Erosion Rates and the Age of the Earth

"The continents are being denuded at a rate that could level them in a mere 14 million years!  Yet, evolutionary-uniformitarian geologists feel certain that the continents have existed for at least 1 billion years." — Stuart Nevins (Steven A. Austin), Institute for Creation Research, 1973¹

"A continent like North America would be worn down to low-lying terrain in geologically short times, of the order of ten million years!" — John Southard, MIT, in an undergraduate geology textbook²

"However, I am not sure that erosion rates are independent of the topography.  Could it be, instead, that erosion rates are high at times of high relief (as today) but become much smaller as the continents approach peneplanation?  This should be considered." — John Woodmorappe, reviewing the newest young-earth book on erosion for the Journal of Creation, 2024³

Summary

Rivers carry roughly fifteen to twenty billion tonnes of rock into the ocean every year.  Spread that loss evenly over the land and the continents lose about forty metres of height every million years.  The land averages just under eight hundred metres above sea level.  Divide one by the other and the continents should be flat in about twenty million years.

That calculation is correct.  It is not a creationist invention, and no mainstream geologist disputes it — it appears in undergraduate textbooks as a standard puzzle.  Young-earth writers have used it since 1973 to argue that the continents cannot be billions of years old.

The conclusion does not follow.  The calculation measures how fast the land would vanish if nothing else were happening, and something else is very obviously happening.  Three corrections apply, and they are of very different sizes:

1. Continents float. Strip rock off the top and the crust rises to replace about five-sixths of what was removed. To lower the surface one metre you must erode roughly six. This is a factor of six. It turns twenty million years into about a hundred and thirty.
2. Flat land erodes slowly. Erosion tracks relief, so the decline is exponential rather than linear. This is real and measured. It does not rescue the old-earth position either — run the exponential out and the continents are still gone long before four and a half billion years.
3. Tectonics rebuilds the continents. This is the correction that does the work, and it is the one the young-earth version leaves out.

It is worth being exact about what the argument's best advocate actually concludes, because it is more modest than the version usually attacked.  Fleming does not say erosion dates the earth.  He says it constrains it, and he goes on using radiometric dating himself: "erosion of course doesn't give us an exact date, but it really does help us constrain the age of the Earth…  I use radiometric dating, I understand it, I've used it to do stratigraphic correlations…  but even when we're generous, erosion rates show that the rock record is hundreds or even thousands of times younger than the radiometric dates would seem to indicate."⁴

That is the claim this article disagrees with.  The honest form of what the erosion data support is not "the earth is young" and not "the rock record is a thousand times younger than dated."  It is "the earth cannot be tectonically dead" — which everybody already agrees with.

The evidence that continents are actively rebuilt is independent and measurable, and the single strongest piece of it is this: in every mountain belt on earth you can walk up to rocks that crystallised twenty to thirty kilometres underground.  They are at the surface because that much rock was removed from over them.  Erosion on the scale the young-earth argument demands is not a problem for mainstream geology.  It is a mainstream prediction, and it has been confirmed by measuring the pressure at which the exposed minerals formed.

The thing that is actually surprising

The relief-dependence objection — the one that sinks the argument in its usual form — was not first raised by a skeptic.  It was raised by a creationist, at a creationist conference, in 1986, in the printed discussion of a creationist paper.  It was raised again, in the flagship creationist journal, in December 2024.  Both times it was stated clearly.  The first was answered once, in print, and then dropped; the second has not been taken up at all.

That, more than any measurement, is the story of this argument, and this article is largely about it.

What this article does

It works the arithmetic in the open using measured numbers, and shows which steps survive.  Where a figure is computed rather than quoted, the computation is shown so a hostile reader can redo it.  The mean elevation of the land, the area of the continents, the volume of rock above sea level, and the mean depth of the ocean used below were all computed for this article from the ETOPO 2022 global relief grid at sixty arc-seconds, using exact ellipsoidal cell areas.⁵

It takes the best available version of the young-earth case rather than the weakest.  That version is Monte Fleming's — a working geologist at Loma Linda University who published a book on it and then laid it out across two episodes, about a hundred minutes in all, on Todd Wood and Paul Garner's podcast Let's Talk Creation in June and July 2024.  Both episodes were transcribed for this article and are quoted throughout, with timestamps.

It grades both sides by the same standard, which means it also documents where the rebuttals are wrong.  The most widely cited skeptical answer to this argument contains no numbers at all.  Two popular university course pages that do give a number use one that is roughly ten times too small for the job.  A common talking point — that modern erosion rates are inflated by farming — turns out to be far weaker than its users think, and Fleming is closer to right about it than his critics.

It follows citations back to their origins, which turns out to matter more than any new measurement.  The man who supplied the argument's headline number published a correction to it four years later that cuts the natural rate by a factor of nearly three.  The USGS paper both sides quote for the disparity between erosion and uplift says, in the same abstract, that there is ample time for continents to be planed flat — and works out the isostatic correction itself.  And the ocean-sediment version of the argument turns on a single wrong denominator, which one number published in 2020 settles.

Finally, it says plainly where the young-earth case is strongest.  The weakest part is the popular coastal-erosion version, though a narrower form of it survives and is stated here rather than buried.  The strongest part is not about rates at all.  It is about the flat, undisturbed contacts between rock layers supposedly separated by millions of years — and about a genuine double edge, since the same erosion rates that embarrass an old earth embarrass a one-year flood considerably more.

Every source it rests on is catalogued, with what each one measured and which way it cuts, in a companion page: the literature.

Where the argument comes from

The erosion argument has a clear paper trail, and following it matters, because almost every number in the modern version was measured by mainstream geologists in the 1950s and 1960s and has been passed hand to hand ever since.

1955-1964: the mainstream sets the numbers.  James Gilluly estimated that 13.6 km³ of solid material reaches the world ocean each year.⁶  Stanley Schumm, in a US Geological Survey paper titled The disparity between present rates of denudation and orogeny, put modern mountain uplift at 7.6 metres per thousand years and noted it was about eight times the maximum rate at which erosion strips a landscape.⁷  Sheldon Judson and Dale Ritter measured sediment and dissolved load in United States rivers and got an average lowering of 6.1 cm per thousand years, or 61 metres per million years.⁸  None of these men were creationists.  All three papers are still cited, and Schumm's turns out to matter enormously — see below.

1973: Stuart Nevins turns it into a young-earth argument.  Writing in the Institute for Creation Research's Impact series under a pen name — he is Steven A. Austin, later of Grand Canyon and Mount St. Helens fame — Nevins added up every route by which material reaches the sea: 20.2 billion tons of suspended river sediment, 4.6 billion dissolved, 2.2 billion from icebergs, and smaller contributions from groundwater, wave erosion and wind, less a small loss to salt spray, for 27.5 billion tons a year.  (His tons are short tons, which is 24.9 billion tonnes; the article's tables convert them.)  He divided the mass of the continents above sea level by that figure and got fourteen million years.  Then he asked the question that has been asked ever since: "Yet—miracle of miracles—the continents are still here and do not appear to have been eroded even one time!"¹

Nevins's numbers have held up remarkably well, and that should be said loudly.  His estimate of the volume of sediment on the ocean floor, 77 million cubic miles, converts to 3.21 × 10⁸ km³.  The 2019 GlobSed measurement, built from modern seismic data, gives 3.37 × 10⁸ km³ — within about five per cent.  His total flux, 24.9 Gt/yr once converted to tonnes, is in the same range as modern estimates.  Redoing every calculation in the paper from his own inputs reproduces every one of his answers.  There is no arithmetic cheat in the founding document, and no bad data.

The problem is structural, and Glenn Morton — who once made the argument himself — put it best.  Dividing a stock by a flux gives you the time until the stock is exhausted, not the time since it began.  A river two hours from the sea did not start flowing two hours ago.  To turn "how long until the continents are gone" into "how long have the continents existed," you need a term for what puts rock back, and Nevins's calculation has no such term: no uplift, no isostatic rebound, no recycling.

1986-1988: Ariel Roth builds the careful version.  Roth, of the Seventh-day Adventist Geoscience Research Institute, is the most rigorous writer in this tradition and the one everything later depends on.  His Origins paper works through six independent rate arguments, quotes the mainstream literature accurately, states the standard counter-explanations, and answers them one at a time.⁶  A companion paper two years later supplies the second half of the case, about stratigraphic gaps.⁹  Roth's figures — 6.1 cm per thousand years, 623 metres of mean continental elevation, 10.2 million years to level — are the ones nearly every later creationist writer reuses.

1990s-2000s: it becomes a standard talking point.  Tas Walker's Creation magazine article "Eroding ages" reproduces Roth's figures and examples almost line for line.¹⁰  It appears in ICR's "Four Geological Evidences for a Young Earth" and "Five Global Evidences," in Henry Morris's Scientific Creationism, and in Paul Ackerman's It's a Young World After All.

2021-2024: Monte Fleming.  Fleming is Assistant Professor in the Department of Earth and Biological Sciences at Loma Linda University, with a PhD in earth science from the same institution.  He is not a credential-free populariser, and it is worth saying so plainly before disagreeing with him.  He has published original field and laboratory research in mainstream, non-creationist journals — two papers in Elsevier's Geomorphology on pothole formation and sediment removal by wind on Rock Window Mesa in Arizona, and one in the SEPM Journal of Sedimentary Research describing a precision sediment feeder he built for flume work.¹¹  His master's thesis was on marine vertebrate burial in the Miocene Pisco Formation in Peru.  He runs a sediment flume at Loma Linda designed to vary depositional parameters independently.  His book Stories About Earth's History: A Geologist's Dissent from Deep Time — self-published in December 2021, fifteen chapters and about 121 pages — gives nine arguments against deep time, of which erosion is the first and longest.¹²  In June and July 2024 he presented it across two episodes of Let's Talk Creation, the podcast run by Todd Wood of Core Academy of Science and Paul Garner of Biblical Creation Trust.^{13 4}  Those episodes are the most detailed public statement of the argument in existence, and this article treats them as the primary text.

2024-2026: the argument goes live again.  John Woodmorappe — a pen name — reviewed Fleming's book for CMI's Journal of Creation under the title "A strange admixture of erosion geology and tendentious theology": warm on the geology, hostile to a theological digression in the book's closing chapter.³  In April 2026 Jake Ramgren wrote a popular version for the New Creation blog, opening with the lighthouses that have had to be moved back from the Cape Cod cliffs.¹⁴  In June 2026 the biologist Joel Duff answered him in detail.¹⁵  That exchange is the current state of the debate.

The arithmetic, done carefully

Start by granting the argument everything it asks for.

The mass.  Land covers 148.5 million square kilometres and averages 794.5 metres above sea level.  At a rock density of 2.7 tonnes per cubic metre that is 3.19 × 10¹⁷ tonnes of rock standing above the waterline.  Both figures were computed for this article from ETOPO 2022.⁵  Fleming's own working figure, "about 800 metres tall on average," is right.¹³

The flux.  Estimates of what rivers deliver to the ocean cluster between about 12.6 and 24 billion tonnes a year depending on what is being counted.  The most useful single number does not come from river gauges at all.  Hella Wittmann and colleagues measured cosmogenic beryllium-10 in sand from more than fifty of the world's largest rivers, covering 32% of the land surface.  For the 65% of those rivers whose sediment is in cosmogenic steady state, the measured flux is 3.07 ± 0.56 Gt/yr, at a source-area denudation rate of 54 m/Myr.  Scaling that rate up to all the world's exorheic basins — which is an assumption, and the authors say so — gives 15.2 ± 2.8 Gt/yr, averaged over the last eleven thousand years.¹⁶  So it is a pre-agricultural number, and it is anchored in measurement, but the global figure is an extrapolation like the others.  It should not be presented as though it were not.

