EC project "Review of Historical Seismicity in Europe" (RHISE) 1989-1993



[Deliverables and queries] [Presentation] [Homepage]


Romano Camassi*, Santi Mirenna*, Alessandro Rebez**,
yMassimiliano Stucchi* and Angela Zerga*
* Istituto di Ricerca sul Rischio Sismico, CNR, via Ampère 56, 20131 Milano, Italy.
** Osservatorio Geofisico Sperimentale, P.O. BOX 2011, 34016 Trieste, Italy.

Problems and pitfalls in the compilation of comprehensive, parametric earthquake catalogues


Foreword
This paper discusses some problems arising in the compilation of comprehensive parametric earthquake catalogues from regional ones, through the analysis of some cases. They are taken from current parametric earthquake catalogues and, mostly, from the European catalogue by Van Gils and Leydecker (1991).
The authors want to make clear that most of these catalogues are to be considered as useful tools, in spite of the errors they might contain. This is definitely true also for the Van Gils and Leydecker catalogue; unfortunately, circumstances did not allow the compilers to follow the recommendations set up at the beginning of their project, which are found in the introduction to the catalogue itself (Van Gils, 1988).
The considerations which follow are aimed to encourage more reliable results; they are not intended, by any means, as a sort of "in pursuit of errors".

Introduction
The compilation of parametric earthquake catalogues covering areas, and/or time-windows, broader than those covered by two, or more, existing catalogues, is a problem well known to many investigators; some aspects are discussed for instance by Van Gils (1988), Stucchi (1994) and others.
The compilation is generally performed converting the input catalogues into a comprehensive format, merging them into a single file and then sorting out possible duplications: this operation is usually accomplished with the help of almost automatic tools. To follow this procedure implies the following, hidden assumptions:
- parameters of input catalogues are homogeneous; that is, each parameter called in the same way represent the same quantity and is expressed in the same unit;
- parameters are derived from primary data with homogeneous procedures;
- compilers are confident that their authomatic routines are so highly performing to be able to master possible duplication or similar problems.
Unfortunately, these assumptions seldom meet the reality. It is well known to most compilers, and it is discussed also in this volume, that parametric catalogues are far from being homogeneous and transparent. Therefore, the compilation of comprehensive, parametric catalogues often results in propagation and amplification of errors and in increasing internal inhomogeneity of the final product. The most frequent problems are:

- earthquake duplications (or losses)
- inhomogeneity and forced parameterisations.

It is to be stressed that most of these problems are also met inside national catalogues, as they are often compiled following the same procedures.

Earthquake duplications
The most common problem is duplication, that is, the presence of double or multiple entries representing the same event, carrying different set of parameters. The sources of duplication cover a wide range: they go from simple mistakes in the compilation of some parameters to different compilation procedures. But the most frequent source is the use of different sets of primary data and/or errors in historical interpretation.
Duplication problems derived from the latest source are well known as they are discussed for instance by Alexandre (1991), Bellettati et al. (1993) and many others. The examples which follow deal with cases covering also the instrumental period and even the most recent one.

The "Italian" catalogues
The Postpischl (1985) catalogue put together many entries of varied origin. The main contributing sets are parametric catalogues, such as: Enel file (1978; about 20600 entries), Carrozzo et al. (1973; about 435), Dell'Olio and Molin (1980; about 5500), Bernardis et al. (1977; about 1800), Carrozzo et al. (1975; about 7600). While the first two sets cover the whole territory, the other sets cover limited areas (Central Italy, Eastern Alps, Calabria and Sicily).
After merging the input catalogues, Postpischl sorted out many problems authomatically; but the huge task and the limited time did not allow avoiding some pitfalls, to be mastered with careful analysis only. Therefore, a limited, but significant number of duplications survived; an example follows.

1939.02.11. The two following entries are compiled on the basis of Bernardis et al. (1977) and Enel (1978) respectively (Fig. 1a):

Date Time Lat Lon M Io Ep.Area
1939 02 11 11 16 46 06 13 06 65 FAGAGNA (B)
1939 02 11 11 16 54 44 04 11 39 52 70 MARRADI (E)


The search for the roots of the two entries leads respectively to Caloi (1939) and to Malaroda and Raimondi (1957), via Karnik (1969). Malaroda and



Fig. 1 - Epicentral location of the duplicated entries.

