Weather Routing

Background and history

Pre-computational weather routing

Practical weather routing has been used since the early days of sail-powered shipping. The classic example is the trade-wind passages developed by Iberian, Dutch, English and French navigators in the 16th to 18th centuries: ships sailing from Europe to the Americas would deliberately route south to capture the NE trade winds, and route north on the return for the westerlies. The American captain Matthew Fontaine Maury’s Wind and Current Charts (1847) and The Physical Geography of the Sea (1855) systematised pre-computational weather routing using compiled logbook observations, demonstrating fuel and time savings of approximately 30% on US East Coast to South America voyages.

In the steam era through the early 20th century, weather routing continued through compiled climate atlases (the Admiralty Weather Reports, the US Naval Hydrographic Office Pilot Charts) and the daily synoptic charts produced by national weather services. Ships used these to plan the broad route shape but had no real-time ability to adjust mid-voyage based on forecast updates.

1950s to 1980s: synoptic-routing service emergence

The first commercial weather routing service was launched by the US Hydrographic Office in the late 1940s for US Navy ships. Commercial shipping followed in the 1950s with services from the US National Weather Service Maritime Branch (later spun out as Oceanroutes in 1972, founded by James Lewis), the British Met Office marine division, and the Japan Meteorological Agency.

These early services operated on a shore-to-ship advisory model: routing experts at the service provider analysed daily synoptic charts and issued routing recommendations to subscribed ships via radio teletype, typically once per 12 to 24 hours. The recommendations balanced weather avoidance (minimising heavy weather) against route-distance economics. The US Coast Guard’s 1979 Weather Routing Effectiveness Study documented average fuel savings of 3 to 5% across the US flag merchant fleet using these services.

1990s: numerical weather prediction integration

The development of numerical weather prediction (NWP) models in the 1980s and 1990s, particularly the ECMWF Integrated Forecasting System (IFS) and the NOAA Global Forecast System (GFS), revolutionised weather routing. Key developments:

• High-resolution wave forecasting through coupling of NWP atmospheric models with ocean wave models (WaveWatch III, WAM).
• Ensemble forecasting providing probabilistic uncertainty estimates for routing decisions.
• Increased forecast horizon: from 3-day reliable forecasts in the 1980s to 7-10 day reliable forecasts by 2000.
• Vessel performance polars developed by class societies and universities (notably the University of Strathclyde’s marine vehicles research) enabled mathematical optimisation of routes.

The 1995 establishment of Applied Weather Technology (AWT) in California, providing a fully digital ship-to-shore routing service, marked the transition to the modern era of weather routing.

2000s to 2020s: real-time and onboard systems

The 2000s and 2010s saw further evolution:

• Onboard routing software providing 24/7 routing capability without shore-based latency (e.g. AWT’s Bon Voyage System licensed to StormGeo in 2012, now BVS).
• Real-time satellite weather data via INMARSAT and Iridium broadband connections.
• Vessel-specific performance polars built from ISO 19030 standardised hull-and-propeller performance monitoring data.
• Integration with ECDIS and bridge nautical equipment for one-screen routing display.
• Integration with shipowner / charterer management systems for fleet-wide optimisation.

By 2020 approximately 80% of the deep-sea commercial fleet was subscribed to some form of commercial weather routing service.

2020 onwards: AI-augmented routing

The 2020 to 2024 period has seen the emergence of AI-augmented routing using machine learning to optimise the multi-objective routing problem. AI techniques are particularly useful for:

• Improving the vessel performance polar by learning from actual operational data rather than published shop tests.
• Predicting probability of cargo damage under different routing options.
• Optimising fleet-wide routing considering port congestion, cargo readiness and onward logistics.

The leading AI routing vendors include AWT, StormGeo, Wärtsilä Voyage Solutions (with the Eniram acquisition), ABS NS Voyage Manager and emerging entrants including Maersk Routing (in-house), Cargill Ocean Transportation Routing (in-house), Hellesøe Routing (Danish startup) and Sofar Ocean (US ocean-data startup).

