You are here: Home Projection Highlights Climate change and migration

Climate change and migration

Since species survival in climate change and resulting future distribution depend notably on their dispersal capacity, a dispersal module has been integrated into dynamic vegetation models. This novel approach is tested through the Holocene postglacial re-colonization of Europe by temperate tree species. Simulations highlight that competition acts as a brake on species dispersion. This finding is important to keep in mind for the future management of ecosystems.


There is now convincing empirical evidence that many animal and plant species have already started to move their ranges in response to the climatic trends of the last century [1,2]. In mountain ecosystems, the predominant trend of such movements faces upwards as organisms try to follow the cool conditions they are adapted to, or because competition from species coming from below forces them to higher elevations [3,4,5]. The conic shapes and limited altitudes of mountains, however, make these upward movements a potential cul-de-sac and inevitably involve range reductions and, in extreme cases, local extinctions. As a corollary, cold-adapted high-mountain plants may face a massive biodiversity crisis. The likely magnitude of habitat loss and extinction processes in alpine environments is nevertheless controversial.


Part of the uncertainty about forthcoming climate effects on mountain plants arises from the predominant use of correlative models, so called species distribution or habitat models, for making the respective forecasts [6,7]. These models, though sophisticated in their internal structures, make some simplifying assumptions and suffer from a couple of shortcomings which render their results disputable [8]. A major such shortcoming is that they are ‘static’ in nature, i.e. they just make predictions about the current and future spatial distributions of the habitats climatically suitable to a species but do not represent the demographic processes that necessarily link the current and future distributions of real populations. This is considered problematic as limited dispersal capabilities may prevent many species from keeping track with the moving climate surfaces and realized future ranges might hence be smaller than suitable ones [9]. The importance of such migration lags is probably scale-dependent: it might be relatively low when focusing on a particular landscape [10] but much more prominent on regional or larger scales. Indeed, evidence from ECOCHANGE funded studies suggests that even the millennia of the Holocene have not been sufficient for many plants to reach an equilibrium with the modern climate when larger mountain systems like the European Alps are considered [11].

To overcome these particular shortcomings of correlative models we have, within ECOCHANGE, developed a so-called ‘hybrid model’, a kind of cellular automaton that couples static projections of habitat shifts with mechanistic simulations of local demographic processes and seed dispersal. We used this hybrid model to forecast the climate-change driven 21st century range dynamics of 150 subalpine and alpine plant species across the entire European Alps which were represented as a contiguous raster of ~ 20 Mio. individual 100 x 100 m sites [12]. In particular, we wanted to know how the results of such transient mechanistic simulations of population processes compare to the predictions of correlative models in terms of the habitat loss rates that high mountain species might face until the end of the century. To account for parameter uncertainty, we ran the model with two different sets of demographic (rates of germination, growth, survival, reproduction) and dispersal (range of propagule dispersal processes) parameters per species, one set to high and one to low values along a plausible range.

Discussion / interpretation of figures

Interestingly, for most of the 21st century, hybrid model predictions of the extent of habitat loss are lower than those of correlative models, even if the latter assume that species would adapt their ranges to the changing climatic conditions instantaneously whereas simulations were ran with low demographic and dispersal parameter values (Fig. 1).

Figure 1: Forecasts of proportional climate-change driven average range reduction of (150) high-mountain species in the European Alps until the end of the 21st century

fig 1 climate change & migration

The bold lines represent predictions of correlative habitat distribution models (either assuming unlimited (black) or no (red) dispersal abilities of the species), the thin lines predictions of hybrid models (ran eihter with high (black) or low (red) values of demographic and dispersal paramters).

Meaning: Hybrid models, which are include demographic and dispersal processes, indicate that range losses will be, apparently (see below) less severe than even the most optimistic habitat distribution model do predict.


Figure 2: Average number of sites predicted to be climatically unsuitable, but occupied; suitable,   but unoccupied; and both suitable and occupied.

fig 2 climate change & migration

The left and right panels refer to simulations with demographic and dispersal parameters set to high and low values, respectively. As in Fig. 1, sites are 100 x 100 cells of a regular grid spanning the entire Alps. In the upper panels the bold black line represents the propotion of sites the hybrid models predict to be occupied although they are not climatically suitable to the species at the respective point of time according to the habitat models. Vice versa, the grey lines represent the proportion of sites which would be climatically suitable, but remain unoccupied according to the hybrid models. Porportions are calculated in relation to the overall number of cells projected to be climatically suitable to the species. In the lower panels, the bold red line represents the number of sites which both the hybrid model predicts to be occupied and the habitat models classifies as climatically suitable. The other lines in the lower panels are taken over from Fig. 1 for comparison and represent the number of sites predicted to be occupied by the hybrid model (thin black line), the static model with unlimited (dashed black line) and with no dipsersal (dashed red line), respectively.

