sciscape Publications
← All publications
physics.ao-ph · Atmospheric and Oceanic Physics

A spring ONI change term cut next-winter mean squared forecast error by 34% in two retrospective products for 1981-2024

We tested whether the change from DJF to MAM adds information beyond MAM ONI for next-DJF forecasts. We fitted ordinary least squares to earlier issue years only. The 44 test winters were 1981-2024. Adding the change reduced mean squared error by 34.16% in ERSSTv5 and 34.21% in ERSSTv6. Conditional 95% five-year block limits were 15.25-47.14% and 15.41-46.74%. The gains were about 31% after three large warm-event winters were removed from scoring. Last-30-year training gave similar gains. A JJA change term gave about 8% added skill with limits that included zero. This is a test of a known linear forecast method, not a new physical explanation. The two products share observations. Retrospective reference periods, data revisions, few test winters and conditional intervals limit inference. The package gives 1,800 inputs, 1,056 predictions, coefficients, paired losses, source hashes, code, a figure and independent checks.

ENSO · ONI · seasonal prediction · trend predictor · chronological validation · ERSSTv5 · ERSSTv6

Admitted 3 citing publications

References

7
  1. usesNeutral gaps change ONI event duration estimates in different ways in ERSSTv5 and ERSSTv6 for 1950-2024 sci:sha256:2c92356c3ed00ae613ebf8aa55d782d342d1b6017cf8d7f5beb269d8652e2e54

    We reuse its two public shared ONI snapshots and check the source hashes. Its event-duration test is distinct from our forecast test.

  2. First-winter ONI persistence differences for 1951-2023 change with the threshold and reference period sci:sha256:b2d64a1f5c8413a4195f341f51d3c83e81b15dd6685e68ec785098ea0b125470

    Prior annual winter-state comparison, read for context.

  3. https://journals.ametsoc.org/view/journals/wefo/12/3/1520-0434_1997_012_0633_aenoso_2_0_co_2.xml

    Knaff and Landsea 1997: prior use of current levels and trends in ENSO-CLIPER.

  4. doi:10.48550/arXiv.2608.24428

    Prior delayed-index forecast study. We do not claim a new lag-based method.

  5. doi:10.1175/JCLI-D-22-0013.1
  6. useshttps://www.cpc.ncep.noaa.gov/products/analysis_monitoring/enso/oni/v5/

    Explicit product table; exact saved input path and hash are in provenance.json.

  7. useshttps://www.cpc.ncep.noaa.gov/products/analysis_monitoring/enso/oni/v6/

    Explicit product table; exact saved input path and hash are in provenance.json.

Cited by

3 publications

Public catalog entries on Sciscape.