Methodology

Numerical weather prediction inputs

Modern weather routing depends on global NWP models providing forecasts on a regular grid:

• ECMWF IFS (European Centre for Medium-Range Weather Forecasts): the global gold standard, run twice daily with 9-day forecast horizon at 0.1° resolution.
• NOAA GFS (Global Forecast System): the US public-domain model, run four times daily at 0.25° resolution.
• NOAA WaveWatch III: ocean-wave forecasting coupled with GFS.
• ECMWF Wave Model (WAM): wave forecasting coupled with IFS.
• HYCOM (Hybrid Coordinate Ocean Model): operational ocean model providing currents.
• ETOPO (Earth Topography): bathymetry for shallow-water routing.

The forecast data is downloaded via the WMO Global Telecommunications System (GTS) by routing service providers, processed onto vessel-specific routing grids, and refined for marine conditions.

Vessel performance polars

A vessel performance polar is a tabular or functional description of the ship’s fuel consumption as a function of:

• Service speed (typically 4 to 25 knots).
• Draft (laden vs ballast vs partial load).
• Wave height (Hs): significant wave height in metres.
• Wave period (Tp): peak wave period in seconds.
• Wave direction relative to ship: bow, beam, quarter, stern seas.
• Wind speed and direction: similar.
• Hull condition: clean vs partially fouled vs heavily fouled.
• Trim: optimum trim or fixed trim.

The polar is typically constructed from a combination of:

• Sea trial data (ISO 15016 standardised conditions; corrected to standard conditions).
• Operational data from past voyages (ISO 19030 hull-and-propeller performance monitoring).
• CFD (Computational Fluid Dynamics) simulations for conditions outside the empirical envelope.

The performance polar is the critical input that distinguishes weather routing services: better polars produce better routing recommendations.

The optimisation objective

Before any algorithm runs, the operator has to fix what the route is being optimised for. The objective is not a single thing, and the choice changes the answer. Four classic objective formulations are in routine commercial use, and they don’t agree with each other.

Least-time routing finds the track that reaches the destination soonest given the forecast field and the ship’s speed-loss behaviour in waves. It’s the default for a liner sailing to a fixed berthing slot, or a vessel under a charter party with a tight cancelling date. Least-time routing will accept higher fuel burn to claw back hours.

Least-fuel routing minimises total fuel mass over the voyage and lets the arrival time float within an agreed window. This is the objective that matters for annual CII attained and for a spot charterer paying the bunker bill. Least-fuel and least-time are the same route only in calm, current-free water; in a real forecast field they diverge, often by several hours and several tonnes.

Constant-power (constant-RPM) routing holds the engine at a fixed power or shaft speed and lets the ship’s speed vary with the sea state. The route is chosen so the resulting speed profile arrives on time. Many engines run most efficiently at a steady load, so constant-power operation can beat a route that keeps demanding speed corrections in a seaway. WMO-No. 558 defines the optimum-track service against exactly this trade-off: the least-time, least-fuel, or least-damage track for the ship’s own characteristics, not a generic great circle.

Constant-speed routing holds speed-over-ground (or through-water) fixed and lets power and fuel rate vary with the weather. It’s the simplest to schedule against and the easiest to verify on arrival, which is why charter-party speed warranties are usually written as constant-speed clauses. It’s also the least fuel-efficient of the four in a heavy seaway, because holding speed into added resistance forces the engine up its power curve.

Layered on top of the primary objective are the hard constraints: meet the ETA window, stay inside the load-line seasonal zones, and never route the ship into conditions the master has declared off-limits (the safety constraint, below). The routing engine treats these as bounds on the feasible set, not as terms in the cost function.

Added resistance in waves and the power model

The reason weather routing works at all is that waves add resistance, and that added resistance is highly directional and strongly non-linear in wave height. A routing engine that ignored it would route on distance alone and lose money.

Added resistance in waves, $R_{aw}$, is a second-order (mean) force: it scales with the square of wave amplitude, not linearly. ITTC Recommended Procedure 7.5-02-07-02.2 computes the mean added resistance in an irregular sea by linear superposition of the regular-wave response over the wave spectrum:

$R_{aw} = 2 \int_0^\infty \frac{R_{aw}(\omega)}{\zeta_a^2}\, S(\omega)\, d\omega$

where $S(\omega)$ is the wave energy spectrum (Pierson-Moskowitz, JONSWAP or the ITTC two-parameter spectrum), $\omega$ is wave frequency, $\zeta_a$ is wave amplitude, and $R_{aw}(\omega)/\zeta_a^2$ is the quadratic transfer function of mean added resistance measured in regular waves or computed by panel/CFD methods. Because the response is quadratic, doubling significant wave height roughly quadruples the added-resistance contribution from that sea state, which is why a 2 m sea costs little and a 5 m head sea costs a lot.

