Artemis II: Scheduled Allocation vs Public Record

Method: every claim carries its evidence class

This audit reports a quantified, previously unpublished ground-segment allocation. It does not argue the allocation was wrong, hidden, or improper — the point is cost transparency by measurement: the ground side of a crewed lunar mission is measurable from public surfaces if someone archives them, so we archived them.

Five claim classes, tagged inline throughout:

• measured our instruments, named artifact with hash
• source-backed dated public document, cited inline
• interpreted inference from measured data
• estimated estimator output under stated assumptions
• conjecture labeled speculation

Instruments: an hourly archive of two public NASA/JPL feeds — the CDSCC (Canberra) forward schedule and the DSN Now realized-tracking feed — running since 2026-04-06. The full archive and every derived artifact behind this audit are published alongside it (Data). The draft was independently red-teamed at high effort before publication; every finding was verified against the artifacts and applied.

1 · The announced baseline

• source-backed Artemis II launched 2026-04-01: a ~10-day crewed lunar flyby (Wiseman, Glover, Koch, Hansen). NASA Artemis hub, as captured in our 2026-07-09 Artemis architecture audit.
• source-backed Post-Artemis II manifest of record: Artemis III = LEO HLS docking demo NET 2027 (NASA crew release, 2026-06-09); Artemis IV = first south-pole landing early 2028; Artemis V = second landing late 2028 + surface-base start (NASA architecture and policy releases, 2026-02-27 / 03-03 / 03-24).
• source-backed The only per-mission DSN figure we located in the Artemis era is retrospective and coarse: uncrewed Artemis I (Nov 2022) used more than 900 hours of DSN support over its 25-day mission — NASA OIG, Audit of NASA’s Deep Space Network, IG-23-016 (July 2023). The same report carries the network-level framing: demand exceeding supply by 40% at times.
• interpreted Absence claim, bounded: within the corpus we reviewed (the NASA pages cited above, the OIG reports in the architecture audit’s source list, and the DSN public web surfaces we archive hourly), no per-mission DSN allocation for Artemis II was published before or during flight. A wider search could overturn this; the falsifier table carries it.

2 · What the instruments measured

Scheduled allocation, one complex

measured 22 unique Artemis II segments at Canberra, 267.1 scheduled antenna-hours over DOY 092–100 (2026-04-02..04-10), segid-deduped at latest version. Dish split: DSS-34 109.3 h, DSS-43 (70m) 108.5 h, DSS-36 49.4 h. Interval-clipped daily allocation (UTC dates): 38.5, 39.0, 33.2, 23.4, 23.2, 23.3, 23.4, 23.9, 39.1 h — a 23.2–39.1 h/day range of daily average equivalent load (roughly 1–1.6 dish-equivalents), heaviest at campaign start and end. DSS-43’s 108.5 h equal 50.2% of a 216-hour nine-calendar-day denominator (53.0% of the 204.9 h support span). This is scheduled allocation as finally published — not billed cost and not proven executed antenna time; the falsifier table carries the distinction.

The schedule’s labels match the announced mission

measured The 22 segments’ activity labels, in chronological order: L1 TLI, L3 OTC-1/2/3, L2 LUNAR CA, L3 RTC-1/2, L1 RTC-3/REENTRY, L1 RE-ENTRY, plus tracking passes; 13 of 22 labels contain B/U; all 22 rows carry work-category 1A1. source-backed DSN scheduling assigns priorities per event, not per mission, on a seven-level scheme with nearly all activities at nominal science level and elevated classes reserved for launches and high-risk critical events (Shouraboura, Johnston & Tran, “Prioritization and Oversubscribed Scheduling for NASA’s DSN”, ICAPS 2016, JPL) — a uniform elevated work category across a crewed campaign is exactly the structure that scheme predicts. interpreted Reading B/U as backup coverage and L1/L2/L3 as criticality tiers (L1 on the irreversible events) is consistent with the label structure; conjecture 1A1 as the literal top-priority booking class — the code-level taxonomy lives in the DSN Service Preparation Subsystem and a public decode was searched for and not found (2026-07-10; the DSN Services Catalog 820-100 and the ICAPS paper carry the structure, not the codes).

Observed corroboration, where coverage exists

measured The DSN Now archive covers 2 of the 9 campaign days (scraper outage 04-08..04-17). In those two days: 75 Artemis II tracking observations across all three complexes (Madrid 37 on DSS-54/56, Canberra 34 on DSS-34/43, Goldstone 4 on DSS-24/26), with two-dish simultaneous support visible at Madrid and Canberra. measured Ranked against every target the network tracked in that window (26 targets, 31 snapshots), Artemis II was first on both metrics: 75 observations against 33 for the next target (Juno), 46.9 in-window credited dish-hours against 20.2, and it was the only target observed at all three complexes. No Goldstone 70m Artemis II track appears in the covered days.

source-backed REV E (2026-07-12). The public DSE downtime feed lists DSS-14 (Goldstone 70 m) under approved downtime from 2026-03-02T16:00:00Z through 2028-05-05T00:00:00Z for concurrent work packages including 20 kW transmitter replacement, US-1/US-2 power work, cooling upgrade, and GSSR obsolescence (status Approved, inSSSasDown: true). Snapshot: dsn-scraper/data/dsn_schedules/downtime_2026-07-10T23-38-11Z.jsonl (sha256 d072b64ed99d6085…, four DSS-14 rows with that window). Source endpoint: https://dse.jpl.nasa.gov/ws/downtimejson. This is the dated cause the falsifier table asked for before any “down for upgrades” wording ships. It does not by itself prove that Artemis II would have used DSS-14 if the dish were up — only that the dish was scheduled out for the entire campaign and through the early Artemis IV window.

estimated Gap-capped credited tracking: 49.9 h across 6 dishes at the 90-minute persistence cap, fully reproducible from the published per-snapshot ledger. Cap sensitivity: 27.7 h at 30 min, 41.2 at 60, 49.9 at 90, 56.7 at 120. The figure is an estimator output and moves −44%/+14% across a 3× cap range — quote the cap with the number. measured Cross-check: all 34 Canberra observations in the two covered UTC dates fall inside a same-dish scheduled Artemis II segment — 34/34 within that coverage window.

3 · Scored comparison

interpreted S3+S4 are the program-watch thesis in miniature: the mission’s public story is complete on the flight side and quantitatively silent on the ground side within the corpus we reviewed — and the ground side is measurable from public surfaces if someone archives them.

4 · Falsifiers and load-bearing assumptions

Each row: what this audit rests on, and what would overturn it.

5 · Limitations, stated up front

• One complex: only Canberra publishes this schedule feed; no extrapolation to Madrid/Goldstone is made (the covered-days data shows the symmetric assumption would be wrong).
• Observed coverage is 2 of 9 campaign days; the return leg is unobserved.
• Feed cadence mechanism undetermined (three batch transitions in 95 days); 16 of 22 segments differ from their April-published versions, edit times unknowable at snapshot granularity.
• Work-category and criticality-tier taxonomies undecoded.
• “Most-tracked target” was withdrawn in an earlier revision for lacking a comparison population; it returns here backed by the all-target ranking artifact with its metrics stated (26 targets, both metrics, Artemis II first on each).

Data: the archive and every derived artifact

Data source attribution: DSN Now data courtesy NASA/JPL-Caltech; schedule data from the NASA Canberra Deep Space Communication Complex (CDSCC, managed by CSIRO for NASA). The underlying feed content is NASA material not subject to copyright. This audit and the dataset curation are not endorsed by or affiliated with NASA, JPL, or CSIRO.