Real-Time Earthquake Intensity Estimation Using Streaming Data Analysis of Social and Physical Sensors

The Predictive Relationship between Earthquake Intensity and Tweets Rate for Real‐Time Ground‐Motion Estimation

“…In order to validate whether the numerical results quantifying hypothesized relationships between logarithmic tweet rates following ten minutes after an earthquake and earthquake intensity on MMI scale, obtained from Kropivnitskaya et al (2016), can be used for other earthquakes in California and in other regions around the world, we selected independent events for testing purposes listed in Table 1. The magnitude scale used in Table 1 is moment magnitude.…”

“…The second limitation is related to the lack of geotagged data within available Twitter data sets. We selected seismic events that both showed a spike in the amount of positive tweets within ten minutes following the earthquake and where at least fourteen of those tweets were geotagged (the minimum number of tweets in the models (Kropivnitskaya et al, 2016)). The geotagged tweets contain the current user location indicator at the time of tweeting and can be directly used in the location-specific intensity estimation algorithms.…”

“…include a named location, 20.4% referenced information that can be used to identify a country, while 15.1% provided humorous or non-spatial information (Takhteyev et al, 2012)). In the approach used by Kropivnitskaya et al (2016), all three types of geotagged technics were used and the same logic has been implemented here. A text-based geolocation algorithm has been improved by employing a location database extension for California, Japan and Chile.…”

“…Previous studies have shown the potential of Twitter data for earthquake detection (Earle et al, 2010(Earle et al, , 2011Sakaki, 2010;Crooks, 2012;Burks et al, 2014) and intensity estimation (Kropivnitskaya et al, 2016). Kropivnitskaya et al (2016) created four empirical predictive relationships (linear, two-segment linear, three-segment linear, exponential) that link the positive tweet rates in the first ten minutes following the earthquake with the instrumental intensity level in MMI scale units from a regression analysis of data from physical and social sensors during the Napa earthquake. Figure 1 shows the ratio between the combined data (instrumental and Twitter) and the MMI intensity values recorded from the USGS in the Napa region.…”

“…One potentially important data source in this category is the widespread micro-blogging platform Twitter, ranked ninth worldwide as of January 2016 by number of active users, ~320 million (Twitter, 2016). In our previous studies, empirical relationships between tweet rate and observed Modified Mercalli Intensity (MMI) were developed using data from the M6.0 South Napa, CA earthquake (Napa earthquake) that occurred on August 24, 2014 (Kropivnitskaya et al, 2016). These relationships allow us to stream data from social sensors, supplementing data from other sensors to produce more accurate real-time intensity maps.…”