Semi-Automated Counts on Drone Imagery of Breeding Seabirds Using Free Accessible Software

Abstract

Long-term monitoring of breeding seabirds is fundamental for assessing the conservation status of their populations. Whereas traditional monitoring is often time consuming and has disadvantages, such as observer bias or disturbance to the breeding grounds, the use of uncrewed aerial vehicles (UAVs or drones) has proven to be an efficient alternative by allowing non-invasive monitoring of inaccessible areas. Nonetheless, the use of drones for monitoring wild populations brings forth a new challenge, namely the handling of large amounts of data (images), usually negating the efficiency of the previous steps. Diverse methodologies have been developed to deal with this issue, but they usually involve the use of commercial software, that reduces the accessibility of users with limited resources. We tested if the popular free software ImageJ could compete in terms of efficiency (i.e. accuracy and processing time) with other commercial software. We obtained similar values of agreement between manual and semiautomated total counts of individuals (99.1%), reducing the analysis duration fivefold. In addition, we propose a correction factor in the detection of incubating individuals based on the assessment of the individual behaviour of 10% of the birds present in each colony. Following this correction, we were able to estimate the total number of incubating birds with a 103.5% agreement with manual counts, reducing the time invested up to threefold. Thus, we show support for the use of free software (ImageJ) as a good low-cost alternative for users of drone imagery in assessing breeding birds and as a conservation tool.