FEASIBILITY  STUDY  ON  EARTHQUAKE  EARLY  WARNING  FOR  SCHOOLS  IN  SOUTHERN  ITALY  Antonio  EMOLO1,  Ma.eo  PICOZZI1,  Aldo  ZOLLO1,  Gerardo  VERDERAME1,  Claudio  MARTINO1,  Luca  ELIA1,  Simona  COLOMBELLI1,  Ortensia  AMOROSO1,2,  Maria  Teresa  DE  RISI1,  Dino  BINDI3,  

Stefano  PAROLAI3,  Tobias  BOXBERG3,  Nicola  MIRANDA4,  Lilly  BUONAIUTO4,  Angelo  AMELIA4  

(1)  AMRA-­‐University  of  Naples,  Italy  ,  (2)  IsQtuto  Nazionale  di  Geofisica  e  Vulcanologia,  Osservatorio  Vesuviano,  Italy  -­‐  University  of  Naples,  Italy,    (3)  Helmholtz-­‐Zentrum  Potsdam  Deutsches  GeoForschungsZentrum,  Germany,  (4)  ITIS  ‘E.  Majorana’,  Somma  Vesuviana,  Italy    

Monitoring  system  

PRESToPlus  System  with  Early  Warning  SenFnel  

EvaluaQng  the  effecQveness  of  Earthquake  Early  Warning  Systems  (EEWS)  and  real-­‐Qme  risk  assessment  procedures   in   reducing   seismic   risk   for   various   industrial   partners   and   end-­‐users   is   one   of   the   main  objecQve   of   the   WP7   (Strategic   ApplicaQons   and   Capacity   Building)   in   the   framework   of   the   REAKT-­‐Strategies  and  tools  for  Real  Time  Earthquake  RisK  ReducTion  FP7  European  project.  In  this  context  our  group   is   engaged   in  a   feasibility   study  on   the  applicaFon  of  earthquake  early-­‐warning  procedures   in  two   high   schools   (Liceo   ‘F.   de   SancQs’,   S.   Angelo   dei   Lombardi,   SALI   and   ITIS   ‘E.   Majorana’,   Somma  Vesuviana,  MAJI)   located   in   the   Irpinia   region   (Southern   Italy).   The   area   has   been   interested   by   large  earthquakes   both   in   historic   and   recent   Qme.   The   last   destrucQve   earthquake   was   the   23   November  1980,   M   6.9,   Irpinia   earthquake,   which   caused   more   than   3000   casualQes   and   produced   huge   and  widespread  damage.  Here  we  present  a  summary  of  the  acQviQes  carried  out  and  the  results  obtained  so  far  at  the  MAJI  School.  

A   high   quality   accelerometric   staQon  consisQng   of   a   24-­‐bit   ADC   (Sigma/Delta)  Agecodagis   Kefren   data-­‐logger   and   a   Guralp  CMG-­‐5TC   accelerometer   with   a   0.25g   full-­‐scale  has  been  installed.  In  order  to  perform  a  conQnuous   seismic   monitoring   of   the   site,   a  total  of  five  accelerometric  staQons  have  been  installed   in   different   parts   of   the   school  building.   In   parQcular,   besides   the   Guralp  CMG-­‐5TC   accelerometer   in   the   school  courtyard,   four   SOSEWIN   sensors   have   been  deployed   at   different   locaQons   within   the  building.   Commercial   ADSL   lines   provide  transmission   of   real-­‐Qme   data   to   the   EEW  centre.  Data  streams  are  now  acquired  in  real-­‐Qme   in   the   PRESToPlus   (regional   and   on-­‐site,  threshold-­‐based   early-­‐warning)   sojware  plakorm  (Satriano  et  al,  2010)*.  

