ISCGEM Global Reference Earthquake Instrumental Catalogue 19002009 D

ISC-GEM Global Reference Earthquake Instrumental Catalogue (19002009) D. Di Giacomo, I. Bondár, E. R. Engdahl, D. A. Storchak, W. H. K. Lee, A. Villaseñor, J. Harris, P. Bormann 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn

Motivation Seismic hazard studies need accurate knowledge of the spatial distribution of seismicity and the magnitude-frequency relation. Existing catalogues for past century, however, are compilations of different sources covering different time periods, and therefore contain inhomogeneous locations and magnitudes. There is the need for an improved global instrumental catalogue for large earthquakes spanning the entire 100+ years period of instrumental seismology. 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 2

ü Project in a nutshell Collecting, digitising and Cut-off magnitudes: ü 1900 -1917: MS≥ 7. 5 worldwide + processing data from a multitude smaller shallow events in stable of historical sources for continental areas earthquakes occurred up to 1970; ü 1918 -1959: MS≥ 6¼ ü 110 years of relocated ü 1960 -2009: MS≥ 5. 5 earthquake hypocenters; ü recomputed MS and mb values This Catalogue is unique for relocated events using because uniform procedures; it contains homogeneous ü MW values (with uncertainty) locations and magnitude based on: estimates with the estimates of 1. seismic moment from uncertainty for the entire GCMT (mainly 1976 -2009); period 1900 -2009 done using 2. seismic moments from the same tools and techniques literature search for to the extent possible. earthquakes up to 1979; 3 3. proxy values based on 43 Nordic Seismology Seminar, 24 -26 October 2012, Tallinn rd

Phase and Amplitude Data Collection Period Body Wave Arrival Times Body/Surfac e Wave Amplitudes 19001917 19181959 19601970 ~10, 000 ~1, 000, 0 00 Quality station bulletins DIGITALLY NOT AVAILABLE BEFORE THIS PROJECT 19712009 DIGITALLY AVAILABLE, ISC database Major Sources of Phase Data: • Gutenberg Notepads (19041917) and BAAS (1913 -1917) • ISS Bulletins (1918 -1963) 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 4

Processing historical seismological bulletin 1906 San Francisco earthquake report from Göttingen (Germany) station bulletin. ~15, 000 individual seismic bulletins from 290 institutions over the period 1904 – 1970 were recovered from ISC storage The same report stored in digital format in the ISC database. Period and amplitude data finally available for magnitude recomputation. 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 5

Amplitude Data from Quality Station Bulletins • ~110, 000 “brand new” amplitudes up to 1970 now available in the ISC database • Effort equivalent to ~70 personmonths Time Coverage: UPP, RIV, and LPZ nearly continuous, gaps for other stations Station timeline 6

Earthquake Location Procedure Location method: 1. Determine event depth using the EHB style of processing (Engdahl, van der Hilst and Buland, 1998): a) comprehensive analysis of near-event surface reflections off the earth surface inland ocean bottom or water surface in the oceans; b) Station patch corrections; 2. Use the new ISC location algorithm (Bondár and Storchak, 2011) with earthquake depths fixed to those from EHB analysis: a) more accurate epicentre locations due to correlated error structure taken into account (removes bias from uneven geometrical positioning of stations) b) independent depth confirmation using depth phase stacking; 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 7

Earthquake Relocation results Before relocation… 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 8

Earthquake Relocation results …. after relocation. 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 9

Earthquake Relocation results Before After 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 10

MS and mb recomputation The recomputed MS and mb benefit from: 1) amplitude data added up to 1970; 2) station magnitudes consistent with newly computed hypocentre solutions; 3) network magnitudes based on several station measurements using alphatrimmed median (α = 20%) of the single station magnitudes (no network magnitude based on one station only). 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 11

Mw from GCMT and literature search MW from GCMT is available from 1976 (plus some deep earthquakes between 1962 and 1975). For 970 relocated earthquakes direct measurements of M 0 were compiled from the literature. For the remaining relocated earthquakes, proxy MW values are obtained from the recomputed MS and mb using new empirical 12

MW proxy based on recomputed MS Num=174 72 • Data population strongly dominated by earthquakes with magnitude below 6; • The relationship between MS and MW is not linear over the entire distribution; • Median values for separated bins (dashed black line) suggest that a nonlinear model could fit well the data. • Histogram equalization scheme: bins of varying width so that each bin contains the same number of data points. • For each bin a randomly chosen 10% of the data is assigned to the validation dataset, while the 90% to the training dataset used to obtain 13 the regression model.

