ManuscriptpreparedforNat. HazardsEarthSyst. Sci. withversion3.2oftheLATEXclasscopernicus.cls. Date: 10January2011 Identification of earthquake precursors in the hydrogeochemical and geoacoustic data for the Kamchatka peninsula by flicker-noise spectroscopy G.V.Ryabinin1,Yu. S.Polyakov2,V.A.Gavrilov3,andS.F.Timashev4,5 1GeophysicalSurvey,KamchatkaBranchoftheRussianAcademyofSciences,Petropavlovsk-Kamchatsky,Russia 1 2USPolyResearch,Ashland,PA,U.S.A. 1 3InstituteofVolcanologyandSeismology,FarEasternBranchoftheRussianAcademyofSciences, 0 Petropavlovsk-Kamchatsky,Russia 2 4InstituteofLaserandInformationTechnologies,RussianAcademyofSciences,Troitsk,MoscowRegion,Russia n 5KarpovInstituteofPhysicalChemistry,Moscow,Russia a J 7 Abstract. A phenomenological systems approach for iden- cal,geodetic,andseismic(Geller,1997;HartmannandLevy, tifying potential precursors in multiple signals of different 2005;Uyedaetal.,2009;Ciceroneetal.,2009).Electromag- ] n typesforthesamelocalseismicallyactiveregionisproposed neticprecursorysignalsarefurtherclassifiedintothesignals a basedontheassumptionthatalargeearthquakemaybepre- believed to be emitted from within focal zones, such as tel- - a ceded by a system reconfiguration (preparation) at different luricandmagneticfieldanomalies,andradiowavesoverepi- t a time and space scales. A nonstationarity factor introduced centralregions(Uyedaetal.,2009).Thelocalizedchangesin d withintheframeworkofflicker-noisespectroscopy,astatis- electricandmagneticfieldsthatreportedlyaccompanysome s. ticalphysicsapproachtotheanalysisoftimeseries, isused seismic events span a wide range of frequencies, including c as the dimensionless criterion for detecting qualitative (pre- ULF,VLF,ELFandRFfields,andwereobservedinthetime i s cursory)changeswithinrelativelyshorttimeintervalsinar- framefrom2-3yearstodozensofminutespriortoanearth- y bitrarysignals. Nonstationarityfactorsforchlorine-ioncon- quake (Cicerone et al., 2009; Uyeda et al., 2009). Hydro- h centration variations in the underground water of two bore- logical/hydrochemicalprecursorysignalsincludewaterlevel p [ holesontheKamchatkapeninsulaandgeacousticemissions or quality changes in the weeks, days, or hours prior to a inadeepboreholewithinthesameseismiczonearestudied number of earthquakes, groundwater temperature changes, 1 togetherinthetimeframearoundalargeearthquakeonOc- and variations in the concentrations of dissolved ions like v tober 8, 2001. It is shown that nonstationarity factor spikes chlorineormagnesiumusuallyinthetimeframeofmonths 3 7 (potential precursors) take place in the interval from 70 to to days before an earthquake (Hartmann and Levy, 2005; 4 50daysbeforetheearthquakeforthehydrogeochemicaldata Cicerone et al., 2009; Du et al., 2010). Gasgeochemical 1 andat29and6daysinadvanceforthegeoacousticdata. precursorysignalscomprisenumerousanomalousgasemis- . 1 sion observations, the majority of which were reported for 0 the concentration of radon gas in the earth (Hartmann and 1 Levy, 2005; Cicerone et al., 2009). More than 100 stud- 1 1 Introduction ies show that changes in radon exhalation from the earth’s : v crust precede a number of earthquakes by months, weeks, i Earthquake prediction in the time frame of several months X ordays(Ciceroneetal.,2009). Geodeticsignalsmostlyin- to less than an hour before the catastrophic event, which is cludesurfacedeformations(tilts,strains,strainratechanges) r often referred in literature as ”short-term” prediction, has a over distances of tens of kilometers that precede some ma- been a subject of extensive research studies and controver- jor earthquakes by months to days (Cicerone et al., 2009). sial debates both in academia and mass media in the past Seismic