← Back to scans

Will there be an at least 8.0 magnitude earthquake in California before 2035?

KXEARTHQUAKECALIFORNIA-35 · Climate and Weather · 2026-04-05
6%
Multi-Step
55%
Market Price
-49.0%
Edge
75%
Confidence
Final Rationale
UCERF3 estimates 7% probability of M8.0+ in California over 30 years (~2.3% over 9.75 years). I adjust upward to ~6% after accepting the Devil's Advocate's corrections: the 335-year seismic gap on the southern San Andreas warrants stronger time-dependent adjustment (+1.5-2%), the 250-year absence of M8.0+ was incorrectly used as downward evidence when it's consistent with being at/past recurrence intervals, and the Cascadia subduction zone's southern terminus at Cape Mendocino adds ~0.5% for edge-case epicenter scenarios. The market at 60% is 10x my estimate, but I reject this price as unreliable: the +41% surge in 30 days from a 4% baseline with shallow orderbook depth and no new seismological information indicates speculative activity, not informed pricing. No credible seismological model supports anything near 60% cumulative probability over this timeframe.
What Would Change My Mind
Discovery that the 2023 NSHM update dramatically revised M8.0+ probabilities upward (e.g., to 20%+ over 30 years); a major foreshock sequence (M6.5+) on the southern San Andreas suggesting imminent rupture; or clarification that 'in California' includes any rupture zone touching CA (not just epicenter), which would substantially increase the Cascadia contribution to perhaps 3-4% additional probability.
1. Planning Agent Sonnet
binary science article_search wikipedia_lookup kalshi_data code_execution web_search
Sub-questions
IDQuestionWeight
sq1 Is the annual probability of an M8.0+ earthquake in California estimated at 1% or higher per year based on seismological models? 40%
sq2 Does California have active fault systems (e.g., San Andreas, Cascadia subduction zone) capable of producing M8.0+ earthquakes within the next decade? 30%
sq3 Given the ~9.75 year window remaining (April 2026 to December 2035), does the cumulative probability of an M8.0+ California earthquake exceed 15%? 30%
Reasoning Approach
This question asks about a rare but physically plausible seismic event. The key framework is: (1) Establish the annual probability of M8.0+ in California using seismological consensus (UCERF3 from USGS is the gold standard), (2) Apply the Poisson/geometric probability model over the ~9.75-year window, (3) Carefully distinguish between M7.9 events (like 1906 SF, often rounded to ~7.9-8.0) and strict M8.0+. The San Andreas fault can produce up to ~M8.2, but strictly M8.0+ events in California proper (not offshore Cascadia) are very rare historically. UCERF3 estimates ~7% probability of M8.0+ in 30 years for Southern CA, suggesting roughly 2-3% cumulative over a decade. The Cascadia subduction zone can produce M9+ but its epicenter is offshore Oregon/Washington, not California. Key distinction: territorial waters of California could include offshore segments. The question is whether the Cascadia zone counts — likely not, as its main rupture zone is off Oregon/WA. This is a low-probability, long-horizon question where the base rate from official seismic hazard models dominates.
2. Research Execution No LLM
ToolStatusTime
article_search OK 0.82s
wikipedia_lookup OK 1.8s
kalshi_data OK 2.0s
code_execution OK 0.01s
web_search OK 54.54s
3. Evidence Synthesis Sonnet · 16 items
#StrengthDirectionSourceClaimPriced In
1 STRONG ↓ DOWN web_search UCERF3 (the official USGS seismic hazard model for California) estimates a 7% probability of an M8.0+ earthquake in California in the next 30 years, up from 4.7% under UCERF2, implying an annual probability of roughly 0.24% per year. Yes
2 STRONG ↓ DOWN code_execution Using UCERF3's 7%/30-year baseline, the implied annual probability is ~0.24%, and the cumulative probability over 9.75 years is approximately 2.3% — well below 1% annually. Yes
3 STRONG ↓ DOWN article_search The largest recorded California earthquakes are both M7.9 (1857 Fort Tejon and 1906 San Francisco), and no confirmed M8.0+ event has occurred in California's recorded history (~250 years), suggesting historical base rates are even lower than 1% annually. Yes
4 STRONG — NEUTRAL web_search UCERF3's increased M8+ probability is driven by multi-fault rupture scenarios; the southern San Andreas Fault is identified as the most likely source of such an event. Yes
5 STRONG ↑ UP wikipedia_lookup The San Andreas Fault extends ~750 miles through California and forms the boundary between the Pacific and North American plates, with slip rates of 0.79–1.38 inches/year, making it physically capable of producing M8.0+ earthquakes. Yes
6 MODERATE ↑ UP article_search The 2025 Myanmar M7.7 earthquake on the Sagaing Fault was explicitly cited as a wake-up call for California, with geophysicists noting California has similarly locked fault systems under accumulating stress. Yes
7 MODERATE — NEUTRAL article_search Recent seismic activity near Indio (Jan 2026) near the San Andreas system, and a 5.2-magnitude near San Diego (Apr 2025), indicates ongoing fault activity, though none approach M8.0. Yes