MW proxy based on recomputed MS We applied a non-linear regression using an exponential model of the form My = exp(a+b*Mx)+c (EXP, purple). • The exponential model follows well the median values curve over the entire population. • Proxy MW vs true MW (=10% of the original population not used for deriving 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn the model). 14

MW proxy based on recomputed mb • The exponential model follows well the median values curve close to the saturation level. Seminar, of m 24 -26 43 Nordic Seismology b. October 2012, Tallinn rd 15

Magnitude composition of the ISC-GEM catalogue Direct MW per year Proxy MW per year 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 16

Magnitude composition: Centennial vs ISC-GEM catalogue Centennial catalogue ISC-GEM catalogue 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 17

Magnitude distribution of the ISC-GEM catalogue 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 18

Frequency-Magnitude distributions • Seismicity rates for large (M>7. 5 -7. 6) earthquakes better assessed considering a long time window (violet) Mc=6. 4 Mc=5. 6 • For moderate earthquakes the modern period (red) is a better basis for magnitude-frequency studies, whereas for strong to major shallow earthquakes the entire ISC-GEM catalogue may be used 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 19

Conclusions ü We collected, digitised and processed an unprecedented amount of phase and amplitude data for earthquakes occurred before 1970; ü In the 110 years covered by the ISC-GEM catalogue, the relocation provided significant improvements especially in the first part of past century; ü We recomputed MS and mb using uniform procedures, and new nonlinear relationships are used to obtain MW proxies when direct computation of M 0 from GCMT or literature is not available; ü The ISC-GEM Global Instrumental Earthquake Catalogue represents the final product of one of the ten global components in the GEM program, and will be available to researchers at the ISC website (www. isc. ac. uk). 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 20

THANK YOU 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 21

Appendix 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 22

ISS bulletins (19181963) (predecessor of the ISC, phase data Converted into digital form by scanning only!) the bulletin pages and applying an optical character recognition (OCR) procedure (Engdahl and Villaseñor, 2002) Biggest source of earthquake data from 1918 to 1963. • Over 1. 1 million phases (~1000 seismic stations between 1918 and 1963) from ISS have been used in the relocation process; over 730, 000 have been inserted into the ISC database during this project for earthquakes occurred between 1918 and 1959. • Over 5000 phases (from ~160 seismic stations) have been added before 1918 (mostly from BAAS and G&R notepads). 23

Earthquake Relocation results Before After 24

MW proxy based on recomputed MS • The relationship between MS and MW is • Data population strongly dominated by not linear; earthquakes with magnitude below 6; • Authors normally perform bi-linear • Median values for separated bins regression splitting the dataset at MS = (dashed black line) suggest that a non 6. 1; linear model could fit well the data over the • This separation, however, is arbitrary entire distribution. 25 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn because slope change occurs in a

MW proxy based on recomputed The histogram MS equalization defines magnitude bins varying width so that each bin contains the same number of data points. For each bin a randomly chosen 10% of the data is assigned to the validation dataset, while the 90% to the training dataset used to obtain the regression model. 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 26

Magnitude composition of the ISC-GEM catalogue 43 rd Nordic Seismology Seminar, 24 -26 October 2012, Tallinn 27

MW proxy based on recomputed mb We applied both the GOR (green) and a non-linear regression using an exponential model of the form My = exp(a+b*Mx)+c (EXP, purple). • The exponential model follows well the median values curve close to the saturation • level Proxyof. Mm. Wb. vs true MW (=10% of the original population not used for deriving the models), show EXP model works better than GOR 28