W for Weather + Earth Lens — 3 data-backed gaps for a new channel

How a gap is scored

gap = demand × openness (1−saturation) × transfer-fit × freshness × new-value × proof quality (× seasonal factor)

A variation only makes the report if it clears the demand gate: at least 2 independent real videos/channels that each pass on views, a strong relative outlier, or genuinely heavy comments — never a single fluke, never 10 comments on 10K views. Each concept is generated by applying a SCAMPER transformation (Substitute / Combine / Adapt / Modify / Put-to-another-use / Eliminate / Reverse) to the reference, and scored on new-value — how genuinely fresh the angle is, not just how uncrowded — so a me-too topic-swap can't outrank a real transformation. Proof is graded strong / solid / thin as a multiplier. Every proof video links to its source.

The gaps, ranked

01The Forgotten Fault — hazard at REGION resolutionsubtopicsolid0.73

02Ground & Water Failure — named US lake, dam, aquiferadjacentsolid0.49

03Consequence Films — the named hazard run forward as a clockformatstrong0.45

The gaps

The three content pillars that survived the evidence, deep-dived. Every proof names the channel and its subscriber count, so relative outliers are visible: a 100K-view video on a 1,770-sub channel is a far stronger signal for you than a 3M-view video on PBS.

Read the novelty penalties in context. Scores below already carry the channel-level novelty multiplier (×0.6 taken, ×0.85 racing), which is why none of them sits above 0.75. No concept in this niche is unoccupied — it is a gold rush where every winning structure is carpet-bombed within 48-72 hours. You did not ask for an untouched idea; you asked which topics pay. So the ranking that actually matters is the lane table below, not these three scores — and the incumbents named here are all the same size class as you, which is the point: they are beatable, and they prove the seats are not locked.

01

The Forgotten Fault — hazard at REGION resolution

'America's most dangerous fault isn't in California' — reject the geography the viewer expects and hand the dread to a multi-state region that thinks it is safe

0.73gap score

proof solidsubtopic axisSCAMPER: Reversesignal: both

Demand88

New value80

Saturation45

Transfer fit92

Freshness82

In-season40

Novelty check passed no channel is running this exact concept yet.

⚠ Others racing 3 channel(s) just launched this: Critical Latitude (3 weeks old, 11 videos), GeoSphere (active in the lane now), ~21 clone channels (all launched in the last 3 weeks, median ~150 views). Nobody has won yet — move fast.

Channels already in this space: Critical Latitude (15.1K), GeoSphere (19.3K), Earthly (103K), TheGeoModels (88.7K)

The channel. Faceless AI-narrated 20-27 minute mini-documentary. Every episode takes a named US seismic zone the audience has never heard of, attaches a hard population and casualty number (86,000 projected casualties across 8 states; 25-40% chance of M6+ within 50 years), and tells them why the ground under their region is loaded. Fully evergreen — no news peg required. USGS names the zones for you, which supplies both the authority and the ignorance hook.

New value. The San Andreas is a commodity — every channel in the niche has run it. The transformation is a Reverse: instead of telling a Californian about the fault they already know, tell 11.6 million people in Tennessee, Missouri, Arkansas and Illinois that the ground under them is loaded. The genuinely surprising fact nobody leads with — old cold intraplate crust transmits shaking roughly 20x further than California crust — is the reveal that separated the 323,621-view winner from nineteen dead clones running the identical title.

Why the gap is open. Every 100K+ small-channel result in the whole fault lane came from this reversal, not from San Andreas coverage. Critical Latitude's video was the first upload of a channel that did not exist on 17 July 2026 and did 323,621 views with 14,562 likes and 949 comments — roughly 47% of that channel's entire lifetime views. Three more independent channels cleared 140K-344K on the East Tennessee variant. Saturation on East Tennessee is ~25 versus ~80 for New Madrid, so the freshest pocket is still open.

Risks. Do not ship the literal "isn't in California" title — 21 near-identical uploads landed in three weeks with a ~150-view median, and new ones are still arriving. Take the subject, write a new sentence. Do not name a city: Memphis-scoped got 25 views against the region-scoped 323,621 on the same fault in the same month. The bench is also shallower than it looks — only New Madrid and East Tennessee are proven; Wasatch (24K ceiling), Charleston (27K) and Hayward (1.5-33K) all fail the 50K floor unpegged.

Flag. Gold-rush timing: Critical Latitude, Breakpoint, Earth Lens, S zone and Planets Uncovered ALL broke out within the last 5-10 weeks. That is the profile of a lane mid-rush — the upside is real and the window is not guaranteed to stay open.

