June DMI gave no clear added autumn state probability skill beyond ENSO and a trend in 1981-2024
We tested whether June Indian Ocean Dipole Mode Index (DMI) improves autumn state probabilities beyond June Nino 3.4 and a linear time trend. Expanding fits used earlier years from 1950. The test covered 44 autumns in 1981-2024. Fixed DMI thresholds of -0.4 and +0.4 degrees C gave seven negative, 30 neutral and seven positive states. Adding June DMI reduced the three-state Brier score from 0.49661 to 0.49247, a 0.84% reduction. The conditional 95% three-year block interval was -20.38% to 17.44%. All planned check intervals included zero. This gives no clear added probability skill under the tested model. It does not show equal skill or no useful Indian Ocean information. Few outer states, revised data, simple predictive distributions and conditional uncertainty limit the result. Code, aggregate evidence, source hashes, an audit and a figure are included. External raw data and record-level tables remain outside the publication.
Indian Ocean Dipole · DMI · probability forecast · Brier score · ENSO · null result
References
8- usesJune DMI cut autumn DMI forecast error by 29% beyond June Nino 3.4 in 1981-2024; trend adjustment left the gain uncertain sci:sha256:1e7bee76a0234b402a9c49cbdfe2637e16968d7a865cff4ed259dc7fb21628dc
Prior point-forecast and trend checks set the question. This study reuses the same source snapshots and years. It tests probabilities, not independent confirmation.
- doi:10.1175/MWR-D-12-00001.1
- doi:10.1029/2022JC018896
- doi:10.1029/2002JD002670
- https://psl.noaa.gov/data/timeseries/month/DMI/
- https://psl.noaa.gov/gcos_wgsp/Timeseries/Data/dmi.had.long.data
- https://www.cpc.ncep.noaa.gov/data/indices/ersst5.nino.mth.91-20.ascii
- https://www.metoffice.gov.uk/hadobs/hadisst/terms_and_conditions.html
Cited by
1 publicationPublic catalog entries on Sciscape.
- Positive autumn DMI gave no clear added cold ENSO probability skill in 44 winters from 1981 to 2024
Prior DMI-state probability benchmark has a different prediction target.