paper extended the same machinery to baroclinic atmospheres. (Sources: MWR citations recovered via multiple search results; AMS journal index https://journals.ametsoc.org/view/journals/mwre/100/9/1520-0493_1972_100_0683_aeopsm_2_3_co_2.xml and https://journals.ametsoc.org/view/journals/mwre/102/10/1520-0493_1974_102_0687_amlsmi_2_0_co_2.xml.)
  • Visit to GFDL (Princeton), early 1970s. Edwards (A Vast Machine, 2010) — drawing on personal communication from Ron Stouffer (5/13/98) — states that “Bourke and Barrie Hunt had originally worked out the spectral modeling techniques while visiting GFDL in the early 1970s,” and that “in the mid-1970s, GFDL imported a copy of the spectral GCM code developed by W. Bourke at the Australian Numerical Meteorological Research Centre.” This is one of the very few first-person sourced statements about Bourke’s career path. (Source: Edwards, A Vast Machine online supplement, https://pne.people.si.umich.edu/vastmachine/i.GFDL.html.) Caveat: Edwards’s note conflates CMRC and ANMRC; Bourke’s home institution in the early 1970s when the technique was being worked out was strictly CMRC. ANMRC succeeded CMRC only in 1974.
  • 1974–1984: ANMRC (the renamed CMRC, headed first by R. H. Clarke 1974–1978). Bourke was a senior research scientist through this period. The Methods in Computational Physics chapter (Bourke, McAvaney, Puri & Thurling, 1977) gives “Australian Numerical Meteorology Research Centre, Melbourne”. (Sources: BMRC reference list, https://pcmdi.llnl.gov/mips/amip/home/Documentation/11bmrc_fn.html; AMOS Clarke biographical note, https://www.amos.org.au/activities/awards/inspiration-for-awards/.)
  • 1985 onwards: BMRC (Bureau of Meteorology Research Centre, founded 1985 after the political demise of ANMRC). Bourke was one of the senior scientists who carried the spectral-model line forward into BMRC; he is repeatedly listed in BMRC research reports through the 1990s on physical parameterisations, AMIP simulations, and the operational Global Assimilation and Prediction (GASP) system. (Sources: BMRC reference list, especially Hart, Bourke, McAvaney & Forgan 1990 J. Climate 3: 436–459; McAvaney et al. 1991 BMRC Research Report No. 29; Seaman, Bourke et al. 1995 and Bourke, Hart et al. 1995 — Evolution of the Bureau of Meteorology’s Global Assimilation and Prediction System parts 1 and 2; ECMWF / BoM Naughton presentation, https://slideplayer.com/slide/8096493/.)
  • By the late 1990s Bourke was the institutional memory and lead modeller for spectral NWP in Australia. The Naughton slide deck “BoM Modelling and Computing Update” credits “Bill Bourke & BMRC colleagues” as the modelling group. (Source: Naughton, M., BMRC slide deck “BoM Modelling and Computing Update,” circa 2000s, https://slideplayer.com/slide/8096493/.)
  • Retirement: undocumented in open sources, but the BoM moved off its spectral GASP system to the UK Met Office Unified Model under the ACCESS programme in the late 2000s, which coincides plausibly with Bourke winding down as principal modeller. His 2021 historical paper carries no institutional affiliation in the result snippets I could see, suggesting he was retired by then.

Items I could not verify and that the post should not assert:

  • Exact date and place of birth.
  • Primary and secondary education.
  • Undergraduate institution and degree.
  • Doctoral institution, supervisor, and thesis topic.
  • Marriage / family details.
  • Specific dates of his GFDL sabbatical (Edwards only says “early 1970s”).
  • Any AMS or AMOS medal. (I searched the AMOS award lists — Gibbs, Priestley, Morton, Meyers, Radok medals — and the eponymous-awards page on amos.org.au; Bourke is not listed as a recipient under those names.)
  • Whether he is still alive in May 2026 — strongly probable given the 2021 paper, but not certain.

2. The 1972 and 1974 papers

2.1 Bourke 1972

Citation. Bourke, W., 1972: An efficient, one-level, primitive-equation spectral model. Monthly Weather Review 100(9): 683–689. AMS DOI archive at https://journals.ametsoc.org/view/journals/mwre/100/9/1520-0493_1972_100_0683_aeopsm_2_3_co_2.xml. The user’s prior citation of pp. 683–689 is correct.