Raimondi initially reported two entries; in their final table, however, only the second one, the origin of which is unknown, survived. Finally, Bernardis et al. (1977), dealing only with the Eastern Alps area, preserved the second entry which, later on, was assumed also by Postpischl (1985).

Neighbouring countries catalogues: Italy and Slovenia

1889.09.02. The following entries are taken from Ribaric (1982) and Postpischl (1985) respectively (Fig. 1):

Date Time Lat Lon M Io Ep. Area
1899 09 02 45 87 13 98 30 BUDANJE (*)
1899 09 02 10 10 45 37 11 15 50 VALDAGNO (o)

The search for the roots of the second record leads to Cancani (1900). The root of the Ribaric (1982) record are Mojsisovics (1900) and Bernardis et al. (1977). Strangely enough, Bernardis et al. (1977) quote an unpublished Slovenian catalogue by Ribaric as a source! It must be said that cross- references between these two catalogues are not an unfrequent case.

The Van Gils and Leydecker (VGL) catalogue
As already remembered, the compilation of this catalogue did not follow the recommendations which are found in the introduction to the catalogue itself (Van Gils, 1988). With respect to duplications, the compilers had the unpleasant task to merge national catalogues without having access to the supporting datasets: therefore, some duplications escaped their control.

The 1295, Central Alps earthquake. This earthquake, still a key one for the assessment of seismicity and seismic hazard in Central Alps, is represented in the VGL catalogue by two entries (Fig. 2), which come from the Swiss and the Postpischl catalogues respectively.

Cat Date Time Lat Lon M Io Ep. Area
CH 1295 09 04 46.83 09.53 8.0 REICHENAU (*)
ITA 1295 09 17 45.75 09.50 7.0 CAPRINO B. (+)

Æt between the two events is 14 days, while the distance between the two epicentres is about 150 km: such a distance usually escape authomatic investigation. The background of the duplication is found in historical interpretation mistakes: Fig. 3 shows the roots of the two entries.
Actually, coeval sources mentioning Monza (Obituary, XIII-XIV) and Chur (Chronicon Osterhoviense, XIII-XIV) agree that an earthquake was felt "die sabbati tertio die mensis septembris" ("on Saturday, the third of September"), or "in octava Sancti Augustini", which means "the eighth day after the feast of St. Augustin" (which falls on August 28), that is September 4 (Albini et al., 1994).
On the contrary, a later Italian author, Corio (1503), made a mistake: he probably interpreted "Saturday, the third of September" as "the third Saturday of September", that is September 17.
Bonito (1691), using sources from the two "trees", list two earthquakes. Baratta (1901) is aware that the two dates correspond to one earthquake only, but he relies mostly on Corio (1503).



Fig. 2 - 1295, September earthquake in Central Alps: epicentral locations (from VGL)
and intensity data points (from Albini et al., 1994).

For these reasons the Italian catalogue compiled a second record; this was also favoured by the fact that the Italian sources mention Italian localities, while German sources mention localities North of the Alps only. The two records then survived the "duplication inspections" performed by VGL.

Fig.3 - Roots of the two entries representing the 1295, September earthquake in Central Alp.

1512-1513

Cat Date Time Lat Lon M Io Ep. Area
CH 1512   46.38 08.97   7.0 BIASCA
CH 1512 02 08   46.30 09.38   6.0 CHIAVENNA
ITA 1513 02 10   46.17 09.00 5.7 9.0 BELLINZONA


The duplication of the first entry with the third one was analysed by Albini et al. (1988). The most striking fact is not that both entries refer to the same event, but that the event itself is a fake quake; actually, what happened in the location proposed by the Swiss catalogue was a landslide. The situation is complicated by the second entry, the reliability of which is questionable.

1542

Cat Date Time Lat Lon M Io Ep. Area
CH 1542 11 08 47.82 10.05 6.0 WIGGENSBACH
D 1542 11 08 47.83 10.00 6.0 BAYERN



Very probable duplication.

1616 and 1623

Cat Date Time Lat Lon M Io Ep. Area
CH 1616 02 19 04 30 00 46.47 08.52 5.0 VAL BEDRETTO
CH 1616 02 29 46.83 08.40 8.0 ENGELBERG
CH 1623 02 20 46.30 09.78 6.0 CHIESA
ITA 1623 03 01 01 00 00 46.22 10.15 5.0 BIANZONE

The first of these possible duplications originated inside the Swiss catalogue, the other one originated when merging the Swiss and the Italian catalogues. They are probably due to different calendars (Julian and Gregorian) used by the sources.

