How Fast Can a Canyon Be Cut?

"A narrow gorge is sometimes inferred to represent slow persistent erosion, whereas Canyon Lake Gorge was formed in a matter of days." — Michael Lamb and Mark Fonstad, Nature Geoscience, 2010¹

"The amount of erosion demonstrates no relationship with flood magnitude or bedload availability.  Instead, structural pattern of the bedrock through faults and joints appears to be the primary control on landscape change." — Loreto Antón and colleagues on the Ricobayo gorge, Nature Communications, 2015²

"If this much damage resulted from a large but local failed lake flowing over a corner of the continent, what damage could be expected from the great Flood of Noah's day?" — John Morris, Institute for Creation Research, 2011³

Summary

Canyons can be cut in days.  This is measured, published, and not in dispute.  In July 2002 a flood on the Guadalupe River in Texas cut a 2.2 kilometre bedrock gorge through Cretaceous limestone in about three days.  Between 1933 and 1940 an unlined dam spillway in northwest Spain cut 100 metres down into unweathered granite at ordinary flood discharges.  In the Channeled Scabland of eastern Washington the mass balance of the coulees implies waterfalls retreating a kilometre or more a day.  Anyone who says a canyon needs millions of years is arguing against the measurements.

But every one of those cases turns on the same condition, and it is not the size of the flood.  It is whether the rock is already broken into blocks the water can lift.  Where joints are spaced less than about a metre apart, the flood hauls the pieces off like gravel.  Where joints are 3 to 5 metres apart, plucking stops and the flood has to grind, which is slower by orders of magnitude.  Rapid incision is a threshold property of jointed rock, not a property of rock in general and not a property of big floods.

Eight conclusions.

1. A bedrock canyon was cut in three days and it was watched. Canyon Lake Gorge, Texas, July 2002: a flood peaking at 1,461 cubic metres per second "moved metre-sized boulders, excavated ∼7 m of limestone and transformed a soil-mantled valley into a bedrock canyon in just ∼3 days."¹ Three neighbours watched it happen and each put most of the erosion in the first three to four days.¹
2. Morphology does not date a landscape, and the geomorphologists say so. Lamb and Fonstad: "inner channels, knickpoints and terraces are often formed slowly over geologic time … but in Canyon Lake Gorge and other megafloods they must have formed rapidly."¹ Amphitheatre-headed canyons, once read as slow seepage erosion, turn out at Box Canyon and Malad Gorge to be flood-cut.⁴⁵ Steven Austin's observation at Mount St Helens, that a branching drainage network formed in a day, is correct.⁶ This is conceded here.
3. Big floods are not what makes the difference. Joints are. At Ricobayo the gorge grew in granite at discharges of 80 to 1,500 cubic metres per second, about one ten-thousandth of the Missoula peak, and the authors found "no relationship with flood magnitude."² Kelin Whipple and colleagues had already measured the controlling variable: "joint spacing, fractures, and bedding planes exert the most direct control."⁷
4. The same rock cuts fast or slow depending only on whether a threshold is crossed. Malad Gorge, Idaho, has three canyon heads in one basalt flow. Two were cut by a single short flood about 46,000 years ago. The third has the Wood River in it and has been retreating at 2.5 centimetres a year for at least 33,000 years.⁵ Same rock, orders of magnitude between the rates.
5. The Channeled Scabland is the best analogue anyone has, and its measured numbers are smaller than the retellings. Upper Grand Coulee lost 67.8 cubic kilometres of basalt, in about six canyon-cutting floods out of at least 104 releases spread over three to four thousand years.⁸⁹ Every remeasurement since 2016 has made the floods smaller, because the old method assumed the water filled the finished canyon.¹⁰¹¹¹²
6. A genuine megaflood ran down the Grand Canyon and did not carve it. About 165,000 years ago the Hyaloclastite lava dam failed and released 11 billion cubic metres of water in 31 hours, one of the ten largest floods known in North America.¹³ Cassandra Fenton, Robert Webb and Thure Cerling report that such floods are "ineffective agents of geomorphic change," and that the reach they crossed has a lower long-term incision rate than the reach they did not.¹³
7. The Flood argument fails on water-to-rock ratio, not on speed. Michael Oard, a young-earth creationist who has published on the Missoula flood, does the arithmetic himself: about 2,200 cubic kilometres of lake water eroded about 200 cubic kilometres of rock, "about a 10 to 1 ratio of water to erosion."¹⁴ Austin's own figures require about 3 to 1, over harder rock, before any of the surrounding plateau is stripped.¹⁵¹⁵¹⁵
8. Width is the part no fast mechanism reaches. A confined torrent cuts a slot. Spreading the same water across a wide front drops the shear stress below the threshold, and Karin Lehnigk and colleagues measured exactly that contrast in the Scabland.⁸ The Grand Canyon is 6 to 29 kilometres wide, and its cliffs retreat at a few hundred metres per million years.¹⁶

The surprising trend

The modern literature runs the opposite way to what a reader would guess.  Four remeasurements of canyon-cutting floods since 2016 have all come out smaller than the estimates they replaced.  Moses Coulee:  five to ten times smaller.¹⁰  The Columbia River Gorge:  30 to 40 per cent smaller.¹¹  Upper Grand Coulee:  about four and a half times smaller.⁸  Iceland's canyon-carving glacier floods:  "at most a third of previous estimates."¹²  The reason is always the same.  The old reconstruction assumed the flood filled the canyon that exists now.  It did not.  It flowed across a much shallower surface and cut the canyon as it went.

That is bad news for both sides.  It removes the need for an unimaginable discharge, which helps the catastrophist.  It also removes the link between discharge and erosion, which is the link the catastrophist needs when he argues from "the Flood had more water."

