Slow Steaming: speed, fuel and carbon intensity

What slow steaming is

Slow steaming is the deliberate, sustained operation of a merchant ship below its contracted design speed, chosen by the operator to cut fuel consumption, cost or emissions rather than forced by weather, draft or defect. It is defined against each ship’s own design point, not against an absolute number of knots, which is why a container ship running a liner string at 18 knots and a Capesize bulker crossing at 11 knots are both doing the same thing.

The design speed is the speed the yard contracted at a stated draft with a stated sea margin. Everything below that is a band of operating practice rather than a single number. A 9,000 TEU ship drawn for 25 knots that runs at 18 is slow steaming; so is a bulker built for 14.5 knots that loads iron ore and crosses to Qingdao at 11.

The reason any of it pays is a single physical relation, and the reason it persists is that three separate systems now push in the same direction: the fuel bill, the CII rating , and the charter party.

The speed-power relation

For a displacement hull in its normal operating range, the power the machinery must deliver climbs with roughly the cube of speed. Cut speed and power falls far faster than the speed itself, which is what makes three knots off a 24-knot ship worth tens of thousands of dollars a day in bunkers.

The Admiralty relation expresses the same idea as a scaling: delivered power varies as displacement to the two-thirds times the cube of speed, divided by a coefficient that is treated as constant for one hull in one loading condition.

Three assumptions sit inside that relation and the article states them because most presentations do not. Displacement to the two-thirds is a proxy for wetted surface, so the relation holds only while the wetted-surface-to-displacement relationship is unchanged. The Admiralty coefficient is not a physical constant and is not comparable between ships, only between conditions of one ship. And it embeds a constant propulsive coefficient, which is not true across a wide speed range.

Two resistance components drive the exponent. Frictional resistance, the larger share on a slow full-form hull, follows roughly the square of speed, though the friction coefficient itself falls slowly with Reynolds number, so friction grows slightly slower than the square. Residuary resistance, dominated by wave-making, grows far faster on the steep flank of its curve but is a small share at merchant operating speeds. Power is resistance multiplied by speed, so a squared resistance gives a cubed power. The background is in ship resistance and powering , hull form design and block coefficient .

The exponent is not always three

The fitted exponent is a measured quantity, not a constant, and the measurements disagree in a way that matters to anyone projecting a saving.

Calm-water and design-condition fits cluster near 3 to 3.5. Wang and Meng put container ships at 2.7 to 3.3, leg-dependent. Kontovas and Psaraftis put them at 4 or above over 20 knots. Against that, Adland, Cariou and Wolff (2020) fitted noon reports from 16 crude oil tankers with a speed-dependent elasticity and found values below 0.5 at low speed, near 2 at mid speed, and above 3 only near design speed.

Psaraftis and Lagouvardou (2023) argue those low in-service figures are artefacts of treating speed and weather as independent regressors, and that calm-water resistance cannot have an exponent below 2 on hydrodynamic grounds. The dispute is unresolved. What a practitioner should take from it is that a cube-law projection is a design-condition estimate, and that the further below design speed the ship runs, the more the real curve flattens toward the square. The measurement problem is treated separately in the limits of the cubic speed-power law .

One correction to a claim that circulates widely. Merchant hulls do not operate anywhere near a wave-making hump. A Panamax bulker at 14.5 knots on 225 m between perpendiculars sits at a Froude number of 0.159; a VLCC at 15 knots on 320 m at 0.138; a 14,000 TEU ship at 22 knots on 366 m at 0.189; a 25-knot 9,000 TEU box ship reaches only 0.230. The fleet is entirely in the friction-dominated regime, and slowing moves it further from the residuary hump. “Hull speed”, a displacement-yacht heuristic at a Froude number near 0.4, has no application here.

Instantaneous power against voyage fuel

The cube governs the rate at which fuel leaves the tank, not the total burned over a fixed distance. Those differ because a slower ship takes longer to cover the same miles, and a longer voyage means more hours of burning. The distinction is where most back-of-envelope estimates go wrong, and it is why the speed decision belongs inside a full voyage estimate rather than in a fuel calculation on its own.

Take a ship dropping from 20 to 16 knots, a 20 percent cut. Instantaneous power and fuel rate fall by 1 minus 0.8 cubed, about 48.8 percent. But voyage time stretches by 20/16, so the fuel burned over the leg is the rate multiplied by the time: a cubed rate against a reciprocal time leaves a squared relation. Voyage fuel falls by 1 minus 0.8 squared, or 36.0 percent.

So the headline saving on a fixed route at a 20 percent speed cut is 36 percent, not 49 percent. Push the cut to 30 percent and voyage fuel falls 51.0 percent; a 40 percent cut gives 64.0 percent; a 50 percent cut 75.0 percent. The marginal saving per knot shrinks as speed falls, which is why there is a practical floor rather than an open-ended race to zero.

Because cargo per voyage is unchanged, that percentage is also the cut in CO2 per voyage and in grams of CO2 per tonne-mile. It is not a cut in well-to-wake intensity , which is measured in grams of CO2 equivalent per megajoule of energy used and is a property of the fuel. Slowing down does not move it by a single gram. That distinction decides how the EU instruments behave and is taken up below.

The auxiliary load does not follow the cube

The cubic law applies to propulsion only. Generators feeding the accommodation, pumps, reefer plugs, cargo systems and bow thrusters draw a roughly steady load per hour that does not fall when the main engine throttles back, so a longer voyage burns proportionally more auxiliary fuel.

Table 17 of the Fourth IMO GHG Study 2020 gives the at-sea figures: 410 kW for a 100,000 to 199,999 dwt bulk carrier with no boiler at sea, 860 kW plus a 300 kW boiler for a 200,000 dwt oil tanker , 2,300 kW for a 14,500 to 19,999 TEU container ship , and 6,750 kW plus a 300 kW boiler for a 200,000 cubic metre gas carrier.

Set against propulsion, that is roughly 4 to 6 percent at design speed, 10 to 13 percent at a 20 percent speed cut, and 30 to 50 percent at literal half speed. The commonly quoted “12 to 20 percent when slow steaming” corresponds to a 20 to 25 percent speed cut, which is what most slow steaming actually is, not to half speed.