1644
Cat Date Time Lat Lon M Io Ep. Area
ITA 1644 02 15 44.00 06.75 5.7 9.0 VAL VESUBIE
FRA 1644 02 15 04 30 00 43.88 07.30 8.5 ALPES-MARIT.

Duplication originated from different epicentral locations, performed from the same dataset, but following different parameterisation procedures.

1670

Cat Date Time Lat Lon M Io Ep. Area
ITA 1670 07 16 15 45.43 10.97 4.4 6.5 VERONA
A 1670 07 17 02 00 00 47.30 11.50 5.3 8.0 TIROL

Well known example, discussed for instance by Guidoboni and Stucchi (1993). The reason for this duplication is the use of different time-systems by the sources, and uncorrect interpretation by the compilers.

1690
Cat Date Time Lat Lon M Io Ep. Area
D 1690 11 24 48.90 10.50 6.0 BAYERN
A 1690 12 04 15 45 00 46.60 13.80 6.2 9.0 KAERNTEN
ITA 1690 12 04 18 46.60 13.85 5.7 9.0

While the first entry is a possible duplication, still under investigation (Eisinger and Gutdeutsch, 1994), the Austrian and Italian entries are just unsorted duplications.

1743

Cat Date Time Lat Lon M Io Ep. Area
ITA 1743 02 20 22 45.55 11.55 3.9 5.5 VICENZA
GR 1743 02 20 39. 0 20.30 6.9 9.0 CORFU

Classical example of duplication due to the use of a limited set of sources (a local compiler, Piovene (1888), from Vicenza, Italy), without data cross-checking from surrounding areas.
The earthquake was actually felt in Vicenza: the distance between the two epicentres is more than 500 km.

1778


Cat Date Time Lat Lon M Io Ep. Area
A 1778 01 28 47.20 09.60 2.5 6.0 VORARLBERG
D 1778 01 28 02 30 47.90 07.60 6.5 BADEN-WUERT.
CH 1778 01 29 00 30 00 47.20 09.82 5.0 GR WALSERT.


Possible duplication.

Lost earthquakes

Exactly in the symmetrical way as for duplications, earthquakes may get lost from parametric catalogues. This can happen when a catalogue covering a large area is partly used because, for many reasons, another catalogue, covering a smaller area, inside the previous one, is preferred for the small area itself.

For instance, in the catalogue of the Balkan region (Shebalin et al., 1974) an entry describes, without parameters, an earthquake of 1556 which, according to Hoff (1840) and his sources, affected "Baiern, Österreich, Windischmark, Ungarn, Croatien, Dalmatien und Mähren" (Fig. 4), damaging 26 localities.
This entry today survives in a Croatian earthquake file only: the epicentre is located in the present Austrian territory and Io is 10! Right or not these parameters, it is a matter of fact that the Austrian part of the VGL catalogue ignores it, while no trace of this earthquake is found in the territories of the neighbouring countries, according to their catalogues.


Fig. 4 - Areas affected by the 1556 earthquake (from Hoff, 1840) and epicentre (j)
from the Croatian earthquake file.

Wrong puzzle solutions
It can also happen that only some "pieces of earthquake" get lost or, simply, that they float around looking for they parent earthquakes and then they stick to the nearest one coming. This might be due to the fact that earthquake parameters are often derived from different sets of data - this is mostly the case with magnitudes - and put together taking origin time as a guide.
But sometimes things go wrong. For instance, a recent, oustanding magnitude re-evaluation of earthquakes of Italy and sorrounding areas (Margottini and Screpanti, 1991), produced a final catalogue attaching the new evaluated magnitudes to the Postpischl (1985) parameters (To, lo, jo, etc). In this way, two large magnitude earthquakes, with very small Io, appeared in nearly aseismic zones (Fig. 5). In this case the VGL catalogue turns out useful, providing the solution:


Cat Date Time Lat Lon M Io Ep. Area
ITA 1943 05 28 00 25 46 04 9 15 40 MENAGGIO
D 1943 05 28 01 24 08 48 16 8 58 80 BADEN-WUERTTEMB.
ITA 1955 11 23 06 24 40 45 54 11 00 40 ROVERETO
FRA 1955 11 23 06 39 12 47 24 5 58 60 HAUTE-SAONE


Fig. 5 - The two couples of duplicated epicentres biasing the puzzle.