The division.

The range from present-day measurements is twelve to twenty-five million years.  Fleming's headline figure of twenty-two million is correctly calculated from his inputs, and Nevins's fourteen reproduces exactly.  Nothing is wrong with this arithmetic.  A reader who came here expecting to catch a creationist in a calculation error will not find one.

Two rows in that table deserve attention, because they come from inside the argument's own source list.

The second is the more awkward.  Sheldon Judson supplied the number the whole creationist tradition rests on — 6.1 cm per thousand years for the United States, published with Dale Ritter in 1964.  Four years later, in American Scientist, Judson published a global figure that corrected for human activity, and the correction is large.  As reported in Nature by B. Gregor: "man's intervention has increased the overall denudation rate from 9 to 24 billion tons a year."¹⁷  Judson's own estimate of the natural, pre-human rate is 9 Gt/yr — roughly a third of the modern figure, and about a third of what a creationist article citing him usually uses.

Ariel Roth, to his credit, reports this: he notes that "Judson (1968), while correcting for human activity, suggests 34 Ma for complete erosion of the continents," and the mass balance reproduces it — 3.19 × 10¹⁷ tonnes divided by 9 Gt/yr is 35 million years.⁶  Others have been less careful.  Tas Walker's Creation article states that "the average height reduction for all the continents of the world is about 60 mm per 1,000 years, which equates to some 24 billion tonnes of sediment per year," and lists the uncertainties that would push that number up — unmeasured bed load, rare catastrophic events — without mentioning the one that pushes it down by a factor of nearly three, which is the human amplification the author of the number identified himself.

There is an older and cleaner measurement available too.  Dole and Stabler's 1909 survey for the USGS, made before mechanised agriculture, put the United States at 783 million short tons a year of dissolved and suspended load.  Their own stated result is "one inch in 760 years," which is 33.4 metres per million years — about half the Judson and Ritter figure.  Recomputing from their tonnage using their own rock density of 165 lb/ft³ gives 33.6, so their arithmetic checks; at the 2.7 t/m³ used elsewhere in this article it would be 32.9.¹⁸

The top row is the odd one out and it matters.  It is not an extrapolation of modern rates at all — it is the denudation rate implied by the volume of Phanerozoic sedimentary rock actually preserved, corrected for what was later eroded or subducted.  Read from the rocks rather than from a river gauge, the average over the last half-billion years comes out at about a third of the present rate.¹⁹

It is also not a creationist number.  Roth was scrupulous about crediting its origin: the ten-million-year figure, he notes, "has been a well-accepted figure" appearing in standard textbooks including Dott and Batten's Evolution of the Earth and Garrels and Mackenzie's Evolution of Sedimentary Rocks.⁶  It is still in textbooks.  Joel Duff, arguing against the young-earth use of it, concedes the point without hesitation: "The mainstream geological literature openly states that if denudation were the only thing happening, the continents would have been planed flat many times over the span of geological history.  This is not a creationist 'gotcha' that someone forgot to address.  It is a standard chapter heading."¹⁵

The question is what the number means.

Correction one: continents float

Take a kilometre of rock off a continent and the continent does not stay put.  Crust floats on denser mantle, and lightening it makes it rise.

The arithmetic is simple.  If crust has density ρ_c and mantle ρ_m, removing a thickness h of rock lets the column rise by h·(ρ_c/ρ_m).  The net drop of the surface is only h·(1 − ρ_c/ρ_m).  With ρ_c = 2.75 and ρ_m = 3.30 g/cm³ that is 0.167 — about seventeen centimetres of lowering per metre of rock removed.

So levelling a continent whose surface averages 795 metres does not mean removing 795 metres of rock.  It means removing about 4.8 kilometres, and multiplying the time by six.

Across plausible densities — ρ_c from 2.67 to 2.85 and ρ_m from 3.2 to 3.4 — the multiplier runs from 4.7 to 9.1.  Six is the value at the central densities, not the midpoint of that range, which is nearer seven.

This is not a modern rescue invented to answer creationists.  Schumm's 1963 USGS paper, the one the creationist literature has been citing for sixty years, says explicitly that isostatic compensation multiplies the time to plane a continent "by a factor of five or much more."⁷

Fleming knows about isostasy and describes it correctly with an ice-cube analogy: chip the top off a floating ice cube and the rest bobs up.⁴  His objection is that you cannot keep doing this because you run out of crust.  Removing 4.8 km is about 12% of the area-weighted continental crustal thickness of 39 km, so the trick works about eight times and stops.  (The run-out-of-crust objection is put to him by one of the hosts rather than raised by Fleming; the arithmetic here is this article's own.)

He is right that isostasy alone does not answer the argument.  A factor of six turns twenty million years into a hundred and thirty.  The earth is held to be 4.5 billion years old and the continents at least 2.5 billion.  Isostasy closes about three per cent of the gap.  Anyone who has answered this argument by saying "isostasy" and stopping has not answered it.

Correction two: flat land erodes slowly

This is Woodmorappe's objection, and it is a real effect with a large literature behind it.  Erosion rate scales with relief.  Franz Ahnert's study of twenty mid-latitude drainage basins found rates as low as 1.6 cm per thousand years in Texas and England.  Against that, Menard reported about 100 cm per thousand years for the Himalaya, and Ollier and Brown about 1,900 cm per thousand years on a volcano in New Guinea — a spread of more than a thousandfold, all of it collected by Roth himself.⁶

If erosion rate is proportional to elevation the decline is exponential rather than linear: h(t) = h₀·e^−t/τ.  Calibrating to the same numbers as above, and including the isostatic factor of six, gives a time constant of 128 million years.  Then:

So the second correction does not rescue the old-earth position either.  Exponential decay is gentler than linear at first and then merciless; a billion years flattens everything.  This is worth stating plainly, because it is exactly where a careless defence of mainstream geology goes wrong.  "Isostasy and diminishing returns" is not a sufficient answer, and a young-earth reader who has been fobbed off with it has been short-changed.

Correction three: something rebuilds the continents

Once the first two corrections are spent, only one option remains.  If the continents are billions of years old and still stand most of a kilometre above the sea, then something must be actively putting rock back up.  There is no way around this.  It is also not a rescuing device: it is the central claim of plate tectonics, arrived at for entirely unrelated reasons decades before anyone applied it here.

The young-earth reply, made by Roth in 1986 and repeated by Fleming in 2024, is that this cannot be true because the evidence would be visible.  Roth's version:

"It has been suggested that mountains still exist because they are constantly being renewed by uplift from below.  However, this process of uplift could not go through even one complete cycle of erosion and uplift without eradicating the layers of the geologic column found in them.  Present erosion rates would tend to rapidly eradicate evidence of older sediments; yet these sediments are still very well represented, both in mountains and elsewhere."⁶

Woodmorappe puts the same objection more sharply in 2024: "tens or perhaps hundreds of kilometre thickness of rock would have to be removed everywhere from every continent.  There would be no Phanerozoic sedimentary rock left anywhere on Earth.  Yet there are."³

This is the strongest form of the objection and it deserves a direct answer.  The answer is that the prediction is correct and has been confirmed — but not everywhere at once, and the two halves of the objection describe two different places.

The refutation is already inside the creationist literature

Before the geology, a matter of record.

In August 1986, at the First International Conference on Creationism in Pittsburgh, Glenn Morton — then a young-earth creationist — presented a paper called "Geologic Challenges to a Young Earth."  He endorsed the erosion argument, calling it "the best geological argument" available.  The proceedings printed four signed responses.  One of them, by John W. DeVilbiss, took the argument apart:

"It is true that if we take an empirical rate of erosion proportional to altitude, a differential equation leads to absurd initial heights for a 225 million year mountain chain.  But the whole reasoning is not realistic.  A mountain chain does not start out at maximum height.  It 'grows' in the presence of erosion, and as it 'grows' higher, the erosion becomes exponentially greater… Thus, in the presence of erosion proportional to altitude, a mountain would never get unreasonably high even if very, very old."²⁰

And then, in one sentence, the whole point:

"The balance is between growth, altitude and erosion rate, not merely between maximum altitude and erosion rate, as Mr. Morton and others have suggested."²⁰

That is the coupled-system answer that mainstream geomorphology has given since Grove Karl Gilbert.  It was published in a creationist venue, by a creationist, forty years ago, aimed squarely at the version of the argument creationist literature still runs today.  No skeptic can dismiss it as an evolutionist talking point.

Thirty-eight years later, reviewing Fleming's book for the Journal of Creation, Woodmorappe raised the same objection again in the passage quoted at the top of this article, and then dropped it — "This should be considered."  As of this writing nobody has taken it up in print, though the review appeared only in December 2024 and online in April 2026, which is not long for a quarterly.

Three things follow, and they do not all point the same way.

In fairness to Morton, his 1986 reply was not stupid.  He answered DeVilbiss by pointing at the Appalachians:

"Most authorities would also agree that the Appalachians have not experienced significant uplift since the terminal Paleozoic orogeny… Since that time there has been no significant orogenic activity in the Appalachians and the subsequent history of the mountain chain has been one of erosion.  Thus Dr. DeVilbiss' argument falls since the Appalachians have been experiencing little but erosion for 200 million years (in his view), and applying an increase in the erosion rate as the mountains were taller in the past is a perfectly valid approach to the problem and leads to absurd results.  Thus the Appalachians could not be 200 million years old."²¹

DeVilbiss's answer works for a growing orogen — the Himalaya, the Andes, Taiwan.  It does not obviously work for a mountain range that mainstream geology says stopped growing.  That is the live crux, and it is dealt with below.

And credit where it is due, because this article's whole standard depends on noticing it.  In the second podcast episode, unprompted and against his own argument, Fleming retracts one of his own book's numbers:

"I did make a mistake in the book, because I assumed equivalent density.  And when I actually went and dug into that later, the sediment that collects behind dams is not compacted like the sedimentary record on the continents is, and so the actual mass is about one in five million."⁴

A self-published author volunteering an arithmetic error that weakens his own case, on the record, is exactly the behaviour this article is about to say the movement needs more of.  It should be said before the criticism, not after.

But the larger record still stands.  The relief-dependence problem has been identified twice, in print, inside the movement, by people arguing in good faith, and neither time did the argument change.  It has never been placed on Creation Ministries International's list of arguments creationists should not use, nor on Answers in Genesis's equivalent — both lists were checked in full for this article, and neither contains any entry on erosion, denudation, ocean sediment, sea salt, deltas, or Niagara Falls.

That matters, because there is a precedent for how these organisations retire a bad clock.  In 1993 Andrew Snelling and David Rush spent forty pages in a creationist journal killing the moon-dust argument, on the ground that its input figure was wrong by three orders of magnitude:

"Should creationists then continue to use the moon dust as apparent evidence for a young moon, earth and solar system?  Clearly, the answer is no.  The weight of the evidence as it currently exists shows no inconsistency within the evolutionists' case, so the burden of proof is squarely on creationists…"²²

"We, of course, take responsibility for the conclusions, which unfortunately are not as encouraging or complimentary for us young earth creationists as we would have liked."²²

The erosion argument has not received that treatment.  It should.  The test Snelling and Rush applied is exactly the right one: is the input rate measured, is it representative, and has the removal or compensation term been accounted for?  For erosion the answers are yes, roughly, and no.

