|
1. |
Debate on evaluation of the VAN Method: Editor's introduction |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1291-1293
Robert J. Geller,
Preview
|
PDF (299KB)
|
|
摘要:
This special section presents a debate on the statistical significance of the results obtained by the VAN method of earthquake prediction. Other topics related to the VAN method are also discussed by some of the papers in this section. This introduction explains the background that led to this debate. Tables of seismicity data are also included in this introduction. These tables are cited in some of the other papers in this special section.
ISSN:0094-8276
DOI:10.1029/96GL00742
年代:1996
数据来源: WILEY
|
2. |
Basic principles for evaluating an earthquake prediction method |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1295-1298
P. Varotsos,
K. Eftaxias,
F. Vallianatos,
M. Lazaridou,
Preview
|
PDF (534KB)
|
|
摘要:
A three year continuous sample of earthquake predictions based on the observation of Seismic Electric Signals in Greece was published byVarotsos and Lazaridou[1991]. Four independent studies analyzed this sample and concluded that the success rate of the predictions is far beyond chance. On the other hand,Mulargia and Gasperini[1992] (hereafter cited as MG) claim that these predictions can be ascribed to chance. In the present paper we examine the origin of this disagreement. Several serious problems in the study of MG are pointed out, such as: 1. The probability of a prediction's being successful by chance should be approximately considered as the product of three probabilities,PT,PEandPM, i.e., the probabilities with respect to time, epicenter and magnitude. In spite of their major importance,PEandPMwere ignored by MG. The incorporation ofPEdecreases the probability for chancy success by more than a factor of 10 (whenPEis taken into account it can be shown that the VAN predictions cannot be ascribed to chance). 2. MG grossly overestimated the number of earthquakes that should have been predicted, by taking different thresholds for earthquakes and predictions. With such an overestimation, MG's procedure can “reject” even an ideally perfect earthquake prediction method. 3. MG's procedure did not take into account that the predictions were based on three different types of electrical precursors with different lead‐times. 4. MG applied a Poisson distribution to the time series of earthquakes but included a large number of aftershocks. 5. The backward time correlation between predictions and earthquakes claimed by MG is due to misinterpretation of the text of some predictions and an incorrect use of aftershocks. Although even the discussion of the first problem alone is enough to invalidate the claims of MG, we also discuss the other four problems because MG violated some basic principles even in the time domain alone. The results derived in this paper are of general use when examining whether a correlation between earthquakes and various geophysical phenomena is beyond chance o
ISSN:0094-8276
DOI:10.1029/96GL00905
年代:1996
数据来源: WILEY
|
3. |
Inaccuracies in seismicity and magnitude data used by Varotsos and Coworkers |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1299-1302
M. Wyss,
Preview
|
PDF (496KB)
|
|
摘要:
The claims of successful earthquake predictions by VAN are based on magnitudes said to be Ms (surface wave). However, they are larger by 1.0 unit, on the average, than those published in the Preliminary Determination of Epicenters (PDE), the authoritative source for Ms values. This discrepancy arises because VAN calculate Ms from ML (local magnitude) using the incorrect transformation Ms(VAN)=ML+0.5. Least squares regressions, based on the PDE data for Greece, yield Ms=l.50ML−2.62 and Ms=1.82mb−4.16 instead. An important part of many author's evaluations of VAN's claims is an assessment of whether their performance is better than that expected by random chance. If the random chance probability is defined using standard (PDE) seismicity data and magnitudes, whereas VAN's claims are based on a non‐standard scale, the significance of VAN's successes will be overestimated strongly. Additional discrepancies are introduced because Ms(VAN) is based on preliminary ML values), or sometimes on newspaper reports, rather than those published in the final bulletin of the Seismological Institute of the National Observatory Athens (SI‐NOA). The correct final magnitude should be used in all studies claiming success and evaluating these claims years after the earthquakes occurred. Inaccuracies in presenting an earthquake catalog also exist inVarotsos et al.