MARPOL Annex I Reg 23: accidental oil outflow

What Regulation 23 is, and what it replaced

Regulation 23 of the revised MARPOL Annex I is the probabilistic oil outflow rule for new oil tankers. It applies to oil tankers of 5,000 tonnes deadweight and above for which the building contract was placed on or after 1 January 2007, or, in the absence of a contract, the keel was laid on or after 1 July 2007, or delivery was on or after 1 January 2010. The 1 January 2010 delivery date is the practical handle most surveyors use: every standard double-hull tanker delivered from the start of that year carries an approved $O_M$ calculation in its damage-stability documentation.

The rule did not appear from nothing. The old Annex I carried deterministic outflow rules in Regulations 22 through 24, which limited the size of individual cargo tanks and the hypothetical outflow from a stipulated damage. The Marine Environment Protection Committee recognized that those deterministic rules “did not properly account for variations in subdivision in general, and longitudinal subdivision in particular,” as the MEPC.122(52) Explanatory Notes put it. A tanker could meet the tank-size cap and still have a poor arrangement for limiting spills. So the MEPC built a performance-based regulation that handles subdivision variation directly, and adopted it as part of the wholesale revision of Annex I under MEPC.117(52).

The technical lineage runs through the Interim Guidelines of 1995 and the Revised Interim Guidelines adopted by resolution MEPC.110(49) in 2003. Those guidelines used a rigorous method: apply five probability density functions each for side and bottom damage, evaluate thousands of unique damage cases, and compute three outflow parameters. Regulation 23 is the simplified production version of that method. It keeps the same statistical basis but reduces the work to one parameter, the mean outflow, computed tank by tank rather than case by case. The MEPC judged the mean outflow “to be the best indicator of overall outflow performance,” so Regulation 23 reports it alone and drops the probability-of-zero-outflow and extreme-outflow parameters that the Revised Interim Guidelines still carried.

That simplification rests on a stated assumption worth keeping in view: the MEPC.122(52) notes say the simplified Regulation 23 method is reasonable “as each design must also meet the provisions of regulation 19.” The deterministic double-hull rule is doing the work of assuring the likelihood of a spill, measured by the probability-of-zero-outflow that Regulation 23 no longer computes. Regulation 23 then governs the expected magnitude. The two are a pair, not alternatives.

The mean oil outflow parameter and its limits

The deliverable of Regulation 23 is a single dimensionless number, $O_M$. It’s the expected oil outflow across all credible collision and grounding casualties, divided by the total oil the ship carries, so it reads as a fraction of cargo lost on average per casualty. A smaller $O_M$ is a cleaner design. The full combination is given in paragraph 8.1 of the Explanatory Notes:

Here $O_{MS}$ is the mean outflow from side (collision) damage in cubic meters, $O_{MB}$ is the mean outflow from bottom (grounding) damage in cubic meters, and $C$ is the total cargo oil volume at 98% tank filling within the cargo block length. The 0.4 and 0.6 weights come from an assumed casualty split of 40% collisions to 60% groundings, carried over from the Revised Interim Guidelines. Note that $O_{MS}$ and $O_{MB}$ are dimensional volumes; dividing by $C$ makes $O_M$ non-dimensional.

The acceptance limit scales with ship size. The regulation sets two anchor points and interpolates linearly between them:

The interpolation band exists because the largest tankers, which carry two longitudinal bulkheads and a finer cargo subdivision, can reach a lower outflow fraction, so the regulation holds them to a tighter standard. Worked at $C = 300{,}000$ m³ the limit is $0.012 + (0.003/200{,}000)(100{,}000) = 0.0135$. At $C = 350{,}000$ m³ it is $0.012 + (0.003/200{,}000)(50{,}000) = 0.01275$. Most VLCCs sit between 320,000 and 380,000 m³ of cargo capacity and fall in this band. The 460,000 DWT ultra-large class in the MEPC parametric series sits above 400,000 m³ and faces the flat 0.012 cap.

