Original ArticleUsing localized Twitter activity to assess harmful algal bloom impacts of Karenia brevis in Florida, USA
Graphical abstract
Introduction
Red tide algal blooms associated with marine dinoflagellates are regular occurrences in coastal environments where oceanographic conditions, nutrient availability, and planktonic community interactions create conditions favorable for growth. Increases in the frequency, duration and severity of these events have occurred over the years (Tester, Steidinger, 1997, Tester, Steidinger, 1997, Kirkpatrick, Fleming, Squicciarini, Backer, Clark, Abraham, Benson, Cheng, Johnson, Pierce, et al., 2004, Kirkpatrick, Fleming, Squicciarini, Backer, Clark, Abraham, Benson, Cheng, Johnson, Pierce, et al., 2004, Glibert, Anderson, Gentien, Granéli, Sellner, Glibert, Anderson, Gentien, Granéli, Sellner), stimulated by multiple factors related to coastal land development and climate change (Glibert, 2020, Glibert, 2020). Toxins (Lin, Risk, Ray, Van Engen, Clardy, Golik, James, Nakanishi, 1981, Lin, Risk, Ray, Van Engen, Clardy, Golik, James, Nakanishi, 1981, Poli, Mende, Baden, 1986, Poli, Mende, Baden, 1986) produced by these species or degradation of water quality conditions with cell decomposition can have multiple negative impacts on marine environments and human communities in coastal areas. Large die-off events of marine organisms are commonly observed with red tides (Sievers, 1969, Sievers, 1969, Steidinger, 1996, Steidinger, 1996), including fish, marine associated birds, turtles, and large mammals (e.g., cetaceans, sirenians). Bioaccumulation of neurotoxins can further alter ecosystem trophic dynamics (Pierce, Kirkpatrick, 2001, Pierce, Kirkpatrick, 2001). Coastal conditions for humans during ride tides are also negatively impacted, as toxins are aerosolized out of the water column contributing to respiratory or skin irritation that can lead to hospitalizations (Backer, Kirkpatrick, Fleming, Cheng, Pierce, Bean, Clark, Johnson, Wanner, Tamer, et al., 2005, Backer, Kirkpatrick, Fleming, Cheng, Pierce, Bean, Clark, Johnson, Wanner, Tamer, et al., 2005, Fleming, Kirkpatrick, Backer, Bean, Wanner, Dalpra, Tamer, Zaias, Cheng, Pierce, et al., 2005, Fleming, Kirkpatrick, Backer, Bean, Wanner, Dalpra, Tamer, Zaias, Cheng, Pierce, et al., 2005, Fleming, Kirkpatrick, Backer, Bean, Wanner, Reich, Zaias, Cheng, Pierce, Naar, et al., 2007, Fleming, Kirkpatrick, Backer, Bean, Wanner, Reich, Zaias, Cheng, Pierce, Naar, et al., 2007, Milian, Nierenberg, Fleming, Bean, Wanner, Reich, Backer, Jayroe, Kirkpatrick, 2007, Milian, Nierenberg, Fleming, Bean, Wanner, Reich, Backer, Jayroe, Kirkpatrick, 2007). Economic impacts are also observed, as tourism or waterfront business revenues are reduced in areas affected by red tide.
Red tides from K. brevis have been observed on the southwest coast of Florida for over a century. Similar to global trends in harmful algal bloom (HAB) occurrences, these events have increased in severity, frequency, and geographic extent in recent years. While blooms originate offshore on the West Florida Shelf, onshore winds and currents can bring red tide blooms inshore to beaches and estuaries along Florida’s Gulf Coast, where they may be exacerbated by coastal nutrient sources. Extensive hydrologic modification of southwest Florida has altered the quantity, location, and timing of nutrient delivery from background conditions. Land use conversion of wetlands to agriculture and urbanization of coastal areas has further altered drainage patterns. The impacts of these changes on coastal water quality in southwest Florida have been exacerbated by climate change, as seasonal temperature and precipitation patterns have deviated from historical averages. Although no single cause has been identified, the synergistic effects of multiple stressors have likely contributed to significant red tide events. One of the longest and most severe events to date persisted on the coast of southwest Florida for 16 months from Fall of 2017 to early 2019.
