What "SHTF" actually means in San Bernardino (92404)

Base date: November 2026. Scope: one zip code, 92404 — north San Bernardino,

Del Rosa / Waterman Canyon side, up against the San Bernardino National Forest.

The honest framing: "SHTF" is not one event here. It is a stack of six events with

radically different probabilities, and the one that is most likely to actually hit

you is not the one people picture.

0. The bottom line first

The highest-probability SHTF in San Bernardino is not a movie disaster. In rank

order of what will actually happen to you over the next ten years:

1. Wildfire + smoke + the 1-3 winters of debris flow that follow it — near

certainty. 92404 is the bottom of the mountain, and this has already happened

three times in twenty years.

2. Extreme heat with grid stress — near certainty, worsening on a schedule

you can literally read off a model: San Bernardino goes from ~6 historical

heat days per year to ~36 by mid-century.[35]

3. Insurance / cost-of-living squeeze — already in progress, non-violent,

and the thing most likely to force a life decision.

4. A big San Andreas / San Jacinto earthquake — genuinely uncertain, but the

highest-consequence item on the list and the one that dictates everything you

should stock.

5. Water curtailment — slow-rolling, structural, mostly a cost/rationing

event until a quake or a bad multi-year drought stacks on top of it.

6. Economic shock — the county is one-industry (logistics) and that industry

is in a freight recession. A port strike, a fuel spike, or a national

downturn hits San Bernardino first and hardest.

The earthquake is the scenario you prepare for because it dominates the others

and it is the only one where the difference between prepared and unprepared is

measured in weeks of water, not in dollars.

1. The earthquake case, with real numbers

Probability, correctly stated

produce a M≥6.7 quake in 30 years (participation probability 0.53).[5]

to 1-in-5 conditional 30-year probability in the time-dependent UCERF3 model.

San Bernardino North: 0.24. San Bernardino South: ~0.20.[6]

M≥6.7 probability but rose in M≥8 probability under UCERF3 — fewer quakes,

bigger ones.[5][7]

destroyed the capilla in the San Bernardino Valley, wrecked Mission San Gabriel,

and killed 40 people at San Juan Capistrano.[8] San Bernardino was also heavily

damaged in 1899 and 1918 on the San Jacinto.[7]

in a single event — in 1812 — and can again.[9]

So: ~200 years of accumulated strain on a fault whose trace runs along the base

of your mountains and through Cajon Pass, which USGS describes as a critical site

for through-going ruptures.[36]

What the ShakeOut model says the region endures

The ShakeOut scenario (M7.8, southern San Andreas) is the best available

empirical model and it is deliberately conservative:

takes 4 to 18 months. MWD's south-of-fault storage covers about 6 months

assuming a brutal 25% demand cut (40% cut initially). Local storage is

insufficient for the repair duration. Some service areas may take 6+ months

to get water back in the taps, and rebuilding the water system outright is

cheaper than finding every break.[3][4]

with debris, tens of thousands still without permanent shelter or jobs.[2]

simply gone. The domino is jobs, not shaking.[2]

A M6 aftershock through Cajon Pass offsets the repaired rail lines.[2]

What that means specifically for 92404

92404 is 63,243 people, median household income $66,701, poverty rate 19.1%,

10.6% of adults with a bachelor's or better, 27.9% under 18, median age 32.5,

50% renters, 22.1% foreign born, 59% of households speaking a language other than

English at home, mostly Spanish.[26] The county has 65,388 SSI recipients.[27]

The county this sits inside is 2.22 million people, 3.20 persons per household,

44.3% speaking a language other than English at home, and a median household

income of $85,478 against a state median far higher.[25] It counted 3,718 homeless

people in the 2026 point-in-time count — 66% unsheltered — and 28,164 K-12 students

(7% of enrollment) were homeless or housing-insecure in the 2024/25 school year.[32]

Read those numbers against the water timeline and you get the real answer to

"what will I endure":

do not have a generator, and do not have a garage. When the water goes out for

weeks, they leave — but they have nowhere to go.

inconvenience at 19% poverty; it is displacement.

means the April-aftershock-era normal takes a full school year to return.[2]

all arrive in English. In 92404, that is a large share of the actual population

missing the actual instructions.[26]

The pattern that ShakeOut describes and every real Inland Empire disaster repeats:

**the people with cars and cash leave; the zip codes like 92404 absorb everybody

who stayed.** Your block densifies, your grocery store shelves empty, and that is

before the supply chain question even arrives.

