Skip to content

title: "HELIOS Phase II — Evidence Package" subtitle: "Curated Phase I deliverables for Phase II proposal embed" author: "577 Industries Inc. — Columbus, Ohio" date: "2026-05-18"

HELIOS Phase II — Evidence Package

Heliophysics-Enhanced Location Integrity and Operations System Calibrated Space-Weather Decision Intelligence for NASA Mission Operations and U.S. Precision Agriculture

Submitted by: 577 Industries Inc. — Columbus, Ohio — https://www.577industries.com Principal Investigator: Thomas Waweru NASA Phase I subtopic: SPWX.1.S26A — Advanced Data-Driven Applications for Space Weather R2O2R Last refreshed: 2026-05-18 (initial assembly)

About this document. This evidence package is a one-shot orientation packet for Phase II proposal reviewers and NASA-center contacts. It curates the Phase I portfolio across 7 evidence categories — every claim has a clickable URL the reviewer can audit independently. It does not duplicate the public companion document (companion/companion.md); the companion is the narrative, this is the evidence index.

Quick map for reviewers

Question a reviewer asks Section
"What did 577 Industries actually deliver in Phase I?" §1 (narrative), §2 (artifact URLs)
"Show me one verifiable, citable claim" §3 (5 Gannon ground-truth observations)
"Are they engaging the community?" §4 (RFC-0001)
"Did the methodology survive contact with real data?" §5 (Sprint C-Training v1 → v2 evolution)
"Does the calibrated fusion beat best-component?" §6 (kill-gate result — post-Sprint D)
"Is there a peer-reviewable artifact?" §7 (arXiv preprint — post-submission)

1. Lead narrative

The Phase I deliverable narrative is in the public companion document:

The companion mirrors the submitted Phase I proposal section-by-section and attaches live citations to every public artifact as it ships. Use the companion as the reading-order entry point; the evidence package below is the index that lets a reviewer jump directly to any specific claim's underlying artifact.

2. Public artifact URLs with current status

Five public Pages sites and four public repositories (plus one private companion for the trained-weights IP). All claims in the Phase I proposal map to one of these.

# Artifact Repo Pages (HTTP 200) Release Cited in proposal Notes
1 helios-program (meta) GitHub docs v0.2.0 (meta) Master plan, companion, specs, operations
2 helios-provenance-spec GitHub docs v0.1.0 (RFC) §1.4 · §4.2 innovation #2 JSON Schema draft 2020-12 + pydantic v2 ref impl
3 helios-spaceweather-connectors GitHub docs v0.2.1 §2 Obj. 1 · §3 T1 6 adapters live (DONKI · SWPC · GOES · DSCOVR · CDDIS · Scoreboards)
4 helios-fusion-engine GitHub docs v0.1.2 §2 Obj. 2 · §3.1 · §4.2 innovation #1 BMA + isotonic + Mondrian conformal; 176 tests; ≥80% coverage
5 gannon-storm-rtk-analysis GitHub docs v0.1.0 §1.3 Gannon · §2 Obj. 4 · §4.2 innovation #4 1,302 station-hours over 2.5 cm (v1 climatological)
6 helios-fusion-internal (private — IP-gated) n/a (private) refit at 00a80eb (proprietary) Trained BMA priors, isotonic calibrators, equipment transfer functions

Machine-readable artifact registry: companion/footnotes.yaml — auto-synced via python -m orchestration.companion_sync.

The 5-public-Pages portfolio is itself the strongest evidence: every claim in the proposal has a click-through path a reviewer can audit without 577 Industries' cooperation.

3. Citable Gannon ground-truth observations

Five verifiable measurements anchored to the May 10-12, 2024 Gannon G5 superstorm. Each has full provenance lineage (record ID + commit SHA + .provenance.json path). The 5th lands after Sprint D ships.

3.1 Kp peak = 9.0 on 2024-05-11

3.2 Bz peak = -59.16 nT on 2024-05-10 (GSE frame)

3.3 TEC peak = 55.1 TECU at Columbus OH at 2024-05-10T20:00 UTC

  • Source: NASA CDDIS Global Ionosphere Maps (IONEX) via CddisGimAdapter (helios-spaceweather-connectors v0.2.1)
  • Resolution: 2-hour temporal, 2.5° latitude × 5° longitude
  • Current evidence-state: synthetic (verified against the v1 climatological model); real-data confirmation pending Earthdata credentials (operator action — OPERATOR_TODO.md item 2). When credentials land, this row is refreshed to "real" and the Gannon analysis is retagged v0.2.0.
  • Record ID: cddis.gim.2024-05-10T20Z.columbus
  • Adapter docs: https://577industries.github.io/helios-spaceweather-connectors/adapters/cddis/
  • Review pack: specs/2026-05-17-Wave2b-CDDIS-review-pack.md

3.4 1,302 station-hours over 2.5 cm across 25 NGS CORS stations

3.5 Kill-gate hold-out HSS / Brier / CRPS — PENDING SPRINT D

Placeholder. This observation lands when Sprint D ships per spec 2026-05-18-Sprint-D-kill-gate-spec.md. Blocked on OSF pre-registration filing (OPERATOR_TODO.md item 3).

