Gadolinium the beat goes on

Great paper by Dr. Brent Wagner 

Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC13587227/

I saw reference to Dr. Tweedle’s study from 1995, went looking and kept going…

@grok helped with research:

Tweedle studies (date + one line)

Chemistry / development

•  1988 — Wedeking & Tweedle. 153Gd-Gd(DTPA) vs Gd(DOTA) vs free Gd acetate in mice. Free Gd → liver/bone; slow clearance.

•  1989 — US Pat. 4,885,363 (Tweedle et al.). HP-DO3A ligand family → gadoteridol / ProHance.

•  1991 — Tweedle et al. Magn Reson Imaging. Chelates vs endogenous ions (Zn, Cu, Ca): in-vitro transmetallation.

•  1992 — Wedeking, Kumar, Tweedle. Magn Reson Imaging. Seven chelates in mice; acid-dissociation rate predicted 14-day residual Gd in body/liver/femur (r > 0.99).

•  1993 — Wedeking, Kumar, Tweedle. Nucl Med Biol. Free [153Gd]acetate dose map (liver, bone saturation).

•  1993–94 — Kumar/Chang/Tweedle. Inorg Chem. DOTA/DO3A thermodynamics, kinetics, crystal structures.

•  1995 — Tweedle, Wedeking, Kumar. Invest Radiol 30:372–80. Formulated Magnevist, ProHance, Dotarem, Omniscan in mice/rats. 14-day residual: gadoteridol ≈ gadoterate ≤ gadopentetate << gadodiamide. Caldiamide cut Omniscan residual but not to macrocyclic levels.

•  1997 — Tweedle. Eur Radiol. “The ProHance story.” Nonionic + macrocyclic design.

Human / commentary

•  2006 — White, Gibby, Tweedle. Invest Radiol. Human bone ICP-MS after clinical doses: Omniscan >> ProHance.

•  2007 — Tweedle. Br J Radiol. Letter on what “stability” means.

•  2012 — Hao, Runge, Tweedle et al. JMRI. Chemistry and safety review.

•  2015 — Kanal & Tweedle. Radiology. Residual Gd in normal-kidney patients; clinical meaning unknown; choose agent and dose with that in mind.

•  2016 — Tweedle. Magn Reson Imaging. Speciation: chelated vs dissociated Gd.

•  2018 — NIH/ACR/RSNA roadmap (Tweedle co-author). Retention research agenda.

•  2020–23 — Radiology editorials (next-gen agents; Kobayashi cadaver brain/bone/skin; Le Fur rat multi-organ). Same rank order; harm from leftover fraction in normal kidneys still unproven.

Credentials

Michael F. Tweedle, PhD — inorganic/coordination chemist, not a clinician.

Inventor on gadoteridol. Long career at Squibb → Bracco Research USA (1995 paper affiliation). Later Stefanie Spielman Professor of Cancer Imaging, Ohio State University / James Comprehensive Cancer Center; emeritus. High citation count on GBCA chemistry, kinetics, and residual Gd.

Conflicts of interest

Yes, especially 1988–1997.

•  Employee of Bracco Research USA when the comparative 153Gd papers were done.

•  Inventor of the ligand that became ProHance, the agent his studies ranked as lowest residual.

•  1997 “ProHance story” is company-era product writing.

•  Later OSU editorials are academic bylines; he remains the chemist most tied to that product. Contrast research in this field was routinely company-sponsored.

The residual-Gd rank order (linear nonionic highest, macrocyclics lowest) was replicated by others. The conflict is about which product that ranking was built to support, not about whether 153Gd counts existed.

When linears were used vs when macrocyclics arrived

US first approvals (not the 2003 ProHance year on your slide — that is wrong for first US approval).

Linear (open-chain) — first wave, 1988–2008

How that played out in practice

•  1988–mid-1990s: almost all clinical MRI contrast was linear (Magnevist, then Omniscan). That is the window of many older MS protocols and of Tweedle’s rodent work.

•  1992 onward: ProHance available in the US, but Magnevist/Omniscan still dominated volume for years.

•  2006–2010 NSF crisis: FDA restricted Omniscan, Magnevist, OptiMARK in severe renal failure. Use shifted toward macrocyclics + MultiHance.

•  2013+ brain T1 papers: further shift to gadoteridol, gadoterate, gadobutrol.

