Active research in environmental data science — seismology, space weather, climate, hydrology, and the cross-domain bits where everything starts to look the same. All studies built on open government data.
Population-level test of the earthquake-lights folklore: 9,379 US M>=4.5 earthquakes (1910-2014) cross-matched against 71,008 NUFORC reports. Observed coincidences (3) sit below the 10,000-permutation null mean (9.4); null survives every sensitivity check.
Local baseline characterization of the geographic environment at Skinwalker Ranch (40.2549°N, -109.8864°W). Phase 1 of GitHub issue #222.
How many hours does southward IMF Bz lead Dst-index minima, and is the predictive relationship Granger-causal? Pure space-weather predictability test on TerraPulse's 71-day Bz/Dst overlap window.
CNEOS fireball day-of-year tested against 12 named meteor shower peaks. Clean null after deduplication: Schuster p=0.22, zero showers survive Bonferroni at any window width. Detection floor at N=355 is rate ratio >=1.96 (Bonferroni-12); optical-bolide Geminid ratios of 1.3-1.5 are below the floor.
Cluster >3sigma AQI anomaly days across 10 cities by their constituent-pollutant signatures (PM2.5, PM10, O3, NO2, SO2, CO).
Do ground-level Safecast background-radiation measurements show a measurable depression during Forbush decreases (galactic-cosmic-ray flux reductions following CME shock arrivals)?
V3 follow-up to #132. Continuous-driver framing replaces small-N storm catalog with daily Bz signal on ~180 daily WSPR corridor samples.
Cross-protocol V2 of PMA #132. Holds input data and Kp event catalog constant with #132 V1; swaps only the analytical protocol to match PMA #82's per-storm scalar-delta one-sample t/Wilcoxon with Bonferroni over 4 corridors. Isolates which of four methodological factors drives the #82 (Bonferroni-significant POLAR -0.29 dB on Dst, 6yr, N=200) vs #132 V1 (clean null on Kp, 6.5mo, N=26) divergence.
How does WSPR SNR on defined long-distance corridors respond to geomagnetic storms? Superposed-epoch analysis centered on Kp excursions.
On 2026-05-15 a M6.6 earthquake struck offshore Sanriku, NE Japan (~39.03N, 141.81E), in the Pacific-Plate subduction zone where the 2011 Tohoku M9.1 mainshock occurred.
V2 re-run of #180 with IGRA sounding data now available (1.2M rows). Test if WSPR 40m precursor timing aligns with CAPE peak timing.
V2 of #181: re-run with full IGRA data (1.2M rows) and live WSPR. Test whether WSPR 40m precursor is independent of LI or proxy.
Is the WSPR 40m precursor independent of LI, or a proxy for atmospheric instability? H1: independent. H0: LI explains all of the WSPR-tornado correlation.
Does atmospheric instability (CAPE / lifted index) ramp on the same 4-6h pre-tornado timeline as the WSPR 40m precursor signal? H1: timing match. H0: independent timing.
V5: are the V3/V4 40m WSPR pre-tornado precursors tornado-specific or generic to deep severe convection? Magnitude-matched non-tornadic hail+wind days as trigger
V4 discriminates D-layer absorption (WSPR leads lightning) from QRN (simultaneous) using GLM historical lightning data on the V3 outbreaks. Data source: GOES-16 GLM L2 LCFA from noaa-goes16 public S3 bucket.
V3 with fixed control design — calm-season controls, SPC-pre-screened, density-matched
V2 verification of the paper #29 40m precursor finding using 7 historical outbreaks + 21 matched controls
Does WSPR HF propagation show a detectable SNR/density depression in grid cells under active severe weather? 10-day CONUS dataset with SPC + NWS overlay. GH issues #133-138.
Research question: What is the empirical distribution of lag times between a southward IMF Bz excursion (DSCOVR L1 measurement) and the resulting Kp index increase indicating geomagnetic storm onset?
Does the lunar gravitational tide produce a detectable periodic signal in inland river streamflow at M2 period (12.42h)?
On 2026-04-16 a M6.0 earthquake struck south of the Kermadec Islands (-32.21, -178.05, depth 10 km). 48 hours later (2026-04-18), a M5.9 struck the same
Fact-check of claims that an earthquake swarm off Washington coast is occurring in the Cascadia Subduction Zone. USGS data shows the swarm at ~129°W on the Juan de Fuca Ridge, ~400 km west of the CSZ trench.
