Noah's Ark: The Logistics

"Based on what we know from other periods of history, a seagoing ship built entirely of wood might safely be 70–75 meters in length, but when the hull got longer than this, naval architects employed metal bracing, like iron or steel girders inside the hull." — William M. Murray, The Age of Titans, p. 173, on the largest ships of antiquity¹

"The Ark is near the maximum size that is known to be possible for a wooden vessel." — Tim Lovett, Answers in Genesis²

"These instructions are much too short to serve as a practical blueprint for construction.  Rather they give the minimum details required for making sense of the flood story." — Gordon Wenham, Genesis 1–15, p. 173³

Summary

The usual objection is that Noah could not possibly have fitted millions of species onto a wooden boat.  That objection is wrong, and it is wrong for a reason that is easy to check:  the median land vertebrate weighs about 19 grams.⁴  You could hold it in a closed hand.  The Ark's problem was never the number of animals.

But almost everything else that gets said in the Ark's defense is either unsourced, arithmetically wrong, or quietly assumes the answer.  This article works the problem from published data — taxonomic databases, veterinary space standards, agricultural engineering handbooks, wood mechanics, and live-export shipping records — and reports what comes out, including where it comes out against us.

Here are the findings, ordered by how much each one decides.

1. The animals fit, but only under two specific commitments, and the commitments have to be stated out loud.  Computed on measured floor-area standards rather than volume, two adults of every extinct and living land-vertebrate genus need 20,687 m² of deck against the Ark's 7,574 m² at the conservative cubit.  That fails by a factor of 2.7.  Switch to measured juvenile masses and the same cargo needs 2,214 m², which fits with room to spare.  Every one of the 81 scenarios tested fits when juvenile masses are used; every one that fails uses adult masses or the traditional "a juvenile weighs half an adult" rule.  The Ark's feasibility rests on juveniles and on kinds coarser than genus.  That is a real position, but it is a position, not a detail.

   And no, stacking the cages does not rescue it.  Grant unlimited free tiering inside each 4.15 m deck and the adult-mass manifest improves by five per cent, still overrunning the ship 2.6-fold.  Anything heavier than about 90 kg cannot be double-tiered, and animals above that line hold 94% of the required area.  The 13,325 genera that can stack eight high are together about one per cent of the problem.

2. The single most-repeated capacity figure in Ark apologetics is inflated about six-fold, and it rests on a misquotation.  "The Ark could hold 125,000 sheep" is volume arithmetic applied to a problem that is limited by floor area.  Fitted against the Australian live-export standard — the closest real-world analogue to the Ark that exists — the honest answer is about 22,000 sheep.  Worse, the underlying "240 sheep per railroad stock car" figure traces through The Genesis Flood to a 1919 livestock textbook that said 120 sheep per deck, called it an average, and added "These figures represent mere averages."⁵  Whitcomb and Morris turned an average into a floor and doubled it.

3. The popular "juveniles weigh half of adults" rule hurts the Ark's own case by a factor of six — though the best creationist treatment never used it.  Fitted against Myhrvold's amniote life-history database, reptiles hatch at 13% of adult mass scaled by M^0.59, which for a sauropod means a hatchling of about 3 kg out of 50 tonnes, not 25 tonnes.  The measured rule is far kinder to the Ark than the shortcut.  It is also brutal in the other direction for birds, which fledge at a median 83% of adult mass, so taking juvenile birds buys almost nothing.  Credit where it is owed:  Woodmorappe already used a mass-dependent juvenile factor in 1996 — 0.45, 0.34 and 0.05 for his small, medium and large mass classes — and its direction matches the measured allometry, with the largest animals discounted hardest.  The half-mass rule belongs to the popular literature and to this site's own old notes, not to him.

4. The ventilation calculation decides an exegetical question, and it decides it against the traditional picture.  Metabolic heat, not oxygen and not carbon dioxide, is the binding constraint in any confined animal building.  A single one-cubit window supplies roughly 300 m³/h of air in a light breeze.  The leanest defensible fauna needs about 86,000 m³/h.  That is a shortfall of about 270-fold, and no amount of husbandry ingenuity closes three orders of magnitude.  The Ark works only on the reading where the tsohar is a continuous cubit-high opening running the full perimeter under the eaves — and even that reading is under half the published cold-weather minimum for a naturally ventilated barn once three decks are counted.

5. Feed is not a problem for living animals and is a fatal problem for adult dinosaurs.  At family-level kinds with living animals only, feed and water together are about 2% of the hull.  Add every dinosaur genus at adult mass and the food alone is 87% of the ship on the generous assumption that dinosaurs metabolized like reptiles, and 153% if they metabolized like mammals.  The consumables budget closes at family-level kinds with juveniles, at about 9% of the hull, and nowhere else.

6. The hull is not impossible, but nobody has demonstrated that it is possible, and the study everybody cites does not demonstrate it.  Run independently, a 137 m wooden box with 30 cm scantlings sees about 5.4 MPa of bending stress under a standard wave bending moment, against green white oak's 57 MPa modulus of rupture — a margin of 10.5.  The skeptical claim that the hull must snap in half is, on the arithmetic, wrong.  But bending was never the tight check.  Seam shear is, and meeting it requires scaling the densest edge-joinery in the archaeological record up by a factor of 4.6 into a joint nobody has ever excavated.

7. The Korean seaworthiness study cannot carry the weight put on it, and we should say so before a critic does.  Hong et al. (1994) modelled the frames as truss elements and the planking as membrane elements joined at rigid nodes.⁶  Those element types are physically incapable of representing joint slip, fastener working, or seam failure — the exact mechanisms that destroy large wooden hulls.  The paper names no wood species, no elastic modulus, no allowable stress, and no computed wave bending moment.  Its famous "30 metre waves" figure is a straight-line extrapolation roughly four times beyond the largest sea state it actually computed, all of whose data points stop at about 11 m.  And on its own combined index the Ark placed second, not first — a fact Answers in Genesis's own Tim Lovett confirmed when he re-ran the numbers.⁷

8. The mortality benchmark is the hardest single fact in the file.  Australian long-haul sheep voyages averaging 15.6 days lost 1.3% of their animals across 1988–2017.⁸  The Cormo Express, a purpose-built mechanically ventilated livestock carrier with a veterinarian aboard, lost 9.82% of 57,937 sheep over 80 days.  Extrapolated at constant hazard to 370 days, modern best practice gives about 3.6%.  On a ship carrying one breeding pair per kind, a 3.6% loss of animals is close to a 3.6% loss of kinds.  The dominant killers are not heat or ammonia; they are inanition and salmonellosis — animals that will not eat, and disease.  That is good news for the engineering and bad news for the assumption that wild animals will simply eat what they are given.

9. Two of the hardest objections are not logistics questions, and this article defers them rather than pretending they are solved.  Post-Flood biogeography — where the difficulty is not that marsupials are in Australia but that there is no dispersal trail anywhere between Ararat and Australia — and the founding-population problem, where a diploid pair carries at most four alleles per locus.  Both are getting their own articles.  Both are acknowledged in section 8 with the strongest version of the objection, including Answers in Genesis's own concession in its journal that the required dispersal is "not only unlikely, it is highly improbable."⁹  Nothing below should be read as answering either.