Opening titles for the new channel

The Fault Under Memphis Rang Church Bells in Boston — And It's Still Fully Loaded
11 Million Americans Live On a Fault They've Never Heard Of
The Appalachians Are Supposed to Be Dead — Something Under Tennessee Didn't Get the Message
This Crust Is So Old It Carries a Quake 20 Times Further Than California's
In 1811 the Mississippi Ran Backwards for Three Hours — The Cause Is Still There
FEMA Modelled 86,000 Casualties Across Eight States — And Almost Nobody Knows Which Ones
Reelfoot Lake Didn't Exist Before One Night in 1812
The Seismic Zone Under Knoxville Is More Active Than New Madrid

Demand proofproof 324K, 344K, 141K, 201K, 224K, 85K

Real videos / channels proving demand

America's Most Dangerous Fault Isn't in California — And It's Far Worse Than You ThinkCritical Latitude · 324K views · 949 comments · 15K subsThe small-channel ceiling proof. 15.1K subs, 11 videos total, lifetime channel views 691,807 — this ONE video is ~47% of everything the channel has ever earned. 14,562 likes. Published 18 July 2026. Incumbents do not own this lane.
East Tennessee earthquakes — are the ancient Appalachian faults waking up?TheGeoModels · 344K views · 89K subsHighest number found on the East Tennessee Seismic Zone, the least-contested pocket in the audit (imitators scored 51 and 271 views).
More Dangerous Than The New Madrid FaultEarthly · 141K views · 103K subsSecond independent proof of the East Tennessee angle at a different channel size — a pattern, not a fluke.
Hundreds of Quakes Shake the Midwest: America's 'Big One' Already Happened HereGeoSphere · 201K views · 19K subsThe swarm-count framing transplanted off California onto a region that pays. 19.3K subs.
200+ Earthquakes Shake The New Madrid Fault: The Big One Could Be CloseEarthly · 224K views · 103K subsSame skeleton as the reference channel's own best video (80 quakes north of SF, 62K) — pointed at a geography that pays 3.6x more.
The Tennessee Hidden Earthquake Threat No One Talks AboutGeoSphere · 85K views · 19K subsThird East Tennessee proof, on a 19.3K-sub channel — clears the 50K floor with near-zero clone pressure.

Who's already there (saturation)

Critical Latitude · @CriticalLatitude · 15K subs — 3 weeks old, 11 videos — the current leader and beatable
Earthly · 103K subs — runs the New Madrid / Tennessee angle well but as one topic among many
GeoSphere · 19K subs — same size class, same lane — proves it is winnable at this scale
~21 clone channels — carpet-bombed the title in 3 weeks; median ~150 views. The topic is not the moat.

02

Ground & Water Failure — named US lake, dam, aquifer

Substitute the hazard: keep the identical dread format and USGS-data register, swap the moving ground for collapsing ground and vanishing water

0.49gap score

proof solidadjacent axisSCAMPER: Substitutesignal: views

Demand78

New value74

Saturation40

Transfer fit85

Freshness70

In-season55

⚠ Already run by a channel Crisis Radar (19.9K subs) — e.g. “Lake Jackson CAVES IN — Shocking 12 Billion Gallons Gone Overnight (546,523)”. Findable in one title search.

Channels already in this space: Crisis Radar (19.9K), MegaBuilds (112K), US Housing Report (2.52K), ViewCation (21.9K)

The channel. Every episode names one American lake, reservoir, dam or aquifer, states a measured number as the headline (12 billion gallons, 42% drop, 2-mile fissure), and tells the story past tense with satellite/InSAR proof. Two modes: live gauge readings for the news half, and a 40-60 episode backlog of historic dam failures with body counts and buried warning reports for the evergreen half.

New value. Same geology identity, same faceless format, same authority register — but the peg supply is refillable weekly straight off USGS gauge data instead of waiting for the earth to move. It also de-risks the whole channel: the fault lane runs dry the moment there is no marquee event, this one never does. Neither reference channel has a single video in it across 28 recent uploads.

Why the gap is open. Crisis Radar is a faceless AI geo-hazard channel with 19,900 subs — 5,000 fewer than the reference's genre twin — and has eleven separate 100K+ videos in this lane. Its best, Lake Jackson at 546,523, beats Earth Lens's all-time best (141,577) by 3.9x. The skeleton has been run at least seven times by the same channel at 103K-252K each, which makes it the most repeatable structure found anywhere in the audit. Historic dam disasters work from literally zero: Vajont did 108,718 on a 611-sub channel.