2. The infrastructure you sit on top of

The city's water pipes are genuinely old

SBMWD runs 700+ miles of mains and 45,000+ service connections for 200,000+

people.[20] Its own capital plan names specific mains from **1924, 1936, 1941 and

1948** being replaced for "chronic leak histories." In one fiscal year the

distribution crews did 621 water main and lateral repairs.[21] A segment

scheduled for replacement (24th/25th/26th St and Davidson Ave) was installed in

1948 and had leaked enough to force replacement.[21] Del Rosa and Foothill sewer

siphons are on the assessment list.[21]

Brittle, aged pipe is exactly what ShakeOut says fails worst: the pipe volume

needed to rebuild is so large that orders get backordered and communities fight

over priority.[2] You are on top of pre-war cast iron in a city that is already

losing water to leaks with the system intact.[21]

The supply itself is more resilient than you'd guess — until it isn't

In an average year the San Bernardino Valley is roughly **75% local

surface/groundwater, 25% imported** State Water Project.[18] Groundwater provides

~80% of the region's supply and the basins hold ~10 million acre-feet of storage,

about 80% full.[19] That is real insulation versus coastal LA.

The catch: imported supply is wildly variable and legally constrained. SWP

allocations have hit 5% in 2014, 2021 and 2022, and a five-year drought is

modeled to average ~15%.[19] And ShakeOut explicitly notes the areas around San

Bernardino proper are served by local agencies and a State water contractor, i.e.

not the MWD storage cushion that the 6-month figure describes.[3]

Practical version: drought years mean rationing and cost, not taps going dry.

An earthquake means taps going dry, because that is a pipe problem, not a supply

problem.

Cajon Pass is your choke point, and it is a known single point of failure

Everything southern-California-bound that isn't on a boat comes through the pass:

~150 BNSF trains/day across three tracks plus a Union Pacific line, 5,500 trucks/day,

~30 million tons of cargo a year, and I-15 congestion costing >$3B/year.[33]

Worse, the 1990s FEMA/EERI lifeline study of Cajon Pass found the lifelines are

co-located — fiber, fuel pipelines, water mains, power and rail all in the same

narrow corridors. Its conclusions: fuel pipeline failure has the largest knock-on

impact on all other lifelines, and co-located fiber means "more severe telephone

communication failures than have been experienced in past earthquakes" is the

expectation.[10]

Translation for you: **the fiber and the fuel and the rail all break together,

in one pass, 15 minutes north of your house.** Expect internet and cell data to be

the slowest thing to come back, not the fastest.

Hospitals: already running hot before anyone is injured

Community Hospital of San Bernardino: 18 ER treatment stations, 61,159 visits —

161% of rated capacity. St. Bernardine: 24 stations, 68,620 visits —

143% of capacity.[34] Both are well inside the city. Loma Linda, Arrowhead

Regional (Colton) and Kaiser Fontana carry the rest.[34]

In a mass-casualty event, the ERs that are already 40-60% over design load become

triage tents. If you have a non-trauma emergency on day one, you wait.

Power

Better than you'd fear, worse than "fine." Summer 2026 statewide planning margin

was projected at 4,100-6,750 MW surplus, but an event like 2020 or 2022 drops that

to roughly 300-3,000 MW, and California's import capability (up to ~11,655 MW) is

itself vulnerable to western wildfires taking lines out.[22][23]

Fuel

California crude refining capacity is down ~18% after the Phillips 66 Wilmington

and Valero Benicia closures, with the state leaning on record marine imports to

fill the gap.[22][24] That means the state's fuel system now has less buffer and

no redundancy in the Inland Empire — every gallon here arrives by truck over the

same freeway corridors that will be broken.

