Development of a novel physiology tag to measure oxygen consumption in free-ranging seabirds: research

Discussion

This study demonstrates the first use of a non-invasive NIRS sensor to investigate muscle oxygen saturation in free ranging European shags. Further, the SmO₂ measurements in the current study provide the first non-invasive muscle oxygen measurements in a free-ranging wild animal. Estimates of VO₂ derived from these are similar to the published range of empirical measures of VO₂ in similar seabird species.

The NIRS system we developed was an adapted off-the-shelf muscle oxygenation sensor (Moxy Monitor). This was selected as it is relatively low-cost, is relatively small and lightweight, and is fully autonomous. The Moxy systems were successfully adapted for the marine environment and to ensure it was less than 3% of the instrumented animals' body mass. From a practical perspective, the loggers required a series of technical modifications which, although relatively minor, required specialised engineering facilities. In particular, a skilled electronic engineer was required to remove the LEO and RPDs from the PCB after having been attached via flow soldering, and to reattach these to the extension leads. Further, the application of the waterproof coating and subsequent waterproof tests also required suitable testing facilities, specifically a shallow (0 to 2 bar) pressure-chamber.

The deployment of the dual tag system (NIRS and GPS/accelerometer) on shags was similar in scope to the attachment of other more commonly used tags such as GPS, with the exception of the time taken between capture and release of the birds which was substantially longer (around 25 minutes compared to the typical 10 minutes for the deployment of two loggers). The additional time was required to attach the three NIRS sensors securely to ensure skin contact between the feathers, which was a delicate procedure. The birds behaved normally at release after tag deployment, returning to the nest within a few minutes. The behaviour of the birds over the next ca. 24 hours (as recorded by GPS and accelerometers) was markedly different, with one bird foraging more extensively at a further distance from the colony, and one foraging less extensively close to the colony. However, these patterns are within the normal range of behaviours of breeding shags, with time activity budgets comparable to those recorded in past work (Wanless et al. 2005) and locations within the typical foraging range of this population during the breeding season (Bogdanova et al. 2014). At tag retrieval, the birds both left the nest as the person approached. They returned to the nest after a few minutes and were captured successfully. A small proportion of birds that have been caught previously respond in this way when they first observe someone approaching the nest. However, there are two mitigating circumstances that may explain why we saw this response in both birds. First, the birds were comparatively late breeders and this subset of the population tends to be more susceptible to disturbance; we had to work on late breeders because of the delay in logger development. Second, they were nesting in locations that were quite distant from the shoreline, and shags are typically more nervous in such circumstances. We had to select birds nesting in these locations because late in the season there is a much smaller pool of birds available for capture. Despite these potential contributory factors, we can't discount the possibility that the short-term adverse effect that we observed was a direct result of this particular deployment. Crucially, however, we have no evidence that these effects were long-term; the birds returned to the nest within a few minutes of release after the loggers had been removed, attended the nest thereafter and fledged their chicks successfully.

In terms of the suitability of the system we developed here for deployment on other species, there are a number of important considerations related to their morphology, behaviour and ecology. The tag weight and shape should be developed taking into account aspects such as body mass, flight type and foraging strategy of the target species (Table 1), to ensure the tags don't add excessive weight and/or substantially increase energetic costs associated with flight and foraging. Particular care should be taken with species such as auks that naturally incur high flight and foraging costs. The overall weight (in air) of the system we used in this project was ~26g; although this would be suitable for deployment on shags and other large species, it is excessive for smaller species of interest such as black-legged kittiwake, common guillemot, razorbill and Atlantic puffin. Following the widely used rule that devices should not exceed 3% of the birds' body weight (Kenward 2001, but see also Vandenabeele et al. 2012), the weight of tags deployed on all species in Table 1 except great black-backed gull and gannet would need to be below 26 g (range 9-25 g; calculation made using the lower weight limit presented). Furthermore, the shape of the device ideally needs to be as streamlined as possible, particularly for diving species, to limit the increased drag (Vandenabeele et al. 2012). Another consideration is the probability of recapturing the birds to retrieve the devices. Although it was possible to attach the NIRS tags and recover them with limited removal of feathers, they did require the attachment of sensors to the down feathers of the birds and this may increase thermal sensitivity. It is, therefore, important to consider the potential consequences of tag deployment in situations where recapture is not successful, and the tag remains attached to the birds for longer than initially intended. Seabird recapture rates are generally high but can vary among and within colonies and years depending on factors such as environmental conditions, locations of nest sites and timing of breeding.

Although the NIRS system did collect biologically relevant SmO₂ measurements, the duration of viable data was relatively short (~15 mins). We assume that after this period, the data appear to have been contaminated by LEO/RPD movements, or inadequate contact between LEO/RPD and the bird's skin. A clear avenue for further research is, therefore, the development of a non-invasive mechanical attachment system to maintain positive downward pressure of optics. Nevertheless, the results appear promising and support that NIRS can be a viable energetics tool for avian research. Further, the initial pilot data has proven that NIRS has the capacity to operate successfully on a free-ranging animal and can provide measures of SmO₂ in birds. However, due to the relatively short deployment duration, it is difficult to formally assess the capabilities of the system in a range of activity states as only resting/sedentary states were captured in the usable data.