Given the potential for large negative effects of red tide events on the environment, economy and public health, additional datasets that can augment conventional monitoring data could improve how local governments plan and respond to these events. Local governments and regional planning councils rely on regularly updated and routine monitoring data to respond appropriately to environmental disasters. These data inform how future response is managed to anticipated events, as well as how real-time response to ongoing events changes relative to where impacts are most observed. For red tide events, local governments may require data on where blooms are most severe to inform clean-up events (e.g., in response to fish die-offs) or to issue public safety announcements regarding beach closures or air quality advisories. Real-time rapid assessment of in situ conditions can inform these decisions, whereas more detailed information about oceanographic conditions and phytoplankton cell counts that require lab processing can be used for forecasting where bloom events are likely to occur in the future (National Oceanic and Atmospheric Administration (NOAA) (2018), University of South Florida Ocean Circulation Group (2010)). Moreover, conducting keyword analysis of social media messages could help in determining the issues of public’s utmost concern (e.g. is it economic impacts? health consequences? general concern for the environment?), hence informing decisions on the optimal distribution of the resources and funding (e.g. assisting waterfront businesses, distributing protective equipment, organizing marine debris cleanup, etc).
The recent bloom in southwest Florida was the first major red tide event since broad public use of social media platforms, offering potentially unique opportunities to assess complementary sources of information that can aid management response to disaster events. During catastrophes like floods, earthquakes, and hurricanes people have been turning to platforms such as Twitter (Eismann, Posegga, Fischbach, Eismann, Posegga, Fischbach, Kryvasheyeu, Chen, Obradovich, Moro, Van Hentenryck, Fowler, Cebrian, 2016, Kryvasheyeu, Chen, Obradovich, Moro, Van Hentenryck, Fowler, Cebrian, 2016, Mukkamala, Beck, 2017, Mukkamala, Beck, 2017, Zou, Lam, Shams, Cai, Meyer, Yang, Lee, Park, Reams, 2019, Zou, Lam, Shams, Cai, Meyer, Yang, Lee, Park, Reams, 2019), Facebook (Bird, Ling, Haynes, et al., 2012, Bird, Ling, Haynes, et al., 2012, Eismann, Posegga, Fischbach, Eismann, Posegga, Fischbach, Bhuvana, Aram, 2019, Bhuvana, Aram, 2019, Jayasekara, 2019, Jayasekara, 2019), Instagram (Sherchan, Pervin, Butler, Lai, Ghahremanlou, Han, 2017, Sherchan, Pervin, Butler, Lai, Ghahremanlou, Han, 2017), Flickr (Liu, Palen, Sutton, Hughes, Vieweg, et al., 2008, Liu, Palen, Sutton, Hughes, Vieweg, et al., 2008, Preis, Moat, Bishop, Treleaven, Stanley, 2013, Preis, Moat, Bishop, Treleaven, Stanley, 2013, Chien, Comber, Carver, 2017, Chien, Comber, Carver, 2017) to share their sentiments and experiences, making social media a valuable source of information. There’s been a variety of aspects studied in cases of utilizing social media within the context of disaster events, ranging from collective behavior analysis (Liu, Palen, Sutton, Hughes, Vieweg, et al., 2008, Liu, Palen, Sutton, Hughes, Vieweg, et al., 2008, Eismann, Posegga, Fischbach, Eismann, Posegga, Fischbach), to risk and crisis communication (Veil, Buehner, Palenchar, 2011, Veil, Buehner, Palenchar, 2011, Bird, Ling, Haynes, et al., 