This week, we take a look at a fascinating study on a migratory bird and how the authors teased out potential drivers of population trends. We also see an interesting experimental study about our tendency to redistribute wealth and our capacity for tolerating inequality. Finally, we look at three methods papers – one that tackles pseudoreplication, and two that address remote sensing.
Birds of a feather don’t necessarily flock together
In a remarkable study reported (with commentary) in PNAS, we learned this week about “migratory connectivity” – a strong association between geographically-distinct breeding and nonbreeding areas. Kramer et al. (2018) determine the cause of population declines in one distinct breeding population of golden-winged warblers, by bringing together several lines of evidence. They used migratory data (70 individuals with geolocators), long-term (since 1966) citizen science population trend data, and land cover data for nonbreeding sites for two congeners, blue-winged and golden-winged warblers. These species are ecologically similar and migrate between approximately similar regions. This setup provided a good phylogenetic control. The breeding population of golden-winged warblers from the Appalachian Region appears to be in decline, while the other breeding populations of this species and breeding populations of the blue-winged warbler appear relatively stable. The blue-winged warblers exhibit weak migratory connectivity, with somewhat distinct breeding populations mixing in their nonbreeding area. The golden-winged warblers, by contrast, exhibit strong migratory connectivity with distinct breeding populations using distinct nonbreeding areas. The only declining population, from the Appalachian breeding site, was also the only population of either species to over-winter in northern South America, where it appears that habitat degradation (forest loss) may be linked to the population decline.
Ed: This study is quite unique in its breadth, depth, and use of phylogenetic controls. Deploying enough geolocators and recapturing 70 of the tagged birds after a year is phenomenal and provided a rich dataset on migratory movements in these two species. Having Breeding Bird Survey data going back to 1966 also provided a clear picture of population trends for both species, which, in turn, represented a neat phylogenetic control.
A little bit of Robin Hood doesn’t go enough of the way
Why do we struggle with inequality? Bechtel et al. (2018) try to answer that question using an experimental game played by a large number of German and United States study subjects (nearly 5000, in total). All participants were placed in random “pairings” whereby both participants recieved a “payout”. The pair partners were unknown to each other. The distribution of the winnings were such that the participant won either USD 25 (versus USD 75, i.e. disadvantageous inequality within the pair), USD 50 (versus USD 50, i.e. equality within the pair), or USD 75 (versus USD 25, i.e. advantageous inequality). Each participant was allowed to give to or take from their partner any sum they pleased. In general, participants tried to redistribute the winnings more equitably when they were in either inequality pairing: participants who received USD 25 took money from their USD 75 partners, and USD 75 participants gave money to their USD 25 partners. Even though the difference in winnings in the inequality pairings was USD 50, participants only gave/took USD 10 to/from their partners (on average). This would suggest that people try to apply some redistribution of wealth, but are generally quite comfortable with retaining relatively large inequalities – even when the starting point for both partners in a pair was essentially a luck-based lottery winning.
Ed: Even though the dollar values in this study were somewhat trivial, this study still gives important insight into human behaviour with regard to tolerance of inequality and redistribution of wealth. Some of the study strengths include a very large sample of participants (who were representative of the adult voting public), and a “survey instrument” that allowed participants to both donate to and take from their partners.
Crying fowl over pseudoreplication
Individual variation in physiological responses to stressors, and individual variation in metabolism of circulating stress hormones are both relatively under-appreciated aspects of non-invasive hormone monitoring. Coppes et al. (2018) address this issue in an extensive study of faecal glucocorticoid metabolites in the capercaillie in Germany. The authors modeled (as the “full” model) the variability in the faecal glucocorticoids using several environmental parameters and animal sex (as fixed effects), and controlled for the individual variability using a random effect for animal ID. The authors then compared this model to a “naive” model that assumed each sample was independent and thus allowed for pseudoreplication (treating non-independent samples from the same animal as independent samples). The full and naive models differed in which parameters were considered significant and differed in how much variance was explained.
Ed: Although treating repeated measures from individuals is an important step in analysis, as the authors highlight, the broader issue that this study exposes, is the problem of pseudoreplication. By removing the random effect for individual ID in the “naive” model, the authors were actually incorrectly increasing the effective sample size by assuming that the multiple samples from a given individual all contained unique information. Naturally, this would change how much variance could be explained by the model, and it could allow additional variables to be selected as significant, simply by inflating the sample size. The important take-home message here might be that researchers need to think clearly about pseudoreplication during study design (incorporating sufficient spatial or temporal separation in sampling to reduce pseudoreplication) and during statistical analysis (for example by using a random effect in a random effects or hierarchical model).
Remote sensing offers the promise of being able to monitor biological populations in areas that are difficult or dangerous to reach, with ease, speed, and costs that might outperform traditional human observers. However, to date, remote sensing has mostly been used successfully when counting large animals across a relatively small spatial scale when the background substrate is relatively homogeneous. As this review in Methods in Ecology and Evolution shows, there are still a number of obstacles to overcome before remote sensing can play a larger role in monitoring biological populations. Some of these obstacles include the cost of high-resolution data, and the ability to automate the organism detection and counting process, which is currently still subject to error.
Ed: This is a field ripe for further development and improvement, both in counting animal populations and plant populations (not addressed in the review).
Drones for clones
As an example of one of the scenarios where remote sensing may prove helpful in counting animal populations, Hodgson et al. (2018) performed rigorous ground-truthing for drone-based remote-sensing of a seabird colony. The authors created replicas of seabirds of known number so that they could compare the accuracy of ground-based counts by human observers with counts based on drone imagery processed manually or with a semi-automated process. The drone-based counts were more accurate than the ground-based counts.
Ed: This study demonstrates an important aspect to any remote sensing study – the need for rigorous ground-truthing. The drone method appeared to work well in this contrived example, but, it should be noted that the background substrate – beach sand – was homogeneous and the replica birds on nests represented large detection targets. If a drone-based count is going to work for any animal population – this would probably be it.
Every dog has its … picture painted
This week’s conservation art profile:
Depicting 4454 wild dogs all individually hand painted. Caitlin O’Gorman’s representation of what is left in the wild. Oil on canvas. 2m x 1.8m
Caitlin grew up mostly in Ireland, and completed her final years of schooling in KwaZulu-Natal. She has sketched and painted for as long as she can remember. Our lives are full of accidental stories which take us in unexpected directions. One of the happiest accidents in Caitlin’s life was discovering the Michaelis School of Fine Art at the University of Cape Town. In 2018, she is in her fourth and final year as an undergraduate, and a key component of the year is a display of what drives and inspires the people in her class. For Caitlin it is conservation and therefore, she plans to work in collaboration with the Landmark Foundation to discover the inspiration for her project. Watch this space.
I am a fine art student currently in my fourth and final year of studying. I have always had a strong passion for wildlife and my art has resulted in always taking the shape of that subjectivity. In my practice in recent years, I attempt to convey to the viewers the emotions that I feel about wildlife and it’s current state. Although there is a lot of good happening with wildlife conservation, there is dramatic negative activity happening and I attempt to portray this human-wildlife co-existence in a way that challenges the viewer and engages them with a more emotive and connected response. As a visual representation of statistical information, I hope to allow the viewer to truly see the stats rather than read, and in turn to draw the viewer in to empathize with the subject.
If you like what you’ve seen here and want to receive weekly digests via email of recent biodiversity and development research, then go ahead and sign up below. If you know of any recent papers, projects, or issues worth sharing, feel free to get in touch.