Risks. The lane may already be cooling. Every 250K+ proof clusters Feb-May 2026, and a filtered sweep for sub-200K-sub channels clearing 30K in this lane in the last 75 days returned zero results. Crisis Radar's own template went 546,523 / 403,885 / 267,052 in early May and 638 recently. Narrow to water and dams only — grid is prepper-owned with a 4-7K ceiling, and bridges have exactly one proven hit in 18 months. Run 3-4 episodes before committing.

Flag. Transfer evidence, not native evidence. Proven on a near-identical peer (Crisis Radar — same format, same size class) but never tested on the reference channel itself.

Opening titles for the new channel

12 Billion Gallons Left This Lake Overnight — Nobody Can Find Where It Went
Satellites Caught America's Biggest Dam Moving — The Ground Inspection Found Nothing
A Crack Two Miles Long Opened in Arizona and It's Still Growing
This Reservoir Fell 42% in Nine Months — The City Below It Has One Water Source
They Wrote a Report Warning About This Dam in 1961 — Two Thousand People Died in 1963
The Florida Retirement Town That Is Slowly Sinking Into Its Own Aquifer
An Entire American Lake Drained Through a Hole in Three Days
The First American City That Will Run Out of Water Already Knows It

Demand proofproof 547K, 843K, 404K, 688K, 385K, 109K

Real videos / channels proving demand

Lake Jackson CAVES IN — Shocking 12 Billion Gallons Gone OvernightCrisis Radar · 547K views · 20K subsBest-in-lane, 11,345 likes, on a channel SMALLER than the reference's genre twin. 3.9x Earth Lens's all-time best.
Texas' Largest Dam Is Sinking — And There Is No Backup PlanMegaBuilds · 843K views · 112K subsThe lane ceiling, built entirely on satellite/InSAR data. The remote-sensing variant is barely worked by small channels.
Three Sisters Caldera CRACKS OPEN — 213,072 Americans On DEATH ZONE AlertCrisis Radar · 404K views · 17K subsSecond 400K+ on the same small channel with NO live event — proves the lane is evergreen, not peg-dependent.
Urgent Warning as Giant Cracks Emerge Across the United StatesViewCation · 688K views · 22K subs31x its own channel average. Won because the body delivered NAMED states (Arizona, California, Texas) — the localisation rule again.
The Villages Florida is collapsing — here's whyUS Housing Report · 385K views · 3K subs2,520 subs. Property-value dread rather than geology dread, on the same subsidence material.
Vajont Dam 1963: The Mistake That Killed 2,000 PeopleGround Zero Science · 109K views · 611 subs611 subscribers. The evergreen historic-dam backlog works from a standing start — with Taum Sauk at 130,813 on 2,580 subs as the second proof.

Who's already there (saturation)

Crisis Radar · 20K subs — owns the lane at the moment but is the same size class as you — and its recent uploads are decaying
MegaBuilds · 112K subs — engineering channel, not geo-hazard — different audience, holds the ceiling
US Housing Report / Giga Builds / CityFall — property-and-infrastructure framing, no geo-hazard identity — the fusion seat is open

03

Consequence Films — the named hazard run forward as a clock

Modify/Magnify: stop reporting what happened and run one named hazard forward against one named place — minute 0, minute 90, hour 6, day 3, week 6

0.45gap score

proof strongformat axisSCAMPER: Modifysignal: views

Demand76

New value78

Saturation60

Transfer fit70

Freshness68

In-season30

⚠ Already run by a channel Vanished Worlds (14.9K subs) — e.g. “SAN ANDREAS: The Big One That Could Destroy California (554,000)”. Findable in one title search.

Channels already in this space: Vanished Worlds (14.9K), Terra Incognita (6.85K), So What If... (2.29K), FAULT LINES (1.04K)

The channel. Evergreen library, not a news feed. One famous named North American hazard (Cascadia, San Andreas, Rainier, New Madrid) versus one named place, run forward as a clock. USGS/FEMA scenario datasets as the spine so the dread is sourced, not invented. 20-30 minutes, faceless, cinematic b-roll.

New value. The reference channels sell anomaly detection — 'something happened, here's what it means'. The payoff is an interpretation, which has a hard ceiling. This sells consequence, which the viewer can picture and argue with. Against the existing consequence videos (PBS Terra, Insane Curiosity) the new value is the clock: nobody in the geo-hazard lane consistently structures the entire film as a named-place timeline sourced from FEMA and state resilience plans — real lahar arrival times, real bridge inventories, real evacuation windows.

