Range filling in European trees
Article first published online: 6 OCT 2006
Journal of Biogeography
Volume 33, Issue 11, pages 2018–2021, November 2006
How to Cite
Svenning, J.-C., Normand, S. and Skov, F. (2006), Range filling in European trees. Journal of Biogeography, 33: 2018–2021. doi: 10.1111/j.1365-2699.2006.01630.x
- Issue published online: 6 OCT 2006
- Article first published online: 6 OCT 2006
A response to Welk, E. & Bruelheide, H. (2006) There may be bias in R/P ratios (realized vs. potential range) calculated for European tree species – an illustrated comment on SVENNING & SKOV. Journal of Biogeography, 33, 2013–2018.
In their comment, Welk & Bruelheide (2006; hereafter, W&B) argue that the results of Svenning & Skov (2004; hereafter, S&S) are too influenced by confounding factors and biases to support our conclusion that the low R/P ratios found for many European tree species to a large extent reflect dispersal limitation. Although we agree that the estimation of potential species ranges is a topic that needs much further study, we believe the S&S results are at least qualitatively robust, and explain why below. Our response follows the structure of W&B's commentary.
Atlas Florae Europaeae (AFE) uses a mapping unit of c. 50 × 50 km (Jalas & Suominen, 1972–94). W&B argue that ‘the coarse distribution data of AFE overestimates the species’ distribution, and thus, probably the climatic niche’. From a practical perspective, we note that at the scale of Europe most distribution data are simply not available at a finer spatial resolution. Furthermore, a key advantage of AFE and other atlas data is the existence of extensive, published documentation of the data and methodology (Jalas & Suominen, 1972–94). W&B use European-wide distribution data for seven tree species at a 10′ × 10′ raster resolution. However, it is difficult to judge the spatial resolution of the data that these maps are build on, since neither W&B nor the source of the maps (Hoffmann & Welk, 1999) provide much methodological detail. According to Hoffmann & Welk (1999) the maps were ‘compiled from a wide array of sources: dot and grid maps, floras covering larger or smaller areas, and distributional data taken from monographic or ecological studies’ (Hoffmann & Welk, 1999, p. 40). Another issue is that it is debatable to what extent increasing the spatial resolution will always improve estimation of climatic niches from distribution data. We agree with W&B that in mountainous regions mean climatic values at the AFE scale may not reflect the climatic conditions under which a species actually occurs, and hence may cause overestimation of its climatic tolerances. On the other hand, climatic range limits need not be caused by physiological tolerances of individuals, but may also reflect larger-scale climatic effects on population persistence and recolonization rates (Carter & Prince, 1981). Furthermore, non-climatic factors such as topography, land use, soils and biotic interactions may modify or over-rule climate as range determinants at smaller scales (Pearson & Dawson, 2003). In the case of Abies alba, emphasized by W&B, anthropogenic fires are probably responsible for its disappearance from much of the south alpine foreland and inner French Alps (Tinner et al., 1999; Carcaillet & Muller, 2005).
W&B provide bioclimatic suitability maps for seven tree species based on their 10′ × 10′ resolution data (W&B maps) that can be compared with the corresponding maps from our paper (S&S maps; their Fig. 1 and S1). In our opinion the W&B maps do not appear to provide better estimates of the species potential range. Notably, the W&B model for A. alba fails to predict its naturalization in the British Isles (Peterken, 2001) and underpredicts to a greater extent than the S&S model its wide naturalization in Norway and Sweden (Lid & Lid, 1994; Jonsell, 2000). The W&B models are also clearly too conservative with respect to climatic suitability for Carpinus betulus, Fagus sylvatica and Taxus baccata in western Scandinavia and the western and northern parts of the British Isles as judged by their naturalized and native distributions in these areas (Lid & Lid, 1994; Jonsell, 2000; Peterken, 2001; Reynolds, 2002). For example, F. sylvatica so successfully invades natural forest areas in northern and western Britain that it is considered a conservation problem (Reynolds, 2002) and novel Fagus–Quercus and Fagus–Fraxinus forest types are developing (Peterken, 2001). Furthermore, based on an analysis of the climatic factors associated with the present range limits of beech (Fagus spp.) throughout the Northern Hemisphere, Fang & Lechowizc (2006) conclude that ‘the northerly distribution of beech in Britain has not reached its potential limit, perhaps due to insufficient time since deglaciation to expand its range’.
S&S estimated the realized (R) and potential (P) ranges as the total number of suitable 10′ × 10′ pixels within the 50 × 50 km AFE cells occupied by a given species and across the whole study area (i.e. Europe), respectively. W&B suggest that a ‘‘‘mixed scales’’ approach’ confounds the R/P estimates ‘with uncontrollable bias’. This conclusion might partly reflect a misunderstanding of our approach, as W&B write that S&S ‘used the AFE resolution to estimate R’. In fact, R, like P, was estimated at the 10′ × 10′ scale, as explained above and clearly stated in the S&S paper. We note that we handled spatial scale for model calibration and projection in a similar fashion as other studies, notably Thuiller et al. (2005). Although the bioclimatic envelope models were calibrated at the 50 × 50 km scale, we think it makes good sense to project them at the 10′ × 10′ scale to provide estimates that are as accurate as possible for where suitable climates occur within the occupied AFE cells (R) or across Europe (P). As an example of the problems with the AFE spatial resolution, W&B highlight a distribution gap for A. alba in the Central Alps that is not apparent in the AFE data (their Fig. 1). However, our bioclimatic envelope model for A. alba nicely estimates zero climatic suitability for A. alba in this area (Fig. 1). Similar patterns were found for C. betulus, F. sylvatica, Quercus pubescens and T. baccata (Fig. 1). Hence, calibration at the 50 × 50 km scale and model projection at the 10′ × 10′ scale worked well in most cases. However, repeating S&S we accept that complex topography might cause problems for model calibration in other cases, as probably exemplified by Pinus cembra (Fig. 1; Fig. S1 in W&B). Furthermore, we note that projecting the models at the 10′ × 10′ scale rather than at the AFE scale will generally produce more conservative (i.e. larger) R/P estimates because unoccupied AFE cells will tend to contain fewer climatically suitable pixels than occupied cells. For example, when we recalculated R/P projecting the models at the AFE scale for the seven species selected by W&B, the resulting values were larger relative to the S&S approach in six cases, albeit slightly smaller for one species, Abies cephalonica (mean change 0.109 ± 0.080 SD).