precursory signals encompass foreshocks that typi- two decades (Geller, 1997; Geller et al., 1997; Wyss et al., callytakeplacelessthan30daysbeforethemainshockand 1997;Uyedaetal.,2009;Ciceroneetal.,2009). Oneofthe high-frequency (acoustic emission) and very low-frequency key areas in this field is the study of earthquake precursors, precursorysignalsthatarenotdetectedbyconventionalseis- physical phenomena that reportedly precede at least some mographs(IhmleandJordan,1994;Reasenberg,1999;Gor- earthquakes.Theprecursorysignalsareusuallygroupedinto dienkoetal.,2008;Gavrilovetal.,2008). Anotherpromis- electromagnetic,hydrological/hydrochemical,gasgeochemi- ingtypeofpossibleprecursorysignalsisanomalousanimal Correspondenceto: YuriyS.Polyakov behavior for very short time frames (within 2-3 days, usu- ([email protected]) allyhours)priortoalargeseismicevent(Kirschvink,2000; 2 G.V.Ryabininetal.: EarthquakeprecursorsinthedatafortheKamchatkapeninsula Yokoietal.,2003;Lietal.,2009). abrupt structural changes in the system generating the sig- Despite the large number of earthquake precursors re- nal, which makes it a promising candidate to be one of the portedinliterature,mostofwhicharesummarizedbyHart- standard criteria. The nonstationarity factor was previously mann and Levy (2005); Cicerone et al. (2009), an Interna- usedtodetectprecursorsinelectrochemicalandtelluricsig- tionalCommissiononEarthquakeForecastingforCivilPro- nals recorded in the Garm area, Tajikistan prior to the large tection concluded on October 2, 2009, ”the search for pre- 1984Dzhirgatalearthquake(Descherevskyetal.,2003;Vs- cursors that are diagnostic of an impending earthquake has tovskyetal.,2005),geoelectricalsignalsatstationGiuliano, notyetproducedasuccessfulshort-termpredictionscheme” Italypriortoseveral2002earthquakes(Telescaetal.,2004), (ICEFCP, 2009). The reports of the International Associa- andULFgeomagneticdataatGuampriortothelarge1993 tion of Seismology and Physics of the Earth’s Interior con- Guamearthquake(HayakawaandTimashev,2006;Idaetal., tain similar findings (Wyss and Booth, 1997). The lack of 2007).Otherapproachestoidentifyingprecursoryfeaturesin confidence can be attributed to several reasons. First, some earthquake-andvolcano-relatedsignals,whicharebasedon fundamentalaspectsofmanynon-seismicsignals,forexam- different monlinear analysis techniques, were discussed by ple,lithosphere-atmosphere-ionospherecouplingandpropa- Telesca et al. (2010, 2009a,b); Telesca and Lovallo (2009); gationofhigh-frequencyelectromagneticsignalsinthecon- Telescaetal.(2008). ductive earth, are unresolved, and many of the proposed In this study, we consider a combined analysis of two physicalmodelsarequestionable(Uyedaetal.,2009). Sec- different types of signals, hydrogeochemical (sampling fre- ond, the experimental data on precursory signals are often quency of 3 to 6 day−1) and geoacoustic (sampling fre- limitedtofewearthquakesandfewmeasurementsites, they quencyof1min−1), recordedontheKamchatkapeninsula, frequently contain gaps and different types of noise (Hart- Russia. mann and Levy, 2005; Cicerone et al., 2009; Uyeda et al., 2009). Third,differenttechniquesofidentifyingtheanoma- 2 Nonstationarityfactor liesareusedfordifferentsignalsorevenindifferentstudies for the same signal. In some cases, the anomalous changes Here,wewillonlydealwiththebasicFNSrelationsneeded are determined by analyzing the signals themselves (Hart- tounderstandthenonstationarityfactor. Theapproachisde- mann and Levy, 2005; Uyeda et al., 2009; Cicerone et al., scribed in detail elsewhere (Timashev, 2006; Timashev and 2009), while in other cases they are identified by studying Polyakov, 2007; Timashev, 2007; Timashev et al., 