IntroducFon  

PRESTo  (PRobabilisQc  and  EvoluQonary  early  warning  SysTem)  is  a  sojware  plakorm  for  Earthquake  Early  Warning,   that   integrates  a  regional,  network-­‐based  approach  with   an  on-­‐site,   single   staFon   approach.   The   sojware   processes   the   real-­‐Qme   3-­‐Component   acceleraQon   data   streams   from   seismic   staQons   of   the   Irpinia   Seismic  Network  (ISNet)  in  Southern  Italy  and  from  schools  involved  in  the  NERA  and  REAKT  European  Projects.  Once  a  P-­‐waves  arrival  has  been  detected  at  a  staQon  and  three  seconds  of  signal  are  available,   PRESTo  measures   the   peak   displacement   and   predominant   period   of   P-­‐waves.   These   values   are   compared   to   thresholds,   set   for   a  minimum  magnitude   6  and  instrumental   intensity  VII,  based  on  the  empirical  regression  analyses  of  strong  moQon  data.   If   thresholds  are  exceeded  and  damage   is  expected,  a  first   rough  but  robust   alert   level   to   the   site   /   school   is   issued.     The  data   streams  and   alert   levels  from  all  the  staQons  are  also  gathered  and  processed  at  regional  scale,  contribuQng  at   improving   the   earthquake   locaQon   and   magnitude   esQmaQons,   peak   ground  moQon  predicQons  at  target  sites  and  regional  PotenQal  Damage  Zone.    Whenever  an  alert  message   is  released,   independently   from  the  EEWS  source  (i.e.  On-­‐site,    Regional,  or  even  from  both  of  them  cooperaQng),  it  is  received  and  taken  in  charge  by  a  EEW  SenFnel   (i.e.  Arduino®).  Arduino   is  a   low  cost  programmable  actuator  that  we  set-­‐up  to  keep  listening  and  interpreFng  the  messages  and  alert  levels  from  PRESToPlus.  Hence,  in  the  case  of  an  event,  this  EEW  senFnel  is  able  to  provide   to   the   school   users   a   real-­‐Fme   P-­‐wave   based   predicFon   of   the   ground  moFon  severity  expected  at  the  target  site.        

T0  Origin  Time  

2013-­‐12-­‐29  17:08:45 17:08:56

DetecQon  Time  

Lead  Time:  7s  M:4,3  

 Mmin:3,9  Mmax:4,6  

PGV:  0,09  cm/s  

 2°  Alert  Sent  to  Target  

 3°  Alert  Sent  to  Target  

1°    Alert    Sent  to  Target  

17:08:57

Lead  Time:  6s  M:4,5  

 Mmin:4,3  Mmax:4,8  

PGV:  0,13  cm/s  

17:08:59

Lead  Time:  4s  M:4,7  

 Mmin:4,5  Mmax:4,9  

PGV:  0,19  cm/s  

1°  Alert  

2°  Alert  

3°  Alert4  

 Weak–  II°-­‐III°  

 Weak–  II°-­‐III°  

 Weak–  II°-­‐III°  

 Green  Light  

 No  Siren  

 

I.T.I.S.  Majorana  School  Somma  Vesuviana  (Naples)  

Off-­‐line  analyses  at  the  school  ITIS  ‘E.  Majorana’  

The  first  tests  were  performed  on  24-­‐hour  long  waveforms  recorded  by   the   five   staQons,   during   four   different   days,  with   school   acQvity  but  no  nearby  earthquakes.  The  aim  of  these  analyses  was  to  verify  the  rate  of  false  events  detected  by  the  EEW  On-­‐site  system  due  to  the  anthropic  acQviQes.  The  P-­‐wave  picking  was  performed  using  the  same  picker  used  in  PRESTo  (FilterPicker5,  Lomax  et  al,  2012)*.  Due  to   the   high   noise   level   incorrect   and   false   phase   pickings   can   be  automaQcally   declared,   up   to   1   pick   every   5  minutes   for   the  most  sensiQve   staQon.   This   suggests   that   the   single   staQon   EEW  approaches   require   a   specific   tuning   of   parameters,   to   avoid   the  repeated   occurrence   of   false   alarms.   Conversely,   when   an   almost  simultaneous   pick   at   a   minimum   number   of   4   staQons   is   required  before   declaring   an   event,   the   rate   of   false   alerts   is   significantly  reduced.  To  account  for  Qme  differences  in  the  picks  due  to  the  low  signal-­‐to-­‐noise   raQo,   we   chose   a   0.5   sec   Qme   window   for  coincidence.  We  observed  that  when  3-­‐staQons  are  considered  only  1-­‐2  events  per  day  are  detected,  while  there  was  never  a  coincidence  of   picks   within   0.5   seconds,   and   thus   no   early   warning   procedure  would  have  been  triggered,  when  4-­‐  or  5-­‐staQons  are  considered.  