Meaning: the number of occupied sites is always higher, the number of suitable and occupied sites is always lower than the number of suitable sites. Over time both deviatons increase indicating the simultaneous built-up of an extinction debt and rising dispersal constraints on realized ranges. As a corollary, long-term range reductions are probably more pronounced than observable mid-term range reductions will suggest.

In combination these population dynamic processes predict an observable pattern with apparently less species (24 and 32%) loosing a critical proportion of their current range (> 80%) than even correlative models with unlimited dispersal would predict (33%). However, when focussing only on those occupied sites which are (still) climatically suitable, predicted loss rates clearly exceed those levels (49 and 55%, Fig. 3). In addition, these loss rates differ among species groups: they are larger for alpine than for subalpine species and, in particular, they are much more severe for species endemic to the Alpine chain than for non-endemics.

Figure 3: Percentage of species predicted to loose or gain a certain proportion of their current range size until the end of the 21st century.

fig 3 climate change & migration

Dark and bright grey values differentiate results from hybrid model simulations with low or high values of demographic and dispersal parameters, respectively. The upper two panel rows represent calculations based on the number of sites predicted to be occupied by the species; the lower panels calculations based on the number of sites which both the hybrid models predict to be occupied and the habitat models classify as still climatically suitable to the respective species. The thick and thin dashed line represent a range size loss of 100% and 80%, respectively

Meaning: Focussing on the number of occupied sites, i.e. the observable patterns, the percentage of species loosing a critical amount of habitat area (> 80%) will appear lower than when focussing on the number of occupied and suitable cells. Again, this implies that long-term effects of climate warming on biodiversity will probably be higher than those detectable during the next decades.

Overall, our simulations suggest that transient range shifts of mountain plants during the 21st century will be characterized by the combined effects of delayed population declines and lagged migration rates. As a result, observable range reductions will probably be less severe than expected from correlative model predictions. However, the persistence of remnant populations under already unsuitable climatic conditions creates an ‘extinction debt’ [13,14] which will have to be paid over time unless climatic conditions improve again, the species manage to survive in some small-scale climatic niches, or are able to adapt to the changing climate and the associated alterations in their biotic environments. In the long run, range reductions will thus likely be considerably more severe than those observable during the end of the century. Overall, these results hence warn against overly optimistic conclusions from lower-than-expected mid term range reductions of high mountain plants.


Further reading / Bibliography

Recently published paper in NATURE:


(1)    C. Parmesan, G. Yohe, G. (2003), Nature 421, 37-42.

(2)    J. Lenoir, J.C. Gegout, P.A. Marquet, P. De Ruffray, H. Brisse H (2008), Science 320, 1768–1771.

(3)    G. Grabherr, M. Gottfried, H. Pauli (1994), Nature 369, 448.

(4)    K. Klanderud, H.J.B. Birks (2003), Holocene 13, 1-6.

(5)    E. Kelly, M. L. Goulden (2008), Proc. Natl. Acad. Sci. U.S.A. 105, 11823-11826.

(6)    W. Thuiller, S. Lavorel, M. B. Araujo, M. T. Sykes, I. C. Prentice (2005), Proc. Natl. Acad. Sci. U.S.A. 102, 8245-8250.

(7)    R. Engler et al. (2011), Glob. Change Biol. 17, 2330-2341.

(8)    H. M. Pereira et al. (2010), Science 330, 1496-1501.

(9)    S.R. Loarie et al. (2009), Nature, 462, 1052-1055.

(10) R. Engler et al. (2009), Ecography 32, 34-45.

(11) F. Essl, S. Dullinger, C. Plutzar, W. Willner, W. Rabitsch. (2011), J. Biogeogr. 38, 604-614.

(12) S. Dullinger et al., in prep.

(13) D. Tilman, R.M. May, C.L. Lehman, M.A. Nowak (2002), Nature 371, 65–66.

(14) M. Kuussaari et al. (2009), Trends Ecol Evol 24, 564–571.

Document Actions