The men behind GeoForecaster.com, which sells earthquake forecasts starting at $9.95 a month, are good, smart people with "over 50 years of collective experience in earth science research and development." Who are they, and are their methods sound?
Well, there's good and bad; first the good.
The Principals of the Firm
CEO Michael Kozuch and Chief Forecaster Lowell Whiteside are versatile, caring people. Both served in the Peace Corps, for instance. During his stint, Kozuch compiled a geologic map for the Central American country of Honduras. Whiteside was in Swaziland in southern Africa, some time after he earned his B.S. in biology/ecology from Minnesota's Hamline University in 1968.
Both men attended the University of Colorado at Boulder and earned Ph.D.'s there, Kozuch in 1995 and Whiteside in 1999. Kozuch's thesis was titled "Earthquake Hazard Analysis of Venezuela Using Site-Specific Attenuation" while Whiteside's was "Short-Term Variations in Seismic Hazard." Both men have worked at the National Geophysical Data Center (NGDC) in Boulder, helping to build the world's largest earthquake catalog. Whiteside is still there, contributing to this important scientific resource.
Both have been at New Zealand's earthquake agency, the Institute of Geological and Nuclear Sciences. Kozuch is still there, contributing to public understanding of seismology and earthquake hazards. He's a co-leader of the innovative Quaketrackers school program.
Neither man is a notable researcher, but like most grad students and postgrads they have made presentations at major scientific meetings. From those and a few other statements they've made, I can say a little about their likely forecasting methods.
The Raw Edge of Prediction Research
Today's would-be forecasters are searching for precursors to earthquakes. It's a challenging problem because
- We don't understand very well how earthquakes work
- The quake record is short, spotty and inconsistent
- The obvious ideas have all failed
- The remaining possibilities are quite subtle effects with large uncertainties
Kozuch and Whiteside know these facts. They therefore rely on the NGDC seismic catalog, the world's best, and mine that data for patterns with bleeding-edge techniques and hypotheses. In particular:
Kozuch has analyzed earthquake data with ApEn or approximate entropy, an experimental mathematical tool used to assess the "patternness" of data (such as heartbeat records or random number tables) that appears to be chaotic.
Whiteside has long studied correlations of earthquakes with solar activity, geomagnetic anomalies, "tectonic stress waves," and global normal modes ("ringing" of the Earth).
Results from these avenues of inquiry haven't appeared in peer-reviewed journals. Usually this means that the data isn't conclusive. Despite the ideals of scientific philosophy, no one publishes vague or negative results.
Flirting with Plausibility
Kozuch has cited other more conventional possibilities as showing promise:
- Earthquake swarms (and their opposite, quiet periods)
- Variations in quake rates or magnitudes
- Foreshocks
We know these things exist, but they only appear significant when you look backwards in time from an earthquake that actually happened. Still, there's a small probability that any seismic event could be a precursor, and you could adjust the odds by minuscule amounts from day to day, like someone who trades stocks around the clock looking for a tiny edge.
So far so good. All these faint signals (or possible signals) could be added up by a suitable program to churn out forecasts. Let's assume that Kozuch and Whiteside are doing something like that. But any results are likely to be difficult to distinguish from randomness. That's where they begin to hide and distort things, as I explain next, and where my respect for them starts to fade.
