Tectonic Shifts, Tweets, & Tactical Data Analysis

The U.S. Geological Survey (USGS) operates a real-time system that detects earthquakes using only tweets. The main benefit of the tweet-based detections is speed, with most detections occurring between 20 and 120 seconds after the origin time. This timeframe is significantly faster than seismically-derived detections in poorly instrumented regions of the world. This talk will present a technical overview of the Elasticsearch and Kibana-based system and describe the potential for rapid characterization of earthquake damage via the added real-time search and visualization tools combined with analysis of the 32 million ‘earthquake’ tweets the system has so far amassed.

Presented by Michelle Guy, US Geological Survey

Transcript

1. Tectonic Shifts, Tweets, & Tactical Analysis Presenter: Michelle Guy – Computer Scientist Co-author: Paul Earle - Geophysicist
2. { } CC-BY-ND 4.0 •  What we do Global earthquake monitoring and analysis •  Our tweet based earthquake detection system System performance and earthquake characterization •  Taking it to the next level Elasticsearch & Kibana for data analysis •  Wrap it up More fun to come Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. { 2 } Our Story
3. { } CC-BY-ND 4.0 { 3 } National Earthquake Information Center
4. { } CC-BY-ND 4.0 Instrument Based Earthquake Monitoring •  Energy Propagation •  Seismic Station Distribution •  Human Review •  Accuracy { 4 }
5. { } CC-BY-ND 4.0 http://www.gly.fsu.edu/~salters/GLY1000/Chapter4/Slide20.jpg { 5 } •  It  takes  (me  for  seismic   energy  to  travel   through  the  Earth   •  There  is  a  (me  delay   between  kinds  of   seismic  waves   Energy Propagation
6. { } CC-BY-ND 4.0 Frank Wattenberg, Department of Mathematics, Montana State University { 6 }
7. { } CC-BY-ND 4.0 { 7 } An Earthquake Information Timeline Global earthquakes 0sec 1min 5-10min 10-20min 1hr What can be done here?
8. { } CC-BY-ND 4.0 M5.8 Offshore Java, Indonesia { 8 } Tweets  /  minute  
9. { } CC-BY-ND 4.0 Tweet Based Earthquake Detection System 9   outputs detect, locate SQL  DB   geocode collect archive time-series Text messages e-mails Interactive maps Tweets/min select preferred tweets earthquake   gempa   temblor   terremoto   deprem   lindol   erdbeben   zelzele   Σεισμός    هزرل‌نیمز      
10. { } CC-BY-ND 4.0 Data Filtering Remove Tweets Containing: http, numbers, more than 7 words, @, predict, drill, song 10   News  Story   Earthquake  
11. { } CC-BY-ND 4.0 Generating Time-series C(t) = STA/(mLTA+b) 11  
12. { } CC-BY-ND 4.0 System Performance { 12 Statistics derived from dataset spanning July 2013 – March 2014 Detec%on   Frequency   ~  3  tweet  events  /  day   Detec%on   Delay   90%  <  2  min   Loca%on   Accuracy   90%  within  200km   Magnitude   Distribu%on   M1.4  through  M8.2  
13. { } CC-BY-ND 4.0 System Sensitivity & Accuracy •  Detector Tunable •  9.5% False Triggers •  Currently, low cost for false detections 13  
14. { } CC-BY-ND 4.0 Global Coverage 14   Where  are  they?   •  Seismically active areas •  Places where people tweet
15. { } CC-BY-ND 4.0 15   Example Alert M5.1 76 km deep ~150 km N of Anchorage, AK Tweet  event  detec%on   NOT  AN  OFFICIAL  USGS  ALERT   NOT  SEISMICALLY  VERIFIED     Detec%on  Time:   2014/04/16  20:25:48     Possibly  felt  in:   Anchorage,  Alaska,  United  States  (3/7)   61.218,  -­‐149.858     City:  Anchorage   Level2:  Anchorage  Borough   Level1:  Alaska   Country:  United  States   2014/04/16  20:25:48   UL:  No  loca(on  string   TXT:  EARTHQUAKE     2014/04/16  20:25:47   UL:  No  loca(on  string   TXT:  Holy  f*ing  s**t  that  was  an  earthquake     2014/04/16  20:25:44   UL:  Alaska.   TXT:  Oh  hi,  minor  earthquake.     2014/04/16  20:25:42   UL:  Anchorage   TXT:  Earthquake?     2014/04/16  20:25:42   UL:  None   TXT:  Earth  quake  ?     2014/04/16  20:25:41   UL:  Anchorage,  Alaska   TXT:  Earthquake!     MMI  IV  
