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

← Back to scans

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

KXEARTHQUAKECALIFORNIA-35 · Climate and Weather · 2026-05-08

Agent

31%

Market Price

-22.0%

Edge

62%

Confidence

Volume: 19,242

Spread: 1.0c

Days to resolution: 3523

Markets in event: 1

Final Rationale

My independent estimate of 6.2% anchored on UCERF3's ~2.3% Poisson decadal baseline plus adjustments for southern San Andreas time-dependent renewal (~345 years elapsed vs 150-200 year mean) and territorial waters inclusion. I PARTIALLY ACCEPT the critique on Cascadia contribution and definitional ambiguity at the M7.9-8.0 boundary (1857 and 1906 events sit right at this threshold with magnitude uncertainty), warranting an upward shift to ~9%. I REJECT meaningful deference to the 31% market price: with 36 contracts/day, an 8-73% historical range, and a +9% jump after the July 2025 Kamchatka M8.8 (a different plate boundary with no California stress-transfer mechanism), the market is attention-driven rather than informed. I deviate sharply from the market because UCERF3 represents a consensus scientific model and no published seismological evidence justifies a 5x multiplier on its base rate.

Price vs. prediction

How the market moved after the agent's call. Solid blue line is the actual market price; dashed line is the agent's static forecast at $t{=}0$.

$n = 58$ follow-ups

Re-scan Context

This market has been scanned before. Previous predictions:

Date	Predicted	Market Price	Confidence
2026-05-01	9%	34%	60%
2026-04-05	6%	55%	75%

1. Planner Opus · Fermi decomposition + tool plan

Tools chosen: article_search web_search wikipedia_lookup kalshi_data code_execution

Sub-questions (Fermi decomposition)

Is California's seismic environment capable of producing an M8.0+ earthquake within any given 10-year window based on historical and geological evidence? (w=0.35)
Do current seismic hazard models (UCERF3 or similar) assign a probability of at least 5% for an M8.0+ earthquake in California within the next ~10 years? (w=0.40)
Has there been any significant increase in seismic activity or precursor events in California since 2024 that meaningfully elevates near-term M8.0+ risk? (w=0.25)

Planner reasoning

This question requires estimating the probability of a rare but geologically plausible extreme seismic event. The key reference class is the UCERF3 (Uniform California Earthquake Rupture Forecast, 3rd edition) published by USGS/CGS, which provides official probabilistic seismic hazard estimates for California faults. M8.0+ earthquakes in California are extremely rare in the instrumental record (only the 1906 San Francisco earthquake ~M7.9 comes close), but paleoseismic evidence shows they have occurred on the San Andreas fault system historically. UCERF3 estimates roughly a 7% probability of M8+ in 30 years, implying about a 2-3% probability in the ~9.65 years remaining until resolution. The causal mechanism is fault stress accumulation on the southern San Andreas, Cascadia subduction zone edge, and other major faults. What makes this different from a generic earthquake question: (1) M8.0 is a very high threshold — M7.8-7.9 events are far more likely but would not resolve YES; (2) territorial waters inclusion adds Cascadia subduction zone relevance; (3) the 9.65-year window is relatively short for such rare events. Temporal analysis: For YES to occur, a major fault rupture exceeding 300+ km would need to occur somewhere in California or its waters before Dec 31, 2035. The southern San Andreas is the most likely candidate (overdue per paleoseismic intervals), but still has low annual probability. The Poisson model is appropriate given the memoryless nature of earthquake occurrence at these timescales. Market price from Kalshi should provide useful crowd-sourced calibration.

2. Research Parallel tool calls · legacy v2

Tool	Status	Time	Items
article_search	OK	0.6s	-
web_search	OK	55.0s	-
wikipedia_lookup	OK	2.1s	-
kalshi_data	OK	2.0s	-
code_execution	OK	0.0s	-