10. Several arguments in wide circulation on our side do not survive checking and should be dropped.  Little Tyke the vegetarian lion ate milk and eggs, which supply the taurine and preformed vitamin A that a cat cannot make.  The Galápagos tortoise fast is documented at four to six months, not a year.  Ussher's Leontifera is dated 280 BC, not 280 AD, and no ancient source gives it any dimension at all.  Zheng He's 450-foot treasure ships trace to a 1597 novel in which the ships are built by an immortal, and the shipyard basin they were supposedly built in was excavated in 2003–04 and is 41 m wide, which forbids the claimed 56 m beam.

The shape of the honest answer

Volume was never the objection.  Everyone on both sides has been arguing about the wrong quantity.  The Ark is not short of cubic metres; it is short of deck area, air, and labour hours, in that order of tightness, and the popular defenses address none of the three.

The load is extraordinarily concentrated.  Between 27% and 87% of total animal mass sits in the top 1% of kinds, depending on the scenario.  For living genera at adult mass, 80% of the mass is in 53 genera out of 5,305.  The Ark problem is not "how do you fit sixteen thousand animals."  It is "what do you do about the elephants and the sauropods."

The strongest pro-Ark results in this article are ones no apologist seems to have run, and the weakest claims against the Ark are the ones apologists spend the most effort answering.  That asymmetry is worth noticing.

What this article does

It works the arithmetic in the open and publishes the scripts, so a hostile reader can rerun everything.  It uses published standards — the USDA Wood Handbook, ASAE manure tables, the NRC laboratory-animal guide, the Australian Standards for the Export of Livestock — rather than plausible-sounding estimates.  Where a number could not be verified at the primary level it is marked as unverified rather than quietly used.  And it grades our own side by the same standard as the critics, which means several sections below are corrections to arguments this site would otherwise have repeated.

The full source audit is in the companion page:  the literature.

1. What the text actually specifies

Genesis 6:15 gives three numbers:  three hundred cubits by fifty by thirty.  Everything else in the specification is remarkably thin, and the thinness is the first thing an honest treatment has to concede.

Four of the five load-bearing words are rare or unique.  Gopher wood occurs exactly once in the Hebrew Bible, and Brown-Driver-Briggs says the word is dubious and reports a proposal that it is a copyist's corruption.¹⁰  Kopher, the pitch, occurs once in this sense — and it is not the ordinary Hebrew word for bitumen, which is ḥēmār and appears at Genesis 11:3, Genesis 14:10, and Exodus 2:3.  Kopher is the Akkadian kupru.  S. R. Driver's judgement, in the Westminster commentary, is that "it is possible therefore that kopher came into Heb., with the story, from Babylonia."¹¹

That is a concession, and it should be made cheerfully.  A text that supplies three numbers, three unglossed technical nouns, and one clause that John Skinner calls "unintelligible as it stands" is not a shipyard drawing.¹²  Wenham, an evangelical, puts it exactly right in the epigraph above.  Apologetics that presents the Ark specification as unusually detailed and technical has it backwards.

The cubit, and why it matters more than anything else in the article

Volume scales as the cube of the cubit, so this one choice moves every downstream number.

The only Iron Age Hebrew length that can be checked against a surviving structure is Hezekiah's tunnel.  The Siloam inscription states the tunnel's length as 1,200 cubits; the tunnel measures about 533 m; the quotient is 0.4442 m.¹³  That is the short or common cubit, and it is the defensible default.

Answers in Genesis instead uses 20.4 inches (51.8 cm), which inflates the Ark's volume by 1.58× relative to Siloam.  That figure has no artifact behind it, does not appear in AiG's own royal-cubit table, and is derived as seven-sixths of a figure in a 1959 journal article.  Woodmorappe works at 45.72 cm.  The creationist literature is not consistent about this, and readers deserve to be told that the Ark's size is a range, not a number.

Honesty runs the other way too.  The cubit evidence is softer than either side admits.  Lepsius says in print that the classic Egyptian cubit rods are votive grave goods with unevenly spaced divisions, and calls one of the ten he catalogues a forgery.  Kha's rod, which is a working tool rather than a grave good, varies by ±3.4% from palm to palm.  The Siloam figure of 1,200 is partly a restoration on the stone and is almost certainly rounded.  There is no Israelite architectural metrology in hand at all.  Nobody on either side knows the cubit of Genesis 6:15 to better than about ±15%.

This article uses the short cubit as its default, because the conservative choice is the one that survives a critic.  Where the royal cubit changes a verdict, that is stated.

The proportions are genuinely good, and the reason is not the one usually given

The Ark's ratios are 6 : 1 : 0.6.  Schneekluth and Bertram's standard merchant-ship design method is calibrated over length-to-beam ratios of 4.7 to 7.63, length-to-depth of 8.12 to 15.48, beam-to-depth of 1.47 to 2.38, for ships between 100 and 250 m.  The Ark falls inside all four bands and inside the length range.  That is a real result and it is worth printing.

But do not overclaim it.  The same reference puts minimum resistance at a length-to-beam ratio of 8 to 9, so 6 : 1 is not speed-optimal.  What 6 : 1 is optimal for is a low-speed hull that wants to minimise wetted surface and maximise capacity — modern bulk-carrier practice, and exactly right for a vessel that does not go anywhere.

The window, and the compartments, are both genuinely uncertain

Two ambiguities in the Hebrew turn out to control physical arguments later in this article, so they have to be laid out before the physics rather than after it.

The tsohar of Genesis 6:16 might be a single window, or it might be a continuous cubit-high gap running under the eaves.  The gap reading is not creationist special pleading:  it is S. R. Driver's, who describes "a kind of casement running round the sides of the ark… a little below the roof,"¹¹ and it is one of Wenham's two preferred options.  But Skinner calls the underlying clause unintelligible, Wenham calls it "the most obscure remark in the flood story," and Wellhausen, Budde, Ball and Gunkel all emended it.  The difference between the readings is roughly 137 m² of opening against 0.2 m².  Section 6 shows that the physics requires the large reading.  That is a case of engineering constraining exegesis, which is a legitimate move, but it should be labelled as what it is.

The qinnim of the same verse are usually translated "rooms" or "nests," and are the basis for every claim that Genesis specifies internal compartments.  Wenham translates the word "reeds," following B. Jacob, G. R. Driver, McEvenue, and the NEB, on a repointing to qanim.  If they are right, Genesis specifies no internal subdivision at all.  An apologetic that says "Genesis specifies compartments for the animals" is choosing one of three readings without telling the reader that it chose.

The Mesopotamian comparison, stated carefully

The other ancient flood boats are worse boats, and this is genuinely interesting.  The Gilgamesh ark is a cube, 120 cubits on each side, with seven levels and nine compartments.  The Ark Tablet published by Irving Finkel describes a giant round coracle.  A. R. George, in the standard critical edition of Gilgamesh, says of the cube in his own voice:  "The configuration of this strange boat was obviously not drawn from practical experience."¹⁴

Genesis is the only account in the set whose vessel has the proportions of a ship.  That is worth saying.  What it is not is evidence that Genesis is earlier.