Combination carriers, ships built to carry both dry bulk and oil, get a separate, more lenient line. They run without a centerline bulkhead and so struggle to meet the standard oil-tanker curve, so Regulation 23.3.1 lets a combination carrier apply its own criterion if its calculations show the heavier hull structure gives environmental protection at least equal to a standard double-hull oil tanker of the same size, to the satisfaction of the Administration. For a 5,000 DWT combination carrier that allowance starts near $O_M$ of 0.021 and converges onto the oil-tanker line by about 200,000 m³.

Side damage outflow: collision

Side damage models a collision: a striking bow opens the side shell of the struck tanker. The mean outflow from side damage, in paragraph 6.2 of the notes, sums over every cargo tank the probability that the tank is breached times the oil that then escapes:

$P_{S(i)}$ is the probability of penetrating cargo tank $i$ from side damage, and $O_{S(i)}$ is the outflow from that tank once breached. For side damage the regulation makes a conservative assumption: there were no reliable data on how much of a tank’s contents stay aboard after a collision, and theoretical estimation of retained liquid was judged impractical, so the method assumes 100% of the oil in a side-breached tank flows out. A collision is treated as a clean loss of the breached tank’s cargo.

The probability of breaching a given tank from the side is built from the casualty-derived distributions as a product of independent terms (paragraph 5.1):

Each factor is the probability that the damage reaches the tank in one dimension. $(1 - P_{Sf} - P_{Sa})$ is the probability the damage penetrates the longitudinal zone of the tank, where $P_{Sa}$ is the chance the damage lies entirely aft of the tank and $P_{Sf}$ the chance it lies entirely forward. $(1 - P_{Su} - P_{Sl})$ is the probability it reaches the vertical zone, with $P_{Su}$ above and $P_{Sl}$ below. $(1 - P_{Sy})$ is the probability the transverse penetration reaches inboard past the tank’s outboard boundary, with $P_{Sy}$ the chance the damage stays outboard of the tank. Multiplying the three treats the dimensions as statistically independent, which the data did not contradict.

The five side-damage probability density functions, plotted in Figures 2 through 6 of the Explanatory Notes, describe longitudinal location (uniform along the length), longitudinal extent (a sharp peak near zero, density 11.95 at the origin falling to 3.5 at 0.1L and to about 0.35 by 0.2L), vertical location (rising from zero at the baseline to 1.50 from mid-depth up to the deck), vertical extent (density 3.83 at zero falling to 0.50), and transverse penetration (density 24.96 at the shell, 5.00 at 0.05B, falling to 0.56 by 0.10B). The transverse curve is the steepest: most collision damage barely penetrates, which is exactly why a double-hull wing tank of even modest width sharply cuts the breach probability of an inboard cargo tank.

The $C_3$ factor corrects a known conservatism. For most arrangements $C_3$ is 1.0. For tankers above 300,000 m³ capacity, which all carry two longitudinal bulkheads (a three-across cargo arrangement), the minimum-clearance shortcut the regulation uses overestimates side outflow, so $O_{MS}$ for those designs is multiplied by $C_3 = 0.77$. The validation in paragraph 6.4 of the notes showed under-200,000 m³ ships with a single centerline bulkhead need no correction, while the largest three-across designs do.

For a symmetrical tanker the side calculation is run on one side only. For an asymmetrical cargo arrangement the calculation is performed from both sides and the two results averaged, $O_{MS} = (O_{MS\text{-port}} + O_{MS\text{-starboard}})/2$.

Bottom damage outflow: grounding

Bottom damage models a grounding, where the tanker sets down on the seabed or a submerged object and the outer bottom is torn. Bottom outflow is governed by the hydrostatic balance principle rather than a clean spill, because the cargo head and the external sea head fight each other across the breach. The mean outflow from bottom damage for a given tide, in paragraph 7.2, is:

$P_{B(i)}$ is the probability of penetrating tank $i$ from the bottom, $O_{B(i)}$ is the hydrostatic-balance outflow from that tank, and $C_{DB(i)}$ is an outflow reduction factor for oil entrapped in non-cargo spaces below the tank. The breach probability uses the same product structure as the side case (paragraph 5.2):

with the transverse pair now port and starboard ($P_{Bp}$, $P_{Bs}$) and $P_{Bz}$ the probability the vertical extent stays below the tank.