The LNG carrier is the segment that gains least, and the usual explanation is wrong. It is not hotel load. Tank heat ingress generates boil-off gas at a rate set by the cargo and the ambient conditions, independent of speed, so it is a sunk energy cost that slowing cannot avoid. The classic design point sized combined gas consumption at 19 to 21 knots to match natural boil-off, and below that the ship makes more gas than it can burn. In one 150,000 cubic metre dual-fuel diesel-electric case with a reliquefaction plant, surplus boil-off for propulsion runs out at about 13.7 knots, and the excess must be reliquefied at roughly 1 kWh per kilogram. The mechanism is set out in boil-off gas management .

A worked voyage

Put numbers on it. Take a Panamax bulk carrier on a 8,500 nautical mile leg, design speed 14.5 knots, main engine burning 38 t/day of VLSFO at that speed, auxiliary a steady 3 t/day, bunkers at USD 600/t. A burn of 38 t/day at 14.5 knots sits at the top of the Panamax range; a modern Kamsarmax is nearer 28 to 33 t/day, and the arithmetic scales.

At 14.5 knots the leg takes 24.4 days. Main engine burns 928 t, auxiliary 73 t, total 1,001 t, costing about USD 601,000.

Slow to 11.5 knots, a 20.7 percent cut. The main-engine rate falls by the cube to 19.0 t/day. The leg now takes 30.8 days. Main engine burns 585 t, auxiliary 92 t, total 677 t, costing about USD 406,000.

Propulsion fuel fell 37.1 percent, which is exactly the squared-relation figure for a 20.7 percent cut rather than approximately it. Total fuel fell 32.5 percent, because the auxiliary term grew from 73 to 92 t as the voyage stretched by 6.4 days. The fuel saving is about USD 195,000, and against it sits 6.4 days the ship cannot earn elsewhere. Note that the headline 36 percent quoted earlier is the generic 20 percent case; this worked example is a 20.7 percent cut with an auxiliary load added, which is why the two figures differ.

The profit-maximising speed

There is an economic optimum and it is not the slowest the ship can run. Slowing saves fuel but loses voyages, and a lost voyage is lost revenue. The balance depends on the bunker price, the daily value of the ship’s time, and the steepness of the fuel curve.

Model daily fuel consumption as a coefficient k multiplied by speed to the power n. Over a leg of distance D, sea time is D divided by 24 times speed, so fuel cost is proportional to speed to the power n minus 1, and time cost is proportional to the reciprocal of speed. Differentiating the total and setting it to zero gives:

where \(C\) is the daily opportunity cost of the ship’s time in USD, \(B\) is the delivered bunker price in USD per tonne, and \(k\) is the fuel coefficient. At \(n = 3\) the denominator is \(2kB\).

Two things about that expression are worth stating plainly, because both are commonly got wrong. The denominator carries \(n - 1\), not \(n\): the exponent that governs a fixed-distance voyage is one less than the exponent that governs the instantaneous fuel rate. And the numerator is the opportunity cost of the ship’s time, not the time charter equivalent , because TCE is already net of bunkers and voyage costs, so putting it on top double-counts the fuel. Where the numerator is instead gross daily earnings before bunkers, the denominator becomes \(nkB\), and the two forms are not interchangeable.

Distance cancels out of the derivation entirely. Leg length does not affect the optimum speed, which surprises people who expect a long ocean passage to justify slower steaming than a short one.

The sensitivity is the part a practitioner actually uses, and it is robust regardless of the fitted exponent. Because the optimum varies as the cube root of the earnings-to-bunker ratio, doubling the daily time value raises it by about 26 percent and quadrupling it by about 59 percent, while doubling the bunker price lowers it by only about 21 percent. That damping is why fleet speeds moved a knot or two across a decade in which freight rates moved by an order of magnitude. The fuller treatment, including the three revenue schedules in Ronen’s original models and the cargo inventory term that Psaraftis and Kontovas add to the objective function, is in optimum speed economics .

How slow steaming became structural

Slow steaming existed as a tactic long before 2008. What changed in late 2008 was scale: the practice went from an occasional choice to a fleet-wide standing policy across container shipping inside a few months.

Three things hit together. The Baltic Dry Index fell from its all-time peak of 11,793 on 20 May 2008 to 663 in early December 2008, about 94 percent in roughly seven months. The pre-crisis orderbook kept delivering newbuildings into that hole, so the fleet grew while the cargo shrank. And bunkers were expensive: Rotterdam IFO 380 cSt averaged USD 679 per tonne in July 2008 against about USD 220 per tonne in January 2007. Slowing absorbed surplus capacity, because a slower loop needs more ships to keep weekly frequency, and cut the fuel bill at the same time. One lever solved both problems.

What Maersk published

A.P. Moller-Maersk drove the shift, and it is worth separating what the company stated from what the trade press reported.

In a presentation to the United States Environmental Protection Agency marine standards forum on 4 May 2010, Maersk described a super slow steaming study begun in 2007 across 110 vessels, run with the engine manufacturers, which established that the engines could be operated as low as 10 percent load against a traditional 40 to 60 percent range, and which prompted the makers to change their recommendations. Reported savings were 10 to 30 percent on fuel.

Separately, Maersk Line’s own sustainability reporting recorded CO2 down more than 16 percent from 2007 levels by the end of 2010, measured as average grams of CO2 per TEU-kilometre across owned and chartered vessels, with fuel oil consumption falling from 10,392 kt in 2009 to 9,792 kt in 2010. By 2013 the company put carbon efficiency at 34 percent better than 2007, attributed substantially to slow steaming.

What “super slow steaming” means

The industry coined gradations and never agreed on them. Trade usage of “super slow steaming” ranges from anything under 18 knots to a specific 15-knot band, and no source supports a fleet-wide threshold in knots.

The definition with contractual effect is by engine load, not by speed. The BIMCO Slow Steaming Clause for Time Charter Parties 2011 draws the operative line at the auxiliary blowers’ cut-out point: its first tier requires compliance only while the engine runs above that point, and its second tier, ultra-slow steaming, requires compliance whether the engine operates above or below it, and may require modifications or additional equipment on the engine designers’ advice. That is the boundary a fixture can be written against.

The barrier was never the hull, which only gets more efficient at lower speed. It was the main engine, designed to run near 70 to 85 percent of its maximum continuous rating and now asked to hold a quarter of that for weeks.

Did fleet speeds actually fall

Yes, but less than service-speed figures suggest, and the two measures must not be mixed.