The coastal version, which is the weakest

Fleming's opening move, and the one that has spread furthest, is not about denudation at all.  It is about coastlines.

The observation is real and vivid.  Lighthouses on the American east coast have had to be picked up and moved inland.  Houses fall off cliffs in Florida and Alaska and on the shore of Lake Michigan.  Coastal erosion costs the United States around half a billion dollars a year in property.  Fleming's numbers, given on the podcast: the east coast retreats at "about a meter a year," the lighthouse-relocation record works out to about two feet a year, and the west coast runs at "a few centimeters per year."¹³

The book's figures are cleaner, and two reviewers quote them: the Atlantic coast retreating at 78 cm a year and the Pacific at 20 cm a year.  Neither review states the time unit, but the arithmetic settles it.  The two coasts close on each other at 0.98 metres a year, and dividing a continental width of about 4,100 km by that gives 4.2 million years, which is his answer, with the Pacific front travelling about 850 km inland — near Salt Lake City, where he says it lands.  So the figures are annual rates of shoreline retreat and his arithmetic is self-consistent.  It is worth noting that the coast-to-coast width he would need is at the narrow end: the 2,800 miles usually quoted, and used in the popular version of his own argument, is 4,500 km and would give 4.6 million years.

Then the extrapolation:

"So if we take these values — a meter a year, several centimeters per year — and then kind of extrapolate what should happen over a few million years, the United States is, you know, a few thousand miles wide, and it would only take about a little over 4 million years if sea level didn't change, and if the ground didn't move and any of that — it'd only take about 4 million years to just kind of level everything.  Now, again, that is simplistic, because there are these other factors in play.  But yeah, but even even so, I mean, that is remarkably rapid."¹³

The emphasis is added, and it matters.  Fleming attaches the two conditions and flags the simplification in the same breath.  Any version of this article that quoted him without them would be doing precisely what it criticises others for.

And the closing move, which is the one that does the real work: continental shelves are only a few tens of kilometres wide, so very little coastal erosion can ever have happened.

This version has five problems, and they compound.

It measures the wrong direction.  Continental denudation is a vertical quantity — how much elevation the whole surface loses.  Coastal retreat is horizontal — how fast a line moves sideways.  They are not convertible, and a continent is not "levelled" by its edges moving inward; it is levelled by its surface coming down.  Duff states it exactly: "The relevant quantity is not how many miles of coastline retreat per year.  It is how many millimeters of elevation the whole surface loses per year.  Geologists have a word for this: denudation."¹⁵

It uses the wrong substrate, and the wrong number for it.  A metre a year is not a rate at which rock erodes.  Rates like that come from unconsolidated material — sandy barrier islands, and the glacial sand and till that make up Cape Cod, which the Laurentide ice sheet dumped there about twenty thousand years ago.  Fleming himself gives the reason the west coast is slower — "the rock type that the waves encounter matters a lot"¹³ — and then runs the extrapolation on the fast number anyway.

The measured numbers for actual rock are much lower, and they are available.  The global database of rocky-cliff recession rates, GlobR2C2, holds 1,680 records from 1,530 sites, and its medians are 2.9 cm/yr for hard rock, 10 cm/yr for medium, and 23 cm/yr for weak rock — with chalk classed as weak.²³  The United States Geological Survey measured California's coastal cliffs across 17,653 transects and got a statewide average of 0.3 ± 0.2 m/yr, or 17.7 metres of retreat over the seventy-year record.²⁴  Fleming's east-coast figure of a metre a year is roughly twice the USGS long-term average for sandy Northeast and Mid-Atlantic shoreline, and about thirty-five times the global median for hard rock.

Where do the book's figures come from?  Fleming says, on camera, and the answer matters more than any objection this article could raise.

"North America East Coast is about a meter per year — that's the last figure, the last figure I pulled off of the top Google search was 1.2, before that it was 0.8, average is about a meter."²⁵

And a year earlier, giving the same argument: "…divide that number by 1.25, which is the average erosion rate according to Google at the moment."²⁶

The 0.8 m/yr he names as his previous figure is the book's 78 cm.  So the load-bearing input of the most widely repeated young-earth erosion argument of the last five years is a search-engine result, which changed underneath its author from 0.8 to 1.2 m/yr, and which he handled by averaging the old snippet against the new one.

He is not hiding this.  He says "according to Google at the moment" on the record, twice, a year apart, and the book itself ends its chapter on continental erosion with a list headed "Search terms:" — nine questions to type into a search engine, plus "(Check Google Earth for the canyons.)"  That is the citation method, stated openly.  Six of the nine footnotes in that chapter point to popular web pages rather than to literature.

This should be said plainly and without sneering, because the alternative charge was worse and this article was about to make it.  Fleming's 20 cm/yr Pacific figure happens to match a specific USGS number exactly — the average over the eroding transects only, in a report whose net California trend across all 14,562 transects is +0.2 m/yr, accretional.²⁷  On the evidence now available that resemblance is a coincidence of round numbers, or Google repeating the erosion-only average; it is not Fleming selecting a subsample.  The real criticism is simpler and he supplies it himself: a number that moves by fifty per cent between searches is not a measurement, and no amount of correct arithmetic downstream can repair it.

There is a further detail in the rocky-coast database worth knowing, because it points the same way.  On France's systematically surveyed rocky coastline the national inventory could not detect retreat slower than 10 cm/yr, and reported no measurable retreat at all on most sections — "these null values represent 67% of the studies of rocky coasts."  The published median of 2.9 cm/yr is computed after those sections are deleted.  Two-thirds of surveyed rocky coast is eroding too slowly to measure over seventy years of aerial photography.

It assumes sea level has always been here.  It has not, and Fleming knows it.  Sea level swings roughly 120 metres through the ice-age cycles, and it has been near its present height for only the last few thousand years.  For most of the Quaternary the shoreline was tens to hundreds of kilometres seaward of where it is now, out on what is today the continental shelf.  Wave attack has not been grinding at one line; it has been sweeping back and forth across a broad zone.

This is not an objection Fleming failed to anticipate.  He states it himself, correctly, in the middle of the same episode:

"One takeaway is that because the sea level is only thought to have been close to its current level for a few thousand years, over the last 300,000 years, most of the coastal erosion for this period would have happened down on the shelf and not at the current coastline."¹³

He then concedes more:

"So at least for part of the Pleistocene and the Holocene, it looks like the mainstream geomorphological narrative of the California coast is fairly coherent."¹³

Having granted that the mechanism spreads erosion across the shelf rather than concentrating it at the modern shoreline, and that the mainstream account of the last three hundred thousand years holds together, he goes on to apply a single-line retreat rate across fifty million years anyway.  That is the pivot on which the argument turns, and it is made in the space of two minutes.

The size of the effect he conceded is worth putting a number on.  Sea level at the last glacial maximum was 134 metres below present, and reached its current height only within the last few thousand years.²⁸  On the 0.1-degree shelf gradient Fleming himself uses, a 134-metre drop moves the shoreline about 77 kilometres seaward.  That is comparable to, or larger than, the entire shelf width he says is available — which means the sea has spent the Quaternary sweeping across the whole shelf rather than grinding at one line, and the shelf is exactly where you would expect the erosion to have gone.

Historic rates also demonstrably overstate long-term ones, and this has been measured at the rock face.  Martin Hurst and colleagues sampled beryllium-10 in flint nodules across the shore platforms below the chalk cliffs at Hope Gap and Beachy Head in East Sussex.  The isotope inventories require retreat of 2 to 6 cm/yr for most of the Holocene, against historically observed rates of 22 to 32 cm/yr — an acceleration of about an order of magnitude within the last few centuries, which they attribute mainly to the thinning of the protective beaches at the cliff foot.²⁹  The bias runs one way: thicker past beaches would mean less isotope accumulation, so correcting for them makes the long-term rate slower still.

In fairness, that acceleration is not universal.  At Hartle Loup in Yorkshire the same technique found retreat steady at 4.5 ± 0.63 cm/yr over seven thousand years, statistically indistinguishable from the 2.7 ± 2.9 cm/yr measured by monitoring over seven years.³⁰  Anyone who wants to answer the coastal argument by saying "historic rates are always inflated" is going beyond the data.

It misreads the continental shelf.  Fleming treats shelf width as a budget: the shelf is what is left of the eroded continent, so a narrow shelf means little erosion has happened.  But the shelf is not an erosional stump.  It is a constructional platform — a wedge of sediment built outward, which is why passive margins like the American Atlantic carry sediment piles kilometres thick.  Its outer edge sits at about 130 metres depth because that is roughly where the sea stood at the last glacial maximum.  The shelf is not the continent's corpse; it is where the continent has been putting the eroded material.  This article's own ETOPO computation puts the world's shelf area between 0 and −200 m at 26.3 million square kilometres — about 18% of the land area, and not a small remnant of anything.

Cliff retreat is self-limiting.  As a sea cliff retreats it leaves a shore platform behind it.  The wider that platform gets, the more wave energy is spent crossing it before reaching the cliff toe, so the retreat slows.  A rate measured today, on a platform that has only existed since sea level stabilised a few thousand years ago, is the fastest that system will ever run.  Extrapolating it is like timing the first minute of a cooling cup of coffee and predicting when it will freeze.

And Fleming's own best evidence dates the coast.  The most striking thing he shows on the podcast is a photograph of the marine terraces on San Clemente Island — a staircase of old wave-cut platforms lifted hundreds of metres above the sea, each one a former shoreline.  He walks through a published reconstruction correlating those terraces with the sea-level curve, re-measures the figure himself, and reports that it holds up:

"It kind of actually nailed it.  So it shows that their diagram was actually quite good, even zoomed in perhaps closer than they thought anyone would ever zoom into it."¹³

"Another takeaway would be that if the diagram is correct, an uplift rate over the last 300,000 years is roughly steady.  So that adds coherence to the story."¹³

A staircase of dated shorelines is a record of hundreds of thousands of years of coastal history, checked by an independent sea-level curve, and it survived being re-measured by a hostile reader.  It is hard to see how that evidence is supposed to support a coast a few thousand years old.

Fleming's own source criticism, graded.  He spends several minutes on the satellite study that reports the world's beaches to be roughly stable — 24% eroding, 28% accreting, 48% stable, with a net global gain of 0.33 m/yr.³¹  Three of his four objections are good ones.  The study really does cover only sandy shoreline, which is 31% of the world's ice-free coast, and his figure of "about 31%" is exactly right.  It really does exclude sea cliffs, which can only retreat.  And the accretion really is concentrated at river deltas and at artificial land reclamation — the paper concedes this itself, noting that "two large-scale land reclamations appear in the top seven accretive beaches in the world," and that Asia's strongly positive mean is "likely due to the artificial development of the Chinese coast."