Part 1: Four canyons cut fast, and what was measured in each

Canyon Lake Gorge, Texas, July 2002

Canyon Lake is an Army Corps of Engineers reservoir on the Guadalupe River in central Texas.  Its drainage took up to 0.9 metres of rain between 30 June and 6 July 2002.  The lake overtopped an emergency spillway that had never been used, a grassy valley with a short concrete canal at its head.  Twelve people died and damage passed a billion dollars.

Michael Lamb and Mark Fonstad surveyed what the water left.  The gorge is 2.2 kilometres long and, beyond the first 400 metres, "approximately constant at ~50±10 m" wide.  Along the channel floor 2.64 metres of rock came out over the first half-kilometre and 7.2 metres over the next 700 metres, more than 12 metres in places.  Waterfalls up to 9 metres high appeared.  The whole gorge is about 4.6 × 10⁵ cubic metres of rock.

The peak daily-average discharge, on 6 July, was 1,461 cubic metres per second.  Lamb and Fonstad's threshold calculation puts the erosive window at "2.8-3.7 days when the flood exceeded this discharge", which matches what the neighbours said.¹

The mechanism is the whole point.  The bedrock is the Cretaceous Glen Rose Formation, limestone and dolostone and marl, and it is jointed.  Bedding planes give near-horizontal joints with "individual bed thicknesses are 0.7±0.5 m", and the Balcones fault system gives near-vertical ones.  Boulders carried down the gorge "often consist of a single limestone bed."  The flood did not break the rock.  It lifted the pieces out.¹

Abrasion did almost nothing.  Scratch marks on the freshly exposed rock are "etched <1 mm into bedrock surface," and the bedding planes "lack flutes or potholes and some show preserved Cretaceous wave-ripple forms with no significant wear."¹  A flood that stripped seven metres of rock in three days left hundred-million-year-old ripple marks intact on the surfaces it uncovered.  That is the clearest statement anywhere of what plucking is and is not.

Lamb and Fonstad state the limit of their result in their closing sentence:  "it seems plausible that erosion of well-jointed rock by large floods might be extremely rapid, such that canyon formation is limited by the capacity of the flood to transport plucked blocks rather than by the plucking processes itself."¹  Note the condition.  Well-jointed rock.

One more thing about this paper is routinely left out when it is cited.  Its first sentence reads:  "Most bedrock river canyons are thought to be cut slowly over millions of years (for example, Grand Canyon, USA)."¹  The authors name the Grand Canyon as the slow end member in the sentence that introduces their fast result.  Any citation that drops that sentence is quoting the paper against its authors.

Ricobayo, northwest Spain, 1933 to 1940

The Ricobayo dam on the Río Esla was built between 1929 and 1933 with an unlined spillway cut into unweathered granite.  When the reservoir first filled, the overtopping began to scour it.

The first scour event, on 10 January 1934, ran at about 80 cubic metres per second.  The second, on 22 March 1934, ran at about 400 and produced 75 metres of headward retreat.  By that summer there was a 230 metre canyon with 369,300 cubic metres of rock gone.  Over six years the total reached about 1.58 million cubic metres, with more than 270 metres of headward retreat, about 100 metres of incision and about 160 metres of widening.²

Loreto Antón and colleagues state the finding plainly:  "The amount of erosion demonstrates no relationship with flood magnitude or bedload availability.  Instead, structural pattern of the bedrock through faults and joints appears to be the primary control on landscape change.  These discontinuities facilitate rapid erosion … principally through fluvial plucking and block topple."²

This case defeats two objections at once.  It defeats "fast cutting only happens in weak rock," because this is granite.  It defeats "fast cutting needs a megaflood," because 80 to 1,500 cubic metres per second is a routine river flood.  What it does not defeat is the requirement for joints, and that requirement is the same in every study in this literature.

Jökulsárgljúfur, northeast Iceland

The Jökulsá á Fjöllum is a glacial meltwater river cutting a canyon through stacked columnar basalt flows.  Three waterfalls stand in a five-kilometre reach:  Selfoss at 13 metres, Dettifoss at 54, Hafragilsfoss at 20.

Edwin Baynes and colleagues dated flood-stripped bedrock surfaces with cosmogenic helium-3 and found what looks like an on-off switch.  The columns will not topple until the water is deep enough:  "a minimum water depth of 8.1 m would be required to initiate toppling of the basalt columns in the study area, corresponding to a minimum discharge of 3,250 m³/s at Selfoss."  That is "6 times higher than the maximum peak discharge recorded between 1973 and 1979 in this region."¹⁷

Below the threshold the canyon stops deepening.  The everyday river carries about 500 cubic metres per second of sediment-rich meltwater and has not cut into the lava flows:  "the canyon floor is always found to coincide with the top of a lava flow, and there is no evidence for active vertical incision into the lava flows other than the knickpoints themselves."¹⁷

Above the threshold the canyon runs.  In three flood periods, about 9, 5 and 2 thousand years ago, waterfalls more than 13 metres high retreated distances that "could exceed 2 km."  Their comparison:  "The rates of knickpoint retreat during the short-lived Jökulsárgljúfur floods (hundreds of meters in days) are far greater than the highest documented long-term knickpoint retreat rates in other rivers of a similar scale, such as the ~1.5 m/y retreat rate of the Horseshoe Falls at Niagara between 1842 and 1905."¹⁷  Averaged over the Holocene the canyon gets "up to 100 m of vertical erosion over the last 8.5 ka, equivalent to an average vertical incision rate of ~12 mm/y", but the average hides the structure.  Nearly all of it happened in three short bursts.

