The Taiwan Roadkill Observation Network Data Set.

Stage 1: Crawling Facebook Posts for Occurrence Data to Upload to Database

This process was in use from August 2011 to the end of 2015 in which the volunteers posted observation data directly to the Facebook group.

Originally, citizen science members uploaded a photo of the road kill specimen with a scale reference directly onto the Taiwan Roadkill Observation Network (Reptile Roadkill Mortality) Facebook page. Pertinent information was included with photos as comments on the published post, including dates, times, species names (if known), and GPS coordinates. We used a web crawler via Facebook’s Graph API to find, retrieve, verify, and input road kill data into the database. The crawler collected posts (along with images, time and location data, and other annotations) on the Facebook Group and put them in an aggregated table of all observations. Participants were also encouraged to collect and send road kill specimens to TESRI.

Due to the unstructured format, contributors often did not upload posts with complete information, or did not post pertinent information uniformly, which made crawling for data more difficult. For example, users would often record the location inconsistently, such as decimal degrees or minutes, seconds, degrees format, qualitatively describe the location, post photos of nearby road-side utility poles with locational references, or take screen shots of a GPS app.

Soon after launching the web app, TaiRON also released an Android app (“Android App 1.0”) for members on the Android platform to use. Similarly to the TaiRON Web App 1.0, users could submit structured data through the Android app, which would publish uniform information directly through the contributor’s Facebook account. The data would be crawled and verified by project managers before being uploaded to the database.

There were several issues in this process. It appeared to encourage users to first publish onto the Facebook group (through Facebook, the TaiRON Web App 1.0, and the TaiRON Android App 1.0). We then crawled the group for data. User and protected species privacy were compromised; and an ever-updating Facebook APIs made application updates to keep up with Facebook necessary. Even though we were able to update the Android app in order to fix bugs that arose when each time Facebook changed its APIs, the Google Play Store’s one week lag-time in releasing the update was prohibitively slow. This prompted the project managers to develop methods to better control the content distributed on the Facebook group and to circumvent Facebook’s constant API updates.

Stage 2: Direct Uploading Observations to Project Website via Android App, and Posting Abridged Information to Facebook Afterwards

This process was in use from late 2015 to early 2017. The volunteers posted structured observation records first to the TaiRON project website. The website then posted abridged records to the Facebook group for user interactions and species identifications.

Due to the privacy issues and difficulty in keeping up with Facebook’s ever-changing interface, project managers worked with app developers to launch a new version of the free Android app (“Android App 2.0”) that members could download. With the TaiRON Android App 2.0, users were still able to log onto the app through their Facebook account, providing an easy transition for members. The main procedural change this app created was that the app allowed citizen scientists to upload their observations directly to the TaiRON website before the information was published in a uniform manner onto the Facebook group through the contributor’s Facebook account. This effectively resolved issues of gathering non-uniform data with a web crawler for database uploading and added an additional layer for information distribution control.
After users uploaded onto the TaiRON website through the app, only the uploader and project managers had access to all the data; the detailed locational data was not published to Facebook. There was also an added option of sending an anonymous observation digest to the Facebook Group. The observer's identity and observation location is kept at the TaiRON website and known by the researchers, but is not available to others at the Facebook Group. This provides some protection of the observer's locational privacy. When the observations are aggregated into datasets, we remove the observers' IDs in the records before the datasets are made available to others. On the other hand, as many participants have opted to release their observation photos under the Creative Commons Licenses, they must be properly attributed when the image files are released. Participants can have photos attributed to their nicknames, or require no attribution at all (by the use of the CC0 Public Domain Dedication). Additionally, this step allowed us to protect the locational data for endangered and protected species, which are not released to the public on the Facebook group or on the project website; users who wished to see this data must be approved by the project managers.

Continuing to publish the observations (sans detailed location) to the Facebook group allowed other members, including experts in the group, to have continued involvement in the verification of the data. After experts from the group verified the identification the observation, project managers would review the observation for accuracy and completeness and changed the status of the observation from “unconfirmed” to “confirmed” in the database. All verified data is available for the wider community to access through the website except that of protected species.

Though this flip in procedural steps smoothed the process of user data input, data mining/collection from Facebook, and privacy issues with public posts on Facebook, the app was not accessible to members using iOS platforms.

Stage 3: Direct Uploading Observations to Project Website via Web App, and Posting Abridged Information to Facebook Afterwards

This process is in use starting from early 2017. The volunteers also posted structured observation records first to the TaiRON project website. This is done by accessing directly to a page at website for uploading observation records. After uploading, the website abridged records to the Facebook group.

Because the Apple App Store was prohibitively difficult to publish an app in, and the Android app was often waylaid by the time it took to release new updates, the TaiRON project team has developed the a web-native method (“Web App 2.0”) that is accessible across all smartphone platforms and does not have to be published by a third party app store. Users access the web app through their smartphone’s web browser and TaiRON’s webpage (https://roadkill.tw/). Members sign onto the webpage with their Facebook account and upload through dedicated observation-reporting pages (https://roadkill.tw/app), which is accessible through smartphones, computers, and other mobile devices with Web browsers. Here they have the option to upload up to five photos of an individual, or up to five photos of separate individuals. Once the photos are uploaded, the web app detects date, time, and locational metadata associated with the photo and automatically fills corresponding sections of the form with the option to edit the information. Users can also indicate species, supplementary explanations, if they are sending a specimen, the quantity of individuals observed, and the supposed cause of death. After an observation is directly uploaded to the website by the user, using the Facebook API, the website then automatically publishes the observation sans locational data to Facebook through the group’s user account, “TW Roadkill.” The published Facebook post tags the contributing user to credit them with the observation. After posting in the group, project managers and other expert members can then review the observation for accuracy and completeness, and once verified, change the observation status to “confirmed.”

Currently, members can contribute observations to TaiRON’s database through several means, listed in decreasing order of ease of data collection method: 1) TaiRON Web App 2.0 accessed from the mobile website 2) directly posting on the Facebook group, and 3) directly messaging observations to a project manager. Less than 5% of observations are now contributed outside of using the Web App 2.0.