NEL2: From Radial Displacement to Claim-Gated Growth Forecasts: Identifiability, Telemetry, Uncertainty, Recovery, and Final Synthetic Closure

NEL2 is an independent technical research report and numerical evidence ledger by Nikita Teslia, Fieldmind Lab. It examines the evidential requirements for converting radial-displacement measurements into admissible forecasts of irreversible tree growth. Extending the claim-gating methodology of NEL1, the report presents twelve linked stages while keeping observed annual data, synthetic known-growth experiments and statistical planning distinct. The annual branch audits 5,693 public ring-width measurements in 60 recorded series from two Kaindy Lake collections and retrospectively reconstructs a memory–climate forecasting comparison. Previously analysed observations are not presented as a new untouched test, and recorded-series identity is not equated with verified independence of biological trees. The synthetic investigation separates growth from reversible water-related deformation, temperature response and instrumental effects. It examines growth extraction, sampling cadence, point-in-time telemetry, storage and delivery failures, future forecasting, prediction intervals, selective abstention, shift detection and controlled recovery. Known synthetic growth is withheld from forecasting-model training and operational decisions, allowing forecast accuracy against algorithmic labels to be distinguished from accuracy against the underlying simulated growth. The final evaluation uses a frozen computational pipeline and a new synthetic cohort of 240 parent realizations with 4,560 linked telemetry replays. In the primary weekly comparison, the research candidate accepts 9,777 of 12,800 scheduled forecasts. Its mean absolute error against known synthetic growth is 7.065 µm, compared with 9.036 µm for Persistence on the same accepted windows—a 21.81% reduction. This predictive improvement does not establish end-to-end validity. The exceedance rate at the fixed diagnostic tolerance is 35.43%; the weakest evaluated group has 75.44% coverage under nominal 90% prediction intervals; and none of the 160 primary drift-challenge realizations receives a first alarm within the prescribed deadline. The integrated uncertainty/risk and shift-detection criteria therefore remain unmet. The separate automatic growth-claim layer remains closed under its prior qualification rule. This refusal is not interpreted as zero prediction error or evidence that every unreliable forecast was detected. The report’s contribution is a bounded numerical evidence chain showing where signal identifiability, predictive skill, calibrated uncertainty and permission to issue a growth claim diverge. Additional virtual instrumental information supports conditional correction, but does not constitute a physically validated reference sensor. Sample-allocation studies distinguish the value of representative conditions and independent contexts from the apparent precision obtained by accumulating dependent records. Public release and availability. The document provides symbolic equations, selected aggregate metrics, 25 aggregate-result charts, one information-flow schematic, statistical definitions and a public claim register. Source code, raw trajectories, per-window predictions, fitted states, exact operating settings and reconstruction-enabling records are not included. Selected non-public materials may be considered for verification only under a separate written nondisclosure agreement and with the author’s approval. This release is not a complete open-source replication package and does not establish biological calibration, hardware certification or field-deployment readiness. Version 1.2 incorporates editorial and presentation corrections without changing experimental results or scientific gate outcomes. Licence: CC BY-NC-ND 4.0.