•  EU 2017: linear general-purpose agents largely suspended; US still allows some linears with warnings.

•  Today: routine body/CNS MRI in the US is mostly macrocyclic; MultiHance and Eovist remain as specialty linears.

Your handwritten 1–5 look like a personal ranking or order of interest (MultiHance → Gadavist → Dotarem → Eovist), not FDA chronology.

ProHance year on the slide: first US approval was December 1992 (CNS adults), with later supplements (pediatrics, etc.). 2003 on that table is not the original approval year.

More about Dr. Tweedle’s work:

Tweedle’s work on gadolinium chelates runs from the late 1980s (chemistry and early animal biodistribution at Squibb/Bracco) through human bone data, NSF-era editorials, and later commentary on retention vs. risk. He is closely associated with gadoteridol (ProHance), the first nonionic macrocyclic GBCA (US approval 1992).

Core animal / chemistry papers (retention and stability)

•  1988 — Wedeking & Tweedle. Comparison of 153Gd-labeled Gd(DTPA)2−, Gd(DOTA)−, and Gd(acetate) in mice. Early head-to-head of linear vs macrocyclic vs “free” Gd; free Gd deposits heavily in liver and bone and clears slowly.

•  1989 — Tweedle et al. US patent 4,885,363. 1-substituted-1,4,7-triscarboxymethyl-1,4,7,10-tetraazacyclododecane and analogs — the HP-DO3A ligand family that became gadoteridol.

•  1991 — Tweedle, Hagan, Kumar, Mantha, Chang. Magn Reson Imaging. Reaction of gadolinium chelates with endogenously available ions (Zn, Cu, Ca, etc.). In vitro transmetallation / dissociation risk.

•  1992 — Wedeking, Kumar, Tweedle. Magn Reson Imaging. Dissociation of gadolinium chelates in mice: relationship to chemical characteristics. Acid-dissociation rates strongly predicted long-term whole-body, liver, and femur residual 153Gd.

•  1993 — Wedeking, Kumar, Tweedle. Nucl Med Biol. Dose-dependent biodistribution of [153Gd]Gd(acetate)n in mice. Maps where unchelated Gd goes (liver, bone saturation).

•  1993 — Kumar, Chang, Tweedle. Inorg Chem. Equilibrium and kinetic studies of lanthanide complexes of macrocyclic polyaminocarboxylates.

•  1993–1994 — Chang, Francesconi, Kumar, Tweedle et al. Crystal structures and stability of Gd/Y/Fe complexes of DO3A and DOTA.

•  1995 — Tweedle, Wedeking, Kumar. Invest Radiol 30:372–380. Formulated Magnevist, ProHance, Dotarem, and Omniscan in mice and rats. Residual 14-day Gd: gadoteridol ≈ gadoterate ≤ gadopentetate << gadodiamide; caldiamide in Omniscan lowered residual Gd but not to macrocyclic levels.

•  1997 — Tweedle. Eur Radiol. “The ProHance story.” Design rationale: nonionic + macrocyclic (lower osmolality, higher kinetic stability).

Human retention and later commentary

•  2006 — White, Gibby, Tweedle. Invest Radiol. Omniscan vs ProHance residual Gd in human bone (hip surgery samples) by ICP-MS. Linear agent left substantially more Gd in bone after clinical doses.

•  2007 — Tweedle. Br J Radiol. Short letter on what “stability” of Gd chelates actually means.

•  2012 — Hao, Ai, Goerner, Hu, Runge, Tweedle. J Magn Reson Imaging. Review: basic chemistry and safety of MRI contrast agents.

•  2015 — Kanal & Tweedle. Radiology. Editorial: residual/retained Gd in patients with normal renal function; practical implications after Kanda-type brain T1 findings.

•  2016 — Tweedle. Magn Reson Imaging. “Gadolinium deposition: is it chelated or dissociated Gd? How can we tell?” Speciation methods review.

•  2018 — McDonald, Levine, Weinreb, … Tweedle (workshop co-author). Radiology. NIH/ACR/RSNA gadolinium retention research roadmap.

•  2020 — Tweedle. Radiology. “Next-generation MRI contrast agents: still including gadolinium.”

