Using Twitter to Map Nuisance Flooding

| |


Share this article:

Along the 3,700 miles of coastline that make up the East Coast and Gulf Coast region of the United States, only 132 tidal gauge stations exist with long-term records. This paucity of measuring stations makes understanding the localized impact of coastal flooding challenging. To add to this, tidal gauge stations use three standard tidal heights to determine minor, moderate, and major flooding. Flooding can be highly variable within a small region, depending on topography and bathymetry. With nuisance, or sunny day flooding, on the rise due to climate change, understanding localized flooding is becoming increasingly important. The impacts of what is considered moderate flooding at one location may not have the same impact at another location. The authors of a recently published paper on measuring localized flooding note, “The same degree of inundation could have substantively different social impacts, depending on the distribution of people, infrastructure and economic activity along the coast.” Interpolating flooding from the sparse network of tidal gauge stations is not fine-grained enough.

To get a better understanding of localized flooding, the researchers mined Twitter data for geolocated tweets to fill in the gaps in data about flooding. The authors used these Tweets for two purposes: to map out localized flood events and to measure the social impact of flooding. The authors provided the example of flooding that covers an uninhabited area such as farmland has a lower social impact than flooding that impacts an important road.

For the study, researchers analyzed Tweets between between March 2014 and November 2016 that occurred in shoreline countries along the Atlantic and Gulf coasts of the United States. Those tweets were classified based on a bag-of-words approach that looked for words associated with flooding. Researchers ended up with 473,000 flood-related tweets from 235 shoreline counties representing 5 million Twitter users. The data was then combined with tide gauge data on maximum daily tide height from stations along the Atlantic and Gulf coasts by matching the county to the closest tidal gauge station. Using a measurement of a 25% increase in Tweets about flooding to signal a remarkable flooding event, researchers were able to use this methodology to identify 22 counties that had experienced nuisance flooding at thresholds lower than the flood threshold standards. These countries include several major cities such as Miami, New York, and Boston, indicating that unidentified nuisance flooding is occurring because it doesn’t reach standard thresholds.

Researchers measure incidents of nuisance flooding by analyzing the increase in frequency of geolocated flood related tweets.  Source: Moore & Obradovich, 2020.
Researchers measure incidents of nuisance flooding by analyzing the increase in frequency of geolocated flood related tweets. Source: Moore & Obradovich, 2020.

The study:

Moore, F. C., & Obradovich, N. (2020). Using remarkability to define coastal flooding thresholds. Nature Communications, 11(1), 530.



Most Tourists Don’t Stray Far From the Road When Taking Photos in National Parks

Counter Mapping and Cultural Preservation