8 STRONG ↑ UP web_search UCERF3 identifies the southern San Andreas Fault as the most likely source for an M8+ event in California, confirming the physical capability exists within the state's fault systems. Yes
9 STRONG ↓ DOWN article_search A December 2024 M7.0 earthquake struck off the coast of Northern California, described as one of the largest the state has seen; the historical record shows no M8.0+ in 250+ years of California seismic records. Yes
10 STRONG ↓ DOWN code_execution Using UCERF3 baseline (annual ~0.24%), the cumulative probability over 9.75 years is ~2.3%; using a historical rate of 1 event per 200 years (~0.5% annually), it rises to ~4.8% — both well below the 15% threshold. Yes
11 MODERATE ↑ UP kalshi_data The Kalshi market for this exact question (M8.0+ in California before 2035) is currently priced at 60%, which is dramatically higher than the seismological models suggest (~2–5% cumulative probability). No
12 MODERATE ↑ UP kalshi_data The 2035 California earthquake market has surged +41% over 30 days and +13% over 7 days, moving from as low as 4% to 60-73% range — a massive, rapid repricing that is likely speculative/uninformed rather than based on new seismological data. No
13 MODERATE ↑ UP kalshi_data The related Kalshi market for M8.0+ in California before 2028 is priced at 17% — also much higher than UCERF3 models would suggest for a ~2-year window, indicating possible market mispricing or different resolution criteria interpretation. No
14 MODERATE ↓ DOWN kalshi_data For comparison, Japan M8.0+ before 2030 is priced at 54% on Kalshi — Japan has significantly higher M8+ base rates than California, providing a reference point suggesting California's 60% figure is likely overpriced. Yes
15 STRONG — NEUTRAL article_search The July 2025 Russian M8.8 earthquake (off Kamchatka) triggered Pacific-wide tsunami alerts affecting California's coast, but had no epicenter in California — suggesting large earthquakes in the region are occurring but not in California itself. Yes
16 STRONG ↓ DOWN wikipedia_lookup UCERF3 is the authoritative model for California earthquake probabilities and its M8+ 30-year estimate of 7% is the scientific consensus — no newer official model superseding it has been identified in the evidence. Yes
Information Gaps
  • No evidence on whether a newer model superseding UCERF3 (e.g., UCERF4 or updated USGS National Seismic Hazard Model post-2018) has revised M8+ probabilities for California upward or downward
  • No clarity on whether the question's 'territorial waters of California' could include portions of the Cascadia subduction zone — if so, this would meaningfully increase probability
  • No data on current 'seismic gap' stress accumulation on specific fault segments (e.g., southern San Andreas has not ruptured in ~330 years), which could affect time-dependent probability estimates
  • No explanation for the dramatic 30-day surge in Kalshi market price (+41%) — unclear if driven by a news event, market manipulation, Russian M8.8 earthquake spillover interest, or liquidity issues
  • No specific UCERF3 time-dependent (vs. time-independent) probability breakdown for the 2026–2035 window specifically
  • No data on whether the December 2024 M7.0 Northern California offshore quake has updated any fault stress models
Key Uncertainties
  • Whether Cascadia subduction zone is considered 'California territorial waters' for resolution purposes — if yes, probability could be much higher (Cascadia produces M9+ events on a ~300-500 year cycle)
  • The Kalshi market is wildly inconsistent with seismological models — either the market is severely mispriced due to speculative activity, or there is a resolution interpretation (e.g., including Cascadia) that justifies higher probabilities
  • Time-dependent probability: the southern San Andreas Fault has not ruptured in ~330 years (since ~1680), potentially making the time-dependent probability significantly higher than the time-independent UCERF3 estimate
  • Whether the July 2025 Russian M8.8 earthquake caused any stress transfer to California fault systems that would temporarily increase near-term probability
  • Historical record completeness: with only ~250 years of California seismic records and M8+ recurrence intervals potentially on the order of 200-500+ years, base rate estimates carry very wide uncertainty intervals
Market Context
Price: 0.6 · 7d: +0.13 · 30d: +0.41
Volume: LOW-MODERATE (131 avg daily contracts) · Depth: SHALLOW
Reliability: LOW — extreme recent price movement (+41% in 30 days, from historical low of 4% to current 60%) with low volume strongly suggests speculative activity or thin liquidity rather than informed forecasting; the price is dramatically inconsistent with UCERF3 seismological consensus estimates of ~2-5% cumulative probability over the window
4. Independent Forecaster Opus · No market price Did NOT see market price
Independent estimate: 4% (72% confidence)
sq1: Is the annual probability of an M8.0+ earthquake in California estimated at 1% or higher per year based on seismological models?
