Space Weather Predictions

Probabilistic forecasts derived from live NOAA SWPC solar wind plasma, magnetometer Bz component, and Kp index data. Updated every page load from primary NOAA feeds.

Bayesian updating Poisson process AR(1) autoregression NOAA SWPC live feeds

Methodology note: Geomagnetic storm probability is derived from solar wind speed and IMF Bz direction using Bayesian inference with a 5%/day historical base rate (NOAA 1996–2024). Poisson process models event-arrival rates from 60 recent samples. All confidence intervals are ±1σ from model uncertainty, not observational error.

Solar Wind Speed — Live with Threshold Bands

2-hour rolling feed from NOAA SWPC. 30-point moving average shown. Storm threshold at 500 km/s.

NOAA Space Weather Prediction Center · plasma-2-hour.json · live

How we compute P(storm)

P(Kp≥5 | SW>500) ≈ 34% vs P(Kp≥5 | SW≤500) ≈ 2%. Bz southward multiplies likelihood by ×5.6. Bayesian update gives posterior probability.

Why this matters

Geomagnetic storms affect satellite operations, GPS accuracy, HF radio, and in extreme cases (Carrington 1859 level) power grid infrastructure.

Kp Index — Empirical Distribution (last 90 minutes)

Observed frequency of each Kp value from recent 1-minute data. Bar height = probability mass.

NOAA SWPC · planetary_k_index_1m.json · live

Poisson process model

Event arrival rate λ estimated from recent samples. P(Kp ≥ k in t hours) = 1 − e^(−λt) · Σ(λt)^n/n! follows directly from Poisson CDF.

AR(1) forecast

Kp(t+1) = φ·Kp(t) + ε. Autocorrelation coefficient φ estimated from last 10 observations. Short-run persistence captured.

Signal Decomposition — Solar Wind Trend / Residual

60-sample moving average isolates trend from noise. Residual = observed − trend. Large residuals indicate anomalous conditions.

NOAA SWPC · derived from plasma-2-hour.json

Climate Signal Predictions

Statistical extrapolation and anomaly detection on the Keeling Curve (CO₂) and NASA GISS temperature record. Linear regression with confidence intervals; Z-score anomaly detection against the 1951–80 baseline.

OLS linear regression Z-score anomaly detection 90% confidence interval NOAA GML + NASA GISS

CO₂ Trend Extrapolation — With 90% Confidence Interval

OLS fit on weekly Keeling Curve data 1958–present. Shaded region = 90% CI. 430 ppm marks the 1.5°C pathway threshold.

NOAA Global Monitoring Laboratory · Mauna Loa Observatory · live weekly data

Regression model

CO₂(t) = β₀ + β₁·t + ε. Fitted on 3,400+ weekly readings. R² > 0.99. Annual increment accelerating: 1.0 ppm/yr in 1960s → 2.4 ppm/yr now.

Scientific consensus

97%+ of peer-reviewed climate science confirms anthropogenic warming. Trajectory to 430 ppm (1.5°C pathway) is within 2–4 years at current rate.

Temperature Anomaly — Z-Score vs 1951–1980 Baseline

Current temperatures expressed in standard deviations from the 1951–80 mean. Values >2σ have <2.3% probability under null hypothesis of no change.

NASA GISS · GISTEMP v4 · GLB.Ts+dSST.csv · live annual data

Anomaly detection

Z = (x − μ) / σ where μ=0.0°C, σ=0.17°C from 1951–80 baseline. Recent years exceed 7σ — the chance of this occurring naturally approaches zero.

What this means for predictions

P(record year) = Φ(Z_current − Z_record). As each year now starts from a higher base, record years become the statistical norm, not the exception.

Air Quality Risk Signals

Probabilistic health risk assessment from live PM2.5 measurements across 12 major cities. Threshold exceedance probability computed via logistic function fitted to WHO dose-response data.

Logistic exceedance model WHO dose-response OpenAQ v2 live

PM2.5 Exceedance Probability — 12 Cities vs WHO Thresholds

Left axis: measured PM2.5 μg/m³. Right axis: P(exceeds WHO 24h limit of 25 μg/m³) from logistic model. WHO annual guideline: 5 μg/m³ (2021).

OpenAQ API v2 · live measurements · coordinates radius 25km

Logistic model formula

P(unhealthy) = 1 / (1 + e^(−k·(x − θ))) where θ = 25 μg/m³ (WHO 24h), k = 0.15 (slope fitted from epidemiological data).

Health burden (Pope et al. 2002)

6% increase in all-cause mortality per 10 μg/m³ PM2.5. Air pollution causes 7M deaths/year globally — more than AIDS, TB, and malaria combined.

Seismic Activity Signals

Earthquake probability estimates using three independent methods: Gutenberg–Richter frequency-magnitude law, Poisson process event-rate modelling, and Omori–Utsu aftershock sequence decay. All fitted in real time to the USGS 7-day feed.

