Can Jim Berkland predict earthquakes? Jim Berkland claims to be able to predict earthquakes. Such earth-shaking claims, however, appear to be groundless

Jim Berkland is a retired Santa Clara, California, county geologist who has issued earthquake predictions since 1974. He has a monthly newsletter and a Web site; Berkland was recently featured in a book by Cal Orey titled The Man Who Predicts Earthquakes (Orey 2006).

Berkland bases his prediction windows (eight-day time periods) on syzygy, the conjunction or opposition of the sun and moon, which happens once a month (each) for a total of two possible windows. It's possible to have three on occasion, since the lunar cycle is shorter than a month. When perigee occurs close to new moon or full moon it determines the primary window, the one used as a prediction for that month. Tide height at various stations (principally the Golden Gate Bridge) also may enter into the formula. Another factor can be the number of lost pets in whatever newspaper Berkland happens to consult. His final window is a hodge-podge of factors, which makes replication difficult. None of these factors are original with him, but the way he combines them may be.

Analyzing Berkland's Data

In order to study the validity of Berkland's claims, I used searches of the earthquake catalogs of the National Earthquake Information Center (NEIC) and the Advanced National Seismic System (ANSS), both available on the Internet. The NEIC file was taken from the Centennial Catalog (Engdahl 2002) and is the best source for larger global earthquakes from 1900 to 1999. The ANSS file contains all known events for the western United States since 1900 and was chosen for its completeness (including lower magnitude levels) compared to the NEIC catalogs.

I also have compiled a list of all Berkland's newsletter predictions taken in part from copies of the newsletters and from information posted on his Web site. I believe the list to be accurate and have sent it to him for verification, but despite his repeated promises to do so he has never certified it in full. The list starts with 1974 and continues to the present. My analysis ends with 2004.

Procedure

Evaluating the success of a prediction should be a straightforward process. You simply compare the parameters of the prediction (date, location, size) against all the earthquakes and count hits and misses. Unfortunately Berkland has made this nearly impossible by shifting the location from time to time and insisting on getting partial credit for near-misses. The best I can do is to analyze the results he would have obtained had he been consistent. His most recent parameters were used for all windows, since they should be the most accurate.

I wrote a program that compares all the earthquakes to all his predictions and counts all the hits and misses for each area and each target ring individually. These totals are compared to chance by two different methods described below.

The statistical significance of the results can be determined by comparing his results to those that would result by chance. This is best done by the z-binomial test that may be used when all predictions have the same probability. This is true for Berkland's four areas, which are the same every month. In order to meet the z-binomial restrictions I have treated each parameter group individually. I used the calculator published as part of VassarStats by Purdue University (available at http://faculty.vassar.edu/lowry/VassarStats.html).

Prediction Parameters

Berkland predicts for four areas, three in the western United States and one global. The bullseye prediction is a 140-mile-radius circle around two particular locations in California, a blanket area covering all of Washington and Oregon, and a global prediction for anywhere on Earth. Each area has an eight-day window and a particular magnitude range.

In addition to the bullseye, he insists on three near-miss rings consisting of an added day at the beginning and end of the date range, a 10 percent increase in the radius and an additional 0.1 unit of magnitude on each end of the magnitude range.

Finally, he insists that all times be expressed in Pacific Standard Time, which is confusing to anyone looking at a list of earthquakes in Greenwich Mean Time.

Berkland's areas are seen in table 1, taken from his Web site:

Table 1. Jim Berkland's earthquake prediction chart

                Ring        Location        Radius     Mag. Range

Mt. Diablo    Bullseye   37.9 N, 121.9 W   140 miles   3.5 to 6.5
              Ring 1                       154 miles   3.4 to 6.6
              Ring 2                       168 miles   3.3 to 6.7
              Ring 3                       182 miles   3.2 to 6.8

Los Angeles   Bullseye   34.0 N, 118.0 W   140 miles   3.5 to 6.5
              Ring 1                       154 miles   3.4 to 6.6
              Ring 2                       168 miles   3.3 to 6.7
              Ring 3                       182 miles   3.2 to 6.8

WA-OR         Bullseye     Anywhere in                 3.0 to 5.5
              Ring 1     the two states                2.9 to 5.6
              Ring 2                                   2.8 to 5.7
              Ring 3                                   2.7 to 5.8

Global        Bullseye     Anywhere in                 7.0 to 9.9
              Ring 1        the world                  6.9 to 9.9
              Ring 2                                   6.8 to 9.9
              Ring 3                                   6.7 to 9.9