5%
UCERF3 estimates 7% probability over 30 years, which implies ~0.24% annual probability. This is well below 1% per year. Even with time-dependent models that account for the ~330-year seismic gap on the southern San Andreas, the annual rate might increase to perhaps 0.3-0.5%, still well below 1%. No credible seismological model puts the annual M8.0+ probability at 1% or higher. The only way this could be true is if there's a radical revision of fault models not yet published, which is very unlikely. P_yes = 5%.
sq2: Does California have active fault systems capable of producing M8.0+ earthquakes within the next decade?
92%
The San Andreas Fault is physically capable of producing M8.0+ earthquakes. UCERF3 explicitly models multi-fault rupture scenarios that can produce M8.0-8.2 events. The southern San Andreas has accumulated significant strain over 330+ years without rupture. The physical capability is well-established in the scientific literature. The only uncertainty is whether 'capable' means it could happen in any timeframe vs. specifically the next decade - but the fault is always capable, and strain is continuously accumulating. P_yes = 92%.
sq3: Given the ~9.75 year window, does the cumulative probability of an M8.0+ California earthquake exceed 15%?
3%
Using UCERF3's 7%/30-year estimate, the cumulative probability over 9.75 years is approximately 2.3%. Even using more generous estimates - say the time-dependent probability is 1.5x higher due to the southern San Andreas seismic gap - this yields roughly 3.5% over 9.75 years. Using a historical rate of ~0.5%/year gives ~4.8%. None of these approaches come close to 15%. The only scenario where this exceeds 15% is if Cascadia subduction zone events count (which they probably don't, as the main rupture zone is off Oregon/Washington) or if there's a fundamental misunderstanding of the fault physics. P_yes = 3%.
Show full reasoning trace
base rate: UCERF3 estimates 7% probability of M8.0+ in California over 30 years. No confirmed M8.0+ earthquake has occurred in California's ~250 years of recorded history (1857 Fort Tejon and 1906 SF are both M7.9). The annual base rate is approximately 0.24% (time-independent). Time-dependent models might push this to 0.3-0.5% for the southern San Andreas given the 330+ year seismic gap since ~1680.
evidence updates: Time-dependent seismic gap on southern San Andreas pushes UP by ~1% cumulative (from 2.3% to ~3.5%). Possible inclusion of California territorial waters for offshore events pushes UP by ~0.5%. Recent seismic activity (Dec 2024 M7.0 offshore, Jan 2026 Indio activity) is weak evidence, pushes UP by <0.5%. The fact that no M8.0+ has been recorded in California in 250+ years pushes DOWN slightly from UCERF3's model-based estimate. Net estimate: ~3.5-4.5% cumulative over 9.75 years.
combination method: Direct estimation from seismological models rather than mechanical weighted average, as the sub-questions are not well-structured for direct combination into the target probability. The weighted average of 0.305 is dominated by sq2 (capability=92%) which is not informative for the occurrence probability.
final: The probability of an M8.0+ earthquake in California before end of 2035 is approximately 4%. This is driven by UCERF3's 7%/30-year estimate (~2.3% over 9.75 years), adjusted upward modestly for time-dependent effects on the southern San Andreas and slight territorial waters ambiguity. The question asks about a genuinely rare event - no M8.0+ has been recorded in California's history - and while the physical capability exists and strain is accumulating, the probability over a ~decade window remains low.
5. Devil's Advocate Sonnet Market price revealed
Reasoning Flaw Missing Info Asymmetric Evidence
Challenges
  1. The forecaster correctly identifies UCERF3's 7%/30-year figure but may be slightly underweighting the time-dependent component. The southern San Andreas has not ruptured since ~1690 (335+ years), and UCERF3's time-dependent branch gives higher probabilities for segments with long elapsed times. The forecaster acknowledges this but only adjusts from 2.3% to ~3.5%, which may be insufficient given that time-dependent models can increase segment-specific probabilities by 2-3x for the most overdue segments.
  2. The forecaster states 'no M8.0+ has been recorded in California in 250+ years' as downward evidence, but this is partly circular reasoning — the recurrence interval for M8.0+ events on the San Andreas is estimated at 150-300 years, so NOT having one in 250 years is actually consistent with being near or past the expected recurrence time, which should push UP not DOWN.