A liner service speed is the sea-passage design point for a scheduled string. A fleet-average speed is speed over ground across the whole year, including port approaches, canal transits, waiting and repositioning, and it comes out several knots lower. Clarksons Research data over 2012 to 2024 shows container ships down 1.50 knots, bulk carriers down 1.01 knots and oil tankers down 0.74 knots across the period. Across 2024 container ships averaged 14.0 knots, up 1 percent on 2023, while oil tankers fell to 11.4 knots and bulk carriers to 10.9 knots. The container fleet’s low point was 13.8 knots in the first quarter of 2023.

The Fourth IMO GHG Study 2020 dates the effect. Overall carbon intensity in 2018 was 21 percent better than 2008 measured in AER and 29 percent better in EEOI on the voyage-based allocation, and more than half of that improvement was achieved before 2012, with the pace slowing since 2015 to annual changes of 1 to 2 percent. The Study names speed reduction as a key driver, especially for bulk carriers, chemical tankers, container ships and oil tankers, and records that most ship types ceased slowing further from 2015 as the market improved and fuel prices fell.

Engine operation at low load

A two-stroke main engine is sized for a contracted maximum continuous rating and tuned to be efficient near 70 to 85 percent of it. Run it at a quarter of that for months and the tuning is wrong, the turbocharger is oversized for the operating point, and combustion runs cold. The responses, in rough order of permanence, are turbocharger cut-out, part-load or low-load tuning, and derating.

How low a two-stroke will actually run

Down to 5 percent engine load, continuously, on modern electronically controlled and dual-fuel engines with precautions. MAN Energy Solutions Service Letter SL2021-714 of June 2021 is titled for the 5 to 40 percent band and records ships having run at 2 to 3 percent load for up to four days. Wartsila states 10 percent to 100 percent of contracted MCR without major modification.

What breaks below the limit is not the cylinder liner. Soot, coke and unburned fuel accumulate in the scavenge air space and raise the scavenge fire risk; exhaust valve heat load rises slightly in the 20 to 40 percent band; the turbocharger gas side fouls; the exhaust gas boiler builds soot and a pressure drop; and the auxiliary blowers, running continuously, wear faster than they were designed for. Where selective catalytic reduction is fitted, it disengages by default below 7 percent load and re-engages at 10 percent on load-up, because the exhaust temperature to run dosing is not available below that.

Turbocharger cut-out and blanking

Cut-out isolates one turbocharger so the remaining units see proper exhaust mass flow. The hardware is two pneumatically operated swing gate valves, one at the turbine inlet and one at the compressor outlet, and over 500 systems are in service.

For planned operation the recommended maximum load is 35 percent with one of two turbochargers cut out, 65 percent with one of three, and 70 percent with one of four. Measured at 25 percent engine load, the fuel consumption reduction is 6 g/kWh cutting one of two, 5 g/kWh cutting one of three and 4 g/kWh cutting one of four. Wartsila measured 8 to 12 g/kWh on an RT-flex96C with three turbochargers. A second benefit is easy to miss: cut-out shifts auxiliary blower activation to a lower load band, which cuts blower electrical consumption and the wear that goes with continuous running.

Blanking is a different measure and is often confused with cut-out. A blinded-off turbocharger is semi-permanent, limits the engine to roughly 10 to 60 percent load, and prevents the periodic high-load run that clears deposits from being carried out effectively. The distinction is drawn out in turbocharger cut-out and blanking .

Part-load and low-load tuning

Variable injection and exhaust valve timing is the baseline capability of an electronically controlled engine. The tuning method that actually moves the efficiency peak is the exhaust gas bypass, with variable turbine area as the alternative on some turbochargers. MAN part-load optimisation covers 50 to 85 percent load and low-load optimisation 25 to 70 percent. WinGD, which took over the Wartsila two-stroke business in 2015, offers Delta Bypass Tuning below 50 percent load and Low-load Tuning below 75 percent, certifiable together as Dual Tuning but not operable simultaneously.

The trade is rarely stated and it is the important part. The IMO NOx limit is a weighted average of emissions measured at 25, 50, 75 and 100 percent load, so a maker can tilt the fuel consumption profile toward low load only by accepting worse consumption at high load. Nothing is free; the weighted cycle is the budget. The mechanisms are covered in exhaust gas bypass and variable turbine area .

The slide fuel valve is often credited to the slow-steaming era and it predates it. MAN introduced it in Service Letter SL02-403/UM of June 2002 for large-bore engines from 80 cm bore upward, to eliminate the sac volume whose low-pressure discharge produced smoke, unburned hydrocarbons and carbon deposits, and to reduce NOx. It was standard on new engines and retrofittable on older ones from that date. Its low-load benefit is real, and MAN cites it as important to the success of low-load operation, but the hardware existed first and slow steaming exploited it.

Derating against power limitation

Engine derating permanently lowers the rated power and re-optimises the engine around the lower figure. Moving the specified MCR off the full-rated mean effective pressure line onto the maximum derated line is worth about 4 g/kWh: a worked G-type example gives 171.5, 169.5 and 170.0 g/kWh at 50, 75 and 100 percent load on the higher line against 167.5, 165.5 and 166.0 g/kWh on the derated line.

Two caveats belong with that figure. The published consumption tolerance is 5 percent at nominal, widening to 7 percent at 50 to 64 percent load, which is larger than the gain. And MAN’s own low-load service letter states that part-load optimisation and engine derating have no significant relevance in the 5 to 40 percent load band, so a ship that genuinely operates that low is not the candidate for it.

Engine power limitation is the regulatory cousin and it is a different thing mechanically. EPL limits engine power: on a mechanically controlled engine by a sealing device that locks the fuel index with a wired mechanical stop screw or an equivalent governor limit, and on an electronically controlled engine by a fuel index limiter or a direct power limit in the control system. ShaPoLI limits shaft power instead, using torque and shaft-speed sensors feeding a control unit. Both must be tamper-proof and both must be non-permanent, requiring deliberate action by the master or the officer in charge to release.

A genuine derate changes the specified MCR in the layout diagram, re-matches the turbocharger, changes the certified NOx Technical File operating point and normally needs a matched propeller to preserve the light running margin. A derated engine cannot afterward be run above the new rating at all. An EPL can be released, and the release is logged. The distinction is developed in engine derating for slow steaming and the engine layout diagram .

Cylinder condition after the 2020 sulphur cap

The low-load risk profile inverted on 1 January 2020, and an article written on the pre-2020 position now gives the wrong instruction.

