← Back to scans

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

KXEARTHQUAKECALIFORNIA-28 · Climate and Weather · 2026-05-10
3%
Agent
11%
Market Price
-8.0%
Edge
75%
Confidence
Volume: 6,094
Spread: 3.0c
Days to resolution: 966
Markets in event: 1
Final Rationale
UCERF3 implies ~0.6-1% probability of M8.0+ in California over the remaining ~2.6-year window, with California having zero confirmed M8.0+ in instrumental history (1906 SF was Mw 7.9). I partially accept the Devil's Advocate critique on magnitude-boundary uncertainty — events like 1857 Fort Tejon and 1906 cluster right at the Mw 7.9 threshold, so I nudge up slightly to ~3% to account for potential post-event magnitude revisions and Cascadia/offshore boundary ambiguity. I reject deference to the 16% market price: with 86 contracts/day volume, a $0.98 bid-ask spread, and a 2-22% historical range, this market is a noise generator rather than a calibrated signal, and tracks public earthquake anxiety rather than seismological base rates. My deviation from the market is justified by direct reliance on USGS UCERF3 hazard modeling.
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:
DatePredictedMarket PriceConfidence
2026-05-02 4% 11% 70%
2026-04-22 2% 11% 78%
1. Planner Opus · Fermi decomposition + tool plan
Tools chosen: wikipedia_lookup article_search web_search kalshi_data kalshi_orderbook code_execution
Sub-questions (Fermi decomposition)
  1. Is the base rate of M8.0+ earthquakes in California consistent with a >5% probability over any given ~2.6-year window? (w=0.50)
  2. Are there currently elevated seismic indicators or recent swarm activity in California that would increase the near-term probability of a M8.0+ event above the historical base rate? (w=0.30)
  3. Does the current state of major California fault systems (San Andreas, Cascadia subduction proximity) suggest a meaningfully elevated probability of a M8.0+ rupture before 2028? (w=0.20)
Planner reasoning
This question requires applying seismic hazard modeling to a specific geographic area and time window. The key reference class is the USGS/UCERF3 California earthquake hazard model, which provides probabilistic estimates for M8.0+ events on California fault systems. Critical distinctions: (1) M8.0+ is a very high threshold — California has had only one or possibly two events at or above this magnitude in recorded history (1906 SF earthquake, debated at ~7.9-8.25); (2) The relevant fault systems include the southern San Andreas (capable of ~M8.2), the Cascadia subduction zone (but its epicenters typically fall off the Oregon/Washington coast, not California), and potentially offshore faults; (3) The time window is ~2.64 years (May 2026 to Dec 31, 2028). Base rate approach: UCERF3 estimates roughly 7% probability of M8.0+ in California over 30 years, implying an annual rate of ~0.0024. Over 2.64 years, Poisson probability ≈ 0.63%. This is quite low. Temporal analysis: For YES to resolve, a M8.0+ earthquake would need to occur within ~966 days with epicenter in California or its territorial waters. There is no predictive capability for specific earthquake timing; this is purely a probabilistic/base rate problem. No sequence of events 'leads up' to this — it either occurs or doesn't, with probability governed by the underlying seismic rate. What makes this question different: The strict M8.0 threshold is critical. California averages much more frequent M6-7 earthquakes, but M8+ events have a recurrence interval of centuries on most fault segments. The territorial waters inclusion matters (offshore faults like the Cascadia, though that zone's epicenters are typically in Oregon/Washington). Market pricing should be anchored near 2-4% given the short time window and rare base rate.