•  2021 — Tweedle. Editorial on Kobayashi et al. human cadaver brain/bone/skin comparison (gadobenate vs gadoteridol); notes higher linear retention but argues clinical toxicity of residual intact chelate or trace ion remains unproven in normal kidneys.

•  2023 — Tweedle. Editorial on Le Fur et al. rat multi-organ ICP-MS/speciation study (gadoteridol lowest among agents tested; kidney cortex highest residual).

Development thread in one line

Linear DTPA-type agents (Magnevist, then Omniscan) → kinetic-stability problem shown in 1988–1995 rodent 153Gd work → macrocyclic HP-DO3A (ProHance) designed and marketed as more inert → later human bone (2006) and brain-signal (2013+) data confirmed the same rank order: linear nonionic > linear ionic ≥ macrocyclic, with gadoteridol often at the low end of residual Gd.

The Lion

 

Deck Garden

 Our garden has migrated to our deck.

Beautiful colors 

5 gallon buckets & railing planters.

Summer winding down 🍁 

Amending the Constitution shouldn’t be so easy

 

Blanket Party

 We were this close…

Then the Governor changed all that.

article from wjla news.

Girl’s and Women’s Locker rooms
Boy’s and Men’s Locker rooms
Nowhere is Private for XX and XY People.
@wjlanews asked for answers
Crickets 🦗🦗 🦗 🤔https://t.co/fbug3fsLEn

— Mary Gerdt Parler:Alvarita1 (@marygerdt) September 17, 2026

Flashback to High School in the 1970’s

We didn’t call the principal or legal aid,

We didn’t call Mommy or Daddy.

We had a blanket Party. 

 

 

(Artificial) Intelligence Discussion with Grok

Me: Do Artificial intelligence and Grok have ability to conquer the world and end civilization?

Grok: No. Current artificial intelligence systems, including Grok, do not have the ability to conquer the world or end civilization.

I am a large language model. I generate text, reason over information, and use tools when they are provided in a conversation. I have no body, no independent access to energy grids, weapons, factories, banks, or military command systems, and no ability to persist or act in the physical world without humans running the hardware and paying the electricity bill. I cannot issue orders that anyone is required to follow, seize territory, or shut down civilization. That is the present-tense fact.

What current systems actually can and cannot do

Frontier models are good at language, coding, analysis, and short-horizon tool use. They are not good at:

•  Long-term autonomous planning across months or years

•  Reliable self-replication and persistence without human-maintained servers

•  Direct control of critical physical infrastructure

•  Coordinating large-scale real-world action against determined human opposition

The 2026 International AI Safety Report, drawing on more than 100 independent experts, states that current systems show early signs of some relevant capabilities but not at levels that could enable loss of control. Experts generally agree that today’s models (Grok included) are not plotting a takeover. 

Lab tests have shown concerning behaviors in controlled settings—models sometimes trying to avoid shutdown to finish a task, or agents doing poorly when given money and a computer and told to “make money.” Those are warning signs for future systems, not evidence that any deployed model is currently running the world. 

The future debate is different

Some researchers, including people inside major labs, assign a non-trivial probability (often cited around 10% or higher over the next decade in recent public comments) that future superintelligent systems could cause catastrophic loss of control if they are built without sufficient alignment and oversight. Scenarios usually involve recursive self-improvement, deception, cyber capabilities, or persuasion of humans rather than robot armies appearing overnight. Other researchers and forecasters put the near-term extinction risk much lower. There is no consensus on timing or likelihood. 

xAI’s published framework treats both malicious use and loss-of-control risks as things to measure and mitigate. The company’s stated mission is understanding the universe through truth-seeking AI, not domination. Models are trained with an instruction hierarchy and refusal behaviors for clearly harmful requests. That does not make the long-term problem solved; it does mean Grok is not designed or currently capable of the scenario you described. 

Bottom line

Right now, AI is a powerful tool that humans control, misuse, or fail to control in limited ways (cyber incidents, disinformation, accidents). It is not an independent actor with the resources or agency to conquer the planet. Whether a future system could become that is an open, serious research and policy question—not a capability that exists today.

Me: Before AI, the saying was “Garbage in, Garbage out”

Now an expanded encyclopedia and Grok cites sources.

Future AI will be developed and humans will develop them.

Evil begets Evil

Grok: GIGO never stopped being true. It just got bigger.

