Data Management and Storage
eMammal has a 4-part data management workflow:
1) Upload via the Desktop App -> 2) Expert Review -> 3) Data Storage -> 4) Website Analysis
Figure 1: eMammal Data Workflow. Solid arrows show the flow of data through the system and dotted arrows show data shared via automatic Application Programming Interfaces (APIs). (Note: Our teams are currently working together to move our storage from Amazon Web Services to Microsoft Azure)
1. Upload via the Desktop App -
We have created desktop software to help people view and tag camera trap pictures. This is available for Macs or PC's, and runs on the Adobe Air platform. The software will work offline, but the user has to first initialize it online to download all their project information. The user can then identify pictures off-line, and upload the pictures and metadata at a later time.
This program automatically groups bursts of photos into sequences if they are <1 min apart, making it easy to look through a data card full of images. Camera trappers use this software to examine each sequence and tag which species, and how many individuals, are found. The species list for each project is based on the mammals and larger birds found in the study area, and is customizable by each project manager. Sequences without animals are tagged as 'No Animal', while those the camera trapper isn't sure of are tagged 'Unknown Animal'. Humans are also tagged as 'Camera Trapper' when setting up and taking down the camera trap, 'Human non-staff', 'Bicycle', or 'Vehicle'.
Once a sequence is tagged it can be uploaded for review by the project manager.
For more details see the help video in the How To page.
2. Expert Review Tool -
eMammal requires that all pictures be reviewed for quality control. We have an online Expert Review Tool (ERT) to make this an efficient process and allows experts to review photos using any computer with an internet connection. Images and data uploaded from the desktop software are stored temporarily on Amazon web servers. Only the project managers have access to these pictures, and they have the responsibility to review them before submitting them for long-term storage at the Smithsonian Digital Repository.
There are two basic aspects of the review process. First, the expert needs to determine if the camera was set in an appropriate way, at the right height, and aimed appropriately. Cameras aimed too high will shoot over the heads of smaller animals and those aimed too low will survey a very small area. These inappropriate sets should be rejected outright.
Second, the expert needs to review the species identifications and animal counts for each sequence. This can be done either by looking at a thumbnail view of each sequence, or by clicking through each frame for a closer look. All frames tagged as 'unknown animals' or ‘no animal’ should be examined carefully. Each study will probably have a few species that camera trappers often misidentify (e.g. red fox vs. gray fox) that should be more carefully examined by the expert. Once all sequences in a deployment are approved they are sent to the Smithsonian Digital Repository and deleted from the Amazon server.
When reviewing pictures it is a good strategy for experts to not approve species identifications they aren't yet 100% sure on. It is more difficult to change ID's later in the process. We recommend leaving the 1st sequence of a deployment (usually the camera trapper setting up the camera) unapproved until all other sequences are approved (this prevents accidental full approval before you are ready). Caution in reviewing is especially important for new study areas where confusing species take a while to distinguish, but become clearer after a larger collection of camera trap images is available later in the study. Reviewers may want to wait to approve all sequences within a deployment until several deployments can be viewed.
Each project is responsible for reviewing their own pictures. This could be done by the project manager, or by staff, students or volunteers trained and supervised by the project manager. New project managers should be extra cautious in approving difficult sequences, checking with their supervisor until they have proven their qualifications as an 'expert' in the wildlife of that particular area.
(Note: Our teams are currently working together to move our storage to Microsoft Azure)
3. Smithsonian Data Repository -
After being approved by project staff all pictures and data are stored in the Smithsonian Data Repository. This database provides free long-term storage.
Pictures and data must eventually be available to the public, but can initially be placed under an embargo. Smithsonian offers data embargoes of 3 years for projects that do not want to immediately share their data so that they have the first opportunity to publish their results. These terms will be set upon the creation of a project, and can be changed later if needed.
In addition, the species location is masked for photographs of threatened and endangered species. The repository contains the original location of all photographs, but each project has a center location identified by the Project Manager. All photographs of endangered species (and all other photographs from that location) are associated with this central location in all publically available data.
The eMammal database will store pictures of people collected by camera traps but will not share these images with the public. Requests to view people photographs for research or security purposes should be sent to project managers or eMammal staff. Additionally, we will be blurring the photos of people to be in compliance with privacy requirements set by the Smithsonian.
4. Data Analysis
Check out the Browse Data tab at the top for a list of analyses you can run on our website!
Costs -
eMammal is supported by the Smithsonian and external grants, ensuring long-term stability of the project and storage of images and data. However, there are costs associated with using cloud-based image processing that we cannot avoid. Therefore, we pass these costs on to project managers of eMammal. See the costs page for details.