2. Research Parallel tool calls · legacy v2
ToolStatusTimeItemsSummary
wikipedia_lookup OK 2.1s -
article_search OK 0.6s -
web_search OK 66.7s -
kalshi_data OK 2.3s -
kalshi_orderbook OK 0.2s -
code_execution OK 0.0s -
3. Evidence Synthesis Sonnet · 12 items · legacy v2
#StrengthCredibilityDirectionSourceClaimPriced In
1 STRONG 85 ↓ DOWN code_execution UCERF3-based Poisson calculation yields a ~0.64% probability of M8.0+ in California over the ~2.64-year window (May 2026–Dec 2028), with a range of 0.36%–1.12% depending on the base rate assumption used. Yes
2 STRONG 80 ↓ DOWN wikipedia_lookup The 1906 San Francisco earthquake is estimated at Mw 7.9 (not 8.0+), suggesting California has had zero confirmed M8.0+ events in recorded instrumental history, consistent with a very long recurrence interval for such events. Yes
3 MODERATE 75 ↓ DOWN wikipedia_lookup The San Andreas Fault is a ~750-mile continental strike-slip fault with slip rates of 0.79–1.38 inches/year; its southern locked segment is capable of ~M8.2, but recurrence intervals are on the order of centuries. Yes
4 MODERATE 45 ↑ UP kalshi_data The Kalshi market for M8.0+ in California before 2035 (a ~9-year window) is priced at 31%, implying roughly 4% per year or ~10% over 2.6 years — substantially higher than the UCERF3 Poisson estimate of 0.64%. Yes
5 MODERATE 65 NEUTRAL web_search Multiple earthquake swarms occurred in California in late 2025–early 2026, including a prolonged tectonic swarm near San Ramon (Nov 2025–Feb 2026+) with ~90 small quakes in November and a M3.8 event, and a Coachella Valley swarm near the southern San Andreas with a M4.9 mainshock in January 2026. Yes
6 STRONG 72 ↓ DOWN web_search Scientists explicitly state that swarms can only slightly increase the statistical likelihood of a larger earthquake by a very small amount, and California's past swarms have not reliably preceded M8+ events. Yes
7 STRONG 70 ↓ DOWN web_search San Ramon swarms are historically recurrent (occurring in 1970, 1976, 1990, 2002, 2003, 2015, 2018, and now 2025), with no volcanic component, confirming these are routine tectonic swarms rather than anomalous precursors. Yes
8 MODERATE 68 ↓ DOWN article_search Recent California seismic events in 2025–2026 are all in the M3–M5 range (M4.6 near Santa Cruz, M4.0 near Santa Rosa, M4.9 near Indio), which is normal background seismicity — none approaching the M6–7+ threshold that might more meaningfully precede an M8+ event. Yes
9 MODERATE 70 ↑ UP wikipedia_lookup The San Andreas Fault's southern locked segment (Coachella/San Bernardino) has not ruptured in ~330 years and is considered overdue for a major earthquake, potentially up to M8.2, though 'overdue' is not a precise predictor of imminent occurrence. Yes
10 MODERATE 60 ↓ DOWN article_search No recent articles report any USGS or scientific advisories, geodetic anomalies, or fault-locking changes that would signal an elevated near-term probability of M8+ rupture on California fault systems. Yes
11 MODERATE 40 ↑ UP kalshi_data The Kalshi market for M8.0+ in California before 2028 is priced at 16%, which is dramatically higher (~25x) than the UCERF3 Poisson base rate of ~0.64%, suggesting significant market overpricing or resolution criteria ambiguity. Yes
12 MODERATE 55 NEUTRAL kalshi_orderbook The orderbook shows very wide spread ($0.01 bid vs $0.99 implied ask, spread of $0.98) despite high nominal depth of 14,068 contracts, suggesting low genuine two-sided liquidity and poor price discovery — market price of 16% should be treated with low confidence. Yes
Information Gaps
  • No UCERF3 (or UCERF3.3) official probability tables were directly retrieved to confirm the exact 30-year M8.0+ probability for California; the 7% figure used in calculations is from secondary sources and the reasoning prompt itself.
  • No data on current GPS/geodetic strain accumulation rates on the southern San Andreas or other major locked fault segments that might indicate stress loading above historical averages.
  • No information on whether the Cascadia Subduction Zone's southernmost extent (northernmost California) could produce an M8+ with an epicenter in California territorial waters, and what probability UCERF3 assigns to that scenario.