Insurance and money

The state's insurer of last resort is the market now: 696,562 FAIR Plan policies

and $768B of exposure by June 2026, ~tripled in five years. San Bernardino County

is second in the state with 65,132 FAIR Plan policies, +109% growth, and four

counties including San Bernardino carry FEMA's "Very High" wildfire rating.[13]

After the January 2025 LA fires the FAIR Plan needed a $1B assessment, 105 insurers

passed the recoupable half onto policyholders statewide, and a judge upheld the

surcharges in June 2026.[14] The structural point: the FAIR Plan only covers fire.

Water damage, theft, and liability need a separate DLC policy.[13]

And the city itself is structurally short: FY2025/26 General Fund revenue $228.2M

against $257.5M of expenditure, gap closed with reserves, with nearly half of

general fund revenue coming from two consumer-spending taxes.[31] A consumer

recession guts the city's ability to respond at exactly the moment it's needed.

The fire case, which is the one that will actually happen

This is not hypothetical. The Old Fire started in Waterman Canyon in October 2003,

burned 91,281 acres and 993 homes, and killed six people — five from stress-induced

heart attacks.[11] It was arson out of the canyon at the bottom of 92404.

In 2019 the Old Water Fire hit the same canyon and the mandatory evacuation covered

Arrowhead Road, Mariposa Drive and David Way north of 40th Street — **92404

addresses** — with the evacuation center at San Gorgonio High School, 2299 Pacific

Street, 92404.[12] Highway 18 was closed between 40th Street and Crestline.[12]

The USFS calls Waterman Canyon one of the most fire-prone spots at the base of the

mountains — "fire likes to travel uphill and you're at the bottom."[12]

The follow-on hazard is the part people forget. After a burn scar, debris flows

trigger on startlingly little rain — USGS thresholds put magnitude-I events at

roughly 0.3 inches of rain per hour-equivalent.[16] In December 2010 an atmospheric

river dumped 10.95 inches at Lytle Creek and 10.57 at Devore and buried cars two

feet deep in mud in Loma Linda and Highland.[17]

So the realistic fire sequence for 92404: smoke and ash for weeks, evacuation

order for the north end, possible structure loss, then **two winters where a

single good storm puts mud across your street.**

The city adopted updated Fire Hazard Severity Zones for Local Responsibility Areas

in June 2025 (Ordinance 4489), which is what drives the insurance and building

consequences.[15] Those maps are the practical reason a 92404 address near the

foothills is harder to insure than one three miles south.

Jobs and the economy

Logistics is roughly one in five workers in San Bernardino County — near double the

state and national share — and it is concentrated in exactly the roles that stop

when freight stops.[28][29] Warehouse work pays about 75% of the regional average,

with the lowest tier at $42,850 against a regional average of $61,930.[30]

That is the economic SHTF: not a crash, a squeeze. The Inland Empire is where the

national freight cycle shows up first and leaves last.

3. What you will actually endure, by timeline

Hour 0 — 72 hours (earthquake or major fire)

Shaking or a wall of smoke and an evacuation order. Power out, gas meters possibly

tripped, highway 18/210/215 jammed. Cell networks congested; SMS works better than

voice. Pipes are still full and pressure still exists for the first several hours

to a day, which produces false confidence — the system is bleeding out through

breaks you cannot see.[3] Water disappears day 1-2.

Week 1

Water is the entire story. Boil-water notices. Sewer damage in places means no

flushing. Fires burn unattended because hydrants are dry.[3] ERs triage. Aftershocks

every hour, dropping off. Grocery shelves empty in 24-72 hours and restock arrives

over roads lined with debris.[2] Gas stations without power cannot pump.