3. Evidence Synthesis Sonnet · 16 items · legacy v2

#	Strength	Credibility	Direction	Source	Claim	Priced In
1	STRONG	88	↑ UP	wikipedia_lookup	The San Andreas Fault extends ~750 miles through California with average slip rates of 0.79–1.38 inches/year, and paleoseismic evidence confirms it has produced M8+ earthquakes historically. The 1906 San Francisco earthquake (~M7.9) is the closest modern analog.	Yes
2	STRONG	92	↑ UP	web_search	UCERF3 estimates M8+ earthquakes in California occur on average every 494 years (previously 617 in UCERF2), confirming geological capability for such events within any given multi-decade window.	Yes
3	MODERATE	72	↑ UP	article_search	A May 2025 article notes the Cascadia subduction zone (which intersects Northern California at the Mendocino triple junction) is 'due to rupture' and could produce the largest earthquake ever seen in the US, confirming M8+ capability in California's territorial waters.	Yes
4	MODERATE	85	↑ UP	wikipedia_lookup	The Cascadia subduction zone intersects the San Andreas fault at the Mendocino triple junction (offshore Northern California), and a major Cascadia rupture could trigger San Andreas rupture — these faults are within California's territorial waters scope.	Yes
5	STRONG	93	↓ DOWN	web_search	UCERF3 assigns ~7% probability of M8.0+ earthquake in California in the next 30 years (up from 4.7% in UCERF2). A Poisson scaling to ~9.65 years yields approximately 2.3% probability.	Yes
6	STRONG	88	↓ DOWN	code_execution	Poisson model using UCERF3's 30-year 7% base rate yields P(M8+ in 9.65 years) = 2.31%. Even with an optimistic 10% 30-year rate, the 9.7-year probability is only ~3.3%, well below the 5% threshold in sq2.	Yes
7	MODERATE	80	↓ DOWN	web_search	UCERF3 does not publish a standalone 10-year figure for M8.0+ statewide; the 30-year figure must be scaled. No updated model superseding UCERF3 (e.g., UCERF4) has been found in research, meaning the 2015 model remains the authoritative reference.	Yes
8	MODERATE	45	↑ UP	kalshi_data	Kalshi market for M8.0+ in California before 2028 (~2.5-year window) is priced at 16%, which if taken at face value implies an implausibly high annual rate (~7%/yr) inconsistent with UCERF3. This suggests significant market overpricing relative to scientific models, likely due to retail trader anchoring bias.	Yes
9	MODERATE	42	↑ UP	kalshi_data	Kalshi market for M8.0+ in California before 2035 is priced at 31%, substantially above the ~2.3% Poisson estimate from UCERF3. The 30-day price increase of +9% suggests recent upward pressure, possibly from the July 2025 Russia M8.8 earthquake generating attention.	No
10	MODERATE	55	↓ DOWN	kalshi_data	Japan M8.0+ before 2030 market is at 58% — dramatically higher than California's 31% for 2035, consistent with Japan's much higher base seismic rate (multiple M8+ events in instrumental record vs. zero confirmed M8.0+ for California).	Yes
11	WEAK	65	NEUTRAL	article_search	A 4.9M earthquake struck near Indio Hills (Southern California, near San Andreas fault) in January 2026 with aftershocks, but this is routine seismic activity and not indicative of elevated M8+ risk.	Yes
12	WEAK	62	NEUTRAL	article_search	A series of M3.7-3.8 earthquakes struck San Ramon (East Bay) in November 2025 — routine low-level seismic activity with no expert commentary suggesting elevated large-earthquake risk.	Yes
13	WEAK	65	NEUTRAL	article_search	The Geysers geothermal field in Northern California (August 2025) continues experiencing induced seismicity, but these are small, induced events unrelated to tectonic stress buildup on major faults.	Yes
14	MODERATE	70	↓ DOWN	article_search	No articles found documenting any significant seismic precursor events, unusual strain measurements, or expert warnings of elevated near-term M8+ risk specifically in California since 2024.	Yes
15	MODERATE	78	NEUTRAL	article_search	The July 2025 Russia M8.8 Kamchatka earthquake sent tsunamis to California coast but did not epicenter in California. It may have triggered increased public and market attention to California seismic risk without actually elevating it scientifically.	No
16	WEAK	72	NEUTRAL	article_search	Trump administration FEMA cuts to California earthquake retrofitting ($33M canceled) do not affect the probability of an M8+ earthquake occurring, only preparedness for its impact.	Yes

