Nesting attempts and success of Arctic-breeding geese can be derived with high precision from accelerometry and GPS-tracking

Abstract

Sensors, such as accelerometers, in tracking devices allow for detailed bio-logging to understand animal behaviour, even in remote places where direct observation is difficult. To study breeding in birds remotely, one needs to understand how to recognise a breeding event from tracking data, and ideally validate this by direct observation. We tagged 49 adult female pink-footed geese ( Anser brachyrhynchus ) with transmitter neckbands in Finland in spring of 2018 and 2019, and in Svalbard in summer 2018, and validated inferences from tracking by field observations of nesting sites and family status in 2018–2020 (54 spring–summer tracks). We estimated nesting locations by taking the median coordinates of GPS-fixes at which the goose was motionless (overall dynamic body acceleration, ODBA \u2009<\u2009 1) on days with a daily median ODBA \u2009<\u2009 1, which approached the real nesting locations closely (within 1.6–3.7\xa0m, n \u2009=\u2009 6). The start of nesting was defined as the first day on which the goose spent \u2009>\u2009 75% of time within 50\xa0m of the nest, because nest site attendances steeply increased within one day to above this threshold. Nesting duration (number of consecutive days with \u2009>\u2009 75% nest site attendance) ranged between 3 and 44\xa0days ( n \u2009=\u2009 28), but was 30–34\xa0days in confirmed successful nests ( n \u2009=\u20099). The prolonged nesting of 39–44\xa0days ( n \u2009=\u20093) suggested incubation on unhatchable egg(s). Nest losses before hatching time occurred mostly in day 3–10 and 23–29 of nesting, periods with an increased frequency of nest site recesses. As alternative method, allowing for non-simultaneous GPS and accelerometer data, we show that nesting days were classified with 98.6% success by two general characteristics of breeding: low body motion (daily median ODBA) and low geographic mobility (daily SD of latitude). Median coordinates on nesting days approached real nest sites closely (within 0.8–3.6\xa0m, n \u2009=\u2009 6). When considering only geographic mobility (allowing for GPS data only) nesting locations were similarly accurate, but some short\xa0nesting attempts were undetected and non-breeding tracks misclassified. We show that nesting attempts, as short as 3\xa0days, and nesting success can be detected remotely with good precision using GPS-tracking and accelerometry. Our method may be generalised to other (precocial) bird species with similar incubation behaviour.