Dust bathing events obtained from video-recordings as well as acceleration vectors recorded with accelerometer obtained in Japanese quail

Annotated base constructed to validate the methodological approach for detecting dustbathing events presented in e publication "Thanks to repetition, dustbathing detection can be automated combining accelerometry and wavelet analysis" (Fonseca et al., 2024. Ethology). For this experiment was performed where the behaviour of Japanese quail (Coturnix japonica), equipped with accelerometers, was captured through video recording within their home boxes. Experimental disign is presented in detail in Fonseca et al [1]. A total of 13 pairs of male/female quail were studied, named boxes 1-13, as stated in file names. The week before the experiment, a backpack was fitted onto the male, to promote habituation to it (Pellegriniet al., 2019). The backpack was 3D-printed in black plastic and had two elastic bands on their sides to be passed around the base of the animal's wings. On the morning of the experiment (between 9-11 am) the TechnoSmArt@ accelerometer-loggers Axyz was inserted into the backpack using a specially designed applicator to ensure synchronization between the acceleration time series and the two video recordings (top and side cameras). Tri-axial accelerometers were set to gather data with a sampling frequency of 25 Hz (i.e. 25 data points per second) based on previous studies that show the possibility of high-speed transitions between behaviours in this species [2]. A wire bar wall partition was inserted in the middle of the home box, dividing it into two separate, equal-sized, compartments. After the placement of accelerometers in the backpack, males were then positioned in one of these two home box compartments. Ten minutes later, the female companion bird was placed into the adjacent compartment. After a second 10-minute period, the wall partition was lifted, allowing the birds to interact. The respective video recordings began before the placement of the accelerometer and ended after the accelerometer was removed. Testing lasted 6h, thus given the 25Hz sampling frequency of the accelerometer, for each male at least 540,000 time points were obtained for the three axial compoenets (x, y and z) of the acceleration vector, corresponding to dataset columns ax,ay and az, respectively. For the second aim of this study, the last group (box 13) was recorded over a week-long period to demonstrate the methodology´s potential for studying dustbathing dynamics, such as daily rhythms, in long-term studies.Front and side video recordings were first scanned by an observer trained to detect dustbathing events. By strict definition, dustbathing can be characterized as a precise and orderly sequence of movements consisting of (a) pecking alternately from side to side with a closed beak; (b) scratching (one foot at a time) while sitting or squatting; (c) tossing the dust with the wings and undulating the body underneath the dust shower; and (d) occasionally rubbing head and/or body parts in the dust. Movements (b) and (c), and sometimes (a) and (d), are repeated a variable number of times. Since the initial pecking and scratching also appear in different behavioural contexts, the dust toss and body roll (undulation) were considered firm indicators of dustbathing. The time of beginning and end of each dustbathing event were determined from video recordings. Pauses longer than 5s, movement away from the site of dustbathing (usually preceded by standing and shaking), and/or the performance of another behaviour, marked the end of each dustbathing event. The annotated database was constructed by assigning the correspondence between accelerometer data and video recordings, with each time point (i.e.column labeled time, expresend in seconds) from the accelerometer annotated with a label indicating whether the bird had been observed dustbathing. Labels for the dustbathing column are indicated with a 0 if the behavior is not being performed and a 1 if behavior is being performed at a given time point.Maximum power of the y-axes component of the acceleration vector (ay) was estimated using the complex Morlet mother wavelet as descibed in [1]. Code provided is [3,4]. The squared absolute value of the complex wavelet coefficient is the power. When depicted in a magnitude scalogram the resulting plot is called a spectrogram (Flesia et al., 2022). The maximum value of power estimated within the range of scales 25 – 60 s is presented in the dataset as power spectrum coefficient (PSC) for the three axes, columns PSC_x, PSC_y and PSC_z.