The indicators we have developed weight species equally and at first sight that might seem to be too simple an approach. But simplicity is one of the reasons why this approach has been adopted. The more complicated the procedure, the less understandable and the more prone to misinterpretation an indicator is. Our focus has been to develop comprehensible biodiversity indices that convey information on bird trends, and the simple geometric mean of species population trends has proven effective in that respect.
But one might, for example, weigh species indices by the habitat specialization of the species included, the desirability of these species, species abundance, species body mass, the precision of the species trend, degree of endemism, conservation value, phylogenetic uniqueness or diversity, cultural preference, or sensitivity to a particular driver (see Gregory & van Strien, 2010). In each case, the selection of weighting variables is subjective, and because weighting might affect the indicator, the resulting index would have a particular meaning, interpretation, and use.
We already explored the idea that biodiversity indices might be weighted by species biomass. One can imagine several reasons why indices of biomass might deviate from indices of abundance because, for example, different drivers may affect large- and small-bodied species differentially. But we found that the European farmland bird indices for biomass and abundance were highly similar (Voříšek et al., 2010). We would welcome studies to other weightings.
Statistically viewed, the graphs should indeed be presented with a logarithmic scale of the Y-axis. That would be in line with using the geometric mean of species´ indices, which is the back-transformed mean of the logarithm of the indices of species. However, we think that it is more convenient for users to apply a normal scale. A decline of the Farmland Bird Index with 30% or 40% may be more easily read from the graph.
Changes in the list of species that contribute to an indicator may affect that indicator. We improved the species classification procedure and finally based it on bio-geographical regions. That has led inevitably to changes in the species lists.
The farmland bird indicator (especially when statistically smoothed) has always shown the same message, i.e. a dramatic decline of farmland birds in Europe, regardless of the species classification used. It appears to be robust and less sensitive to changes in species lists. The index of common forest birds shows a higher sensitivity to changes in species lists, probably linked to the fact that forest habitats are much more diverse and different driving forces operate at a national or regional scale. This requires further study.
To make the indicators less sensitive to species composition changes, we include as many species in the indicators as possible. Furthermore, criteria used for species selection and classification for the indicators are chosen to prevent too much influence on one or a few species.
Learn more about PECBMS species habitat classification in Methods, chapter 3, Box Species selection and classification, or in chapter 2, in Box Data checks.
See also related question “How did PECBMS classify the species that are characteristic for each habitat type?“
The indicator uses the logarithm of the indices of species, and taking the logarithm usually dampens the effect of changes in one or a few species on the indicator substantially. Also, criteria used for deciding whether a species is to be included in an indicator are aimed to prevent too much effect of a single species on it. Nevertheless, if the indicator contains only a handful of species, population changes in one or a few species may considerably affect the indicator. It is possible – and useful – to test statistically whether the change in the indicator is only based on one or a few deviating species or reflects an overall change.
Learn more in Methods, chapter 4. Quality control.