SOI → Townsville wet-season rainfall
JADE ANALYSISSouthern Oscillation Index → Townsville wet-season rainfall. Townsville, Queensland.
Identity
- Relationship id
- soi-tsv-wet
- Season / context
- Nov–Apr rainfall; Sep–Feb SOI
- Lag
- 0–3 months
- Period
- 1952–2026
- n
- 75
- Metric
- Pearson r (seasonal means)
- Full-period r
- +0.39 p=<0.001
- Historical r
- +0.37
- Recent r
- +0.33
- Stability
- STABLE RELATIONSHIP
- Structural break
- Not established
- Updated
- 2026-08-18T12:42:11Z
Li Index
Remaining usable order in a relationship — operational negentropy, not thermodynamic entropy.
Λ = (S · P · K · E)^(1/4). If K is missing, three-term geometric mean.
- S — How organised the association is (|r| / 0.60, capped at 1).
- P — How much of that organisation remains through time.
- K — Out-of-sample usefulness versus a naïve climatology.
- E — How much the sample and robustness justify trust in S, P and K.
The paper treats conservation as the structuring of flows that preserve biospheric coherence, and names Confucian Lǐ as principled pattern. The Li Index is that idea made operational for one environmental relationship: remaining usable order. Grier, N. (2025). Conservation is Coherence: Introducing the Negawatt Philosophy of Lawful Design. EarthArXiv preprint.
- · Not a physical entropy measurement.
- · Not a regime-shift detector.
- · Not a 0–100 product score. Components are always shown.
- full (1952–2026) Λ=0.8543 Strongly coherent
- historical_to_1999 (1952–1999) Λ=0.8333 Strongly coherent
- recent_20y (2007–2026) Λ=0.8054 Strongly coherent
- last_30y (1997–2026) Λ=0.7592 Strongly coherent
If the band flips across reasonable baselines, that sensitivity is shown. It is not a regime shift.
Uncertainty
Bootstrap 95% CI +0.21 to +0.58
Δr (recent − historical) = -0.04. Bootstrap interval -0.39 to +0.41. Zero is inside— this is how drift is distinguished from noise.
Confidence is an evidence judgement, not effect size.
Strength through time
20-year rolling Pearson r. The dashed line is zero. Rolling-window wiggles are not a story; only a change that survives the historical-versus-recent bootstrap is treated as drift. Pettitt is not applied to these overlapping windows.
Li Index through time
Rolling Λ from the same 20-year windows. A band crossing is not a regime shift.
0.2-provisional · geometric mean of S, P, K, E on each window
Predictive skill
- Train 1952–1988: r = +0.46
- Test 1989–2026: r = +0.35
- Test MAE 405.45 mm vs climatology MAE 441.41 mm
- Beats climatology on MAE.
Decision relevance
Same as ONI relationship; atmospheric limb of ENSO.
Evidence drawer
Methods
- · Pearson and Spearman on seasonal aggregates
- · 20- and 30-year rolling windows
- · Bootstrap 95% CI
- · Out-of-sample linear skill vs climatology
- · Bootstrap test of recent-versus-historical r
Datasets
- · GHCN-Daily ASN00032040 (Townsville Aero)
- · PSL DMI
Counter-evidence
- · A full-period association can remain while rolling-window r fluctuates inside bootstrap uncertainty.
- · ENSO is not the only driver of Townsville wet-season rainfall; a weaker correlation is not proof that ENSO ceased to matter.
What would weaken this
- · Recent-window r inside the historical bootstrap interval.
- · Direction call reverses under a reasonable alternative baseline.
Linear, seasonal-mean association only. No partialling of IOD, MJO or monsoon onset. ONI and SOI are overlapping representations of ENSO; they are not independent tests.