Willem van der Bilt and colleagues redated the floods in 2021 and made them smaller:  3.5, 1.5 and 1.35 thousand years ago, with peaks "at most a third of previous estimates."  Their conclusion is the one that matters:  "the extent of bedrock erosion is not necessarily controlled by discharge."¹²  The waterfalls still retreated more than two kilometres.

Mount St Helens, the Toutle River, 1980 to 1982

This is the case young-earth writers cite most, and the honest version is Austin's own technical page, not the popular retellings.

Austin describes an eruption on 19 March 1982 that melted the crater snowpack and sent a mudflow north.  It "downcut the rockslide and pyroclastic flow deposits north of the volcano," and produced "A miniature 'Grand Canyon' over 100 feet deep."  His Figure 6 caption says what was cut:  "The flat plain of pumice deposited on May 18, 1980, was eroded to a depth of more than 100 feet by August 1984."⁶

So the material was 22-month-old unconsolidated pumice, ash and landslide debris, and Austin says so himself.  The depth is about 30 metres.  The Grand Canyon is up to about 1,800 metres deep.  An Answers in Genesis restatement gives the eruption date as 19 May and says "Bedrock was eroded up to 600 feet deep," which Austin's account does not support.  The accurate version is the next clause of the same sentence:  "Individual canyons up to 140 feet deep were cut through the landslide debris and volcanic ash deposits."¹⁸  Quote Austin, not the popularisation.

Two things in the Toutle case are real, and both cut in the creationist's favour.  A branching tributary network formed in days.  Amphitheatre-headed side canyons formed in days.  Both shapes are routinely read as slow.  Austin is right that the appearance of a landscape does not date it.

The measured cases side by side

No discharge is printed for the Toutle row because none could be found for the March 1982 mudflow.  The 0.26 × 10⁶ cubic metres per second that the US Geological Survey's world flood compilation lists for the Toutle is a different event, a landslide-dam failure about 2,500 years ago.¹⁹  The Grand Canyon volume in the bottom row is Austin's own figure with no peer-reviewed source behind it;  see the unanswered questions.

Part 2: Morphology cannot date a landscape

This is the strongest thing the catastrophist has, it comes from mainstream sources, and it should be said before anything is argued against him.

Landforms have long been read as clocks.  A narrow gorge with an inner channel, a set of terraces, a knickpoint partway up the profile, an amphitheatre-shaped canyon head, a tree-like network of tributaries:  each has been used as evidence that a landscape took a long time to make.  None of them works.

Lamb and Fonstad say it in Nature Geoscience, about a canyon whose formation was witnessed:

"It is difficult to identify morphologic features in Canyon Lake Gorge that indicate canyon formation during a 3 day event, versus a longer-lived flood or multiple events.  For example, inner channels, knickpoints and terraces are often formed slowly over geologic time in response to shifting climate or tectonic forcing, but in Canyon Lake Gorge and other megafloods they must have formed rapidly through intrinsic instabilities in the erosion processes.  A narrow gorge is sometimes inferred to represent slow persistent erosion, whereas Canyon Lake Gorge was formed in a matter of days."¹

The amphitheatre head is the best documented casualty.  A canyon that ends in a rounded, cliff-walled bowl used to be read as the signature of groundwater seepage undermining the headwall, a slow process by definition.  Box Canyon, Idaho, was the type example:  35 metres deep, 120 metres wide, with the eleventh-largest spring in the United States pouring about 10 cubic metres per second out of its headwall.  Lamb and colleagues showed the boulders on its floor need "Q > 220 m³/s" merely to move, "a factor of 22 greater than the modern spring discharge", and put the canyon-forming flood at 800 to 2,800 cubic metres per second.⁴  The seepage reading was wrong.

The branching network is the other casualty, and here the creationist made the observation first.  Austin's account of the North Fork Toutle after March 1982 records that the mudflow "formed anastomosing channels over much of the hummocky rockslide debris allowing cataracts to erode headward and established for the first time since 1980 a dendritic integration of the North Fork Toutle drainage."⁶  The textbook product of slow landscape maturation appeared in a day.

So the shape of a canyon does not tell you how long it took.  That is granted without hedging.  What it does not do is transfer.  Morphology is a bad clock because fast and slow processes make similar shapes, and that works in both directions.  It does not license the inference that any particular canyon was cut fast.  Lamb and Fonstad apply the point to megaflood reconstruction on Earth and Mars, not to the Colorado Plateau.

Part 3: The threshold, and why it is a property of joints

What controls the rate

Kelin Whipple, Gregory Hancock and Robert Anderson surveyed ten rivers across welded tuff, granite and high-grade metamorphic rock, with joint spacings ranging over two orders of magnitude.  Their conclusion:

"the relative efficacy of the various processes of fluvial erosion … is a strong function of substrate lithology, and that joint spacing, fractures, and bedding planes exert the most direct control.  We find that plucking dominates wherever rocks are well jointed on a submeter scale … In more massive rocks, abrasion by suspended sand appears to be rate limiting."⁷

And they measured where the switch sits.  "Minimum joint spacings observed in abrasion-dominated reaches range from 1-1.5 m in the Sierra Nevada drainages studied up to 3-5 m on the Indus River in Pakistan."⁷  Below about a metre the river plucks.  Above about three metres it has to grind.

The difference is not one of degree.  Plucking means the rock is already cut into pieces and the water only has to carry them.  Whipple and colleagues put it as a change of category:  where the rock is pervasively jointed, channels "tend toward a transport-limited rather than detachment-limited condition."  The river stops being a rock grinder and becomes a gravel truck.