Before the cap, on 2.5 to 3.5 percent sulphur heavy fuel oil at low load, the cylinder liner wall could fall below the sulphuric acid dew point of the combustion gases and condensed acid attacked the liner. That is why high base number cylinder oils were introduced as the design basis, and why makers developed jacket cooling water bypass, load-dependent cylinder liner cooling and rating-dependent liners, which raise liner wall temperature by about 12 degrees C.

On 0.50 percent sulphur fuel the acid load is roughly a fifth of that. MAN’s position is that at low fuel sulphur the engine is excessively protected against corrosion, to the point that the Alpha Cylinder Control factor cannot be assessed at all, because lubrication never enters the active area. The whole of the current low-load service letter, written for low-sulphur operation, gives one lubrication instruction and it is to cut the feed rate, preferably to 1.2 g/kWh in the 5 to 40 percent band, because below roughly 30 percent load excessive cylinder lubrication occurs.

One consequence is frequently stated backwards. It is not low base number oil on low sulphur fuel that leaves deposits when overdosed. It is high base number oil on low sulphur fuel, where the unreacted calcium carbonate detergent package has no acid to neutralise and deposits as ash on the piston topland and in the ring grooves. Base number is selected against fuel sulphur, not against a legacy default. The mechanism and its monitoring are in cylinder lubrication and feed rate and cold corrosion in two-stroke engines .

The cold-end risk did not disappear; it migrated downstream. The exhaust gas boiler still needs its gas temperature held above the sulphur condensation level, which is a constraint on how low the engine can be run without fouling the boiler.

Periodic high-load running to clear deposits is real practice, and the two makers disagree on it, so there is no single industry interval to quote. Wartsila recommends raising load twice a week, as high as possible and at least to 70 percent, for a minimum of one hour, with turbocharger washing and soot blowing done at that load. MAN takes the opposite emphasis for its own engines: load-up must be minimised when running at low load because frequent load-up could jeopardise cylinder condition, and load-ups should follow the turbocharger maker’s cleaning intervals rather than a calendar. Where a load-up is done after an extended low-load period, MAN’s manual programme is 5 to 40 percent over 30 minutes and 40 to 75 percent over 60 minutes.

Overhaul intervals are the other claim to treat carefully. No published maker or class source states that part-load operation extends piston ring overhaul intervals, and the current low-load guidance asks for more frequent inspection rather than less. What is defensible is that sustained part load lowers peak firing pressure and thermal load, that both makers report positive long-run service experience since the 2009 to 2012 learning period, and that intervals are in practice set condition-based against scavenge drain oil iron content rather than by a running-hour figure. The subject is treated in two-stroke engine overhaul intervals and slow steaming and engine cleanliness .

Hull, propeller and fouling

Hull and propeller condition erodes the slow-steaming saving faster than most operators expect, and the widely quoted ranges for it are looser than the measured ones.

Schultz (2007) gives the increase in required shaft power by fouling condition, which is more useful than a single range because it tells the reader which state their hull is in:

Demirel and others (2017) showed that roughness height dominates coverage: 10 percent coverage of 5 mm barnacles costs about the same added power as 50 percent coverage of 1.25 mm barnacles. Full-scale before-and-after cleaning trials measured biofilm alone raising the powering requirement by up to 18 percent. The subject is developed in hull roughness and added resistance .

Whether slow steaming itself causes fouling is more subtle than the common claim. Most fouling organisms have an adhesion threshold around 4 to 5 knots, and settlement is much more likely below that, so operating at 11 to 18 knots is not itself a settlement regime. What does hurt is the second-order consequence: a slower ship spends more time at anchorage and in port, and a self-polishing copolymer coating renews itself through flow, with the polishing rate falling by roughly half when stationary compared with 14 knots. The coating stops renewing exactly when biofouling pressure is highest.

The current instrument is MEPC.378(80) , the 2023 Biofouling Guidelines adopted 7 July 2023, which replaced the 2011 guidelines and introduced a four-level fouling rating and explicit idle-time provisions. Acceptable idle time is a biofouling management plan parameter, typically 18 to 30 days, and it is often written into the charter party, which makes it a commercial term rather than only a technical one. Cleaning under that framework is condition-based against the fouling rating, not calendar-based.

The margins built into the ship at design are the industry’s own answer to the same question. Sea margin is traditionally about 15 percent of propeller design power and engine margin a further 10 or 15 percent, giving a service rating near 90 percent of the specified MCR. Light running margin was revised upward to 4 to 10 percent for new projects from 1 May 2015, applicable at all intended draughts and verified by averaging measured values across 50 to 100 percent load. Both are covered in sea margin and engine margin and propeller light running margin .

A propeller drawn for 25 knots is not the right propeller at 17. Planning figures for a retrofit to a lower design point are 2 to 6 percent of main engine fuel on IMO GreenVoyage2050 numbers and 3 to 10 percent for high-efficiency propellers on Lloyd’s Register’s, at roughly USD 400,000 to 850,000 against a propeller life of about 15 years. The same logic applies to the bulb: the only onboard-measured bulbous bow retrofit case is the HMM 8,600 TEU series, designed for up to 27 knots and slow steaming at 15 to 18, where the first conversion cost USD 680,000 and cut fuel by close to 1,000 tonnes a year, verified at around 5 percent or above. The gain grows the further the ship is from the speed the bulb was drawn for.

Measuring the saving and proving it

A claimed saving is only as good as the baseline it is measured against, and the baseline drifts as the hull fouls and the propeller roughens. Compare this year’s slow-steaming burn against the sea-trial figure and the saving looks larger than it is, because part of the gap is fouling penalty rather than speed benefit.

ISO 19030 is the method, and the part matters. The standard has three parts, all first edition, all published November 2016, none revised. Part 1 carries general principles and the definitions of the performance indicators. Part 2 carries the default method: how to build a speed-power reference from sea-trial or model data, how to filter noon-report or high-frequency sensor data for weather, draft and current, and how to express the residual as a percentage performance change over time. Part 3 carries alternative methods. Citing Part 1 for the method, which is common, points the reader at the wrong document.

Two scope limits are stated in the standard itself and they constrain what can be claimed from it. It applies to displacement ships with conventional fixed-pitch propellers only. And it states expressly that the methods are not intended for use in a regulatory framework and are not for comparing different ships. It is a self-comparison tool over time, which makes it the right instrument for tracking a hull and the wrong one for asserting a compliance position. It also cannot separate hull fouling from propeller roughness, because it measures hull and propeller performance as a single indicator by design. The standard is treated in ISO 19030 hull and propeller performance .