The bottom-damage distributions (Figures 7 through 11) differ in shape from the side set. Longitudinal location is skewed forward (density 0.2 at the aft perpendicular rising to 2.6 at the forward perpendicular), reflecting that a ship runs aground bow-first. Longitudinal extent peaks at the origin (4.5) and flattens to 0.5. Transverse location is uniform across the breadth. Transverse extent is U-shaped, high at the centerline (4.0), dipping to 0.4 mid-beam, rising again to 1.6 at the full beam. Vertical penetration is sharply peaked near the baseline (14.5 at zero, 1.1 beyond about 0.05D), which is why a double bottom of even modest height intercepts most grounding penetrations before they reach the inner bottom.

Three corrections shape the bottom outflow. First, cargo tanks bound by the bottom shell lose some oil even when in hydrostatic balance, from initial impact exchange and from wave and current action; the method assumes outflow of at least 1% of such a tank’s volume. Second, when bottom damage breaches both the outer and inner bottom into a cargo tank, part of the escaping oil is retained in the double-bottom void. Model tests in the IMO comparative study found roughly one-seventh of the outflowing oil was retained when the pressure differential was large and penetration small, and the breached double-bottom space at equilibrium was taken to hold 50% oil and 50% seawater. To capture this without tracing each tank combination, the regulation multiplies the hydrostatic-balance outflow by $C_{DB(i)} = 0.60$: 40% of the computed outflow is assumed entrapped by the non-oil tanks below. The reduction factor for ten actual double-hull tankers fell between 0.50 and 0.70, and 0.60 was selected as representative.

The tide cases: 0 m and 2.5 m fall

A grounded tanker that stays aground through a falling tide loses more oil as the external sea head drops and the hydrostatic balance shifts in favor of outflow. Regulation 23 captures the fall of tide with two calculations and a weighted combination (paragraph 7.4):

$O_{MB(0)}$ is the bottom outflow computed at zero fall of tide, and $O_{MB(2.5)}$ is the outflow computed at a 2.5 m fall of tide. The 70/30 weighting comes from the tidal probability work in section 7.5 of the notes. The fall of tide was modeled with two probability density functions, one for the relative fall during the long-period tidal motion and one for the tidal double amplitude at the moment of grounding, derived from the casualty data. The combined fall-of-tide density shows a meaningful effect up to about 3 m; the probability of an actual fall exceeding 3 m is under 5%, and grounding at high tide is less likely because under-keel clearance is larger then. The MEPC determined that representing the tidal effect by calculating at 0 m and 2.5 m, then combining 70% to 30%, captured the distribution well enough for uniform application.

Note the asymmetry between the two damage modes. Side outflow ignores tide because a side breach above the waterline spills under gravity regardless of tidal state; bottom outflow lives or dies by the hydrostatic balance, so the tide is the dominant driver and the regulation models it explicitly.

The damage probability distributions and their basis

Every probability in Regulation 23 traces to a fixed dataset. The side and bottom probability density functions were derived from historical damage statistics covering 52 collisions and 63 groundings, compiled by the classification societies at IMO’s request. The casualties were oil tankers, chemical tankers, and combination carriers of 30,000 tonnes deadweight and above, over the period 1980 to 1990. The MEPC.122(52) notes flag the limitation plainly: these statistics consist largely of damages to single-hulled tankers, and “should be periodically reviewed as new data becomes available.”

The densities are normalized so the same curves apply to any hull size. Longitudinal location and extent are normalized by ship length, transverse location and extent by breadth, and vertical location and extent by depth. The variables are treated as independent of one another, “for the lack of adequate data to define their dependency,” which is the assumption that lets the breach probability factor into a product of one-dimensional terms.

