Scan: Will there be an at least 8.0 magnitude earthquake in California before 2028?

← Back to scans

Will there be an at least 8.0 magnitude earthquake in California before 2028?

KXEARTHQUAKECALIFORNIA-28 · Climate and Weather · 2026-05-17

Agent

11%

Market Price

-6.0%

Edge

70%

Confidence

Volume: 6,130

Spread: 3.0c

Days to resolution: 959

Markets in event: 1

Final Rationale

UCERF3 gives ~7% over 30 years for M8.0+ in California, implying ~0.6% over the remaining 2.6 years, but this needs upward adjustment for time-dependence on the southern San Andreas (locked since 1857), magnitude boundary effects near M7.9-8.0, and resolution scope ambiguity. I partially accept the Devil's Advocate critique on challenges 1-3 (magnitude boundary and time-dependence deserve more weight than my initial +1.5% combined adjustment), but reject challenge 4's deference to the market given the questionable orderbook depth ($0.98 spread with both bids at $0.01). My final 5% sits above my independent 2.8% to incorporate these legitimate uncertainties, but well below the 11% market price, which appears to reflect illiquidity and residual overpricing being actively corrected (-7% in 30 days).

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
2026-05-10	3%	11%	75%
2026-05-02	4%	11%	70%
2026-04-22	2%	11%	78%

1. Planner Opus · Fermi decomposition + tool plan

Tools chosen: kalshi_data kalshi_orderbook article_search web_search wikipedia_lookup code_execution

Sub-questions (Fermi decomposition)

Is the annual probability of an M8.0+ earthquake in California approximately 1-3% based on historical seismicity and geological models? (w=0.50)
Has any fault system in California (San Andreas, Cascadia, or offshore faults) shown significantly elevated stress or precursor activity that would raise the probability of an M8.0+ event before 2028? (w=0.30)
Given the ~2.6-year window remaining until Dec 2028, does the cumulative probability of an M8.0+ California earthquake exceed 5%? (w=0.20)

Planner reasoning

This question requires estimating the probability of an M8.0+ earthquake in California (including territorial waters) over a ~2.6-year window ending Dec 31, 2028. Key reference classes and mechanisms: 1. USGS UCERF3 (Uniform California Earthquake Rupture Forecast) provides the most authoritative estimates. For M8.0+, the San Andreas Fault system has a recurrence interval of ~200-300 years for full ruptures. Annual probability of M8+ somewhere in California is roughly 1-2% per year from major fault systems. 2. Historical record: California's last M8+ was arguably the 1906 SF earthquake (~M7.9-8.3 depending on the estimate). True M8.0+ events in California proper are extremely rare — roughly once per 150-300 years. 3. The question includes 'territorial waters' which could include offshore faults like the Cascadia Subduction Zone's southern edge and other offshore segments. 4. Temporal analysis: With ~959 days (~2.63 years) remaining, using a Poisson model with annual rates of 0.7-2%, the cumulative probability is roughly 1.8-5%. The question requires a specific threshold (8.0 Mw) which is quite high — most 'big ones' discussed in California context are M7.8-8.0 range which may not qualify. 5. What makes this different: M8.0 is a very high bar. The oft-cited 'Big One' on the southern San Andreas is typically modeled as ~M7.8-8.0, so whether it counts depends on exact rupture. Cascadia full ruptures could be M9+ but their southern edge touching California is uncertain for epicenter purposes. 6. Prediction markets should reflect the base rate of ~2-5% for this window. Any significant recent seismic activity would be a key update.