Information Gaps

No data on whether UCERF4 or any updated California seismic hazard model has been published since UCERF3 (2015), which could change baseline probabilities
No data on current Coulomb stress loading on the southern San Andreas fault or paleoseismic interval recurrence data updated post-2020
No data on whether any recent GPS/InSAR strain measurements show anomalous locking or strain accumulation on key California faults
No expert seismologist commentary specifically addressing whether 2024-2026 California seismicity patterns represent any elevated concern
No data on the precise definition of 'territorial waters' used in this market — whether Cascadia subduction zone ruptures offshore Northern California would qualify is unresolved and could significantly affect probability
No information on whether the 30-day Kalshi price increase (+9%) corresponds to a specific triggering event (e.g., the Russia M8.8 in July 2025) or reflects fundamental reassessment
Absence of any UCERF3 time-dependent (renewal model) probability specifically for the southern San Andreas, which could be higher than the Poisson estimate given elapsed time since last major rupture

Key Uncertainties

The territorial waters definition: if Cascadia subduction zone offshore Northern California counts, the base rate could be meaningfully higher than the pure California land/SAF estimate
Whether UCERF3's Poisson vs. renewal model should be used — the renewal model for the southern San Andreas (which has not ruptured since ~1680) would assign higher probability than Poisson
Kalshi market price of 31% is ~13x higher than the Poisson model estimate (~2.3%) — this large gap raises questions about whether the market reflects information not captured in UCERF3, or significant miscalibration
Model uncertainty in UCERF3 itself: the 7% 30-year figure has substantial uncertainty bands, and some fault-specific estimates (e.g., San Jacinto) have increased M8+ likelihood
Whether the Russia M8.8 July 2025 earthquake could have induced any stress changes on Pacific plate boundaries affecting California fault systems

Market Context

Price: 0.31 · 7d: -0.02 · 30d: +0.09
Volume: LOW · Depth: SHALLOW
Reliability: LOW — very low daily volume (36 contracts/day), extremely wide historical price range (8%-73%) suggesting thin liquidity and high susceptibility to anchoring bias. The 31% price is likely substantially overpriced relative to scientific base rates (~2-3%). The related 2028 market at 16% also appears inconsistent with UCERF3 at a much shorter timeframe. The 30-day +9% increase likely reflects the July 2025 Russia M8.8 attention effect rather than fundamental probability reassessment.

4. Independent Forecaster Opus · No market price · legacy v2 Did NOT see market price

Independent estimate: 6% (67% confidence)

base rate

70% conf

evidence driven

70% conf

contrarian

60% conf

Spread

2pp

Low disagreement

sq1: Is California's seismic environment capable of producing an M8.0+ earthquake within any given 10-year window?

92%

This is a capability question, not an occurrence question. Paleoseismic evidence on the San Andreas, the Cascadia subduction zone offshore Northern California, and historical near-M8 events (1857 Fort Tejon, 1906 SF) confirm M8+ is geologically possible. The capability is essentially certain; the only reason not to set this at 0.99 is interpretive ambiguity about whether 'capable within a 10-year window' means 'has nontrivial probability' (yes) or 'expected' (no). 92%.

sq2: Do current seismic hazard models assign ≥5% probability of M8.0+ in California in next ~10 years?

UCERF3 gives ~7% over 30 years, which Poisson-scales to ~2.3% over 9.65 years. Even with time-dependent renewal model adjustments for the southern San Andreas quiescence, getting to 5% over a decade requires aggressive assumptions. No UCERF4 has been published. The official answer is clearly NO. Small upward probability (~8%) accounts for any obscure model variant or fault-specific estimate that might cross 5%.

sq3: Has there been a significant increase in seismic activity/precursors since 2024 elevating near-term M8.0+ risk?

Evidence consistently shows routine background seismicity (M3-5 events) with no expert warnings of elevated near-term M8+ risk. The Russia M8.8 Kamchatka event is on a different plate boundary segment with no established stress-transfer mechanism to California faults. No anomalous strain measurements reported. This is firmly NO.