2012, Bird, Ling, Haynes, et al., 2012), to relief effort coordination (Gao, Barbier, Goolsby, 2011, Gao, Barbier, Goolsby, 2011, Purohit, Castillo, Diaz, Sheth, Meier, 2014, Purohit, Castillo, Diaz, Sheth, Meier, 2014, Bhuvana, Aram, 2019, Bhuvana, Aram, 2019), among others. Meanwhile, in this work we intend to focus on studying the spatio-temporal correlations between localized social media activity and disaster impacts to respective areas (Preis, Moat, Bishop, Treleaven, Stanley, 2013, Preis, Moat, Bishop, Treleaven, Stanley, 2013, Kryvasheyeu, Chen, Obradovich, Moro, Van Hentenryck, Fowler, Cebrian, 2016, Kryvasheyeu, Chen, Obradovich, Moro, Van Hentenryck, Fowler, Cebrian, 2016, Chien, Comber, Carver, 2017, Chien, Comber, Carver, 2017, Chen, Wang, Ji, 2020, Chen, Wang, Ji, 2020), along with conducting keyword and sentiment analysis for gauging public’s situational awareness during the disaster (Verma, Vieweg, Corvey, Palen, Martin, Palmer, Schram, Anderson, 2011, Verma, Vieweg, Corvey, Palen, Martin, Palmer, Schram, Anderson, 2011, Yin, Karimi, Robinson, Cameron, 2012, Yin, Karimi, Robinson, Cameron, 2012, Mukkamala, Beck, 2017, Mukkamala, Beck, 2017). That would serve to provide an additional use case and partially confirm some of the research done in Preis, Moat, Bishop, Treleaven, Stanley, 2013, Preis, Moat, Bishop, Treleaven, Stanley, 2013, Kryvasheyeu, Chen, Obradovich, Moro, Van Hentenryck, Fowler, Cebrian, 2016, Kryvasheyeu, Chen, Obradovich, Moro, Van Hentenryck, Fowler, Cebrian, 2016, when user activity on Twitter and Flickr was shown to be a good proxy for evaluating local damages during Hurricane Sandy, or the study conducted in Chen, Wang, Ji, 2020, Chen, Wang, Ji, 2020, where tweets containing certain keywords were shown to correlate with road closures due to Hurricane Harvey flooding.
Our study focused on Twitter - a "microblogging" service used by over 330 million people worldwide to share short strings of text, or "tweets", to convey opinions or ideas on any topic. Besides its efficiency of communication (due to character restrictions imposed on tweet content), a key component of Twitter is the ability to share or "retweet" information from other users such that a single message can reach a much wider audience than the user’s immediate followers. Tweets can also be used to assess a response in time (i.e., how do people react after an event) that can be traced to a geographic location if, for example, a tweet is spatially referenced by the user. Thus, interpretation of tweet content across multiple users, relative numbers of retweets, spatial, and temporal information can provide an assessment of topic importance that can guide response and recovery efforts. The use of Twitter to both gauge the impacts and inform disaster response has received some attention in the recent literature. For example, links between Twitter activity and localized storm damages (Kryvasheyeu, Chen, Obradovich, Moro, Van Hentenryck, Fowler, Cebrian, 2016, Kryvasheyeu, Chen, Obradovich, Moro, Van Hentenryck, Fowler, Cebrian, 2016) and community response to storm events when power outages prevented use of conventional communication outlets (Pourebrahim, Sultana, Edwards, Gochanour, Mohanty, 2019, Pourebrahim, Sultana, Edwards, Gochanour, Mohanty, 2019) have both been demonstrated with an analysis of tweet data during Hurricane Sandy. While storm events unfold over days, red tide events can last several months and thus introduce a significant temporal as well as spatial component to the impacts.