The mainstream Assyriological position is that Genesis 6–9 is later than and dependent on the Mesopotamian material.  Finkel states it plainly:  "the dependence of one upon the other is unavoidable," and proposes Judaean exiles learning cuneiform in the Babylonian school curriculum as the mechanism.¹⁵  Kitchen's conservative reply — that "the respective boats differ totally; Noah's has shiplike proportions, but the Mesopotamian one was a cube!" — is exactly the argument of the paragraph above, and used as evidence of priority it is circular.¹⁶  "The later text improved the shape" explains the data just as well.  Millard's argument that Genesis is more concise and therefore not an elaboration is real but weak, since an epitomiser produces the same signature.¹⁷

Claim only that Genesis's specification is better engineering.  Never that it is therefore earlier.

One more concession belongs here, because it will otherwise be found.  The 2014 NOVA reconstruction of Finkel's coracle, built on Lake Vembanad in Kerala, floated.  NOVA's verdict was that "the boat-building team never managed to duplicate Babylonian bitumen, but the structure of the ark is sound."¹⁸  The leak was a waterproofing-materials failure, because Iraqi bitumen could not be exported.  Do not claim the coracle sank.  The same programme also puts the Mesopotamian flood-layer dates on screen — Ur 3,500 BC, Uruk 2,900 BC, Shuruppak 2,700 BC, Kish 2,500 BC — and notes that they were "caused by different events, spread over a thousand years."  That is fatal to the "Woolley found Noah's flood" apologetic, and conceding it costs nothing.

2. Could a wooden hull that size exist?

This is the hardest technical objection, and it deserves the most careful treatment.

Bending is not the problem

The intuitive skeptical claim is that a 137 m wooden hull must break its back in a seaway.  Work it out and that claim fails.

Model the Ark as a wall-sided box girder with 30 cm scantlings, apply the standard IACS wave bending moment for a hull of that length, and the peak fibre stress is about 5.4 MPa.  Green white oak has a modulus of rupture of 57 MPa in the USDA Wood Handbook's Table 5-3a.¹⁹  That is a margin of 10.5.  Put the ship on the steepest non-breaking wave of its own length — a 19.6 m monster — and the stress rises to 16.4 MPa, still 3.5 times below rupture.  A rectangular box is also the best possible shape for this job:  it has the maximum section modulus for its dimensions, and when evenly loaded it has essentially no still-water bending moment at all.

The Ark's section modulus is roughly 32 times what a steel ship of the same length would need.  There is a great deal of margin in bending.  Any critic who argues from bending alone has picked the wrong failure mode.

Shear at the seams is the real constraint, and it is close

Wood's shear strength along the grain is only about 15% of its bending strength, so the shear check is tighter.  Hull-girder shear works out at 1.95 MPa against green oak's 8.6 MPa — a factor of 4.4, which is 2.4 times tighter than bending.

The binding requirement is not the plank; it is the seam between planks.  Every seam must be continuously keyed over about 23% of the plank thickness, which for a 300 mm plank means a 68 mm oak spline running the length of the joint.

Priced against measured archaeology, this is the most interesting number in the article, and it cuts both ways.  A 2025 finite-element study of ancient ship joints measures a reconstructed Uluburun seam at 84% of the bending strength and 80% of the tensile strength of an unjointed plank of the same thickness.²⁰  That is a genuinely strong joint, and Uluburun-grade joinery — 62 × 20 mm tenons with the mortises nearly touching — delivers 0.19 to 0.27 against the 0.23 requirement, meeting it within a factor of about one either way.

But that result is about proportions, not about attested joints.  To use Uluburun's proportions on a 300 mm plank, the tenons must scale up 4.6× to roughly 285 × 92 mm at 370 mm centres, and no joint that size has ever been excavated.  Bronze-Age-sized tenons dropped into an Ark-sized plank fall 4.4× short.  The sparser Kyrenia joinery of a thousand years later is about 15× short.  The variable is joinery density, not joinery type — which is at least a specific, checkable claim.

There is a further problem that the seam calculation does not touch.  The same 2025 study measures ancient keel scarf joints at only 10% to 40% of the strength of an unjointed beam.  A 137 m keel needs at least six scarfs, and each one is a hinge.  The excellent 84% seam efficiency is a good number attached to the wrong joint.

The study everyone cites assumed the answer

Hong et al., "Safety Investigation of Noah's Ark in a Seaway," is the paper behind every creationist claim about Ark seaworthiness.⁶  It is real naval architecture by real naval architects at a real institution, and it should be cited.  It should also be cited accurately, and it usually is not.

Its structural finite-element model represents the frames as truss elements and the planking as membrane elements, joined at perfectly rigid nodes.  Truss elements carry only axial load.  Membrane elements carry only in-plane load.  A model built from those two element types cannot represent joint slip, fastener working, plank sliding, or seam failure — which are precisely the mechanisms that destroy large wooden hulls, and precisely the mechanisms the previous subsection identifies as binding.  The paper never states a numerical allowable stress, never names a wood species, never gives a modulus of elasticity, and never publishes the wave bending moment it computed.  Its entire structural conclusion is one sentence:  "Because the maximum stress was smaller than the allowable stress, the Ark could be said to have had safe structural performance."  As published, that is unfalsifiable.

The famous "30 metre waves" number is worse.  Every computed data point in the study's figures stops at a significant wave height of about 11 m.  The 30 m, 43 m and 47.5 m figures are read off straight lines drawn beyond the data, using linear strip theory whose small-amplitude assumption is void long before then, for a hull whose total depth is 13.5 m.

And the "optimal proportions" claim is false as usually stated.  Answers in Genesis's own Tim Lovett re-ran Hong's numbers and found the Ark seventh of thirteen in seakeeping, sixth in strength, fifth in stability, and third or fourth overall under most sensible weightings.⁷  "A good all-round compromise" is what the data support.  The study was also fully supported by the Korea Association of Creation Research, and its comparison set was generated by scaling the Ark itself.

The historical precedents mostly do not survive checking

Creationist literature carries a standard list of large ancient ships offered as precedent.  The list traces to a single article:  Larry Pierce, "The large ships of antiquity," Creation 22(3):46–48 (2000).²¹  Checked against the primary texts, most of it fails.

Two patterns in that table are worth more than the individual verdicts.  First, when antiquity needed to float something enormous, it yoked several ordinary hulls together.  The "forty," the Leontophoros, and the obelisk carrier were all multi-hulled on the mainstream readings.  That is a coherent engineering culture, and it is the opposite of a single box.  Second, every one of these ships was a one-off that was then retired.  The "forty" made a trial run and was never used.  The Syracusia made one voyage.  Caligula's obelisk ship was filled with concrete and sunk.  A skeptic can fairly say the pattern is that the biggest hulls ever built were not economic and everyone stopped.

The naval-archaeology consensus figure is the epigraph to this article, and it comes from the scholar of exactly these giant Hellenistic ships.  Murray puts the safe limit for an all-wood seagoing hull at 70–75 m — roughly half the Ark's length — and he states it knowing all about the "forty."¹  That figure has to be read for what it is:  a limit on antiquity's construction methods, not a limit on wood.  Every large hull discussed below exceeds it — the Great Republic at 102 m, the Wyoming at 107 m of hull, the Dunderberg at 115 m, and 101 completed First World War wooden steamers at 86 m that made 305 voyages over 490,000 statute miles "with substantially no loss to the shippers."  The Ark at 137 m still exceeds all of them, and that concession stands.