His fourth objection is wrong, and specifically so.  He says the measurements have a resolution of "about 0.5 metres," so a global mean of +0.33 m/yr is below the noise floor.  The paper's actual positional accuracy is about half a Landsat pixel — 15 metres, not 0.5.  The 0.5 m/yr figure he is reaching for is a rate class boundary, derived from 15 m of position over 30 years, not a resolution.  And because each transect's rate is a regression through up to thirty-three annual shorelines rather than a difference between two images, the uncertainty on the trend is much smaller than the positional error: the authors report that over 60% of their 2.2 million transects have an uncertainty band under half the trend.  The error runs in the direction that would help him if it were true — a 15-metre positional error sounds worse for the study than a half-metre one — so this is a misreading rather than a thumb on the scale.  What actually answers the objection is the regression: because each transect's rate is fitted through up to thirty-three annual shorelines rather than differenced between two images, the uncertainty on the trend is far smaller than the uncertainty on any single position.

One caution about the arithmetic in the popular version.  Jake Ramgren's 2026 blog post takes the three-feet-a-year Cape Cod figure, divides 2,800 miles of continent by it, and gets five million years; the arithmetic is right on its own assumptions.  It also attributes to Fleming a claim that the crust would need to be raised fifty kilometres on the west coast and four hundred on the east, "the same height above sea level as the International Space Station."¹⁴  Those figures come from feeding a horizontal coastal-retreat rate into a vertical uplift budget.  Duff's verdict is correct and this article endorses it: "The '400 kilometers' is not a finding about geology.  It is an artifact of the wrong input, and a model is only as trustworthy as the number you feed into it."¹⁵

The shelf-width clock

There is one more version, made in Fleming's own videos rather than on the podcast, and it is the most striking thing in the whole argument because it looks like a successful prediction.

He measures the width of the continental shelf in Google Earth, averages the American east coast at 131.8 km, divides by his 1.25 m/yr retreat figure, and gets 105,440 years.  He repeats it for the west coast, with its much narrower shelf and much slower rate, and gets 105,158.7 years.  The two independent coasts agree to within 281 years out of a hundred thousand.  His conclusion: "one obvious interpretation of this data is that Pangea split up only thousands of years ago."²⁶

The agreement is real arithmetic and it means nothing, for a reason that has nothing to do with geology.  Both inputs are known to two significant figures at best — one of them is a search-engine snippet that he watched change by fifty per cent.  Two quantities each uncertain by tens of per cent cannot produce a quotient reliable to four significant figures.  The 281-year match is not a measurement agreeing with a measurement; it is a coincidence of rounding presented at a precision the inputs cannot carry.  Had he propagated the uncertainty, both answers would read as "roughly a hundred thousand years, give or take a factor of two," and the striking agreement would disappear.

What the calculation actually shows is that the two coasts' shelf widths happen to stand in about the same ratio as the two coasts' quoted retreat rates — roughly six to one.  There is a simpler reason for that than a common start date: the east coast is wide and soft, the west coast is narrow and hard, and both facts follow from one being a passive margin and the other an active one.  Substrate and tectonic setting control both numbers, so the ratio carries no information about elapsed time.

The version of the coastal argument that survives

The popular form should be dropped.  But something narrower does survive, and it would be dishonest to close the section without printing it.

Grant the tenfold historic-to-millennial correction.  The best-measured millennial rates for rocky coast are still 2.6 to 5.7 cm/yr at Sussex and 4.5 cm/yr at Yorkshire.  Over ten million years those put 260 to 570 kilometres of retreat on a rocky coastline.  Over the two hundred million years assigned to the American Atlantic margin, the global hard-rock median of 2.9 cm/yr gives 5,800 kilometres — wider than North America.  That number is enormous and it should be printed rather than buried.

Reconciling it requires the true long-term rate to be another one to two orders of magnitude slower still.  There is now direct evidence that such rates exist: measuring beryllium-10 in the debris shed off cliffs, rather than on the platform below them, gives millennial retreat of 0.3 to 0.6 mm per year on Mediterranean and Peruvian coasts — fifty to a hundred times slower than the published global median.³²  At that rate a coast retreats only thirty to sixty kilometres in a hundred million years, which is the same order as a real shelf width, and the problem closes.

But that result rests on eleven samples from two low-energy coasts.  It is not yet a global measurement.  So the fair statement is this: the mainstream account of rocky coastlines over deep time depends on a rate correction that is real, is measured at a handful of sites, and has not been established globally.  A skeptical geomorphologist would accept that framing.  It is a much smaller claim than "erosion proves the earth is young," and it is the only part of the coastal argument that a careful reader should still find interesting.

There are also three mechanisms that make slow long-term retreat physically expected rather than merely convenient.  A retreating cliff leaves a widening, shallowing platform that dissipates wave energy before it reaches the cliff toe.  The debris it sheds becomes a beach that armours it — and modelling shows a rocky coastline can settle into a stable plan-view configuration of headlands and bays through that feedback alone, with no uplift and no difference in rock hardness.³³  And of the four feedback loops identified in cliff backwearing, three are self-limiting — wave-energy depletion at the cliff toe, pore-pressure reduction, and protection by the cliff's own debris — against one that reinforces, abrasion.³⁴

Where both sides have been sloppy

Grading only one side would be dishonest, and the rebuttal literature turns out to be in worse shape than its confidence suggests.

The most-cited skeptical answer contains no numbers.  Mark Isaak's Index to Creationist Claims is what nearly every later rebuttal is downstream of.  Its entry on mountain erosion answers a quantitative argument with three assertions and not a single figure: new mountains are being built, present rates are unusually high, and the claim contradicts another creationist claim about volcanism.³⁵  The second point is a factual claim about earth history offered with no citation.  The third is an observation about creationist consistency, not about the earth.

Two widely-copied university pages use a rate that is ten times too small.  A course site at the University of South Dakota and a physics-outreach page both put continental erosion at 0.005 mm/yr and then declare the creationist budget baseless.³⁶  Whether one copied the other could not be determined, and the South Dakota page attributes its figure to a book by Hugh Ross rather than deriving it.  That figure is close to the median rate for bare rock outcrops — 5.4 m/Myr in the global compilation — not the median for drainage basins, which is 54 m/Myr and is the relevant quantity for a whole-continent sediment budget.³⁷  Using the outcrop number makes the problem disappear by arithmetic sleight rather than by physics.

The "modern rates are inflated by farming" reply is much weaker than its users think, and Fleming is closer to right about it.  Asked on the podcast whether human activity accounts for the discrepancy, he answers: "the estimates that I've seen is that maybe it's twice as high, so it's not the orders of magnitude that you need."⁴  Three separate lines of evidence say he is substantially correct for the quantity his own calculation uses.

• Cosmogenic beryllium-10 measures the pre-human rate directly, because the isotope inventory is set tens of centimetres down and integrates over thousands of years. Two centuries of ploughing is a rounding error in that window. As Portenga and Bierman put it, "shallow, human-induced soil erosion does not typically affect cosmogenic estimates of basin-scale erosion rates."³⁷
• The measured pre-agricultural global flux from that method is 15.2 Gt/yr against 24 Gt/yr for the modern sediment and dissolved load reaching the ocean — a factor of about 1.6, not a factor of a hundred.¹⁶
• Where the hundredfold figure is real, it does not do the work people want. In the southeastern United States, peak post-settlement hillslope erosion ran about 950 m/Myr against a beryllium-10 background of about 8 m/Myr — more than a hundredfold. But river sediment yields over the same basins rose only five- to tenfold, because the rivers were carrying only about six per cent of the eroded material; the rest is still sitting at the bottom of hillslopes and in valley floors.³⁸

The flux that matters for "how fast is the continent losing mass" is the flux that leaves the continent, and that one has roughly doubled, not multiplied by a hundred.  Anyone rebutting this argument by pointing at agriculture is overstating their case.

The reason the two numbers differ so wildly is geographical, and Wilkinson and McElroy state it in one pair of sentences: about 83% of the global river sediment flux comes from the highest ten per cent of the earth's surface, while about 83% of cropland erosion happens on the lowest sixty-five per cent.¹⁹  The natural sediment machine is mountains.  Agriculture is on the plains.  They barely overlap, so ploughing a great deal of flat land multiplies hillslope soil loss enormously and moves the global export figure hardly at all.

The same paper supplies a number that does more work than anything else in this section, and it is not derived from modern rates at all.  Reading the Phanerozoic sedimentary rock record itself — the volume of sediment actually preserved, corrected for later erosion and subduction — Wilkinson and McElroy get a mean continental denudation rate over the last half-billion years of 16 metres per million years, with epoch-by-epoch values ranging from about 4 m/Myr in the Middle Triassic to about 53 m/Myr in the Pliocene.  That is a rate measured from the rocks, not extrapolated from a river gauge, and it is roughly a third of the present-day figure.

Put it through the same arithmetic and the continents level in about fifty million years without isostasy, or three hundred million with it.  This does not rescue an old earth by itself, and this article will not pretend it does.  What it does show is that the present is not a good guide to the Phanerozoic average — which is a point creationist and mainstream writers can agree on, arrived at from the rock record rather than from anybody's model.

In steep terrain the bias runs the other way entirely.  In thirty-two Idaho catchments where the Forest Service had been gauging sediment for decades, cosmogenic rates came out on average seventeen times higher than the gauged rates, because most sediment moves in rare events no gauging programme is likely to catch.³⁹  The same paper found that 70 to 97 per cent of sediment delivery must happen in episodes too infrequent to be measured conventionally — a result creationists could legitimately cite, and generally have not.

One defence that sounds good and does not work.  Measured geological rates famously get slower as the measurement window lengthens — Sadler's effect — so it is tempting to answer the erosion argument by saying that any short-term rate overstates the long-term one.  That defence has to be given up here.  Sadler himself, with Douglas Jerolmack, reports that upland denudation rates derived from sediment yield show little or no dependence on the measurement interval, and that the global sediment flux to the oceans has been roughly constant from months to tens of millions of years.⁴⁰  A 2024 compilation of about 6,400 measurements finds no timescale dependence in fluvial erosion rates at all, apart from an artefact specific to the cosmogenic method.⁴¹  The effect is real for accumulation at a single locality; it is not a licence to discount continental denudation.

The ally this article leans on hardest is not exempt either.  Joel Duff's reply to the coastal argument is the best short answer in print and is quoted here four times, but two things in it should not be repeated.  His stated range of ten to twenty-five million years for planing North America does not reproduce from his own inputs — six hundred metres divided by his own 20 to 60 m/Myr gives ten to thirty.  And his claim that isostatic uplift on passive margins like the American east coast runs at "fractions of a millimeter to a few millimeters per year" is uncited and appears to be wrong in sign for the coast under discussion: much of the mid-Atlantic margin is subsiding, at one to two millimetres a year, as the glacial forebulge collapses.  That matters, because uplift on passive margins is precisely the mechanism his argument and this one both need.

And the creationist side has misused its own best source.  In a 2019 ICR article, Jake Hebert cites Schumm's 1963 USGS paper in support of the erosion argument, and dismisses episodic uplift and isostatic compensation as "imagined rescuing devices."⁴²  Both of those devices are argued for, quantitatively, inside the paper he is citing: Schumm's entire thesis is that uplift outruns denudation by roughly eightfold, and Schumm states that isostatic compensation multiplies peneplanation time "by a factor of five or much more."⁷  A reader who follows the citation finds the opposite of what it was offered for.

A second ICR article, by Brian Thomas, describes beryllium-10 as "an isotope of the element beryllium that becomes radioactive with exposure to the sun."⁴³  Beryllium-10 is produced by cosmic-ray spallation, not by sunlight, and nothing "becomes radioactive" by being illuminated.  It is the kind of error a hostile reader finds in thirty seconds, and it is on our side of the ledger.