The two magnitudes are to be attached to two well assessed earthquakes in Germany and France respectively, of which the Italian entries are just duplications. Unfortunately, neither Postpischl, nor VGL, nor even Margottini and Screpanti sorted these duplications out.

Inhomogeneity and forced parameterisation
If the assumptions formulated in the introduction are not met by data, the comprehensive catalogue may result formally homogeneous, but its content may still be deeply inhomogeneous.
For instance, the VGL catalogue amalgamated national catalogue which might be internally homogeneous, but which are surely not homogeneous one to the other with special reference to: time-window, historical sources, intensity scales, parameterisation procedures, magnitudes, completeness.

More in detail, the VGL catalogue amalgamated epicentral intensities expressed in different macroseismic scales and assessed by varied "hands", including also figures like 6.3, and so on.
In the same way, magnitudes of different nature, both instrumental and macroseismic, were taken from the input catalogues and put all together in the same column, without qualification or warning for the users. For instance, Fig. 6 clearly indicates the presence of macroseismic magnitudes mixed together with instrumental ones in the Italian data.



Fig. 6 - Distribution of magnitude versus time for the Italian entries in the VGL catalogue.

On the contrary, Fig. 7 shows that, in the data for Spain, magnitudes cover a limited time-window and they are nearly only instrumental.

One result of this situation is that, when a user, unaware of such problems, plots magnitudes against epicentral intensities, he gets the picture shown in Fig. 8. With the current regression procedures a M = M(Io) correlation can certainly be obtained: but, what value will it have?


Fig. 7 - Distribution of magnitude versus time for the Spanish entries in the VGL catalogue.


Fig. 8 - Magnitude versus epicentral intensities for the Austrian (A), Greek (GR) and Italian (I) entries in the VGL catalogue.

The depth column shows even more puzzling problems. Fig. 9 shows that, for the Italian data, many depths are "fixed", according to current procedures. Even more striking is that, for the Greek entries, depth is constant at 70 km for most entries before 1960 (Fig. 10). Asked about the reason, the Greek colleagues said that the material handled to Van Gils contained only the information that "most Greek events are intermediate".


Fig. 9 - Distribution of depth versus time for the Italian entries in the VGL catalogue.


Fig. 10 - Distribution of depth versus time for the Greek entries in the VGL catalogue.

Apparently, this information was turned into "70 km" for all events which did not carry a depth, probably because the information did not fit in the space available in the depth column.
Now the problem is that, if an user wants to investigate earthquake depths in Greece, he might be encouraged to assume the VGL data as an evidence of a very large, active layer, at a depth of about 70 km, covering all Greece and sorrounding areas.

Conclusions
The problems discussed show that handling those useful tools called parametric catalogues with little care, might result in a further step of earthquake data spoiling, after the main step represented by the parameterisation of historical records.
Problems of duplication appear in some way relatively easier to detect, with respect to homogeneity problems. For mastering all these problems properly, it is necessary to retrieve and process the primary data: playing with parameters usually increases the confusion.

Finally, it is to be stressed that a further problem is represented by the use which can be made of catalogues. Though this problem is not compiler's responsibility, however, compilers should be aware of it; it must be admitted that compilers sometimes do not help in this direction.

Acknowledgements
Working hard on this topic, authors are well aware of how time-consuming and unrewarding the catalogue compilation may be; they want to express respect and appreciation for the work of the colleagues.

