Script Flashcards
Hello
and thank you for attending my presentation on the Flight Response, and the Effect of Predatory Stimuli on Bornean Birds.
To introduce you to the background of our project:
Sabah, Borneo, is home to rich and eclectic avifaunal assemblages; with a high proportion of endemic species. In Figure 1 you can see a map of Borneo, alongside our experimental sites. Notably, Sabah is vulnerable to anthropogenic intervention – a point which I will refer to in the conclusions of this presentation.
Our study,
conducted in Sabah, Borneo, is important because there currently exists a knowledge gap in the response of Bornean birds to predatory stimuli in Danum Valley. As such, we chose to study this.
Our question was
whether bird abundance or diversity (here measured as species richness), would change upon exposure to a predatory stimulus, and if so, how? To test this, I generated 3 hypotheses: hypothesis 1: that bird abundance would change on exposure to predatory stimulus; hypothesis 2: that bird diversity would change on exposure to predatory stimulus, and hypothesis: 3 that there would be less “predatable” birds after a predatory stimulus, where I define “predatable” as a bird belonging to a Family that makes up a significant proportion of that predatory bird’s diet.
So, now that we had our hypotheses,
how did we carry out our study?
We used a “point count” methodology,
an established protocol in Sabah used for both hornbills and parrots (Marsden, 1999) with a variety of advantages and disadvantages. Usually, point count studies would use at least 20 point count stations. However, limited to a small group of investigators and only a small amount of camera trapping equipment that was being utilised in other study efforts, we used 6 systematic, boundaried locations, above the recommended distance of 200m apart. Further, we counted in 3 “open” and 3 “closed” sites, to begin to interrogate whether habitat type may have an effect. In Figure 2 you can see an example Open and Closed site; across the poster, Open sites are colour-coded and blue, and Closed sites as green. On initial sampling, the order in which these sites were sampled was random. It was important that they were then balanced over the following days.
Upon arriving at each point count station
we allowed a 2 minute habituation period, followed by a 10 minute fixed sampling time, across which we enforced a rule of minimal disturbance. At 5 minutes, we played a 1 minute standardised vocalisation recording. The time 1 minute was chosen so as to maximise the exposure of the local bird community, but to minimise the effect on our ability to auditorially identify the birds present. At half of the sites, this recording would be the call of a predatory bird, and at the other half, a non-predatory control and the calls would be balanced over the course of the experiment. With regards to the specific species’, it was important that we used a species call that would be familiar within the regional territory of Sabah. Hence, we were initially going to use a pied hornbill for the non-predatory bird, before finding out from our RA Didi that pied hornbill’s predate on the wing. Ultimately, we used the great argus as our non-predatory call and the crested goshawk as our predatory call; the sonogram of which I built in RAVEN, and can be seen, alongside an identification image aid, in Figure 3.
We noted bird species’
presence, abundance, distance (which was important to factor into our downstream distance modelling), and further notes. We also noted covariates such as the weather, detector identity, hours after sunrise, and ambient noise such as cicadas, car engines and one particularly disruptive volleyball game. Birds were resampled after the 5 minute call to determine retention or flight behaviours. (279, 487)
With regards to the results of our experiment,
I thought it best to contextualise our findings with a species accumulation curve, seen here in Figure 4. This shows a gradual accumulation of 90 species’ identified across 444 observations; 6 of which were endemic. However, notably, when modelling my data with a half-normal key function there was an expected abundance of 3367 birds; hence we did not sample every bird, due to detectability decay across distance. It is also notable that the best fitting half-normal key function with the lowest AIC for my data included detector identity and hours after sunrise as covariates.
With regards to our investigation itself,
initially, I tested my hypothesis 1; the results you can see in Figure 5. Colour-co-ordinated site type is on the x axis. On the y axis we have the change in abundance, in response to the non-predatory stimulus shown in orange, and predatory stimulus in red. The lines represent 95% confidence intervals. As you can see, all of the bars’ 95% confidence intervals overlap with one another, indicating that no site was significantly different with regards to change in abundance compared to one another. However, interestingly, the closed site change in abundance on exposure to a predatory stimulus appears to show a decrease from 0 – confirmed using a one-sample, two sided t-test.
When analysing the change in diversity,
a very similar pattern emerges. Here in Figure 6, we can see on the x-axis the colour-coded open and closed sites, but on the y we have the change in species richness. Once again, all 95% confidence intervals overlap, but it appears the change in diversity at closed sites to red predatory calls is significantly decreased from 0, again confirmed using a one-sample, two-sided t- test. So far, birds appear to decrease in abundance and diversity at closed sites after exposure to a predatory stimulus.
Finally, to test my third hypothesis,
that less “predatable” birds would be present after predatory stimulus exposure, I searched the external literature to determine the diet of the crested goshawk, and narrowed this down to the 5 most prevalent Families. I then partitioned our dataset into the relative abundance of those 5 “predatable” families, and the remaining families assigned as “unpredatable”.
The mosaic plot of Figure 7 shows
before coloured in dark green, and after in light green. As you can see, “predatable” species are much more abundant before compared to after in closed sites; a trend that is not recapitulated for “unpredatable” species in Closed habitats – this pattern was not observed in Open ones. When running a chi-squared test of association on this data, we see a p-value of 0.0546; not quite statistically significant, but tantalizing close; and potentially indicative of interesting biological activity warranting further investigation.
My suggested explanation for my results observed is that
in my dataset, “predatable” birds exhibited a greater proclivity for flight responses in closed habitats due to the specialisation of the closed-habitat avian predator that is the crested goshawk. As such, these “predatable” species would have evolved in an arms race to escape its predation by flight.
With regards to improvements, further analysis and future work,
I think it may be important, would I conduct this study again, that we more firmly address the biases and limitations of our collection techniques. These took forms both human and experimental – for example our experiment was unblinded; so potentially our awareness of which bird call treatment was applied may have subconsciously skewed our sampling effort; this could be solved by blinding our experiment. Moreover, birds we found trickier to identify may have skewed our focus, further complicated by a spectrum of identification skills within the group, causing us to miss other potentially present birds; this could be solved via camera trapping and computer identification.