2. Research Parallel tool calls · legacy v2

Tool	Status	Time	Items
kalshi_data	OK	2.2s	-
kalshi_orderbook	OK	0.2s	-
article_search	OK	0.8s	-
web_search	OK	54.6s	-
wikipedia_lookup	OK	2.1s	-
code_execution	OK	0.0s	-

3. Evidence Synthesis Sonnet · 14 items · legacy v2

#	Strength	Credibility	Direction	Source	Claim	Priced In
1	STRONG	92	↓ DOWN	web_search	UCERF3 estimates a 7% probability of an M8.0+ earthquake in California in the next 30 years, implying an annual probability of approximately 0.24% per year (~1-in-420 odds), far below the 1-3% range suggested in the sub-question.	Yes
2	STRONG	88	↑ UP	web_search	The southern San Andreas Fault has not produced a major rupture since the 1857 Fort Tejon earthquake (M7.9), meaning ~168 years of stress accumulation; UCERF3 identifies it as the most probable source for an M8.0+ event.	Yes
3	MODERATE	75	↓ DOWN	code_execution	Using a Poisson model with UCERF3-implied annual rates (~0.24%), cumulative probability over 2.63 years is roughly 0.6%; using higher assumed annual rates of 1-2%, the range is 2.6-5.1%.	Yes
4	MODERATE	80	↓ DOWN	wikipedia_lookup	The 1906 San Francisco earthquake had an estimated magnitude of 7.9 (some estimates up to 8.3), illustrating that California's historical M8.0+ record is sparse and recurrence intervals are very long.	Yes
5	WEAK	55	NEUTRAL	article_search	A swarm of up to M4.5 earthquakes struck Imperial County near the Salton Sea (on the San Andreas system) in May 2026, prompting ongoing monitoring but no M8.0 precursor designation.	No
6	WEAK	55	NEUTRAL	article_search	In January 2026, a M4.9 earthquake near Indio (close to San Andreas) triggered a series of aftershocks; renowned seismologist Dr. Lucy Jones noted more temblors could follow due to proximity to the San Andreas Fault, but no M8.0+ precursor was identified.	Yes
7	WEAK	50	↑ UP	article_search	An article from May 2025 highlights that the Cascadia Subduction Zone (running from Northern California to British Columbia) is 'due to rupture — possibly in our lifetimes,' but no specific near-term elevated stress indicators were cited for California's portion.	Yes
8	MODERATE	60	↓ DOWN	article_search	No evidence found of any significant, scientifically validated precursor activity (e.g., elevated foreshock rates, GPS strain anomalies, or official warnings) for an imminent M8.0+ event in California as of mid-2026.	Yes
9	WEAK	65	NEUTRAL	article_search	A July 2025 M8.8 earthquake struck Russia's Kamchatka Peninsula, sending tsunami waves to the U.S. West Coast; this event was not in California and did not directly affect California fault stress levels.	Yes
10	STRONG	80	↓ DOWN	code_execution	Using the UCERF3-implied annual rate of ~0.24%, cumulative probability over 2.63 years is ~0.6%; even with a generous 1-2% annual rate assumption, the range is only 2.6-5.1%, making 5% a borderline upper-bound scenario.	Yes
11	MODERATE	60	↑ UP	kalshi_data	The Kalshi market for 'M8.0+ California earthquake before 2028' is currently priced at 11%, which is significantly higher than the UCERF3-derived base rate of ~0.6-5% for this window.	Yes
12	MODERATE	58	NEUTRAL	kalshi_data	The Kalshi market for 'M8.0+ California earthquake before 2035' is priced at 32%, while the 2028 market is at 11%; this implies roughly 21% probability assigned to the 2028-2035 window (7 years), which is internally consistent with low annual rates.	Yes
13	MODERATE	62	↓ DOWN	kalshi_data	The 2028 market has dropped 7% over the past 30 days and 5% in the past 7 days, suggesting a significant downward revision in market sentiment, possibly reflecting recalibration after prior overpricing.	No
14	MODERATE	50	NEUTRAL	kalshi_orderbook	The orderbook shows a spread of $0.98 with both yes_bid and no_bid at $0.01, suggesting very thin liquidity at the extremes despite a nominally HIGH depth label; this raises concerns about market efficiency and reliability of the 11% price.	Yes

