Research series
Eleven documents follow a tropical-Pacific field forecaster from system validation to its final deployment handoff.
Series summary
Nine sequential reports develop and test an exactly invertible forecasting system; a standalone manuscript and a shared metric charter sit beside the main sequence.
Outside the arc
The standalone manuscript reports the final products to an external audience; the charter governs how everything else is measured. Neither belongs to the chronology below.
The arc
Nine reports in reading order, each stating its incoming baseline, the single factor it changes, its predeclared gates, and the artifact it hands forward. Each entry opens the complete report, with its prose, figures, tables, and references preserved.
Builds the thing and proves it is built correctly: lossless coefficient packing, exact inverse reconstruction, auditable checkpoint surgery on the official backbone, and fail-closed access to the sealed period. The scope is the pipeline; skill claims come later.
Read reportNine controlled families ask whether direct adaptation on 335 observed windows can forecast at all. It cannot: low error is bought with near-climatological damping, amplitude costs correlation, and residual learning spends itself cancelling persistence.
Read reportMasked coefficient pretraining over four SST products improves reconstruction monotonically and forecast transfer not at all. The durable products are a coefficient-energy explanation of why, and the statistical governance the rest of the series runs under.
Read reportPretraining on the actual forecast task across eleven CMIP6 models transfers where reconstruction did not, and the resulting model spends the project's one pristine hash-gated read of the sealed period. This is the strongest confirmatory result in the series.
Read reportA controlled test of whether ocean heat content adds anything. Inside simulation it helps; through the observation pathway it resolvably hurts. The zeroed-input twin design is what makes that a causal statement rather than a guess.
Read reportGaussian, latent-noise, and diffusion heads surround the frozen center. Diffusion fixes the smudgy individual member that started the programme, but the regional calibration wall survives every tested change, including sixty-four times more data.
Read reportSix rungs of explicit calibration establish what post-processing can and cannot buy. Average reliability can be corrected and lifted into sharp fields, but flow-dependent width does not transfer from simulation and cannot be manufactured after the fact.
Read reportRe-weighting the objective across wavelet bands is what unlocks temporal coherence — the joint sequence architecture alone does not reach it. The adopted stack is the best validation product of the series, and its one authorized test read shows the advantage only partly transferred.
Read reportThe control the earlier reference finding was missing: five pipelines differing only in the anomaly reduction, each retrained from scratch. Retraining, not climatology updating, turns out to explain most of the degradation. Ends by freezing the deployment contract.
Read report