Methodology

Observed means an AIS message was received: position, speed over ground, timestamp. Published specification means a static figure from builder, registry or class records — length, tonnage, engine configuration, maximum guest berths. Estimated means a model output with an uncertainty range: fuel burn, CO₂, market value.

Ranges are never replaced by a single tidy number. If the range is wide, the underlying uncertainty is wide.

Each vessel's CO₂ estimate carries its own uncertainty range, reflecting how confident the underlying fuel-burn model is for that vessel. That uncertainty is not fully independent across vessels: many vessels on the same model tier share the same underlying assumptions (fuel density, generator load factors, class-based scaling), so if one of those assumptions is systematically off, it shifts multiple vessels in the same direction — not randomly in different directions.

Treating vessel uncertainties as fully independent (as root-sum-of-squares does) would understate this shared risk; simply adding every vessel's bound (assuming full correlation) is the opposite, maximally conservative case. Wakeprint uses a correlation coefficient of ρ=0.5, treating fleet-level uncertainty as half shared/systematic and half vessel-specific — a deliberately stated, moderate assumption rather than an implicit extreme. The fully independent (ρ=0) and fully additive (ρ=1) bounds remain available alongside this figure for transparency.

AIS reception is uneven. Terrestrial receivers cover coastal areas densely and open water poorly; satellite passes leave gaps of minutes to hours. Transponders can be switched off, misconfigured, or report stale static data.

Positions published here are delayed by at least 15 minutes and the snapshot refreshes at most once per hour. This is deliberately not a realtime tracker.

Gaps are shown as gaps. We do not draw a route through unobserved water, and we do not infer a port call from an absence of signal.

For each observed interval, the model takes the vessel's installed power, hull type and published speed profile, and derives a load factor from the observed speed over ground. Fuel burn per hour follows from that load factor and a specific fuel consumption band for the engine class, producing a low and a high figure rather than a point value.

Stationary time is modelled separately at auxiliary/hotel load, which is why an anchored vessel still carries a non-zero estimate.

Method tiers describe the quality of the inputs: tier A uses vessel-specific published consumption data, tier B a class-level curve, tier C a length-and-tonnage fallback.

Litres are converted to mass using the density of the reported fuel type, then to CO₂ with a fixed well-recognised emission factor per tonne of fuel (MGO and HFO differ). The conversion adds no uncertainty of its own; the range you see is inherited from the fuel estimate.

Flight comparisons are computed from the great-circle distance between two airports, multiplied by a 1.08 route uplift, and a fixed 0.0000616 tonnes CO₂ per passenger-kilometre factor for a long-haul economy seat. The default route is Amsterdam (AMS) to New York JFK, but both origin and destination can be changed. The result is an illustration of scale, not part of the emissions calculation.

The flight factor uses direct CO₂ only — no radiative forcing or other non-CO₂ climate effects — to remain consistent with Wakeprint's yacht CO₂ metric, which likewise excludes non-CO₂ effects. It is derived from the UK DEFRA/DESNZ 2026 Government GHG Conversion Factors long-haul economy figure (0.11704 kg CO₂e per passenger-kilometre, including radiative forcing), with the RF multiplier (~1.9x) backed out: 117.04 / 1.9 ≈ 61.6 g CO₂ per passenger-kilometre.

The snapshot reports three independent metrics and never merges them:

• provider_availability_pct — how much of the window the upstream AIS providers were actually delivering data.
• target_detection_rate_pct — the share of vessels in the target population that were detected at all.
• emissions_temporal_coverage_pct — the share of the window for which a given vessel had usable observations. This is the one the public 24h headline depends on.

Below 70% emissions temporal coverage, no emissions figure is published for that vessel. It appears as insufficient observational coverage instead of a confident-looking estimate.

Fuel in litres is the source datapoint, split into propulsion and onboard services. CO₂ is derived from it: fuel_mass_kg = litres × fuel_density_kg_per_liter and co2_kg = fuel_mass_kg × 3.206 (EU MRV factor for diesel / gas oil). Density is an explicit per-vessel parameter, sourced where published, with a stated default of 0.84 kg/L (source: superyacht_master_batches_1-5_fuel_speed.xlsx, sheet “Methodology”, row “Default diesel density” — modelling assumption only, to be replaced with a fuel-specific density when available) — never a silent constant. Each source datapoint also carries a fuel scope (main engines only, total onboard, or unknown) so generator consumption is not double counted.

Emissions are reported per operational state — underway, at anchor, in port — as time actually spent in that state × the best available state-specific fuel rate × the fuel CO₂ factor. The anchor rate comes from a strict preference order: vessel-specific published generator consumption, then a vessel-specific generator/load model, then a calibrated class model, then a broad fallback. It is never calculated as a percentage of the underway rate; published fleet splits (YETI 10/34/56, Heesen 42–58%) are sanity checks only.

Every estimate carries a status — Observed, Estimated — vessel specific, Estimated — class model, Estimated — broad benchmark — and a method tier. A broad benchmark is never shown with the same visual certainty as a vessel-specific datapoint, and low-confidence figures appear as ranges rather than falsely precise single numbers.

The cruise comparison uses the ICCT figure of 250 g CO₂ per passenger-kilometre for the most efficient cruise ships, derived from EU MRV-verified reporting (for one specific ship, 317 g CO₂ per passenger-nautical-mile). Both sides of the comparison use the same unit: kilograms of CO₂ per passenger over the same observed distance in the same 24h window.

AIS does not report persons on board. The yacht side therefore divides total CO₂ by a charter-certified guest capacity taken from published charter data — 12 guests for the reference vessel, not the 22–24 cabin berths some sources quote — and is labelled Estimated — vessel specific, never observed.

This site previously showed a comparison between cumulative fleet CO₂ and the EU single-use plastics directive's expected savings. It was removed: the two figures used different time bases (a cumulative running total vs. a 2030 projection) and could not be reconciled into a fair comparison. Removed 2026-08-29 — see Data status changelog.

European Commission, COM(2025) 109 final, Table 1 (18 March 2025) — review of Regulation (EU) 2015/757 on the potential inclusion of ships below 5,000 GT but not below 400 GT. Source of the quoted 2023 figures: 896 yachts of 400–4,999 GT calling at EU ports, an estimated 1.45 million tonnes CO₂, equal to 1.1% of all shipping emissions monitored under EU MRV that year. Quoted verbatim on the Impact page and labelled as EU Commission data, never as an output of this model. That population (400–4,999 GT, EU port calls) is not the ≥60 m LOA fleet tracked here, so the two sets of figures are never combined.

Gross tonnage thresholds shown on vessel pages — 300 GT (EU Vessel Traffic Monitoring Directive), 400 GT (proposed MRV/ETS extension) and 5,000 GT (current MRV/ETS scope) — are regulatory reference points, not an inclusion criterion of this tracker. Length alone (≥60 m LOA) decides which yachts are followed here.

No person tracking, no ownership claims, no statements about who is aboard. Capacity is always described as maximum guest berths. The dataset is described as observed superyachts, never as all superyachts.

Figures shown reflect the live observation pipeline, running since 25 August 2026. Coverage is still building toward full 24h windows for the full tracked fleet.