liJadeRelationship intelligence

TSV-001 · Wet/Dry Pattern Persistence

JADE ANALYSIS

Does Townsville show statistically meaningful recurring wet/dry regimes, and if so, have their characteristics changed?

Townsville Aero (BOM 032040 / GHCN ASN00032040) · 85 seasons · 1942–2026 · EXPLORATORY SIGNAL

Median
940.6 mm
Mean
1,030.6 mm
σ
486.3 mm
Range
301.6–3,070.8

Observational integrity

Townsville Aero — GHCN ASN00032040 / BOM 032040

GHCN-Daily does not carry a homogenisation or relocation flag in this extract. Townsville Aero (032040) has operated at the airport since 1940. The Pettitt candidate near 1960 on wet-season totals is statistical only. No instrument-change or site-move note for 1960 is present here; an observing-system break can be neither confirmed nor ruled out from this file.

Thin years: 1 · 2026

Missing months are flagged, not filled. Completeness is days present over the calendar year in the GHCN extract.

Wet-season totals

Tercile split of complete wet-season totals (≤Q1 dry, ≥Q3 wet, else near-median). Rule is statistical, not synoptic.

Proposed 10–11-year cycle

Not supported as a dominant cycle

Strongest periodogram shares: 9y (10.6%) · 12y (7.3%) · 25y (7.2%) · 24y (6.9%). Autocorrelation at lag 10 = +0.11, lag 11 = -0.01.

Structural break

NO SIGNIFICANT CHANGE DETECTED

Pettitt test on wet-season totals. Candidate season 1960, p≈0.1677. Early mean 1,171 mm, later mean 990.2 mm. Not a regime shift.

Era comparison

PeriodnMeanMedianσ
1942–1970291,043.81,028.9362.7
1971–200030971.6903.2478.8
2001–2026261,083.9912.1611.6

Second-half variance 582.1 mm versus first-half 369.3 mm. Mean almost unchanged; tails are not.

Persistence

0.7368

Only consecutive years that are both wet or both dry enter this rate.

Transitions used: 19

Interpretation

  • Wet-season (Nov–Apr) rainfall at Townsville Aero is available for 85 complete seasons (1942–2026). Median wet-season total is 940.6 mm.
  • The proposed ~10–11-year wet/dry cycle is not a dominant spectral feature in this record. Strongest periodogram peaks are at 9y, 12y, 25y. Autocorrelation at lag 10 is 0.1102; at lag 11 is -0.0087.
  • A Pettitt change-point candidate falls at season 1960 (p≈0.1677). This is not statistically significant at 0.05; no structural break is claimed.
  • 2019 and 2024–25 are shown because they are recent comparison periods requested in the brief. Recency is not evidence of significance.

Limits and falsification

  • · Single station (Townsville Aero). Airport siting, instrumentation and observing-practice changes can create apparent breaks.
  • · GHCN-Daily is used as a reproducible public extract of BOM station data, not a homogenised climate-reference series.
  • · A periodogram peak is not a cycle. Persistence and out-of-sample usefulness are required before any operational use.
  • · No cyclone-removal or urbanisation adjustment is applied.

What would weaken this

  • · An independent nearby station series showing no comparable periodicity or break.
  • · The 10–11-year peak disappearing under reasonable record-length or window choices.
  • · A documented observing-system change coinciding with any candidate break.