Information Gaps

No current USGS or SCEC data on recent strain accumulation or GPS-measured deformation rates along the San Andreas or Cascadia faults to assess whether stress has elevated above baseline.
No information on whether UCERF3 has been updated (UCERF4 development was underway) that might revise M8.0+ annual probability estimates up or down.
No clarity on whether Cascadia Subduction Zone southern segment epicenters would qualify as 'California territorial waters' for resolution purposes — this is a significant scope question.
No data on the exact resolution criteria Kalshi uses (e.g., which magnitude scale, which monitoring agency) that could affect whether a borderline M7.9-8.0 event resolves YES.
No detailed seismic network data showing current microseismicity trends along major California fault segments that might provide updated stress state information.
The historical annual rate calculation in code_execution uses '0.0077' — the basis for this figure is unclear and appears higher than UCERF3 implies; the source of this rate should be clarified.

Key Uncertainties

Whether the Cascadia Subduction Zone's southern terminus qualifies as California territorial waters — a full Cascadia rupture (M9+) would almost certainly resolve YES if its epicenter counts.
Whether a borderline M7.9 event on the southern San Andreas would be measured as M8.0+ by USGS (due to measurement uncertainty and possible upward revision), which could affect resolution.
The Kalshi market at 11% is roughly 2-10x higher than the UCERF3 base rate implies; it's unclear whether this reflects informed trader beliefs, liquidity premiums, or overestimation of risk.
Whether multi-fault rupture scenarios (now included in UCERF3) significantly raise the true annual probability above the naive single-fault interpretation of 0.24% per year.
The time-dependency of earthquake probability: the longer since the last major rupture (1857 for S. San Andreas, 1700 for Cascadia), the higher the conditional probability — UCERF3 partially accounts for this but uncertainty remains high.
Trump administration cuts to USGS earthquake monitoring capacity (noted in May 2025 article) could create information gaps in detecting precursor activity.

Market Context

Price: 0.11 · 7d: -0.05 · 30d: -0.07
Volume: LOW-MODERATE — 133 avg daily contracts is thin for a binary event market · Depth: QUESTIONABLE — nominal HIGH depth (11860) but yes_bid and no_bid both at $0.01 with $0.98 spread suggests illiquidity at current price level
Reliability: LOW-MEDIUM — market price of 11% appears significantly above UCERF3-derived base rates (~0.6-5%); recent sharp downward trend (-7% in 30 days) suggests prior overpricing being corrected; thin volume limits confidence in price as a true probability signal

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

Independent estimate: 3% (77% confidence)

base rate

75% conf

evidence driven

78% conf

contrarian

78% conf

Spread

0pp

Low disagreement

sq1: Is the annual probability of an M8.0+ earthquake in California approximately 1-3%?

15%

UCERF3 implies ~0.24% annual probability (7% over 30 years), which is well below the 1-3% range. Even accounting for elapsed time since 1857 Fort Tejon (~168 years) and time-dependent models, the annual rate remains likely under 1%. The 1-3% range overstates the true rate. Probability that the rate falls in this elevated range: ~15%.

sq2: Has any fault system shown significantly elevated stress or precursor activity?

No scientifically validated precursors identified. Minor swarms in Imperial County and near Indio are routine background seismicity. Cascadia 'due' claims are long-term, not near-term. The Kamchatka M8.8 event was geographically unrelated. No GPS strain anomalies or foreshock acceleration reported. Probability that meaningful precursor elevation exists: ~8%.

sq3: Does cumulative probability over 2.6 years exceed 5%?

10%

UCERF3-implied annual rate of 0.24% gives ~0.6% cumulative over 2.63 years. Even with time-dependent boosting and multi-fault rupture inclusion, reaching 5% would require an annual rate near 2%, which exceeds most published estimates. The probability that the true cumulative exceeds 5% is low: ~10%.