Gutenberg-Richter b-value Poisson λ estimation Omori-Utsu R(t)=K/(t+c)^p USGS GeoJSON live

Gutenberg-Richter Law — Fitted to Live USGS Data

log₁₀(N ≥ M) = a − b·M. b-value fitted by OLS. Deviation from fit indicates unusual seismic conditions.

USGS Earthquake Hazards Program · M5+ feed · 7-day window

What b-value tells us

Global average b ≈ 1.0. Higher b → more small quakes relative to large. Lower b → elevated large-event risk. Real-time fit shown above.

Extrapolation to rare events

G-R law predicts frequency of any magnitude. Combined with Poisson process, gives P(M≥7 in next 30 days) = 1 − e^(−λ₇·30) where λ₇ = N(M≥7)/7.

Omori-Utsu Aftershock Sequence — R(t) = K/(t+c)^p

Hourly event counts vs Omori-Utsu model fit. p ≈ 1.1, c ≈ 0.5 hr (global averages). Excess above fit may indicate foreshock sequence.

USGS · derived from timestamps · 7-day window

Near-Earth Object Risk Signals

Impact risk assessment using geometric collision cross-section, estimated kinetic energy, and Torino Scale mapping. Today's close approaches from NASA JPL Center for Near Earth Object Studies.

Geometric cross-section P ∝ σ/d² Torino Scale: P × Energy KE = ½mv² NASA NeoWs live

NEO Risk Matrix — Today's Close Approaches

X: miss distance (log scale). Y: velocity. Circle size: estimated diameter. Red = Potentially Hazardous Asteroid (PHA). Risk zones from Torino Scale.

NASA Center for Near Earth Object Studies · NeoWs API · today's date

Impact energy calculation

E = ½ρV(πd³/6)v² where ρ = 3,000 kg/m³ (rocky), v = approach velocity. 1908 Tunguska = ~10–15 megatons. Chelyabinsk 2013 = 0.5 MT.

Honest context

Virtually all NEOs in NASA's catalogue score Torino 0 — no concern. 99.994% of known NEOs present zero 100-year threat. Monitoring is a precaution, not an alarm.

Economic Signal Predictions

Regime detection, Z-score stress indicators, and FX volatility signals from live ECB exchange rates and World Bank macro data. All probabilities are statistical — economic prediction has large inherent uncertainty.

Z-score stress index Volatility EWMA Frankfurter ECB live World Bank historical

Disclaimer: Economic predictions are among the least reliable of any forecasting domain. These are signals and base-rate estimates, not investment or policy advice. GDP forecast error typically exceeds ±2% at the 1-year horizon.

World GDP Growth — Z-Score Deviation from Long-Run Mean

Annual GDP growth expressed in standard deviations. Recessions visible as negative Z spikes. GFC 2009, COVID 2020 are clear outliers.

World Bank historical data · NY.GDP.MKTP.KD.ZG · verified 2005–2023

EUR/USD Exchange Rate and Annualised Volatility

Spot rate and rolling 20-day annualised volatility. Volatility spikes precede regime changes.

European Central Bank via Frankfurter API · live daily rates 2024

Health & Epidemiological Signals

Exponential decay projection for child mortality, Poisson-based epidemic threshold estimation, and trend-break detection in global health indicators.

Exponential decay U5MR(t) = U5MR₀·e^(−λt) SDG probability via normal CDF World Bank · disease.sh

Child Mortality — Exponential Decay Projection to 2040

U5MR fitted to exponential decay model. Shaded band = 95% CI. SDG 3.2 target: ≤25 per 1,000 by 2030. Projection shows whether current trend is on track.

World Bank · SH.DYN.MORT · historical 1990–2022 · projection 2023–2040

Model: exponential decay

U5MR(t) = U5MR₀ · e^(−λt). λ fitted from log-linear regression on 1990–2022 data. λ ≈ 0.027/year (R² ≈ 0.98). Confidence interval propagated from σ of regression.

SDG target probability

P(SDG target met) = Φ((25 − projected) / σ_forecast). At current rate, projected 2030 value is ~28–30, just above the ≤25 target. Accelerated investment needed.

All Signal Intelligence

Every live prediction across all domains, ranked by signal probability. All estimates derive algorithmically from public data fetched at page load.

BayesianPoissonZ-scoreRegressionLogisticTorino ScaleGutenberg-Richter

Signal Correlation Matrix

Pearson correlation between prediction scores across domains. Positive = signals tend to be elevated together; negative = inversely related.

Derived from all prediction outputs · updated each session

All Predictions — Ranked by Signal Strength

All domains · live session

Question	Domain	P(%)	±CI	Level	Algorithm	Source
loading all signals…