Development of the printed circuit board (PCB), reading hardware and software
A PCB was designed to hold the required components and compose the circuit mechanisms of the collar. Because the PCB was intended to be installed on a collar, it had to be flexible. This was achieved by first forming a 4-layer PCB circuit board fabricated with polyimide substrates. Although the final flexibility was somewhat limited by the rigid components installed on board, we reserved ‘keepout’ areas without components to allow specific bends that would fit with the round shape of the collar.
The device was centered around a microcontroller, Texas Instruments MSP430FR2355, a real-time clock (RTC) AB0815 from Abracon with a quartz crystal reference, and an ambient light sensor LTR-308 ALS Lite-On corporation. Light intensities measured by the optic sensor were proportional to ambient light (luxes). The custom-made PCB with all components weighs 282 mg. When the two lightweight chemical batteries (330 mg each; Energizer® Zinc Air [ZN/O2]) were fixed on the PCB, the total mass was 942 mg. The board dimensions were 27.9 x 5.3 mm and the PCB was a flex board with a thickness of 200 µm. The RTC upkeeps the time and date while sustained by diminutive currents in the µA range. In addition, the microprocessor, a part of the microcontroller, also remained in a dormant mode to reduce battery consumption. General architecture design is given in Fig. 1A.
Exits and entries from the burrow were detected and monitored by keeping track of ambient light transitions. Abrupt changes in light intensity creates a signal that triggers further analysis by the microcontroller. After each light transition (e.g. from dark to light) and if the new state is maintained for at least 4 seconds, the transient amplitude (i.e. light intensity right after crossing a transient threshold), real time (from 00:00 to 23:59), and date of the event is stored in random access memory (RAM). Fleeting events that occurred within 4 seconds are ignored because they are generally assumed to be noise events such as passing under an object or in a small, illuminated portion of a tunnel. Very slow transitions are also ignored because they can be too easily triggered from changing weather (e.g. overcast vs. sunny, day to night or passing clouds, see algorithm in Fig. 1B). Figure 1C presents the external structure and shape of the PCB and completed collar.
When a change in light intensity crosses the amplitude transient threshold levels, the system logs the real-time clock data and transfers SRAM buffered data to the 32kB ferroelectric permanent memory. To avoid high power consumption from the permanent memory, it is only actuated when such transitions that last > 4 seconds occurs. A miniature magnetic switch mounted on the PCB bestows the possibility of in field activation with a simple 3 magnet swipes performed within 10 seconds. The redundancy affords the prevention of false activations and battery economy by avoiding actuation of the system prior to its final installed deployment time. The microcontrollers were programmed with a custom host firmware. This configuration allows parameter modification in the module via a RS-232 to USB terminal interface and a simple terminal software on a personal computer.
Assembling the collar
To assemble the collar, the PCB was first slid into a transparent 2-cm heat shrink sleeve. A tie-wrap was then slid under the PCB inside the heat shrink sleeve. Only then was the shrink heated with a heat gun, which fixed the PCB on the tie wrap. To keep away any water or humidity from the PCB, both ends of the reduced heat shrink were filled with acetic acid free silicon without touching the PCB itself. Due to the heatshrink sleeve covering the ambient light sensor, light intensities that are recorded do not represent direct sunlight, but the transparency of the sleeve allowed a reliable proxy.
Impact of collars on lemmings in captivity
Adult brown lemmings have a minimum weight of ~ 30 g, whereas collared lemmings start at ~ 40 g [20, 21]. The mass of the collar (1.59g) was ≤ 5% of the body mass of adult lemmings. Keeping tracking devices below a 5% threshold is recommended , but could still negatively impact behaviors and vital rates (see  for a review; , ). We evaluated how collars impacted lemmings by comparing body mass changes, a proxy of body condition, and recapture rates, a proxy of survival or behavioral alteration, between lemmings with and without collars. In November 2019, 4 brown lemmings (Lemmus trimucronatus) and 2 collared lemmings (Dicrostonyx groenlandicus) were held in captivity in Cambridge Bay, Nunavut, Canada. They were provided ad libitum food and water before and during the experiments. For more details about how the lemmings were live-trapped in the field, for the housing conditions and care given to the captive lemmings, see .