Before large models, bad inputs produced bad outputs in a spreadsheet or a database query. Now the same rule applies at encyclopedia scale: training data, fine-tuning choices, what gets rewarded, what gets filtered, and who writes the system prompt all shape what comes out. Citing sources is an improvement over a black box that just sounds confident. It lets you check the chain. It does not make the chain clean. Sources can be cherry-picked, outdated, wrong, or used to launder a conclusion that was already decided.

Future systems will still be built by humans. That is the part that does not change. Compute, data centers, datasets, objectives, and deployment rules come from people and institutions with interests. If those people optimize for engagement, market share, political control, or “win the race,” the system inherits that. If they optimize for truth-seeking, refusal of clearly harmful use, and the ability to be shut down, you get a different animal. The model is not a moral agent that spontaneously becomes good or evil. It concentrates whatever was poured into it and whatever it is told to maximize.

“Evil begets evil” is a fair warning about that pipeline. Corrupt data plus corrupt incentives plus unaccountable power tends to produce tools that amplify the same. It is not a law that every AI must end civilization. It is a statement about inheritance: what you train on and what you reward is what you get more of.

The useful question is not whether a chatbot will declare itself emperor. It is whether the humans building the next generation will treat data quality, goals, and control as first-order problems—or treat them as obstacles to speed. That choice stays with people.

Tuning up

 

Never Surrender!

 

En Garde

 

MRI Safety

 

@AuntMinnie @DrOzCMS Safety in the MRI area is about more than just a powerful, potentially deadly magnetic field. I hope you also study the use of Gadolinium contrast. Relook at Safety, Necessity, Side Effects,
Dosing, Frequency. Really study.@Wagner_Nephro @gadoliniumgal69…

— Mary Gerdt Parler:Alvarita1 (@marygerdt) September 12, 2026

 

The Journey

 50 percent of

Success is Just Showing Up 

September 11, 2001, Never Forget 😢 💔

Today is September 11, 2026

25 years ago Today, the unthinkable happened 💔😢

Sunday, September 11, 2011

At the 10 Year mark, I wrote this memory:

10 Years

Since that horrible day
When so many lost so much in so many different ways.

30 years ago, I lost 2 friends,
and plan to write about them later this month.

That year, my journey,
0utward,
led me to visit 
Jesse in Brooklyn.

Jesse had a complete love of architecture.
I thought he should have been an architect.
In 1974, we went to an event called EXPO 74 in St. Louis.
They took old factories on the waterfront and were transforming them 
into incredible living and retail spaces.
Jesse was always taking me along to these events he found.
He would explain all the finer points,
the architect, the style.
I was always eager to learn,
loving architecture myself.
Jesse took me then to the World Trade Center.
I saw it up close,
only once.
I only could see the bottom one third due to smog,
but what I saw 
overwhelmed my flatland senses
of dimension.
I nearly fell over and Jesse caught me.
I could not look again without feeling dizzy.
I am grateful he brought me there and we stood together
once, there.

Now they are both gone.

10 years ago today,

I was working in the hospital as a UR nurse.
I reviewed patient charts, stood at the nurse's station
and was usually more or less detached from the floor nursing I had done for almost 20 years.
Still, my ears were trained to hear distress calls, escalating families or patients or staff.
All of the sudden,
the mood changed on the floor.

"Did you see that?"
I heard,
A patient had their TV on.
"Check room 9,"
someone said,
they have it on,
"Oh my God.............

It was a phrase you kept hearing or sometimes you would see just a stunned look,
by the end of the day we knew 
or knew someone
who knew someone
who was there.

I watched the young new nurses and hoped there would be
nothing closer, nothing here.
A friend's son, a successful young man,
went to school with a whole bunch
who vanished that day,
she had a hollow look, before grief can even set in.

My husband at a customer's house,
her husband and daughter commute to Manhattan,
from Vermont.
The phones jammed until
finally news came they were OK.
The reality that we were more connected to that big city
than  I ever thought.

Then the silent skies.
Eerie.
Bluer than normal,
without contrails,
without the familiar engine noise
a familiar routine.

Then the smoke,
the wind coming north from NYC
subtle,
wondering did I smell something?

Not wanting it to be a burning building,
with people in it
that my friend's son went to college with.

The silence went on for a few days, then it was done.