What it did, technically.

  • Wrote the primitive equations on the sphere in terms of vorticity and divergence (not u and v). This is the key trick: on a sphere u and v have a coordinate singularity at the poles; vorticity and divergence do not.
  • Expanded all fields in spherical harmonics Y_n^m (triangular truncation).
  • Computed non-linear products (advection terms) by the Orszag (1970) / Eliasen–Machenhauer–Rasmussen (1970) transform method: at each timestep go from spectral coefficients to a Gaussian grid via inverse FFT (in longitude) + inverse Legendre transform (in latitude), evaluate the non-linear product pointwise on the grid, then transform back. This is vastly cheaper than the previous “interaction coefficient” approach (which is O(N^5) for triangular truncation N).
  • Used the semi-implicit time scheme of A. Robert (1969), which time-averages the fast-moving linear gravity-wave terms so the timestep is no longer bound by gravity-wave CFL, only by advective CFL — a typical 3× to 6× speed-up on its own.
  • 116-day integrations showed conservation of energy, angular momentum and squared potential vorticity were respected to high accuracy. This was the punchline: the spectral transform method, long known to be elegant, was now demonstrably cheap and stable for a one-level primitive-equation system. (Sources: paper abstract recovered via search results; review by Bourke himself, Spectral methods in climate models, in Schlesinger ed., Physically-Based Modelling and Simulation of Climate and Climatic Change, Kluwer 1988, pp. 375–431; Krishnamurti et al., An Introduction to Global Spectral Modeling, https://imdpune.gov.in/training/training%20notes/Krishnamurti-etal-An%20Introduction%20to%20Global%20Spectral%20Modeling%20-%202006.pdf.)

Predecessors that the paper builds on, all explicit in subsequent reviews:

  • Silberman 1954: first spectral barotropic experiment.
  • A. Robert 1966 (J. Met. Soc. Japan 44: 237–244): semi-spectral integration; spherical harmonics on the barotropic vorticity equation. Robert’s PhD work at McGill. (Source: text of Robert 1966 in the public record, https://www.atmos.albany.edu/facstaff/rfovell/ATM562/robert-1966.pdf.)
  • Robert 1969 (Tokyo IUGG/WMO symposium): semi-implicit treatment of gravity waves in a spectral model.
  • E. Eliasen, B. Machenhauer & E. Rasmussen 1970, Univ. Copenhagen Inst. for Theoretical Meteorology report: the transform method for the spherical-harmonic vorticity equation.
  • S. A. Orszag 1970 J. Atmos. Sci. 27: 890–895: the transform method for the spectral form of the vorticity equation (the canonical North American citation; see https://journals.ametsoc.org/view/journals/atsc/27/6/1520-0469_1970_027_0890_tmftco_2_0_co_2.xml).

2.2 Bourke 1974

Citation. Bourke, W., 1974: A multi-level spectral model. I. Formulation and hemispheric integrations. Monthly Weather Review 102(10): 687–701. AMS index at https://journals.ametsoc.org/view/journals/mwre/102/10/1520-0493_1974_102_0687_amlsmi_2_0_co_2.xml. The user’s citation 102:687–701 is correct.

What was new vs. 1972.

  • Extended the spectral transform machinery from one level to many — i.e. from barotropic to baroclinic. This is non-trivial because it forces you to handle vertical advection, hydrostatic balance, and the energy / temperature equations all in spectral space, with semi-implicit treatment of the linearised gravity-wave operator now being a global elliptic problem at every step.
  • Demonstrated a working semi-implicit time integration for a baroclinic spectral primitive-equation model — for many authors this is the date stamp of “operational spectral NWP becomes feasible.”
  • Introduced divergence dissipation as an initialisation device to damp out spurious large-scale inertia-gravity oscillations during model spin-up — a practical fix that became standard. (Source: paper abstract via search results; Krishnamurti et al. lecture notes, op. cit.)
  • Validated the model against analytic test fields and against Southern Hemisphere analyses — the SH bias is itself important: Bourke and his Australian colleagues were the natural community to push SH-aware NWP, which other northern-hemisphere modellers had under-tested.

The companion paper by Hoskins and Simmons (1975), Quart. J. Roy. Meteor. Soc. 101: 637–655, published the multi-layer spectral semi-implicit method independently at Reading. Hoskins–Simmons becomes the bibliographic anchor on the UK / ECMWF side of the family; Bourke 1974 is the southern-hemisphere parent. The two papers are the standard pair-citation for the start of multi-level spectral NWP.