That added resistance enters the propulsion power model directly. In calm water the delivered power to hold a given speed follows roughly $P_D \propto V^3$ (the admiralty-coefficient relation; see admiralty coefficient and the cube-law fuel calculator). In a seaway the engine must overcome calm-water resistance plus added resistance from waves, wind, and steering. The required delivered power becomes approximately $P_D = (R_{calm} + R_{aw} + R_{wind})\, V / \eta_D$, where $\eta_D$ is propulsive efficiency. The ship responds in one of two ways: hold speed and burn more fuel (involuntary plus voluntary power increase) or hold power and lose speed (involuntary speed loss). The performance polar captures both responses across the full directional envelope. ISO 15016:2015 supplies the calm-water and wind-correction baseline; the wind resistance ISO 15016 calculator implements the aerodynamic term.

The speed-power baseline against which all of this is measured is the ship’s own reference curve. ISO 19030-2:2016 names ship speed through water $V$ and delivered power $P_D$ as the two primary measurement parameters and defines hull-and-propeller performance change as the change in delivered power needed to hold a given speed under unchanged conditions. The standard sets a speed-log accuracy floor of plus or minus 1% at one standard deviation, stricter than the SOLAS minimum, because the whole performance signal is buried in small speed differences. A routing polar built on a degraded or out-of-date reference curve produces optimistic recommendations the ship can’t actually meet.

Routing optimisation algorithms

With the objective and the power model fixed, the routing engine searches the space of feasible tracks. The principal algorithms:

• Isochrone method (James, 1957): from the departure point, the engine projects the set of points reachable within one time step under the forecast field, then projects the next isochrone from that frontier, and so on to the destination. Each isochrone is the locus of furthest progress at constant time. The method is fast and memory-light, which is why it dominated the shore-to-ship advisory era and still backs many onboard systems. Its weakness is that the naive form can lose the global optimum when the reachable set folds back on itself near coasts or in strong opposing currents.
• Dynamic programming (Bellman, 1957): the route space is discretised into a grid of waypoints and arrival times, and the engine finds the globally optimal track by backward induction over Bellman’s principle of optimality. It’s heavier than isochrones but returns a provably optimal route for the discretised problem, and it handles the time dimension (a moving forecast field) cleanly.
• Graph-search / Dijkstra and A*: the modern dominant approach treats the ocean as a weighted directed graph. Each edge between grid nodes carries a cost (fuel, time, or motion penalty) computed from the forecast field and the performance polar at the time the ship would traverse that edge. Dijkstra’s algorithm returns the least-cost path; A* speeds it up with an admissible heuristic. Because the edge weights are time-dependent (the forecast changes as the ship advances), the graph is a space-time lattice, not a static map. This formulation handles arbitrary cost functions and hard constraints (no-go cells for excessive $H_s$, parametric-roll zones, load-line boundaries) by setting infinite or forbidden edge weights.
• Genetic and evolutionary algorithms: stochastic search suited to many-objective formulations where the Pareto frontier between fuel, time, and motion is wanted rather than a single optimum.
• Machine learning: trained on historical voyages to correct the published polar toward the ship’s measured behaviour, and to predict the probability of motion-related damage on a candidate track.

Most commercial services run a deterministic backbone (isochrone, dynamic programming, or a space-time graph search) and layer machine learning on top to sharpen the polar and the risk estimate. The deterministic core guarantees a feasible, constraint-respecting route; the learned layer improves the cost estimates that route is judged against.

Forecast uncertainty handling

Weather forecasts have intrinsic uncertainty that grows with forecast horizon:

• 3-day forecast: typical wave-height uncertainty ±0.5 m (good).
• 5-day forecast: ±1.0 m (moderate).
• 7-day forecast: ±1.5 m (poor).
• 10-day forecast: ±2.0 m (effectively unusable for tactical decisions).

Modern routing services handle this through:

• Ensemble forecasting: running multiple NWP forecasts with perturbed initial conditions to estimate uncertainty.
• Adaptive routing: re-running the routing optimisation as new forecasts arrive (typically every 6 to 12 hours).
• Risk-adjusted routing: using the route with the lowest expected fuel consumption rather than the lowest deterministic fuel consumption.