2010b). thederivedstatisticsorfunctions,suchasFisherinformation The FNS procedures for analyzing original signal V(t), or scaling parameters (Telesca et al., 2009a,b). Moreover, where t is time, are based on the extraction of information seasonal changes and instrumentation or other background containedinautocorrelationfunction noise often need to be filtered out prior to the identification ofprecursors. ψ(τ)=(cid:104)V(t)V(t+τ)(cid:105), (1) Inviewoftheabovethreeproblems,webelievethatearth- quake precursor research can be advanced by employing a where τ is the time lag parameter. The angular brackets in phenomenological systems approach to the analysis of sig- relation(1)standfortheaveragingovertimeintervalT: nals of different types in the same local geographic region. 1(cid:90) T/2 Weassumethatalargeearthquakemaybeprecededbyasys- (cid:104)(...)(cid:105)= (...)dt. (2) temreconfiguration(preparation)atdifferenttimeandspace T −T/2 scales, whichmanifestsitselfinqualitativechangesofvari- To extract the information contained in ψ(τ), the follow- oussignalswithinrelativelyshorttimeintervals. Forexam- ing transforms, or ”projections”, of this function are ana- ple, such anomalous hydrogeochemical signals may be ob- lyzed: cosine transforms (power spectrum estimates) S(f), served months to weeks before the impending earthquake, wheref isthefrequency, anomalous geoacoustic emissions - only days prior to the event, and anomalous behavior of animals - only hours be- (cid:90) T/2 forethecatastrophe. Inordertotestthisapproachandiden- S(f)= (cid:104)V(t)V(t+τ)(cid:105)cos(2πft1)dt1 (3) tify different signals that may be related to a specific large −T/2 seismic event, one needs to have a standard criterion or a anditsdifferencemoments(Kolmogorovtransientstructural set of standard criteria to detect signal anomalies in virtu- functions)ofthesecondorderΦ(2)(τ) ally arbitrary signals. In this study, we will use a nonsta- (cid:68) (cid:69) tionarity factor introduced within the framework of flicker- Φ(2)(τ)= [V(t)−V(t+τ)]2 . (4) noise spectroscopy (FNS), a statistical physics approach to the analysis of time series (Timashev and Polyakov, 2007; Toanalyzetheeffectsofnonstationarityinrealprocesses, Timashev, 2007; Timashev et al., 2010b). This dimen- westudythedynamicsofchangesinΦ(2)(τ)forconsecutive sionless criterion is practically independent from the indi- ”window” intervals [t ,t +T], where k = 0, 1, 2, 3, and k k vidual features of source signals and is designed to detect t =k∆T,thatareshiftedwithinthetotaltimeintervalT k tot G.V.Ryabininetal.: EarthquakeprecursorsinthedatafortheKamchatkapeninsula 3 of experimental time series (t +T <T ). The time inter- coastofKamchatka,whichisboundedbyadeep-seatrench k tot valsT and∆T arechosenbasedonthephysicalunderstand- ontheeast(Fedotovetal.,1985). ing of the problem in view of the suggested characteristic Specialized measurements of underground water charac- time of the process, which is the most important parameter teristics were started in 1977 to find and study hydrogeo- ofsystemevolution. Thephenomenonof”precursor”occur- chemical precursors of Kamchatka earthquakes. Currently, renceisassumedtoberelatedtoabruptchangesinfunctions theobservationnetworkincludesfourstationsinthevicinity Φ(2)(τ)whentheupperboundoftheinterval[t ,t +T]ap- of Petropavlosk-Kamchatsky (Fig. 1). The Pinachevo sta- k k proachesthetimemomentt ofacatastrophiceventaccom- tion includes five water reservoirs: four warm springs and c paniedbytotalsystemreconfigurationonallspacescales. one borehole GK-1 with the depth of 1,261 m. The Mo- The analysis of experimental stochastic series often re- roznaya station has a single borehole No. 1 with the depth quirestheoriginaldatatobeseparatedintoasmoothedand