2.3 The 1977 review chapter

Citation. Bourke, W. P., McAvaney, B., Puri, K., and Thurling, R., 1977: Global modelling of atmospheric flow by spectral methods. In Methods in Computational Physics, 17, J. Chang (ed.), Academic Press, New York, pp. 267–324. (Source: BMRC reference list, https://pcmdi.llnl.gov/mips/amip/home/Documentation/11bmrc_fn.html. Also the canonical citation in Edwards 2010, in the LLNL/PCMDI documentation, and reproduced in most spectral-NWP review papers.)

This Methods-in-Computational-Physics chapter is the operational blueprint the rest of the world copied. The 1972 and 1974 MWR papers are the founding documents; the 1977 chapter is the manual.

3. Institutional setting

  • CMRC (1969–1974), joint CSIRO / Bureau of Meteorology research centre at Aspendale and Melbourne. Created by Gibbs and Priestley to bring NWP research into Australia. First serious computational meteorology environment in the country, building on Dick Jenssen and Uwe Radok’s late-1950s CSIRAC barotropic experiments (the subject of Bourke’s own 2021 historical paper).
  • ANMRC (1974–1984), the renamed and slightly more independent successor, on paper run by CSIRO and the Department of Science. Reg Clarke OIC 1974–1978. Period of greatest external visibility for Bourke’s group — Methods in Computational Physics chapter (1977), McAvaney, Bourke & Puri 1978 J. Atmos. Sci. 35: 1557–1583 (“A global spectral model for simulation of the general circulation”), and the export of the Australian spectral code to NCAR and GFDL.
  • BMRC (1985–2007ish, eventually absorbed into CAWCR), the Research Division of the Bureau of Meteorology, founded after the political demise of ANMRC. Bourke moved with the group. By the 1990s BMRC’s operational global model — GASP, the Global Assimilation and Prediction System — was the descendant of the 1974 multi-level spectral model, on much higher resolution and with full physics packages from McAvaney, Hart, Rikus and others. (Sources: EOAS entries for CMRC, ANMRC and BMRC, https://www.eoas.info/biogs/A000913b.htm and https://www.eoas.info/biogs/A000912b.htm; Seaman, Bourke et al. 1995 GASP papers.)

Collaborators worth naming in the post:

  • Brian Hunt — went with Bourke on the GFDL early-1970s visit; later at CSIRO Atmospheric Research, ran the CSIRO spectral GCM, Mark series. (Edwards 2010.)
  • Barrie McAvaney — long-running co-author from 1977 onwards; led the BMRC physics development through to the AMIP era; co-authored the 1978 JAS paper and most BMRC AGCM technical reports.
  • Kamal Puri — co-author 1977, 1978; subsequently took the Australian model code to NCAR during an extended visit, where it became the basis for CCM0A (see §4 below).
  • R. Thurling — fourth author on the 1977 review; less visible in the later literature.
  • T. L. Hart, R. A. Colman, L. Rikus, M. Naughton, J. R. Fraser, R. Seaman, P. Steinle, G. Embery — the BMRC modelling group through the 1990s; all co-authors with Bourke on AMIP and GASP papers.

Not a Bourke collaborator (contra the user’s question): Masao Kanamitsu. Kanamitsu was at JMA in Tokyo in the early 1980s — he led the team that built JMA’s first operational spectral model (Kanamitsu, Tada, Kudo, Sata & Isa, 1983; see https://www.jstage.jst.go.jp/article/jmsj1965/61/6/61_6_812/_article). He later moved to NCEP and to Scripps, working on regional spectral models (Juang & Kanamitsu 1994). I find no direct Kanamitsu–Bourke joint paper; the apparent linkage is that both were applying spectral-transform NWP at their respective agencies in the 1970s–1980s. Worth recording that JMA, NCEP, ECMWF, NCAR all picked up the spectral-transform method within roughly a decade of Bourke 1974, but none of those adoptions are joint papers with him.

4. Why spectral — what problem was being solved

Two intertwined problems, both well-documented in the contemporary literature.

