Citations:
Davidson, J., & Andrewartha, H. G. (1948). The influence of rainfall, evaporation and atmospheric temperature on fluctuations in the size of a natural population of Thrips imaginis(Thysanoptera). The Journal of Animal Ecology, 200-222.
Harrison, J. G., Shapiro, A. M., Espeset, A. E., Nice, C. C., Jahner, J. P., & Forister, M. L. (2015). Species with more volatile population dynamics are differentially impacted by weather. Biology letters, 11(2), 20140792.
Blog author: Miranda Salsbery
Author Background:
Prof. J. Davidson was head of the Department of Entomology, Waite Agricultural Research Institute, and a professor of entomology in the University of Adelaide. Dr Davidson actually passed away and this paper was published posthumously.
Dr. Josh G. Harrison works on interactions between hosts and symbionts and the ensuing evolutionary consequences for both interactors, focusing in microbial communities and uses lab and field experimentation with large-scale surveys of natural variation using data-intensive methods. He is currently a post-doc in Alex Buerkle's lab at the University of Wyoming. This was his first publication.
Davidson & Andrewartha (1948)
The main goal of this paper was to investigate density-independent factors that control populations, focusing primarily on how temperature and rainfall correlates with adult thrip population estimates.
From 1932-1946, adult thrips were counted in 10-20 roses daily (expect for Sundays and holidays) thought the spring and summer in the garden at the Waite Institute in Adelaide, Australia. Weather measurements were taken from a meteorological station close by. The authors performed a series of regressions to determine which weather condition correlated with daily and annual population changes in the thrips. They also tested several dependent variables, including the logarithum of the geometric mean daily counts for 15, 30, and 60 days before maximum daily counts as well as the logarithm of the geometric mean for October and November.
For changes in population density day to day, maximum temperature was found to the three times as important as rainfall in influencing the number of thrips in the roses. When looking a population year to year, 78% of the variance of the population could be related to environmental factors. The most important of which was the total thermal units (maximum daily temperature minus 48 divided by 2) accumulated between the inferred start date of development for the host plant (the ‘break’ of the season) and August 31st. Thus, maximum density for the population was largely determined by weather during the proceeding autumn. The author also observed a potential rhythm in the annual maximum density of the population, with a pattern of progressive increase for three to four years followed by a rapid decline for a year, then a progressive increase and so on.
The real goal of this paper was to show that density-independent factor, like weather, can explain population fluctuations. The author found little evidence to support the idea that competition or other density-dependent factors played a role in determining population density in this system.
Harrison et al. 2015
Rather than looking at how weather affects a single species, these authors looked to see, and found, that species with high population volatility (population that commonly fluctuate as opposed to remaining stable) are impacted more by certain weather events such as El Nino. They used a large long-term population dynamic dataset of co-occurring butterfly populations and a Bayesian framework. Changes in precipitation had very dissimilar responses between volatile and stable species, most likely by affecting the host plants. Temperature is more complex as it affects more aspect of species ecology such as behavior. This paper is interesting in that it shows species with similar dynamics may respond to climate change similarly.
My thoughts:
For starters, the thrip paper had way too many tables (23). It also wasn’t written in the most understandable way but more like a stream of consciousness. The variables were confusing. I do find it very interesting that they look multiple dependent variable as well as independent variable. We don’t do that very often anymore. I also think this paper is really important today as it laid out the initial evidence that the environment can affect populations, which as now grown into a large portion of climate change ecology, which the Harrison paper builds on. I though both papers were a bit short and speculative on ecological explanations for why they saw the patterns they did, especially in the 2015 paper. Even though it was a short a paper, I would have liked to see a more in-depth explanations for their results.
I agree with Miranda in that there was a whole lot of information to process in the Davidson paper, but the conclusions drawn from the data seem mostly valid. I did have a hard time reconciling with their idea that the effect of competition was negligible in this system, and the fact that the word competition was in quotes made it seem like Davidson didn’t quite buy into the idea of intraspecific competition as an ecological principle. The companion paper built off Davidson’s ideas nicely and expanded the concept to apply to a much larger sample of species. My understanding of statistics is quite limited at this point, but with a few exceptions (notably the MEI PC1), the trend lines in Figure 2 were not very convincing for Harrison’s arguments. Even some of the plots that had greater than a 95% chance of non-zero correlation seemed to have a very wide spread in data points.
