The Oscillarium

Now: active regimes

A fast entry point for the latest unusual states in the bundle: strongest monthly anomalies, daily teleconnection tails, and the current RMM state. Click any card to open the fuller diagnostic view.

Strongest monthly anomalies

Top-ranked latest monthly indices, standardised against the current Cabinet baseline and shown over their recent history.

The cabinet wall

Latest value of every monthly index, with a five-year sparkline and a standardised anomaly against the baseline chosen below. Tap a tile to open it in Timelines. ° marks series whose record can't cover the chosen baseline, so the full record is used instead.

Timelines

Overlay up to six series — monthly, daily, or MJO amplitudes — on one time axis. Drag on the chart (mouse) to zoom; double-click to reset. On touch, use the range buttons and date fields.

Same mode, different hands

Curated groups where several providers measure the same (or closely related) phenomenon. Overlay them, then pick any pair for a scatter, a difference series, and where they disagree most.

Pairwise agreement

Largest disagreements

Thirteen dialects of the MJO

Each index has been rotated into the RMM (RMM1, RMM2) basis empirically — the rotation and its fit r are disclosed per index. Amplitude correlation and day-by-day phase agreement are computed on days where both indices are active (amplitude ≥ 1).

All dialects vs RMM

Wheeler–Hendon phase space

The MJO's wandering through (RMM1, RMM2). Faint where amplitude < 1; the trajectory shades from grey to violet through time, with a dot each week. Geographic labels appear only for indices that rotate cleanly onto the RMM basis.

A day in the system

Pick any date and see the whole cabinet's state: the MJO wheel with a 40-day tail, the daily teleconnections ranked against their calendar-month history, and every monthly index for that month sorted by how anomalous it was.

The month around it

Every monthly index with a value that month, sorted by |z| against the chosen baseline.

Extremes & spells

Rank the record's highest or lowest values for any index, filtered by calendar month, with a declustering window so one event doesn't fill the table. Below, the longest unbroken spells past a threshold.

Longest spells

Compound states

When do two (or three) modes line up? Conditions are evaluated on monthly series over their common valid months. The lift compares observed co-occurrence with what independence would give — it measures association, not causation.

Occurrence strip

Year × month tapestry

The whole record of one index as a wall of months. Raw values centre on zero (or the record median for non-anomaly series); z-by-calendar-month removes the seasonal cycle first. Hover or tap for the cell value.

Lag laboratory

Correlation between two monthly indices as a function of lead. Positive lag means A leads B by that many months. Filter which calendar months of B count — DJF-only tightens many teleconnection relationships considerably.

Provenance & validation

Cross-source checks (computed at build)

MJO dialects: empirical rotation onto the RMM basis

Series manifest

Bundle files

Caveats worth keeping

Never mix indices without recording provider, version, baseline climatology, filtering and update date — the manifest above carries name, source, units, baseline, coverage and URL for exactly that reason. AMO/AMV comes detrended here (Kaplan, unsmoothed); other AMO flavours answer different questions. MEI v2 rows are two-month seasons labelled by their first month. TNH and PT are only defined in part of the year, hence their large gaps. GLAAM ends in 2014 and the all-India monsoon series in 1999 — kept for history, not for monitoring. ONI's centred 30-year climatology means it is deliberately not the 1991–2020-based Niño 3.4 smoothed (r = 0.978 here, not 1). Daily-index percentiles in the Day view are ranks within that calendar month's full history; the z-baseline setting applies to monthly series only.