The industrial-age record is the other half of the objection, and it turns out to have been reported badly by nearly everyone, including us.

The Wyoming was not 450 feet long.  Her hull was 350 ft over all and 329.5 ft on the register, 50.1 ft in beam, 30.4 ft depth of hold, and 304 ft on the keel.²²  The 450 ft figure — which our own notes repeated — is jib-boom tip to spanker-boom tip, adding about 100 ft of overhanging spars to a 350 ft hull.  This matters in the skeptic's favour as often as ours:  anyone claiming "the Ark was the same size as the largest wooden ship ever built" is wrong, because the Ark at 137 m is 28% longer than the Wyoming's entire hull and roughly four times her internal volume.

She was not cheaply built, and this article previously conceded that she was.  That concession was false.  She was built under American Bureau of Shipping supervision, iron-strapped and salted, and classed A1 for sixteen years — the longest term the Bureau granted any wooden vessel.  She was opened up for survey at Bath in 1916 and passed, and still held that class in 1923.  Her published maintenance record shows the deck caulked once in ten years.  Percy & Small was not the cheap yard; the standard historian of the trade names a different yard for that.  The honest sentence is therefore stronger than the one it replaces:  she was built to the best standard the industry had, and still had a fourteen-year life.  That moves the argument from "she was cheap" to "twelve to sixteen years is what wood gave you, and the Ark needs one."

She did not break up at sea.  She was lost in March 1924 at anchor on Nantucket Shoals, loaded with 6,000 tons of coal, in a north-east gale; the wrecks lie in thirty feet of water, and a loaded Wyoming drew nearly thirty.  Contemporary accounts all give a bottom-contact mechanism.  Her whole crew of twelve to fourteen was lost.  That is a stranding, not a mid-ocean hull failure, and "foundered" invites the wrong picture.  A twenty-two-year-old five-master rode out the same gale in company with her and came home.

But the leaking claim is true of the class, and that is the part that counts against us.  W. J. Lewis Parker, the standard historian, writes that "all the large schooners leaked badly and were forced to depend heavily upon their steam pumps," and W. H. White, the Royal Navy's Director of Naval Construction, states the principle:  "in wood ships of large size the entire prevention of such working is practically impossible."  The measured hogging figure that everyone half-remembers is not the Wyoming's at all — no hog measurement for her exists in any serious source — it is the Edward J. Lawrence, "an exceptionally well-built schooner," crowned a full thirty inches on her keel after only nine years.

And the class did not die of broken backs.  Of the nine wooden six-masters, five were lost by grounding, two by fire, and two foundered — and the two that foundered were the two oldest, at 19 and 20 years.  Average life 13.2 years.  The steel seven-master Thomas W. Lawson lasted five.  What killed these ships was lee shores, shoals, collision and fire, not the hull girder.

On the Great Republic, cite Hall, not Casson.  She was the largest wooden merchant hull ever built by tonnage and internal volume — four decks, 4,555 tons old measurement — though at 334½ ft she is exceeded in length by both the Wyoming and the Dunderberg.  Henry Hall's federal census monograph gives the actual iron:  "The frames on the inside were strapped with iron by braces 4 inches wide, 1 inch thick, and 36 feet long, there being 90 straps on each side of the ship."²³  Note for anyone checking our sources:  that 90-straps-per-side figure circulates misattributed to the Wyoming.  It is the Great Republic's.  Hall also answers the "no tree is 137 m long" objection in passing:  "There was little timber in the Great Republic longer than 50 and 60 feet, but the oak was in much shorter lengths."

Does the schooner record transfer to an unrigged box?

Partly, and less than an apologist would like.

Parker's own explanation of why the big schooners hogged does not transfer.  His causes are length without corresponding depth of hold, ends that "were quite fine and very heavily weighted," anchors and windlass forward, a great overhanging stern aft, so that "the ends therefore lacked the buoyancy which supported their long flat midsections."  A rectangular unrigged box is missing every single term in that list.

But the rig matters less than the argument needs.  White ranks the four classes of strain on a ship in order of importance and puts longitudinal hogging and sagging first and strains from propulsion by sail or steam third, calling the steady rig-induced racking moment "so small in amount as to be practically unimportant in its effect upon the ship considered as a whole."  Deleting the masts deletes local damage, not the dominant load.  The Ark's real structural advantages are its full ends, uniform loading, three continuous decks, enormous depth, and a one-year life instead of twenty-five.

There is also now one piece of genuine experimental data where this article previously had only reasoning.  Seppings' three-ship trial is a controlled experiment:  the Nelson, St Vincent and Howe were 120-gun ships of the same form, dimensions, frames, beams and planking scantlings, differing only in diagonal bracing.  Launch hog was 9½ in, 9¼ in, and 3⅝ in.²⁴  Construction method changed longitudinal behaviour by a factor of 2.6 at fixed length and fixed scantlings.  That supports the article's core structural claim — how a long wooden hull is built matters as much as how long it is — and it is period-primary rather than inference.

What remains inference, and must be labelled as such:  nobody has ever built a 137 m unrigged wooden box.  The longest wooden hulls ever built and operated were the Dunderberg at 377 ft, the Great Republic at 334½, the Wyoming at 350 over all, and the Great Lakes tow-barge Santiago at 324.  The Ark exceeds every one.

Verdict on the hull

The pro-Ark technical case is real: a box is the best shape for the job, there is an order of magnitude of margin in bending, and the densest attested Bronze Age joinery is at or just above the seam-shear requirement when scaled.  The anti-Ark case is also real: no unrigged wooden hull of this length has ever been built, the seam requirement is met only by a joint nobody has excavated, the keel scarfs are hinges, and the one study everybody cites assumed the failure mode away.

This is an inference, not a demonstration, and the article says so in those words.  Answers in Genesis's own posture — "near the maximum size that is known to be possible for a wooden vessel" — is the right one.

3. How many animals

The living counts, from the databases themselves

These were counted from the source databases in August 2026, not quoted from a summary.

The marine exclusion is a rounding error, and we should say so rather than lean on it.  Cetaceans (102), sirenians (4), pinnipeds (35), sea turtles (7), sea snakes (~80) and the marine iguana come to about 227 species, or 0.6% of living tetrapods.  Only 81 amphibian species are obligately aquatic, which means amphibians are aboard — a class Woodmorappe omitted entirely.  Bats are 22% of mammal species.

Extinct land tetrapods add roughly 10,000 to 12,000 valid named genera from the Paleobiology Database.  That has to be printed as a range, because Benton showed that 51.8% of dinosaur species named through 2004 are invalid, with 24.3% nomina dubia.²⁹  A creationist cannot treat the fossil record as a straightforward inventory of pre-Flood life, and neither can a critic treat the genus list as a list of biological entities.

What is a "kind"? The evidence, and where it stops

The entire feasibility argument depends on kinds being coarser than species.  The best empirical evidence is hybridization, and it gives a genuinely split verdict that neither side reports.

