The use of small Unmanned Aircraft Systems (UAS; also known as "drones") for professional and personal-leisure use is increasing enormously. UAS operate at low altitudes (<500 m) and in any terrain, thus they are susceptible to interact with local fauna, generating a new type of anthropogenic disturbance that has not been systematically evaluated. To address this gap, we performed a review of the existent literature about animals' responses to UAS flights and conducted a pooled analysis of the data to determine the probability and intensity of the disturbance, and to identify the factors influencing animals' reactions towards the small aircraft. We found that wildlife reactions depended on both the UAS attributes (flight pattern, engine type and size of aircraft) and the characteristics of animals themselves (type of animal, life-history stage and level of aggregation). Target-oriented flight patterns, larger UAS sizes, and fuel-powered (noisier) engines evoked the strongest reactions in wildlife. Animals during the non-breeding period and in large groups were more likely to show behavioral reactions to UAS, and birds are more prone to react than other taxa. We discuss the implications of these results in the context of wildlife disturbance and suggest guidelines for conservationists, users and manufacturers to minimize the impact of UAS. In addition, we propose that the legal framework needs to be adapted so that appropriate actions can be undertaken when wildlife is negatively affected by these emergent practices.

ABSTRACT:To get around UAS limitations and propose a viable solution for wildlife monitoring, the development of new inventory methods is needed. However, most authors use the classic systematic transect method as data processing and statistics are easier. We thus created an application to process data from every type of flight plan and to help detect and compare observations on large datasets. WiMUAS is a small software compatible with the open-source QGIS© that allows the creation of visual maps compatible with geographical information systems based on telemetry data and payload parameters to estimate the covered area. The application also has a slider for animal detection that allows multiple observers to record and compare their results for accurate counts. We then tested it on data from a trial realized on savannah animal populations in Democratic Republic of Congo using the Falcon UAS. We created a new type of flight plan, a rosette-shaped design that can be covered in three flights,.and repeated it twice. More than 5000 images were collected during the six flights. Image projection gives an area of 12,4 km 2 for the first trial and of 12,1 km 2 for the second. The mean sampling rate for both test is 6,1 %. Observers spotted buffaloes, hippos, warthogs and various antelopes with different success over an average rate of 8 images reviewed per minute. Resulting densities observed by the three observers are similar for each test (coefficient of variation 6,9 and 8,6 % respectively) but mean densities vary a lot between the two trials (23,8 and 6,5 animals/km2 respectively).