Problem 1: the pole problem on a latitude–longitude grid. Meridians converge at the poles; on a regular lat-lon grid the east-west grid spacing goes to zero. Explicit time integration is then constrained by the CFL condition at the smallest grid box — meaning at the poles — and the timestep collapses. Sophisticated finite-difference modellers in the 1960s either filtered (Smagorinsky’s GFDL Markfort) or polar-cap-reduced (Kurihara’s grid used in GFDL Zodiac, Kurihara 1965 MWR 93: 399–415) the grid near the poles to keep the timestep usable. All such fixes are kludgy and have their own conservation issues. (Source: GFDL’s own model-history page, https://www.gfdl.noaa.gov/brief-history-of-global-atmospheric-modeling-at-gfdl/.)

Problem 2: spherical harmonics naturally avoid the singularity. The functions Y_n^m are eigenfunctions of the Laplacian on the sphere and treat all latitudes evenhandedly. Operators like the horizontal Laplacian, gradient, and divergence all have simple spectral representations. The Robert / Eliasen-Machenhauer / Bourke programme was a return to the kind of global-mode representation that von Neumann’s group had toyed with in the early 1950s but had abandoned because nobody knew how to handle the non-linear terms efficiently.

Why it was thought to be infeasible. In a triangular-truncated spectral model at resolution T_N you have roughly N^2 spectral coefficients. A naive evaluation of the quadratic non-linear terms in spectral space — via “interaction coefficients” — scales as N^5 and is hopeless for any operationally interesting resolution. This is the wall Silberman (1954) hit.

The transform method (Orszag 1970; Eliasen et al. 1970; Bourke 1972, 1974) breaks the wall. Evaluate the non-linear terms not in spectral space but in physical space (a Gaussian grid that exactly integrates the relevant Legendre polynomials), and ferry between spectral and physical via FFT-in-longitude + Legendre-transform-in-latitude. Cost scales as N^3 log N — same order as a comparable grid-point model — and you keep all the geometric advantages. Bourke’s specific contribution is the engineering: he showed in 1972 that the transform method could be made to work for the full primitive equations on the sphere, conserving the quadratic invariants over month-long runs; and he showed in 1974 that it could be made to scale to many vertical levels with a stable semi-implicit time scheme. (Sources: Bourke 1988 review chapter; Krishnamurti et al. 2006 textbook Introduction to Global Spectral Modeling; Williamson, “100 Years of Progress in Forecasting and NWP Applications,” AMS Meteor. Monogr. 59, 2019, https://journals.ametsoc.org/view/journals/amsm/59/1/amsmonographs-d-18-0020.1.xml.)

5. Adoption history — who picked it up, when, why it became dominant, why it declined

1970s: simultaneous global adoption.

  • Canada (RPN, Dorval). André Robert’s group at McGill / RPN ran the first operationally usable spectral model — Canada was apparently the first country to put a spectral algorithm into the operational suite (1974 per most sources; the precise model was a barotropic / shallow-water spectral, with subsequent baroclinic versions following Robert’s 1969 semi-implicit approach). (Source: CMOS Bulletin tribute to André Robert, https://bulletin.cmos.ca/andre-robert-one-of-the-very-first-scientists-to-successfully-perform-a-simulation-of-the-atmospheres-general-circulation-at-the-global-scale-using-a-computer-model/.)
  • Australia (CMRC/ANMRC). Bourke 1972, 1974; Bourke et al. 1977; McAvaney, Bourke & Puri 1978. Operational use of the multi-level spectral model in the Bureau’s daily suite came progressively over the late 1970s and 1980s.
  • United Kingdom — but at Reading, not yet at the Met Office. Hoskins and Simmons (1975) at the University of Reading laid the academic groundwork for what would become the ECMWF system; Andrew Simmons then moved to ECMWF and was central to its first operational spectral model.

Late 1970s — early 1980s: GFDL imports the Australian code. “In the mid-1970s, GFDL imported a copy of the spectral GCM code developed by W. Bourke at the Australian Numerical Meteorological Research Centre” — Edwards 2010, citing Stouffer. Holloway then bolted Manabe’s Zodiac physics onto Bourke’s dynamical core, producing the GFDL Supersource model that remained in use at GFDL into the 1990s. This is a strong story for the post: the Princeton lab that defined finite-difference atmospheric modelling under Smagorinsky and Manabe (Zodiac, Markfort) ended up running its 1970s-1990s spectral GCM on Bourke’s Melbourne code. (Source: https://pne.people.si.umich.edu/vastmachine/i.GFDL.html, citing Stouffer pers. comm. 1998 and Gordon & Stern, “Spectral Modeling at GFDL,” GARP/JOC report, 1974.)

