IESKO 2019, Special Symposium for the 20th Anniversary of Kocaeli and Düzce Earthquakes, Kocaeli, Turkey, 25 - 27 September 2019, pp.859-863
Earthquake can be determined by two sources. The first source is taken from earthquake
monitoring stations formed by physical sensors. the second source is Online Social Networks
(OSN). Tweeter users might think of people as some kind of sensor. People do not express their
feelings in a numerical way like electronic sensors. But they express what they perceive in
words. For this reason, the necessary information should be extracted by processing the texts
taken from the OSN.
In this study, Twitter was chosen as the OSN. Twitter, the popular social media world in 2019
according to the eleventh of June data is fifth in Turkey. Turkey ranks seventh in the world in the
use of Twitter. Turkey has a population of 2.4 million. 72% of this population is an internet user.
63% of the population uses Social Networks and 53% of them use mobile. Therefore, we may
think that we have many sensors in the vicinity of the earthquake epicenter. The Twitter message
includes the geographical coordinates of the place where the user sends the message. The latitude
and longitude information in the twitter message can give information about how Mercalli's
intensity is felt in places different from the earthquake epicenter. USGS 's record that the last
three years and greater than five MMI (Modified Mercalli Intensity) magnitude in Turkey
earthquake simultaneous tweets were analyzed together. Tweets were collected with the word
"earthquake". This Tweet time series is associated with the physical earthquake event. Shortly
after the earthquake, the earthquake can easily be seen in the Tweet. They write their own
impressions of the earthquake and provide additional information as well as data from the
seismographs. Geotagged tweets were used to directly validate the earthquake. For Turkey
Earthquake Warning System based on Twitter messages, to determine the abnormal event
throwing a tweet, an algorithm was proposed two methods consist of MAD and STL.