of 600 m. The Khlebozavod station also includes a single fluctuation components. In this study, we apply the ”re- boreholeG-1withthedepthof2,540m,whichislocatedin laxation” procedure proposed by Timashev and Vstovskii Petropavlosk-Kamchatsky. The Verkhnyaya Paratunka sta- (2003) based on the analogy with a finite-difference solu- tion comprises four boreholes (GK-5, GK-44, GK-15, and tion of the diffusion equation, which allows one to split the GK-17)withdepthsintherangefrom650to1208m. originalsignalintolow-frequencyV (t)andhigh-frequency The system of hydrogeochemical observations includes R V (t)components. Theiterativeprocedurefindingthenew the measurement of atmospheric pressure and air temper- F values of the signal at every relaxation step using its val- ature, measurement of water discharge and temperature of ues for the previous step allows one to determine the low- boreholes and springs, collection of water and gas samples frequency component V (t). The high-frequency compo- fortheirfurtheranalysisinlaboratoryenvironment. Forwa- R nentV (t)isobtainedbysubtractingV (t)fromtheoriginal ter samples, the following parameters are determined: pH; F R signal. This smoothing algorithm progressively reduces the ion concentrations of chlorine (Cl−), bicarbonate (HCO−), 3 localgradientsofthe”concentration”variables, causingthe sulfate (SO2−), sodium (Na+), potassium (K+), calcium 4 points in every triplet to come closer to each other. Such (Ca2+), and magnesium (Mg2+); concentrations of boric splitting of the original signal V(t) into V (t) and V (t) (H BO ) and silicone (H SiO ) acids. For the samples of R F 3 3 4 4 makes it possible to evaluate the nonstationarity factor for gases dissolved in water, the following concentrations are each of the three functions V (t) (J = R, F, or G), where determined: methane (CH ), nitrogen (N ), oxygen (O ), J 4 2 2 indexGcorrespondstotheoriginalsignal. carbon dioxide (CO ), helium (He), hydrogen (H ), hy- 2 2 TheFNSnonstationarityfactorC (t )isdefinedas drocarbon gases: ethane (C H ), ethylene (C H ), propane J k 2 6 2 4 (C H ), propylene(C H ), butane(C H n), andisobutane QJ−PJ T 3 8 3 6 4 10 CJ(tk)=2×QJk+PkJ ×∆T, (5) (C4H10i). The data are recorded at nonuniform sampling k k intervals with one dominant sampling frequency. For the Pinachevo, Moroznaya, and Khlebozavod stations, this av- 1 (cid:90)αTt(cid:90)k+T erage sampling frequency is one measurement per 3 days; QJ= [V (t)−V (t+τ)]2dtdτ, (6) k αT2 J J fortheVerkhnyayaParatunkastation, onemeasurementper 0 tk 6 days. Multiple studies of the hydrogeochemical data and correspondingseismicactivityfortheKamchatkapeninsula 1 (cid:90)αTtk+(cid:90)T−∆T reportedanomalouschangesinthechemicaland/orgascom- PJ= [V (t)−V (t+τ)]2dtdτ. (7) k αT2 J J position of underground waters prior to several large earth- 0 tk quakesinthetimeframefrom1987to2001(Kopylovaetal., 1994;Bellaetal.,1998;Biagietal.,2000,2006;Khatkevich Here, J indicates which function V (t) (J = R, F or G) is J and Ryabinin, 2006). In this study, we analyze the varia- used. Expressions(6-7)aregivenindiscreteformelsewhere tionsofchlorine-ionconcentrationdeterminedbyatitrimet- (Timashevetal.,2010a). NotethatfunctionsΦ(2)(τ)canbe J ricmethod(relativeerrorfrom1to10%). reliablyevaluatedonlyontheτ intervalof[0,αT],whichis Geoacousticemissionsinthefrequencyrangefrom25to lessthanhalfoftheaveragingintervalT;i.e.,α<0.5. 