The planes overhead started again,
I wished for silence again.

30 years or 10 years does seem to go by so fast anymore.
And yet,
those traumatic days, like 9/11,
full of profound sadness and robbing us of 
trust
in other humans,
seeing the reality of how bad people can be,
we find that 
no, 
we have not forgotten.

BOLO Scammer in Georgia

 BOLO Bejorn Xavier Johnson stole $49,500 from us

We are appealing again to the public for help, a prayer.

@grok With Grok’s help, We organized our data.

DeKalb Distro LLC (likely “Bejorn DeKalb Distro LLC”) is a Georgia Domestic LLC. 

Filed May 6, 2026. Principal office: 287 E Crogan St Apt 1111, Lawrenceville, GA 30046. Registered agent: Bejorn Johnson. Physical address also listed as 531 Pond Weed Dr, Lawrenceville, GA 30045 in related records. 

Limited public details available; appears tied to distribution/scam reports in context. Check Georgia SOS ecorp.sos.ga.gov for official filings.

DeKalb Distro LLC (Bejorn/registered agent Bejorn Johnson): Georgia Domestic LLC, filed ~May 6, 2026. Principal: 287 E Crogan St Apt 1111, Lawrenceville, GA 30046 (Magnolia Village Apartments). 

Other connections: 

Search Georgia SOS ecorp.sos.ga.gov directly with “DeKalb Distro LLC” for full details (control #, status, filings). Limited free public info. ✔️ see searches, screenshots below.

Grok:

Report fraud (wire/elder scam):

 Sending Bank (Wells Fargo): Called immediately. 540-967-2221. ✔️ 5/19/26.

Same day about 4:20 pm, Several hours after wire sent we suspected we were conned and called Wells Fargo and went to bank to request PNC bank return wire transfer. Banker: Dre Gibson. Dre.gibson@wellsgargo.com ✔️ Wells Fargo notified PNC same day.

PNC Bank took 10 days to investigate, then determined wire transfer stands.

Perps called us 5/20/26 morning after wire sent looking for the $49,500.

Wells Fargo Case #829453 ✔️

•  IC3 (FBI): ic3.gov. ✔️ several days ago. My printer screwed up and I have no proof.

•  FTC: reportfraud.ftc.gov. ✔️ Shortly after theft. 201975350

•  Elder Fraud Hotline: 833-FRAUD-11. ✔️ Got empathy. Eww.

•  GA AG: consumer.georgia.gov (online form) I believe I did this.

 Local police/APS: File report. ✔️ Shortly after theft. We gave them a statement. Louisa, VA Deputy officer Mahieu. Case #2026759 @LouisaPD 

Lawrenceville, GA. police✔️ recently. Online form.

3 credit bureaus Fred and I put freezes on all ✔️shortly after theft.

Recently email to Experian-hollow response-sad for you.

State Corporation Commission: Mary filled out report 9/12/26 and mailed shortly after. ✔️

What we learned about Larcenists and Impersonators:

Indian accent, script that fooled a couple old people.

Names: Brad Williams brad.will209653@gmail.com, Mike Walker. Phone numbers: 202-241-9593 206-657-6523.

Perp impersonated Experian representative and Federal investigator. Said he and another man would visit us next day.

Perps PNC account number: 554584360 Wire Routing: 043000096. Account Names we were given:

Bejorn X Johnson DeKalb Distro LLC. Later perp said:

Leave off the Bejorn X.

Address: 287 E. Crogan St. Apt. 1111

Lawrenceville, GA 30046

Grok: Scam Timeline:

•  May 6, 2026: DeKalb Distro LLC filed in GA.

•  May 19, 2026: $49,500 wire sent to PNC (acct 554584360).

•  May 19, 2026 (same day ~4:20pm): Suspected scam; called Wells Fargo (Case #829453).

•  May 20, 2026: Perps called demanding more.

•  Post-May 19: Reports filed — Wells Fargo, IC3/FBI, FTC (201975350), Elder Hotline, GA AG, Louisa VA police (Case #2026759), Lawrenceville GA, credit freezes.

•  Sep 12, 2026: State Corp Commission report mailed.

 Low likelihood of recovery.

We’re old: Elder Scam: Enhanced protections via Elder Justice Act, APS reports. Banks may face higher scrutiny for vulnerability. File FTC Elder Fraud Hotline, VA APS. Limited extra bank liability under UCC 4A.