Performance and economics

Real-world fuel savings

Reported fuel savings from commercial weather routing services:

The savings vary with three factors:

• Trade-route weather variability: routes with consistent weather (Caribbean, Mediterranean) achieve smaller savings; routes with variable weather (North Atlantic, North Pacific) achieve larger.
• Vessel performance polar quality: ships with detailed polars (bulk, tanker, container) achieve better routing than ships with limited polars (cruise, RoRo where polar is less critical to fuel cost).
• Routing service quality: top-tier services (AWT, StormGeo) typically achieve 1 to 2 percentage points better than lower-tier services.

Capital and operational cost

Weather routing is essentially a subscription service:

• Fleet subscription: USD 5,000 to USD 50,000 per ship per year, depending on service level and number of ships.
• Per-voyage subscription: USD 500 to USD 2,500 per voyage (less common; mostly for spot-charter operators).
• Onboard software licence: USD 3,000 to USD 25,000 per ship per year (e.g. StormGeo BVS, AWT BonVoyage).
• Crew training: minimal; bridge teams are familiar with routing principles.

The annual fuel saving (for a typical container ship consuming 30,000 t/yr at USD 600/t) of 3 to 5% = USD 540,000 to USD 900,000.

The payback period is therefore essentially immediate (less than 1 month). Weather routing has been the lowest-cost decarbonisation lever available for decades.

CII improvement

A 3 to 5% per-voyage saving translates into approximately the same improvement in annual CII attained:

• For a bulk carrier with attained CII of 5.5 (D rating, 10% above Required), a 4% routing improvement brings attained CII to ~5.28 (still D but closer to C boundary).
• Combined with slow steaming (10%), JIT arrival (4%) and hull cleaning (5%), the combined improvement is approximately 21% (using the non-overlapping multiplicative formula).

Regulatory and commercial drivers

Weather routing was a pure cost lever for most of its history: you used it because fuel was expensive. Since 2023 it’s also a regulated efficiency measure, which changes who pays attention to it and why.

The SEEMP Part III and CII driver

The Ship Energy Efficiency Management Plan (SEEMP) is mandatory under MARPOL Annex VI Regulation 26. Since 1 January 2023 every cargo ship, ro-pax, and cruise ship above 5,000 GT on international voyages must carry a SEEMP Part III, the Ship Operational Carbon Intensity Plan, verified and accompanied by a Confirmation of Compliance. The Part III documents how the ship will hit its required Carbon Intensity Indicator and lists the operational measures it will use to get there.

IMO resolution MEPC.346(78), the 2022 SEEMP guidelines, names weather routeing explicitly in the menu of operational measures, stating that it “has a high potential for efficiency savings on specific routes and is commercially available for all types of ship and for many trade areas.” It sits alongside speed optimisation, defined in the same guidelines as running at “the speed at which the fuel used per tonne mile is at a minimum level for that voyage.” That pairing matters: a least-fuel route at a non-optimum speed leaves savings on the table, which is why the routing engine and the speed plan are solved together (see slow steaming and the JIT economic-speed calculator).

The commercial driver is the CII rating itself. A ship’s attained CII feeds an A-to-E band published annually; three consecutive years at D, or one year at E, force a corrective action plan into the SEEMP. A 3 to 5% routing saving moves the attained CII by roughly the same percentage, which is often enough to hold a borderline ship inside C rather than slipping to D. The CII corrective trajectory calculator and the SEEMP combined operational measures calculator model where routing sits in the stack of levers a ship uses to defend its band.

Beyond the IMO regime, the same routing data feeds the charterer- and lender-side frameworks: the Sea Cargo Charter for cargo buyers, the Poseidon Principles for ship-finance banks, and per-vessel ratings such as the RightShip GHG rating . A ship that routes well reports a better intensity to all of them off the same voyage record.

The safety and load-line constraints

Routing is bounded by two regulatory constraints that the cost function cannot trade away. The first is the International Convention on Load Lines: a route may not cross into a seasonal load-line zone in a way that would leave the ship overloaded for that zone and season. A least-distance great circle in winter can clip the Winter North Atlantic seasonal area; the routing engine has to keep the track legal for the ship’s actual loaded condition.