Here, we present a multi-level analysis of Twitter activity as it relates to the Florida 2017–2019 red tide event. Our overall goal was to evaluate the use of Twitter as a potential complement for data on real-time conditions as measured by in situ observations (e.g., phytoplankton cell counts) or reported conditions relevant to public health (e.g., beach conditions). The intent is not to supplant existing methods for tracking status and trends of red tide conditions, but to understand social responses to these events that may have management implications for improving communication about red tide, both general education and real time conditions, and for prioritizing prevention or mitigation actions by local governments. The use of Twitter as a reflection of red tide conditions could also augment existing datasets by providing either confirmatory, or in some cases anticipatory, information in the absence of in situ data. For example, this may have value if regular monitoring programs for tracking conditions do not exist or lab processing times prevent rapid response to current conditions. Lastly, we analyzed sentiment and topical nature of the tweets for the purposes of gauging public’s situational awareness during the event, determining the issues of concern (environment, health, economy, government), and potentially improving the accuracy of Twitter metrics as proxies for local red tide conditions.
Section snippets
Red tide local conditions data
For local beach conditions data (Mote Marine Laboratory, 2018), such as dead fish and respiratory irritation levels, we used daily reports from 12 main beaches, which included six in Sarasota county (Manasota, Venice, Venice/North Jetty, Nokomis, Siesta Key, Lido Key beaches), two in Manatee (Coquina, Manatee beaches), and four in Pinellas (Pass-a-Grille, Treasure Island, St Pete, Clearwater beaches). When accumulating the data by counties, the observed dead fish and respiratory irritation
Cumulative Twitter activity on ”red tide” topic across metro areas.
One of the main goals was to study localized Twitter activity in five Florida Gulf coast counties located in the Tampa Bay surrounding area: Pasco, Hillsborough (including Tampa metro), Pinellas (Clearwater and St Petersburg metro areas), Manatee (including Bradenton metro), and Sarasota (including cities of Sarasota, Venice and Englewood). First, evaluating each area’s total Twitter activity - throughout the entire duration of the Florida red tide event, including the anticipation and
Things that didn’t work: User account types, lead/lag correlations
Among several other things we have tried that didn’t come to fruition were the considerations to focus on correlations of red tide conditions with tweets from specific account types (e.g. only media, only citizens), and studying the delayed reactions and anticipation of red tide impacts via lag and lead correlations, respectively (Spriggs, Kaylen, Bessler, 1982, Spriggs, Kaylen, Bessler, 1982, Dajcman, 2013, Dajcman, 2013). For the former, we hypothesized that tweets from citizen accounts
Discussion
As one of the main conclusions to our study, we discovered that, during a lengthy disaster event such as Florida red tide of 2017–2019, Twitter activity can be used to gauge community perceptions of localized coastal impacts, in both comprehensive and discrete spatiotemporal scales. One can consider several distinct tweet metrics - counts or sentiments of tweets that were either explicitly geo-tagged or geo-matched to the area in any way - for improved representation of the relative damage to
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This project was supported in part through the Tampa Bay Environmental Restoration Fund (TBERF) and a USEPA Region 4 Cooperative Agreement Grant to the Tampa Bay Estuary Program (CE-00D89319-0). TBEP funding for this project stems from USEPA Section 320 Grant Funds, and the TBEP’s local government partners (Hillsborough, Manatee, Pasco, and Pinellas Counties; the Cities of Clearwater, St. Petersburg, and Tampa; Tampa Bay Water; and the Southwest Florida Water Management District) through
References (60)
- et al.
Facebook and whatsapp as disaster management tools during the Chennai (India) floods of 2015
International Journal of Disaster Risk Reduction
(2019) - et al.
25 years of time series forecasting
International journal of forecasting
(2006) - et al.
Aerosolized red-tide toxins (brevetoxins) and asthma
Chest
(2007) - et al.
Harnessing the crowdsourcing power of social media for disaster relief
IEEE intelligent systems
(2011) Harmful algae at the complex nexus of eutrophication and climate change
Harmful algae
(2020)- et al.
Literature review of florida red tide: implications for human health effects
Harmful algae
(2004) - et al.
Florida red tide knowledge and risk perception: is there a need for tailored messaging
Harmful algae
(2014) - et al.
Florida red tide perception: residents versus tourists
Harmful algae
(2010) - et al.