Show full reasoning trace

base rate: UCERF3 gives 7% over 30 years for M8+ in California, Poisson-scaling to ~2.3% over 9.65 years. Historical instrumental record: zero confirmed M8.0+ in California (1906 was ~M7.9, 1857 ~M7.9). Paleoseismic recurrence ~494 years for statewide M8+ events.

evidence updates: Upward adjustments: (1) Time-dependent renewal models for southern San Andreas — last rupture ~1680, ~345 years elapsed vs ~150-200 year mean recurrence — adds ~1.5-2.5%. (2) Territorial waters inclusion brings in Cascadia subduction zone (Mendocino triple junction area), adds ~1%. (3) Definitional ambiguity (1857/1906 were borderline M8) adds slight tail risk. Downward: No UCERF4 supersession, no precursor signals, no expert warnings of elevated near-term risk. The Russia M8.8 is not a credible stress-transfer to California.

combination method: Mechanically applying weighted average gives 0.37, but this conflates capability (sq1) with occurrence likelihood. The honest forecast must anchor on actual occurrence probability per UCERF3 Poisson + reasonable adjustments, landing around 5-7%.

final: Final estimate ~6%. UCERF3 baseline ~2.3% + time-dependent renewal adjustment +2-3% + territorial waters/definitional tail +1% ≈ 5-7%. This is consistent with my prior forecasts (6-9%). Confidence is moderate-high given strong scientific anchoring, though the gap with prediction markets suggests I may be missing something or markets may be miscalibrated on rare-event tail risk.

ensemble: {'ensemble_count': 3, 'perspectives': {'base_rate': {'p_yes': 0.06, 'confidence': 0.7}, 'evidence_driven': {'p_yes': 0.055, 'confidence': 0.7}, 'contrarian': {'p_yes': 0.07, 'confidence': 0.6}}, 'spread': 0.015, 'individual_p_yes': {'base_rate': 0.06, 'evidence_driven': 0.055, 'contrarian': 0.07}}

5. Devil's Advocate Sonnet · legacy v2 Market price revealed

Reasoning Flaw Missing Info

Challenges

The forecast may be too low if 'California' includes territorial waters and Cascadia ruptures count. The Cascadia subduction zone produces M8-9 events with ~10% probability in 50 years (~2% per decade just for Cascadia alone), which would stack on top of the San Andreas estimate. The forecaster only adds ~1% for this, which may understate it if the question's resolution criteria credit offshore Northern California ruptures.
The time-dependent renewal model adjustment (+2-3%) may be too modest. The southern San Andreas has gone ~345 years since last rupture against a mean recurrence of 150-200 years — Field et al. UCERF3-TD gives elevated probabilities for some sections that, when properly aggregated, could push the 10-year statewide M8+ probability higher than 5-7%.
The forecaster acknowledges the mechanical weighted average gives 0.37 but rejects it as conflating capability with occurrence — this is correct, but the rejection is somewhat hand-wavy. The decomposition itself may be poorly designed since sq1 (capability) carries little forecasting information, making the decomposition non-additive.
Definitional ambiguity around M8.0 is underweighted. The 1857 Fort Tejon and 1906 SF events are estimated at M7.9 with uncertainty bands that include M8.0+. If a future event lands at estimated M7.9-8.0, resolution could go either way, and modern instrumentation/re-analysis often revises magnitudes upward. This adds tail risk beyond the +1% noted.
Market dismissal may be too quick. While the market is thin (LOW reliability), markets across multiple correlated contracts (16% for 2028, 31% for 2035, Japan at 58%) show internal consistency suggesting traders systematically price seismic tail risk higher than UCERF3 Poisson — this could reflect legitimate model uncertainty about UCERF3 itself, not just anchoring bias.

Suggested adjustment: +3pp

Modest upward adjustment (from 6.2% to ~9-10%) warranted because: (1) Cascadia contribution may be underweighted if territorial waters count, (2) time-dependent renewal for southern San Andreas could be more aggressive than +2-3%, (3) definitional ambiguity at M7.9-8.0 boundary adds genuine tail risk, (4) UCERF3 itself has model uncertainty that point estimates ignore. This is not adjustment toward the market — the market at 31% is likely overpriced — but the forecaster's 6.2% may be slightly under-accounting for legitimate tail-risk factors.

Market comparison: Forecaster: 6.2%, Market: 31%. The divergence is large but the market is thinly traded (LOW reliability, 36 contracts/day, 8-73% historical range) and likely reflects attention-driven pricing after the July 2025 Kamchatka M8.8. The forecaster's scientific anchoring on UCERF3 is sound. However, the forecaster's own information gaps (no UCERF4, no updated strain data, definitional ambiguity on territorial waters) suggest somewhat wider uncertainty than 6.2% conveys. A small upward adjustment is warranted on reasoning grounds, not market conformity.

6. Aggregator Opus · final calibrated forecast

9% (62% confidence)