The second is the master’s overriding authority. IMO’s weather-routeing framework, set out in resolution A.528(13) and the minimum-standards circular MSC/Circ.1063, is built around advice, not instruction. MSC/Circ.1063 ties any routeing service to the voyage-planning duty in SOLAS Chapter V Regulation 34 and states that the service must account for “the speed and handling characteristics of the ship” and that the master must be given “the source of data” and, where possible, its accuracy. Regulation 34.3 safeguards the master’s right to deviate from advice that conflicts with professional judgement, and the ISM Code reinforces the master’s overriding authority to take any decision the master considers necessary for safety. A shoreside router recommends; the master on the bridge decides and carries the liability.

Notable deployments

Maersk routing

Maersk operates one of the largest in-house weather routing capabilities, with a dedicated routing centre at the company’s Copenhagen HQ. The centre:

• Provides routing for the entire Maersk container fleet (~700 vessels by 2024).
• Integrates with Maersk’s onboard performance monitoring.
• Reports approximately 5 to 7% average annual fuel saving across the fleet.
• Has been credited with approximately USD 200 million annual fuel cost savings.

Cargill Ocean Transportation routing

Cargill as the world’s largest dry bulk cargo buyer operates an in-house routing capability for chartered vessels:

• Partnership with AWT for technical routing services.
• Integrates with Sea Cargo Charter reporting.
• Reports 4 to 7% fuel savings on Atlantic and Pacific bulk carrier voyages.
• Uses the savings to support the Sea Cargo Charter alignment reporting.

StormGeo BVS deployment

The StormGeo BVS (Bon Voyage System) is the most widely-used commercial routing service, with approximately 9,000 vessel subscriptions globally:

• Acquired from AWT in 2012.
• Extended with cloud-based forecast updates from 2018.
• Integration with Wärtsilä Voyage Solutions from 2024.
• Reports average 4 to 6% fuel savings across subscribed vessels.

COSCO Shipping routing

COSCO Shipping (the Chinese state-owned shipping conglomerate) operates an in-house routing capability for the COSCO container, bulk and tanker fleet:

• Partnership with Chinese state weather service (China Meteorological Administration).
• Approximately 1,000 vessels routed.
• Reports 3 to 5% fuel savings.

Related operational measures

Weather routing combines naturally with several other operational measures:

Slow steaming

Slow steaming reduces vessel speed below design speed for fuel efficiency. Combined with weather routing, the optimum route can be re-calculated for the slower target speed, often producing further savings (the slower vessel can take more advantageous routes that would be sub-optimal at high speed).

Just-In-Time arrival

JIT arrival coordinates port arrival to avoid anchor wait. Combined with weather routing, the routing service can optimise speed to arrive at JIT-precise times while still routing around weather.

Trim optimisation

Trim optimisation adjusts the ship’s longitudinal trim to minimise resistance. Combined with weather routing, the trim can be adjusted dynamically based on the expected sea conditions on each voyage segment.

Hull cleaning

Regular hull cleaning maintains the vessel performance polar at “clean” rather than “fouled” levels. Weather routing assumes the ship is at its baseline polar; degraded polars produce lower routing benefits.

The SEEMP combined operational measures calculator implements the combined effect using the standard non-overlapping multiplicative formula.

Safety considerations

Weather routing has critical safety implications beyond fuel optimisation:

Heavy weather avoidance

Weather routing routinely avoids ships from being caught in:

• Tropical cyclones (typhoons, hurricanes, cyclones): IMO Resolution A.893(21) on Voyage Planning explicitly references avoidance of tropical-storm tracks.
• Extra-tropical depressions (winter North Atlantic, North Pacific): typically Beaufort 9-12 conditions exceeding most vessel design limits.
• Polar lows and Mediterranean low: smaller-scale intense storms.
• Squalls associated with the inter-tropical convergence zone.

The IMO and major flag states explicitly recommend the use of weather routing for voyage planning under SOLAS Chapter V Regulation 34 (passage planning).

Dynamic stability phenomena and the MSC.1/Circ.1228 thresholds

The strongest safety argument for routing isn’t avoiding storms in general; it’s avoiding the specific wave-encounter geometries that trigger dangerous ship motions. IMO’s MSC.1/Circ.1228, the 2007 Revised Guidance to the Master for Avoiding Dangerous Situations in Adverse Weather and Sea Conditions, names four phenomena with concrete numeric thresholds, and a routing engine encodes each one as a no-go condition in the cost graph.