Why the gap is open. Four independent channels under 15K subs put up 54K, 287K, 554K and 1.13M on this framing with cheap or AI visuals, and two did it in the last six weeks. Vanished Worlds is ten weeks old, has 14.9K subs and 1.97M total views, led by San Andreas at 554,000 and Cascadia at 414,000 — with visibly imperfect AI visuals that top commenters openly mock. Terra Incognita's 2018 Cascadia film still carries a 6,850-sub channel at 1,129,123 views seven years later: this is the one lane where the asset compounds instead of decaying.

Risks. Read the conflict in this report honestly. The lane audit graded 'hour-by-hour simulation' a trap (demand 58) and the concept deep-dive graded it conditional-go — because they tested different things. The literal framing dies: titles containing 'hour by hour' on small channels scored 3,085 / 377 / 78 / 43 / 37 / 36 / 11 / 10 / 8, and Earth Lens's own Cascadia what-if did 10,212 against its 141,577 best. The consequence film framing wins. Same material, different sentence. Also: the lane was flooded in the last 60 days, and there are only ~8-12 canonical scenarios before the bench runs dry.

Flag. Conditional-go, and the condition is real: this is a different business from the reference channels — an evergreen library on a slow cadence, not a daily news feed. Do not run it as your only pillar.

Opening titles for the new channel

Cascadia Ruptures at 2:14 AM — Every Minute Until the Wave Reaches Seaside
Mount Rainier Gives Orting 40 Minutes — This Is What Those 40 Minutes Look Like
New Madrid Lets Go Under the Midwest — Until Every Mississippi Bridge Is Gone
FEMA's Cascadia Rising Exercise Assumed 13,000 Dead — Here's the Page They Don't Publicise
The San Andreas Snaps on a Tuesday — The First 90 Seconds in Los Angeles
Seattle Has 15 Minutes and Two Evacuation Towers — It Needs Eighty
The Wave Reaches the Oregon Coast Before the Shaking Stops
Week Six After Cascadia — When the Water Still Isn't Back On

Demand proofproof 554K, 414K, 1.1M, 287K, 55K, 2.4M

Real videos / channels proving demand

SAN ANDREAS: The Big One That Could Destroy CaliforniaVanished Worlds · 554K views · 15K subs14.9K subs, channel TEN WEEKS OLD, 1.97M total views. AI visuals visibly imperfect — top comments call out a levitating bridge — and it still won.
CASCADIA: The Megaquake That Could Break America's West CoastVanished Worlds · 414K views · 15K subsSecond 400K+ on the same ten-week-old channel. Not a one-off.
The Next Cascadia Earthquake: Worst Case ScenarioTerra Incognita · 1.1M views · 7K subs6,850 subs, published 2018, still earning. The evergreen compounding case — the opposite of the news treadmill.
So What If Yellowstone Supervolcano Erupted? (1 Second - 1 Month)So What If... · 287K views · 2K subs2,290 subs. Proves the small-channel ceiling in the lane without a production budget.
CASCADIA: The Megaquake America Forgot It Was Waiting ForFAULT LINES · 55K views · 1K subs1,040 subs, two weeks old. Clears the 50K evidence floor from a standing start.
What If The Big Earthquake Hits The Pacific Northwest Tomorrow?Geography By Geoff · 2.4M views · 1.1M subsThe topic ceiling, built on Pexels stock — included to show the format does not require a studio, and to keep expectations honest about scale.

Who's already there (saturation)

Vanished Worlds · 15K subs — 10 weeks old, currently winning — beatable, and proof the seat is not locked
PBS Terra / Geographics / NOVA · 2.0M subs — own the evergreen explainer at 3-4M views; structurally cannot ship inside 48 hours of an event
EarthquakeSim — ~125K/video on genuinely bespoke ShakeMap visuals — the one sub-lane stock footage cannot enter

Competitive landscape

The real channels already in this niche — discovered by similarity search, not guessed. Direct competitors and close-adjacent players, with what each wins with that the reference hasn't done (those became seed candidates above).

Critical Latitude @CriticalLatitude15K subs

Breakpoint9K subs

Earth Lens @Earthlens.202614K subs

Crisis Radar20K subs

Vanished Worlds @vanishedworlds15K subs

S zone25K subs

GeoSphere19K subs

The Field Report7K subs

Planets Uncovered @PlanetsUncovered2K subs

FAULT LINES @FAULT_LiINES1K subs

GeologyHub @GeologyHub397K subs

dutchsinse @dutchsinse683K subs

Discover Tube @DiscoverTube2024227K subs

PBS Terra2.0M subs

Shawn Willsey @shawnwillsey186K subs

~21 clone channels

In-season now — launch to catch the window

Qualified concepts whose demand is seasonally timely to seed right now (build now, rank as the season arrives). Separate from the all-time top so timing isn't lost in the ranking.