Larsen and Lamb reached the same conclusion from the Scabland side:

"well-jointed rock that is subject to megaflooding may behave more like a bed of sediment with limited or no cohesion in which individual blocks can be easily entrained and transported, rather than as massive crystalline rock that erodes slowly through abrasion."¹⁰

The sensitivity to slope

Erosion models write incision rate as E = K A^m S^n, where A stands in for discharge and S is channel slope.  The exponent n is where the process shows up.  Whipple and colleagues put plucking at n of about 2/3 to 1, suspended-load abrasion at about 5/3, and cavitation as high as 7/3.  A high exponent means a steep, fast reach erodes disproportionately harder than a gentle one, which is why a knickpoint can run away while the reach below it does nothing.

Malad Gorge: the cleanest natural experiment there is

Malad Gorge in southern Idaho has three canyon heads cut into the same basalt.  Two are amphitheatre-headed.  Cosmogenic helium-3 dating puts their formation at about 46,000 years ago in a flood "approximately sevenfold the largest historic discharge of the Wood River (181 m3/s over a 98-y record)."  The third is a 7 per cent knickzone with the Wood River still running through it, and it has been retreating at 2.5 centimetres a year for at least 33,000 years.⁵

Hold those against each other.  Two and a half centimetres a year is 25 metres in a thousand years.  The two amphitheatre heads next door, Woody's Cove and Stubby Canyon, were cut in "a short-duration flood event" and end in 50-metre vertical headwalls.⁵  The rock did not change.  The discharge did, and it crossed a threshold.

This settles the framing of the whole question.  "Can rock be cut fast?" is the wrong question, and both sides have argued it for fifty years.  The right question is whether a given flow, in a given rock, crosses the threshold, and how long it stays across.  Baynes and colleagues measured the same step at Dettifoss:  no column toppling below 8.1 metres of water depth, kilometres of retreat above it.¹⁷

The third variable: confinement

Discharge and joints are two of three.  The third is whether the flow is squeezed.

Lehnigk and colleagues measured this in two Scabland channels that received similar predicted discharges.  Moses Coulee averages 2.1 kilometres of inundated width;  Wilson Creek averages 5.1.  "the shear stresses generated in the 10-km-long reach of Moses Coulee were several times higher than those generated for the equivalent section of Wilson Creek, likely because the inundation in Moses Coulee is narrower."⁸  Moses Coulee lost 14.5 cubic kilometres of rock.  Wilson Creek lost 1.6.

Spreading water out is how you stop it eroding.  That fact carries most of the weight in Part 6.

Cavitation: possible, proposed, and unmeasured

Steven Austin, Edmond Holroyd and David McQueen propose cavitation as the Grand Canyon's mechanism.  Their description of it:

"especially the process of cavitation, whereby fluid vacuum bubbles form in high-velocity floods.  Paul pointed out how shallow, high-velocity flow creates very low-pressure fluid around channel obstructions, creating vapor bubbles within the fluid.  Those bubbles are vitally important because they implode inflicting explosive forces on rock spillways.  The process of cavitation can produce greatly accelerating bedrock erosion on a colossal scale."²⁰

Their real-world case is real.  In June 1983 an emergency release at Glen Canyon Dam sustained 93,000 cubic feet per second through the left spillway tunnel.  "During that emergency release, earthquakes were felt within the dam, and the 'rooster tail sweep' exiting the tunnel turned distinctly red!  After shut-down of the spillway tunnel, engineers found at an elbow within the tunnel a new 63,000-cubic-feet-volume hole penetrating concrete, steel and bedrock."²⁰  Austin's 1994 account adds the rate:  "It is possible that cavitation was pulverizing concrete, steel, and sandstone at a rate in excess of 1,000 cubic feet per minute, during the peak period of erosion."¹⁵

The mainstream literature agrees with him further than he may realise.  Whipple and colleagues argue cavitation is "more likely to occur in natural systems than previously argued" and is possible "For flow depths in excess of 10 m."  Then they add the sentence that decides the matter:  "Direct field evidence of cavitation erosion is, however, lacking."⁷

No field measurement overturning that sentence was found in the literature searched for this article.  The Glen Canyon damage happened inside a smooth-walled engineered tunnel, at engineered velocities, past an elbow that concentrated the effect.  A natural channel does not obviously reproduce any of those conditions.  Cavitation is possible, unmeasured in the field, and carries no evidential weight for either the Grand Canyon's speed or its slowness.

Part 4: The Channeled Scabland, measured

What Bretz found

J Harlen Bretz spent the 1920s arguing, against his whole profession, that eastern Washington had been scoured by a catastrophic flood.  He was right, and he won.  Nothing here disputes that, and the young-earth writers who tell the story tell it accurately.

Bretz's 1923 paper defines the terrain:  "areas where denudation has removed or prevented the accumulation of a mantle of soil, and the underlying rock is exposed or covered largely with its own coarse, angular debris … They are scored by thousands of channels eroded in the underlying rock."²¹

Then he lists what he called universal relations.  Two of them are the diagnostic he used to map the flood:

"11. The pattern of scabland tracts, where hills of the older topography are isolated in them, is anastomosing or 'braided.'  12. Scabland tracts invariably contain 'channels.'  These are gorges or canyons or elongated basins eroded in the basalt … in most cases, the channel pattern is anastomosing or braided."²¹

A few lines earlier, item 9(a), comes the negative:

"The areas surrounded by scabland invariably have the dendritic drainage pattern, mature topography and loessial soil of the plateau."²¹

Read the two together.  Bretz mapped where the flood had been by looking for braided, soil-stripped rock.  He mapped where it had not been by looking for branching drainage on loess.  That contrast is his, and it is the sharpest test available for asking whether any other landscape was flood-cut.