Before evaluating any WSPR-aircraft tracking hypothesis (e.g., MH370), we must answer the prior question: can WSPR detect commercial aircraft at all? This pre-registered analysis tests whether WSPR SNR variations correlate with known flight crossings of station-pair great circle paths.
Complete statistical census of the WSPRnet raw spot archive: 11.53 billion spots across 221 months, aggregated in 64 minutes of wall time on a single workstation. Foundational reference for all WSPR papers.
Test whether rapid drops in atmospheric pressure trigger small earthquakes. The folk-science hypothesis says storm fronts reduce confining stress on shallow faults and should elevate seismicity by 0.1-1%. Controlled studies generally find no effect.
V4 extends the TerraPulse Granger-causality network analysis (V1-V3) by adding atmospheric composition (CO2, CH4), a lunar tidal-potential proxy, and barometric pressure to the 8-metric space-weather+surface network from V3. Emphasis is on honest per-pair overlap accounting, Bonferroni correction based on tests actually run, joint F-tests over all lags, stationarity checks, a shuffle null, and AR self-predictability baselines.
Do flood warnings predict landslide advisories? Do heat advisories predict air quality alerts? Build a temporal cascade network from the NWS alert stream.
Do near-Earth object close approach patterns correlate with fireball events? Are there temporal clusters in flyby distance or velocity?
Can simple time-series methods (ARIMA, spectral decomposition, or ML) on our 208-year sunspot record produce a competitive forecast for Solar Cycle 25? Benchmark against NOAA predicted range.
Can we detect a lunar tidal signal in inland river streamflow? The gravitational tidal force acts on all mass, not just oceans -- but the effect on rivers should be extremely small.
Does background radiation (Safecast citizen-science network) show detectable anomalies in the days before significant earthquakes?
Starlink attribution to NUFORC sightings is impossible: NUFORC ends 2014, Starlink launched 2019. Pre-2014 satellite phenotype analysis: 826 sightings (1.03%) with explicit satellite mentions, 3,862 (4.81%) matching broad satellite phenotype. Formation shape increased 1.6x post-Iridium (chi2(1)=23.0, p=1.6e-6, N=76,097). Even broadest filter reduces unexplained from 75.9% to 71.1%. Starlink (10,119 sats) would dominate post-2019 reports but cannot be tested against this dataset.
Does ENSO modulate NWS alert rates? Monthly alert counts stratified by ONI phase. La Nina -> more tornadoes? El Nino -> more floods? Fire weather vs drought index.
The magnetic north pole has accelerated toward Siberia (~55 km/yr). Our WSPR data shows -0.1 dB/year across all HF bands over 21 years.
We now ingest from 4 seismic networks. Compare: detection thresholds by region, magnitude agreement for shared events, temporal coverage gaps.
POLAR paths degrade 0.29 dB during geomagnetic storms (t=4.9, p=2e-6, d=-0.35, N=200 per-storm deltas). EQUAT null (Wilcoxon p=0.088, Bonferroni p=0.35) validates corridor-differential method. Differential POLAR-EQUAT = -0.23 dB. Effect monotonically increases with storm severity (d=-0.62 at Dst<-100). NA_EU nominally significant (p=0.024) but does not survive Bonferroni correction.
80K NUFORC sightings filtered: 6.2% self-classified fireballs (exclusive), 16.3% storm/aurora (exclusive), 0 CNEOS matches. 75.9% unexplained with strong summer seasonal bias (chi2(11)=3466, days-per-month null). Storm match rate (19.5% of covered) below random baseline (31.2%).
3 of 32 Cat C events reclassified (G5 storm, solar eclipse, Grimsvotn eruption). 29 remain unexplained: 79% positive enhancement (p=0.002), 55% at solar minimum, upper HF dominant. Sporadic F-region density enhancements.
Cycle removal + anomaly detection on 108K WSPR rows finds 202 transient events. 17 attributed to known storms, 66 unexplained (statistically identical to attributed, p=0.71). Method validated by blind detection of May 2024 G5 superstorm (z=6.7σ).
11.5B WSPR spots (2004-2026), 108K daily band aggregates. 40m/20m improve in solar max (r=+0.45), 10m degrades (r=-0.29). Network grew from 4 TX to 771. The ionosphere breathes with the Sun.