[1981b], who claimed a one‐to‐one correlation of seismic electric signals (SES) with the list of earthquakes with mostly ML>2.9 published by the SI‐NOA. In fact, VAN had used instead a list called “Ten Days Preliminary Seismogram Readings at Athens” which is not a bulletin and contains no epicenters. As a consequence 37% of the earthquakes reported in the SI‐NOA Monthly Bulletin (including the largest one) were not contained in the list of VAN, and 40% of the earthquakes VAN claimed had occurred, were not in the SI‐NOA Bulletin. In addition, VAN added another 25% of events to the list. These latter events were identified as added, but there is no evidence that they were not explosions or other cultural disturbances. It is concluded that some of VAN's claims of correlations between SESs and earthquakes, and some of the evaluations of these claims, are based on incor
ISSN:0094-8276
DOI:10.1029/95GL03545
年代:1996
数据来源: WILEY
|
4. |
Reply to “Inaccuracies in seismicity and magnitude data used by Varotsos and Co‐workers,” by M. Wyss |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1303-1306
P. Varotsos,
K. Eftaxias,
M. Lazaridou,
E. Dologlou,
V. Hadjicontis,
Preview
|
PDF (581KB)
|
|
摘要:
A direct comparison of the predicted magnitude values (Mpred) to the actual magnitude values (MEQ) of the earthquakes (EQs) is allowed only when both values, i.e., Mpredand MEQ, refer to the same scale. In view of the fact that the Seismological Institute of the National Observatory of Athens (SI‐NOA) publicly announces as MEQthe ML+0.5 value (where MLthe local magnitude), VAN made it clear long ago, that the predicted values Mpred(after a proper calibration) referred to ML+0.5. Therefore, a self‐consistent evaluation of VAN‐predictions should consist of a direct comparison of Mpredwith the actual ML+0.5. Unfortunately,Wyss[1996] confuses the discussion by proceeding to a direct comparison of Mpredwith Ms(PDE); this is not allowed because the values of ML+0.5 exceed, on the average, Ms(PDE) by 1.0 unit. An additional confusion arises from the fact that the relation suggested byHamada[1993], i.e., ML+0.5=mb+0.3, is misinterpreted by Wyss as saying Ms(PDE)=mb+0.3. These two alterations by Wyss reveal that his Figures 1 and 2 are erroneous.Wyss[1996] also criticizes VAN, because (in an early publication)Varotsos et al.[1981b] used the Preliminary Bulletin of SI‐NOA, instead of the final one. First of all, the final bulletin could not be used by VAN at that time, because it appeared (more than one year) after the publication of the paper byVarotsos et al.[1981b]. Secondly, the correlation between SESs and EQs is evident, when we use consistently, either the preliminary, or the final bulletin of SI‐NOA. On the other hand,Wyss[1996] claims that he could not find any correlation between EQs and SESs; we show that this is due to the fact that Wyss included, in his study, small EQs that occurred several hundreds km away from the measuring VAN station (i.e., in Albania, western Turkey, etc.), but he simultaneously deleted the small magnitude EQs that occurred very close to that station. Wyss's procedure is, of course, not acceptable and hence his Appendix B is wrong. Furthermore, Wyss's claim that VAN added 25% of events to the list, is shown to be untrue.Beyond the unusual fact that Wyss quotes “VAN's statements” that have never been published by VAN, the following is also noted: althoughWyss[1996] uses quotation marks (in order to indicate that he reproduced exactly what VAN said), he adds critical wording to VAN statements and hence their true meaning is drastically changed. For example, Wyss states: “Varotsos et al.[1981a] had first formulated that SESs “occurred a few minutes before each earthquake [related to that SES]” (Varotsos et al.[1981a]).” Thus, Wyss leads the reader to the wrong conclusion that VAN initially claimed that SES have a lead time of a few minutes, and that VAN changed it later. However, we show that this lead time (published by VAN) referred to another type of precursor, and not to SES, but the words in brackets (which are added by Wyss) alter the true mea