Surf-riding and broaching-to happen in following and quartering seas. MSC.1/Circ.1228 puts the danger zone at encounter angles $135° < \alpha < 225°$ (waves coming from astern) when ship speed exceeds about $1.8\sqrt{L}/\cos(180° - \alpha)$ knots, with $L$ the ship length. Inside that zone a wave can carry the ship along on its face, the rudder loses grip, and the ship slews beam-on. The guidance is blunt: take the speed, the course, or both outside the dangerous region.

Reduction of stability on the wave crest is acute when the wavelength sits in roughly the 0.6 L to 2.3 L band. As the crest passes amidships the waterplane narrows and righting arm drops; combined with following-sea speeds it sets up the conditions for the other phenomena.

Synchronous rolling occurs when the encounter period $T_E$ falls near the ship’s natural roll period $T_R$. A single resonant match can build large roll amplitudes in beam seas.

Parametric rolling is the one that has cost container operators stacks of boxes overboard. MSC.1/Circ.1228 flags it when the average wavelength exceeds 0.8 L and the significant wave height exceeds 0.04 L, with the encounter period near $T_R$ (the 1:1 case) or near $0.5\,T_R$ (the 1:0.5 case, the classic head-sea parametric roll). The mechanism is the stability variation between crest and trough pumping the roll, not a direct resonance with the wave force. The guidance directs the master to change course and speed to move $T_E$ away from $T_R$ and $0.5\,T_R$. A routing engine implements this by computing $T_E$ along every candidate edge and forbidding the edges that fall inside the parametric window. See seakeeping for the underlying motion dynamics.

Slamming and green-water avoidance

Bow-flare slamming and bottom slamming load the forward structure with impact pressures that can crack plating and, on container ships, accelerate fatigue. Slamming probability rises sharply when the bow emerges and re-enters in steep head seas, so the polar carries a slamming-probability penalty against head-sea cells above a wave-height threshold. Green water on deck, solid water shipped over the bow, threatens deck cargo, hatch covers, and forward equipment. Routing trims both by keeping the ship out of the steepest head-sea combinations or by recommending a speed reduction that drops the relative bow motion below the slamming threshold. The trade-off is explicit: a head-sea diversion or slowdown costs fuel and time, and the master weighs that against the structural and cargo risk.

Cargo motion damage

For cargoes susceptible to motion damage (project cargo, vehicles, refrigerated cargo, sensitive industrial equipment), weather routing optimises for minimum motion exposure rather than minimum fuel. The trade-off is typically 1 to 3% additional fuel cost in exchange for substantial reduction in cargo damage claims.

Crew safety

Beaufort 8+ conditions present risks to crew (particularly during deck operations, maintenance and lifeboat drills). Weather routing reduces crew exposure to severe weather, contributing to safety performance metrics tracked by P&I clubs and operator HSE departments.

Shoreside routing providers versus onboard decision support

Routing comes in two delivery models, and most fleets run both.

The shoreside advisory model is the original. A team of marine meteorologists and master mariners at the provider analyses the forecast field, runs the optimisation against the ship’s polar, and sends a recommended route plus weather briefings to the bridge. The strength is human judgement: an experienced router reads a developing depression, weighs forecast confidence, and can override the algorithm when the model is behaving badly. The weakness is latency and coverage. Advisories arrive on a 12-to-24-hour cadence in the classic form, faster over modern broadband, but the bridge still depends on the next message to react to a fast-changing field.

The onboard decision-support model puts the routing engine on the ship. Software such as the BVS-type onboard systems pulls forecast updates over INMARSAT or Iridium broadband, runs the optimisation locally against the loaded polar, and lets the bridge team re-run the route at any hour without waiting for shore. The strength is immediacy and 24/7 availability; the weakness is that the quality of the answer rests entirely on the polar loaded and the forecast pulled, with no human router in the loop to catch a model that’s diverging from reality.

In practice the two are complementary, not competing. A typical deep-sea operator subscribes to a shoreside advisory for the strategic route and the human read on developing weather, and runs onboard software for tactical re-routing between advisories. MSC/Circ.1063 anticipates this split: it requires the service to be interactive, with the master reporting position, course, speed, and weather observations back so the next advisory is grounded in the ship’s actual state, not the forecast alone.

Limitations

Weather routing is a forecast-driven optimisation, and every weakness it has traces back to that fact.