ReplyDeleteI thought Davidson and Andrewartha (DA) was an interesting read, since these authors (along with Birch) strongly disagreed with the idea that population are regulated. While the consensus in population ecology is not in favor of the idea of populations maintained by density independent factors, I enjoyed reading such detailed and well-reasoned arguments against the major theories of the time.
ReplyDeleteI appreciated the clarity of presentation in the Harrison et al. paper and think it was a good match with the DA paper. Something that I wish was the authors had spent more time on was showing that their result that volatile species were more responsive to weather wasn’t just a consequence of volatile species having greater population variability (by definition). Also, it was unclear to me why they used the number of sightings as the response variable in their second model rather than abundance estimates like they used count data earlier to calculate CVs.
David
The Davidson and Andrewartha paper was, admittedly for me, a bit hard to process in the beginning. I struggled a bit in the beginning to fully grasp how they were calculating out all their variables. By the time I got to the discussion, the material began to make a bit more sense. I found their conclusions very fascinating, especially since they deviated away from the density-dependent, competition-based theories of population increases and decreases we have read so much about this semester. I was most impressed with the sheer amount of work that went into the project (collecting nearly every day for fourteen years!!) and wondered if any modern ecology studies still do this type of mass collection over time. Overall, I found this paper dense but worth it because of the conclusions they made.
ReplyDeleteThe companion paper, Harrison et al. (2015), was a good follow up paper not only because it mentioned the impact climate change could have on populations, but also because it asked the important follow up question: how could multiple taxa be affected by weather variables, not just one single taxon. In saying this, I found myself confused after reading the paper. I am not quite sure what their conclusions truly were. This paper feels, to me personally, like a lot of information is missing (which is most likely due to it being found in the supplementary info.). I wish the authors had spent a bit more time discussing their results and future directions this work could be taken in (especially when it comes to climate change research).
I appreciate the amount of work that went into the Davidson and Andrewartha paper and I enjoyed the simplicity of the project design (count thrips! record weather!). I thought the paper was nicely detailed but still readable. I do think their argument would have benefited from the inclusion of other density dependent or independent factors like predation, which may also be influenced by weather.
ReplyDeleteThe Harrison paper was a good match to the classic paper. It took the results a step further, making links from weather to climate. Of course, climate is a pressing question in modern ecology than it was for Davidson and Andrewartha. I do think the results from the Harrison paper are slightly cyclical (i.e. populations that have a tendency to fluctuate are more likely to fluctuate in response to weather conditions).
The thrip paper tried to figure out what environmental factors are responsible for the changes of thrip population. It is amazing that the author collected data for 14 years. The way the paper layout makes it a little difficult for me to follow, especially trying to figure out what are those “x”. Given the lack of modern computational tools, it was still a lot of efforts to build that model. The author mentioned that factors such as x2 could affect plant growth and so affect the thrip population, but I am curious if there is a way to separate the direct effect of weather and the effect of plants (the indirect effect of weather), on the population of thrips, since the plant performance (such as the number of flowers on a tree) is also affected by weather. I don’t think it will change the results a lot, but the mechanisms are different. Because Thrips imaginis are pests for multiple plant species that may have different blossoming triggers, I think separating the effect of weather and plants could help for better predictions.
ReplyDeleteI like the companion paper because it used more than one species (50 butterfly species) from a long studying period. However, it would be great to incorporate plant data since the author discussed a lot about the effects of precipitation on hosts. Also, the trend in the figure 2 is very vague to me.
The hand calculations of regression coefficients and detailed explanation of models in the Davidson and Andrewartha paper were very impressive! In reading papers on population fluctuations of other outbreaking insects (read: bark beetles), the conclusions have often been that temperature has a greater effect on populations than rainfall, consistent with the findings presented here. One of the main limitations was that the authors only sampled adult thrips in one rose garden, thereby not assessing the spatial variability that is known to occur in populations of outbreaking insects. On the other hand, it was a major advantage to have 14 years of count data while minimizing the variation in garden structure (and availability of plants) and other climatic variables that would have influenced population growth in other ways, had they sampled over numerous locations.