Prager and Wilson measured how deep hybridizability runs by calibrating against molecular divergence.  Hybridizable bird pairs average about 22 million years of divergence and frogs about 21 million — but placental mammals only 2 to 3 million, a result Fitzpatrick's later work confirms.³⁰

Read that honestly and it means:  hybrid data licenses coarse kinds in birds and amphibians, and does not license them in mammals.  Answers in Genesis's 137 mammal kinds therefore do not rest on the hybrid evidence they call their best evidence; they rest on the "cognitum," which is looking at photographs.

It gets worse on inspection.  The AiG method paper licenses its central rule — one documented hybrid merges two whole genera — with a coyote × red fox example.³¹  Canis has 2n = 78 and Vulpes 2n = 34.  The example is given with no citation, nothing in the reference list supports it, and no accepted Canis × Vulpes hybrid exists.

The wider hybridization rate is also lower than apologetics suggests.  Schwenk, Brede and Streit put the animal hybridization rate at about 1% (range 0.1–3%), explicitly "at odds with" the 10% figure often quoted from Mallet.³⁰  eBird's data give 0.064% of individual birds.

The two creationist estimates disagree by twenty-seven fold

Two things in that table are ours to own.  Redoing Woodmorappe's own method with current databases roughly doubles his answer.  And the Ark Encounter's 1,398 kinds is not derivable from the published work:  the extant Answers Research Journal papers sum to 628, and the remaining 770 come from treating fossil families as kinds.  The arithmetic implies exactly 329 "clean" kinds that nobody has ever listed.

Two claims in our own notes should also be retracted.  The "Rattus developed 137 species since the Middle Ages" line has no primary source.  And Lake Victoria's cichlid radiation should never be cited at 15,000 years without Meier's 150,000-year figure alongside it.

The honest statement of the diversification problem is this.  Getting from 137 mammal kinds to 6,650 land species in about 4,400 years needs each line of descent to double its species about every 790 years.  That is about twice as fast as Lake Victoria's cichlids on the radiocarbon date for the lake's refilling, and about 20 times as fast on Meier's 150,000-year figure.  Counting species per year instead gives 1.5 a year worldwide and a gap that looks like 300 times, but that measure ignores branching:  a group with many lines of descent makes more species a year than a group with few, at the same pace in each line.  The rate per lineage is worked in The Creation Model of Rapid Speciation.  Every observed rapid case — Darwin's finches, Pod Mrčaru lizards, apple maggot flies, the Faeroe house mouse — is within a genus or a species.  The model needs family-level diversification, and nothing observed comes close.

4. Do they fit?

Mass is the wrong currency

Nearly every capacity argument in print, on both sides, computes an animal's volume and compares it to the ship's.  Critics are right that this is the wrong quantity.  Animals need floor area.  A 40 kg sheep occupies about 0.04 m³ of space as a solid object and needs roughly 0.34 m² of deck to stand on.

Floor-area requirements were fitted from 234 rows of published husbandry standards — the NRC Guide for the Care and Use of Laboratory Animals (8th ed.), USDA regulations at 9 CFR Part 3, EU Directive 2010/63/EU Annex III, EU Regulation 1/2005, and the Australian Standards for the Export of Livestock.³⁴

The striking result is that ASEL 3.3 alone refits to A = 0.03144 · W^0.6601 with R² = 0.994.  The Australian live-export regulation is, to three decimal places, the two-thirds surface-area law.  That is exactly the standard the Ark should be judged against:  a long sea voyage, animals confined for survival rather than display, real regulatory numbers, real mortality data attached.

Juveniles decide everything

The popular shortcut — a juvenile weighs half an adult — is unsourced, and it is wrong in both directions.  It is worth saying at once that it is not Woodmorappe's: his 1996 food arithmetic carries an explicit juvenile factor that varies with mass class, 0.45 for his small class, 0.34 for the medium and 0.05 for the large, which is the right shape and anticipates the result below.³⁵  The half-mass rule is what the popular literature and this site's own notes used.  Fitted against Myhrvold's amniote database:

The reptile relation has an independent check that is worth stating, because it is the kind of thing that ought to persuade a skeptic.  Extrapolate it to sauropod adult masses and it predicts hatchlings of 1.20 kg and 2.91 kg.  The actual fossil eggs give 1.22 kg and 3.66 kg.  Two entirely independent routes, one from living reptiles and one from Mesozoic eggshell, agree within 25%.

The consequence is that the half-mass rule is worse for the Ark than reality by a factor of six in floor area.  Dinosaurs are 53.5% of the adult floor-area load and 1.9% of the juvenile load.  Creationists have spent decades using a number that damages their own case.

For birds it runs the other way, and no creationist source appears to have noticed:  a fledgling bird is 83% of an adult, so the juvenile move buys almost nothing for the 43% of genera that are birds.

The result

Deck area available:  7,574 m² at the short cubit, 10,543 m² at the royal cubit.

Across the full grid, 64 of 81 scenarios fit at the short cubit.  All 17 failures use adult masses or the half-of-adult rule together with extinct taxa.  Every scenario using measured juvenile masses fits, at every kind level, every scope, and every cubit — the largest needing 5,057 m².  This survives a sensitivity test that re-imputes unknown extinct masses at the 25th and 75th percentiles.

"Just stack the cages"

The obvious objection to everything above is that a deck is not a single layer.  Deck-to-deck spacing at the short cubit is 4.45 m, leaving about 4.15 m of clear height under the beams.  A hamster does not need 4.15 m of headroom.  Put in shelves, run catwalks between them, and the effective floor area multiplies.  This is exactly how a laboratory animal facility works, and the article would be wrong to ignore it.

So it was computed.  Enclosure height was fitted against body mass from the same 234 rows of published husbandry standards used for the floor-area fit, giving h = 0.3773 · M^0.3732 (R² = 0.849, n = 87, spanning 0.01 to 4,000 kg).  The exponent recovers the theoretical cube-root scaling, which is a good sign the fit is real.  Tiers per deck are then floor(4.15 m ÷ (cage height + 5 cm shelf)), and the whole manifest is re-costed.

Stacking is legitimate, and it buys almost nothing.  Given unlimited free tiering, the adult-mass scenario improves by 5% and still overruns the ship by a factor of 2.6.

The reason is the concentration noted above.  An animal heavier than about 90 kg cannot be double-tiered in a 4.15 m deck, and animals above that line hold 94% of the required area.  Of the 16,182 genera on the manifest, 13,325 can be stacked eight high — and all of them together are about one per cent of the problem.  Under the adult scenario, 961 of 962 dinosaur genera are single-tier-only and hold 99.5% of that class's area; 1,524 mammal genera are single-tier-only and hold 89.5% of theirs.

This is worth stating plainly because it cuts in our favour.  The floor-area result is robust against the most natural objection to it.  Someone can grant the Ark free shelving, free catwalks, and no structural or ventilation penalty for any of it, and the adult-mass manifest still does not fit.  The case for the Ark rests on the size of the animals, not on the cleverness of the cabinetry.

Two caveats.  Tiering is not actually free — shelves are timber weight, catwalks are floor area removed, and stacked cages make the ventilation problem in section 6 worse, not better, by concentrating more metabolic heat into the same perimeter.  And the height fit is extrapolated well past its 4,000 kg ceiling for the largest taxa, though since those animals are single-tier under any plausible fit, the extrapolation does not affect the conclusion.