Understanding communication dynamics on twitter during natural disasters: a case study of hurricane sandy
International Journal of Disaster Risk Reduction
(2019) - et al.
The compassion fatigue scale: its use with social workers following urban disaster
Research on social work practice
(2008)
Occupational exposure to aerosolized brevetoxins during florida red tide events: effects on a healthy worker population
Environmental health perspectives
Flooding facebook-the use of social media during the queensland and victorian floods
Australian Journal of Emergency Management, The
Time-series forecasting
Assessing disaster impacts on highways using social media: Case study of hurricane harvey
Construction Research Congress 2020: Infrastructure Systems and Sustainability
Does flickr work in disaster management?–a case study of typhoon morakot in Taiwan
Proc. of GIS Research UK (GISRUK)
Interdependence between some major european stock marketsa wavelet lead/lag analysis
Prague economic papers
Initial evaluation of the effects of aerosolized florida red tide toxins (brevetoxins) in persons with asthma
Environmental health perspectives
Language matters in twitter: a large scale study
Proceedings of the International AAAI Conference on Web and Social Media
Compassion fatigue, compassion satisfaction, and burnout: reactions among employee assistance professionals providing workplace crisis intervention and disaster management services
Journal of workplace behavioral health
Role of facebook as a disaster communication media
International Journal of Emergency Services
Timeline of severe red tide events on the West Florida Shelf: insights from oral histories
Technical Report
Rapid assessment of disaster damage using social media activity
Science advances
Public perceptions of florida red tide risks
Risk Analysis: An International Journal
Risk in daily newspaper coverage of red tide blooms in southwest florida
Applied Environmental Education & Communication
Isolation and structure of brevetoxin B from the” red tide” dinoflagellate Ptychodiscus brevis (Gymnodinium breve)
Journal of the American Chemical Society
In search of the bigger picture: The emergent role of on-line photo sharing in times of disaster
Proceedings of the Information Systems for Crisis Response and Management Conference (ISCRAM)
Geo-temporal twitter demographics
International Journal of Geographical Information Science
A Twitter-lived red tide crisis on Chiloé Island, chile: what can be obtained for social-ecological research through social media analysis?
Sustainability
Cited by (7)
What's in a name? Political and economic concepts differ in social media references to harmful algae blooms
2024, Journal of Environmental ManagementRegional geographical and climatic environments affect urban rainstorm perception sensitivity across China
2022, Sustainable Cities and SocietyCitation Excerpt :Accordingly, we constructed the urban rainstorm attention index (URAI) from the microblogs that were posted by local users only. The definition and calculation approach to the URAI metric has been widely used in current research as it facilitates fair comparisons between cities (Skripnikov et al., 2021; Zou, 2019). For example, Kryvasheyeu et al. (2016) used a ratio between the number of daily messages and the number of local active users during the observation period as a proxy to measure the difference of tweeting activities among cities.
Initial estuarine response to inorganic nutrient inputs from a legacy mining facility adjacent to Tampa Bay, Florida
2022, Marine Pollution BulletinCitation Excerpt :Occurrence of this species has historically been spatially distinct, with blooms originating in subsurface water offshore on the West Florida Shelf (Liu et al., 2016; Steidinger, 1975; Weisberg et al., 2014, 2019) and occasionally occurring at bloom concentrations in lower and middle Tampa Bay. Although bloom concentrations in 2021 were extreme, historical blooms have been observed in Tampa Bay with notable events occurring in 1971 (Steidinger and Ingle, 1972), 2005 (Flaherty and Landsberg, 2011), and recently in 2018 (Skripnikov et al., 2021). Seasonal persistence in Gulf waters in southwest Florida can vary between years, with some blooms lasting as short as a few weeks, while others have been present for longer than a year (the 2018 bloom lasted sixteen months, Skripnikov et al., 2021).
Algorithmic uncertainties in geolocating social media data for disaster management
2023, Cartography and Geographic Information Science