After over a year in and out of review I am pleased to announce the publication of my first article Consequences of inconsistently classifying woodland birds in Frontiers in Ecology and Evolution, co-authored by Georgia Garrard, Libby Rumpff, Cindy Hauser and Michael McCarthy. It’s been a long slog; I have battled constructive and un-constructive reviewers, self doubt and fatigue but I am finally triumphant! I am now a published scientist! All that remains is to ensure that people read our article (and finish my PhD and get a job…) I’ll give a short run down of the article here and then you can make up your mind about whether to read it. Cindy has also written up a nice short summary of the work which you can find here
I blogged about my preliminary findings many moons ago but our article has come some way since then.
This project arose because, during my Masters, I had wanted to study the effect of restoration on woodland birds but was stymied by my inability to determine which species to include in this group. I didn’t have the expertise to determine which species to include and trawling the literature unearthed a multitude of non-identical lists. I wanted to get to the bottom of this and discover how and why these lists differed and whether this impacts the results of research.
We systematically reviewed the literature and compiled a set of 38 lists of woodland birds which allowed us to work out how consistently each species was classified as a woodland bird. The graph below shows the frequency distribution of species classification (as seen in Fraser et al. 2015) where the bars represent the number of species falling into a classification bin (e.g. 0-10: classified as a woodland bird in between 0 and 10% of studies).
Frequency distribution of the percentage of studies in which individual species are classified as a woodland bird (total number of species = 165). Complete consistency in classification would appear as a binary distribution, where species are either regarded as woodland species 100% of the time or 0% of the time. Maximum inconsistency would occur if all species were classified as woodland birds in 50% of lists
You can see that, although there are peaks at either end of the spectrum, indicating that 56% of species are mostly classified as either woodland or non-woodland birds, there is little consistency in the classification of the remaining species. My supervisors and I felt that this was a fairly high degree of inconsistency so we wanted to find out if it’s just woodland bird researchers who classify things inconsistently or whether its a broader problem.
It turns out that ecologists have been noticing and writing about the inconsistent use of terms for as long as the field has existed. A few examples of this are; Mason and Langenheim’s 1957 article about the use of the term ‘environment’, Peet et al.’s 1974 investigation of the term ‘species diversity’, Hall et al.’s 1997 study of the term ‘habitat’, and most recently Herrando-Perez et al.’s 2014 call for ecology to develop a convention for nomenclature. Despite this, no one has ever comprehensively discussed the reasons for this or quantified it’s effects which is where my article comes in.
Why do researchers us the term ‘woodland bird’ differently?
We surveyed woodland bird authors and found that the main reasons that researchers classify different species as woodland birds were
- different aims of research. Researchers tailor their list of ‘woodland birds’ to include species that they expect to respond most strongly to the phenomenon they are interested in.
- disagreement about what a woodland is. Bird researchers variously consider woodlands to be i) any area with trees, ii) low density treed areas, iii) areas with vegetation matching Specht’s classification or iv) areas designated as woodlands by vegetation maps (e.g. EVC maps)
- disagreement about how to determine which birds depend on woodlands. Researchers variously determine which species are woodland birds based on i) whether they saw them in a woodland, ii) whether they occur more often in woodlands than in other habitats, iii) their nesting and foraging traits, iv) what types of habitat they require or avoid (e.g. do they need large areas of habitat or do they avoid degraded areas), v) process of elimination where certain types of species like water birds are excluded from the category, or vi) whether they have been classified as a woodland bird in another article/book
What impact does classifying species differently have on results
Garrard et al. (2012) modeled the effect of habitat aggregation (which is roughly the inverse of habitat fragmentation) on the occurrence of woodland birds, using a subset of species that they classified as woodland birds. We re-ran their model first using the entire complement of species and then being increasingly more selective about which species qualify as woodland birds. We excluded species based on the % of lists in which they were classed as woodland birds. The graph below shows the results where at 10 on the x axis, all species that are classified as woodland birds in more than 10% of articles are included in the analyses and at 80 only the species which are classified as woodland birds in more than 80% of lists are included. The horizontal line and shaded area represent the original model estimate and credible intervals.
The predicted effect of tree cover aggregation on species prevalence, for different subsets of species representing frequency thresholds of 10, 20, 30,…80%. At 80 on the horizontal axis, only species which are regarded as woodland birds in 80% or more of studies are included in the model. Error bars represent 95% credible intervals. Mean estimate from the original Garrard et al. (2012) model is represented by the line and the 95% credible intervals by the grey shaded area.
It is evident that the estimated effect of habitat aggregation differs depending on which species are included. If all species found in woodlands are included (as they are in some studies) the estimated effect of habitat aggregation is substantially lower than the original model estimate and that achieved above the 40% classification frequency threshold (where all species are classed as woodland birds in 40% or more of lists). Furthermore there is an upward trend in estimates as the woodland birds classification becomes more selective. This is an indication of a systematic bias in results where studies which are less selective about which species are woodland birds are likely to always obtain different results (probably with lower effect sizes) than those that are very selective about their classification.
In this case we looked at how classification effects results when the data collection, survey area and analyses are identical and found significant differences. When comparing results from studies using different classifications it is impossible to know whether differences are attributable to data collection, survey area or analyses or whether they are due to differences in classification. The more different lists of woodland species are the less comparable their results will be.
This is particularly problematic when you’re trying to understand woodland bird ecology or predict how they will react to management. Only a small subset of research uses identical lists of woodland birds, so researchers must choose between including all available information (which risks differences in classification confounding results) or only including studies which use the same list of woodland birds (which risks excluding valuable insights from other studies).
We propose that woodland bird researchers unite behind a single definition and list of woodland birds. Our research shows that this is likely to be an unpopular idea but we believe that standardization will make it easier to find and compile relevant evidence in the literature, avoid redundant scientific investigations and ease communication of research (MacGregor-Fors, 2011; Herrando-Perez et al., 2014)
Fraser, H., Garrard, G.E., Rumpff, L., Hauser, C.E. & McCarthy, M.A. (2015) Consequences of inconsistently classifying woodland birds. Frontiers in Ecology and Evolution, 3, 1–8.
Garrard, G.E., McCarthy, M.A., Vesk, P.A., Radford, J.Q. & Bennett, A.F. (2012) A predictive model of avian natal dispersal distance provides prior information for investigating response to landscape change. The Journal of Animal Ecology, 81, 14–23.
Hall, L.S., Krausman, P.R. & Morrison, M.L. (1997) The habitat concept and a plea for standard terminology. Wildlife Society Bulletin, 25, 173–182.
Herrando-Perez, S., Brook, B.W. & Bradshaw, C.J.A. (2014) Ecology needs a convention of nomenclature. BioScience, 64, 311–321.
MacGregor-Fors, I. (2011) Misconceptions or misunderstandings? On the standardization of basic terms and definitions in urban ecology. Landscape and Urban Planning, 100, 347–349.
Mason, H.L. & Langenheim, J.H. (1957) Language analysis and the concept ‘environment’. Ecology, 38, 325–340.
Peet, R.K. (1974) The measurement of species diversity. Annual Review of Ecological Systems, 5, 285–307.