Earthquake prediction: state of the art

Aftershock rumours are going around and the Seismological Bureau has issued a warning of a possible 6 to 7 magnitude quake. Apparently Chengdu, Chongqing and Xi'an are mostly closed for business. School has been cancelled so I have too much time on my hands.

I found an Eartquake Forecast study presentation (PPT file) on the Chinese Institute of Geophysics website. It gives an interesting look at earthquake forecasting methodology used in China, unfortunately using data from 1990 to 1998.

Firstly, it's important to note the word "forecasting" is used, and not "prediction". The one time in history when an earthquake was successfully "predicted" was in Haicheng, in 1975. A number of precursors led to the evacuation of the city's 1 million residents 5 hours before a Mag. 7.3 earthquake. Prediction is short term, and the author of this presentation, calls this prediction "a lucky (and therefore unrepeatable) successful prediction".

He cites foreshock activity as a significant precursor, but also abnormal animal behaviour and ecological "anomalies". He also implies it was basically very good timing for the earthquake to have come so quickly after the evacuation, when people were still willing to accept waiting around outside. A few hours later they would have become impatient and would have returned to their homes. Not really a factor in the prediction, but it emphasises the good luck on that day.

Apparently seismologists in China have an annual "National Consultative Meeting on Seismic Tendency". Given the impossible task of predicting an earthquake, they tend to focus on three general questions for a specific (1 year) time frame:

1. Will there be any M>7 earthquakes occurring in western China?
2. Will there be any M>6 earthquakes occurring in eastern China?
3. Which areas seem to be likely to have a M>5 earthquake?

The difference in magnitude between Western and Eastern China is based on the quality of infrastructure. Eastern China is more developed and buildings are more likely to withstand a Magnitude 6 quake. Damage will more likely be financial.

In Western China there are still many underdeveloped areas, as the Sichuan earthquake demonstrated, were casualties are more likely to be the predominant concern.

This annual consultation, therefore, has a narrow and manageable focus: probability of destructive seismological activity within the period of one year in the future. Because of this their forecasts are easier to analyse statistically.

The following GIF animation shows the success rate of forecasts in China from 1990 to 1998. The yellow areas are the areas where earthquakes were likely to occur. Every red dot represents an earthquake during that year. The success rate is expressed as a number between -1 and 1, where 1 is complete success.



They use the following formula:
Success rate = successful predictions divided by total number of earthquakes minus false alarm area divided by total aseismic area

(I take "aseismic" to mean dormant.)

Thus: hit rate minus false alarm rate.

An interesting side note, made by the authors, is that OF COURSE if you say China will have an earthquake this year, you can be 100% sure of being correct. This is why the false alarm rate has to be subtracted in the above formula.



According to this formula, the trend has been climbing, albeit very slowly, since 1990.