Keywords

1 Introduction

On September 2017, two major earthquakes hit Mexico: one on the 7th with an epicenter close to the southern state of Chiapas (marked with number one in Fig. 1 and the other on the 19th (number 2) in the central plateau of the country, which includes Mexico City and other densely populated centersFootnote 1. The epicenters were located approximately 800 km apart from each other with very strong seismic forces: 8.2 Mw the first one, and 7.1 Mw the latter. Despite the very high intensity of these earthquakes, casualties remained under 320 persons [5, 6], although there were severe damages in buildings and roads on those areas.

Fig. 1.
figure 1

September 7th, 2017 earthquake (1) and September 19th, 2017 earthquake (2).

Full size image

Mexico is geographically located in one of the most active seismic zones in the world. Particularly on its western side, the country rests over the Cocos, Pacific, Rivera and North American plates [9] which are in constant motion. The aforementioned tectonic plates are regularly causing earthquakes, most of negligible effect, almost every day of the year.

In contemporary Mexico, there have been massive affections caused by earthquakes. Coincidentally, also on September 19th, 1985 there was an earthquake that has the record of the most destructive in the history of Mexico, with a magnitude of 8.1 Mw and which caused more than 9,000 deaths and major infrastructure affectations, particularly n Mexico City [8].

2 Social Media Activity During the September 7th and 19th Earthquakes

According to [3], in 2016 47% of Mexican homes had an internet connection and 73.6% of people over 6 years old in the country had a smartphone. However, the states where the September 7th earthquake occurred (Oaxaca and Chiapas) are the ones with the least internet access in the country, whilst two of the states most affected by the September 19th earthquake (Mexico City and Morelos) are above the national average (Fig. 2). It must also be noted that both Oaxaca and Chiapas are at the bottom of GDP distribution in Mexico [1].

Fig. 2.
figure 2

Percentage of mobile internet access by state in Mexico

Full size image

As expected, social media activity was very intense in the aftermath of those events and was vital in disseminating very important information for everyone involved. Social networks such as Facebook, Twitter and WhatsApp were flooded with posts regarding the earthquakes. They were particularly useful for the coordination of all the support networks and on a personal level to keep track of friends and family living in the affected areas. For our research purposes, we focused in analyzing the information generated on Twitter. As of 2017, Twitter is one of the most prominent social network in the country [7] with more than nine million active users in Mexico alone, and was extensively used during the earthquakes.

However, as in other regions of the world that have experienced natural events as these ones there are several problems attached to all this flow of data coming from social networks, such as information overload, questionable speed of information delivery, difficulties of processing information in a non-standard format from different sources and in various languages, the complexity of managing volunteer communities and the very limited value of using information at the street level [2].

3 Tweets Filtering

Immediately after the September 7th earthquake we began collecting tweets that were pertinent to this event. We used a linux machine to constantly monitor tweet activity originated in Mexico. As [4] mentions, when a disaster occurs, time is limited and safety is in ques- tion, so people need to act quickly with as much knowledge of the situation as possible. It is becoming more common for affected populations and other stakeholders to turn to Twitter to gather information about a crisis when decisions need to be made, and action taken.

The trending topic for these tweets were by far related to the seismic phenomenon. Those tweets were clustered based on their hashtagsFootnote 2, and after that we ordered them based on their frequency. We decided to let the computer gather tweets during the rest of the month to check on the persistence of the hashtags selected. Unfortunately, during the same month the second earthquake happened. We then proceeded to apply the same procedure to this event to identify the most frequent hashtags (Table 1).

Table 1. Hashtags most used during the two earthquakes in Mexico during September 2017.
Full size table

During the first event, most tweets were sent to express solidarity (like those with the #PrayforMexico or #TacoLaArepaEstaContigo hashtags. Tweets coming from the affected areas were less prominent and sparse. Reports on telecommunication disruptions suggest that most inhabitants with mobile or wired internet access before the earthquake could not immediately communicate their well being or current needs.

Even when the state of emergency in southern Mexico was still active, the focus on Twitter turned completely to the event in Central Mexico on September 19th. Right after the event the tweets began to be sent, overwhelmingly coming from Mexico City, and the states of Puebla and Morelos (the states more affected during this event). The hashtags associated with those tweets were noticeably different in the sense that they served as the basis for the creation of a support network for people in distress (#AyudaCDMX). There was also an immediate call for government accountability (#revisa mi grieta, #PartidosDenSuDinero) through the tweets.

4 Future Work

The first phase of our work consisted in gathering tweets that would be filtered to search for those related to earthquakes in Mexico. Our server has been retrieving tweets ever since the September 7th earthquake, and during the course of our research another major seismic event hit Mexico: on February 13th a 7.2 Mw earthquake with an epicenter close to the Oaxaca coast, affecting the same zones as the September 7th one. We applied the same heuristics to the new bank of data, identifying hashtags and ordering based on their frequency. The next phase in our research aims to automatize this process and embed sentiment and trend analysis for tweet classification, which we hope will help for better decision making in the event of an emergency.