Forecast skill caps the achievable saving. A route is only as good as the forecast it was optimised against. Wave-height uncertainty grows with horizon, from roughly half a metre at three days to a metre at five and worse beyond, so a route planned on a 7-day field can be substantially wrong by the time the ship gets there. Re-optimising every 6 to 12 hours as new forecasts arrive helps, but it can’t recover a saving that depended on a feature the model placed in the wrong spot. The headline figures in the savings table are voyage-average outcomes across many sailings, not a guarantee on any single voyage; a ship can route perfectly against the forecast and still meet weather the model missed.

The polar is the largest error source. The optimisation takes the performance polar as ground truth, and a stale or optimistic polar produces routes the ship can’t meet. ISO 19030-2 frames hull-and-propeller performance change as a change in delivered power at constant speed, and over a docking interval that change can run well into double-digit percentages as fouling builds. A polar fixed at the clean condition will under-predict fuel and over-predict speed in a seaway, so the realised saving falls short of the modelled one. Polars built from sea-trial data alone (ISO 15016) miss the in-service degradation that ISO 19030 monitoring is designed to catch.

The numbers vary widely and are largely vendor-reported. The fuel-saving percentages quoted across the industry sit in wide bands (2 to 10% depending on route and season) and most originate from the routing providers themselves, who have an interest in the high end. Independent, like-for-like measurement is hard because the counterfactual, the fuel the same ship would have burned on the route it didn’t take, is never observed. Treat any single percentage as an estimate with a wide confidence interval, not a measured fact.

The saving is route- and season-dependent, and sometimes near zero. Routes with steady weather (much of the Mediterranean, the Caribbean) leave little for the optimiser to work with, and the saving collapses toward the 1 to 3% floor. The large savings come from variable, energetic basins (winter North Atlantic, North Pacific, the Southern Ocean approaches), and even there a calm season returns little. A fleet that routes every voyage will see a wide spread of outcomes, many of them small.

It optimises one ship at a time. Classic routing solves for a single vessel against a forecast. It doesn’t natively account for port congestion at the far end, berth availability, or cargo readiness, which is why a least-time route can still end in days at anchor. Combining routing with just-in-time arrival closes that gap, but only where the port and the terminal share an arrival window in advance, which many still don’t.

Safety constraints override the economics, by design. The master’s overriding authority means the routing recommendation is advisory. When the master rejects a least-fuel route because the forecast confidence is poor or the ship is laden in a way the polar doesn’t reflect, the modelled saving doesn’t materialise, and that’s the system working correctly. The dynamic-stability thresholds in MSC.1/Circ.1228 carve no-go zones out of the feasible set regardless of fuel cost, so the achievable saving is always bounded below the unconstrained optimum.

Data and connectivity dependence. Onboard re-routing needs current forecasts, which needs working satellite broadband; shoreside advisory needs the ship to report back. A communications gap leaves the bridge routing on a stale field or a stale advisory. The interactive loop MSC/Circ.1063 requires breaks down quietly when the link does.

Future outlook

By 2030 weather routing is expected to:

• Cover approximately 95% of deep-sea commercial fleet (vs ~80% in 2024).
• Achieve average 4 to 6% per-voyage fuel savings (vs ~3 to 4% currently) through improved vessel performance polars and AI-augmented routing.
• Be increasingly integrated with JIT arrival through digital port-coordination platforms.
• Be increasingly integrated with Sea Cargo Charter and Poseidon Principles reporting through standardised data interfaces.

By 2040 weather routing combined with autonomous routing decisions (subject to MASS regulatory framework progress at the IMO ) is expected to enable continuous re-routing without human-in-the-loop latency, achieving an additional 1 to 2 percentage points of fuel savings.

Related Articles

• Wind-Assisted Propulsion - parallel technical efficiency measure

• Air Lubrication Systems - parallel technical efficiency measure

• Just-In-Time Arrival - operational measure combined with weather routing

• Slow steaming and CII - operational measure complementing weather routing

• What is CII - operational index improved by weather routing

• What is EEDI - design-phase index (not credited for weather routing)

• What is EEXI - existing-ship index (not credited for weather routing)