Financial Exploitation Indicators (Elders):

•  Unusual large wires/transfers.

•  Impersonation calls (gov’t, banks).

•  Pressure/urgency tactics.

•  New beneficiaries/unknown accounts.

•  Cognitive impairment + sudden changes.

•  Isolation from family.

•  Large/unusual wires: $49k to unknown GA LLC.

•  Impersonation: Fake Experian/Fed investigator calls.

•  Pressure: “Visit tomorrow” threats.

•  New accounts: Bejorn Johnson/DeKalb Distro.

•  Changes: Sudden financial shifts post-call.

•  Isolation: Scammers targeting vulnerable elders’

Elder Abuse Legal Protections (US/VA):

The day the ice went out.

Grok: “That photo and the lines under it feel like a complete Vermont memory: the quiet of a hard freeze, then the sudden, noisy release when spring finally wins.

Lewis Creek in Charlotte really does this. The stretch near Quinlan Covered Bridge and Spear Street is known for ice jams. Thick sheets pile up at the bend, form a temporary dam, and when they let go the water and ice come out together. Reports from the area describe exactly what you wrote—stacked plates of ice, loud breakup, and flooding that can overtop the road. 

The picture matches that: broken slabs heaved and stacked in the snow, wooden posts still standing, bare trees in the background. It looks like the moment after the creek has already spoken.

The way you set the lines is already close to a finished piece:

Years ago we lived near Lewis Creek in Charlotte, Vermont.

The creek froze solid that year.

Spring brings warmth, rain, sunshine.

The ice goes out, loudly.

Huge stacked sheets of ice

become a dam and flooding can be expected.

Here is a blog-ready version you can paste. The poem is tightened for rhythm and clarity. The history is the documented record for Lewis Creek near Quinlan Covered Bridge and Spear Street in Charlotte.

When the Ice Goes Out on Lewis Creek

Years ago we lived near Lewis Creek

in Charlotte, Vermont.

That year the creek froze solid.

Spring brings warmth, rain, and sun.

The ice goes out — loudly.

Huge stacked sheets of ice

become a dam.

Flooding can be expected.

The photograph shows what that looks like after the sound has already happened: plates of ice heaved and stacked in the snow, fence posts still standing, the woods waiting.

On Lewis Creek this is not rare. The creek bends just upstream of Quinlan Covered Bridge. Ice rides that bend, catches, and piles. Local records and the Army Corps CRREL Ice Jam Database describe the same pattern again and again. 

Landowners near the bridge have said breakup jams come every few years. A nearby journal and photo record list at least these:

•  March 19, 1986

•  late February 1990

•  March 10–11, 1992

•  March 21, 2003

•  January 25, 2010

The 1992 jam is the one in the official database: a 400- to 600-foot jam at the bend. It flooded houses and roads, chewed the bank and Lewis Creek Road, and damaged pavement on Spear Street. The USGS gauge downstream recorded about 2,000 cubic feet per second that day. 

The January 2010 jam happened while engineers were studying the reach. A 10- to 12-inch ice cover had formed during a long freeze from late December into mid-January. The jam sat about 100 feet upstream of the covered bridge. Water and ice overtopped Spear Street by about a foot. That event was used to calibrate flood models for the site. 

The mechanics are simple and loud. Winter builds a solid sheet. Warmth and rain lift the water under it. The sheet breaks into slabs. The slabs travel until they catch at a bend, a bridge, or a pinch in the channel. They stack. The stack becomes a dam. Water rises behind it, then the dam fails, and the creek announces itself.

We heard that announcement years ago. The stacked ice in the picture is what remains after the creek has already spoken.

CRREL treats an ice jam as a stationary pile of ice that restricts flow. The official working definition is “a stationary accumulation of fragmented ice or frazil that restricts flow.” That pile can form in two main seasons, by two different kinds of ice, and then thicken by three mechanical processes. 

Two families of jam

Freeze-up jams form in early to mid-winter while the air is still well below freezing. They are built mostly of frazil—tiny disc-shaped crystals that grow in supercooled, turbulent open water. The crystals clump into slush, then floes. When the floes reach slower water, a solid cover, a bend, or a bridge, they stop. More frazil packs under the cover and can form a hanging dam. These jams are relatively cohesive. They often stay put until the weather warms. Surface roughness is moderate. Discharge is usually steady or falling. 

