Com­plex world of quake fore­cast­ing

Kaikoura Star - - YOUR LOCAL NEWS -

The 7.8 mag­ni­tude Kaiko¯ura Earth­quake that struck just af­ter mid­night last Novem­ber 14 was a mon­ster that tore great scars in the land­scape and ripped up the seabed.

Thank­fully, it wasn’t fol­lowed by a ter­ri­fy­ing pro­ces­sion of ma­jor af­ter­shocks.

There was the har­row­ing mag­ni­tude 5.7 quake that hit par­tic­u­larly hard the town of Scargill in North Can­ter­bury on Novem­ber 22, set­ting res­i­dents in the north­east of the South Is­land on edge again.

That turned out to be one of only two quakes classed as ‘se­vere’ by GeoNet in the past year that didn’t hap­pen on Novem­ber 14. The other was a 5.5 mag­ni­tude quake cen­tred near Sed­don on De­cem­ber 4.

A month af­ter the Kaiko¯ura quake, GeoNet pub­lished an ar­ti­cle, which said the af­ter­shock se­quence was ‘‘fall­ing within or just be­low the lower end of our fore­cast range’’.

It helped that the bot­tom of the ranges for dif­fer­ent size quakes and time pe­ri­ods was of­ten zero.

In its lat­est af­ter­shock fore­cast - for the two months from Oc­to­ber 19 - GeoNet puts the num­ber of 5.0-5.9 mag­ni­tude quakes ex­pected at from 0-6.

For mag­ni­tude 6.0-6-9 and mag­ni­tude 7.0 and above, the range is 0-1.

GeoNet makes it very clear these are fore­casts.

They are prob­a­bil­i­ties of earth­quakes hap­pen­ing in a spec­i­fied mag­ni­tude range, spa­tial area and cer­tain time pe­riod.

They are not pre­dic­tions giv­ing spe­cific times and lo­ca­tions. ‘‘At present there is no sci­en­tific way to ac­cu­rately and

‘‘At present there is no sci­en­tific way to ac­cu­rately and re­li­ably pre­dict when and where a big earth­quake is go­ing to hap­pen next.’’

re­li­ably pre­dict when and where a big earth­quake is go­ing to hap­pen next,’’ GeoNet said.

It is us­ing ob­ser­va­tions of pre­vi­ous earth­quake pat­terns in New Zealand and around the world and the vast amounts of data from the Novem­ber 14 earth­quake to try to im­prove the fore­cast­ing of af­ter­shock se­quences fol­low­ing fu­ture big quakes.

GNS Science haz­ard and risk sci­en­tist Dr An­nemarie Christo­phersen said the Kaiko¯ura af­ter­shock se­quence ‘‘seems a bit less pro­duc­tive than the av­er­age’’ New Zealand af­ter­shock se­quence.

If some peo­ple felt the af­ter­shock se­quence fol­low­ing Kaiko¯ura was less ac­tive than af­ter the Can­ter­bury earth­quakes, one rea­son might be that many Can­ter­bury af­ter­shocks hap­pened in ur­ban ar­eas where many peo­ple felt them, Christo­phersen said.

Also, the Can­ter­bury se­quence was in­vig­o­rated with fur­ther large af­ter­shocks about ev­ery six months for a cou­ple of years.

To com­pli­cate the count­ing of earth­quakes above a cer­tain size for com­par­i­son of the Kaiko¯ura and Can­ter­bury se­quences, the mag­ni­tude of earth­quakes of some sizes used to be slightly over­es­ti­mated be­fore the in­tro­duc­tion of an au­to­mated earth­quake pro­cess­ing sys­tem in 2012.

Ob­vi­ously af­ter­shock fore­cast­ing is com­pli­cated. The mod­els used to fore­cast af­ter­shocks use pre­vi­ous earth­quakes as in­put, and are up­dated as new earth­quakes hap­pen.

For the best re­sults, all earth­quakes would have ideally been pro­cessed in the same way, and all earth­quakes above a cer­tain thresh­old size would be recorded. But that’s not the case. Af­ter a big earth­quake, the seis­mic net­work gets overloaded and smaller earth­quakes are ini­tially not de­tected.