The speed-power reference itself now rests on a moved standard. ISO 15016:2025, published 3 February 2025, superseded ISO 15016:2015, and Resolution MEPC.403(83), adopted 11 April 2025, requires either that edition or ITTC Recommended Procedure 7.5-04-01-01.1 (2024) for EEDI speed trials from 1 May 2026. The 2024 ITTC procedure supersedes the 2017, 2021 and 2022 revisions. The 2025 ISO edition changed the wave correction method, replaced the shallow-water correction, moved the wind profile exponent and made the maximum trial wind speed depend on ship length, and the practical effect is that reported trial speed generally rises by up to 0.2 knots on the same measured data. Anyone comparing a 2024 trial against a 2027 trial is comparing two different analyses. See speed and power trials and ISO 15016 .

The data underneath is the noon report or, increasingly, continuous shaft-power and flow-meter telemetry. A noon report logs speed over ground, speed through water, observed weather, fuel consumed and shaft revolutions once a day. Its weakness is that it is coarse and self-reported, and that speed and weather are cross-correlated in ways that bias a naive regression, which is precisely the criticism levelled at low in-service speed exponents. High-frequency monitoring samples power and flow every few seconds and is what makes a defensible trend possible.

What the regulation actually requires

No IMO instrument sets a speed limit for merchant shipping. MARPOL Annex VI Chapter 4 regulates a design index, a management plan, fuel consumption reporting and an annual carbon intensity rating, and none of them states a number of knots. Power limitation caps power and therefore caps top speed as a consequence, but that is downstream of the rule rather than the rule itself.

The chapter, as revised by MEPC.328(76) and in force since 1 November 2022, runs regulations 19 to 29. Regulation 22 is the attained EEDI and regulation 24 the required EEDI; regulation 23 is the attained EEXI and regulation 25 the required EEXI; regulation 26 is the SEEMP ; regulation 27 the data collection system; and regulation 28 the operational carbon intensity.

EEXI and power limitation

EEXI applies from 400 gross tonnage, under regulations 19.1, 23 and 25. Verification took place at the first annual, intermediate or renewal survey, whichever came first, on or after 1 January 2023, so the practical deadline fell at different points through 2023 depending on the ship’s anniversary date rather than on a single calendar day.

The cheapest route to compliance for most existing ships is a power limitation, which formalises slow steaming in the ship’s papers. MEPC.335(76) , as amended by MEPC.375(80) of 7 July 2023 and MEPC.390(81) of 22 March 2024, governs the arrangement and the override.

The override is narrow. Paragraph 3.1 permits use of the power reserve only for securing the safety of the ship or saving life at sea, with adverse weather, ice-infested waters, search and rescue, avoidance of pirates and engine maintenance given as examples. There is no commercial override, and a master asked to make up schedule cannot use it for that. What must be recorded is extensive: the ship type and IMO number, the limited and maximum unlimited power, the position and timestamp of use, the reason, the Beaufort number and wave height or ice condition where weather was the reason, the system records during use, and the position and timestamp when the limit was reactivated, signed by the master on the record page of the onboard management manual. Where the override is activated but the reserve is not used, the event still goes in both the bridge and engine-room logbooks. The Administration and the port of destination are notified without delay, and the Administration reports uses to IMO annually.

CII: thresholds, metrics and bands

The CII applies from 5,000 gross tonnage, not 400. Regulation 28.1 covers twelve ship types: bulk carrier, combination carrier, containership, cruise passenger ship, gas carrier, general cargo ship, LNG carrier, refrigerated cargo carrier, ro-ro cargo ship, ro-ro cargo ship (vehicle carrier), ro-ro passenger ship and tanker. Offshore vessels, tugs, barges, fishing vessels, dredgers, cable layers, research ships and yachts are outside it, as are Polar Code Category A ships under regulation 19.3 and ships trading solely within their own flag State’s waters under regulation 19.2.1.

Two metrics are in use, and MEPC.352(78) paragraph 2.5 assigns them. The supply-based indicator using deadweight capacity is the AER, applying to bulk carriers, tankers, container ships, gas carriers, LNG carriers, general cargo ships, refrigerated cargo carriers and combination carriers. The one using gross tonnage is cgDIST, applying to cruise passenger ships, vehicle carriers, ro-ro cargo ships and ro-ro passenger ships. EEOI and the remaining indicators are voluntary trial metrics. The AER is treated in the Annual Efficiency Ratio .

Rating boundaries come from Table 1 of MEPC.354(78) , derived by quantile regression on 2019 data and expressed as multipliers on the required CII. For a bulk carrier they are 0.86, 0.94, 1.06 and 1.18; for a tanker 0.82, 0.93, 1.08 and 1.28; for a container ship 0.83, 0.94, 1.07 and 1.19. The bands were calibrated so that in 2019 the middle 30 percent of a segment fell in band C. As the reduction factor tightens, the boundaries move with the required CII and the relative spacing between them does not change, so a ship is rated against predetermined boundaries rather than against its peers in that year. See CII rating boundaries .

The consequence of a poor rating is narrower than it is often described. Regulation 28.7 requires a ship rated D for three consecutive years or rated E once to develop a plan of corrective actions; regulation 28.8 requires the SEEMP to be revised to carry that plan and submitted for verification no later than one month after the attained CII is reported; regulation 28.9 requires the actions to be undertaken. There is no fine, no detention and no trading prohibition in the regulation, and port state control carries no CII enforcement hook. The mechanism is set out in the CII corrective action plan .

The reduction factor trajectory

The single largest change to slow-steaming economics since 2023 is one many articles have not caught up with. MEPC.400(83) , adopted 11 April 2025, replaced Table 1 of the G3 guidelines and filled in four years that had been blank since 2021:

Note the reference year: 2019, not 2008. The G2 guidelines define the reference line as the median attained operational carbon intensity of a defined group of ships in 2019. The step from 11 to 13.625 percent between 2026 and 2027 is the largest single-year tightening in the series and it is now fixed rather than pending. The trajectory is treated in the CII reduction factors .

