It took us three years from the initial analyses until the publication, but the idea for this paper dates back to 10 years ago. This is how it all started.
Red-necked tanager. Photo by José Carlos Morante-Filho
Cris Banks-Leite: My research is about understanding what drives species sensitivity to deforestation and what happens to ecosystems once species go extinct. Back in 2009, a colleague and I were discussing about bird sensitivity to habitat loss in the Atlantic Forest. I remember telling him that a particular site I had studied was so pristine that I was able to find many individuals of Habia rubica (red crowned ant tanager). He looked at me as if I was from another world and said: but this species is only found in disturbed areas. We started cross-checking our perceptions about certain species, and what became quickly apparent was that species that I considered to be very sensitive to habitat changes, he regarded them as disturbance-specialists and vice-versa. We did a quick and rough analysis of our data together, and indeed we found no correlation in how Atlantic Forest birds were impacted by habitat loss. We never went on from there, but this lack of consistency in species responses has always bothered me. Three years later, Jack Hatfield, then doing a Masters project under my supervision, looked into which morphological or behavioural traits were most predictive of species sensitivity to habitat loss, and we did something slightly different to previous papers – we tested for the transferability of our results. We found that although we could identify traits that explain why species are sensitive to habitat loss, we cannot reliably use these traits to predict how the same species would respond elsewhere (Hatfield et al. 2018 Ecological Applications). The findings that sensitivity was not intrinsically driven by species traits, and the observation that species have very plastic responses to habitat loss led to the idea of a Masters project to understand whether sensitivity to habitat changes is driven by species interactions, distance to range edge or climatic suitability.
Cris Banks-Leite doing field work for her PhD in the Atlantic Forest (2007)
Sarah Mayor: This is when I came into the research and started this study as part of my Masters. I emailed as many people as possible who had studied the effects of habitat fragmentation on birds in the Atlantic Forest of Brazil. We already had one dataset, and we received another five datasets from co-authors Pedro Develey, Luiz dos Anjos, Alexandre Uezu and José Carlos Morante. We derived sensitivity of the bird populations in each region and then checked for a correlation, to test the null hypothesis that populations of the same species will respond in the same way to habitat loss regardless of where they are found. Again, we found no correlation in bird sensitivity across the Atlantic Forest! Species that only occurred in high amounts of forest cover in one dataset were then occurring mostly in highly degraded areas in another dataset (see Supplementary Material Fig 1). However, none of the variables we analysed – distance to range edge, local interspecific competition and local climatic suitability – seemed to really explain this pattern. I wrote up my Master thesis without a good mechanistic explanation.
The following year, Jack Hatfield and Cris had obtained more data and, when we re-ran our analysis, we found that there was now strong support for a mechanism where distance from range edge modified species' sensitivity to forest loss.
We realised we had just found a very interesting result that disputes the ability of sensitive species as indicators of ecosystem integrity. And now we could explain such variability: populations occurring near their range edge were always more sensitive than those occurring in their range core.
The results we obtained and the ideas we had for taking the work forward however quickly overwhelmed both me and Cris, which is when David jumped in to do some statistical and GIS wizardry.
David Orme: A lot of my research has been in macroevolution and macroecology, studying the large scale patterns of diversity in phylogenetic trees and across space at global scales. That means I’ve spent a lot of time figuring out how to use large geographic datasets in research, and how to build statistical models that account for complex structure within the data, such as trying to differentiate intraspecific and interspecific responses to forest cover and distance to range edge.
Putting it together, we had clear hypotheses and preliminary results but then a large amount of validation and data processing and some new models to run. The details of this saga are in our Open Science Framework project for the paper (https://osf.io/4pbzt/wiki/home/), but we spent a long time and weeks of processing time matching taxonomy between field researchers and the species’ distributions, finding inland range edges, calculating distances between edges and species occurrences and estimating forest cover at many scales! That finally gave us a completed dataset and - with Jason Tylianakis helping ride herd on the statistics - a robust test of our hypotheses, confirming Sarah’s initial results that species sensitivity to habitat loss is variable and modulated by distance to range edge.
The excellent Birdlife International bird distribution maps then made it possible for us to look at these results from a macroecological perspective: if this pattern holds true for all species in the Atlantic Forest avifauna, then what are the spatial patterns of sensitivity likely to be across the ecoregion? That meant going back to the species maps for the whole avifauna and calculating range edges and distances for all species at high resolution across the whole ecoregion: time to go back to the High Throughput Computing service at Imperial College London again. Those maps, though, revealed the important conservation-relevant structure in patterns of sensitivity across the ecoregion that we report in the paper.
Cris, Sarah and David: In 2018, we submitted the paper to Nature Ecology and Evolution and got grilled by four highly critical, but very helpful, reviewers. Their comments led to a substantial increase in the length of the Supplementary Materials, but also increased our confidence in our results. We are certain that distance to range edge drives bird sensitivity to habitat loss in the Atlantic Forest, but we don’t know whether this holds true for other taxa and other regions. Other scientists have since told us that our results resonate with some inconsistencies they have seen in the field or in their data. We hope that our findings will inspire others to test these hypotheses, and will ultimately generate a better understanding of the mechanisms driving species sensitivity.