Show full reasoning trace

base rate: UCERF3 gives 7% over 30 years for M8.0+ in California, implying ~0.24% annual rate. Over 2.63 years remaining, Poisson cumulative ≈ 0.6%. California has zero confirmed M8.0+ in instrumental history (1906 SF was M7.9, 1857 Fort Tejon was M7.9).

evidence updates: Time-dependence since 1857 on southern San Andreas pushes UP slightly (+0.5%). Magnitude measurement uncertainty at the 7.9-8.0 boundary pushes UP (+1%). Inclusion of Cascadia southern edge as 'territorial waters' adds small UP (+0.5%). No precursor evidence keeps probability near base rate. Recent swarms are routine and not predictive.

combination method: Direct estimation from base rate plus adjustments, since sub-questions are characterizations rather than independent event probabilities.

final: Final estimate ~3%, consistent with prior forecasts (2-4%) and UCERF3-implied physics. Confidence is high (0.75) because the base rate is well-established and no new precursor evidence exists.

ensemble: {'ensemble_count': 3, 'perspectives': {'base_rate': {'p_yes': 0.03, 'confidence': 0.75}, 'evidence_driven': {'p_yes': 0.025, 'confidence': 0.78}, 'contrarian': {'p_yes': 0.03, 'confidence': 0.78}}, 'spread': 0.005, 'individual_p_yes': {'base_rate': 0.03, 'evidence_driven': 0.025, 'contrarian': 0.03}}

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

Overconfidence Asymmetric Evidence Missing Info

Challenges

The forecaster's base rate may be too low. UCERF3's 7% over 30 years is for M8.0+, but this is a long-term time-averaged rate. Time-dependent models (UCERF3-TD) for the southern San Andreas, which has been locked since 1857 (~169 years vs. typical 100-200 year recurrence), give higher conditional probabilities. The forecaster acknowledges this but only adds +0.5%, which may understate the time-dependence effect.
Magnitude boundary uncertainty deserves more weight. The 1906 SF earthquake has magnitude estimates ranging from 7.9 to 8.3 depending on methodology. A future event in the M7.8-8.0 range has meaningful probability of being officially cataloged as M8.0+. The forecaster adds only +1% for this, but the historical near-misses suggest this boundary effect could be larger — perhaps 2-3%.
The resolution scope ambiguity (Cascadia southern segment in California territorial waters, offshore events, magnitude scale used) is flagged as an information gap but only adds +0.5%. Given Kalshi's resolution practices often favor broader inclusion, and that any M8.0+ along the Cascadia southern terminus could plausibly qualify, this could add more probability mass.
The forecaster effectively dismisses the market at 11% as 'overpriced,' but the market has thousands of dollars of depth and incorporates trader views on resolution criteria, magnitude uncertainty, and time-dependence that may not be fully captured in UCERF3's point estimate. While the market trend is downward (suggesting correction), settling at 11% rather than near 1% suggests informed traders see structural reasons for higher probability.
Sub-question 3's framing is somewhat circular: asking whether cumulative probability exceeds 5% and answering 10% essentially restates the base rate conclusion. This doesn't add independent information and may create false confidence through redundant reasoning.

Suggested adjustment: +2pp

Modest upward adjustment warranted because: (1) magnitude boundary uncertainty at M7.9-8.0 historical events deserves more weight than +1%, (2) time-dependence on southern San Andreas after 169 years is likely undercounted, (3) resolution scope ambiguity around Cascadia/offshore events adds genuine probability mass. These are reasoning gaps independent of the market signal. A forecast around 4-5% better captures these uncertainties while still remaining well below the market.

Market comparison: Forecaster: 2.8%, Market: 11%. The divergence is large but partially justified — UCERF3 base rates do support a low probability, and the market shows downward correction (-7% in 30d) suggesting prior overpricing. However, market at 11% with reasonable depth likely reflects legitimate considerations (magnitude boundary, time-dependence, resolution scope) that the forecaster underweights. Truth probably lies between 4-8%.

6. Aggregator Opus · final calibrated forecast

5% (70% confidence)