The double edge

Here is the part that ought to worry a young-earth reader most, and it was first pointed out by a creationist.

In the same 1986 paper where Glenn Morton called continental denudation "the best geological argument" for a young earth, he added this:

"However, the young earth creationists, like the old earthers, have generally ignored the problem presented by these same erosion rates when regional unconformities are discussed."²⁰

The point is simple, and only one person has seriously taken it up.  Michael Oard's answer, across a four-part series in the Journal of Creation, is that the kilometres of continental erosion happened during the receding phase of the Flood rather than at present rates — on his own accounting, present processes have removed an average of only 0.275 m from the continents in 4,500 years.  That is a real answer and it is the right shape of answer.  What it costs is the argument this article is about: if present rates are unrepresentative by three orders of magnitude, they cannot also be used to date the continents.  If observed erosion rates are the yardstick, they have to be applied everywhere — including inside the rock sequence that flood geology says was laid down in a single year.

Morton's own example came from seismic data he had worked with professionally.  Beneath the coastal plain of the southern United States lie buried Appalachian folds ten to twenty thousand feet high, with up to ten thousand feet planed off their tops, and then twenty-five hundred feet of undisturbed Cretaceous strata lying flat across the truncated surface.  Using the fastest erosion rate in his own list — 7.5 feet per thousand years, from the Transverse Ranges of California — stripping that ten thousand feet takes 1.3 million years.²⁰  That is one erosional episode, in the middle of the sequence, on the most generous rate available, and it does not fit in a year.

The problem gets worse when creationist authors state their own erosion requirements.  Andrew Snelling and Tom Vail put the volume removed from the Colorado Plateau at 400,000 km³ over an area of 647,000 km² — which works out to 618 metres of rock stripped off the whole plateau, in or shortly after the flood year.⁴⁴  At the global background rate of 38 m/Myr, that is sixteen million years' worth of erosion.  Their own figure, incidentally, is sourced not to a measurement but to another creationist popular book.

And the same arithmetic runs in the depositional direction.  Timothy Helble, a retired hydrologist, took Steven Austin's own numbers for the Coconino Sandstone — 42,000 km³ of sand, a source boundary 1,600 km long, and about twelve days available for it within the early-flood window on Austin's own proportional accounting — and asked what sediment transport rate would be needed.  He then granted flood geology a transport rate of 30 kg per second per metre of front, ten times the highest value on the published curve, and ran it for the full twelve days across the entire 1,600 km boundary.  The result is 26 km³ of sand.  The requirement is 42,000 km³.  The most generous assumptions available deliver one part in sixteen hundred of what the model needs; at that rate the Coconino takes about fifty-three years.⁴⁵

Helble also quotes a 1994 paper by six leading flood geologists conceding the underlying difficulty in print:

"It may not be possible to derive all the Flood sediments from igneous and/or metamorphic precursors by physical and chemical processes in the course of a single, year-long Flood."⁴⁵

Joel Duff puts the resulting bind in one sentence, and it is the sentence this whole section exists to make:

"You cannot simultaneously argue that observed erosion rates prove the earth is young and that observed erosion rates are a misleading remnant of a vanished catastrophe.  One of those claims has to go."¹⁵

That is exactly right, and it is a problem of the argument's own making.  The erosion argument works only if present rates are representative of the past.  Flood geology works only if they are wildly unrepresentative of the past.  Running both at once is not a strong position held twice; it is one position that cancels the other.

There is a published reply, and it concedes the point.  A 2025 presentation at the International Conference on Creationism is titled "Helble's Critique of Creationist Research on the Coconino Sandstone Fails."  It was obtained and read for this article.  Three things about it matter.

It is a one-page conference abstract rather than a paper — no figures, no equations, four references — and its authors call it "an initial critical response."  Its lead author, John Whitmore, edited the proceedings volume his own reply appears in.  Its second author is Paul Garner, the co-host of the podcast that is this article's primary text.

And on the arithmetic it does not fight.  It contests no number in Helble's chain: not the 42,000 km³, not the 1,600 km front, not the twelve days, not the transport rate, not the resulting shortfall.  It offers no counter-calculation of any kind.  What it offers is this:

"Helble is correct that the processes he outlines could not have formed the Coconino during the Flood, but the Flood model requires extrapolations beyond conditions that prevail in the present-day.  If a Flood model is correct, we will never be able to observe the depositional processes that made the Coconino in the laboratory or in nature."⁴⁶

That is a fair statement of the position, and it should be taken at face value rather than mocked.  But notice what it costs.  The reply grants that no observable process can deposit the Coconino in the time available, and appeals instead to conditions that are unobservable in principle.  An argument that rests on present erosion rates being a reliable guide to the past cannot also rest on the past being unobservable in principle.  The reply is honest, and it concedes the double edge rather than escaping it.

The abstract's real dispute with Helble is a different one — whether the Coconino is wind-blown or underwater — and on that it lists eleven lines of evidence and argues none of them, promising fuller work "still being prepared for publication and peer review."  As of this writing that work has not appeared.

What erosion actually removed

The objection Roth and Woodmorappe raise — that tectonic replenishment would have stripped tens of kilometres off the continents and left no Phanerozoic sedimentary rock — makes a testable prediction.  It is worth taking seriously, because the prediction is right.  Tens of kilometres have been stripped off, in specific places, and there is a way to measure how much.

Minerals record the pressure at which they formed.  A metamorphic rock's mineral assemblage depends on the temperature and pressure of its last equilibration, and pressure at depth in the crust is simply the weight of the rock column above.  Certain minerals are diagnostic: kyanite forms only in a restricted high-pressure field, and coesite — a dense form of silica — requires pressures found only at depths beyond about eighty kilometres.  When a rock containing kyanite is sitting at the surface where you can hit it with a hammer, the column that was once above it is no longer there.

One caution, because it is a trap and a knowledgeable reader will spring it.  Coesite-bearing rocks, like those in the Western Gneiss Region of Norway, prove burial past a hundred kilometres and return to the surface — but they were not unroofed by a hundred kilometres of erosion.  Ultra-high-pressure terranes come back up buoyantly, flowing along the subduction channel, and most of their ascent is tectonic rather than erosional.  They are evidence about how deep rock can travel, not about how much rain has fallen.  The numbers this article uses are crustal-depth figures from ordinary orogenic belts, where the removal is much better attributed to unroofing and erosion.

That is not an inference from erosion rates.  It is a measurement of depth, made on the rock itself, and it is entirely independent of how long the removal took.  The conversion is straightforward: pressure at depth is the weight of the column above, so at a mean crustal density of 2,750 kg/m³ one kilobar corresponds to about 3.7 kilometres of rock.

The southern Appalachians are the test case, because they are the range Morton used against DeVilbiss.  Rocks in the Blue Ridge record equilibration pressures from about 5.8 kbar at staurolite grade to 11.5 kbar in the Taconic core — that is, burial under roughly 21 to 43 kilometres of overburden, none of which is there now.⁴⁷

Removal is not the same as erosion, and the difference matters here.  A good part of that overburden was taken off tectonically: thrust sheets stacked during the Taconic, Neoacadian and Alleghanian events were later carried away by faulting, not by rainfall.  So the honest statement is that 21 to 43 km of rock was removed from above these rocks by a combination of tectonic unroofing and erosion, and only the second part is denudation in the sense this article has been measuring.  Spread across the ~450 million years since the Taconic that is 48 to 95 metres per million years, which is an ordinary orogenic figure — but it is an average across the orogenic interval, not a rate for the quiet time afterwards.

And that distinction is exactly where Morton's objection bites.  His point was about the interval after the Appalachians stopped growing: "since that time there has been no significant orogenic activity… and the subsequent history of the mountain chain has been one of erosion."  Averaging the orogenic and post-orogenic intervals together answers a different question from the one he asked.  The post-orogenic numbers are much smaller — Cenozoic exhumation of the southern Appalachians has likely not exceeded about 1.2 km, and the removal was front-loaded — and no study closing the full 300-million-year budget was located for this article.  What can be said is that the range is still standing after 300 million years because erosion slowed as relief fell, because isostasy kept lifting the root as the load came off, and because there is evidence of Miocene rejuvenation of relief.  That is three mechanisms rather than a measurement, and Morton is entitled to say so.

This is why the two halves of the objection describe two different places.  In orogenic belts the geologic column has been eradicated, exactly as Roth predicted it would have to be — you are standing on rocks that were once twenty or thirty kilometres beneath the surface, and the sediments that used to lie over them are somewhere else, as sediment.  On cratons and stable interiors, where the flat-lying Phanerozoic cover survives, the measured erosion rate is not the global average at all.

The global compilation of beryllium-10 measurements makes the size of that difference plain.  Across 1,599 recalculated rates, bare rock outcrops erode at a median of 5.4 m/Myr and a mean of 12 m/Myr, and not one outcrop in the entire global set exceeds 140 m/Myr.³⁷  Drainage basins, which include soil-mantled slopes and steep terrain, come in ten times faster: median 54, mean 218 m/Myr.  On the oldest and quietest surfaces the numbers get slower still.  Australian granite on the craton has, in the authors' careful wording, "in some places eroded only decimeters over the past million years."  The full measured range across their sixty-one samples is 0.3 to 5.7 m/Myr, and where this article names a single Australian figure it uses the bottom of that range, which the reader should discount accordingly.⁴⁸  Namibian bedrock runs 3.2 to 3.6 m/Myr.⁴⁹  The southeastern United States, a passive margin, ran about 8 m/Myr before European settlement.³⁸

Put those two facts together and the objection dissolves.  The arithmetic is worth seeing:

Those cratonic figures are the right order for what cratons have actually lost.  The Colorado Plateau's entire Palaeozoic and Mesozoic sedimentary section is about three kilometres thick, sitting on a crust forty to forty-five kilometres deep, and the plateau has shed only four to eight hundred metres of it through the Cenozoic.⁵⁰  Australia at the slow end of its measured range loses three hundred metres in a billion years, which is why surfaces and inselbergs there survive at all — though at the fast end of the same dataset a billion years removes 5.7 km, and the honest reading is that the craton spans both.  And there is independent confirmation that the quiet places really were quiet: a GSA Bulletin study using thermochronology and stratigraphy, with no stake in this argument, concludes that the Colorado Plateau "remained near sea level for 500 m.y."⁵¹

The young-earth argument reaches its contradiction by applying mountain rates to cratons and craton preservation to mountains.

Laid out on a log scale, the spread is the whole answer.  Measured rates run from 0.3 metres per million years on Australian craton granite to nearly a thousand on ploughed hillslopes — more than three orders of magnitude.  There is no such thing as "the" erosion rate to divide the continents by.  The red line marks the rate at which the present land surface would survive four and a half billion years on isostasy alone with no tectonics at all: about one metre per million years.  Only the very slowest cratonic surfaces sit below it.  That is precisely why active replenishment is not an optional extra in the account — it is required by the measurements.

Four things about that reply are weaker than they look, and a skeptic will find them.  First, the basin median of 54 m/Myr does not reconcile as easily as the outcrop median; the answer — that basin rates are dominated by steep, tectonically active, soil-mantled terrain, and that the sampled basins cover only 2.3% of the land surface with known geographic bias — is a good argument but it is an argument, not a measurement.  Second, cosmogenic rates integrate over thousands to a million years, and stretching them to hundreds of millions assumes a stationarity that cannot be directly tested.  Third, cratons are not closed systems: they subside and receive sediment as well as shedding it, so "net exhumation over 500 Myr" is a balance, not a running total.  Fourth, and most awkwardly, multiplying an average rate by a long time is precisely the move this article criticises the erosion argument for making.

