Thought, it remains difficult to globally estimate the impact of human isolation on wildlife activity. There are a few local initiatives such as determinobs.fr or ispotnature.org but these datasets presents some limitations to address this specific subject :
Then, how can we improve global knowledge about wildlife activity ? Where could we find more data to complete these already existing but partial datasets ?
We believe the answer lies in the frenetic global usage of social media.
Twitter for instance displays no less than 500 millions tweets per day. If only a mere fraction of this activity would be useful for wildlife observation, it would mean quite a deal for naturalists and environment enthusiasts !
The idea was to look for the information where it already is, and benefits from people's enthusiasm for wildlife and nature. Better I than IA !
So we tried to catch some data in order to show up that is was possible to enrich existing maps where volunteers reports animals activity by the thousands who do it without thoughts on helping science.
We started studying Twitter available public data, its developer APIs and relevant hashtags to gather data. We developed a small Python HTTP Client to interact with the standard API to retrieve as much data as we could from the past seven days, coupling these first results with the use of Twitter's streaming API to retrieve live tweets on the hashtags we identified (wildlifefrommywindow, wildlifethroughmywindow, aworldwithoutus and birdwatch).
ELK (Elasticsearch Logstash Kibana) stack was a simple solution we knew to gather some information and show that trending tags spread across countries with pictures and videos of wildlife which could be valuable for scientists.
In the end, we have been able to gather thousands of tweets data, among media content, creation date and some location information. Many of these data needs filtering and more post-processing to be useful though. Quite surprisingly, most of Twitter data does not seem to be geo-located which has eventually been the major drawbacks of our solution.
An ideal solution would gather media content (photos or videos) to allow wildlife experts to identify the species appearing in the original tweet, as well as accurate location data to be visualized on a map and correlated with earth observation data.
We believe that social media can play a huge part in wildlife data gathering, facilitating the publishing of relevant information for future observation of human activity impact on Nature.
https://drive.google.com/file/d/1aZ9EVSOy514wR-sMXGlEQI1tUJFiT5nq/view?usp=sharing
https://drive.google.com/file/d/1wLLWqqzgs3kY44RwH5pMc-kzswwr4Hlx/view?usp=sharing
https://drive.google.com/file/d/1ikTv-T-aLlDsLyaE-ktnWD5u-3VIzlmi/view?usp=sharing
As the datasources provided in example, mainly maps, were less accurate than what we expected to catch wildlife, the idea is to focus on people initiative with example from twitter.