PS 87-258
Twitter as a data resource for understanding human-wildlife interactions in urban ecosystems
Wildlife formerly excluded from human settlements is re-establishing a foothold in urban settings in unprecedented numbers. While many welcome nature on their doorsteps, consequences can be detrimental to the larger ecosystems in which urban areas are embedded. Efforts are increasing to control species that so successfully coexist with humans that they threaten the species diversity necessary to healthy ecosystems. However, rapid expansion of wildlife ranges into urban settings is a relatively new phenomenon. It is often unclear what factors favor one species over another or how to rectify imbalance without alienating the public.
Our survey studies (AGU2011/AMS2012/AGU2103/WDAFS2014) identify Twitter as an intriguing, largely untapped source of observations of the natural world. Users spontaneously mention sightings of a wide variety of wildlife throughout the day and year. Users notice and report on birds, fish, crustaceans, mammals, reptiles, amphibians, insects & bugs moving, migrating, mating, nesting, hunting, foraging, defecating and playing in a broad range of urban, suburban, rural and wild contexts, outdoors and in, during all types of human activities. They include details on appearance/behavior and express a spectrum of emotional reactions to the observation/interaction. These studies establish that Twitter provides an intimate, immediate observational window on human-wildlife interactions. Our current research pursues this insight to investigate the specific utility of the data resource to inform decision-making processes for managing urban wildlife and ecosystems impacts.
Results/Conclusions
This poster presents the results of our analysis of Twitter observations extracted for a representative sample of urban animal species. We find that observation frequency ranges from ~1/month to greater than 10/day. We present frequency analyses of the environments in which animals are observed (e.g., yards, parks, roads, roofs), animal behaviors as well as observer activities, reactions and responses. We discuss challenges in distilling information from crowd-sourced observations (140-character 'tweets') and strategies to surmount them. Determining accurate fine-scale geolocations for individual tweets is a particularly complex problem. Despite widespread use of smart phones, explicit lat-long coordinates are included in only 1-3% of observations. We discuss technical strategies for overcoming this challenge and present examples of how the geolocatable tweet fraction can be increased by using user profile and contextual geolocation algorithms. We identify potential outlier observations and map observations by urban area. Based on these results we offer suggestions on how the ecological and environmental science communities can use this type of tweet-observations data to inform decision-making processes for managing urban wildlife.