That last point is why the barometry matters so much.  It does not depend on any rate at all.  A rock whose minerals cannot form under less than thirty kilometres of overburden, sitting in a roadcut, is evidence of removal regardless of how fast or slow or steady it was.  Everything else in this section is supporting arithmetic; that is the measurement.

One further step is needed to answer Woodmorappe's version, and it is worth spelling out.  His objection assumes that removing tens of kilometres would consume the crust — that a continent 37 kilometres thick cannot afford to lose thirty.  It could not, if the crust were a fixed slab.  But mountain building thickens crust before erosion thins it.  Continental collision stacks and shortens the crust to fifty, sixty, in places seventy kilometres — the present crust beneath Tibet is roughly double the continental average — and erosion then unroofs that thickened pile while isostasy lifts its base.  The sequence is not "start at 37 km and subtract 33."  It is "thicken to 60, strip 30 off the top while the base rises, end near 37."  The rock that was removed was not borrowed from a fixed budget.  It was manufactured by the collision that raised the mountains, which is the same event the erosion is evidence of.

Fleming's hardest argument

There is a stronger version of this objection than Woodmorappe's, and Fleming makes it on the podcast.  It deserves an answer rather than a mention.

He grants the reply, then tries to falsify it on a mass budget.  If the American east coast has been retreating at a metre a year for two hundred million years, he says, and if uplift is what keeps it in place, then you would have to rebuild the crust about thirteen times over — once every fifteen million years — and therefore "you shouldn't have crust older than 15 million years on the East Coast."¹³  Then he brings a test: the measured age distribution of continental crust, which does not look like that at all.  Most of it formed before the Phanerozoic.  And he quotes a peer-reviewed conclusion at it: "regardless of sampling approach or geographic scale, igneous rock quantity does not decrease exponentially with increasing age as predicted by most basic models of rock cycling."

This is a real argument with an empirical test attached, and it is the best thing in either episode.  It fails, but not for a trivial reason.

The premise is the coastal number, which the previous section has already dealt with.  Set that aside; the argument still fails, and the reason is worth having.

Erosion does not destroy continental crust.  It relocates it.  When rock is stripped off the Appalachians it does not leave the continent.  It travels a few hundred kilometres east and settles on the continental shelf and the coastal plain — and the shelf and coastal plain are continental crust, thinned and buried but not gone.  Nothing needs to be "reformed," because nothing was consumed.  Only the fraction that reaches the deep ocean and rides a plate into a trench is genuinely removed from the continental budget, and that fraction is small.

And the crustal age distribution does not measure what the argument needs it to measure.  A crustal formation age records when rock crystallised from melt.  Erosion does not reset it.  Strip five kilometres off a Precambrian shield and the zircons in the rock you expose still give Precambrian ages, because they were Precambrian before you started.  The age spectrum is a record of when crust was made, not of how much surface has since been carried away.  The quoted sentence about igneous rock quantity is about whether the volume of preserved igneous rock decays with age, which is a real and interesting question and a different one.

The test Fleming proposes would work against a model in which erosion destroys crust and magmatism replaces it wholesale.  Nobody holds that model.  What holds the continents up is thickening at convergent margins, magmatic addition at arcs, and isostatic rebound as load comes off — none of which resets a zircon clock.

The global mean is the wrong statistic, and this is measurable.  When Jane Willenbring and colleagues corrected their own earlier extrapolation — a public self-correction of a factor of five, which is worth noting in a debate about intellectual standards — they found that the steepest ten per cent of the continental surface supplies close to half of all global denudation.⁵²  Wilkinson and McElroy, working from preserved rock volumes rather than from cosmogenic nuclides, put the share of the highest ten per cent at about 83% — the two estimates use different measures of terrain and different data, and the honest summary is that a small fraction of the land supplies somewhere between half and four-fifths of the sediment, with the exact fraction unsettled.

Either way the point stands.  Averaging an actively rising orogen together with a Precambrian shield, and then treating the average as a depletion rate for the shield, is the actual error in the calculation.

And the clock has been checked against other clocks.  Cosmogenic erosion rates have been tested against the volume of sediment physically piled up in alluvial fans over sixteen thousand years — agreeing within one standard error in the study that first did it,⁵³ and within twenty per cent in a later one — and against apatite fission-track exhumation over roughly ten million years and 3.6 km of rock removal, agreeing within a factor of two.³⁹  Three chronometers with no physics in common, spanning four orders of magnitude in time, give the same answer.  That concordance is the thing a young-earth account has to explain, and this article is not aware of any attempt to.

The young-earth response to this dataset is worth reporting exactly.  It is not silence.  Michael Oard addresses cosmogenic methods in the Journal of Creation and rules them out before looking at them: "Some of them are indirect, such as chemical and cosmogenic isotopic methods, which rely on radiometric dating.  Since these methods are based on uniformitarian assumptions, other more scientific methods will be used."⁵⁴

That is a presuppositional refusal rather than a technical rebuttal, and it is a weaker position than silence would have been, because the concordance described above is not an assumption — it is three independent clocks agreeing, and one of them, the alluvial fan, can be measured with a shovel.  Beyond that refusal no technical engagement with the beryllium-10 denudation data was found in the Journal of Creation, Answers Research Journal, or the Creation Research Society Quarterly.  That search was not exhaustive and the CRSQ archive is poorly indexed, so the claim here is limited to what was looked for and not found.

Surfaces that should not have survived

Ariel Roth's strongest single move was not a rate calculation.  It was to point at flat land surfaces that mainstream geologists themselves say are extremely old, and ask how they can still be there.  His authority for this is C.R. Twidale, an Australian geomorphologist who wrote about the problem at length and in his own words called it an embarrassment:

"The survival of these paleoforms is in some degree an embarrassment to all of the commonly accepted models of landscape development."⁵⁵

And:

"Even if it is accepted that estimates of the contemporary rate of degradation of land surfaces are several orders too high… there has surely been ample time for the very ancient features preserved in the present landscape to have been eradicated several times over.  Yet the silcreted land surface of central Australia has survived perhaps 20 m.y. of weathering and erosion under varied climatic conditions, as has the laterite surface of the northern areas of the continent.  The laterite surface of the Gulfs region of South Australia is even more remarkable, for it has persisted through some 200 m.y. of epigene attack."⁵⁵

Both quotations are accurate, and Roth quotes them accurately.  A two-hundred-million-year-old land surface, never buried, is a striking thing to have to account for.

Twidale accounts for it, in the same paper, in three ways that the creationist retelling leaves out.

First, he draws the distinction that does most of the work himself, on the second page: surfaces that were buried and later re-exposed are not evidence of anything.  "They have been preserved as a result of burial and have only recently been reexposed to subaerial attack.  Their survival thus poses few problems."⁵⁵  A great many of the world's "ancient surfaces" — Precambrian ones in Canada, Greenland, India, and beneath the Great Artesian Basin — are of that kind.

Second, and more importantly, the genuinely old Australian surfaces are not preserved ground surfaces at all.  They are etch surfaces, and Twidale says so explicitly:

"There is no evidence to suggest that these paleoplains and other paleoforms have been buried.  But many are of etch type: the present surface represents either an exposed indurated horizon of the weathering profile, or the weathering front, that is, the base of significant weathering, or the junction between the regolith and the intrinsically fresh bedrock."⁵⁵

That changes the object being dated.  A laterite or silcrete surface is a hardened layer that formed inside a weathering profile, several metres below the ground of its day, under a porous sandy horizon that Twidale describes as "an absorbent cushion protecting the ferruginous horizon."  Stripping that cushion away and exposing the duricrust is not the survival of an old landscape.  It is the creation of the present one.  The old thing is a chemical horizon, not a topography.

Third, Twidale lists five mechanisms he thinks preserve what does survive: aridity, repeated uplift, resistant structure and lithology, the concentration of river erosion into trunk streams while the ground between them barely lowers, and armouring of gullies by coarse debris shed from the caprock.  None of these is a rescue invented later; they are in the paper Roth reviewed.

And the measurements since have gone a long way towards resolving it, though not all the way.  The Australian surfaces erode at about 0.3 metres per million years, which removes three hundred metres in a billion years — quite slow enough for a landform to keep its shape.  Crucially, they are eroding: not one of the sixty-one Australian samples was saturated in cosmogenic nuclides, which means every surface measured is losing mass.⁴⁸  These landforms are old in form while being continuously lowered, which is a different claim from "nothing has happened here."

Fleming cites a real paper on exactly this, and it does not say what he takes it to say.  In his chapter on continental erosion he notes that "I have seen rare mentions of the topic in scientific literature, but the authors assume that long ages are true, and therefore erosion rates have been extremely, impossibly low."  His citation is Flowers, Bowring and Reiners, who used ancient helium dates on the western Canadian shield to report long-term unroofing of 2.5 micrometres a year or less since about 1.7 billion years ago.⁵⁶  That is indeed an extremely low rate.  But the authors do not present it as a rescue: their own sentence is that these rates, while far below what earlier thermochronology suggested for continental interiors, "are within the range of short-term erosional estimates."  The rate is low because a Precambrian shield is the flattest, most tectonically dead terrain on the planet — which is the relief-dependence point again, arriving from a source Fleming chose himself.

The honest residual is real, and larger than the paragraph above implies.  The cosmogenic rates measured on those very surfaces span 0.3 to 5.7 m/Myr, a factor of twenty, so arithmetic that makes a 200-million-year surface comfortable at the bottom of the range strips more than a kilometre off it at the top.  Twidale did not accept the cosmogenic result quietly either; he published a formal Comment against it in Quaternary Research.  Several Scandinavian and Greenland workers — Green, Lidmar-Bergström, Japsen, Bonow and Chalmers among them — push the other way, arguing that he over-dates and that no never-buried surface older than the Miocene is likely to survive.  That is an active professional disagreement running in both directions, not a settled matter with a creationist holdout.  And Twidale's word was "embarrassment," which he wrote himself, about his own field, and which no amount of context makes disappear.

Mountains, and the source that says the opposite

The third form of the argument is about uplift, and it has the cleanest resolution of the three, because the paper both sides cite settles it.

The claim, from Roth in 1986 and Fleming in 2024, is that measured uplift rates multiplied by deep time give impossible mountains.  Roth quotes Schumm's figure of 7.6 metres of uplift per thousand years, notes it is about eight times the maximum rate of denudation, and reports the textbook consequence that for erosion to keep pace with a typical uplift rate a mountain would have to stand about forty-five kilometres high.  Fleming works the same argument on Everest: erosion in the Everest region of about 2.1 mm/yr would level it in 4.2 million years, published uplift rates run from 4 mm/yr up to 65, and over the forty million years he takes for the range you would need "somewhere between 100 miles and 1,400 miles of uplift."⁴

Schumm's paper says the opposite of what it is cited for, in the same abstract.  Immediately after the sentence about the eightfold disparity comes this one:

"Calculations of the time required for planation of 5,000 feet of relief suggest that ample time exists between orogenic periods for the development of peneplains."⁷

Quoting the first sentence without the second is quoting half of a pair.  And Schumm goes further inside the paper, doing the isostatic correction himself, forty years before this article did it:

"If the denudation rate at 12,500 feet (1.6 feet per 1,000 years) is the average of these rates, 3 million years are required for the reduction of 5,000 feet of relief.  If isostatic compensation requires a 5-fold adjustment, then peneplanation will occur in 15 million years."⁷

A factor of five, from the 1963 source, independently reproducing the factor of six derived above.  Schumm's own curves also sit awkwardly with the forty-five-kilometre claim, though they do not refute it cleanly.  What he writes is that maximum denudation rates "do not reach the average rate of uplift, 25 feet per 1,000 years, until relief is well above 20,000 feet" — a floor of about six kilometres, with no ceiling stated.  He does not say where the crossover is.  And the two figures are not the same quantity: Roth's forty-five kilometres comes from Blatt and colleagues comparing Ahnert's basin-average rates against a 10 m/kyr "typical" uplift, where Schumm is comparing maximum rates against 7.6 m/kyr.  The honest statement is that the forty-five-kilometre number is not Schumm's and is not supported by him.