ISSN:0094-8276
DOI:10.1029/96GL01415
年代:1996
数据来源: WILEY
|
5. |
Probability of chance correlations of earthquakes with predictions in areas of heterogeneous seismicity rate: The VAN Case |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1307-1310
M. Wyss,
A. Allmann,
Preview
|
PDF (533KB)
|
|
摘要:
Evaluations of 22 claims of successful earthquake predictions in Greece byVarotsos and Lazaridou[1991] were performed using the Ms (surface wave) as well as the ML (local) magnitude scales. If we assume that the predicted magnitudes were Ms (the scale was not specified in the prediction telegrams), and use the Preliminary Determinations of Epicenters (PDE) to estimate the seismicity rate expected at random, we find that 74% were false (they did not correlate with an earthquake as specified), 9% correlated by chance (probability, p, of one or more earthquakes at random>.50), and 17% are uncertain (0.15
ISSN:0094-8276
DOI:10.1029/95GL03548
年代:1996
数据来源: WILEY
|
6. |
Reply to “Probability of chance correlations of earthquakes with predictions in areas of heterogeneous seismicity rate: The VAN Case,” by M. Wyss and A. Allmann |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1311-1314
P. Varotsos,
K. Eftaxias,
M. Lazaridou,
E. Dologlou,
V. Hadjicontis,
Preview
|
PDF (520KB)
|
|
摘要:
All conclusions of Wyss and Allmann [1996] (hereafter cited asWA) are wrong, because their methodology is false. For example,WA'smain conclusion reads: “the probability [P] that the observed correlations of [VAN]predictions with earthquakes (…11 out of 23 attempts) was due to chance is estimated as…96%…” However, when followingWA'sprocedure exactly, and assuming that all 23 predictions (out of 23 attempts) are correct, we find a paradox, i.e., values of the probability P larger than unity. In view of this example, any further discussion onWA'sclaims becomes unnecessary. However, we proceed to detailed replies, point by point, in order to show thatWAhave also made several mistakes and major misinterpretations of the true content of VAN's statements.Characteristic examples of the various misinterpretations (and mistakes) made byWAinclude: (i) a direct comparison of predicted magnitude values with Ms(PDE), while VAN had clearly stated that the magnitude values mentioned in the predictions correspond to Ms(ATH), i.e., to ML+0.5. Such a comparison is not allowed because ML+0.5 significantly differs (i.e., on the average by 1.0 unit) from Ms(PDE), (ii) an addition (or deletion) of critical wording to the VAN statements (and predictions) so that they distort VAN's true meaning, (iii) the use of 22 day prediction time window in the large majority of predictions which, however, correspond to single SES (and hence to an 11 days prediction time window), (iv) an incorrect statement thatVarotsos et al.[1993a,b] define the acceptable uncertainty as ΔM ≤ 1.0, while VAN repeatedly published that a prediction is accepted as successful only when ΔM ≤ 0.7, (v) an erroneous claim that when using SI‐NOA “12 out of 22 VAN predictions fail to conform to the error limits,” while the reader can easily check that only 6 (or 7) out of 23 cases deviate from the error limits. Furthermore,WAgrossly overestimated the number of the earthquakes (EQs) that should have been predicted, i.e., while VAN clearly stated that predictions are issued only when the expected magnitude is larger than (or equal to) 5.0 units,WAerroneously demand that VAN should predict all EQs with Ms≥ 4.3 or Ms≥ 4.0. Hence they characterize as a “missed earthquake” any event with Ms≥ 4.3 (or Ms≥ 4.0 respectively) for which prediction was not issued.Last but not least, we recall thatWyss and Baer[1981] published long term predictions in Greece (for the same time period discussed in this debate) ‐referring to expected EQs with magnitude 7.75‐ which tur
ISSN:0094-8276
DOI:10.1029/96GL00908
年代:1996
数据来源: WILEY
|
7. |
VAN earthquake predictions‒An attempt at statistical evaluation |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1315-1318
Y. Y. Kagan,
Preview
|
PDF (528KB)
|
|
摘要:
