Optimal migration theory prescribes adaptive strategies of energy, time or mortality minimization. To test alternative hypotheses of energy and time minimization migration we used multisensory data loggers recording time-resolved flight activity and light for positioning by geolocation in a long-distance migratory shorebird, little ringed plover Charadrius dubius. We could reject the hypothesis of energy minimization based on a relationship between stopover duration and subsequent flight time as predicted for a time minimizer. We found seasonally diverging slopes between stopover and flight durations in relation to the progress (time) of migration, which follows for a time minimizing policy if resource gradients increase and decrease, respectively. Total flight duration did not differ significantly between autumn and spring migration, although spring migration was 6% shorter. Overall duration of autumn migration was longer than that in spring, mainly due to a mid-migration stop in most birds, when they likely initiated moult. Overall migration speed was not significantly different between autumn and spring. Migratory flights often occurred as runs of 2-7 nocturnal flights on adjacent days, which may be countering a time minimization strategy. Other factors may influence a preference for nocturnal migration, such as avoiding flight in turbulent conditions, heat stress, and diurnal predators.

The data were collected using custom designed micro data loggers (MDLs) that recored acceleration, light during specified periods, air pressure, and temperature. The loggers were deplyed on little ringed plovers Charadrisu dubois, during the breeding season (May and June) in southern Sweden, and retrived 1-2 years later. 

The MDLs were pre-programmed with a calendar defining when to run the light level measurements for position estimates. This approach is different from conventional geolocators in that our loggers only measured sequences of daily light cycles for a limited number of consecutive days. The reduction of measurement periods of light-level substantially reduces the amount of data to be stored and prolongs operation time by minimizing the power consumption. In our study, we ran six measurement periods each of 5 days distributed over one year, starting on 15 July, 1 September, 1 November, 1 February, 15 April, and 5 May. The timing of measurement sequences was selected to match periods of residence previously identified by conventional geolocators (Hedenström et al. 2013). 

We used the R software GeoLigth  for all steps in the analyses of light level data (Lisovski and Hahn 2012). To annotate twilight events we used the twilightCalc() function and a threshold value of 2 lux. Each twilight event was visually inspected before proceeding. Positions were translated from light measurements into geographical positions using the coord() function. Because light intensity was only measured during a few short periods, appropriate calibrations were not possible. Consequently, we set sun elevation angles to -6° to -5.5° for all but three loggers, which were set to -4°. Adjustments were done by visually inspecting the discrepancy in latitude estimates between the two measurement periods in November and February and for the positions derived in May, when we know that the birds had returned to the breeding sites, adjustment were made to yield positions corresponding to south Sweden. This approach gives a somewhat less certain estimate of wintering positions compared to geolocators recording light continually, but we obtain reasonable estimates of wintering areas and major stopovers provided limited data.

Total migration distances were estimated as the sum of great circle distances between well-defined consecutive stopovers and the breeding/wintering sites. To define stopovers we used light data during periods of residency as revealed by the MDLs (see above).

Acceleration in the Z-axis was sampled every 5 minutes with runs of 10 measurements at 5 time points, each separated by 5s. Each measurement is a sample during 100 ms at 100 Hz in the range ±4g. For each run, the mean of the values was subtracted from each of the 10 measurements to compensate for static gravity, and the recorded acceleration was considered as indicative of flight if at least 3 of the 10 values were greater than g/3, where g is acceleration due to gravity. Each 5-minute sample was assigned the number of runs that indicate flight behavior, i.e. (0, .., 5). Every hour a summary of results from all 12 runs were stored according to the distribution of the samples across the different activity categories (0, .., 5). If the bird is perched and motionless the data stored will be (12, 0, 0, 0, 0, 0), and if it is flying with continuous wing beats the data are (0, 0, 0, 0, 0, 12). To illustrate flight activity in graphical actograms the hourly recordings are coded as ‘black’ to represent continuous flapping flight (0, 0, 0, 0, 0, 12), and ‘white’ to represent no flight (12, 0, 0, 0, 0, 0), with shades of grey representing intermediate levels of activity (Hedenström et al. 2016; Bäckman et al. 2017).

To identify flight periods based on the accelerometer data we derived weighted hourly activity scores by calculating the sum of each score multiplied by the number of events within each hour. Thus, the lowest score becomes 0 (0 * 12) and the highest possible score is 60 (5 *12). We then identified all hours with a weighted score > 49, which we defined as an activity level corresponding to flapping flight. In almost all cases these hours were associated with sequences of high scores that could be identified as periods of flight. Start and end times of a flight were defined around the above defined flight period by subtracting all measurements (5 min periods) falling under a score of 3. Specifically, for a flight to be defined as ended zero-scores must be present during that hour. Thus, the starting and end points of flight periods, and hence flight duration, were calculated to a 5-minute resolution, provided that all scores < 3 and all scores >2 were recorded in sequence, respectively. Periods between well-defined flight periods were recorded as stopovers or winter/breeding site residency, depending on season.

In some cases, particularly at the end of flight periods, hourly weighted scores were below 50. In such cases we examined the distribution of the raw activity scores (0-5) and defined the hour as in flight as long as no zero-scores were recorded. If a zero-score was recorded, we looked at the next hour to asses if the bird had landed by summing the number of zeros between the two hours. If that sum was > 11 (corresponding to more than or equal to 1 hour) the bird had landed, if it did not then the bird had continued, and we considered the full sequence as a continuous flight. However, when calculating the duration of the flight all 5 min scores < 3 were omitted.

References

Bäckman, J., A. Andersson, T.  Alerstam, L. Pedersen, S. Sjöberg, K. Thorup, and A. Töttrup. 2017. Activity and migratory flights of individual free-flying songbirds throughout the annual cycle: method and first case study. Journal of Avian Biology 48:309-319.

Hedenström, A., R. H. Klaassen, and S. Åkesson. 2013. Migration of the little ringed plover Charadrius dubius breeding in south Sweden tracked by geolocators. Bird Study 60:466-474.

Hedenström, A., G. Norevik, K. Warfvinge, A. Andersson, J. Bäckman, and S. Åkesson. 2016. Annual 10-month aerial life-phase in the common swift Apus apus. Current Biology 26:3066-3070.

Lisovski, S. and S. Hahn. 2012. GeoLight - processing and analysing light-based geolocator data in R.  Methods in Ecology and Evolution 3:1055–1059.