The tables of probability are built by integrating each density between the planes that bound a tank. For a region bounded below by $d_1$ and above by $d_2$, the probability that the region is damaged is $p = 1 - p_b(d_1) - p_a(d_2)$, where $p_b$ is the probability the damage is restricted to below $d_1$ and $p_a$ the probability it’s restricted to above $d_2$. The three-dimensional breach probability is then the product across the three dimensions, with each compartment modeled as an equivalent rectilinear block described by six boundaries. The boundary symbols in section 8 of the notes, $X_a$ and $X_f$ for the aft and forward longitudinal terminals, $Z_l$ and $Z_u$ for the lowest and highest points, and $y$ for the minimum horizontal distance to the side, are the inputs a designer pulls off the tank plan for every cargo tank.

Two methodology points matter for accuracy. The compartment dimensions are taken at the extreme boundaries of each tank rather than averaged over sloping bulkheads, which the validation found “generally provided more consistent and usually slightly conservative results.” And for symmetrical ships the calculation runs on one side; for asymmetrical arrangements it runs from both port and starboard and averages. The cargo-tank permeability for the outflow calculation is taken as 0.99 (Regulation 23.4.5), tighter than the 0.95 used in damage-stability work, because a double-hull cargo tank is relatively free of internal structure.

Working the parameter in practice

A designer runs Regulation 23 in three steps, as paragraph 2.1 of the notes lays out: determine the probability of penetrating each cargo tank for both side and bottom damage; assess the expected outflow from each damaged tank; then compute $O_M$ and compare it to the permissible value. In production this is a software task. The same class stability platforms that handle Regulation 28 damage stability carry a Regulation 23 module that reads the tank capacity plan, builds the six-boundary block for each cargo tank, looks up the integrated probabilities, applies the side and bottom outflow rules with the $C_3$ and $C_{DB}$ factors and the two tide cases, and reports $O_M$ against the size-dependent limit.

The cargo density input follows a stated convention. Regulation 23.4.3 and 23.4.4 set cargo density by dividing total deadweight by total cargo volume, and the calculation is run on a hypothetical zero-trim, zero-heel condition rather than an actual loaded condition, to keep the result uniform across yards. For combination carriers seeking the separate criterion of Regulation 23.3.1, the work is heavier: a series of finite-element collision and grounding calculations evaluating dissipated plastic deformation energy at each damage position, with the combination carrier as the struck ship at full load, to demonstrate the heavier hull reduces damage extent enough to match a standard tanker.

The result lives in the ship’s documents. The approved $O_M$ calculation report is part of the damage-stability documentation issued by the class society at delivery and held aboard, alongside the Regulation 28 survival calculations. A Type 3 approved stability instrument under the harmonized loading-instrument rules can run the probabilistic Regulation 23 outflow alongside the deterministic damage-stability check, so most modern newbuilds carry one. Unlike damage stability, which the master re-verifies each voyage against the actual loading condition, $O_M$ is a fixed design property of the as-built tank arrangement: it doesn’t change with how the ship is loaded, so it’s certified once at design approval and re-checked only when the cargo-tank subdivision is altered by conversion or structural modification.

How Regulation 23 sits with Regulation 19 and Regulation 28

Regulation 23 is one of three rules that together govern the cargo-area pollution defense of a modern oil tanker, and keeping them straight is the first thing a design reviewer checks. Regulation 19 is the deterministic double-hull rule. It prescribes the protective distances: minimum wing-tank width and minimum double-bottom height as functions of deadweight, plus an upper cap on individual cargo-tank volume. It’s a geometry rule, satisfied by measuring the plan, and it keeps cargo physically away from the shell so that a moderate breach hits ballast space, not oil.

Regulation 23 is the probabilistic outflow check on that geometry. It takes the arrangement Regulation 19 permits and asks whether the expected spill across the casualty distribution actually stays low. The MEPC.122(52) notes are explicit that Regulation 23’s simplification is acceptable only because Regulation 19 is also in force: the double-hull provisions assure the likelihood-of-spill side that Regulation 23 stops computing, while Regulation 23 governs the expected magnitude. A tanker designer who meets Regulation 19’s clearances but arranges the cargo block poorly can still fail the $O_M$ limit, which is the whole point of having a performance check on top of a geometry rule.