One older number that does hold up

Most of this article corrects figures that circulate on our own side, so it is worth recording one that survives.

The estimate of 2,301,307 kg for the total mass of animals aboard, computed in 2013 from a purpose-built database of 25,453 tetrapod species and their masses, reproduces.³⁶  Re-running the same method — genus-level kinds, adult masses, cetaceans excluded, two animals per genus with fourteen birds per genus for Genesis 7:3 — gives 2,333,330 kg, within 1.4%.  The dinosaur term, which is 79% of the total, matches to one kilogram.  The residual is a genus-filtering difference in the mammals and amphibians.

What that figure is not is a full manifest.  Its 5,368 genera come from a 2012 dataset that samples roughly a third of the extinct genera now recognised, against the 16,182 land genera used elsewhere in this article.  So the two numbers are correct about different cargoes, and the older one is smaller for a reason that has nothing to do with arithmetic.

One convergence is worth noting because the two routes share no data.  That 2013 manifest is 34,652 animals.  Redoing Woodmorappe's genus-level method against 2026 taxonomic databases, as section 3 does, gives about 34,700.

What this costs us

The popular capacity figures are wrong.  "125,000 sheep" and "522 boxcars" are volume arithmetic on a floor-limited problem.  The floor-limited answer at the short cubit is about 22,124 sheep — a factor of 5.7 smaller.  Any article repeating the old numbers is repeating an error, and this one will not.

The juvenile assumption buys deck space and pays for it in repopulation time.  Griebeler and colleagues put sauropod sexual maturity at 19 to 25 years.³⁷  A 3 kg hatchling boards, grows to perhaps 5–15 kg in a year, and disembarks still hatchling-sized into a devastated world needing two more decades to breed.  Only 5.7% of reptile species reach maturity within a year, against 54% of mammals.  The space problem is solved by juveniles; the survival problem is made considerably worse by them.  That trade is real and it is not usually disclosed.

Eighty-five per cent of extinct genera have no measured mass.  Their masses are imputed, and they carry 64% of the extinct adult mass.  The conclusion is stable under the sensitivity test, but the individual numbers are not precise.  The fossil record is also biased toward large animals, since big bones get found and named, so imputing the rest at the measured median probably overstates the load.  That bias runs in our favour, which is exactly why it should be declared.

The area allometry is extrapolated far past its data.  The transport fit spans 1 to 750 kg.  Applying it to a 50-tonne sauropod is a 67-fold extrapolation, and no standard exists because nobody has ever shipped one.

5. Feed, water, and metabolism

Correcting our own arithmetic first

The calculation in the site's own notes scales food linearly with body mass:  one human's annual food supply, 0.22 m³ of dry goods for a 62 kg person, multiplied by total animal mass over 62 kg.  Done properly the feed volume is seven to nine times larger.  The note's figure is not a conservative estimate; it is an underestimate.

The interesting part is why, because the obvious explanation is not the main one.

Note what this means for the standard objection.  "Small animals eat far more per kilogram, so Kleiber's law wrecks the calculation" has the sign right and the magnitude wrong.  It is true that the 3,331 genus-kinds under 1 kg are 1.2% of the live mass but 10% of the feed energy.  It is also true that switching from linear scaling to M^0.744 reduces the total by 31–46%, because elephants and rhinos dominate the mass sum and eat proportionally much less.  The real error in the popular calculation is factor C, which is a storage-engineering question, not a physiology question.

The numbers

Family-level kinds, adult masses, living animals only, 370 days, against a 43,006 m³ Ark:

At genus-level kinds the same calculation gives 2,124 m³ of feed (4.9%) and 1,453 m³ of water (3.4%).  For living animals, consumables are not what sinks the Ark.

Ectotherms are a genuinely strong point and the numbers are large.  Solved from Nagy's own equations, a 100 g mammal needs 12.0× and a 100 g bird 20.6× the daily energy of a 100 g reptile — reproducing exactly the "about 12 and 20 times" Nagy states in words.³⁸  Reptiles and amphibians are roughly a quarter of the kinds aboard and 1.3% of the energy budget.

Dinosaurs break the budget

Run the same calculation on the merged mass table, which adds extinct mammals, dinosaurs and pterosaurs with measured masses:

At genus-level kinds with adult dinosaurs, the food alone will not fit in the ship — before a single animal is housed, and even granting dinosaurs reptilian metabolism.  Dinosaurs are 89% of the live mass and 67% of the feed energy in that scenario.

So the feasibility case rests on two specific commitments:  kinds at roughly family level, and dinosaurs taken as juveniles.  Take both and consumables come to about 9% of the hull.  Take neither and there is no voyage.  Say it as a commitment rather than smuggling it in by choosing the parameters that work.

Note also that dinosaur thermal physiology is unresolved and moves the answer by a factor of 1.8, and that hay is the whole storage problem:  chopped hay is 52% of feed mass and 87% of feed volume.  Every conclusion is hostage to the assumed bale density, and a pre-industrial builder cannot bale.  Ruminants also cannot be fed grain alone without acidosis, so the eighteen-fold grain advantage cannot be fully exploited.

Hibernation is a much weaker lever than our side assumes

Modelled generously — 20% of mammal kinds hibernating, 10% of bird kinds in daily torpor, 70% of reptiles brumating, all for six months — torpor saves 7.9% of the total energy budget.  The reason is arithmetic, not physiology:  86% of the demand comes from mammals, and the mammals that dominate it are large tropical and temperate herbivores that do not hibernate at all.  None of it applies to dinosaurs.

Hibernation should be presented as a modest help, not a solution.  It is also seasonal and cued by photoperiod and temperature, not something a keeper can induce on demand.

The hard cases are about perishability, not volume

A pair of hummingbirds needs about 1.2 litres of honey for the entire voyage.  Volume is not the problem.  The problem is that they must be fed several times an hour, and that a koala needs 133 kg of fresh eucalyptus leaf per animal, which cannot be dried, stored, or substituted.  The same applies to obligate myrmecophages, nectarivores, vampire bats, and snail-specialist snakes.

Perishability and substitutability are the real difficulty here, and no creationist source has a quantified answer.

Two claims in our own notes do not survive the veterinary literature and should be cut.  Little Tyke, the supposedly vegetarian lion, ate milk and eggs — which supply the taurine and preformed vitamin A that felids cannot synthesise — so the diet was not plant-only, and the source is a single self-published memoir.  The "vegetarian dog that lived to 27" is Bramble, whose age is contested, who has no Guinness certification, and who was a dog, which is not an obligate carnivore.  Crocodilian frugivory is real, but the authors say fruit is eaten deliberately only "on occasion."  And the Galápagos tortoise fast is documented by Townsend at four to six months, not a year; Townsend's own word is "months."³⁹

6. Waste, air, and eight people

This is the section the apologetic literature waves at, and it is where the numbers bite hardest.

Manure is a volume problem

At the ASAE D384.1 mid-range rate of 65 kg of fresh manure per tonne of live animal per day,⁴⁰ a 370-day voyage carrying 800 t of animals produces about 19,000 m³ of manure — 45% of the entire hull.  Cutting to maintenance rations drops it to 27%.