What the SEEMP says about speed

The current guideline is MEPC.395(82) , adopted 4 October 2024, which revoked the 2022 guidelines. Part I applies to any ship of 400 gross tonnage and above, Part II to any ship of 5,000 gross tonnage and above, and Part III to ships of 5,000 gross tonnage and above in the twelve CII types, which had to carry it with a confirmation of compliance before 1 January 2023.

Speed optimization appears in the guidance on best practice, not in mandatory content, and the guideline explicitly rejects the reading that slower is always better. Paragraph 5.2.6 states that optimum speed means the speed at which fuel used per tonne mile is at a minimum for the voyage, and that it does not mean minimum speed, because sailing at less than optimum speed will consume more fuel rather than less. It names increased vibration and soot deposits in combustion chambers and exhaust systems as possible adverse consequences of slow-speed operation, and notes that for LNG carriers optimization often means a higher speed at the start of laden passages to control tank pressure.

The review, and what changes in 2027

The review of the regime sits in regulation 28.11, not in any SEEMP guideline. Phase 1 closed at MEPC 83 in April 2025, and its binding output was the reduction factor table above. Phase 2 runs from Spring 2026 to Spring 2028 across three sessions and covers enhancement of the SEEMP framework, further development of the CII metrics, correction factors and reference line adjustments, and synergy with the Net-Zero Framework. Nothing about the metric, the ship-type scope or the rating boundaries has been changed. See the short-term GHG measure review .

MEPC.407(84) , adopted 1 May 2026 and due in force 1 September 2027, replaces regulation 28.11 with an open-ended obligation to keep the regulation under review in light of the 2023 IMO GHG Strategy. After that date there is no fixed review deadline in the treaty text. The same resolution opens the fuel oil consumption database, giving Parties non-anonymized access and the public an anonymized version, and designates the North-East Atlantic Emission Control Area.

The IMO Net-Zero Framework is not part of this picture yet. MEPC 83 approved the draft text in April 2025 as a new Chapter 5 to MARPOL Annex VI, and the extraordinary session in October 2025 adjourned its adoption. It is neither adopted nor in force, and its first attained calculation in the draft falls after the end of calendar year 2028. Because its metric is a well-to-wake fuel intensity, speed would not move it in any event.

The EU overlay pulls two ways

The European instruments are a regional overlay on the IMO baseline, they apply to any flag calling at an EEA port, and they do not point the same direction on speed.

EU ETS, under Directive 2003/87/EC as amended by Directive (EU) 2023/959, brought maritime obligations from 1 January 2024 for ships of 5,000 gross tonnage and above, covering 100 percent of emissions on intra-EEA voyages and at berth and 50 percent on voyages to or from a port outside the EEA. Surrender is phased at 40 percent of verified 2024 emissions, 70 percent of 2025 and 100 percent of 2026 onward, with methane and nitrous oxide added from 2026. Because it prices the tonne of CO2, it acts through the same lever as the bunker price and slowing reduces the liability. See EU ETS for shipping and EU MRV .

FuelEU Maritime, Regulation (EU) 2023/1805, applicable from 1 January 2025 to ships above 5,000 gross tonnage calling at EU ports, works on an entirely different quantity. Its metric is well-to-wake greenhouse gas intensity in grams of CO2 equivalent per megajoule, covering CO2, methane and nitrous oxide, with the Article 4 trajectory running minus 2 percent from 2025, minus 6 percent from 2030, minus 14.5 percent from 2035, minus 31 percent from 2040, minus 62 percent from 2045 and minus 80 percent from 2050. Slowing down does not improve a FuelEU balance at all, because burning the same fuel more slowly leaves its intensity unchanged. The two instruments are routinely conflated, and an operator who treats them as one will misprice the position. See FuelEU Maritime and FuelEU penalties, pooling and multipliers .

The UK ETS , China DCS and the CARB at-berth rule are further overlays with their own scopes. None of them mandates a speed either.

Who orders the speed

Slow steaming runs into contracts long before it runs into physics, and the contractual question is not “how slow” but “who decides, and who carries the consequence”.

Utmost despatch and the employment order

Under a time charter party the charterer buys the bunkers and wants the ship slow. The instruction binds because speed and routing are matters of the employment of the vessel rather than of navigation. In The Hill Harmony [2000] UKHL 62, [2001] 1 AC 638, the House of Lords held exactly that on an NYPE charter where the master took the rhumb-line route against charterers’ great-circle orders, costing about USD 89,800 in delay and extra fuel. The owners’ error-in-navigation defence failed because the breach was a refusal to obey employment orders, not a mistake of seamanship. Lord Bingham also treated utmost despatch as an independent duty that, absent a usual route or a maritime reason, itself requires the shortest and quickest route.

That produces the structural tension the whole clause library exists to resolve: the same charter that lets a charterer order slow steaming contains the clause obliging the ship to hurry. The master keeps an overriding discretion for the safety of ship, crew, cargo and environment, and that discretion is written expressly into the BIMCO clause.

Under a voyage charter party the owner buys the fuel, so the decision is the owner’s, bounded by the cancelling date, laytime and demurrage , and any speed warranty given to the charterer. Arrive late and demurrage on the cargo can consume the fuel saving.

The BIMCO clauses

Five clauses matter and each does a different job. The full set is compared in the BIMCO slow steaming clauses .

Two points that are easy to get wrong. Neither the Virtual Arrival Clause nor the Just in Time Arrival Clause exists in a time charter form, because under a time charter the charterer already controls speed by employment order and the problem those clauses solve does not arise in the same shape. And a CII Clause for Voyage Charter Parties 2023 exists alongside the 2022 time charter clause; it lets owners or the master adjust course and reduce speed subject to an inserted minimum speed measured in an inserted good weather definition, and both blanks must be completed or the mechanism does not function. Its guidance notes suggest pairing it with the 2013 or 2021 clause, which is the cleanest link between the carbon layer and the arrival layer.

Under the 2011 time charter clause, compliance with the charterer’s speed instruction is deemed to be compliance with any obligation to proceed with utmost or due despatch. That deeming is the whole point of the clause, and its despatch and bill of lading sub-clauses are marked inapplicable in liner trade because a liner bill such as CONLINEBILL 2000 already confers the liberty at clause 5.

Separately, the BIMCO ETS Allowances Clause for Time Charter Parties 2022 allocates emission allowances to whoever provides and pays for the fuel, by transfer of allowances rather than reimbursement of their cost, with owners notifying monthly within the first seven days.