Finally, Schumm says where his 7.6 m/kyr uplift figure came from, and it is not a global rate.  It is the average of five levelling stations around Los Angeles, sitting alongside measurements of glacial rebound in the Gulf of Bothnia and of the ground sagging under Lake Mead.  His own caution: "The above rapid rates of uplift were measured in tectonically active areas or those adjusting to increase or decrease of load."

A measured uplift rate is mostly not permanent, and this has been quantified.  Peter England and Peter Molnar made the distinction in Geology in 1990, aimed at professional geologists rather than creationists: "Many reports of surface uplift in mountain ranges are based on mistaking exhumation of rocks or uplift of rocks for surface uplift, and provide no information whatsoever on the rates of surface uplift."⁵⁷  Ground measured moving upward between earthquakes is mostly elastic strain that the next earthquake gives back.  For the Himalaya the permanent fraction has been measured: Stevens and Avouac found convergence of 13 to 21 mm/yr across the range, of which the anelastic — that is, permanently retained — part is at most about ten per cent.⁵⁸  Multiplying the whole geodetic rate by tens of millions of years is multiplying by roughly ten times too much, and then compounding the error by assuming none of it erodes away.

And where anyone has actually closed the mass balance, it closes.  In Taiwan, erosion runs at three to six millimetres a year on every timescale from decades to millions of years — the mountains are being built and removed at the same speed.⁵⁹  For the Himalaya the sum has been done over twenty million years: about 7.1 million cubic kilometres eroded off the range against about 7.4 million deposited in the Bengal Fan and its neighbours — a four per cent difference, against stated uncertainties of twenty per cent on each figure.⁶⁰  The rock that came off the mountains is not missing.  It has been located, and there is the right amount of it.

Everest

Fleming's Everest numbers deserve individual treatment, because one of them is wrong and one of them identifies something real.

The 65 mm/yr uplift figure has no published basis — though it should be said at once that Fleming does not rest anything on it.  On air he explicitly runs the argument on the slowest published rate instead: "back to the small measurement, if even the 4 millimeters per year, even at that rate, when you stretch it out over the 40 or 60 million years of Everest's supposed history, it does create a problem."  Taking the weakest number that still makes your point is good practice, and the criticism below is aimed at the inference, not at the 65.  No study reports vertical motion of Everest at anything like that rate; the number is the right size for horizontal India-Eurasia convergence, and appears to be a horizontal rate used as a vertical one.  The measured GPS rock uplift is about 2 mm/yr, and the long-term rate from thermochronology is about 1 mm/yr.

But that gap is real, and mainstream geology finds it interesting rather than embarrassing.  A 2024 paper in Nature Geoscience opens by naming it:

"Further evidence suggests a disconnect between the short-term rock uplift rate of Chomolungma measured from GPS data (~2 mm yr⁻¹) and its long-term rock uplift rate recorded by thermochronology (~1 mm yr⁻¹), which raises the question, is there an underlying mechanism raising Chomolungma's anomalous elevation ever higher?"⁶¹

Their answer is a specific, dated, local transient: about ninety thousand years ago the Arun river captured a neighbouring drainage, the accelerated incision that followed unloaded the crust, and the resulting isostatic rebound is lifting Everest an extra 0.16 to 0.53 mm/yr — accounting for perhaps fifteen to fifty metres of its 8,849.  That is erosion raising a mountain, which is worth sitting with for a moment, since the young-earth argument treats erosion as purely subtractive.

So Fleming is right that the short-term uplift rate is about twice the long-term rate.  Where the argument fails is in what follows: the gap is the subject of a recent paper in a leading journal, and its resolution is that the fast rate is a local transient rather than the steady state — which is exactly the kind of answer the argument needs to rule out and does not.

One claim could not be checked, and the review that reports it garbles the subject.  On air Fleming says: "I ran into one article on Everest specifically, and on the erosion rate, the article said in the conclusion that the erosion rate of Everest must have been 50 times lower in the past than it is now.  No mechanism — [it] just must have been that way."⁴  Woodmorappe's review renders this as a claim about "the volumes of sediment removed by rivers today," which is a different and much broader assertion.  Fleming's version is narrower and about one mountain.  Neither names the paper, and the book could not be obtained, so the source remains unidentified.  A fiftyfold discrepancy asserted without a mechanism in a mainstream paper would be worth seeing; until someone produces the citation it cannot be assessed either way.

The ocean-sediment version, and one wrong denominator

The sister argument runs the same calculation in the depositional direction.  If rivers deliver twenty-odd billion tonnes a year, the ocean floor should be buried in sediment.  It is not.  Nevins divided the mass of ocean sediment by the annual input and got thirty million years, against an ocean he took to be a billion years old.  Andrew Snelling has made a version of this Answers in Genesis's single strongest young-earth evidence, listing it first in "The 10 Best Evidences from Science That Confirm a Young Earth."⁶²  Fleming's version: there is "somewhere between one and two and a half times more sediment on the floor of the ocean than the continental mass above sea level," which at present rates is, in his words, fifteen to at most fifty million years' worth, against an ocean basin "supposed to be a couple hundred million years old."⁴

Everything in that paragraph is arithmetically fine.  The denominators are not.

Fleming's is the closest to right of the three, and the distinction is worth being precise about: "a couple hundred million years" is roughly the age of the oldest surviving ocean floor, which is a real number.  His error is using a maximum where an average belongs.

The ocean floor is not as old as the ocean.  Oceanic crust is manufactured at spreading ridges and destroyed at trenches, so it has a residence time, and sediment older than the crust it sits on cannot exist.  The area-weighted mean age of the world's ocean crust has been measured: 64.2 million years.⁶³  Not one billion, and not three.

Redo the comparison with that number and the problem evaporates.  Take Fleming's own bracket — ocean sediment at one to two and a half times the 3.19 × 10¹⁷ tonnes of continent standing above sea level — and divide by the measured pre-agricultural denudation flux of 15.2 Gt/yr.  The answer is twenty-one to fifty-two million years' worth of continental erosion, sitting on a seafloor whose mean age is sixty-four million years.  Those are the same number to within the precision either side can claim.

Three further complications all point the same way and none of them rescues the young-earth reading.  A large majority of river sediment never reaches oceanic crust at all — it stops on continental shelves, in deltas, and in nearshore basins, which sit on continental crust.  A large fraction of deep-ocean sediment was never continental in the first place: pelagic ooze is mostly biogenic calcite precipitated in surface waters.  And sediment is continuously removed from the ocean, by subduction at trenches and by accretion into mountain belts.  Jonathan Baker's point is the sharpest of these: the "average ocean sediment thickness" figure the argument uses is a deep-ocean pelagic number, so the calculation compares the wrong reservoir to the wrong source.⁶⁴

The mass balance actually closes, and closes tightly.  Take the mass of sediment on the ocean floor — 2.62 × 10¹⁷ tonnes — and divide it by the rate at which sediment is destroyed by subduction, about 1.5 km³ a year or 4.05 Gt/yr.  The result is a residence time of 65 million years, against a measured mean seafloor age of 64.2 million.  Those agree to within two per cent.⁶⁵  Using Hay's own stated minimum removal rate of 1.0 Gt/yr instead stretches the residence time to 262 Myr, which brackets the 180-to-200-million-year age of the oldest surviving ocean floor.  There is no combination of published numbers that gets near six thousand years: six thousand years at 20 Gt/yr delivers about one two-thousandth of the sediment that is actually there.

Nevins was not even wrong about the removal term.  His figure of 2.75 Gt/yr destroyed by sea-floor spreading is close to the modern long-term estimate of about 2.7 Gt/yr.  He had the right numerator and the right sink.  He divided by the wrong clock.

So the correction is not the same size for everyone.  Fleming is out by a factor of about three, from using the oldest seafloor age instead of the mean.  Nevins is out by a factor of about fifteen, from using the age of the ocean.  The version that compares sediment against a three-billion-year ocean is out by a factor of nearly fifty, and is comparing against an age no geologist assigns to the seafloor at all.  In every case the numerator is fine and no care taken with it repairs the divisor.

And the sister argument: is there too little sediment altogether?

Roth's other mass-balance claim is broader.  Citing Gregor, he argues that at present denudation rates roughly twenty-three times more sedimentary rock should exist than does.

His arithmetic is right, and this is the most interesting fact in the section.  Recomputing the ratio from modern measurements of how much sediment survives from each era — about 20% of Phanerozoic, 2% of Proterozoic, and 0.5% of Archean — reproduces his figure to two significant digits: 23.0.  Roth and the geologists are computing the same quantity.  They disagree about one thing only: whether that factor of twenty-three is missing mass or recycled mass.

The evidence that it is recycled is direct rather than theoretical.  About two-thirds of exposed bedrock on the continents is already sedimentary, so a river eroding a random square kilometre is, two times in three, eroding rock that has been through the cycle before.  Detrital zircon grains carry U-Pb ages that identify their source, and the same age populations turn up in formation after formation — the grains are demonstrably second-hand.  And Gregor, the source Roth cites for the twenty-three, concluded the opposite of the use made of him: his own estimate of the pre-human global denudation rate slightly exceeds the modern one, which is the same correction Judson made.

One honest concession, because it is the place the standard picture genuinely fails.  The simple model in which preserved sediment decays away exponentially with age turns out to be false for continental shallow-marine sediment, which is about three-quarters of the surviving volume.  The decay model holds for deep-sea and non-marine sediment and not for the largest category.  That is a real defect in the textbook account, published by mainstream stratigraphers, and it should be reported by anyone using the recycling answer.

Where the argument is strongest

The rate calculation does not work.  The second half of the case is much better, and it is not about rates at all.

It is this: contacts between rock layers that are supposed to be separated by millions of years are often knife-sharp and flat over enormous areas, with no channels cut into them, no soil developed on them, and no weathering.  If erosion happens on the timescale of a human life — and it plainly does — then a surface left exposed for six million years should look like something.  Often it does not.

Fleming's examples, from the second episode: the Hermit Formation and the Coconino Sandstone in the Grand Canyon, separated on the standard timescale by about six million years and meeting at a flat contact; the Moenkopi and Shinarump, separated by fifteen to forty million; another case at fifty million; and the Great Unconformity itself, where roughly 1.2 billion years is missing.⁴  He adds two details that make it worse.  Cobbles at the base of the Shinarump have pressed down into the underlying Moenkopi as load casts, which requires the Moenkopi to have still been soft mud when they landed on it.  And John Whitmore's work at the Hermit-Coconino contact identifies clastic injectites — sand forced down into cracks — which likewise requires unlithified material.