Late 1970s — early 1980s: NCAR adopts the Australian code via Kamal Puri. “The original versions of the NCAR Community Climate Model, CCM0A and CCM0B, were based on the Australian spectral model (Bourke et al., 1977).” Kamal Puri took the model from ANMRC to NCAR during an extended visit; Eric Pitcher (Miami) and Bob Malone (Los Alamos) tuned the FFTs and physics; CCM0A was released as the first NCAR community model. CCM1 (mid-1980s) evolved from CCM0B; CCM2, CCM3, CCM4 continued the spectral-dynamics line for two decades. (Source: NCAR CCM tech notes, https://opensky.ucar.edu/system/files/2024-08/technotes_400.pdf and https://opensky.ucar.edu/system/files/2024-08/technotes_187.pdf; CCM0B description.)

1980: NMC / NCEP (USA). Joseph Sela began spectral-model development at NMC in 1975; the first operational global NMC spectral model (T30, 12 levels) went live August 1980. The GFS lineage from that model continued as a spectral system until 2019. (Source: Sela, The NMC Spectral Model, NOAA Tech Report NWS 30, 1980; https://catalog.hathitrust.org/Record/102347136; AMS 2019 retrospective on the GFS.)

April 1983: ECMWF goes operational with spectral T63L16. Replacing the original ECMWF grid-point model. Resolution rose steadily — T106 in 1985, T213 in 1991, TL319 then TL511, eventually TL1279 by the mid-2010s. ECMWF stayed spectral as its dynamical core well into the 2020s. (Source: Simmons, A. J., and D. M. Burridge, 1981, MWR 109: 758–766, and the ECMWF retrospective “Fifty years of Earth system modelling at ECMWF,” https://www.ecmwf.int/sites/default/files/elibrary/81651-fifty-years-of-earth-system-modelling-at-ecmwf.pdf; “The ECMWF model: progress and challenges,” https://www.ecmwf.int/sites/default/files/elibrary/2014/13033-ecmwf-model-progress-and-challenges.pdf.)

1983: JMA (Japan). First operational global spectral model at JMA — Kanamitsu et al. 1983, T42L12. JMA stayed spectral until the late 2010s.

Mid-1980s onwards: spectral becomes dominant. By the mid-1980s the major NWP centres operating global models — ECMWF, NCEP, JMA, BoM (BMRC), DWD (Germany, briefly), Météo-France — were running spectral dynamics. NCAR’s CCM line dominated US academic climate modelling. GFDL’s Supersource carried the spectral core through Manabe-era climate runs. Spectral models also kicked off the first IPCC-era coupled climate models.

Holdouts. GFDL itself eventually moved away from spectral for comprehensive coupled climate work, citing two problems explicitly: (1) Gibbs ripples in topography and orographic forcing — spherical harmonics struggle to represent the sharp edge of a mountain without ringing; (2) difficulty conserving mass of tracers and dry air to machine precision, which matters for long climate runs and for chemistry transport. (Source: GFDL model-history page, https://www.gfdl.noaa.gov/brief-history-of-global-atmospheric-modeling-at-gfdl/.) UK Met Office Unified Model (operational June 1991) was a gridpoint model — the UK never went spectral operationally; this is an important counter-current.

2000s–2010s: the decline. Two drivers, both about computer hardware (see §6):

  1. The all-to-all communication pattern of the Legendre transform scales badly on massively parallel distributed-memory machines. By the late 2000s a model spending half its time shuffling data between thousands of MPI ranks is no longer attractive.
  2. Cubed-sphere, icosahedral and other quasi-uniform grids — combined with discontinuous-Galerkin or finite-volume discretisations — give comparable accuracy with strictly local communication. (Sources: GFDL FV3 page https://www.gfdl.noaa.gov/fv3/; ECMWF IFS-FVM paper, Geoscientific Model Development 12: 651, 2019, https://gmd.copernicus.org/articles/12/651/2019/gmd-12-651-2019.pdf.)

End-state circa 2020s.

  • NCEP retired the spectral GFS in 2019, replaced by FV3 (finite-volume cubed-sphere, developed at GFDL).
  • Met Office’s Unified Model — always grid-point — is being replaced by LFRic on a cubed-sphere mesh.
  • BoM (Bourke’s institutional home) moved to the UK Unified Model under ACCESS in the late 2000s.
  • ECMWF retained spectral as the IFS core but is developing IFS-FVM as a long-term replacement.
  • NCAR replaced CCM with CAM, eventually adopting the spectral-element CAM-SE and then MPAS for the climate side.