Performance claims and the bill of lading holder

A charterer’s slow order does not automatically cancel the owner’s speed warranty, and the 2011 clause addresses the point directly: existing speed and consumption warranties survive, but performance falling below the warranted speed because the charterer asked for it sits outside the performance guarantee. Without that wording, a charterer who orders slow steaming and then claims for underperformance is claiming for a shortfall it caused.

The claim itself is not an off-hire claim, and conflating the two is a common pleading error. Off-hire suspends hire for a listed cause under the off-hire clause. Failure to meet a warranted speed is a breach of warranty sounding in damages, measured as time lost plus the value of bunkers overconsumed. See off-hire and performance claims and charter party speed and consumption warranties .

The method is a sampling exercise. Extract the periods meeting the contractual good weather definition , compute performance across those periods only, and extrapolate. In The Ocean Virgo [2015] EWHC 3405 (Comm), Teare J held that nothing required a good weather period to run 24 consecutive hours from noon to noon, but that the sample must be large enough to represent the voyage as a whole. The commercially standard definition on the record there was wind up to Beaufort force 4 and Douglas sea state 3, no adverse currents and no negative influence of swell.

Owners and charterers disagree systematically about the weather, because charterers build the claim on a weather routing company’s hindcast gridded data and owners defend on the master’s observed noon reports, and force 4 against force 5 decides whether a period is admissible at all. The 2011 clause makes the obligation to use charterer-supplied routing and performance monitoring bite only where agreed in advance, which is the contractual fix.

The remaining exposure is the one a charterparty cannot close. A third-party bill of lading holder is not a party to the charter and takes on the bill’s own terms, which is why the clause requires charterers to procure bill of lading wording and, recognising that this is unrealistic on every bill, to indemnify owners against liabilities arising from documents issued as presented. Whether deliberate delay for fuel economy can amount to unreasonable deviation under Hague-Visby Article IV rule 4 has not been decided by any reported English judgment. The exposure is evidenced by the drafting rather than by a holding, and an unjustified deviation can cost the carrier the exceptions and package limitation. See deviation in charter parties and bill of lading .

Just-in-time arrival and the port call

Sailing fast and then waiting at anchor burns the saving twice over, and closing that loop is a separate discipline from slowing the sea passage.

The reference document is the Just In Time Arrival Guide: Barriers and Potential Solutions, published in 2020 by the GloMEEP Project and the Global Industry Alliance to Support Low Carbon Shipping. Its definition is the useful part and it is narrower than generic slow steaming: the ship maintains an optimal operating speed to arrive at the pilot boarding place when the availability of berth, fairway and nautical services is assured. All three conditions, not just a berth time.

The Guide identifies two barrier classes, contractual and operational, and the contractual barrier is precisely the despatch obligation and the laytime and demurrage regime, which is why the Guide and the BIMCO clauses are one story rather than two. It documents barriers and implementation cases rather than quantifying an emissions saving, and it should not be cited as though it measured one. The number that does show how the industry prices the trade is the 50 percent of demurrage rate default in the 2013 Virtual Arrival Clause.

The data layer is being standardised in parallel: the IMO Expert Group on Data Harmonization agreed to add just-in-time operational data elements to the IMO Reference Data Model, and DCSA published complementary standards for a just-in-time port call in October 2020. The International Taskforce Port Call Optimization maintains the port and terminal master data those standards depend on. See just-in-time arrival and port call optimisation .

Liner strings and the cost of a knot

For a scheduled service the cost of slowing is not the fuel saving minus a little. It is a step function, and the step is a whole ship.

A weekly service needs one ship on berth in each port every seven days, so the string size is the round-voyage time in days divided by seven, rounded up. Take an Asia to North Europe loop of about 21,000 nautical miles round trip with about 14 days of port and canal time:

This table is a Shipping-Wiki.com construction from the stated inputs rather than a published figure, and the inputs are given so it can be rebuilt for any trade. A 20 to 14 knot cut on this loop costs two extra ships.

The consequence is the non-obvious part. Because the ship count is an integer, there are speed bands within which a further knot costs no extra tonnage at all, and boundaries at which one further knot costs a whole ship’s capital and operating cost. That is why liner operators cut speed in coordinated moves across an entire string rather than ship by ship, and why the same fuel saving is worth wholly different amounts on either side of a boundary. See liner service string design .

The cargo pays for the transit time and, under a liner bill, is not compensated for it. For a TEU of goods worth USD 50,000 at a 10 percent annual cost of capital, one extra day in transit costs about USD 13.70 per TEU, or roughly USD 192,000 per extra day across a 14,000 TEU ship. Those are stated assumptions rather than a published figure; the structure is the point, and it is why Psaraftis and Kontovas argue that the cargo’s inventory cost belongs in the speed objective function alongside the owner’s fuel bill. See transit time and inventory carrying cost .

The rebound effect and the CII critique

Slow steaming is not a permanent gain. It is a function of the freight market and it reverses when the market turns.

The optimum tracks the cube root of the earnings-to-bunker ratio, so any sustained jump in freight rates pulls speed back up. The container boom of 2021 to 2022 is one case, and the 2023 to 2024 Red Sea diversions are the clearer recent one, because they moved speeds in opposite directions across segments. By mid-2024 tonnage crossing the Gulf of Aden was down 76 percent and Suez Canal tonnage down 70 percent while Cape of Good Hope arrivals rose 89 percent, adding roughly 12 extra days at 16 knots on the Asia to North Europe leg. Carriers recovered schedule partly by raising speed and partly by adding tonnage, and container fleet average speed rose in 2024 while every other segment fell. See the Red Sea crisis and Cape diversions and the Cape of Good Hope route .

Congestion does the same work from the other side. Sea-Intelligence measured the share of global container fleet capacity unavailable through delay peaking at 13.8 percent in January 2022 against a 2011 to 2019 baseline of about 2.2 percent, falling to 7.9 percent by August 2022; Linerlytica put the physically congested tonnage peak at about 4 million TEU, roughly 15.7 percent of the fleet. Congestion and slow steaming are both capacity sinks, which is why the boom years tightened the market even as ships sped up. See port congestion .

The critique of the CII metric is separate and it is often stated loosely. The sourced version has three distinct limbs.