The seasonal reverse — pre-build winter nowthinformat0.60

Wildfire and hurricane season both peak in the next 6 weeks and both are traps for this format. The reverse — winter hazard — is 4-5 months out. Script and template it during the quiet weeks so the pivot in November is instant rather than a cold start.

demand 60 · saturation 40

Equinox solar windows (September)thinsubtopic0.55

The one schedulable solar opportunity: two pre-datable spikes a year. Space-news did 81,716 on 1 Aug 2026 riding exactly this on 5,280 subs, naming the US states where the aurora would be visible. Prep the asset in advance; ship only if a real NOAA G3+ watch opens.

demand 55 · saturation 60 — Space-news (5,280 subs) 82K

News & trend-pegged

Qualified concepts pegged to a live news / trending signal found this run — higher risk, shorter shelf-life, but ride a current moment.

Campi Flegrei — day one only, or skipthinsubtopic0.50

The live swarm produced a 30+ channel pile-up in four days. Silki (227K subs) got 134,956 — about 0.6x subs, a par result — and most competent small channels landed under 5,000 (GeoSphere 360, Proofs Unfiltered 127, Earth Pulse 27, Weather Guy 9). Storm Signal decayed 44,679 → 4,399 → 158 in 48 hours. Ship within hours or do not ship.

demand 72 · saturation 88 — Storm Signal (1,100 subs) 45K

Foreign-megaquake misdirection — the peg-supply solutionthinsubtopic0.79

The most useful structure in the whole audit: it converts ANY large overseas earthquake into a US-fault video, which solves the peg-supply problem without breaking the marquee-noun rule. Independently proven at three channel sizes in one week.

demand 90 · saturation 55 — Breakpoint (8,870 subs) 233K · S zone (32,100 subs) 88K · Storm Signal (1,100 subs) 47K

The unclaimed San Andreas stress-loading studythinsubtopic0.72

CBS LA's 3-minute clip on this paper did 604,305; KTLA 196,378; ABC7 74,286; Bright Side 213,644. No faceless channel has built the definitive 10-25 minute treatment. Proven mass appetite, empty long-form slot, marquee noun already in the title.

demand 85 · saturation 30 — CBS LA 604K

Qualified runners-up

Passed the demand gate but did not make the deep-dived top 3. Real proof exists — worth a look.

Marquee named fault + 'something just moved'Substitutethinsubtopic0.79

The only lane where cold-start channels beat every incumbent on the same news cycle — Breakpoint was created 25 May 2026 and did 232,967 views with 324 comments. But it is bimodal: name a marquee fault or do not publish. Cap it at ~35% of uploads; peg decay is savage.

demand 90 · saturation 80 — Breakpoint (8,870 subs) 233K · Breakpoint (8,870 subs) 186K · The Space Desk (22,500 subs) 220K

Caldera as a BEHAVIOUR CHANGE — 'not-Yellowstone' + a population numberReversesolidsubtopic0.74

Highest single ceiling reachable from a small channel with no news peg at all — S zone 409,000 at 25.2K subs, Crisis Radar 403,885 at 17.3K subs. Untapped names with the same shape: Valles, Newberry, Socorro, Medicine Lake, Norris. Never phrase it as an eruption prediction — that framing measurably dies (1,667 and 4,752 on the same channels).

demand 78 · saturation 60 — S zone (25,200 subs) 409K · Crisis Radar (17,300 subs) 404K · The Field Report (6,610 subs) 111K

De-named historical megadisasterEliminatestrongadjacent0.72

The single highest-leverage title rule found in the whole audit. Critical Event's de-named Krakatoa video did 710,119 on 4,100 subs; the SAME channel's named Krakatoa video did 20,062. Naming the historic event is now effectively a de-ranking signal — the 1906 San Francisco cohort ran 25+ channels in six months at a 20-view median.

demand 72 · saturation 50 — Critical Event (4,100 subs) 710K · Dark Horizon (38,300 subs) 641K · Dark Horizon (38,300 subs) 1.7M

Tsunami preparedness-scandal, state by statePutToAnotherUsethinsubtopic0.70

Critical Latitude's '2 towers exist, 58-80 are needed, and Oregon repealed its ban' did 52,219 views and 277 comments with zero news peg. Repeatable state by state (Oregon, California, Hawaii, Puerto Rico, Alaska boroughs). Thin evidence — one strong data point — but a clean, uncontested structure.

demand 70 · saturation 45 — Critical Latitude (15,100 subs) 52K · (877 subs) 37K

Named-fire forensic reconstructionAdaptstrongsubtopic0.68

The wildfire lane is not dead — the body-count recap framing is. 'The Tragic Mistake That Killed 19 Firefighters' did 307,913 on a 2,310-sub channel using only free USDA/Wikimedia stills. Fire-tornado footage sets the ceiling at 1.3M, but it must name the event and date: the generic version got 101 views and 4 views.

demand 74 · saturation 50 — Dark Horizon (38,300 subs) 1.3M · (2,310 subs) 308K · (3,120 subs) 59K

Rejected for weak data

Variations that failed the demand gate — no 2 independent proofs cleared the views / outlier / comment bar. Shown so the gate is auditable.