16. { } CC-BY-ND 4.0 How are these detections and alerts used? 16   •  Frequently NEIC’s first indication of a widely felt event •  Detection of small felt events in sparsely instrumented places in the world •  Provides earthquake detections from an independent source
17. { } CC-BY-ND 4.0 Taking it to the next level •  Describing shaking intensity and effects •  How significant is an earthquake to the people that experienced it? •  How many people experienced the shaking? •  Was there damage? 17   Earthquake Characterization
18. { } CC-BY-ND 4.0 18   Follow on alerts @5min (0,523,695) Viña del Mar, Valparaiso, Chile: high, increasing High  interest:  609  tweets/min  average  in  first  5  minutes.   Increasing  interest  with  %me:  +33%  change  in  first  5  minutes.   Ini%al  repor%ng  loca%on:  Viña  del  Mar,  Valparaiso,  Chile     Tweet  Volume  details   0  tweets/min  pre  trigger  "noise”                                                                   523  tweets/min  in  first  2.5  minutes                                               695  tweets/min  between  2.5  and  5  minutes     Popula%on  Informa%on   From  ini(al  loca(on  es(mate:  Viña  del  Mar,  Valparaiso,  Chile  (-­‐33.02394,  -­‐71.55219)   Popula(on:   20km:  635,000   50km:  1,173,000   100km:  5,504,000     From  updated  loca(on  es(mate:  (-­‐33.200874837098496,  -­‐71.27149415535366)   Popula(on:   20km:  143,000   50km:  2,439,000   100km:  8,687,000  
19. { } CC-BY-ND 4.0 M5.2,  Depth  95  km    2014-­‐04-­‐04  9:52   19   Map  from  Centro  Sismologico  Nacional,  Chile   Max  MMI  IV     00:00 00:30 01:00 01:30 02:00 0 50 100 150 200 250 300 350 Time [Hours:Minutes] Count terremoto   temblor  
20. { } CC-BY-ND 4.0 00:00 00:30 01:00 01:30 0 50 100 150 200 Time [Hours:Minutes] Count M8.2,  Depth  20  km    2014-­‐04-­‐01  23:46   20   terremoto   temblor   Map  from  Centro  Sismologico  Nacional,  Chile   Max  MMI  VII    
21. { } CC-BY-ND 4.0 Tweet Based Earthquake Detection System 21   outputs detect, locate SQL  DB   geocode collect archive time-series Kibana Text messages e-mails Interactive maps Tweets/min select preferred tweets Elasticsearch
22. { } CC-BY-ND 4.0 22  
23. { } CC-BY-ND 4.0 23   Example Alert M6.0 6 km deep ~9 km SSW of Napa, CA Tweet  event  detec%on   NOT  AN  OFFICIAL  USGS  ALERT   NOT  SEISMICALLY  VERIFIED     Detec%on  Time:   2014/08/24  10:21:13     Possibly  felt  in:   Alameda County, CA, U.S. (3/13) 37.680, -121.906 City: Level2: Alameda County Level1: California Country: United States 2014/08/24  10:21:13   UL:  Bay  Area,  Ca.   GEO:  37.879,  -­‐122.371  CA,  United  States   TXT:  EARTHQUAKE!     2014/08/24  10:21:09   UL:  Oakland   GEO:  37.805,  -­‐122.273  Oakland,  Alameda  County,  CA,   United  States   TXT:  EARTHQUAKE  F**     2014/08/24  10:21:06   UL:  esports,  california   GEO:  37.272,  -­‐119.270  CA,  United  States   TXT:  MOTHERF**ING  EARTHQUAKE     2014/08/24  10:21:04   UL:  OAKLAND  STRAIGHT  EDGE   GEO:  37.792,  -­‐122.228  Oakland,  Alameda  County,  CA,   United  States   TXT:  EARTHQUAKE    
24. { } CC-BY-ND 4.0 24   M6.0 Depth 6km Napa, CA August 24, 2014 10:20:44 UTC MMI  VIII  
25. { } CC-BY-ND 4.0 25   M6.0 Depth 6km Napa, CA MMI  VIII  
26. { } CC-BY-ND 4.0 M 6.0 Napa, CA - Photos of Damage 26  
27. { } CC-BY-ND 4.0 More Fun To Come •  Migration to Kibana 4 •  Faster access to links in data •  Need better map •  Sharing of Kibana visualization components in alerts to mobile devices •  We need to improve integration between visualization components •  Adding in seismically confirmed data •  New ideas from this conference, let’s talk!!! 27  
28. { } Thank you for your interest and time. With special thanks to my team members: P. Earle, J. Allen, S. Horvath, J. Turner Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. Michelle Guy mguy@usgs.gov
29. { }
30. { } CC-BY-ND 4.0 This work is licensed under the Creative Commons Attribution-NoDerivatives 4.0 International License. To view a copy of this license, visit: http://creativecommons.org/licenses/by-nd/4.0/ or send a letter to: Creative Commons PO Box 1866 Mountain View, CA 94042 USA { 30