Regulation 28 is the deterministic damage-stability rule, and it answers a third question: after a prescribed side or bottom damage applied at the worst location, does the tanker stay afloat and upright within a survival envelope? Regulation 28 is about the ship not sinking; Regulation 23 is about the cargo not spilling. They use different machinery, a single worst-case deterministic damage for Regulation 28 versus an integrated probability distribution for Regulation 23, and they’re satisfied independently. A common review error is to treat one as covering the other; neither does. The smaller Regulation 18 segregated ballast tank rule is the historical fourth piece: by giving the tanker dedicated ballast tanks in protective locations, it both removes the dirty-ballast discharge problem and contributes the wing and double-bottom volume that the Regulation 19 and Regulation 23 calculations then exploit.

The load-line and subdivision context binds these together. The intact draft fed into the outflow and stability work runs up to the assigned summer freeboard under the load line rules, which set the maximum loaded draft and the reserve buoyancy above it, and the cargo-tank subdivision that Regulation 23 rewards is the same finer subdivision that improves the Regulation 28 survival margin. The size bands that drive the $O_M$ limit map directly onto the standard tanker size classes: Aframax and Suezmax tonnage sits below the 200,000 m³ flat-limit threshold, while VLCC and ULCC tonnage runs into the interpolation band and the tighter 0.012 cap. A transverse cofferdam added to limit a flooded cargo block for damage stability also cuts the breach probability and the per-event outflow that Regulation 23 sums, so the two pull in the same direction even though their mathematics differ.

Worked illustration of the limit

Take a Suezmax-class crude tanker with a total cargo oil capacity at 98% filling of $C = 175{,}000$ m³, which is below the 200,000 m³ threshold, so the permissible mean outflow is the flat $O_M \leq 0.015$. From the MEPC parametric series a representative 150,000 DWT design with a five-long by two-wide tank arrangement and standard double-hull clearances returns $O_M$ near 0.014, inside the limit with a thin margin.

Now scale up to a VLCC with $C = 350{,}000$ m³. The capacity sits in the interpolation band, so the limit is $O_M \leq 0.012 + (0.003/200{,}000)(400{,}000 - 350{,}000) = 0.012 + 0.00075 = 0.01275$. The MEPC series shows 350,000 DWT designs with finer subdivision (six-long by three-wide, with two longitudinal bulkheads) reaching $O_M$ around 0.008 to 0.009, comfortably under the tighter cap, because the three-across arrangement plus the $C_3 = 0.77$ side correction pulls the side contribution down. The pattern across the parametric series is consistent: larger tankers must hit a lower $O_M$, and the finer subdivision they carry lets them do it.

The arithmetic is reproducible by hand once $O_{MS}$ and $O_{MB}$ are in hand. Suppose for the Suezmax the side calculation yields $O_{MS} = 2{,}300$ m³ and the combined two-tide bottom calculation yields $O_{MB} = 2{,}050$ m³. Then $O_M = (0.4 \times 2{,}300 + 0.6 \times 2{,}050)/175{,}000 = (920 + 1{,}230)/175{,}000 = 2{,}150/175{,}000 = 0.0123$, which clears the 0.015 limit. The volumes themselves come from the probability-weighted sums over every cargo tank, which is the part only software does at scale, but the final combination and comparison are plain arithmetic that a surveyor can spot-check.

The casualties that drove the rule

The probabilistic approach to tanker oil outflow grew out of a run of large spills that the deterministic Regulations 22 to 24 had not prevented. The 1978 grounding of the Amoco Cadiz off Brittany put roughly 223,000 tonnes of crude into the sea, the worst tanker spill on record to that point, and made the case that tank size and a single hull were not enough. The 1989 Exxon Valdez grounding in Prince William Sound, about 37,000 tonnes from a single-hull tanker, pushed the United States to the Oil Pollution Act of 1990 and its double-hull mandate, which in turn pressed the IMO to introduce the double-hull Regulation 13F under resolution MEPC.51(32) in 1992. The MEPC.122(52) notes name the 1992 amendments and Regulation 13F as the direct technical ancestor of the modern double-hull and outflow rules.