Variation	Axis	Why it was dropped
Unnamed geography — INCLUDING YOUR CURRENT LIVE PEG	subtopic	Breakpoint published exactly this on 3 Aug 2026 and got 9,140 views / 14 comments — against its own San-Andreas-NAMED swarm video at 186,065 / 274 comments nine days earlier. Same channel, same momentum, same seismology, 20x apart. An unnamed geography is a dead peg even when the news is real.
City-scoped titles (Memphis, Seattle, Denver, Portland)	subtopic	'25,000 Buildings in Memphis Cannot Survive What's Coming' got 25 views — same fault, same month as the region-scoped version at 323,621. Six independent Seattle attempts: 19,467 / 16,462 / 10,672 / 8,689 / 6,223 / 432 / 24. ZERO city-named videos in this format cleared 50K anywhere in the audit.
Hurricane season, El Nino and any pre-season forecast	subtopic	Planets Uncovered: hurricane-season video 8 views, San Andreas video 115,389 — same channel, same format, six days apart. Second hurricane attempt: 382. Peers at 54 / 27 / 5. Forecast content is a TRUST purchase; viewers want a named meteorologist. Earthquakes are a MYSTERY purchase, which is what a synthetic narrator can sell. Four orders of magnitude, not fixable with a better title.
Cloning a winning title — in ANY lane	format	New Madrid title: 21 uploads in 3 weeks, median ~150 views. Rainier title: 11 channels in 9 days, median ~900. Doomsday Glacier: 15+ clones at 291/168/59/30/25/11/10/6/4/1 against the original's 533,512. If a title is working you are already too late — take the SUBJECT and write a new sentence.
Debunk / reassurance in the TITLE	format	Deep-dived and returned NO-GO. Same asteroid, same week: threat framing 74,912 views, relief framing 3 views. Earth Insight's Yellowstone debunk got 70. GeologyHub's own 'Is Yellowstone About to Erupt? (HINT: NO)' did 20,265 in three years against his 85,231 average. Keep the honest payload in the SCRIPT — the reference channel's 62K winner is a dread hook with a truthful body — but never in the title.
Historical retellings with the event name and year	adjacent	Earth Lens ran exactly this and got 4,914 against its own San Andreas video at 142,755 published the SAME DAY — a 29x loss. The 1906 San Francisco cohort: 25+ channels in six months at 94/89/52/50/46/44/38/35/29/23/21/19/13/12/11/10/9/4/3/2/2/1. The de-named version of the same content does 710,119 (see runners-up).
Carrington Event / grid collapse / 'what if the power went out'	adjacent	Best small-channel Carrington result found: 3,829. Then 451, 190, 165, 118, 115, 50, 31, 23, 14. Grid variants: 7,241 / 42 / 27 / 14 / 9 / 4. Twelve-plus attempts, not one clears 4K. It also drags the channel into the prepper vertical — a different audience with different intent.
Literal 'hour by hour' as the title premise	format	Small-channel results for literal hour-by-hour titles: 3,085 / 377 / 78 / 43 / 37 / 36 / 11 / 10 / 8. Earth Lens's own 'What a Cascadia Megaquake Would Really Do' got 10,212 against its 141,577 best. The consequence-FILM framing wins (554K); the simulation-premise framing dies. Same material, different sentence.
Solar storms without a live NOAA G3+ watch	subtopic	Strom network, same channel, near-identical titles: 61,784 (21 Jul, live event) then 1,379 (31 Jul) then 150 (1 Aug). Title craft controls nothing — the only variable is whether a real dated event window is open. Situational at best, roughly 1-2 windows a month.
Foreign geography as the SUBJECT	demographic	Crisis Radar '700 Sinkholes SWALLOW Turkey' 5,358 vs its US Lake Jackson video 546,523 — same channel, same lane, same period. Critical Latitude's Mexico M8.2 video 3,200 vs its US region video 323,621, same week. A ~100x penalty for the disaster happening outside the US. Usable ONLY as the misdirection setup for a US fault story.
Kilauea and Iceland	demographic	Owned beats. Kilauea: GeologyHub 19,142-110,507 per upload, Big Island Video News 2,419,232, while the AI-narrated challenger on the identical eruption got 522. Iceland: Shawn Willsey (a working geologist) 22,707-54,749 per update vs faceless attempts at 1,163 / 1,013 / 480 / 454 / 157 / 122 / 68 / 61. That audience is expert-primed and actively rejects narrated stock footage.
Hayward, Charleston and Wasatch as 'forgotten fault' targets	subtopic	Hayward: 1,498 on a full dread treatment; best result found anywhere 33,474. Charleston: best faceless result 27,556, everything else sub-2K; its one 55,465 was event-pegged. Wasatch: best modern faceless result 24,136 off a 227K-sub channel, small-channel attempts at 50 / 14 / 4. No national name recognition — the forgotten-fault angle needs a zone millions can claim.
Live wildfire body-count / evacuation-tally recaps	subtopic	Earth Lens ran this at 33:47 and got 14,811, while Weather Impact USA — 391 subs — did 57,210 on the same fire in the same 24 hours by framing an open, unresolved number. The tally is already in every wire headline, so the video promises nothing. This is a FRAMING trap, not a topic trap.