An independent check kills the maintenance discount, and it is worth stating because it comes from shipping rather than from farming.  The Australian live-export feed standard of 2.0–2.5% of liveweight per day implies about 55 kg of wet manure per tonne per day for animals merely standing in a pen.  So the working range is 55–65, not the lower figure a sympathetic analyst would reach for.

The waste does not have to be stored, but it does have to be moved, and Genesis gives one door and one opening.  Dumping overboard does not close:  it is roughly 2,600 bucket-loads a day carried up two decks.  So the waste stays aboard, which forces deep litter, which in turn forces the low-mass stocking models from section 4.

Deep litter is the right answer and it is also the biggest single consumer of cargo volume.  Published bedding rates span a factor of fifteen.  The dairy compost-bedded-pack rate demands more bedding volume than the ship has.  Only the lean end — about 1.5–1.7 kg per tonne per day, the broiler deep-litter rate derived from the USDA handbook's own worked example⁴¹ — fits, and only if the bedding is chopped or shavings rather than loose straw.

Ventilation is the calculation that decides the reading of the text

The binding constraint is metabolic heat, not oxygen and not carbon dioxide.  Per 1,000 W of animal heat you need about 72 m³/h of air to hold CO₂ at the EU livestock limit of 3,000 ppm, but 418 m³/h to hold a 5 K temperature rise.  Ammonia sits between them and is live:  at 25% volatilisation of excreted nitrogen you need about 400 m³/h per tonne of live mass to hold 20 ppm.

Whole-fauna demand runs from 86,000 m³/h at 100 t of animals to 1,512,000 m³/h at 1,750 t.

Now the two readings of the tsohar:

Three independent methods agree within a factor of two on what a real livestock ship of this size needs, which is the best evidence that the arithmetic is sound.  Australian Marine Order 43 requires 20 air changes per hour in enclosed livestock spaces, which for the Ark is 860,000 m³/h.⁴²  Measured live-export practice is about 50 m³/h per m² of pen, giving 470,000.  The heat balance for 800 t of animals gives 692,000.  The perimeter slot delivers 223,000 — a quarter to a half of requirement.

Against Ohio State's natural-ventilation design rule, which works out to 3.33% of floor area as the cold-weather minimum for a single-storey barn,⁴³ the Ark's perimeter slot is 1.56% of its three-deck floor area — 0.47× the winter minimum and 0.31× the warm-weather figure.  Reaching the winter minimum needs a slot 2.1 cubits high; reaching a summer figure needs 3.2.

And this lands directly on Woodmorappe's own number, which is the strongest form of the test available.  Working independently in 1996 from the same reading of the text, he puts the window at "a total area of 144 square meters, which comes out to 4.6% of the Ark floor area."³⁵  Our perimeter-slot geometry gives 146 m².  Two calculations done thirty years apart, from different starting assumptions, agreeing to within about one per cent is a good sign both are reading the text the same way and doing the geometry right.

The disagreement is not about the opening.  It is about what the opening has to do.  His 4.6% is measured against one deck's area; the ventilation load is three decks of animals.  Measured against the three-deck floor area that actually generates the heat, the same 144 to 146 m² is 1.56%, which is under half the published cold-weather minimum — and the cold-weather minimum is the easy case.  The most detailed defense of Ark ventilation in print therefore arrives at an opening this article agrees with, and understates what it has to ventilate by a factor of three.  His companion figure at p. 41, that the opening admits about 420 m³ of fresh air per hour per square metre, is also roughly eight times the 50 m³/h per m² of pen that live-export carriers actually measure.

There is a further problem the apologetic literature never raises.  The slot has no stack term, because inlet and outlet sit at the same height and Genesis describes no ridge opening.  The ventilation is therefore wind-driven only, and it stops in a calm.  The lower two decks get nothing at all without inter-deck hatches that the text does not describe.

And here is a consequence nobody on our side seems to have paid.  Heat production per tonne scales as M^−0.25, so every "the animals were small or juvenile" move that eases the food and floor-area problems makes the ventilation problem harder in the same proportion.  The two arguments pull against each other.

Labour: the eight-person problem

Eight people working fourteen-hour days have 6,720 crew-minutes.  For 2,000 animals that is 202 seconds each per day.  For 16,000 animals it is 25 seconds each per day.

Only the broiler figure clears 16,000 animals, and it exists because a broiler grower never touches an individual bird.  So the labour budget closes if and only if the Ark is run as a broiler house rather than as a zoo:  bulk feeders, gravity water, litter left in place, no individual handling.

That is a coherent picture — Woodmorappe's own instinct that the right analogue is a factory farm rather than a zoo is correct — but it must be labelled.  Genesis describes no interior fittings at all:  no stalls, no gutters, no feeders, no waterers, no hatches, no ridge vent.  Every arrangement proposed here is our reconstruction, not the text's claim.  We also could not verify a pre-mechanisation hand-labour standard, so the break-even above is a lower bound on the difficulty, and we should not yet claim ancient attestation for gravity-fed automatic watering.

The mortality benchmark

The closest real-world analogue to the Ark is long-haul live-export shipping, and the data are public.

Extrapolated at constant hazard to 370 days:  3.6% on modern best practice, about 25% on 1990s practice, about 39% on 1891 practice.  On a ship carrying one breeding pair per kind, a 3.6% loss of animals is close to a 3.6% loss of kinds.

Two things about this cut in opposite directions and both should be printed.  The nineteenth-century transatlantic figures are the best pro-Ark datum in this whole article:  they show that naturally ventilated shipboard animal transport, without machinery, is not intrinsically lethal, and that mortality fell fourfold within a decade as husbandry improved.  Against that, the Cormo Express is the closest thing to a long-voyage test that exists, and a purpose-built modern carrier with a veterinarian aboard still lost nearly a tenth of its animals in 80 days.

The most useful detail is the cause of death.  Inanition and salmonellosis account for 60–76% of mortality — animals that will not eat, and disease — not heat, ammonia, or trampling.  That is encouraging for the engineering and discouraging for the assumption that wild animals confined for a year will simply eat what is put in front of them.

7. What was left behind

The capacity argument depends on excluding most of life from the manifest, so the exclusions have to be defended rather than assumed.

Plants

Nothing in Genesis puts plants aboard.  The best evidence here is Darwin's own, and it has to be quoted with his retraction attached.  In his seawater germination experiments, 64 of 87 seed kinds germinated after 28 days in artificial seawater, with Capsicum still viable at 137 days.⁴⁶  But on 24 November 1855 Darwin wrote:  "permit me to cry peccavi… all sink within a month, and most of them within 14 days… my experiments are of little or no use."⁴⁷  He rescued the result using dried material and settled on a conservative estimate of about ten plants per hundred in a flora crossing 900 miles.  That conservative figure is the one to use.

The field confirmation is Ernst's 1908 Cambridge monograph on Krakatau's new flora, and it is better than expected:  of the flowering-plant flora, "39—72% have been introduced by sea-currents, 10—19% by birds, and 16—30% by air-currents."⁴⁸  Sea drift was the single largest source, measured by a mainstream botanist.

Mount St. Helens cuts the other way, and creationist retellings hide it.  Del Moral and Wood measured 1.1% plant cover on the Pumice Plain after eleven growing seasons, and under 0.1% on the Plains of Abraham.  Recovery happened where biological legacies survived — surviving roots, buried seed, protected pockets — and a global Flood leaves none.  The Spirit Lake log mat is genuinely useful (still 20% of the lake surface after 46 years), and the erect sunken stumps really are reported in Geology, but no one has ever published a count, so "thousands" must be written as uncounted.

