T08 — Findings: The Resolution Tax
Supplied working report; original verdicts not validated.
Review status: Needs review. This is an archival source note, not a passed proof.
FINDINGS.md
# T08 — Findings: The Resolution Tax **Verdict: Supported.** ## Headline numbers (2023, 15-min USGS data, three gages) | Gage | h | Missed clusters (of 44/8/6) | Median peak underestimation in flagged bins | |---|---|---|---| | Flashy (Brays Bayou, urban TX) | 1 / 6 / 12 / 24 | 2 / 7 / 16 / 22 | 1.8% / 12.4% / 29.3% / 57.9% | | Mid (Guadalupe, TX) | 1 / 6 / 12 / 24 | 0 / 0 / 1 / 1 | 0.6% / 4.0% / 7.4% / 13.9% | | Smooth (Peace River, FL, spring-fed) | 1 / 6 / 12 / 24 | 0 / 0 / 0 / 2 | 0.2% / 0.8% / 1.3% / 2.4% | ## Support-rule comparison - **R1 (missed events): met.** Missed clusters are non-decreasing in sampling interval for all three gages; at 12h the flashy creek misses 16 of its 44 event clusters while the smooth river misses 0 of 6; at 24h it is 22 vs 2. A manager reading daily averages on the flashy basin would miss half of the year's high-flow events entirely. - **R2 (magnitude distortion): met.** Within flagged bins the coarse mean understates the true 15-min peak by 29% (12h) and 58% (24h) on the flashy creek, versus 1.3% and 2.4% on the smooth river — the claimed "tax scales with the resolution ratio" and the "rare-event processes are taxed more" prediction both hold, and the gap widens with coarseness. - False alarms stayed near zero for all gages at all intervals (0–2 bins): the distortion in these data is one-sided (under-detection and under-estimation), matching the entry's "never a neutral loss of detail" clause. ## Interpretation The mechanism in the entry is borne out: coarse observation averages over within-bin variability, and for a spiky process the average is systematically unrepresentative — it misses the event and understates its size. The smooth groundwater-fed river is nearly immune to resolution coarsening; the flashy urban creek pays a tax that grows steeply with the sampling interval. ## Limitations - Three gages, one year: a demonstration, not an exhaustive survey. The "proportional to the ratio of the two" scaling is visible but not fitted as a law. - The management rule (flag when bin mean exceeds the 15-min 90th percentile) is one plausible coarse policy; other rules (e.g., thresholds on 3-h maxima) would show smaller taxes. The claim is about *a* representative coarse rule, and this one is the natural "same threshold at coarser sampling" reading. - Exceedance clustering is defined at 30-min spacing; results are robust to reasonable variations (±15 min). - The earlier run (Guadalupe as flashy) gave weaker, non-monotone taxes — retained in results/ as a robustness check and a reminder that the tax depends on how spiky the process actually is relative to the sampling interval. ## What would change the conclusion A systematic version would sample dozens of gages spanning flashiness (e.g., by coefficient of variation of daily flows) and fit tax vs (process scale / resolution). If taxes were flat across gages at equal resolution, R2 would fail; the monotone pattern across the three gages here argues the gradient is real.
hypothesis.md
# T08 — The Resolution Tax (Vol XXVIII #03) **Source:** https://incoprea.com/research-library/original-theories-vol28-20261030/theories/resolution-tax.html **Quoted claim:** "Any system that measures or manages a process at a resolution coarser than the process's own scale pays a tax proportional to the ratio of the two, and the tax appears as either missed events or false alarms — never as a neutral loss of detail." Prediction: "Rare-event processes should show systematically larger taxes than smooth processes at equal resolution." Refutation: "If management outcomes are invariant to resolution once measurement error is controlled, the tax is imaginary." **Operationalisation (decided before analysis):** - Process = 15-minute USGS streamflow, 2023. Three gages: Brays Bayou at Alief TX (08074760, flashy urban creek; median 28 cfs vs max 1650), Guadalupe Rv at Victoria TX (08176500, mid), Peace River at Arcadia FL (02298110, groundwater-fed, smooth). - Event = 15-min value above the gage's 90th percentile (clustered at 30-min spacing). - Coarse manager = resample to h-hour means; flag a bin when its mean exceeds the 15-min 90th percentile. Missed cluster = an event cluster overlapping no flagged bin. - Tax metrics: (i) missed clusters; (ii) median relative peak underestimation (bin_max − bin_mean)/bin_max within flagged bins that contain events. **Support rule (pre-written):** - R1: missed clusters non-decreasing in h for every gage, AND at 12h/24h flashy misses ≥ 2× smooth. - R2: flashy median underestimation non-decreasing in h, AND at 12h/24h flashy ≥ 2× smooth. - Verdict: Supported = R1 and R2; Conditional = one; Not supported = neither. - An earlier run used the Guadalupe as the "flashy" gage; it is retained as a robustness check.