Victor Baker's modern summary lists the same inventory:  the Scabland is "a complex of anastomosing rock-cut fluvial channels, cataracts, loess islands, rock basins, broad gravel deposits, and immense gravel bars."²²

How big the floods were

Jim O'Connor and Victor Baker reconstructed the peak discharge from flood-marked high water:²³

"Within the Spokane Valley, near the point of release, the peak discharge probably exceeded 17 ± 3 million m³·sec⁻¹.  Downstream at Wallula Gap … peak discharge was about 10 ± 2.5 million m³·sec⁻¹.  Flow duration was on the order of several days.  These are the largest known terrestrial fresh-water flows."²³

The lake held 2,184 cubic kilometres of water, and the hydrograph that satisfies both the peak and the volume needs about 70 hours in the Spokane Valley.²³  These are the largest freshwater floods known on Earth.  Nothing in this article shrinks them below that.

How many floods, and how much rock

This is the part the popular retellings omit.

The rhythmite record says there were many floods, not one.  O'Connor and colleagues count them:⁹  "at least 31 rhythmites overlying the tephra were counted … summing to at least 59 floods.  Adding these 31 post-Set-S beds in the Pasco Basin to the 72 preceding Set S at Manila Creek indicates at least 103 floods … at least 104, possibly 109 floods."  They add that "Most or all of these records are likely incomplete so the actual flood total is likely greater," and that the lake was "extant and producing floods for at least 3-4 ky during 20-14 ka."⁹

Karin Lehnigk, Isaac Larsen, Michael Lamb and Scott David then measured how much rock came out and how many floods it took.  They reconstructed the pre-flood land surface from hanging tributaries still graded to the old base level, and differenced it against modern topography.⁸

Their headline:  "rates based on these discharges imply that the largest canyons could have formed in only about six or fewer floods … Overall, our results indicate that a small number of outburst floods, with discharges much lower than commonly assumed, can cause extensive erosion and canyon formation in fractured bedrock."⁸

So about a hundred floods were released, and perhaps six of them cut the big canyons.  Lehnigk and colleagues, citing Brian Atwater's 1986 stratigraphy, count 89 floods through upper Grand Coulee whose deposits post-date the canyon's incision.  Against a total of 104 or 108, that leaves at most 15 to 19 that could have preceded it, and the mass balance needs six.⁸

How fast the cutting itself went

Divide the eroded volume by the number of floods and the hours of above-threshold flow in each, and the retreat rate comes out enormous.  Lehnigk and colleagues get 2,280 to 2,440 metres per day for upper Grand Coulee and 2,320 to 3,430 for Moses Coulee:  "If all the erosion in Grand and Moses coulees occurred by headward waterfall or cataract retreat, we predict retreat rates on the order of a kilometer or more per day."⁸

A kilometre a day.  That number is real and it belongs in any honest account of what water can do.

Why the basalt was cheap to cut

John Morris's short ICR piece on the Scablands is accurate about Bretz, and right that the coulees are cut into hard basalt.  What it does not mention is why that basalt was cheap.³

Columbia River flood basalt is columnar-jointed.  It cools into hexagonal columns and arrives at the flood already cut into blocks measured in tens of centimetres.  Larsen and Lamb measured the basalt clasts in the flood deposits at 0.13 to 0.83 metres across, and put the shear stress needed to pluck them at 117 to 1,242 pascals.¹⁰  Above the basalt lies loess, wind-blown silt, which moves at a small fraction of that.  Stripping loess and plucking pre-cut columns is the cheapest erosion job in geology.  That is not a reason to dismiss the Scabland.  It is the reason the Scabland is fast, and it is the same reason Canyon Lake Gorge, Ricobayo, Dettifoss and Malad Gorge are fast.  Five field studies, four rock types, one answer.

The floods keep shrinking

Four remeasurements since 2016 have all brought the discharges down, for the same reason each time.

• Moses Coulee. Larsen and Lamb ran hydraulic simulations over four reconstructed stages of the canyon's own incision¹⁰ and found that "accounting for erosion thresholds during canyon incision results in near-constant discharges that are five- to ten-fold smaller than full-to-the-brim estimates."¹⁰ The brim-full method predicted 3.0 million cubic metres per second; the threshold method needs at most 0.6 million.
• The Columbia River Gorge. Scott David and colleagues corrected for the floods widening the gorge they are reconstructed from, and got peaks 30 to 40 per cent lower.¹¹
• Upper Grand Coulee. The high-water marks demand 14 million cubic metres per second over today's topography and 3.1 million over the reconstructed pre-flood surface.⁸
• Iceland. Van der Bilt and colleagues cut the Jökulsá á Fjöllum peaks to "at most a third of previous estimates."¹²

The old reconstructions assumed the flood filled the canyon that exists now.  It ran across a shallower surface and cut the canyon as it went.  Larsen and Lamb's threshold argument explains why the shear stress ends up only slightly above the plucking threshold:  "Such bedrock channels cannot withstand large discharges or shear stresses that greatly exceed the threshold for plucking."¹⁰

Both sides should notice what this does.  It weakens the mainstream reflex that only a colossal flood could do this.  It also severs the link between discharge and erosion, which is the link the Flood argument needs when it says the Flood had more water than Missoula.

Part 5: A megaflood ran down the Grand Canyon and did not carve it

The Grand Canyon has hosted real megafloods.  Lava dams built across the western canyon repeatedly during the last 630,000 years, impounded the Colorado, and failed.