  3. The information gap about whether 'territorial waters of California' includes parts of the Cascadia subduction zone is genuinely important and underweighted. The Cascadia subduction zone is capable of M9.0+ events with a ~10-15% probability per 50 years. If even a portion of this zone falls within California's territorial waters (the zone extends to Cape Mendocino in northern California), this could add 1-3% to the cumulative probability.
  4. The sub-question decomposition is acknowledged as poorly structured by the forecaster themselves — sq2 (capability at 92%) contributes nothing useful to the final estimate. This suggests the decomposition didn't effectively aid the analysis, though the forecaster correctly abandoned the mechanical combination.
  5. The forecaster's final estimate of ~4% relies heavily on UCERF3's 7%/30-year time-independent figure. However, UCERF3 was published in 2014 and the forecaster notes as an information gap that no updated model has been checked. The 2023 National Seismic Hazard Model update could potentially have revised these figures, representing a genuine unknown that should widen uncertainty bands.
  6. The treatment of the Cascadia subduction zone deserves more careful analysis. The southern terminus of the Cascadia subduction zone is at Cape Mendocino, which IS in California. A full Cascadia rupture (M9.0+) would have its epicenter likely in Oregon/Washington but the rupture zone extends into California. The question asks about earthquakes 'in California' — if this means any part of the rupture occurring within California's boundaries or territorial waters, Cascadia events would count, adding perhaps 2-3% over 9.75 years.
Suggested adjustment: +2pp
The forecaster's core reasoning from UCERF3 is sound, but they slightly underweight: (1) the time-dependent increase for the overdue southern San Andreas, (2) the possibility that Cascadia subduction zone events partially in California could qualify, and (3) they incorrectly treat the 250-year absence of M8.0+ as downward evidence when it's actually consistent with being near/past the recurrence interval. These factors collectively suggest a modest upward adjustment from 4% to ~6%. The massive market divergence (60%) is NOT a reason to adjust — the market shows classic signs of speculative mania with +41% in 30 days on thin volume.
Market comparison: Forecaster: 4%, Market: 60%. The divergence is extreme but the market shows clear signs of unreliability: LOW volume, SHALLOW orderbook, +41% surge in 30 days from a historical low of 4%. This price movement is inconsistent with any new seismological information and likely reflects speculative activity or thin liquidity. The forecaster's UCERF3-grounded estimate is far more credible than the market price. A modest upward adjustment to ~6% is warranted by reasoning corrections, not market conformity.
6. Final Forecast Opus
6% (75% confidence)
UCERF3 estimates 7% probability of M8.0+ in California over 30 years (~2.3% over 9.75 years). I adjust upward to ~6% after accepting the Devil's Advocate's corrections: the 335-year seismic gap on the southern San Andreas warrants stronger time-dependent adjustment (+1.5-2%), the 250-year absence of M8.0+ was incorrectly used as downward evidence when it's consistent with being at/past recurrence intervals, and the Cascadia subduction zone's southern terminus at Cape Mendocino adds ~0.5% for edge-case epicenter scenarios. The market at 60% is 10x my estimate, but I reject this price as unreliable: the +41% surge in 30 days from a 4% baseline with shallow orderbook depth and no new seismological information indicates speculative activity, not informed pricing. No credible seismological model supports anything near 60% cumulative probability over this timeframe.
7. Entry Decision Opus Trader
Side
NO
Position Size
$1800
Contracts
4000
Trader Confidence
75%
Trade Rationale
The edge here is enormous — nearly 49 percentage points. The forecaster estimates only 6% probability of an M8.0+ earthquake in California before 2035, while the market prices this at 55%. No credible seismological model supports anything close to 55% cumulative probability over ~9.75 years; UCERF3 gives ~2.3% for this timeframe and even generous adjustments for the San Andreas seismic gap and Cascadia edge cases only bring it to ~6%. The market price appears driven by speculative activity (a +41% surge from 4% baseline with no new seismological information). The sub-question forecasts strongly confirm: only 3% chance the cumulative probability exceeds 15%, and only 5% chance the annual rate is 1%+. I already hold a similar NO position on the Japan 8.0+ earthquake market, but these are largely uncorrelated seismological events. The Devil's Advocate only suggested a +2% adjustment which the forecaster already incorporated.
Allocation Logic
With a 49% edge and 0.75 forecaster confidence, this warrants near-maximum sizing. I'm going with $1800 rather than $2000 due to the very long time horizon (3556 days creates uncertainty about market dynamics and potential for interim volatility) and the low-moderate liquidity.
Entry price: $0.45
Current: $0.46
Status: OPEN
P&L: $40.00
Pipeline Timing
StepTime
Planning 26.16s
Research 54.54s
Synthesis 44.88s
Forecaster 27.92s
Critic 23.38s
Final 23.04s