Data from: Selective foraging behaviour of seabirds in small-scale slicks

Data from: Selective foraging behaviour of seabirds in small-scale slicks by Lieber et al (2022) contains the HMM input file (.csv) and R analysis code (.txt) and the metadata (.pdf). <br> All data was collected in the Narrows tidal channel, Strangford Lough, Northern Ireland, UK. The field survey data collection was performed using Uncrewed Aerial Vehicle (UAV) surveys to capture seabird (tern species) foraging over slick line manifestations in this tidal channel. UAV missions using the DJIMini2 were conducted on 02/08/2021 during flood tide and low wind speeds (2.1 ms-1). Two UAV hovers (h) were performed at a 120 m altitude with a vertically downward-facing camera and aligned with the mean direction of the flow. All missions were completed in accordance with local regulations and flown by the same qualified (UK Civil Aviation Authority) pilot. The UAV camera was calibrated in the laboratory using a standard checkerboard method. Automatic video analysis techniques were used to track terns foraging over evolving slicks. This resulted in a dataset comprising of a timestep (t), an associated x and y variable for each tern’s position at a given timestep, a calculation of speed, acceleration magnitude, and tortuosity. Further, slick positions were extracted for each instantaneous position along each tern track using three-dimensional interpolation in space and time. The slick parameter was matched to tern tracks where every track position had a binary variable (‘slick=1’ or ‘no slick=0’), along with an associated continuous variable ‘time-to-slick’ (in seconds, hereafter slick proximity), calculated from the timestep a tern was spatially over a slick until it next encountered a slick. All track positions were corrected for camera lens distortion and scaled according to the UAV’s altitude. Apart from the extracted binary slick variable (above), tortuosity was used in the modelling of the data set. For this, the instantaneous tortuosity along each track was calculated using an 11 element window (±5 frames, centred on each position), where the raw positions within this window were smoothed by fitting a cubic spline. This represents a low-pass filtering operation with a cut-off frequency at 2.73 Hz. The tortuosity was calculated as the total distance travelled (sum of the distances between the 11 points) divided by straight-line distance between the first and last position. For the modelling, the log of tortuosity (log(tortuosity) was computed and used as a variable in the hidden Markov model.