Weeks 2-8

The divergence point. Most gas and electric comes back, most major roads reopen —

and the people who could leave have left.[2] Renters are the first out. Families

double up. The city, already running a $29M structural deficit,[31] is now doing

emergency response on reserve funds.

Months 3-6

Water returns to most of the valley; some pockets wait longer, up to 6 months.[3]

Water is not certified safe for a year or more in places.[2] Small businesses that

lacked reserves are permanently closed, and their jobs go with them.[2] This is

the phase that decides whether your neighborhood recovers or sheds population.

Month 6 - Year 2

Normalization, at a lower level. And if there was a fire, the first two winters

bring debris flow risk every storm.[16][17]

Non-earthquake version (heat or wildfire smoke or drought): far shorter, far

more likely, and cumulative. Repeated smoke seasons, repeated Flex Alert evenings,

rising insurance, rising water bills.

4. What is actually in your favor — do not skip this

of storage.**[19] This is the single biggest structural advantage you have over

coastal LA.

hostage to the Delta in a normal year.

former Norton Air Force Base, with a distribution center the chain claimed as

the largest under-one-roof grocery DC in the industry — 2.1 million sq ft

originally, built out to 2.5 million sq ft under a ~$300M modernization that

consolidated eight area facilities into it, running 24/7.[37][38] A locally

based grocer with local cold storage is a materially better food position than

a city served only by national chains. Note the shape of that dependency too:

one campus, one metro.

had a line open within 48 hours of the 2024 pass fires, with I-15 reopened in

two days. National economic interest guarantees engineering priority.

has been adding resources (36 GW nameplate since 2020).[23]

from the coast, which keeps the tax base growing even as it strains services.[*]

5. Ranked prep, by leverage per dollar

1. Water, 2 weeks minimum, 4 weeks better, per person. Water is the only

failure mode that is both certain in a quake and un-fixable by you once it

starts. Aqueduct restoration is 4-18 months;[4] your pallet of water is the

bridge. Add a bathtub bladder and a way to filter/boil.

2. Sanitation. Five-gallon bucket, bags, bleach. ShakeOut's own narrative

flags continuing sewer failure as a scattered but real condition.[2]

3. Two weeks of shelf-stable food that does not need water to prepare, plus a

camp stove and fuel. Grocery restock is 24-72h+ and depends on intact roads.

4. Go-bag for an evacuation order, not a quake. In 92404 you are far more

likely to be told to leave than to be trapped. Documents, meds, cash, pet

carrier, phone battery, and a route that does not use Highway 18 west.

5. Cash, small bills. Card readers need power and network.

6. A generator or a big battery + a way to run a fan. Heat is the actual

killer in this valley, not cold.

7. Get the insurance right now. FAIR Plan + DLC if you are in the foothills

anyway.[13] Do it before a fire, not after — non-renewal is the norm here, not

the exception.[13]

8. Debris-flow awareness after any burn scar near you. Know what drainage you

are downhill of before the first storm, not during it.[16]

9. A physical paper map and a printed contact list. Co-located fiber in Cajon

Pass means data comes back slowly.[10]

10. Know your two hospitals are at 143-161% ER load already.[34] For a

non-trauma emergency, plan to wait or go further afield.

6. The thing worth internalizing

San Bernardino is not a fragile place. It is a **hard, self-sufficient valley

sitting directly on top of two things that periodically reset it** — a major fault

system and a very fire-prone mountain front. It has been reset before: 1812

destroyed the valley's only structure,[8] 2003 burned a thousand homes out of

Waterman Canyon,[11] and 2010 buried cars in mud two towns over.[17]

Every one of those events was absorbed by the same thing: a valley that has

groundwater, agricultural capacity, a locally headquartered grocer, and a

population that is young, hardworking, and used to getting by. The failure mode

that actually hurts here is not dramatic — it is a long, grinding water outage in

a majority-renter, 19%-poverty zip code, where the people with the fewest options

wait out the longest repair.[2][3][26]

Prepare for the water first. Everything else is survivable and most of it is

already happening slowly.