The tradition behind this is older and better documented than the rate argument.  Ariel Roth compiled the cases in 1988,⁹ and before him George McCready Price and Whitcomb and Morris made the same point.  Woodmorappe's version is the sharpest: a Devonian limestone grading imperceptibly into an Ordovician limestone beneath it, with the entire Silurian missing and no physical trace of its passing.

This deserves a straight answer, and this article does not have a complete one.  What it has is five partial ones, and their honest status is worth stating.

First, a flat contact is not the same as a surface that sat exposed.  The claim in the literature is not that a landscape lay unchanged for six million years.  It is that no rock of that age is preserved at that spot — material may have been deposited and removed, or bypassed the site entirely.  The distinction matters and is often lost in the retelling.

Second, Fleming supplies the counter-example himself, names the inference, and answers it — and his answer is where the disagreement actually lies.  Discussing the six-million-year Hermit-Coconino gap, he shows a diagram from his own book explaining how such gaps are constructed, and points out that a thick unit called the Schnebly Hill Formation appears between the same two formations a short drive south, near Sedona:

"The Coconino and the Hermit, for example — if you drive south from Grand Canyon, so you go to kind of the area of Sedona, there's a thick formation between the Hermit Formation and the Coconino called the Schnebly Hill, which must have been deposited in that gap."⁴

His own conclusion from this is that the Schnebly Hill did not take six million years to deposit either, so the gap can be collapsed along with the rest of the column.  He is right about the stratigraphy.  The stratigraphy does not support the conclusion he draws from it.  The Schnebly Hill is about three hundred metres thick 125 km south of the canyon and over six hundred metres in the Holbrook Basin, thinning northward to a zero edge.  Its top contact with the Coconino is gradational and intertonguing, with no major unconformity.⁶⁶  That means the base of the Coconino is diachronous: the Coconino at Grand Canyon is the younger part of a sand sea whose older equivalents lie to the south.  The contact in the canyon is a facies boundary that climbs through time, not a single surface that lay exposed for six million years.

Whitmore, whose Coconino work Fleming cites, concedes the pivot in a single clause: the missing time exists only "assuming the top of the Hermit is isochronous."⁶⁷  Blakey's intertonguing relationship is direct field evidence that it is not.

The six-million-year figure itself is soft.  No primary source for it could be located; Whitmore writes "5–10 Ma?" with a question mark, and the National Park Service describes Permian deposition in the canyon as "nearly continuous."⁶⁸  A number quoted to one significant figure and used to carry an argument should have a citation, and this one does not.

Where the reply does not reach: Grand Canyon lacks the Schnebly Hill entirely, so something there was neither receiving those sediments nor being deeply eroded while hundreds of metres piled up two hundred kilometres away.  A depositional pinch-out on a very low-gradient surface near base level is a coherent account of that, but it is an inference from the geometry rather than an observation of non-erosion.

Third, where a surface really was exposed for a long time, the evidence is often there — and it is in the Grand Canyon.  The Redwall Limestone carries palaeokarst: sinkholes and collapse breccia, formed by rock dissolving subaerially.  The Surprise Canyon Formation fills stream channels cut as much as four hundred feet down into the Redwall.  The Precambrian surface beneath the Grand Canyon Supergroup has up to 240 metres of erosional relief on it, and the Esplanade Sandstone up to 16 metres.⁶⁹  So the prediction — expose a surface long enough and you will find channels, karst, and soils — is met in the same canyon, at other contacts.  What has to be explained is why some contacts show it and others do not, and that is a narrower and more tractable question than the one the argument poses.

Fourth, the injectite argument is real research, and it says less than Fleming says it says.  The work he cites is Whitmore and Strom's study of sand-filled cracks at the base of the Coconino, and it deserves respect on two counts: it is original fieldwork at thirty locations under a Park Service research permit, and it was published in Sedimentary Geology, a mainstream Elsevier journal, not in a house organ.⁷⁰  Its case that these are injectites rather than desiccation cracks is strong — flow banding at hand-lens and thin-section scale, sandstone-clast breccias, cracks up to fifteen metres deep, a depth distribution zoned around the Bright Angel fault, and a preferred crack orientation roughly perpendicular to the fault zones.

But the authors state the difficulty against themselves, in their own words:

"A problem with no readily apparent solution is that it appears the basal Coconino was water saturated and unlithified (or only partially lithified) at the time of Laramide faulting which we suspect caused dike intrusion.  It is unlikely the Coconino could have remained uncemented in excess of 250 million years."⁷⁰

And they then offer two conventional escapes themselves, one of which — earlier movement on the Bright Angel Fault, shortly after the basal Coconino was deposited — is entirely compatible with deep time.  Fleming does not conceal any of this — he lays out the conventional chronology accurately on air, including that the injection most likely happened around the Laramide, tens of millions of years after both formations were deposited.  The disagreement is over the inference, not the reporting.  What can fairly be said is that presenting it as a settled time conundrum states the case more strongly than its own authors do.  Sand injection tens of millions of years after burial is documented worldwide and requires fluid overpressure rather than youthful sediment; what Whitmore and Strom have identified is a real question about cementation timing, not a demonstration.

Fifth, and most important: mainstream stratigraphy agrees that the time is in the contacts, and says so in numbers.  This is the part of the answer most creationist treatments never encounter, and it is not a hand-wave.

Peter Sadler compiled nearly 25,000 measured rates of sediment accumulation in 1981 and found they span at least eleven orders of magnitude, with the average rate falling systematically as the measurement interval lengthens.⁷¹  The reason is that sediment does not accumulate steadily; it accumulates in bursts separated by long periods of nothing.  Working the carbonate-platform data through, Sadler's own conclusion is that the bedding-plane surfaces stratigraphers use to bound metre-scale cycles record eighty to ninety per cent of the time in each cycle.⁷²  The time is not in the rock.  It is in the surfaces between the rock.

A 2018 review in the Annual Review of Earth and Planetary Sciences pushes this much further, with a worked example.  Take a twenty-metre body of cross-bedded fluvial sandstone that represents twenty thousand years on the long-term accumulation rate.  Count the cross-sets: about a hundred, each recording the passage of a single dune face, which at measured sand fluxes takes under four hours.  A hundred sets is therefore under four hundred hours of actual deposition inside twenty thousand years — a temporal completeness of two parts in a million.  The authors' own summary: "the fraction of time actually represented in the strata is so vanishingly small that recorded time is best thought of as a kind of dust."⁷³

So the mainstream position is not that the six million years left a record which creationists have overlooked.  It is that almost all geological time is missing from the rock, everywhere, at every scale, and that this has been quantified for forty-five years.  The missing time, the same review argues, is mostly stasis — surfaces where nothing was happening, neither depositing nor eroding — which is physically the exact condition a flat, unweathered contact records.

This concession is enormous, and a young-earth reader should sit with it before deciding who it helps.  It grants that the rock record is mostly absence of record.  What it does not grant is that the absence can be squeezed to zero.  The same review considers the strong-catastrophist reading — that if completeness is that low, the preserved beds should be the deposits of enormously rare mega-events — and rejects it on the grounds that measured palaeotransport conditions say otherwise: the record is, in their phrase, "strangely ordinary."⁷³  The beds look like ordinary rivers and ordinary dunes doing ordinary things, not like a global cataclysm.

And there are two real limits to how far this cuts, both from inside the same literature.  John Tipper notes that the proportion of stasis in a system cannot be reliably recovered from the strata it leaves behind — an honest admission of a parameter that is not measurable in practice, and one that weakens the mainstream account exactly as much as the young-earth one.⁷⁴  And stasis may not even be needed: Schumer and colleagues showed that a surface performing a random walk reproduces the Sadler scaling on its own, with no quiet intervals required at all.  That is the most deflationary result in the field, and it cuts against the stasis answer this section just gave.

What is left over is real.  Sharp, laterally extensive contacts across long claimed hiatuses are a genuine feature of the record; the explanations are partly inferential; and a reader who finds them unsatisfying is not being unreasonable.  This is the part of the young-earth geological case that is worth working on.  It is also, notably, not an argument about erosion rates — it is an argument about the absence of erosion, which is why the double edge described earlier bites so hard.  Every flat contact that counts as evidence against deep time is also a place where the flood was supposedly moving continent-scale volumes of sediment without leaving a scratch.

Verdict

The arithmetic is right and the conclusion is wrong.

The measured numbers are not in dispute.  Rivers really do deliver enough sediment to level the continents in twelve to twenty-five million years, mainstream textbooks really do say so, and the creationist writers who first published the calculation — Nevins in 1973, Roth in 1986 — did the sums correctly on good data.  That should be conceded without hedging, because it is true and because pretending otherwise makes the rest of the reply look evasive.

The inference fails at three points, in increasing order of importance.  Isostasy multiplies the required erosion by about six, which buys a factor of six and no more.  Relief-dependence makes the decline exponential, which does not help at all over billions of years.  And the continents are actively rebuilt, which is the answer — an answer independently confirmed by walking up to rocks that formed twenty kilometres underground and are now at the surface, and by three unrelated chronometers agreeing on the rates.

Two things ought to be said beyond that.

The popular coastal version should be dropped.  It confuses horizontal retreat with vertical lowering, uses the retreat rate of twenty-thousand-year-old glacial sand as a rate for continental rock, reads the continental shelf backwards, and ignores the fact that modern shorelines are only a few thousand years old — a fact its own author states correctly, on the record, before setting it aside.  A geologist reading it will stop reading.

But one narrow version of it is still open, and pretending otherwise would be the same failure in the other direction.  Even after granting the measured tenfold correction from historic to millennial rates, the best long-term rocky-coast numbers still imply hundreds of kilometres of retreat in ten million years.  Closing that gap requires the true deep-time rate to be another fifty to a hundred times slower again.  Rates that slow have now been measured — but at eleven sites on two quiet coastlines, which is a thin basis for a global claim.  The mainstream account of rocky coasts over a hundred million years currently rests on a correction that is real, is physically expected, and is not yet established at global scale.  That is a legitimate thing for a skeptic of deep time to press on, and it is the only part of the coastal argument that is.

And the argument's own literature has twice contained its refutation.  DeVilbiss said it at the first International Conference on Creationism in 1986.  Woodmorappe said it in the Journal of Creation in December 2024.  Neither was followed up, and the argument has never been placed on any creationist organisation's list of claims to retire — although the precedent for exactly that exists, in the same organisations, in the moon-dust retraction of 1993.

The most useful thing a young-earth reader can take from this subject is not a proof.  It is a standard.  Snelling and Rush spent seven years and forty pages killing an argument they wanted to be true, and wrote that the conclusions were "not as encouraging or complimentary for us young earth creationists as we would have liked."  Answers in Genesis states the same principle in its own words, on its own site, on the page explaining why some creationist arguments have to be dropped:

"We would do a disservice to our witness for Christ by knowingly using bad argumentation—effectively bearing false witness—even if those arguments are used to support the truth of the Bible… We must subject creationist claims to the same high standards of peer review we would demand of evolutionists."⁷⁵

Applied here, that standard retires the erosion clock.  It leaves the flat-contact problem standing, which is the part actually worth the work.

Companion pages

• The literature — every source consulted, catalogued with what it measured, over what timescale, and which way it cuts.
• Natural Arches, Rock Bridges, and the Age of the Earth — the same style of argument at the scale of a single landform.

Sources