The experiment consisted of all lemmings being monitored daily without a collar for several days (lemmings were monitored since August 2019 after their initial capture ), and then equipped with a 1.5 g dummy collar between 24h and 108h (i.e. a tie-wrap with a mass fixed by a heat shrink). Each individual was kept under observation for the first 15 minutes and then checked every 2 hours for the first 8 hours, then every 12 hours, to ensure the collars were not causing drastic changes in behavior (e.g. constantly scratching or trying to take off the collar) or choking. The body mass of each lemming was monitored with an electronic scale (± 0.01 g) every day to every week before the collar was installed on it. Once the collar was fit on the lemming, the body mass was measured every day. To determine if collars had an impact on the body conditions of lemmings, we compared the daily mass change of equipped lemmings to their daily mass change before they had the collars on. Two different (non-overlapping) pre-experimental periods of 12 or 8 days were chosen as controls, because lemmings either continuously gained or had a stable mass during these periods (Fig. 2A). We performed a one-sided t-test, weighting for the duration of the monitoring in each period, to test the hypothesis that equipped individuals had a lesser daily mass gain than when unequipped.
Impact of collars on lemmings in the field
We deployed light-sensitive collars on small mammals in three locations of the Canadian Arctic where populations are monitored every year and assessed the impact of collars on body condition and recapture probability. Rodents fitted with collars were brown lemmings (n = 5 & 36) and northern collared lemmings (Dicrostonyx hudsonius, n = 11 & 0) in respectively in Cambridge Bay and Bylot Island, Nunavut, whereas Ungava collared lemmings (n = 6), an Eastern meadow vole (Microtus pennsylvanicus) and a Northern Bog Lemming (Synaptomys borealis) were fitted with a collar in Salluit, Quebec. All rodents were monitored at these sites with live-trapping and capture-mark-recapture methods as part of multi-annual surveys. At all sites, trapping grids made of 96 to 144 live-trapping stations, each station being separated by 30 m, and arranged according to a cartesian plane were used (Bylot Island: 3 grids; Cambridge Bay: 4 grids Salluit: 2 grids). Longworth and Little Critter traps were used at all these sites. Capture-mark-recapture methods consisted of opening and baiting traps followed by visits of traps every 12 hours until 6 visits were completed. All lemmings captured were marked with a PIT- or ear-tag, weighed and sexed. During live-trapping, adult lemmings with a minimum body mass of 34 g (to ensure that collars accounted ≤ 5% of the total body mass) were fitted with collars. The total number of collars deployed at each site differed due to low lemming densities in both Cambridge Bay and Salluit (< 1 ha− 1), while lemming densities were high on Bylot Island (15 ha− 1; unpublished data). All manipulations were approved by the Animal Care Committees of the Canadian Museum of Nature (2018.02.001) and Université Laval (2019 − 253, VRR-18-050), Parks Canada (SIR-2021-39399), Department of Environment of Nunavut (WL2019-038), Kitikmeot Inuit Association (KTX119N006), and Ministère des Forêts, de la Faune et des Parcs du Québec (SEG 2021-05-31-125-10-S-F).
Recapture probabilities of individuals with and without collars were calculated for each trapping grid. Here, recapture probabilities were calculated as the total number of recaptures across all individuals divided by the total number of captures (i.e. sum of first captures and recaptures). To test if recapture probabilities of equipped individuals were lower than those of unequipped individuals, we used a one-sided t-test with weighted observations to account for the number of deployed collars per grid. This was done to reduce the influence of grids with low sample size on the statistical test.
Using exclusively the data of Bylot Island, where a peak lemming abundance yielded many more captures than at the other sites, we also evaluated the difference in daily mass change between equipped and unequipped lemmings. The relative daily mass changes and 95% confidence intervals (CI) of each group were weighted for the time between captures. We used a one-sided t-test weighted for the time between captures to evaluate if daily mass changes of equipped individuals were lesser than those of equipped individuals.