7. This document is now watched, not static

The ranked hazards above are monitored live for 92404 by

~/.hermes/scripts/shtf_92404_watch.py, run every 30 minutes as the Hermes cron

job shtf-92404-sentinel. It is silent unless something actually crosses a

threshold, so it costs nothing on a normal day. On a real event it emails

cuellargarrett@gmail.com through the host's shared notify path (which owns

the AgentMail daily-quota cooldown) and prints the alert to the gateway.

What it watches, and the tripwire on each:

FeedSource (keyless)Fires when
QuakesUSGS FDSN, M≥3.0 within 50 kmany new event
NWS alertsapi.weather.gov, point=34.1458,-117.2613severe/extreme/moderate: heat, wind, red flag, smoke
Smoke/airOpen-Meteo air qualityPM2.5 ≥ 35.4 µg/m³ or US AQI ≥ 100
Heat + windOpen-Meteo 3-day daily max≥100 °F or ≥25 mph
WildfireCalFire incident feedany incident within ~66 km
GridCAISO rtm_forecast.csv + fuelsource.csvreserve margin < ~3,000 MW / 6% of demand, or import share both above 14% and ≥1.15× its own trailing median
Space weatherNOAA SWPC Kp forecastany G1+ predicted in the 72 h window; escalation to G3+ re-announces

Design notes worth knowing: each hazard dedupes independently, so a heat warning

cannot mask a new fire; the fire feed re-announces a still-active incident once a

day; NWS ids are forgotten once they drop out of the feed, so a re-issued warning

re-alerts; and a total feed outage reports once then goes quiet for 6 hours

instead of paging every 30 minutes. State lives in

~/.hermes/state/92404-watch/; delete state.json to re-announce everything.

Three bugs found and fixed on 2026-11-16

All three were silent, and each is worth knowing about because each one looked

like it was working.

1. The import-share arm was pure noise. It used an absolute 12% threshold.

Measured against a full day of CAISO fuelsource.csv rows, the share sat at

17–22% (median 19.1%) and was over 12% in 100% of intervals — so the

original arm fired on essentially every tick, and 19 of the sentinel's first

25 alerts were grid. Replaced with a dyadic arm: absolute floor 14% and

≥1.15× its own trailing median (~8 days of samples), 6 h hysteresis.

2. A permanently dead gauge masquerading as a live one. 11055500 (Plunge

C) is catalogued for the area but publishes no current IV data, so every run

logged water: ... unparseable. That constant error line also masked the

real one. Replaced with 11058600 (Waterman Canyon C) and 11063680 (Devil

Cyn C) — the two canyons whose mouths are 92404, both probe-verified with

90+ intraday points.

3. api.open-meteo.com was failing intermittently with

_ssl.c:1064: handshake operation timed out on ~4 runs in 5 — an ALPN

negotiation failure, not a network fault. The shared fetch helper now pins

ALPN to http/1.1 and retries once; verified 3/3 OK at 0.12 s where the

default context timed out 3/3 at the full 25 s. **This was the weather

sensor — the heat/wind tripwire — silently dark.**

The one thing it cannot see is what the report calls the real danger — a

long, grinding water outage. That stays a human read: if the taps go dry, the

ranked prep list in section 5 is the plan.

---

\* Two claims still rest on weaker provenance than the rest of this document:

"cheap land in a labor-rich valley keeps absorbing people from the coast" (a

demographic generalization, directionally consistent with [25][26] but not

stated that way by any single source), and the "~$3B/year" I-15 Cajon congestion

figure in [33], a secondary aggregator rather than a state source. Everything

else is cited to a primary or near-primary page.