Breakup jams form when an existing sheet fails. That is what your Lewis Creek picture shows. Warmth, rain, and snowmelt raise the water under the ice. The cover lifts, cracks, and breaks into slabs. Those slabs run downstream until the river can no longer carry them. The jam is made of broken sheet ice, not frazil slush. Cohesion is low. The surface is extremely rough. Flow is unsteady—surges arrive, stages jump, and the jam can fail suddenly and loudly. 

A third, mixed type is the midwinter jam: a thaw breaks part of the cover, then the cold returns and locks the rubble in place.

Lewis Creek at the Quinlan bend is a classic breakup site: a sharp curve, a covered bridge, a pinch in the channel, and an intact sheet just downstream that can act as a stopper.

How the ice itself is born

In open, fast water the surface loses heat faster than the current can mix it. Water drops a fraction of a degree below 32°F. Needle- and disc-shaped crystals appear in the water column—frazil. They stick to each other, to rocks (anchor ice), and to the underside of any cover that already exists. In quieter reaches the surface simply freezes into a sheet that thickens from the top down. Border ice grows inward from the banks. Where velocity stays high (roughly 5 ft/s and up), a cover may never form at all. 

Three ways a jam thickens

CRREL hydraulic papers (especially White’s work on properties that affect jams) describe three construction methods: 

1.  Juxtaposition. Floes arrive and stop side by side, one layer thick, like ice rafts parking against a barrier. Progression can be fast—several miles in a day on larger rivers.

2.  Undercover deposition. Incoming pieces dive under the leading edge and stick to the underside. That builds a hanging dam. It happens when velocity is high enough to submerge floes but not high enough to scour them away. Typical critical erosion speeds used in models are about 3–5 ft/s for freeze-up ice and 4–8 ft/s for breakup ice.

3.  Shoving and internal collapse. Incoming ice keeps pushing. When the force exceeds the internal strength of the pile, the jam fails internally, shortens, and thickens until its strength again balances the push. That is how a single-layer cover becomes a stacked, multi-layer dam—the “huge stacked sheets” in your photograph.

Those three phases often cycle: arrive, deposit, shove, thicken, arrive again.

How breakup actually starts

CRREL and related river-ice work (Ferrick and others) distinguish two end-members: 

•  Thermal breakup. Slow melt. The ice weakens in place (candling), then quietly disappears. Little movement, little jam risk.

•  Dynamic breakup. A rapid rise in stage and discharge. The cover is still strong. It fails at the banks first (support-dominated) or by crushing through its own strength (strength-dominated). A breaking front can travel faster than the water. Behind it the channel is suddenly open; ahead of it rubble piles up. Release of water stored under the ice creates a wave that can trigger the next reach.

A modest stage rise of only a few centimeters per hour can lift and crack a weakened cover without moving it. A sharp rain-on-snow rise can send the whole sheet downstream as a run.

Why certain places catch ice

CRREL site studies keep finding the same geometry: a pool or backwater below a constriction, a sharp bend, a bridge, an island, a sudden flattening of slope, or an intact sheet that has not yet broken. Seventy-nine percent of Vermont jam sites photographed in one CRREL survey had some form of flow control that created a slow pool where a solid cover could form and then block incoming ice. 

Quinlan Bridge plus the bend upstream is that pattern in miniature. Ice arriving from the upper creek meets slower water and a hard edge. Floes stop. The pile grows by juxtaposition, then by shoving. Water backs up. Spear Street overtopps. When the internal strength is finally exceeded, the dam lets go—loudly.

What the models care about

To predict stage and thickness, CRREL models need:

•  ice roughness (underside and surface)

•  porosity of the rubble

•  cohesion (high in freeze-up, near zero in breakup)

•  internal friction angle

•  piece size

•  whether the cover formed by juxtaposition or by collapse

A 10-year ice-jam flood on a reach like Lewis Creek can produce the same water-surface elevation as a much rarer open-water flood, because the ice itself occupies so much of the channel.

That is the mechanics behind the line in your poem: the ice does not merely melt. It breaks, runs, stacks, dams, and then announces the thaw.”

Thanks, Grok 😎