1,400Hz(0.7level)havealsobeenrecordedinthedeepG- 1boreholeoftheKhlebozavodstationunderthesupervision 3 ExperimentaldatafortheKamchatkapeninsula of V. A. Gavrilov since August, 2000. The data analyzed in this paper were obtained by a geophone with crystal fer- Thedatawererecordedinthesouth-easternpartoftheKam- romagnetic sensors (Belyakov, 2000). The output signal of chatkapeninsulalocatedattheRussianFarEast.Theeastern suchasensorisproportionaltothethirdderivativeofground partofthepeninsulaisoneofthemostseismicallyactivere- displacement,andthegainslopeis60dBperdecadeoffre- gionsintheworld. Theareaofhighestseismicitylocalized quency change. The geophone was set up at the depth of inthedepthrangebetween0and40kmrepresentsanarrow 1,035m,whichisenoughtoreduceanthropogenicnoiselev- stripewiththelengthofapproximately200kmalongtheeast els by more than two orders of magnitude (Gavrilov et al., 4 G.V.Ryabininetal.: EarthquakeprecursorsinthedatafortheKamchatkapeninsula 2008). The geophone body was fixed inside the borehole Figures2and3showthevariationsofC forCl-GK1and J casing by a spring. The vertical channel sensitivity of the Gl-CK44togetherwithlargestseismicevents. Itcanbeseen geophone is 0.15 V × s3/m. The sensitivity of horizontal thatspikesinC precedeseverallargeearthquakes.Itshould J channelsis0.60V×s3/m.Thesensoroutputsignalsaresep- be noted that the low-frequency component C shows the R arated by third-octave band pass filters into four frequency most number of precursors for Cl-GK1 and high-frequency bands with central frequencies 30, 160, 560, and 1,200 Hz, component is most informative for Cl-CK44. The first fact whichisfollowedbyreal-timehardware/softwaresignalpro- is in agreement with the study of Khatkevich and Ryabinin cessing. The value of postprocessed output signal for each (2006). The second fact implies that the use of the high- channel is proportional to the average value of input signal frequencycomponenteliminatedseasonalchangesfromthe forone-minuteintervals. Moredetaileddescriptionofgeoa- analysisandmadeCl-GK44aprecursorysignal. Therefore, cousticemissionobservationsandexperimentalsetupforthe the FNS nonstationarity factor together with the procedure G-1boreholeispresentedelsewhere(Gavrilovetal.,2008). for separating out high-frequency and low-frequency signal componentscanbeusedtoanalyzedifferentsignalsdespite majordifferencesintheirspecificfeatures. Figure4showsacombinedanalysisofhydrogeochemical 4 Results andgeoacousticvariationsinthetimeframearoundtheOc- tober8,2001earthquake(M =6.3,H =24km,D=134km l To illustrate the nonstationarity factor and proposed phe- from Petropavlovsk-Kamchatsky), which was the strongest nomenological method, we analyze the hydrogeochemical earthquake (based on local magnitude and distance to the dataforchlorine-ionconcentrationsatGK-1(Pinachevosta- epicenter)recordedforthewholetimeintervalofgeoacous- tion) and GK-44 (Verkhnyaya Paratunka station) and geoa- tic observations in the G-1 borehole. Nonstationarity fac- coustic emissions at the output of geophone vertical fre- torsC forCl-GK1andC forCl-GK44showspikeswith R F quencychannelwiththecentralfrequencyof160Hz(Z160) highest values (precursors) in the time frame from 50 to 70 in G-1 (Khlebozavod station). Chlorine-ion concentration daysbeforetheearthquake. C forG-1(thesignalisahigh- G time series for GK-1 (Cl-GK1) was selected because it is frequencyonebyitsnature)showsprecursors29and6days characterized by a unidirectional long-period trend without before the event, which is in agreement with the results re- seasonalvariations(Fig. 2)andwasalreadytreatedasapre- portedbyGavrilovetal.