In a pa­per writ­ten for the New Zealand So­ci­ety for Earth­quake En­gi­neer­ing on op­er­a­tional earth­quake fore­cast­ing (OEF), Christo­phersen and col­leagues at GNS Science said care­ful fil­ter­ing and seis­mic pro­cess­ing al­lowed some of the smaller quakes to be de­tected.

‘‘How­ever, this is a time­con­sum­ing man­ual process. For ex­am­ple, it took about 18 months for the first 24 hours to be pro­cessed fol­low­ing the M7.1 Darfield earth­quake.’’

Sec­ondly, earth­quake mag­ni­tude has been mea­sured in dif­fer­ent ways, and there have been step changes in the way earth­quakes have been pro­cessed in New Zealand.

For ex­am­ple, from the mid 1980s un­til 2011 a seis­mic pro­cess­ing sys­tem called CUSP was used, then in 2012 GeoNet in­tro­duced the more au­to­matic pro­cess­ing sys­tem called SC3.

‘‘We are still do­ing work to un­der­stand the ef­fect of the change of the pro­cess­ing soft­ware on the mag­ni­tudes,’’ Christo­phersen said.

An­other wrin­kle is that the record­ing of smaller earth­quakes has im­proved over the decades. The den­sity of seis­mic sta­tions started to in­crease in the 1960s, and from that point earth­quakes of about mag­ni­tude 4.0 and above could gen­er­ally be de­tected.

If that doesn’t make things dif­fi­cult enough, the mag­ni­tude of the Novem­ber 14 earth­quake was up­graded from M7.5 to M7.8 two days af­ter the event.

The OEF pa­per said that for one sim­ple af­ter­shock model that im­plied the ex­pected num­ber of earth­quakes dou­bled in any given time pe­riod.

Un­der the ini­tial mag­ni­tude, the num­ber of af­ter­shocks had agreed ‘‘rea­son­ably well’’ with the num­ber of af­ter­shocks fore­cast, but the num­ber of ac­tual earth­quakes fell be­low the fore­cast when the mag­ni­tude went up.

‘‘It is not clear whether the ini­tial agree­ment be­tween the data and the model is due to in­com­plete de­tec­tion of many large af­ter­shocks, or whether the se­quence is much less pro­duc­tive than the av­er­age New Zealand se­quence,’’ the pa­per said.

Christo­phersen said even tak­ing into ac­count any M5-M5.9 quakes missed in the first cou­ple of days, the af­ter­shock se­quence still seemed a bit less pro­duc­tive than av­er­age.

‘‘But we do not un­der­stand the dif­fer­ence in mag­ni­tude well yet that might ex­plain a large com­po­nent of that dif­fer­ence,’’ she said.

The GNS sci­en­tists are de­vel­op­ing hy­brid mod­els for fore­cast­ing earth­quakes at short, medium and long time pe­ri­ods.

The lat­ter is also known as time-in­vari­ant and is used for such things as en­gi­neer­ing de­sign codes.

Ba­si­cally the mod­els rely on two types of earth­quake clus­ter­ing in time and space: the de­cay of af­ter­shocks fol­low­ing a large earth­quake, and the in­crease of seis­mic­ity build­ing up be­fore a large earth­quake.

Mod­els are de­vel­oped by fit­ting data from past earth­quakes. The mod­els are then tested against other data to see how well they work.

‘‘We have learned that hy­brid mod­els that com­bine dif­fer­ent sources of data ... work bet­ter than mod­els that are based on one con­cept and source of data only,’’ Christo­phersen said.

‘‘We still try new data sources and ways of com­bin­ing them, and then test them against each other and the data.’’

Re­search done in New Zealand has also found that the rate and size of smaller earth­quakes can in­crease over years or decades be­fore a large earth­quake. That’s the clus­ter­ing be­fore a large earth­quake men­tioned above.

A model, known as EEPAS, based on that ob­ser­va­tion has been de­vel­oped by Dr David Rhoades of GNS Science. It is be­ing tested in this coun­try and other earth­quake-prone ar­eas around the world. Two ver­sions of EEPAS are part of the Kaiko¯ura hy­brid fore­cast.

‘‘The EEPAS model has been ap­plied to a num­ber of re­gional earth­quake cat­a­logues and con­sis­tently fore­casts ma­jor earth­quakes bet­ter than timein­vari­ant mod­els,’’ the OEF pa­per said.

Newspapers in English

Newspapers from New Zealand

© PressReader. All rights reserved.