SEA of 1,809 DONKI events: storms peak Kp at T+6h (+47%), HSS +6%, radiation belt lags by 38h. CME speed negatively correlates with Kp (r=-0.18) due to catalog selection bias.
Baseline analysis of 550K NWS alerts (45 types, 8 days) + 69K historical tornado reports (1950-2023). 80% marine, 63x diurnal swing, May tornado peak, Kp cross-domain null. Foundation for dashboard and cross-domain work.
X1.5 flare (largest in 12.6M GOES readings) + 1,845 km/s CME. 6-entry lab notebook tracking the event over 60h. CME did not impact Earth. Successful infrastructure test of 7-stream monitoring pipeline.
Do geomagnetic storms degrade gravitational wave detector sensitivity? Can we turn LIGO/Virgo duty cycle data into an accidental geophysical measurement?
Does the Keeling Curve predict temperature anomaly? 52yr CO2 vs temperature data.
Can we detect Pinatubo (1991), El Chichon (1982) in the Keeling Curve? Cross-reference CO2 with 5K eruption records.
Schuster test: do volcanoes erupt at New/Full Moon?
When Kp spikes, what else moves? Trace the cascade through AQI, earthquakes, tides, radiation, power systems, and atmospheric chemistry.
Superposed epoch analysis: do near-Earth object close approaches produce detectable changes in earthquake rate, Kp, or tides?
Is tornado season extending later into the year? 73 years of SPC data.
Cross-domain distribution analysis: power law vs exponential vs lognormal vs Weibull across 10 environmental metrics
Do earthquakes cluster at Full Moon, New Moon, or lunar perigee? Testing the 1-2% tidal triggering hypothesis with 172K earthquakes and 5K hours of JPL ephemeris.
Do solar flares, CMEs, and the solar cycle directly drive earthquakes, volcanic eruptions, and severe weather? The definitive test.
Do earthquakes cluster near active volcanoes? Can seismic swarms predict eruptions?
Do earthquakes, ocean microseisms, and geomagnetic storms correlate with gravitational wave detector sensitivity?
73-year analysis of tornado patterns, Tornado Alley migration, EF trends, and solar-tornado coupling
Test whether temperature patterns predict PM2.5 spikes via inversion-like conditions
Test which environmental signals Granger-cause others — the big connectivity map
Detect anomalous streamflow patterns across 472 USGS California gauging stations
Testing whether 3 years of NASA POWER daily temperature data (10 cities) shows detectable warming trends. Using linear regression, seasonal decomposition, and anomaly scoring.
Testing whether the March 2026 fireball cluster (11 events in 20 days, 3 meteorite falls) is statistically significant against the CNEOS historical baseline. Cross-correlating with TerraPulse space weather data to test solar activity connection.
Separating physical radiation patterns from observational campaign clustering in 679K Safecast measurements across 45 countries. Can citizen science data establish a reliable global radiation baseline?
Derivative study expanding on: Spatial clustering. nearest-neighbor distance analysis Parent study: Temporal Clustering of Global M4+ Earthquakes (earthquake-temporal-clustering)
Detecting days where multiple environmental metrics are simultaneously anomalous using Z-score analysis and Isolation Forest. When earthquakes, space weather, and air quality all spike together — is it coincidence or connection?
Applying the coefficient of variation (CV) inter-event time method proven on earthquakes (CV=1.545) to CMEs, radiation spikes, geomagnetic storms, AQI exceedances, and flood events. Building a clustering atlas across domains.
Replicating Georgieva & Kirov (2010) claim that the correlation between solar activity and geomagnetic activity varies with a changing time lag. Testing with NOAA SWPC solar flux and Kp index data.
Replicating and extending published claims that El Niño/La Niña phases modulate global earthquake rates. Testing crustal loading hypothesis via sea level redistribution. Using 76 years of ONI data and 5 years of USGS M4+ earthquake records.
Testing the controversial hypothesis that solar activity (CMEs, solar flux, geomagnetic storms) correlates with earthquake frequency or magnitude. Using NOAA SWPC, NASA DONKI, and USGS earthquake data.
Analyzing whether significant earthquakes cluster in time beyond what random (Poisson) occurrence would predict. Testing the hypothesis of earthquake triggering — does one large quake increase the probability of subsequent events?
Investigating whether El Niño/La Niña phases predict drought severity in US states. Using 76 years of ONI data and 16 years of USDM drought records.