The largest is dated to about 165,000 years ago.  Cassandra Fenton, Robert Webb and Thure Cerling reconstructed it:

"The Hyaloclastite Dam was up to 366 m high … Failure of the Hyaloclastite Dam released a maximum 11 × 10⁹ m³ of water in 31 h.  Peak discharges … ranged from 2.3 to 5.3 × 10⁵ m³ s⁻¹ … This discharge is an order of magnitude greater than the largest known discharge on the Colorado River (1.4 × 10⁴ m³ s⁻¹) … and in the top ten largest floods in North America."¹³

It left evidence, and geologists know what that evidence looks like.  The flood's own deposits are 20 to 110 metres thick, "greater than 82% basalt", and they "decrease in elevation from 200 to 13 m above modern river level with distance downstream from the damsite," traceable 32 kilometres down the canyon.¹³  Across all five failed dams the same group traced flood benches as far as 53 kilometres downstream.²⁴

And here is the result that matters.

"Floods with large peak stream power, but of short duration and low total energy expenditure are ineffective agents of geomorphic change in alluvial or bedrock channels … lower incision rates are reported … in western Grand Canyon relative to eastern Grand Canyon for the past 630,000 yr, the time during which lava dams periodically existed in the western canyon.  The difference in incision rates may reflect the ineffective erosive power of lava-dam outburst floods in western Grand Canyon."¹³

The measured contrast is stark.  Karl Karlstrom and colleagues, working from cave deposits,²⁵ get 50 to 80 metres per million years in the western canyon against 175 to 250 in the east.²⁵  The reach that received at least five genuine megafloods has been cutting more slowly than the reach that did not.

Three things follow.

1. A megaflood in this canyon deposits more than it excavates. Fenton and colleagues suggest the western reach may have been slowed by "overloading of the river system with volcanic debris," which is the transport-limited condition Whipple predicted. Once the bed is armoured with plucked blocks, the river spends its energy hauling gravel instead of cutting rock.
2. The signature is findable. Basalt-rich flood benches 200 metres up the walls, traceable tens of kilometres downstream, chemically fingerprinted to a named dam.²⁴¹³ That is what a real megaflood in the Grand Canyon leaves behind.
3. The comparison does not depend on the absolute dates. The western and eastern incision rates are measured on the same clock by the same method. A reader who rejects the timescale still has to explain why the flooded reach is the shallower-cutting one.

Part 6: What the Flood argument needs

Water against rock

The most useful arithmetic on this question was done by a young-earth creationist, and it should be quoted from him rather than from a critic.

Michael Oard is a retired National Weather Service meteorologist with a published monograph on the Missoula flood.  Reviewing Walt Brown's model, he wrote:¹⁴

"the total volume of rock eroded by sheet flow would be 8,500 km³ (2,000 mi³).  The volume of Brown's two lakes east of Grand Canyon was only about 12,000 km³ (2,930 mi³).  That is a little more than a cubic km of water to erode a cubic km of sediment, which seems like too little water, just to erode the 300 m (1,000 ft) of sediments before Grand Canyon was even carved!"

And he supplies the benchmark:  in the Missoula flood "that much water only eroded about 200 km³ (50 mi³) of soft silt and hard basalt along the flood path.  That is about a 10 to 1 ratio of water to erosion."¹⁴

Ten cubic kilometres of water for one of rock, in the most favourable material on the planet, in the largest freshwater floods known.  That is the observed exchange rate.

Now run the same test on Austin's model, using only Austin's own numbers from one chapter.¹⁵  He gives the Missoula flood "up to 500 cubic miles of water (half the present volume of Lake Michigan) … Fifty cubic miles of sediment and bedrock were eroded," the same 10 to 1.¹⁵  He gives his own breached lakes 3,000 cubic miles of water,¹⁵ and the canyon "almost 1,000 cubic miles (4,000 cubic kilometers) of sediment."¹⁵  That is 3 to 1.

Austin's flood must therefore be about three times more erosive per unit of water than the Missoula flood, while flowing over layered limestone, sandstone and shale and through 1,750 metres of schist and granite in the Inner Gorge instead of over columnar basalt and loess.  And it must do all that before stripping the surrounding plateaus, which Austin himself says amounted to "not just 1,000 cubic miles of canyon erosion, but many times that volume."¹⁵

Nothing in this arithmetic depends on the age of the earth.  Both ratios are volumes divided by volumes.

Width

Depth is what people photograph.  Width is the hard problem, and it is where the physics from Part 3 lands.

The Grand Canyon is 6 to 29 kilometres from rim to rim.¹⁵  A confined torrent cuts a slot.  It does not cut a 16-kilometre-wide stepped valley, because spreading the same water across a wide front drops the shear stress below the threshold for moving anything.  Lehnigk and colleagues measured the effect:  the narrower channel, 2.1 kilometres against 5.1, removed nine times as much rock.⁸

So the width has to come from cliffs falling back, over and over, as the slot deepens.  That rate has been measured.

Kenneth Cole and Larry Mayer worked it out from packrat nests.  Packrats build debris piles in caves and cement them with urine into a hard mass that survives for tens of thousands of years if rain never reaches it.  Fresh nests sit near the cave mouth, where a packrat is safest from predators.  As the cliff face falls back, the mouth eats into the cave and destroys the nests nearest it.  So the oldest surviving nest in a cave should be the deepest one, and the slope of oldest-nest depth against radiocarbon age is the rate the cliff is retreating.  Cole and Mayer ran that regression on 14 radiocarbon-dated middens from caves in the Redwall Limestone.  The slope is 0.45 metres per 1,000 radiocarbon years, with a 95 per cent confidence interval of 0.18 to 0.72, which is 180 to 720 metres per million years over the last 13,000 years.¹⁶

Two other measurements agree, at 600 to 800 metres per million years over 500,000 years and 500 to 600 over 19 million years for the Kaibab cliff.²⁶²⁶  But those two assume the mainstream timescale, and a reader who rejects it is entitled to set them aside.  The argument here rests on the packrat figure alone, because it needs nothing but 14 radiocarbon dates inside the last 13,000 years and a tape measure.

