Research question
Can locally useful one-step forecasts remain portable and earn permission for automatic operation after leakage-free, source-shifted, uncertainty-aware, and sequential validation?
MODEL SCOPE
Public model description
A leakage-controlled one-step forecasting stack combines grouped and source-held-out evaluation, baseline and memory gates, incremental features, series-balanced uncertainty, versioned replay, selective routing, and fail-closed admission.
PROTOCOL
Declared evidence chain
NEL1.1–NEL1.9 audit two public tree-ring collections through provenance, leakage, portability, uncertainty, prequential integrity, abstention, recovery, and final service closure. Negative gates remain operative in the final result.
Series: NEL1.1–NEL1.9 · 60 series · 5,393 one-step rows · fail-closed closure
- NEL1.1–NEL1.9
- 60 series
- 5,393 one-step rows
- fail-closed closure
The labels below preserve the public protocol identity. They are an index to the publication, not a substitute for its full methods or an open reproducibility bundle.
EVIDENCE
Results inside the model
INTERACTIVE MODEL LENS
Fail-closed routing lens
A source-aware admission path from leakage control through portability and uncertainty to the retained NO_UNIVERSAL_SAFE_STACK result.
All automatic-service admission layers remain closed; audit and replay stay enabled.
- Grouped and source-held-out designs eliminate forbidden target/future leakage; the row-random diagnostic contaminates approximately 98.4% of test targets or future labels.
- Finite endogenous memory and selected local feature or uncertainty gains are supported, but no predictor passes the universal baseline gate and no interval method passes every admissible evaluation.
- The final classification is NO_UNIVERSAL_SAFE_STACK: neither source is authorised for automatic output, while replay, artifact integrity, tamper rejection, and audit infrastructure remain enabled.
BOUNDARY
CLAIM BOUNDARY
- 01Declared numerical protocol
NEL1.1–NEL1.9 · 60 series · 5,393 one-step rows · fail-closed closure
- 02Supported model-level finding
Grouped and source-held-out designs eliminate forbidden target/future leakage; the row-random diagnostic contaminates approximately 98.4% of test targets or future labels.
- Claim boundary
- 03Requires a separate validation chain
Only two collections from one species and one site are studied; living/dead status is perfectly confounded with source identity. The task is univariate, endogenous, and retrospective. It does not establish climate causality, biological mechanism, transfer across Kazakhstan, universal uncertainty validity, field deployment, or autonomous ecological forecasting.
Numerical evidence is not physical validation. Transfer to a material, device, organism, environment, or operational service requires a separate validation chain whenever such a transfer is relevant.
Publication record
- Author
- Nikita Teslia
- Programme
- Applied environmental modelling
- Claim type
- Numerical evidence ledger
- Year
- 2026
- Series
- NEL1.1–NEL1.9 · 60 series · 5,393 one-step rows · fail-closed closure
- License
- CC BY-NC-ND 4.0
- DOI
- 10.5281/zenodo.21812690
Citation
Teslia, N. (2026). NEL1: Claim-Gated Tree-Ring Forecasting under Source Shift and Operational Constraints. Zenodo. https://doi.org/10.5281/zenodo.21812690Open publication recordThe DOI is the canonical external record. The site condenses the public publication and does not replace it.
PROGRAMME TOPOLOGY
Research lineage
Lineage records publication sequence and explicit revision or supersession links. It does not assert empirical causation or an otherwise unverified cross-branch dependence.
No predecessor is asserted in the current research graph.
No published successor is encoded in the current research graph.