Two later casualties accelerated the wholesale revision of Annex I that produced Regulation 23. The 1999 sinking of the Erika off the French coast and the 2002 break-up of the Prestige off Galicia, both single-hull product or heavy-fuel tankers, spilled tens of thousands of tonnes and triggered the accelerated single-hull phase-out under resolution MEPC.111(50) in 2003. That political momentum carried through to MEPC 52 in October 2004, where the committee adopted the revised Annex I (MEPC.117(52)) and the outflow Explanatory Notes (MEPC.122(52)) together. The lesson the regulators took from the casualty record was that a performance metric, not a geometry cap, was the way to reward good subdivision, and the historical damage data from those decades became the statistical base the metric runs on.

The relevant data point for Regulation 23 specifically is the IMO comparative study on oil tanker design, the OTD study, which the notes cite repeatedly. It assembled the collision and grounding statistics, ran model tests on bottom-damage retention, and demonstrated that the mid-deck design (a horizontal partition keeping cargo head below the external sea head by the hydrostatic balance principle) achieved outflow performance at least equivalent to a double hull. That equivalence finding is why Regulation 13F paragraph 5, and later Regulation 19, left the door open to alternative arrangements assessed against a reference double-hull ship, and why the outflow methodology had to be general enough to score any arrangement, not just a conventional double hull. The mid-deck tanker, though rarely built, is the reason the outflow rule exists as a performance test rather than a prescriptive geometry.

Reading the probability tables

The breach-probability factors are not guessed; they’re read off integrated tables built from the five density functions per damage mode. The construction in section 4 of the Explanatory Notes is worth following because it explains why the probability tables don’t run to 1.00 and why the dimensions multiply. For one dimension, the function $p_b(d)$ gives the probability the damage is restricted to less than the normalized location $d$, found by integrating the location density $h(x)$ against the extent density $g(y)$ over the region below $d$:

and $p_a(d)$, the probability the damage is restricted to more than $d$, integrates over the region above. The functions define damage location as the center of damage, so a damage zone near an end or side can extend beyond the hull, which is why the cumulative probabilities stop short of 1.00. The probability that a band bounded below by $d_1$ and above by $d_2$ is breached is then $p = 1 - p_b(d_1) - p_a(d_2)$, and this counts every damage that includes the band, not only damage confined to it.

Each cargo tank is reduced to an equivalent rectilinear block with six boundaries, three pairs of planes. For side damage the boundaries are $X_a$ and $X_f$ (the aft and forward longitudinal terminals measured from the aft end of the rule length $L$), $Z_l$ and $Z_u$ (the lowest and highest points above the molded baseline, with $Z_u$ capped at the molded depth $D_s$), and $y$ (the minimum horizontal distance from the tank to the side, measured at right angles to the centerline). The designer pulls these five-plus-one numbers off the capacity plan for every cargo tank, looks up the marginal probabilities in each dimension, and multiplies. The independence assumption is what makes the lookup-and-multiply shortcut valid; the notes are candid that it was adopted for lack of data on the dimensions’ correlation, not because correlation was shown to be zero.

The integration also explains the conservatism the $C_3$ factor corrects. Using the minimum horizontal distance $y$ to the side for the transverse breach probability is exact only where the side is vertical and parallel to the tank boundary. In way of the forward and aft cargo tanks, where the hull narrows and the shell slopes inboard, the true clearance varies and the minimum understates the average, so the method over-counts breach probability and outflow there. For the large three-across designs the cumulative effect of that over-count across many tanks is enough to warrant the flat 0.77 multiplier on side outflow; for smaller single-centerline-bulkhead ships the effect is small and $C_3$ stays at 1.0. A reviewer who sees a suspiciously high side contribution on a full-form bow section is usually looking at this artifact, not a real design weakness.