The  second  test  was  carried  out  considering  the  only  relevant  earthquake  recorded  so  far  by  the  system,  that  is  the  Matese  earthquake  (2013/12/29  17:08:43,  ML  4.9),  for  which  only  4  staQons  were  available.  The  distance  from  the  school  to  the  hypocenter  was  about  55  km.  On-­‐site  Alert  (Leh):  we  applied  a  low-­‐pass  filter  (10  Hz)  to  the  waveforms  to  reduce  the  high  frequency  noise  contaminaQon  and  we  verified  that  the  seismic  staQons  correctly  picked  the  P-­‐wave  onset  of  the  event.  For  the  less  sensiQve  SOSEWIN  staQons,  that  use  an  internal  MEMS  sensor,  it  was  also  necessary  to  decrease  the  threshold  parameter  of  the  picker.  All  the  staQons  picked  within  0.5  seconds  from  the  arrival  of  the  P-­‐wave.  The  accelerometric   records  have  been  double   integrated  to  get  displacement  waveforms  and  a  high-­‐pass  filter  with  cut-­‐off  frequency  of  0.075  Hz  has  been  applied  to  reduce  the  noise  introduced  by  the  double  integraQon  operaQon,  as   it   is  usually  done  in  Early  Warning  applicaQons.  Then,  the  iniQal  peak  displacement  and  dominant  period  have  been  measured  at  each   staQon,  on  3   seconds  of   signal  ajer   the  P-­‐wave  detecQon,   corresponding   to   the  correct  alert   level  of  0   (No  Local  Damage,  M  <  6).   Integrated  Regional  /  On-­‐site  Alert  (Right):  The  playback  in  PRESToPlus  of  the  waveforms  of  MAJI  and  of  all  the  ISNet  staQons  was  performed  fixing  the  epicenter  since  the  earthquakes  is  external  to  the  network.  PRESToPlus    predicts  a  magnitude  within  0.5  points  to  the  real  value,  and  the  expected  on-­‐site  alert  level  of  0  at  all  staQons,  including  MAJI.  The  SenQnel  then  outputs  non-­‐criQcal  visual  and  audio  signals  based  on  the  correctly  predicted  Intensity  (II-­‐III)  at  MAJI.  

Concerning  the  EEW  On-­‐Site  approach,  the  basic  idea  is  to  use  a  stand-­‐alone  operaQng  system  for  the  rapid  esQmate  of  the  local  damage  at  the  school.  In  this  system,  the  early  porQon  of  the  recorded  P-­‐wave  signal  is  used  to  predict  the  Peak  Ground  Velocity  (PGV)  and  Instrumental  Intensity  (IMM)  expected  at  the  site  We  performed  off-­‐line  analyses  to  test  the  feasibility  of  such  an  on-­‐site  early  warning  procedure  at  the  school  ITIS  ‘E.  Majorana’.  

 *Satriano  C,  Elia  L,  MarQno  C,  Lancieri  M,  Zollo  A,  Iannaccone  G  (2010)  "  PRESTo,  the  earthquake  early  warning  system  for  Southern  Italy:  concepts,  capabiliQes  and  future  perspecQves"  ,  Soil  Dyn  Earthq  Eng  ,doi:  10.1016/j.soildyn.2010.06.008  

*Lomax  A,  Satriano  C,  Vassallo  M  (2012)  “AutomaQc  picker  developments  and  opQmizaQon:  FilterPicker  –  a  robust,  broadband  picker  for  real-­‐Qme  seismic  monitoring  and  earthquake  early-­‐warning”  Seism.  Res.  Le..    ,  83,  531-­‐540,  doi:  10.1785/gssrl.83.3.531  

ON-­‐SITE  ALERT  LEVEL  ZERO:  •   No  Local  Damage  •   M  <  6  

The  ML  4.9  Matese  earthquake@MAJI  Noise  analysis  on  24-­‐hour  waveforms  

REGIONAL  ALERT:  •   No  Local  Damage  •   4.5  ≤  M  ≤  4.9