  • No expert seismological commentary specifically addressing the near-term (2026–2028) probability of M8+ on California faults was found in the articles.
  • The Kalshi market price history (range 2%–22%) is unusual and its drivers are unclear — no information on what caused the historical spike to 22% or the current elevated price of 16%.
  • Lack of direct UCERF3 fault-specific rupture probabilities (e.g., probability of full southern SAF rupture in next 30 years) which would allow better segmentation of the base rate.
Key Uncertainties
  • The extreme gap between the Kalshi market price (16%) and the scientifically-derived base rate (~0.64–1.12%) is a critical unresolved tension — either the market is dramatically overpriced due to behavioral biases, or there is a resolution criteria interpretation (e.g., including magnitude estimates that are later revised) that inflates the apparent probability.
  • Whether the Cascadia Subduction Zone's southernmost segment could produce an M8.0+ with an epicenter technically within California's territorial waters — this could materially change the base rate.
  • The San Andreas southern segment being 'overdue' is real but scientifically contested as a near-term predictor; the 2.6-year window captures perhaps 0.8–1.5% of the ~330-year recurrence cycle.
  • Whether any currently undetected aseismic slip, tremor, or geodetic signals on major faults could indicate elevated stress — this data was not retrieved.
  • Market pricing at 16% vs. 31% for before-2035 implies roughly 57% of the before-2035 probability is packed into the 2026-2028 window, which is inconsistent with Poisson assumptions and suggests strong market mispricing or confusion.
Market Context
Price: 0.16 · 7d: -0.01 · 30d: +0.03
Volume: LOW — 86 contracts/day average is very thin · Depth: MISLEADING — nominally 14,068 depth but spread of $0.98 indicates extremely poor two-sided liquidity
Reliability: LOW — thin volume, extreme spread, dramatic deviation from scientific base rates (16% vs ~0.64%), price range of 2%–22% suggests high volatility from uninformed trading. Market price should NOT be used as a calibrated probability estimate.
4. Independent Forecaster Opus · No market price · legacy v2 Did NOT see market price
Independent estimate: 2% (79% confidence)
base rate
2%
80% conf
evidence driven
2%
78% conf
contrarian
2%
78% conf
Spread
0pp
Low disagreement
sq1: Is the base rate of M8.0+ earthquakes in California consistent with a >5% probability over any given ~2.6-year window?
3%
UCERF3 30-year M8+ probability ~7% implies annual rate ~0.0024, yielding ~0.6% over 2.64 years. Even using upper-bound assumptions (e.g., 10% over 30 years), we get ~0.9%. To exceed 5% over 2.6 years would require an annual rate ~2% — over 8x higher than UCERF3 estimates. The question asks if base rate is consistent with >5%, which it clearly is not. Setting p_yes very low at 3%.
sq2: Are there currently elevated seismic indicators or recent swarm activity that would meaningfully increase near-term probability above base rate?
5%
Recent swarms (San Ramon, Coachella) are routine recurrent tectonic activity with no anomalous signals. Scientists explicitly state swarms only marginally increase larger-quake probability. No M6-7 foreshock activity. No USGS advisories. The evidence overwhelmingly indicates background-level activity, not elevated indicators. Small probability (~5%) that some indicator could be argued to elevate risk modestly.
sq3: Does state of major fault systems suggest meaningfully elevated probability of M8+ rupture before 2028?
8%
Southern San Andreas being 'overdue' (~330 years since last rupture vs ~150-200 year recurrence) provides a weak upward signal — but 'overdue' captures only a small fraction of the recurrence cycle in 2.6 years (~1%). No geodetic anomalies reported. The 'meaningfully elevated' threshold is interpretive; I lean toward no, with some allowance for the overdue argument.