(2008). Inotherwords,anomalous cursory signal due to a gradual chlorine-ion concentration changesinthegeoacousticsignalhappenclosertotheearth- declinedowntoalocalminimum30to60daysbeforesev- quakethaninthehydrogeochemicalones,whichimpliesthat eral earthquakes (Khatkevich and Ryabinin, 2006). On the precursory signals of different nature may take place at dif- otherhand,chlorine-ionconcentrationatGK-44(Cl-GK44) ferenttimescalesbeforealargeearthquake. is not considered as a precursory signal because it is domi- natedbyseasonalconcentrationchangesonthebackground ofaslowlyvaryinglocalmean,theminimumvalueofwhich 5 Conclusions is reached shortly after the strong earthquake on December 5, 1997 (Ml =7.0). The Z160 signal was selected from the The above example shows that precursory signals of differ- whole set of geoacoustic time series because it contains the enttypesmaybeobservedinthesamelocalseismicallyac- lowestlevelofnoise(highestsignal-to-noiseratio). tivezoneatdifferenttimespriortoalargeearthquake,which Tokeepthestatisticalstructureofsourcetimeseriesprac- maybeattributedtosomesystempreparationprecedingthe tically intact, the signals were subjected only to minimal seismic event. In the studied case, the qualitative changes preprocessing, which included the removal of single-point mayberelatedtoasystem-widestructuralmediumreconfig- spikes,reductionofthehydrogeochemicaltimeseriestouni- urationatthepreparatoryphaseoftheearthquake. form sampling intervals using linear interpolations, and ex- This study also shows that the FNS nonstationarity fac- th tractionofevery30 pointinthegeoacoustictimeseriesto tor can be used as the standard criterion to detect qualita- formanewtimeserieswiththefrequencyof30min−1.Then tive changes within relatively short time intervals in virtu- thetimeseriesV (t)wereseparatedoutintolow-frequency allyarbitrarysignals,evenifthesignalscontainstronglypro- G V (t) and high-frequency V (t) components, which were nouncedperiodiccomponents,aswasthecaseforCl-CK44. R F used to calculate the nonstationarity factors. In evaluating It should be noted that the nonstationarity factor should be C (J =R, F orG)forthehydrogeochemicalseries, aver- analyzed not only for the original signal, but also for its J agingtimeintervalsT intherangefrom50to900dayswere smoothed (low-frequency) and fluctuation (high-frequency) used. Forthegeoacoustictimeseries,theintervalT wasvar- components. ied from 3 to 20 days. Our analysis showed that the values In order to validate the proposed phenomenological sys- ofT equalto600and20daysaremostadequateforlocating temsapproach,comprehensivemonitoringofseismicallyac- precursors in the hydrogeochemical and geoacoustic series, tiveregionssuchastheKamchatkapeninsulashouldbeper- respectively. formedandthedatashouldbeanalyzedwiththeFNSnonsta- G.V.Ryabininetal.: EarthquakeprecursorsinthedatafortheKamchatkapeninsula 5 tionarityfactor. Themeasuredcharacteristicsshouldinclude ICEFCP:OperationalEarthquakeForecasting:StateofKnowledge geoacoustic, hydrological/hydrochemical, gasgeochemical, andGuidelinesforUtilization,http://www.iaspei.org/downloads/ geodetic,andelectromagneticsignalssummarizedinthein- Ex Sum v5 THJ9 A4format.pdf,access:1October2010,2009. troduction. Ida, Y., Hayakawa, M., and Timashev, S.: Application of differ- entsignalanalysismethodstotheULFdataforthe1993Guam Acknowledgements. ThisstudywassupportedinpartbytheRus- earthquake,Nat.HazardsEarthSyst.Sci.,7,479–484,2007. sianFoundationforBasicResearch,projectnos. 09-05-98543and Ihmle,P.andJordan,T.: Teleseismicsearchforslowprecursorsto 10-02-01346. largeearthquakes,Science,266,1547–1551,1994. Khatkevich, Y. and Ryabinin, G.: Geochemical an ground-water studies in Kamchatka in the search for earthquakes precursors References (inRussian),VulkanologiyaiSeysmologiya,(4),34–42,2006. Kirschvink,J.L.:Earthquakepredictionbyanimals:Evolutionand Bella,F.,Biagi,P.F.,Caputo,M.,Cozzi,E.,Monica,G.D.,Ermini, sensoryperception,B.Seismol.Soc.Am.,90,312–323,2000. A., Gordeez, E. I., Khatkevich, Y. M., Martinelli, G., Plastino, Kopylova, G., Sugrobov, V., and Khatkevich, Y.: Variations in W., Scandone, R., Sgrigna, V., and