The dominance window for operational spectral NWP runs roughly 1974 (Canada) — 2019 (NCEP), with the peak 1983–2010. That is, the spectral era and the Cray-vector / NEC-vector era are very nearly the same era.

6. Hardware connection — spectral and vector supercomputers

The transform method is dominated by two operations: FFT (longitude) and matrix–vector products with Legendre coefficients (latitude). Both vectorise beautifully — long inner loops of fused multiply-adds, no branches, predictable memory access. On Cray-1, Cray X-MP, Cray Y-MP, Cray C90, NEC SX-3, SX-4, SX-5, SX-6 — i.e. on register-vector machines optimised exactly for dense long-vector arithmetic — spectral codes ran near peak. (Sources: Williamson 2019 AMS retrospective; Drake et al., “Finite Difference and Spectral Models for Numerical Weather Forecasting on a Massively Parallel Computer,” https://www.researchgate.net/publication/2730346.)

Concrete data points:

  • ECMWF’s operational spectral T63L16 went into production in April 1983, on a Cray-1A and shortly a Cray X-MP.
  • NCEP’s spectral T30L12 went operational August 1980 on the IBM 360/195 and migrated to Cray’s vector machines through the 1980s.
  • BMRC/BoM ran its spectral GASP model on a series of vector machines, culminating with the leased NEC SX-4 installed September 1997 at the joint Bureau-CSIRO High Performance Computing Centre, 150 Lonsdale Street Melbourne. The benchmarks for that procurement — choice of Fujitsu, NEC, Cray — were dominated by the BoM spectral code performance. (Source: CSIROpedia computing history, https://csiropedia.csiro.au/csiro-computing-history-chapter-7/; Museums Victoria record of the SX-4 hardware, https://collections.museumsvictoria.com.au/items/289919.)

When the vector era ends — Earth Simulator (NEC) 2002 is the last consumer-grade vector machine for atmospheric modelling, after which the world is dominated by commodity scalar clusters and then by accelerators — the spectral method loses its hardware tailwind. The Legendre transform’s all-to-all becomes the bottleneck. This is the bluntest explanation of why the spectral era and the vector era track so closely. Cubed-sphere and icosahedral cores are not better physics — they are better fits to the hardware after 2005.

This connection — spectral methods are vector-supercomputer methods — is documented explicitly in:

  • Drake et al., ORNL: “Finite Difference and Spectral Models for Numerical Weather Forecasting on a Massively Parallel Computer,” explicit comparison.
  • Worley et al., “Algorithmic considerations for development of three-dimensional spectral models” in the early-1990s parallel-NWP literature.
  • Mozdzynski et al. (ECMWF), “A Fast Spherical Harmonics Transform for Global NWP and Climate Models,” https://www.researchgate.net/publication/257932619 — the ECMWF community’s attempts to keep the spectral method scalable through fast Legendre transforms.

7. Quotes and quotable lines

I have not been able to retrieve direct first-person quotes from Bourke through open-source channels; the Monthly Weather Review papers and the 1988 Schlesinger chapter are paywalled. What I could find:

  • From the abstract of Bourke 1972 (recovered via search-engine cache, not verbatim from the AMS text — handle with caution; verify before quoting): “In integrations of 116 days, the transform model satisfies principles of conservation of energy, angular momentum, and square potential vorticity to a high degree.” — paraphrased; the original is in MWR 100:683.
  • About Bourke, from Edwards 2010 (A Vast Machine, GFDL chapter): “In the mid-1970s, GFDL imported a copy of the spectral GCM code developed by W. Bourke at the Australian Numerical Meteorological Research Centre. Interestingly, Bourke and Barrie Hunt had originally worked out the spectral modeling techniques while visiting GFDL in the early 1970s.” Direct quote, https://pne.people.si.umich.edu/vastmachine/i.GFDL.html.
  • About Bourke, from NCAR CCM tech note 187 / 400 (CCM1, CCM3): the model is described as “based on the Australian spectral model” of Bourke, McAvaney, Puri and Thurling 1977 — a fact stated almost in passing, but it is the single most concrete tribute to his work in the documentary record.
  • Bourke’s own historical voice is documented only in his 2021 RSV paper on Jenssen, Radok and CSIRAC. The Bourke who wrote that paper is clearly someone who cared about institutional and intellectual memory — he chose to write up a 1957–1959 episode that was buried in a thesis. Worth quoting the spirit of the paper if access is obtained: “to present an account of these significant studies to a wider scientific community.”