Reference teardown & notes

The lane ranking — all sixteen, graded

verdict	lane	demand	satur.	ceiling	typical	type
DOUBLE DOWN	US fault lines & megaquake dread	90	80	3,369,207	30,000	both
DOUBLE DOWN	What is directly beneath a named American city / is my region at risk	88	70	323,621	35,000	both
DOUBLE DOWN	The science of prediction	88	78	232,967	20,000	both
DOUBLE DOWN	The forgotten / non-California fault	85	80	344,246	25,000	evergreen
DOUBLE DOWN	What fails first when the disaster hits	78	74	843,179	45,000	both
secondary	Supervolcanoes & caldera unrest	78	84	409,000	14,000	both
secondary	Sinkholes, subsidence, land deformation, cracks in the earth	76	84	546,523	15,000	both
secondary	Wildfires & megafire season	74	62	1,318,146	45,000	both
secondary	Worldwide volcano watch & eruption precursors	72	84	175,296	8,000	news
secondary	Antarctica, Greenland, glaciers and hidden-under-the-ice discoveries	72	84	533,512	30,000	both
secondary	Historical megadisasters retold with modern data	72	88	710,119	15,000	evergreen
secondary	Tsunami risk, coastal inundation and landslide-generated waves	70	78	1,181,040	35,000	both
situational	Solar storms, space weather and grid vulnerability	55	88	111,364	3,000	news
TRAP	Worst-case simulation: hour-by-hour / second-by-second of what the big one actually does	58	93	287,050	9,500	evergreen
TRAP	Hurricane season, forecasting and rapid intensification	28	93	163,636	300	news
TRAP	Debunking / counter-programming the panic content itself	14	22	67,868	200	both

“Ceiling” = the best result found for a video squarely in that lane. “Typical” = what a competent small channel lands. Both are weighted toward channels under 25K subs, and both carry survivorship bias — the unfiltered base rate is visible in the clone graveyards (New Madrid title median ~150 views).

The single sharpest finding: proper-noun resolution, not topic

Almost none of the variance in this niche is explained by topic, by news, or by production quality. It is explained by how precisely the title names a place, and the winning band is narrower at both ends than instinct suggests.

Too abstract dies. Bright Side — 44 million subscribers — published “Giant Crack Beneath America Could Trigger a Megaquake” on 4 August 2026 and got 13,093 views. Its siblings: 11,845, 6,661, 2,060.
Too specific also dies. “25,000 Buildings in Memphis Cannot Survive What's Coming” got 25 views — same fault, same danger, same month as the video that got 323,621 by scoping the identical material to a multi-state region.

The band that works is a name roughly 5-15 million Americans can claim as theirs and already fear by reputation. The cleanest proof is one channel in one fortnight: Breakpoint's San-Andreas-named swarm video did 186,065 with 274 comments, and nine days later the objectively bigger real news event — the 80-quake swarm north of San Francisco, which is your current live peg — did 9,140 with 14 comments. Same channel, same momentum, same seismology, 20x apart. The only variable was whether the title contained a noun the audience already feared.

The operational consequence: the first editorial decision on every video is not “what happened today” but “whose region am I naming”. The news is the excuse to publish, not the product.