Zilpimiani, D.: Hydrogeo- the regime of springs and hydrogeological boreholes in the chemicalanomaliesinKamchatka(Russia),Phys.Chem.Earth, Petropavlovsk polygon (Kamchatka) related to earthquakes (in 23,921–925,1998. Russian),VulkanologiyaiSeysmologiya,(2),53–70,1994. Belyakov,A.S.: Magnetoelasticacousticgeophonesforgeophys- Li, Y., Liu, Y., Jiang, Z., Guan, J., Yi, G., Cheng, S., Yang, B., icalresearchandearthquakeprediction(inRussian), Seismich- Fu, T., and Wang, Z.: Behavioral change related to Wenchuan eskiePribory,33,27–45,2000. devastating earthquake in mice, Bioelectromagnetics, 30, 613– Biagi,P.F.,Ermini,A.,Cozzi,E.,Khatkevich,Y.M.,andGordeev, 620,2009. E. I.: Hydrogeochemical precursors in Kamchatka (Russia) re- Reasenberg,P.A.: Foreshockoccurrencebeforelargeearthquakes, latedtothestrongestearthquakesin1988–1997, Nat.Hazards, J.Geophys.Res.,104,4755–4768,1999. 21,263–276,2000. Telesca,L.andLovallo,M.: Non-uniformscalingfeaturesincen- Biagi,P.F.,Castellana,L.,Minafra,A.,Maggipinto,G.,Maggip- tralItalyseismicity:Anon-linearapproachininvestigatingseis- into, T., Ermini, A., Molchanov, O., Khatkevich, Y. M., and mic patterns and detection of possible earthquake precursors, Gordeev,E.I.:Groundwaterchemicalanomaliesconnectedwith Geophys.Res.Lett.,36,L01308,2009. the Kamchatka earthquake (M=7.1) on March 1992, Nat. Haz- Telesca, L., Lapenna, V., Timashev, S., Vstovsky, G., and Mar- ardsEarthSyst.Sci.,6,853–859,2006. tinelli, G.: Flicker-noise spectroscopy: a new approach to in- Cicerone, R.D., Ebel, J.E., andBritton, J.: Asystematiccompi- vestigatethetimedynamicsofgeoelectricalsignalsmeasuredin lationofearthquakeprecursors,Tectonophysics,476,371–396, seismicareas,Phys.Chem.Earth.,29,389–395,2004. 2009. Telesca,L.,Lapenna,V.,Macchiato,M.,andHattori,K.:Investigat- Descherevsky,A.V.,Lukk,A.A.,Sidorin,A.Y.,Vstovsky,G.V., ingnon-uniformscalingbehaviorinUltraLowFrequency(ULF) and Timashev, S. F.: Flicker-noise spectroscopy in earthquake earthquake-relatedgeomagneticsignals,EarthPlanet.Sci.Lett., predictionresearch, Nat.HazardsEarth Syst.Sci., 3, 159–164, 268,219–224,2008. 2003. Telesca, L., Lovallo, M., Ramirez-Rojas, A., and Angulo-Brown, Du, J., Amita, K., Ohsawa, S., Zhang, Y., Kang, C., and Ya- F.:Anonlinearstrategytorevealseismicprecursorysignaturesin mada, M.: Experimental evidence on formation of imminent earthquake-relatedself-potentialsignals,PhysicaA,388,2036– andshort-termhydrochemicalprecursorsforearthquakes,Appl. 2040,2009a. Geochem.,25,586–592,2010. Telesca, L., Lovallo, M., Ramirez-Rojas, A., and Angulo-Brown, Fedotov, S.A., Gusev, A.A., Shumilina, L.S., andChernyshova, F.:Scalinginstabilityinself-potentialearthquake-relatedsignals, V.G.: TheseismofocalzoneofKamchatka(inRussian),Volka- PhysicaA,388,1181–1186,2009b. nologiyaiSeismologiya,pp.91–107,1985. Telesca, L., Lovallo, M., and Carniel, R.: Time-dependent Fisher Gavrilov,V.,Bogomolov,L.,Morozova,Y.,andA.,S.: Variations InformationMeasureofvolcanictremorbeforethe5April2003 ingeoacousticemissionsinadeepboreholeanditscorrelation paroxysm at Stromboli volcano, Italy, J. Volcanol. and Geoth. withseismicity,Ann.Geophys.-Italy,51,737–753,2008. Res.,195,78–82,2010. Geller,R.J.: Earthquakeprediction: acriticalreview,Geophys.J. Timashev,S.,Polyakov,Y.,Lakeev,S.,Misurkin,P.,andDanilov, Int.,131,425–450,1997. A.:Fundamentalsoffluctuationmetrology,Russ.J.Phys.Chem. Geller,R.J.,Jackson,D.D.,Kagan,Y.Y.,andMulargia,F.:Earth- A,84,1807–1825,2010a. quakescannotbepredicted,Science,275,1616–1617,1997. Timashev,S.F.: Flickernoisespectroscopyanditsapplication: In- Gordienko,V.,Gordienko,T.,Krasnopistsev,N.,Kuptsov,A.,Lari- formation hidden in chaotic signals, Russ. J. Electrochem., 42, onov,I.,Marapulets,Y.,Rutenko,A.,andShevtsov,B.:Anomaly 