SmO₂ depletion has been shown to provide an accurate estimate of total body oxygen consumption (Crum et al. 2017). This is based on the rationale that during exercise locomotor muscles are the predominant consumer of body oxygen stores. During resting periods, SmO₂ is high (<90%) as the muscles are inactive and baseline metabolic costs are the major contributor to VO₂. During physical exertion, where locomotor muscles are the dominant consumer, the magnitude and rate of SmO₂ depletion correlate and provide an estimate of the rate of total body oxygen consumption (VO₂). Using published values for SmO₂ depletion (as measured with a Moxy Monitor in humans) to estimate total body oxygen depletion, median baseline (inactive) VO₂ in this study was estimated to be 10.9 (95% CIs = 8.9-20.0) ml.min.kg⁻¹ (Figure 5). This is similar to the published range of empirical measures of resting VO₂ in similar seabird species. For example, White, Martin and Butler (2008) measured resting VO₂ rates of 30.9 ± 1.5 ml min^-1 (which equates to 13.4 ml.min.kg⁻¹) and Hayama and Yamamoto (2011) measured resting VO₂ rates of between 10.5 and 16.9 ml.min.kg⁻¹ in great cormorants (Phalacrocorax carbo). Similarly, Schmid, Gremillet and Culik (1995) measured VO₂ values of 11.5 ml.min.kg⁻¹ (calculated from the allometric relationship VO₂ =0.691 M^0.755 where VO₂ is measured in l.h^-1 and M is mass in kg) in great cormorants standing on land. Unfortunately, a lack of reliable SmO₂ measurements during other behaviours (flight and diving) precluded comparisons of the SmO₂ values in the current study to published empirical estimates of VO₂ during these behaviours'.

From an applied perspective, changes in the behaviour or activity of birds in response to anthropogenic activities or structures may have significant energetic consequences for individuals (Masden et al. 2010). However, information on the true energetic consequences of behavioural responses or displacement to seabirds has been limited due, in part, to a lack of suitable technology for measuring energy expenditure in free-ranging seabirds at an appropriately high resolution.

A number of techniques can be used to measure energy expenditure in wild birds but are either at insufficient temporal resolutions (e.g. doubly-labelled water techniques: Gabrielsen, Mehlum & Nagy 1987; Weimerskirch et al. 2003) or requires species-specific calibration in controlled laboratory conditions (e.g. heart rate measurements: Butler et al. 2004). More recently, the rapid development of high-resolution accelerometers has led to Overall Dynamic Body Acceleration (ODBA) becoming a routinely used proxy for energy use. This provides a useful high-resolution measure; however, a key issue with ODBA estimates of V̇O₂ is the lack of any physiological (cardiovascular) input. This has the potential consequence that, during high exertion or diving, reliance on muscle anaerobiosis can make the same biomechanics exponentially more metabolically costly. In other words, if birds were required to operate anaerobically, even briefly, ODBA would fail to capture these energetic costs. For example, it has been shown that biomechanics in isolation is a poor predictor of metabolic acidosis in penguins (Williams, Meir & Ponganis 2011). This is because when muscle switches to anaerobic metabolism there is a significant reduction in the rate of oxygen depletion in the muscle tissue as the oxidative pathway for ATP production is abandoned. However, measuring muscle metabolic dynamics such as those measured in the current study (e.g. SmO₂) provide a direct measure of muscle VO_2, and can detect muscle anaerobiosis. This is, therefore, robust to anaerobic metabolism and its associated lower rate of recovery (greater metabolic cost), providing a more direct and realistic measure of energetic expenditure in birds. While ODBA is still the most frequently used method of estimating energy expenditure in free-ranging birds, the capacity of NIRS to directly measure tissue-specific oxygenation in real-time, and to capture metabolic changes which ODBA fails to capture, could make NIRS a potentially more reliable and biologically robust method for measuring V̇O₂.

Future development

This study confirms that an animal-borne NIRS logger can directly measure energy expenditure in seabirds with a high temporal resolution. Despite the deployment of the system providing only SmO₂ measurements over a relatively short period, there were a number of important, novel and ultimately successful steps achieved here. Ultimately, to ensure that this technology can be utilised practically as a tool for reliably measuring energy expenditure over periods of hours or days, a series of required refinements and developments have arisen from the work thus far.

1) A key area of refinement is the stable attachment of the LEO/RPD to the skin of the birds. In the current study, the LEOs and RPD were independently glued to the down plumage. It seems likely that this provided insufficient contact between the optical hardware and the skin for periods more than several tens of minutes. The development of a mechanical attachment system to maintain positive downward pressure of optics is essential before further deployments. Further, deployments on birds in a more controlled environment (e.g. a captive facility), would be an ideal paradigm to develop attachment mechanisms and to diagnose any potential issues.

2) In its current format the NIRS hardware (Figures 2 and 3) was undesirably configured and housed for free-ranging birds. Both the shape and currently compartmentalised design make the system bulky and resulted in poor fluid-dynamic performance. Effort should be made to create a custom light-weight housing for the components (main PCB and batteries).

3) The NIRS system used in the current study does not provide raw voltage data which would provide a measure of the quality of the NIRS optical signal. Other NIRS systems, provide raw data from which ambient light leak and total photon recovery rate can be interrogated; however, these are larger and significantly more expensive. These two factors provide a direct measure as to whether there is sufficient contact between the skin and optical hardware and would ultimately provide a measure of the reliability of the measured data. Raw HbO₂ and HHb data would also provide more detailed information on tissue specific perfusion. While in their current form such systems would appear to be too bulky for deployment on seabirds, development could generate a suitably sized and weighted tag.