ReplyDeleteFor the Harrison et al. paper, this is somewhat nit-picky, but I really disliked their use of the word “volatile” to describe butterfly population fluctuations – it has an inherently negative connotation (i.e., change for the worse) and is not a commonly used term in population dynamics. That matter aside – in contrast to the Davidson and Andrewartha’s detailed explanation of their models, I would have liked a bit more information on the species-level models they used, as well as how they calculated/incorporated phylogenetic independent contrasts (while I recognize that the manuscript had a strict word limit). I found it interesting that the authors saw a stronger effect of precipitation on populations as opposed to temperature – perhaps this variable is more important for butterfly populations that rely on host plants that are sensitive to drought. However, as David mentioned, I was also confused on how they were able to definitively conclude that the effects shown were not just an artifact of some species having larger population fluctuations (and thereby different responses to weather).
The Davidson & Andrewartha paper results were fairly straightforward and simple. I wondered whether this was the first paper that showed fluctuations in populations can be tied to fluctuations in weather patterns, especially because other papers we have read have discounted the effects of density-independent factors on population sizes. For the daily fluctuations, despite the significance of their analysis, their effect size was very small (which they did acknowledge in the daily fluctuation results), which reminded me of our past discussion about the distinction between statistical and biological significance. I think that because the daily variation was not as strongly tied to weather, they should have focused more of the paper on the annual variation. I did not like the sections describing the dependent variable selection because this came off as very post-hoc rather than hypothesis driven. The authors also seemed very preoccupied with assuring the reader that calculating the statistics was not too cumbersome.. "with little labour...with remarkably little extra labour (p. 201)...in order to avoid the labour of calculating (p.202)..." which I found amusing.
ReplyDeleteThe Harrison paper was a good fit to accompany Davidson & Andrewartha because it addressed a similar basic question but using more advanced Bayesian stats and a broader scope. I was curious whether they were really arguing for density-independent effects here because they discuss in the intro that insects are likely tied to density-independent effects like weather indirectly by density-dependence on plant resources. And yet they claim this is one of the only papers tying density-independent factors to population regulation. I just don't think they can have it both ways.
The Davidson paper is really interesting from a historical stats perspective. As a modern ecologist, we would probably do this analysis (if we did it at all) with a Poisson regression. Because Poisson regressions use a log link function, it's pretty similar to what they did here, except that they used the traditional sum of squares approach rather than the likelihood-based approach most of us would probably use now. I don't know if it was possible/feasible to do a Poisson regression by hand at the time, but it's really interesting to watch them work through the rationale behind fitting a regression to integer counts via a log link. It's also interesting because it so obviously precedes the multiple hypothesis paradigm in which multiple models are fitted to a priori hypotheses and then compared.
ReplyDeleteThe Harrison et al. paper is the other end of the spectrum, presenting the Bayesian analysis methods that are becoming more and more prominent in the field of species-habitat interactions. I really enjoyed the Harrison paper because it demonstrated how large datasets can be leveraged for macro-ecological studies using innovative modeling techniques.
I agree that the classic paper did prove to be a bit difficult to read at times, especially when it came to contending with the sheer number of data tables they provided us. That said, I appreciated their discussion section which was explained clearly and made the overall interpretation of the paper fairly simple, despite the vast number of tables to take in while reading the body of the paper. I found it very interesting to look at such long-term data for a population study like this. Although aspects of the experimental design were simple, I still enjoyed looking at a relatively long-spanning study of weather effects.
ReplyDeleteI generally enjoyed the Harrison paper and the comparison of the population dynamics seen in numerous butterfly species. However, there were a few parts I felt I had information coming at me very fast and I could have benefited from a little bit more explanation and additional information in the text, although I understand with word-limits likely necessitated that any extraneous information, data, or analysis/results were put into supplementary data.
-Elizabeth