Astrophyzix Risk Index ARI V3 Scoring Model & Full Methodology
Last revised: 09 October 2026
| Miss distance (up to 40 pts) | 40 · e^(-0.1386 · missLD) — half-life 5 LD. CAD uses the reported approach. Scout uses caDist only. No caDist → this term is 0. |
| Relative velocity (up to 20 pts) | min(20, 10·log10(speed)). CAD: v_rel, else v_inf. Scout: vInf when supplied; a missing vInf scores 0 and the card shows a dash. |
| Estimated diameter (up to 20 pts) | min(20, diameter_m × 0.05). JPL diameter when CAD supplies one; otherwise H-derived. |
| PHA-class designation | +18 if the flag used on the card is true: JPL CAD pha filter, the derived proxy only if that request fails, or Scout phaScore ≥ 50. |
| Maximum | 98 (40 + 20 + 20 + 18). The dial and bar are scaled to 98, not 100. |
Sources. Confirmed approaches: JPL SBDB CAD API for the selected window. A second CAD request with pha=true supplies the official PHA list for that same window. Unconfirmed NEOCP: CNEOS Scout summary, the current candidate list, not filtered by the CAD date window or distance limit. Official impact-risk data: CNEOS Sentry API (Palermo and Torino, not ARI). Orbital elements and physical parameters: JPL SBDB API, fetched per object on demand. JPL SSD endpoints do not permit direct browser embedding, so requests are relayed through a group-aware CORS-proxy fallback chain with circuit breaking — see System Health for live status.
Time scale. CAD close-approach epochs are reported in TDB and shown as-is. The observatory clock is UTC. TDB−UTC (about 69 s) is not applied. It is small next to the uncertainty for most objects, but not for tightly constrained ones. JPL gives the time uncertainty (t_sigma) and the minimum/maximum distance as 3-sigma values. A Scout card date is that object’s last Scout run (UTC), not a predicted close-approach epoch.
Distance. 1 AU = 149,597,870.7 km, 1 LD = 384,400 km. CAD distances are AU. Scout caDist is lunar distances, the unit on the Scout table (the API sample 0.13 with a 0.0003 au MOID is 0.13 LD, not 0.13 AU). Scout moid is AU.
Scout miss distance. Where Scout supplies a close-approach distance (caDist, lunar distances) it is shown as the miss distance. Where it does not, the card shows the object’s Earth MOID instead, labelled as such. MOID is not a predicted approach, so it does not feed ARI, the headline statistics, the radar or the charts. Those views use approaches only; if no approach is in the window, the hero falls back to the nearest listed object and labels a MOID as a MOID. Velocity is shown as a dash when Scout supplies none. CSV export marks each row as “close approach” or “Earth MOID”.
Diameter. Measured CAD diameter used when present (source = JPL); otherwise estimated from H via D(km) = 1329/√pV · 10^(−0.2H), pV = 0.14 nominal (source = H-derived). The nominal is what ARI uses.
PHA-class. For confirmed CAD objects the flag comes from JPL’s CAD pha filter (the official PHA list for the same window). If that request fails, a derived proxy is used instead: H≤22.0 AND minimum approach≤0.05 AU, which can miss PHAs that do not pass within 0.05 AU this window. The card basis line says which of the two was used. Scout objects use Scout’s own phaScore≥50. Sentry and the per-object SBDB pha flag remain the cross-checks; ARI does not read them.
Scout arc unit. The summary arc value is the same quantity Scout’s table heads as Arc (h). The API does not label the unit; this page displays it in hours, not days.
Scout uncertainty. For Scout objects that have a caDist, rather than treating that single estimate as precise, expanding a Scout card offers an on-demand fetch of Scout’s Monte-Carlo sampled-orbit family and reports the spread (inclination, eccentricity, impact-flag %). The impact-flag share is not an impact probability. Objects with no caDist are labelled by Earth MOID and are not treated as a point flyby.
Orbital viewer. A real ecliptic-plane projection computed from SBDB’s osculating elements via the standard perifocal-to-ecliptic rotation and a Newton-Raphson Kepler solver for current position — not a decorative animation.
Caching. IndexedDB only (no localStorage/sessionStorage), store astrophyzix-neo-v10. V9 cached Scout rows used the wrong caDist unit and are not reused. CAD/Scout use stale-while-revalidate: cached data paints instantly while a live refresh runs in the background. SBDB/Sentry/Scout-orbit lookups use longer TTLs since that data changes far less often.
Validation. A benchmark sanity check runs on every render (1 LD = 384,400 km; Apophis-scale inputs land in the expected high ARI band) — failures surface in the provenance line. It does not re-check the Scout unit; that conversion is fixed in the parser.
ARI scores are recalculated on each load and are not stored or published as impact predictions. Objects with missing parameters may receive lower ARI scores because unavailable values do not contribute to the calculation.