424,2006. in high-frequency geoacoustic emission as a close earthquake Timashev, S. F.: Fliker-Shumovaya Spektroskopiya: Informatsiya precursor,Acoust.Phys.,54,82–93,2008. vkhaoticheskikhsignalakh(Flicker-NoiseSpectroscopy: Infor- Hartmann,J.andLevy,J.K.:Hydrogeologicalandgasgeochemical mationinChaoticSignals),Fizmatlit,Moscow,2007. earthquakeprecursors-Areviewforapplication,Nat.Hazards, Timashev,S.F.andPolyakov,Y.S.: Reviewofflickernoisespec- 34,279–304,2005. troscopyinelectrochemistry,Fluct.NoiseLett.,7,R15,2007. Hayakawa,M.andTimashev,S.F.: Anattempttofindprecursors Timashev, S. F. and Vstovskii, G. V.: Flicker-noise spectroscopy in the ULF geomagnetic data by means of flicker noise spec- foranalyzingchaotictimeseriesofdynamicvariables: Problem troscopy,NonlinearProc.Geophys.,13,255–263,2006. 6 G.V.Ryabininetal.: EarthquakeprecursorsinthedatafortheKamchatkapeninsula ofsignal-to-noiserelation, Russ.J.Electrochem., 39, 141–153, 2003. Timashev,S.F.,Polyakov,Y.S.,Misurkin,P.I.,andLakeev,S.G.: Anomalousdiffusionasastochasticcomponentinthedynamics ofcomplexprocesses,Phys.Rev.E,81,041128,2010b. Uyeda,S.,Nagao,T.,andKamogawa,M.: Short-termearthquake prediction: Currentstatusofseismo-electromagnetics,Tectono- physics,470,205–213,2009. Vstovsky,G.V.,Descherevsky,A.V.,Lukk,A.A.,Sidorin,A.Y., andTimashev,S.F.:Searchforelectricearthquakeprecursorsby themethodofFlicker-noisespectroscopy,Izv.-Phys.SoildEarth, 41,513–524,2005. Wyss,M.andBooth,D.C.: TheIASPEIprocedurefortheevalu- ationofearthquakeprecursors,Geophys.J.Int.,131,423–424, 1997. Wyss,M.;Aceves,R.L.,Park,S.K.;Geller,R.J.,Jackson,D.D., Kagan, Y. Y., and Mulargia, F.: Cannot earthquakes be pre- dicted?,Science,278,487–490,1997. Yokoi, S., Ikeya, M., Yagi, T., and Nagai, K.: Mouse circadian rhythmbeforetheKobeearthquakein1995,Bioelectromagnet- ics,24,289–291,2003. Fig.1.Schematicofthemeasurementarea(smallrectangularframe ontheleft)andepicentersoflargestearthquakes(M ≥6,H≤50 l km, D≤ 350 km) from 1985 to 2009, where M – local earth- l quakemagnitude,H –depth,D–distancefromtheepicenter.The largeframeontherightshowsazoomed-inviewofthepositionsof hydrogeologicalstations:1–Pinachevo,2–Moroznaya,3–Khle- bozavod, 4 – Verknyaya Paratunka. The solid circles denote the earthquakes reportedly preceded by hydrogeochemical anomalies. Thedashedlineistheaxisofthedeep-seatrench. Theearthquakes wereselectedusingthecatalogofGeophysicalSurvey,Kamchatka BranchoftheRussianAcademyofSciences. G.V.Ryabininetal.: EarthquakeprecursorsinthedatafortheKamchatkapeninsula 7 Fig.2.ComparisonofnonstationarityfactorC (T=600days,∆T=3days)fortheGK-1chlorine-ionconcentrationtimeserieswithseismic J activity: V –sourcesignal;C –nonstationarityfactorforV ,C –nonstationarityfactorforthelow-frequencycomponentofV ,C G G G R G F –nonstationarityfactorforthehigh-frequencycomponentofV ,M –localearthquakemagnitude,D–distancefromtheepicenter. Solid G l trianglesdenotesampleC spikesprecedinglargeearthquakes.CrossesdenotesampleC spikesnotrelatedtolargeseismicevents. R R Fig. 3. Comparison of nonstationarity factor C (T=600 days, ∆T=3 days) for the GK-44 chlorine-ion concentration time series with J seismicactivity: NomenclatureasinFig. 2. SolidtrianglesdenotesampleC spikesprecedinglargeearthquakes. Crossesdenotesample F C spikesnotrelatedtolargeseismicevents. F 8 G.V.Ryabininetal.: EarthquakeprecursorsinthedatafortheKamchatkapeninsula Fig. 4. Nonstationarity factors for GK-1 and GK-44 chlorine-ion concentrations and Z160 G-1 geoacoustic emissions in the time frame aroundthe8/10/2001earthquake.M –localearthquakemagnitude,D–distancefromtheepicenter.Thedouble-headedarrowsdenotethe l timeintervalsbetweenthenonstationarityfactorspikesandearthquakeitself.