Outflow from a breached bottom tank

The hydrostatic-balance outflow from a single bottom-breached tank is the physical heart of the bottom calculation, and it’s worth stating the mechanism rather than treating $O_{B(i)}$ as a black box. When the outer and inner bottoms are torn and the sea reaches a cargo tank, oil and seawater redistribute until the pressure at the breach from the oil column inside equals the pressure from the sea column outside. Cargo oil, lighter than seawater, floats on top, so the tank ends up with a layer of retained oil above a layer of ingressed seawater, and the oil lost is the difference between the original cargo volume and the retained volume at the new equilibrium. The lower the cargo density relative to seawater and the greater the breach depth below the waterline, the more oil the balance retains.

Two refinements sit on top of the bare balance. First, even a tank that the balance says should retain its oil loses about 1% of its volume to initial impact exchange and to wave and current action, so the method floors the bottom outflow from a breached tank at 1% of tank volume; a tank in perfect hydrostatic balance is not treated as a zero spill. Second, the oil that does flow down through the breach is not all lost to the sea, because some is caught in the double-bottom void below. The $C_{DB(i)} = 0.60$ factor encodes that retention: 40% of the balance outflow is assumed entrapped in the non-oil tank below, based on the model-test finding that retention ran between 0.50 and 0.70 for ten real double-hull tankers, with the breached double-bottom space taken as 50% oil and 50% seawater at equilibrium.

Regulation 23.7.3.2 adds a careful instruction for heavy cargoes. In a real grounding, a cargo denser than seawater can drain almost entirely from a bottom-breached tank, because the heavier oil head wins the balance. But for the regulation’s bookkeeping, even when the nominal cargo density computed from deadweight over volume exceeds seawater density, the cargo level and remaining oil are still calculated on the hydrostatic balance of paragraph 7.3.2. The point is uniform application: the outflow number is a standardized design index, not a prediction of a specific casualty, so the same balance rule applies regardless of the density the deadweight-over-volume convention assigns.

Combination carriers and the separate criterion

Combination carriers, the OBO and ore-bulk-oil types built to carry dry bulk one leg and oil the next, get the one explicit relaxation in Regulation 23. Built without a centerline bulkhead so the holds can take bulk cargo, they can’t match the standard oil-tanker $O_M$ line on geometry alone, so Regulation 23.3.1 lets a combination carrier apply its own outflow criterion. The price is a demonstration, to the Administration’s satisfaction, that the heavier hull structure delivers environmental protection at least equal to a standard double-hull oil tanker of the same size.

Part B of the Explanatory Notes spells out what that demonstration takes. The reference standard tanker must itself comply with MARPOL 73/78, including the wing-tank width and double-bottom height rules, with scantlings sized for a tanker of the same size as the combination carrier and the same loading conditions apart from the dry-bulk cases. The combination-carrier analysis is then a series of finite-element collision and grounding calculations, each developing the dissipated plastic deformation energy at the damage position, with the combination carrier modeled as the struck ship at full load against striking ships at draft differences that define the strike geometry. The output has to show the stronger structure cuts the damage extent enough that the resulting outflow matches or beats the standard tanker. It’s a far heavier submission than the standard tank-by-tank probability sum, which is why most owners build conventional double-hull tankers and take the standard criterion rather than chase the combination-carrier allowance.

The lenient combination-carrier line is not open-ended. It starts above the oil-tanker line at small sizes (near $O_M$ of 0.021 at 5,000 m³) and converges onto the standard 0.015 line by about 200,000 m³, so for large combination carriers the allowance largely disappears and the standard limit governs anyway.

Survey, documentation, and port-state control

Regulation 23 compliance is verified at the design stage and recorded, not re-checked each voyage, which sets it apart from the operational rules elsewhere in Annex I. The recognized organization acting for the flag state reviews and approves the $O_M$ calculation as part of the newbuild damage-stability documentation, and the approved report is held aboard alongside the Regulation 28 survival calculations and the stability booklet and loading computer records. The classification society that performs the review applies the IACS Unified Interpretations in the MPC series so that the methodology reads the same across member societies, which matters because $O_M$ is a single number that a port-state inspector in any jurisdiction can check against the size-dependent limit.