Rafting

The best rafting evidence is mainstream and it contains the worst news in this section.  Carlton and colleagues documented 289 living species on 634 objects crossing 7,000 km of Pacific over six years after the 2011 Japanese tsunami.⁴⁹  Rafting works.  But the wooden debris lasted only two to three years because shipworms ate it, and no Western Pacific wood has ever arrived in North America with a living community in 150 years of observation.

Insects

Roughly a million described insect species, and the textual argument for excluding them is weaker than our notes assume.  Genesis 7:21 already lists "all the swarming things that swarm on the earth" among the dead, so if insects died they needed saving.  The Septuagint of 7:22 has no "in his nostrils" at all.  And the spiracle-versus-nostril distinction is nineteenth-century entomology read back into an idiom that BDB simply glosses as "every breathing thing."

The argument is not worthless — the absence of šereṣ from every boarding list is a real datum — but it is a probabilistic argument from a list, not a demonstration, and should be presented at that strength.

In any case volume was never the objection.  Even the maximalist version, all seven million estimated terrestrial arthropod species with two individuals each in their own box, is 1,750 m³, or 4% of the hull.  Host specificity is the real problem, not space.

Fish

A global flood mixes fresh and salt water, and stenohaline fish die in the wrong salinity.  The best available response is that fresh and salt water do not mix instantly:  haloclines are stable and persistent, and the Black Sea, the Baltic, fjords and river plumes all maintain sharp stratification.  The honest counter is that a year of storm, sediment loading and turbulence is not a calm estuary.

We could not obtain the family-level euryhaline tally or measured salinity tolerance data within this session, so this section is the weakest-sourced in the article and is flagged as such rather than filled with plausible assertion.

8. After the door opens

Does anything eat?

Forage area turns out not to be the constraint.  At Mount-St.-Helens-grade productivity, 2,000 large herbivores need a circle of only about 24 km radius, and animals walk.

The real gap is large carnivores in year one — the interval between carrion running out, which takes weeks, and prey populations rebuilding, which takes about a year.  Three answers are available (stored provisions, fish and insects, extended fasting), each is real, and none is quantified.  This is an unclosed gap and it should be named as one rather than papered over by stacking the three.

Our own notes' mouse calculation has to go.  The claim that two mice reach 230 billion in a year comes from 6^(365/25) = 2.3 × 10¹¹.  Run a daily age-structured model with real Mus musculus life history — gestation about 19–21 days, first breeding at about six weeks, not 25 days — and zero mortality, and the answer after one year is 81,842.  That is wrong by a factor of 2.8 million.  The corrected intrinsic rate of increase, about 0.029 per day, matches measured Australian mouse plagues.  81,842 is still a lot of mice, and the corrected number is the one to print.

Two problems this article does not try to settle

Two of the hardest objections to the Ark are not logistics questions at all, and each is large enough to need its own treatment.  They are named here rather than omitted, because an article that quietly skipped them would deserve to be distrusted.

Post-Flood biogeography.  Everything disembarks at one point in eastern Anatolia, yet marsupials are overwhelmingly Australian and South American and lemurs are Malagasy.  The difficulty is not the modern distribution, which rafting and land bridges could in principle produce.  It is that there is no dispersal trail:  no marsupial remains between Ararat and Australia, no lemur remains between Ararat and Madagascar, along the most heavily excavated ground on earth.  The Wallace Line is a factor-of-two miss for any land-bridge answer, since the glacial lowstand reached about −134 m against a Lombok sill near 250 m.  Chad Arment, writing in Answers Research Journal, calls the required dispersal "not only unlikely, it is highly improbable" and dismisses the usual escape as "untestable and… simply God-of-the-gaps theorizing."⁹  A separate article is planned on this, and nothing in the present one should be read as answering it.

The founding population.  A diploid pair carries at most four alleles per locus, which is arithmetic rather than a model, and MHC allelic depth is where it bites.  The Isle Royale wolves are the closest run experiment — two to three founders in the 1940s, extinct as a breeding population by 2018, with 33% vertebral malformations.⁵⁰  The northern elephant seal is the better case and still shows an eightfold permanent loss of heterozygosity.  Woodmorappe's reply, that "a single pair of founders most definitely can have the same genetic diversity as fifty founders," is false.  Minimum viable population and the genetic bottleneck are also getting their own article, where the human case and the coalescent evidence can be given the space they need.

What belongs here is only the logistical consequence, and it is the one from section 4:  the juvenile assumption that makes the animals fit is the same assumption that makes repopulation hardest, because a sauropod that boards as a 3 kg hatchling disembarks needing two more decades to breed.

9. Arguments we should stop making

Collected in one place, because this site's credibility depends on retiring them before a critic finds them.

That last row deserves emphasis.  Every TalkOrigins claim from CH500 to CH505.8 attacks material our own movement has already disowned.  Conceding it up front costs nothing and buys real credibility.

10. Verdict

The logistics work, under conditions that are considerably more specific than the popular defenses admit.

The Ark holds the animals if kinds are around family level and the animals are juveniles at the measured juvenile mass rather than the invented half-of-adult rule.  Feed and water then come to roughly 9% of the hull.  Waste is manageable only as deep litter, not by removal.  Labour closes only at broiler-house management rates.  Ventilation closes only on the continuous-perimeter-opening reading of the tsohar, and even then only at the low end of the plausible animal mass, with no stack effect and with inter-deck hatches the text does not mention.  The hull is plausible in bending with an order of magnitude to spare, and marginal in seam shear against a joint nobody has excavated.

Change any one of those and it stops working.  That is a much narrower result than "there was plenty of room," and it is a much stronger one, because every step of it can be checked.

One logistical question remains genuinely unanswered:  the survival of large carnivores in the first year on a sterilised planet, in the gap between carrion running out and prey populations rebuilding.  Three answers are on offer, each is real, and none is quantified.

Two larger objections are deferred to their own articles, not because they are weak but because they are too big to dispose of in a section:  post-Flood biogeography, where our own movement's journal has already conceded the point, and the founding-population problem.  A reader who wants to know whether the Flood account survives should treat this article as one of three, not as the whole case.

And one thing should be said about the state of the argument.  The strongest pro-Ark results in this article — that the measured juvenile allometry beats the creationist shortcut sixfold, that reptile hatchling scaling independently predicts sauropod egg masses, that the two-thirds surface law underlies the live-export standard, that nineteenth-century naturally ventilated animal transport achieved 0.31% mortality — are results nobody in apologetics appears to have run.  Meanwhile the arguments the movement has spent decades defending are, disproportionately, the ones that do not survive a check.  There is a lesson in that about where the effort should go.

Appendix: Browse the species-mass dataset

The table below is the article's full working dataset:  34,743 taxa with the body-mass estimates, GBIF taxonomy matches, and aquatic/terrestrial flags used in sections 3-5.  It is served live from species_masses.csv.  Click a column header to sort, open a column's menu to filter it or pick from its most common values, use the search box to search every column at once, or open the SQL button to run your own queries.

Sources