Show full reasoning trace
base rate: UCERF3 7% over 30 years → Poisson ~0.6% over 2.64 years. Range 0.4-1.2% across reasonable assumptions. California has had zero confirmed M8.0+ events in instrumental history (1906 SF was Mw 7.9).
evidence updates: Recent swarms: neutral/slightly down (routine activity, no precursor signal). Southern SAF overdue: small upward push (~+0.3%). No geodetic anomalies or USGS advisories: confirms baseline. Magnitude resolution uncertainty (events could be revised up to 8.0): small upward push (~+0.5%).
combination method: The sub-questions are framed as conditions/indicators rather than direct probability estimates. I anchor primarily on the Poisson base rate (~0.6-1%) and add small adjustments for tail factors, arriving at ~2%.
final: Final estimate 2%. This reflects the very rare base rate, short window, lack of elevated precursor signals, and small allowance for fault-specific overdue factors and magnitude uncertainty.
ensemble: {'ensemble_count': 3, 'perspectives': {'base_rate': {'p_yes': 0.02, 'confidence': 0.8}, 'evidence_driven': {'p_yes': 0.025, 'confidence': 0.78}, 'contrarian': {'p_yes': 0.025, 'confidence': 0.78}}, 'spread': 0.005, 'individual_p_yes': {'base_rate': 0.02, 'evidence_driven': 0.025, 'contrarian': 0.025}}
5. Devil's Advocate Sonnet · legacy v2 Market price revealed
Reasoning Flaw Missing Info
Challenges
  1. The forecaster correctly identifies the UCERF3 base rate (~0.6-1% over 2.64 years) but then arrives at 2.3% — more than double the base rate — without clearly justifying the 1.3-1.7 percentage point upward adjustment. The 'small allowance' for overdue SAF (+0.3%) and magnitude resolution uncertainty (+0.5%) only sums to ~0.8 pp, not the full adjustment applied. Either the base rate should anchor closer to 1-1.5%, or the upward factors need stronger justification.
  2. The 1906 SF earthquake at Mw 7.9 is treated as evidence DOWN, but it actually illustrates magnitude-boundary uncertainty: events can be revised, and the threshold is a sharp cliff. Historical estimates of pre-instrumental events (e.g., 1857 Fort Tejon often estimated Mw 7.9) cluster near the threshold. This argues for somewhat more weight on magnitude uncertainty than the +0.5% allowed.
  3. Information gap on Cascadia southernmost extent producing an M8+ with epicenter in California waters is flagged but not quantified. Cascadia M8+ events do occur and the boundary question is non-trivial — this should widen uncertainty modestly upward.
  4. The forecaster does not address whether 'in California' includes offshore events near the coast or only land-based — resolution criteria ambiguity could matter for marginal cases.
  5. Confidence of 0.787 on a forecast of 2.3% may be slightly overconfident given the acknowledged information gaps (no direct UCERF3 tables, no geodetic strain data, no expert near-term commentary).
Market comparison: Forecaster: 2.3%, Market: 16%. The 13.7pp divergence is large but the market shows clear signs of being uninformed/illiquid: 86 contracts/day, $0.98 spread, and 2-22% historical range suggest noise rather than signal. UCERF3 science strongly supports the low forecast. Divergence is justified.
6. Aggregator Opus · final calibrated forecast
3% (75% confidence)
UCERF3 implies ~0.6-1% probability of M8.0+ in California over the remaining ~2.6-year window, with California having zero confirmed M8.0+ in instrumental history (1906 SF was Mw 7.9). I partially accept the Devil's Advocate critique on magnitude-boundary uncertainty — events like 1857 Fort Tejon and 1906 cluster right at the Mw 7.9 threshold, so I nudge up slightly to ~3% to account for potential post-event magnitude revisions and Cascadia/offshore boundary ambiguity. I reject deference to the 16% market price: with 86 contracts/day volume, a $0.98 bid-ask spread, and a 2-22% historical range, this market is a noise generator rather than a calibrated signal, and tracks public earthquake anxiety rather than seismological base rates. My deviation from the market is justified by direct reliance on USGS UCERF3 hazard modeling.