Sources

[1] https://www.usgs.gov/publications/shakeout-scenario-a-hypothetical-mw78-earthquake-southern-san-andreas-fault

[2] https://pubs.usgs.gov/circ/1324/c1324.pdf

[3] https://doi.org/10.3133/ofr20081150

[4] https://doi.org/10.1193/1.3571563

[5] https://southern.scec.org/ucerf

[6] http://wgcep.org/resources/UCERF3-TD_2015_BSSA_Paper.pdf

[7] https://en.wikipedia.org/wiki/San_Jacinto_Fault_Zone

[8] https://earthquake.usgs.gov/earthquakes/eventpage/ushis31/impact

[9] https://pmc.ncbi.nlm.nih.gov/articles/PMC4803493

[10] https://mitigation.eeri.org/wp-content/uploads/fema_226.pdf

[11] https://en.wikipedia.org/wiki/Old_Fire

[12] https://fire.ca.gov/incidents/2019/10/24/old-water-fire

[13] https://moneygeek.com/insurance/homeowners/california-wildfire-fair-plan-insurer-retreat

[14] https://latimes.com/business/story/2026-07-02/home-insurer-surcharges-for-wildfires-is-legal-judge-rules

[15] https://sbcfire.org/fire-hazard-severity-zone-map

[16] https://pubs.usgs.gov/of/2010/1039/pdf/OF10-1039.pdf

[17] https://www.pressenterprise.com/2010/12/21/rain-flood-warning-slowing-commute

[18] https://www.sbvmwd.com/news/pr-san-bernardino-valley-municipal-water-district-board-reduces-property-tax-rate-for-state-water-project

[19] https://www.sbvmwd.com/app/uploads/2026/05/Part-2-SBVMWD-UWMP-Public-Draft.pdf

[20] https://www.sbmwd.org/226/Service-and-Infrastructure

[21] https://sbmwd.org/AgendaCenter/ViewFile/Agenda/_06092026-283

[22] https://autl.assembly.ca.gov/system/files/2026-07/tn271113_20260630t142353_cerro-2026-draft-staff-report-1.pdf

[23] https://efiling.energy.ca.gov/GetDocument.aspx?DocumentContentId=109457&tn=271805

[24] https://s.giannini.ucop.edu/uploads/pub/2026/02/27/v29n3_3_0ZBYSxj.pdf

[25] https://www.census.gov/quickfacts/fact/table/sanbernardinocountycalifornia/DIS010224

[26] https://censuseasy.com/zip/92404

[27] https://www.ssa.gov/policy/docs/statcomps/ssi_sc/2023/ca.pdf

[28] https://www.amtec.co/blog/inland-empire-workforce-report

[29] https://www.sbsun.com/2026/07/16/what-is-the-inland-empires-true-economic-challenge

[30] https://news.ucr.edu/articles/2025/09/08/report-inland-empire-warehouse-workers-make-gains-through-collective-action

[31] https://sanbernardino.gov/DocumentCenter/View/10390/City-Managers-Adopted-FY2025-26-Operating-Budget-and-CIP-2026-2030

[32] https://indicators.sbcounty.gov/housing/homelessness-housing-insecurity

[33] https://trucker-route.com/passes/cajon-pass

[34] https://oag.ca.gov/sites/all/files/agweb/pdfs/charities/nonprofithosp/dignity-chi-san-bernardino-health-impact-report.pdf

[35] https://calmatters.org/environment/climate-change/2024/09/california-extreme-heat-population-growth-inland-communities

[36] https://www.usgs.gov/publications/cajon-pass-and-southern-san-andreas-fault-system-earthquake-cycle-stress-accumulation

[37] https://www.freightwaves.com/news/stater-bros-opens-300-million-grocery-dc — Stater Bros. opens $300M grocery DC (FreightWaves)

[38] https://progressivegrocer.com/stater-bros-open-21-million-sq-ft-facility — Stater Bros. 2.1M sq ft facility (Progressive Grocer)

Linked from: AI data centers and the grid, Hermes original, AI data centers and the grid, with a second opinion, Home