At the renewal survey for the International Oil Pollution Prevention certificate, the surveyor confirms the damage-stability and outflow documentation is aboard, current, and matches the as-built subdivision. A port-state control officer’s interest in Regulation 23 is narrower than the operator’s: the officer checks that the document exists and corresponds to the ship, not that the underlying $O_M$ arithmetic is right, because the figure is a design property certified at approval. The detainable items in this corner of Annex I are the operational ones, a missing or invalid IOPP certificate, an out-of-date loading manual, an inoperative approved stability instrument, rather than the outflow index itself.

The one event that re-opens the $O_M$ calculation is a change to the cargo-tank subdivision. A conversion that moves a bulkhead, combines or splits cargo tanks, or alters the double-hull clearances changes every breach probability and outflow in the sum, so the original approval no longer applies. The ship needs a re-run and a re-approval before the documentation is valid again, and a surveyor who finds an undocumented structural change to the cargo block treats the outflow and damage-stability documentation as suspect until it’s reconciled. This is the same discipline the Regulation 28 damage-stability calculation carries, and in practice the two are re-approved together because they share the tank-arrangement input.

Limitations and practitioner notes

Regulation 23 measures expected outflow, not worst-case spill, and that’s its first limit. It reports the mean across the casualty distribution; the probability-of-zero-outflow and extreme-outflow parameters that the Revised Interim Guidelines carried are dropped. A tanker can hold a low $O_M$ and still suffer a large single-casualty spill in the upper tail. The regulation accepts this because the Regulation 19 double hull is presumed to handle the low-probability, high-consequence end, and because the mean was judged the best single indicator. A reviewer who treats $O_M$ as a cap on the worst spill has misread it.

The statistical base is fixed and aging. The probability density functions rest on 52 collisions and 63 groundings from 1980 to 1990, predominantly to single-hull tankers, of 30,000 DWT and above. The MEPC.122(52) notes themselves call for periodic review as new data becomes available, and that base has not been broadly refreshed. The distributions also treat the damage dimensions as statistically independent for lack of data on their correlation, which is a modeling convenience, not a measured fact. For an unusual hull, this matters more than for a standard prismatic double-hull tanker that the dataset resembles.

The simplified method has stated geometric blind spots. Paragraph 1.9 of the notes says the minimum-clearance shortcut suits regular tank arrangements but that designs with steps or recesses in decks, sloping bulkheads, or pronounced hull curvature may need the more rigorous sub-compartment calculation of Regulation 23 paragraph 10. The minimum-horizontal-distance assumption for transverse breach probability over-estimates outflow in the forward and aft cargo tanks where hull curvature is sharpest, and the $C_3 = 0.77$ correction only addresses the large three-across case, not curvature generally. A designer relying on the simplified method for a non-rectangular arrangement can produce a result that’s conservative in some tanks and not representative in others.

Scope and exemption boundaries trip people up. Regulation 23 applies only to oil tankers of 5,000 DWT and above; the probabilistic outflow criterion does not extend to smaller vessels, where, per Regulation 23.3.2, tank size is instead governed by the 700 m³ tank-size limit and maximum tank length of Regulation 19 paragraph 6.2. Combination carriers can use a separate criterion only by demonstrating equivalent protection through finite-element collision and grounding analysis to the Administration’s satisfaction; the allowance is not automatic. And $O_M$ is a design property, not an operational one: it’s fixed at approval and stays valid until the cargo-tank subdivision is changed, so a conversion that alters tank boundaries voids the original calculation and forces a re-approval before the result can be relied on.

See also

• MARPOL Annex I oil pollution prevention
• MARPOL Annex I Regulation 28 damage stability
• MARPOL Annex I Regulation 12A oil fuel tank protection
• MARPOL Annex I Regulation 33 crude oil washing
• MARPOL Annex II Regulation 16 cargo tank arrangement
• Probabilistic damage stability
• Oil tanker
• MARPOL Convention
• Accidental oil outflow calculator (MARPOL Annex I/23)