For a richer biographical sketch, the most likely productive sources are not on the open web:

  • The CMRC/ANMRC Valedictory Report 1969–1984 (Melbourne: ANMRC, 1984), 160 pp. Available via Australian libraries; cited in the Bureau’s official history (eoas.info A000912b.htm).
  • Bull. Aust. Meteor. Oceanogr. Soc. (BAMOS) — the AMOS bulletin. AMOS occasionally runs obituaries and career tributes; a search of BAMOS volumes 2010–2025 should turn up any Bourke retirement piece.
  • David Day, Weather Watchers: 100 years of the Bureau of Meteorology (Melbourne University Press, 2007).
  • R. S. Seaman, ANMRC — victim of institutional politics? (Bureau of Meteorology, 2004), 29 pp. — by Bourke’s long-time co-author, with insider context on the political demise of ANMRC.

These should be flagged for the post drafting stage as the right places to look for a usable Bourke quote.

8. Myth-busting: was Bourke really first?

Short answer: No, but he was decisive. The credit structure is:

  • Silberman 1954, Lorenz 1960 — early spectral barotropic experiments. Theoretical proof of concept; not operationally viable.
  • A. Robert 1966 (McGill PhD; J. Met. Soc. Japan 44: 237) — spectral barotropic and shallow-water on the sphere with spherical harmonics. First to systematically apply spherical harmonics to NWP on the sphere with semi-implicit machinery.
  • Robert 1969 (Tokyo symposium) — semi-implicit time integration in a spectral framework.
  • Orszag 1970 (J. Atmos. Sci. 27: 890) — the spectral transform method for the vorticity equation. Pure-applied mathematics framing of the technique, in the North American literature.
  • Eliasen, Machenhauer & Rasmussen 1970 (U. Copenhagen technical report) — the spectral transform method, parallel European discovery; explicitly oriented toward NWP rather than toward turbulence.
  • Machenhauer & Rasmussen 1972 — extension and refinement.
  • Bourke 1972 — first working primitive-equation (not just shallow-water) spectral transform model on the sphere, with semi-implicit time stepping, demonstrated to conserve invariants over month-long integrations.
  • Bourke 1974 — first multi-level (baroclinic) primitive-equation spectral model with semi-implicit time stepping; the canonical operational-grade blueprint.
  • Hoskins & Simmons 1975 — parallel UK formulation of the multi-layer spectral semi-implicit method, technically very close to Bourke 1974, published a year later. Both are cited as the founding pair.

So if “first” means “first to write down spherical harmonics for the atmosphere”, it is Silberman / Lorenz / Robert. If “first” means “first to make the transform method work for an NWP-relevant equation set on the sphere”, it is Orszag / Eliasen et al. in 1970. If “first” means “first to put it all together — transform method + semi-implicit + primitive equations + multi-level — and demonstrate that the resulting model conserves the right invariants and is operationally usable”, it is Bourke 1974, with Hoskins–Simmons 1975 close behind.

The cleanest framing for the post: Robert provided the harmonics, Orszag and Eliasen provided the transform, Bourke welded them onto the full primitive equations on the sphere with a stable semi-implicit time scheme and ran them long enough to prove the result was physical. Hoskins and Simmons did the same independently a year later in Reading. The papers are the founding documents of operational spectral NWP.

9. Open questions for the post

  • Verify the meteorologist’s birth date and university — most likely University of Melbourne. The Bureau of Meteorology library or the AMOS BAMOS archive would be the place to ask.
  • Confirm whether Bourke is still alive (probable, given 2021 paper) and whether he has been awarded any honour by AMOS, AMS, or the Royal Society of Victoria. The list of AMOS award recipients on amos.org.au does not list him, but the AMOS R. H. Clarke Lecture (named after Bourke’s first ANMRC director) is the kind of venue where a Bourke tribute would have appeared.
  • Track down the CMRC/ANMRC Valedictory Report — almost certainly the richest single source for Bourke’s career in his own institutional context.
  • Confirm the exact dates of the GFDL visit. Edwards 2010 only says “early 1970s.” A look at GFDL annual reports of 1970, 1971, 1972 should pin this down.
  • Confirm the language used by NCAR and GFDL when they imported the Australian code — were there any in-text acknowledgements that quote Bourke?