The denominator problem is the root: the AER divides annual CO2 by deadweight-miles rather than by cargo-tonne-miles, so it measures capacity moved rather than cargo moved. IMO Secretariat document MEPC 82/INF.25 found the AER to be the least accurate of the metrics considered, because it assumes vessels are loaded to full capacity in all situations including ballast legs, with the consequence that ships with higher payload utilisation are penalised and lightly loaded ships benefit. The Fourth IMO GHG Study 2020 records the same effect from the data side: an increase in payload utilisation generally reduces EEOI but increases AER.

Ballast-leg exposure follows from that: because the denominator is capacity multiplied by distance, a ballast leg adds transport work while burning less than a laden leg. Idling and engine-running perversity is the third limb, and INTERCARGO’s MEPC 82 submission put it concretely, warning that the framework may encourage running main engines unnecessarily, for example at anchorage, raising absolute emissions while improving the rating, and reporting that E-rated vessels often have lower average absolute CO2 emissions than A to D rated ones across an analysis of more than 5,600 bulk carriers.

These are attributable positions, not commentary. On 9 July 2024 BIMCO, CLIA, ICS, INTERCARGO, InterManager and INTERTANKO issued a joint policy statement calling on the IMO to amend the CII system to avoid unintended consequences contradictory to reducing overall emissions. MEPC 81 in March 2024 acknowledged that possibly inaccurate or misleading ratings could produce unintended adverse consequences, and the IMO had received 78 proposals for amendment by mid-2024.

None of this makes slow steaming a poor fuel choice. It makes the regulatory accounting of it imperfect, and the imperfection sits in the denominator rather than in the speed term.

Limitations

The cube law that makes slow steaming pay is a design-condition approximation and it degrades exactly where slow steaming operates. In-service fits at reduced speed return exponents well below 3, and whether that reflects real hydrodynamics or a regression artefact from cross-correlated speed and weather is unresolved between the published studies. A projected saving built on an assumed exponent of 3 is an upper bound at low speed, not an estimate.

Every figure in the worked examples on this page depends on inputs that move. The Panamax voyage assumes 38 t/day at 14.5 knots, a steady 3 t/day auxiliary load and USD 600/t bunkers; the string table assumes 21,000 nautical miles and 14 days in port and canals; the inventory figure assumes USD 50,000 per TEU at a 10 percent cost of capital. They are worked to show the structure and the sensitivities, not to be quoted as fleet values.

The engine sets a floor the economics ignore, and it is lower than commonly stated but it is real. Below the continuous minimum the constraints are scavenge-space deposits, gas-path and boiler fouling, and auxiliary blower wear rather than liner corrosion, and the countermeasures cost fuel and attention that the bare fuel-curve arithmetic does not price.

Two measurement caveats limit what can be claimed. ISO 19030 cannot separate hull fouling from propeller roughness, because it measures them as one indicator by design, and it states that it is not intended for use in a regulatory framework, so it supports a trend and not a compliance assertion. And the speed-trial baseline itself moved in 2025, so trials analysed under different editions are not directly comparable.

On the commercial side, schedule integrity, speed warranties, laytime and demurrage windows and time-sensitive cargo cap how slow a ship can actually run regardless of the fuel curve, and the optimum moves between fixtures so a ship cannot be tuned once and left. The regulatory position on the CII metric is itself under review to 2028, and the Net-Zero Framework’s status is adjourned rather than settled, so any statement about the post-2028 incentive structure is provisional.

Where slow steaming sits among the other levers

Slow steaming rarely runs alone. The SEEMP treats it as one entry in a stacked efficiency plan alongside weather routing , trim optimisation , hull and propeller maintenance and retrofits.

A ship re-faired with a bulbous bow retrofit for its new operating point, capped by engine power limitation , fitted with energy-saving devices ahead of the propeller, wind-assisted propulsion to add speed-independent thrust and air lubrication to cut friction, stacks several single-digit gains on the speed cut. They do not add linearly, because each measure changes the operating point the next one optimises against, so a combined figure has to be modeled rather than summed.

Alternative fuels change the arithmetic without ending the practice. LNG , methanol , ammonia and biofuels cost more per unit of energy than heavy fuel oil , so each unit saved by slowing is worth more, which argues for more slow steaming. Pulling the other way, their lower well-to-wake intensity eases the FuelEU pressure, though not the ETS pressure, which continues to price the tonne of CO2 regardless of fuel. Onboard carbon capture and shore power sit outside the speed lever entirely.

The speed band has been broadly stable since about 2015 rather than falling further, and the reasons are structural: the marginal saving per knot is already small at today’s speeds, the engine floor is close, newbuildings are now designed around a lower service speed so the gap to close is smaller, and the CII rewards the intensity rather than the absolute burn. What changes the picture next is the 2027 step in the reduction factor and the outcome of the Phase 2 review, not a further collapse in speed.

Related Articles

• Slow steaming and CII
• Engine derating for slow steaming
• Slow steaming and engine cleanliness
• Admiralty coefficient
• Specific fuel oil consumption

Regulation and reporting

• MARPOL Annex VI
• What is CII
• What is EEXI
• What is EEDI
• EEXI, EPL and ShaPoLI
• SEEMP I, II and III
• CII corrective action plan
• IMO Net-Zero Framework
• IMO GHG Strategy
• EU MRV Regulation
• EU ETS for shipping
• FuelEU Maritime
• UK ETS for shipping
• China DCS
• CARB at-berth rule
• Ship efficiency indices

Chartering and claims

• Time charter party
• Voyage charter party
• Charter party speed and consumption warranties
• Off-hire and performance claims
• BIMCO CII clauses
• Laytime
• Demurrage
• Time charter equivalent
• Deviation in charter parties
• Hague-Visby Rules and bills of lading
• Bill of lading
• Baltic Dry Index and freight indices

Hull, propeller and machinery

• Ship resistance and powering
• Hull form design
• Block coefficient
• Marine propeller
• Bulbous bow retrofits
• Energy-saving devices
• Air lubrication systems
• Trim optimisation
• Wind-assisted propulsion
• Two-stroke marine diesel engine fundamentals
• Marine diesel engine
• Bow thruster and stern thruster

Operations and routing

• Just-in-time arrival
• Weather routing
• Suez Canal
• Panama Canal
• Canals and straits

Ship types

• Container ship
• Bulk carrier
• LNG carrier
• Oil tanker

Fuels

• Heavy fuel oil
• Marine gas oil
• LNG as marine fuel
• Methanol as marine fuel
• Ammonia as marine fuel
• Biofuels in shipping
• Well-to-wake intensity
• Onboard carbon capture
• Cold ironing and shore power