Testing the potential of Twitter mining methods for data acquisition: Evaluating novel opportunities for ecological research in multiple taxa

Abstract

Social media provides unique opportunities for data collection. Retrospective analysis of social media posts has been used in seismology, political science and public risk perception studies but has not been used extensively in ecological research. There is currently no assessment of whether such data are valid and robust in ecological contexts. We used “Twitter mining” methods to search Twitter (a microblogging site) for terms relevant to three nationwide UK ecological phenomena: winged ant emergence; autumnal house spider sightings; and starling murmurations. To determine the extent to which Twitter‐mined data were reliable and suitable for answering specific ecological questions the data so gathered were analysed and the results directly compared to the findings of three published studies based on primary data collected by citizen scientists during the same time period. Twitter‐mined data proved robust for quantifying temporal ecological patterns. There was striking similarity in the temporal patterns of winged ant emergence between previously published work and our analysis of Twitter‐mined data at national scales; this was also the case for house spider sightings. Spatial data were less available but analysis of Twitter‐mined data was able to replicate most spatial findings from all three studies. Baseline ecological findings, such as the sex ratio of house spider sightings, could also be replicated. Where Twitter mining was less successful was answering specific questions and testing hypotheses. Thus, we were unable to determine the influence of microhabitat on winged ants or test predation and weather hypotheses for initiation of murmuration behaviour. Twitter mining clearly has great potential to generate spatiotemporal ecological data and to answer specific ecological questions. However, we found that the types and usefulness of data differed substantially between the three phenomena. Consequently, we suggest that understanding users' behaviour when posting on ecological topics would be useful if using social media is to generate ecological data.

Applications for deep learning in ecology

“…Also, in order to reduce illegal trafficking, it has been suggested to use deep learning algorithms to monitor such activities on social media to automatically detect pictures of illegal wildlife products (Di Minin, Fink, Tenkanen, & Hiippala, ). Using deep learning for data mining could easily be extended to other areas, as social media mining has proven to be useful for ecological research such as phenological studies (Hart, Carpenter, Hlustik‐Smith, Reed, & Goodenough, ).…”

Los foros y las redes sociales ayudan a descubrir especies exóticas en Europa y monitorear su propagación: el caso de <em>Exaireta spinigera</em> (Wiedemann, 1830) (Diptera, Stratiomyidae) en la península italiana y Sicilia

“…The use of social media to gather relevant data on biological and ecological issues is likely to increase in the future (e.g. Hart et al, 2018).…”

iEcology: Harnessing Large Online Resources to Generate Ecological Insights

“…This showcases the potential of using voluminous search engine data to explore species distributions in many regions. Others have explored species occurrences and distributions using various sources, such as Flickr, news articles, Twitter, YouTube, Facebook, and Google Trends [13][14][15][16][17][18][19][20][21][22][23][24][25], as well as population dynamics and phenology [14,20,23,[26][27][28][29][30][31]. A particular illustration comes from assessing seasonal migration patterns of sockeye salmon (Oncorhynchus nerka) and Atlantic salmon (Salmo salar) from Wikipedia pageview frequencies (Figure 2B) [32].…”

“…The vast majority of iEcology studies that we have identified have used multiple data sources to validate results, including data from field research, citizen science, online databases, scientific literature, or their combination [12,13,[16][17][18][19][20][21][22][23][24][25]31,32,35,43,44]. In most cases, authors report a satisfying to excellent level of consistency among data sources.…”