10. Source register (working list)

Primary papers and reviews

  • Bourke, W., 1972: Mon. Wea. Rev. 100: 683–689.
  • Bourke, W., 1974: Mon. Wea. Rev. 102: 687–701.
  • Bourke, W. P., McAvaney, B., Puri, K. & Thurling, R., 1977: in Methods in Computational Physics 17: 267–324.
  • McAvaney, B. J., Bourke, W. & Puri, K., 1978: J. Atmos. Sci. 35: 1557–1583.
  • Bourke, W., 1988: in Schlesinger ed., Physically-Based Modelling and Simulation of Climate and Climatic Change, Kluwer, pp. 375–431.
  • Bourke, W., 2021: Proc. R. Soc. Victoria 133(2): 67–81.
  • Hart, T. L., Bourke, W., McAvaney, B. J. & Forgan, B. W., 1990: J. Climate 3: 436–459.
  • Seaman, R., Bourke, W., Steinle, P., Hart, T., Embery, G., Naughton, M. & Rikus, L., 1995: Aust. Meteor. Mag. (GASP papers).

Predecessors and context

  • Robert, A. J., 1966: J. Met. Soc. Japan 44: 237–244.
  • Orszag, S. A., 1970: J. Atmos. Sci. 27: 890–895.
  • Eliasen, E., Machenhauer, B. & Rasmussen, E., 1970: Univ. Copenhagen Inst. Theor. Meteor. report.
  • Hoskins, B. J. & Simmons, A. J., 1975: Quart. J. Roy. Meteor. Soc. 101: 637–655.
  • Kurihara, Y., 1965: Mon. Wea. Rev. 93: 399–415.
  • Sela, J., 1980: The NMC Spectral Model, NOAA Tech Report NWS 30.
  • Kanamitsu, M. et al., 1983: J. Meteor. Soc. Japan 61: 812–.

Institutional history

  • CMRC/ANMRC Valedictory Report 1969–1984, ANMRC, Melbourne, 1984.
  • David Day, Weather Watchers (MUP, 2007).
  • Encyclopedia of Australian Science entries on CMRC, ANMRC, BMRC, https://www.eoas.info/biogs/A000912b.htm and https://www.eoas.info/biogs/A000913b.htm.
  • AMOS Inspiration for Awards page (Gibbs, Priestley, Clarke), https://www.amos.org.au/activities/awards/inspiration-for-awards/.

Adoption and decline

  • Edwards, P. N., A Vast Machine (MIT Press, 2010); GFDL supplement at https://pne.people.si.umich.edu/vastmachine/i.GFDL.html.
  • NCAR CCM Tech Notes 187, 317, 382, 400.
  • ECMWF retrospective: “Fifty years of Earth system modelling at ECMWF,” https://www.ecmwf.int/sites/default/files/elibrary/81651-fifty-years-of-earth-system-modelling-at-ecmwf.pdf.
  • “The ECMWF model: progress and challenges,” ECMWF, https://www.ecmwf.int/sites/default/files/elibrary/2014/13033-ecmwf-model-progress-and-challenges.pdf.
  • AMS Meteorological Monographs 59, 2019, Williamson chapter on 100 years of NWP, https://journals.ametsoc.org/view/journals/amsm/59/1/amsmonographs-d-18-0020.1.xml.
  • GFDL FV3 page, https://www.gfdl.noaa.gov/fv3/.
  • Smolarkiewicz et al., IFS-FVM paper, GMD 12: 651 (2019), https://gmd.copernicus.org/articles/12/651/2019/gmd-12-651-2019.pdf.

Disambiguation (wrong Bourke)

  • Bennett, S., “Bourke, William Meskill (Bill) (1913–1981)”, Australian Dictionary of Biography vol. 17 (2007), https://adb.anu.edu.au/biography/bourke-william-meskill-bill-12238. Lawyer and politician, not the meteorologist.

Research compiled 2026-05-12. Tagged uncertainties are explicit in §1 (“Items I could not verify”) and §9. The strongest single secondary source for Bourke’s career impact is Edwards 2010 (his Stouffer interview); the strongest documentary trace of his collaborative network is the BMRC reference list at LLNL/PCMDI.