Expected fields when filled: - Kill-gate hold-out HSS vs. best-component-model HSS (PASS threshold: ≥15% relative improvement on the 3-event hold-out) - Reliability-diagram slope per Kp stratum (PASS threshold: within 0.15 of 1.0) - Brier score + CRPS with bootstrap 95% CIs - Decision-tree branch routing (full paper / honest ablation / no paper) - Result JSON at results/<date>-killgate.json - OSF pre-registration URL on file at orchestration/osf_preregistration.url - helios-fusion-engine prereg-v1.0 tag at the locked commit (currently ac53eb6)

4. RFC community engagement

The provenance schema is filed as an open RFC, demonstrating community-engaged standards development before commercial lock-in.

The novelty claim — first feature-level lineage standard for heliophysics fusion systems — is testable: the RFC composition (SPASE + W3C PROV-JSON + RO-Crate) is novel relative to any single-standard predecessor, but the RFC is open so the community can challenge the framing before v0.2.

5. Sprint C-Training v1 → v2 — methodology evolution

Demonstrates the program's rigor + adaptability. When v1's blanket synthetic-proxy assumption broke under empirical probing, the program documented the deviation, re-fit, and shipped — without claiming a result it didn't have.

v1 review pack

specs/2026-05-17-Sprint-C-Training-review-pack.md — initial Table 3-1 training run; honest synthetic-proxy disclosure for pre-2018 events.

v2 review pack

specs/2026-05-17-Sprint-C-Training-v2-review-pack.md — exhaustive ISWA probe, hybrid Path A+B decision, real NOAA SESC archive truth labels, per-(component, event) source labeling.

Key empirical finding: the ISWA Jan 2017 cutover — documented in results/2026-05-17-iswa-coverage-matrix.md — established that 6 of 7 Table 3-1 training events have zero real ISWA Scoreboard coverage. v1 had assumed blanket synthetic-proxy fallback; v2 confirmed empirically and labeled per-(component, event) as iswa_real (12 tuples for Sept 2017) or synthetic_proxy (the remainder).

The OSF "Deviations" methodology note for the pre-registration is drafted at the bottom of the v2 review pack — operator drops it verbatim into the OSF filing.

v2 release artifacts

The discipline signal for reviewers: an assumption broke, the program empirically probed, documented the deviation in a public review pack, and re-fit against a different fitness function. The top-weighted component for Sept 2017 shifted from v1's iPATH (synthetic-only) to v2's SAWS_ASPECS/1.X_Nowcasts_Probability (real ISWA stream) — a different model entirely, reflecting the honest change in evaluation methodology.

6. Kill-gate result — PENDING SPRINT D

Placeholder. Refreshed when Sprint D ships per specs/2026-05-18-Sprint-D-kill-gate-spec.md.

The pre-registered kill-gate is the headline Phase II result. Hypotheses (binding, OSF-filed before any hold-out run): - H1: fused all-clear-revocation HSS on the 3-event hold-out exceeds the best-component-model HSS by ≥15% (relative) - H2: reliability-diagram slope within 0.15 of 1.0 across all three Kp severity strata - Hold-out events: 2022-01-20 (M5.5), 2023-02-17 (X2.2), 2024-05-11 (Gannon G5) - Decision rules: PASS both → full arXiv paper · PASS one → honest ablation paper · FAIL both → no paper

Expected artifacts when filled: - results/<YYYY-MM-DD>-killgate.json — full metrics manifest - Reliability diagrams per Kp stratum (PNG) - Bootstrap CI distributions - Decision branch (paper / ablation / no paper) — documented either way

7. arXiv preprint URL — POST-SUBMISSION

Placeholder. Refreshed when the arXiv preprint is submitted per specs/2026-05-18-arXiv-preprint-draft-spec.md. The arXiv draft agent works in a parallel worktree on helios-fusion-engine/paper/ — no overlap with this evidence package's workspace.

Expected fields when filled: - arXiv ID (e.g., arXiv:2026.NNNNN) - DOI - Cover letter (PDF) - Pre-registration linkage (OSF URL referenced in the paper's pre-registration section) - Companion repo tag at the submitted commit

Appendix A — verification checklist for this evidence package

The Phase II evidence package itself is a deliverable; this checklist confirms it stays honest as the portfolio evolves.

  • Every URL in §2 returns HTTP 200 (Pages-checker pattern; same as companion-check.yml lychee step)
  • The 4 (currently; 5 after Sprint D) Gannon ground-truth observations each have provenance lineage documented (record ID + adapter + commit-SHA pointer + .provenance.json path)
  • commercialization-plan-refined.md cites ≥3 specific Phase I learnings with cross-links to source artifacts (5 cited)
  • phase-ii-proposal-draft.md preserves the Phase I proposal's §1-8 section structure
  • mkdocs build --strict passes locally with phase-ii/ content in nav
  • evidence-package.md renders to PDF without errors via .github/workflows/phase-ii-pdf.yml
  • §6 kill-gate placeholder replaced with real numbers (post-Sprint D)
  • §7 arXiv placeholder replaced with preprint URL (post-submission)
  • §3.3 TEC observation upgraded synthetic → real (post-Earthdata creds + TFT-TEC ship)

This evidence package is maintained at github.com/577Industries/helios-program. Auto-built PDF artifact attached to every workflow run on main via .github/workflows/phase-ii-pdf.yml. Contact: engineering@577industries.com.