That figure was contested at once.  Three Comments followed in Geology in 1983.¹⁶  Louise Hose showed that middens sit far deeper in caves than Cole and Mayer's illustration assumed.  Haman argued the sample is biased low, because all 14 sites are on cliff salients while "the cliff retreat at embayments is four to five times more than salients."  Bernard Smith made the catastrophist's objection:  erosion is "accomplished primarily by high-magnitude, low-frequency events," so an average rate describes nothing real.  Cole and Mayer's reply:  "An average rate does not mean a constant rate," and seven separate caves were used so that one catastrophic site could not drive the result.

Take the criticisms at full strength and the arithmetic still does not close.  Work it per wall, since a slot widening into a canyon retreats both sides.  The canyon is 6,000 metres wide at its narrowest, so each wall has to move 3,000 metres, and 14,500 at the widest.

At 0.45 metres per thousand years, 4,500 years of retreat is 2 metres, short by a factor of 1,500 at the narrowest point.  Grant Haman his whole factor of five and it is 10 metres, short by 300.  Now also assume radiocarbon runs ten times fast inside this window, a larger correction than any young-earth calibration proposes, and it is 100 metres, still short by 30 at the narrowest point and 145 at the widest.

That is the honest size of the gap:  not three orders of magnitude once the objections are granted, but not closeable by adjusting this rate either.  Only a different mechanism would close it, and none has been proposed.

The two creationist models, and their disagreement

There is no single young-earth account of the Grand Canyon, and the difference bears on rates.

Austin, Holroyd and McQueen hold to a post-Flood lake breach.  Their 2020 restatement reports that Austin "reported evidence that Grand Canyon was eroded in just a few weeks just thousands of years ago by catastrophic drainage of lakes," with cavitation as the mechanism.²⁰  Holroyd's version:  "Perhaps the catastrophic release of three Great Lakes-worth of water carved the Grand Canyon in a few weeks by means of cavitation process rather than in millions of years."²⁰  Walt Brown's independent version²⁷ says the lakes "breached their boundaries and carved the Grand Canyon in weeks."²⁷

Oard rejects all of that, and calls two of its problems fatal:  the missing lake sediments and shorelines, and the long tributary canyons.²⁸²⁹  Kanab Canyon "enters Grand Canyon as a narrow gorge about one mile (1.6 km) deep and a quarter mile (0.4 km) wide … Havasu Canyon is of similar dimensions," and because they descend to the exact depth of the main canyon from basins 80 and 97 kilometres away, "they could not have formed before Grand Canyon.  They had to form simultaneously with Grand Canyon."²⁸

His alternative is late-Flood channelized erosion:  "It is a straight forward deduction that the Grand Canyon was carved by late Flood channelized erosion.  The Grand Canyon is a long water gap."²⁹  He first requires the whole region to be stripped of thousands of feet of sediment by an eastward sheet flow, then requires the current to reverse and channelize westward.³⁰³⁰

Two creationists, two incompatible models, each calling the other's fatal.  Neither has published a discharge, a duration and a sediment budget together.

Objections and responses

"The Flood had far more water than Lake Missoula, so the Missoula numbers are a floor, not a ceiling."  Total water is not the variable.  Three things are:  whether the rock is pre-jointed, how long the flow stays above the entrainment threshold, and how confined it is.  More water spread over a wider front erodes less, and that is measured:  Moses Coulee and Wilson Creek got comparable discharges, and the narrower one lost nine times the rock.⁸  Van der Bilt and colleagues put the general form plainly:  "the extent of bedrock erosion is not necessarily controlled by discharge."¹²  And the exchange rate that matters, water to rock, is Oard's own:  10 to 1 in the best material available, against the 3 to 1 the breached-dam model needs in worse material.¹⁴

"The rock was still soft.  Flood strata deposited months earlier would cut like mud."  This is the strongest form of the reply.  Three answers.  First, softness is not needed for speed:  Ricobayo cut a hundred metres into unweathered granite.²  Second, soft rock does not hold the shapes the canyon has.  Oard's own Kanab Canyon is "about one mile (1.6 km) deep and a quarter mile (0.4 km) wide."²⁸  A four-to-one depth-to-width ratio in unlithified sediment is a slump, not a landform, and the canyon's cliff-and-bench staircase tracks the strength of individual formations, which is what differential erosion of lithified rock produces.  Third, softness makes the volume problem worse.  Weak material should have been swept clear across the whole plateau, leaving a broad braided scabland rather than one trunk canyon with tributaries graded to it.

"The 165,000-year date on the lava-dam flood assumes the mainstream timescale."  The date is not doing the work.  A ratio measured on one clock is:  the western reach that took at least five megafloods is cutting slower than the eastern reach that took none.¹³²⁵  Rescale the clock however you like and the flooded reach is still the shallower-cutting one.  And the physical deposits are observations, not dates:  basalt-rich flood benches 200 metres above the river, traceable 53 kilometres downstream, chemically matched to identified dams.²⁴¹³

"Morphology cannot date a landscape.  You conceded it.  So the Scabland comparison proves nothing either."  Correct that it proves nothing about time.  The argument is narrower.  Bretz mapped the flood's edge by the transition from braided-and-scoured rock to branching drainage on loess.²¹²¹  If a flood many times larger crossed the Colorado Plateau, the burden is on that model to point to the marks the smaller one left everywhere it went, because a search of the literature for this article found none reported there.  That is an argument from a missing observation, and it can be answered by producing the observation.