Statistical verification of the VAN or SES (seismic electric signals) predictions from 1987–1989 is considered. The test is carried out with the updated rules proposed byVarotsos et al.[1996]. Although for the Greek (SI‐NOA) earthquake catalog the VAN method formally is successful, this high rate of success is due either to the retroactive adjustment of prediction rules or to the non‐randomness of seismicity. A simple prediction algorithm accounting for earthquake clustering (foreshock‐mainshock‐aftershock sequences), yields similar or even better forecast results. If we remove dependent events from the catalog, the ‘prediction effect’ becomes statistically insignificant. For the PDE (NOAA) catalog the test shows that the VAN predictions' rate of success can be attribu
ISSN:0094-8276
DOI:10.1029/95GL03417
年代:1996
数据来源: WILEY
|
8. |
Reply to “VAN earthquake predictions‒An attempt at statistical evaluation,” by Y.Y. Kagan |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1319-1321
P. Varotsos,
M. Lazaridou,
Preview
|
PDF (401KB)
|
|
摘要:
A number of successful VAN predictions were omitted inKagan's[1996] calculations. Furthermore, Kagan assumed an arbitrary cutoff in the magnitude of the earthquakes. We show that, for the SI‐NOA catalog, the VAN high rate of success is not due to a retroactive adjustment of prediction rules, or to the non‐randomness of seismicity. If we remove dependent events from the catalog, the prediction effect still remains statistically significant when considering the prediction rules suggested by VAN since 1986, and not omitting a number of successful predictions referring to main shocks. The simple prediction algorithm, suggested byKagan[1996], cannot be considered as giving similar results with VAN; this is so because, as we show with simple examples, the number of successes is not a safe criterion to compare two methods.Kagan's[1996] algorithm predicts almost exclusively aftershocks, while VAN predicts a comparable (or even larger) number of earthquakes most of which are main shocks. For the PDE(NOAA) catalog, a re‐analysis shows that the VAN predictions' rate of success cannot be attributed to c
ISSN:0094-8276
DOI:10.1029/96GL00909
年代:1996
数据来源: WILEY
|
9. |
Precursor candidacy and validation: The VAN Case so far |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1323-1326
Francesco Mulargia,
Paolo Gasperini,
Preview
|
PDF (483KB)
|
|
摘要:
A reliable identification of seismic precursors requires a two‐step statistical approach: a retrospective “learning” step to establish the candidate precursors on the basis ofad hocchosen laws of the game, and a “validation” step which evaluates candidate precursors on an independent data set according to the previously established laws of the game. No published work has so far provided a sound support to the candidacy of the VAN signals as earthquake precursors. The problem is complicated by a chronic fuzziness in the text of VAN's predictions, which often allows different interpretations, and by the absence of any clear statement by VAN of the laws of
ISSN:0094-8276
DOI:10.1029/95GL03456
年代:1996
数据来源: WILEY
|
10. |
VAN: Candidacy and validation with the latest laws of the game |
|
Geophysical Research Letters,
Volume 23,
Issue 11,
1996,
Page 1327-1330
Francesco Mulargia,
Paolo Gasperini,
Preview
|
PDF (421KB)
|
|
摘要:
We reassess the set of VAN predictions in the period 1987–1989 with the latest laws of the game [Varotsos et al.,1996]. This reassessment does not modify our previous conclusions: VAN predictions have alarm and success rates too low to accept them as candidate precursors, and a degree of time association with earthquakes which suggests rejection of any precursor nomination. At the same time, this reassessment confirms the tendency of predictions in the period 1987–1989 to significantly follow earthquakes, with 13 earthquakes tracked by a spatially close prediction within 11 days. A validation study of the independent set of VAN predictions in the period 1990–1992, leaving the laws of the game unchanged while restricting the operative region to account for network reduction, denies any significant association of VAN predictions with earthquakes, but confirms the tendency of VAN predictions to follow rather than precede earthq
ISSN:0094-8276
DOI:10.1029/95GL03455
年代:1996
数据来源: WILEY
|
|