Research communication / Editorial record
What has been tested, and what has not?
This page records the evidence status and writing checks for four articles developed from Adrian Dunkley's conference research. An article can be mathematically detailed without having a completed empirical validation. Readers should be able to tell the difference.
Scientific evidence status
Coupled atmospheric volumes. Preliminary Jamaican monthly CHIRPS rainfall-conservation results are source-reported: 15.36 versus 14.23 mm/month RMSE for two neural configurations, while the seasonal allocation baseline was stronger at 9.73 mm/month. These figures have not been independently reproduced in this website review. The proposed full three-dimensional, inverse–forward coupled model and any prospective forecast skill remain unverified.
Climate evidence and public decisions. A human-governed decision framework has been designed. No government deployment, independent comparative evaluation or measured public-budget benefit is claimed.
Learning through sport. This is an intervention and evaluation protocol. No completed youth pilot, causal improvement in literacy, or independently measured environmental impact is claimed.
Before-impact disaster assessment. This is a method for future study. No calibrated damage model, validated pre-landfall damage estimates, mortality predictions or operational forecasts are claimed.
Editorial checks
The October 30 review examined the four original site articles for scientific-status language, internal citation references, overstatement, repeated phrasing, adjacent-word repetition, unsupported detector claims, and whether proposed results were mistakenly described as measured. Two descriptions were corrected: historical storm reconstruction is presented as a proposed method, and loss estimates are described as requiring validation rather than already calibrated. Several abstract sentences were also shortened to state their scientific propositions more precisely.
Automated pattern and citation checks can help editors find passages worth reviewing; they are not independent verification of the underlying datasets or physical equations. An AI detector cannot establish who authored a manuscript. The intended pretrained-model test, using independently labelled scientific-abstract controls, has not yet produced an executed, calibrated result in this review. No authorship probability or purported “human-written percentage” is reported.
Revisions were made to correct meaning and evidence status, not to manipulate an AI detector. Paper titles remain distinct from the conference abstract titles. The publication date remains 11 October 2026; this editorial review is dated 30 October 2026.
Authorship and research use
Author: Adrian Dunkley, Climate Studies Group Mona, The University of the West Indies, Mona. AI assistance has been used in preparing and editing these web research communications. Such assistance does not establish experimental results or replace substantive scientific verification. The articles are publicly available working research communications, not peer-reviewed publications. Where study methods are proposed, illustrative physical animations should be understood as explanatory models rather than operational weather outputs.
For the limitations of automated authorship classification, see Maghami, Ho and Patrick (2026), a systematic review of false positives, and Hadra, Cambridge and Mesbah (2026) on detector variation across academic genres. The editorial method is stored in the research site's version-controlled source repository. Final scientific and venue-specific disclosure responsibilities remain with the author.