W&B criticize rectangular envelope box models in general and suggest that ‘any rectangular envelope-box overestimates species distributions in environmental space by adding unoccupied corners’. This conclusion hinges strongly on the assumption that species ranges of most species are significantly influenced by interactions among their tolerances for different climatic factors. Although there are some examples of this, we believe it is premature to generally assume such interactions. Furthermore, a general allowance for tolerance interactions when estimating species tolerances from their distribution patterns increases the risk of overfitting.
In their paragraphs on ‘spatial covariance’ W&B suggest that the R/P ratio–latitude correlation ‘might simply be due to a covariation with topography’ due to biased modelling of species occurring in regions with complex topography, i.e. mainly southern species. However, as discussed above, for five of six species evaluated our model provided seemingly reasonable predictions of climatic suitability within the topographically complex Swiss region (Fig. 1).
W&B discuss how prevalence effects might constrain the possible values of the N/T ratio (ratio of native to total of naturalized and native occurrences) for a given range size. We accept that this is a good point, which is equally relevant for the R/P ratio. Species with large native ranges relative to the extent of the study area will unavoidably have high N/T and R/P ratios, which, however, also makes ecological sense. W&B suggest the approach of Araújo & Pearson (2005) as a promising alternative to the S&S approach. This approach is based on estimating the degree of range–climate equilibrium as the degree of co-variation between species distributions and climate. However, it is not immune to prevalence effects, because the scale of the variation for occurrences of species with intermediate range sizes will coincide more with the scale of climate variation than that of small- or large-range species. An analysis of the range size dependency of the correlation between species composition and climate for > 1500 European plant species confirms that the correlation peaks at intermediate range sizes (Fig. 2). Hence, using this approach one is likely to reach the somewhat counterintuitive conclusion that mid-range species are more in equilibrium with climate than large-range species.
Finally, we note that despite the differences in the approaches of S&S and W&B the resulting differences are rather modest. As also accepted by W&B, the results of both studies are consistent with the qualitative conclusion that a non-trivial proportion of European tree species exhibit low filling of their climatic potential range.
We thank Jesper Bladt and Thomas Fabbro for computational assistance, Leo Junikka of the Atlas Florae Europaeae project for access to the distribution data, David Viner from the Climatic Research Unit, University of East Anglia for letting us use their climate data sets, and the Danish Natural Science Research Council for economic support (grant no 21-04-0346 to J.C.S.).
- 2005) Equilibrium of species’ distributions with climate. Ecography, 28, 693–695. & (
- 2005) Holocene tree-limit and distribution of Abies alba in the inner French Alps: anthropogenic or climatic changes? Boreas, 34, 468–476. & (
- 1981) Epidemic models used to explain biogeographical distribution limits. Nature, 293, 644–645. & (
- 2006) Climatic limits to the present distribution of beech (Fagus L.) species in the world. Journal of Biogeography, 33, 1804–1819. & (
- 1999) A method for the estimation of the global population sizes of plant species – the area–abundance-index. Global Ecology and Biogeography, 8, 39–46. & (
- Jalas, J. & Suominen, J. (eds) (197294) Atlas Florae Europaeae: distribution of vascular plants in Europe, Vols 1–10. The Committee for Mapping the Flora of Europe and Societas Biologica Fennica Vanamo, Helsinki.
- Jonsell, B. (ed.) (2000) Flora Nordica 1. Lygopodiaceae to Polygonaceae. The Bergius Foundation, The Royal Swedish Academy of Sciences, Stockholm.
- 1994) Norsk flora, 6th edn. Det Norske Samlaget, Oslo. & (
- 2003) Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecology and Biogeography, 12, 361–371. & (
- 2001) Ecological effects of introduced tree species in Britain. Forest Ecology and Management, 141, 31–42. (
- 2002) A catalogue of alien plants in Ireland. National Botanic Gardens, Glasnevin. (
- 2004) Limited filling of the potential range in European tree species. Ecology Letters, 7, 565–573. & (
- 2005) Climate change threats to plant diversity in Europe. Proceedings of the National Academy of Sciences of the United States of America, 102, 8245–8250. , , , & (
- 1999) Long-term forest fire ecology and dynamics in southern Switzerland. Journal of Ecology, 87, 273–289. , , , & (
- 2006) There may be bias in R/P ratios (realized vs. potential range) calculated for European tree species – an illustrated comment on SVENNING & SKOV. Journal of Biogeography, 33, 2013–2018. & (
Jens-Christian Svenning is an associate professor at the Department of Biological Sciences, University of Aarhus. His main research interests include macroecology of plants and predictive modelling of global change impacts on species distributions and biodiversity.
Signe Normand is a PhD student at the Department of Biological Sciences, University of Aarhus. Her research mainly addresses macroecological questions on the distribution and diversity patterns of European plant species.
Flemming Skov is a senior research scientist at the Department of Wildlife Ecology and Biodiversity, National Environmental Research Institute of Denmark. His research interests include spatial modelling of species distributions and biodiversity in relation to global climatic change and human land-use.
Editor: Ole Vetaas