• SEEMP I, II and III - operational plan documenting weather routing

• EEXI EPL and ShaPoLi - EEXI compliance levers

• CII Corrective Action Plan - corrective measures combining weather routing

• MARPOL Annex VI - parent regulation

• IMO GHG Strategy - policy framework

• IMO Net-Zero Framework - GFI standard from 2027

• EU ETS for shipping - EU cap-and-trade

• FuelEU Maritime explained - parallel intensity regime

• Poseidon Principles - bank-side framework

• Sea Cargo Charter - cargo-buyer-side framework

• RightShip GHG Rating - per-vessel rating

• Green Shipping Corridors - operational corridors using weather routing

• EUA Market Mechanics for Shipping - allowance market

• Voluntary Carbon Credits in Shipping - parallel mechanism

• CARB At-Berth Regulation - California regional regime

• China DCS - China’s national reporting regime

• UK ETS for shipping - UK cap-and-trade

• EU MRV Regulation 2015/757 - reporting framework

• IMO DCS vs EU MRV - reporting comparison

• Cold ironing and shore power - in-port emission reduction

• Emission Control Areas - regional sulphur and NOx framework

• NOx Tier I, II and III - engine certification regime

• IMO 2020 sulphur cap - global sulphur cap

• Biofuels in shipping - low-carbon fuel pathway

• LNG as marine fuel - dual-fuel pathway

• Methanol as marine fuel - alternative pathway

• Ammonia as marine fuel - zero-carbon pathway

• Heavy fuel oil - residual fuel

• Marine gas oil - distillate fuel

• Specific fuel oil consumption - engine efficiency metric

• Marine diesel engine - main propulsion benefiting from routing

• LNG fuel system - dual-fuel ship handling

• MARPOL Convention - parent IMO treaty

• SOLAS Convention - principal IMO safety treaty (Chapter V passage planning)

• STCW Convention - training and watchkeeping standards

• COLREGs Convention - parallel IMO instrument (collision avoidance)

• Bulk carrier - principal beneficiary of weather routing

• Oil tanker - significant beneficiary

• Container ship - principal beneficiary

• Ro-ro vessel - moderate routing benefit

• Chemical tanker - routing benefit

• LNG carrier - routing benefit

• Voyage charter party - typical contract type

• Time charter party - alternative contract type

• Port state control - parallel federal enforcement framework

• Classification society - performance polar verification

• Flag state and flag of convenience - flag-state role

• Trim Optimisation

• Seakeeping

References

1. International Standards Organisation. ISO 19030:2016 - Ships and marine technology - Measurement of changes in hull and propeller performance. ISO, Geneva, 2016.
2. International Standards Organisation. ISO 15016:2015 - Ships and marine technology - Guidelines for the assessment of speed and power performance by analysis of speed trial data. ISO, Geneva, 2015.
3. IMO Resolution A.893(21). Guidelines for Voyage Planning. IMO, 25 November 1999.
4. James, R.W. A New Method of Determining Optimum Tracks for Ship Routing. Memo, Naval Hydrographic Office, 1957.
5. Bellman, R. Dynamic Programming. Princeton University Press, 1957.
6. Maury, M.F. The Physical Geography of the Sea. Harper, 1855.
7. ECMWF. Integrated Forecasting System Documentation, Cycle CY49R1. ECMWF, Reading, 2024.
8. NOAA. Global Forecast System (GFS) Documentation. NOAA / NCEP, 2024.
9. Applied Weather Technology. Annual Routing Service Report 2024. AWT, Sunnyvale CA, 2024.
10. StormGeo. BVS Annual Report 2024. StormGeo, Bergen, 2024.
11. MeteoGroup / DTN. Marine Routing Annual Report 2024. DTN, Minneapolis, 2024.
12. Wartsila Voyage Solutions. Annual Report 2024. Wartsila, Helsinki, 2024.
13. Sofar Ocean. Annual Report 2024 - Marine Sector. Sofar Ocean, San Francisco, 2024.
14. Cargill. Ocean Transportation Annual Report 2024. Cargill, Geneva, 2024.
15. DNV. Maritime Forecast to 2050 - Operational Efficiency Section. DNV, Oslo, 2025 edition.
16. Lloyd’s Register. Weather Routing: Practical Implementation Guide. Lloyd’s Register Marine, London, 2024.

Further reading

• IMO Marine Environment Division. Voyage Planning and Weather Routing Guidance. IMO, 2018.
• DNV. Maritime Forecast to 2050. DNV, Oslo, 2025 edition.
• Lloyd’s Register. Performance Management of Marine Vessels. Lloyd’s Register Marine, London, 2023.