References
Albini, P., Morelli, G. and Stucchi, M., 1994. Alcuni terremoti importanti per l'Alta Valtellina. In: Studi sismici in Alta Valtellina, ISMES-IRRS (CNR), pp. 24-50.
Albini, P., Bellani, A. and Stucchi, M., 1988. Terremoti e frane nelle Alpi Centrali. Atti del 7° Convegno Annuale del Gruppo Nazionale di Geofisica della Terra Solida, Roma, I, pp. 129-146.
Alexandre, P., 1991. The Seismic Cataclism of 29 March 1000: Genesis of a Mistake. Tectonophysics, 193: 45-52.
Baratta, M., 1901. I Terremoti d'Italia. Saggio di Storia, Geografia e Bibliografia Sismica Italiana con 136 sismocartogrammi. Torino, 950 pp.
Bellettati, D., Camassi, R. and Molin, D., 1993. Fake quakes in Italy through parametric catalogues and seismological compilations: case histories and typologies. Terra Nova, 5: 488-495.
Bernardis, G., Giorgetti, F., Nieto, D. and Slejko, D., 1977. Earthquakes Catalogue for Eastern Alps Region. OGS Trieste [computer printout].
Bonito, M., 1691. Terra tremante overo continuatione de' terremoti dalla Creazione del Mondo fino al tempo presente Ö Napoli, 822 pp.
Caloi, P., 1939. Il terremoto dell'Appennino tosco-romagnolo dell'11 Febbraio 1939, La Ricerca Scientifica, X, 11.
Cancani, A., 1900. Notizie sui terremoti osservati in Italia durante líanno 1899. Bollettino della Società Sismologica Italiana, V.
Carrozzo, M.T., Cosentino, M., Ferlito, A., Giorgetti, F., Patané, G. and Riuscetti M., 1975. Earthquakes Catalogue of Calabria and Sicily (1783-1973). Quad. Ric. Scient., 93, 216 pp..
Carrozzo, M.T., De Visentini, G., Giorgetti, F. and Iaccarino, E., 1973. Catalogue of Italian Earthquakes. CNEN, RT/PROT (73) 12, Roma.
Dell'Olio, A. and Molin, D., 1980. Catalogo macrosismico del Lazio dall'anno 1000 al 1975. ENEA, Roma [internal report].
Eisinger, U. and Gutdeutsch, R., 1994. The Villach earthquake of December 4th, 1690 in the German sources. This volume.
ENEL, 1978. Catalogo dei terremoti italiani dallíanno 1000 al 1975. Roma [computer printout].
Guidoboni, E. and Stucchi, M., 1993. The contribution of historical records of earthquakes to the evaluation of seismic hazard. Annali di Geofisica, XXXVI, 3-4: 201-213.
Hoff, K.E.A. von, 1840. Chronik der Erdbeben und Vulcan-Ausbrüche, IV. Gotha.
Karnik, V., 1969. Seismicity of the European area. I, Dordrecht, 220 pp.
Malaroda, R. and Raimondi, C., 1957. Linee di dislocazione e sismicità in Italia. Bollettino di Geodesia e Scienze Affini, 16, 3: 273-323.
Margottini, C. and Screpanti, A., 1991. La magnitudo dei terremoti italiani dal 1900 al 1986. ENEA, Roma [internal report].
Mojsisovics, E. V., 1900. Allgemeiner Bericht und Chronik der im Jahre 1899 innerhalb des Beobachtungsgebietes erfolgten Erdbeben. Mitteilungen Erdbeben Commission der kaiserlichen Akademie der Wissenschaften in Wien, 18.
Piovene C., 1888. Cronaca dei terremoti a Vicenza. Annali dell'Ufficio Centrale di Meteorologia e Geodinamica, s. II, VIII: 46-57.
Postpischl, D. (Editor), 1985. Catalogo dei terremoti italiani dall'anno 1000 al 1980. Quad. Ric. Scient., 114, 2B, Bologna, 239 pp.
Ribaric, V., 1982. Seismicity of Slovenia. Catalogue of Earthquakes (792 A.D.-1981). Ljubljana, 649 pp.
Shebalin, N.V., Karnik, V. and Hadzievski D. (Editors), 1974. Catalogue of earthquakes. UNDP/UNESCO Survey of the seismicity of the Balkan Region. Skopje, 600 pp.
Stucchi, M., 1994. Recommendations for the compilation of a European parametric earthquake catalogue, with special reference to historical records. This volume.
Van Gils, J.M., 1988. Catalogue of European earthquakes and an atlas of European seismic maps. CEC, Report EUR 11344 EN.
Van Gils, J.M. and Leydecker, G., 1991. Catalogue of European Earthquakes with intensities higher than 4. CEC, Nuclear Science and Technology, Report EUR 13406 EN.


[Top]
[Vol.1] [Vol.2] [Deliverables and queries] [Presentation] [Homepage]