"Cavitation could do what plucking cannot."  Possibly.  Nobody knows.  Whipple and colleagues grant that cavitation is plausible at flow depths over 10 metres and then record that "Direct field evidence of cavitation erosion is, however, lacking."⁷  The one well-documented case of large-scale cavitation damage is the 1983 Glen Canyon spillway tunnel, an engineered conduit with an elbow that concentrated the effect.²⁰  Until somebody measures cavitation eroding a natural channel, it carries no weight for either side.

"Mount St Helens cut 600 feet of bedrock in a day."  It did not, and the correction comes from Austin's own technical page.  What was cut was 22-month-old pumice, ash and landslide debris, to a depth of "more than 100 feet."⁶  The 600-foot bedrock figure appears in a popular retelling that also misdates the eruption by two months.¹⁸  The real Toutle result, that a branching network and amphitheatre heads formed in days, is a better argument than the exaggeration.

"Mainstream geology denies rapid erosion, so any admission of it is a retreat."  It has not since Bretz won the argument.  John Morris's own ICR account puts it at 1960:  "By 1960, Bretz had won the day and convinced geologists that, at least on this occasion, catastrophic processes had ruled."³  Every measurement in this article comes from mainstream geomorphologists in Nature, Science, PNAS, Nature Geoscience, Nature Communications and the GSA Bulletin.  Telling a young-earth audience that secular geologists reject fast canyon cutting is misinforming them, and so is telling a skeptical audience that fast canyon cutting is a creationist invention.

"Canyon Lake Gorge shows a canyon in three days, therefore the Grand Canyon in weeks."  Canyon Lake Gorge removed 4.6 × 10⁵ cubic metres.  The Grand Canyon is of order 4 × 10¹² cubic metres, a factor of roughly ten million.  Upper Grand Coulee, the largest canyon cut by the largest freshwater floods in Earth's known history, is about 1.7 per cent of the Grand Canyon's volume.⁸

Issues and unanswered questions

No peer-reviewed figure for the Grand Canyon's excavated volume could be found.  The "almost 1,000 cubic miles (4,000 cubic kilometers)" repeated by Answers in Genesis, Oard and Clarey traces to one sentence in Austin's 1994 book.¹⁵  A geometric estimate, 446 kilometres long by 16 kilometres mean width by 1.2 kilometres mean depth with a V-shaped section, gives about 4,280 cubic kilometres, with a range of 3,400 to 5,100.  The order of magnitude is safe.  The number is not a measurement, and this article does not treat it as one.

Whether Scabland-type features exist on the plateau surfaces around the Grand Canyon is unresolved here.  Giant current ripples, dry cataracts, streamlined islands and braided channel patterns are the marks of a megaflood, and no published report of them on the Kaibab Plateau or the Marble Platform was found in the literature searched.  That is a negative from a targeted search, not a proof of absence.  A single published map or field description would overturn it.

Oard's 2004 monograph on the Missoula flood has not been read.  The Missoula Flood Controversy and the Genesis Flood is print-only.  Its central claim, that there was one Missoula flood rather than dozens, is contradicted by the rhythmite record and by Lehnigk's mass balance, but it should not be characterised in detail from second-hand accounts.

Two abrasion papers could not be obtained.  Sklar and Dietrich 2004 and Chatanantavet and Parker 2009 refused every retrieval route tried.  Their content reaches this article only through Whipple 2000, Lamb 2008 and Beer and Lamb 2021.

No copy of Lamb and Fonstad 2010 with the journal's own pagination could be found.  The article works from the authors' open advance copy, which paginates 1 to 5.  Each locator gives the advance page and the published page it maps to, and the mapping is an inference.

Three source attributions in earlier drafts were wrong, and are recorded here rather than quietly fixed.  The 200-metre and 53-kilometre flood-bench figures were attributed to Fenton, Webb and Cerling 2006;  they are in Fenton et al. 2004.  Oard's line about "about a dozen serious geological problems" was attributed to his 2010 CRSQ part II;  it is in his later chapter 38.  A discharge of 0.26 million cubic metres per second was listed as the 1980 Toutle event;  the U.S. Geological Survey dates that flood to about 2,500 years ago, and it has been removed.

The Glen Canyon cavitation numbers have not been checked against the engineering source.  Falvey's Cavitation in Chutes and Spillways, Bureau of Reclamation Engineering Monograph 42, was not read.  The tunnel diameter, velocity and damage geometry come from Austin's accounts.

Verdict

Canyons can be cut in days.  That is measured, repeatedly, in four rock types on three continents, and anyone still arguing that a bedrock gorge requires geological time is arguing against Nature Geoscience.  Morphology does not date a landscape, and the inference from an amphitheatre head or a branching network to a long history is dead.  Both points belong to the catastrophist and are conceded without reservation.

What the measurements also show is that fast cutting is a threshold property of jointed rock.  It happens when a flow crosses the entrainment threshold for blocks the rock has already been cut into, and it stops when the flow drops below it, in the same rock, in the same year.  Malad Gorge holds both rates side by side.  Discharge is not the controlling variable;  joints, threshold and confinement are.

Applied to the Grand Canyon, the objections that survive are not about rate.  They are about volume, about the water-to-rock ratio that Oard himself computes at 10 to 1 against the 3 to 1 the model needs, about 1,750 metres of crystalline basement that no documented rapid case resembles, about a 6-to-29-kilometre width that confinement physics says a single flood cannot produce, about tributaries graded to the trunk from 90 kilometres away, about a missing flood signature on the plateau, and about the measured fact that five real megafloods inside this canyon left it cutting slower rather than faster.

The short answer to the title question:  as fast as the joints allow, for as long as the water stays above threshold, in a channel narrow enough to keep the stress up.  That can be three days.  It has never been shown to be four thousand cubic kilometres.

Sources