Evergreen vs news — the recommended split is 60/40, and 70/30 for your first 20 videos

Why evergreen carries the majority: every one of the five biggest small-channel results in this audit was published with no live event. Crisis Radar's Lake Jackson 546,523 (19.9K subs). S zone's Yellowstone 409,000 (25.2K subs). Crisis Radar's Three Sisters 403,885 (17.3K subs). Critical Latitude's New Madrid 323,621 with 949 comments — and that was the channel's first upload. Ground Zero Science did 108,718 on a 1963 dam failure at 611 subscribers. Evergreen also compounds: Terra Incognita's 2018 Cascadia film still carries a 6,850-sub channel at 1,129,123 views seven and a half years later.

Why news stays at 40% and not zero: the news half is the only route into the 48-hour information vacuum where PBS Terra, NatGeo and GeologyHub have not yet published, and it is demonstrably winnable cold — Breakpoint built 8,880 subs and 2,365,390 views in ten weeks on a two-day peg cadence.

Why news must not go higher: peg decay is savage and measurable. Storm Signal went 44,679 → 4,399 → 158 on Campi Flegrei across 48 hours. S zone's four Campi Flegrei videos decayed 16,268 → 9,692 → 8,601 → 4,837. The live Campi Flegrei swarm drew 30+ channels in four days and most competent small entrants landed under 5,000 (GeoSphere 360, Proofs Unfiltered 127, Climate Report 88, Earth Pulse 27, Weather Guy 9).

A weekly schedule at daily cadence (7 uploads): 4 evergreen — 2× forgotten-fault/region reveal, 1× water/dam/ground-collapse, 1× caldera behaviour-change with a population number. 3 news-pegged — 1× marquee named fault moved, 1× foreign-megaquake misdirection converted into a US fault story, 1× flex (day-one caldera only if genuinely first, otherwise a fourth evergreen).

Hard rule on the news half: if today's event does not map onto a fault or region the audience already fears by name, do not make the video. Substitute evergreen.

Recommended upload mix

35% — Marquee named fault + something just moved, framed as forecast payoff ("here's what scientists are watching next")
30% — The forgotten intraplate fault, told at STATE/REGION resolution (New Madrid, East Tennessee Seismic Zone, "the Midwest")
20% — Named US water body, dam or ground collapse — past tense, with a measured number
15% — Volcano / caldera as a BEHAVIOUR CHANGE — non-super and US-domestic preferred

Reference teardown — what the two channels actually are

@WforWeather — 1,730 subs, 7 videos, all published this week. Average ~13,900 views; best is the 80-quake San Francisco swarm at 62,000.
@Earthlens.2026 — 14,200 subs, 31 videos, and it is a 19-year-old dormant channel repurposed roughly three weeks ago (music uploads from 2007 are still on it). Average ~22,900 on its geo videos.

Their own data already contains the answer. Earth Lens's fault videos: 141,577 / 129,000 / 60,000 / 45,000. Its wildfire recaps: 6,600-14,800. Its solar alerts: 3,500-24,000. Its El Niño forecasts: 100-10,000. Its “strongest storm” warnings: 50 and 57 views. The topics with the heaviest news coverage are the ones that perform worst.

One structural warning about the whole niche: the identical title “The San Andreas Fault Just Did Something Scientists Have Never Seen Before” fired on Earth Lens (142,413), Planets Uncovered (115,357, on 1,770 subs) and a third channel within 48 hours. This is a swipe-copy ecosystem. It is simultaneously the proof that the format works and the reason no title is defensible for more than about two days.

What this evidence does NOT support

Runtime is unresolved. Every cold-start six-figure proof here is long-form — Critical Latitude 25 min, Earth Lens 27 min, Vanished Worlds 28.6 min, The Field Report 18-19 min. The reference format is 8-12 min. That conflict was not settled.
Gold-rush timing. Breakpoint (created 25 May), Critical Latitude (18 Jul), Earth Lens (repurposed ~3 weeks ago), S zone (25 Jun), Storm Signal and Planets Uncovered all broke out in the last 5-10 weeks. That is the profile of a lane mid-rush, which cuts both ways.
Hit rates are lottery-shaped even for winners. Critical Event went 1-for-11. Planets Uncovered has one 115,357 spike against 12 videos sharing ~19K. So What If went 1-for-12. Do not model the channel on the outlier.
Survivorship bias in every “typical views” figure — most queries used view floors of 30-100K, so the sample is mostly winners.
Cold-start attribution is unproven. Several breakouts came from channels under ten weeks old; topic performance cannot be cleanly separated from a generic algorithmic push on a fresh channel.
AI-voice friction is real and unquantified. The second-highest-liked comment on the 323,621-view winner (382 likes, 77 replies) is “As soon as I heard that famous AI voice, I knew I had to escape the video.” It did not stop that video — but it is the loudest objection in the lane's biggest hit.