Get Complete Project Material File(s) Now! »
KNOWLEDGE ON OBSERVED VEGETATION SHIFTS RELATED TO CLIMATE WARMING
Effects of climate warming on species’ or community shifts have been observed for animals as well as plants (Walther et al. 2002; Parmesan & Yohe 2003; Root et al. 2003; Walther 2004) from different types of ecosystems.
Plant species shifts along latitudinal gradients have been documented mainly at high latitude as a frequency increase of herbaceous species (Smith 1994) or increasing shrub abundance (Sturm et al. 2001) in arctic or Antarctic tundra. A progress of native evergreen broad-leaved species (the so-called lauriphyllous species) in forests was expected since they should be favoured by an increase in late winter and spring temperatures. While budburst of deciduous trees has not yet started, leaves of these evergreen species should benefit from an increase in early spring temperature. Indeed, a northward shift of Ilex aquifolium was reported from northern Europe (Walther et al. 2005a). An increase in frequency of this species was also noted within its distribution range, in north-eastern France (Cluzeau et al. 2001a; Cluzeau et al. 2001b). In Mediterranean area, Ferula communis L. subsp. glauca was also shown to shift northward (Mandin 1993).
Response of animal species to climate warming was mostly observed for butterflies (Parmesan 1996; Parmesan et al. 1999; Parmesan & Yohe 2003; Warren et al. 2001; Franco et al. 2006) or birds (Thomas & Lennon 1999; Devictor et al. 2008), but also for some other taxonomic groups (Hickling et al. 2006). Both northern and southern ranges limits were affected.
However, most studies on the climate change impact on species distribution have focused on mountain areas where species are expected to respond more strongly.
Observed species shifts in mountain areas
Here we concentrate on plant migration responses to climate warming.
Extension at the upper limit of species distribution
First studies focusing on vegetation shifts in mountains were conduced on alpine summits, by reinventory of summit flora. The oldest ones were the reinventory of earlier studies realised by Braun-Blanquet on Piz Languard in Bernina and Piz Linard in Engadin, Switzerland (Braun-Blanquet 1955; Braun-Blanquet 1957, respectively), and although climate warming was not yet highly topical, increase in species richness on these two summits already suggested a signal of climate warming and were seminal for the following studies.
Later, based on the same kind of phytogeographical monograph of a region (e.g. Rübel 1912), series of summits were reinventoried in the Alps (Hofer 1992; Grabherr et al. 1994; Grabherr et al. 1995). Species newly recorded at the summits were considered to have migrated from lower elevation, responding to a global temperature increase. This was confirmed by more recent studies in the same area (Walther et al. 2005b; Walther et al. 2005c; Holzinger et al. 2008). According to Jurasinski & Kreyling (2007), this upward shift led to an increase in homogeneity among summits. Homogenization process could also have occurred in Norwegian mountain since species with wide altitudinal and ecological ranges were both increasing in abundance and shifting upward, while more specialist species declined (Klanderud & Birks 2003). The same kind of increase in species richness was observed along an elevation gradient in Italy, in front of the valleys previously studied by Braun-Blanquet in Switzerland (Parolo & Rossi 2008). Additionally, species which were already present at lower elevation along the transect at the first sampling date were recorded at higher elevation during the resampling study, confirming the observed trend of upward species shift. In order to get results homogenised on a worldwide scale, a project of long-term summits monitoring and reinventory has been designed with a homogeneous methodology. First results are consistent with other summits reinventory, and show that new species from lower belts have colonised the summits (Pauli et al. 2007; Erschbamer et al. 2009). In arctic mountains, only an increase in the abundance of one shrub and two herbaceous species was reported, but no global upward shift (Wilson & Nilsson 2009), whereas an upslope expansion was reported on a sub-Antarctic Island (le Roux & McGeoch 2008).
Within forest belts, anthropogenic disturbances other than climate change could be at a lower level than in above-forest areas, mainly because livestock and pasturing are most often more intense above than below the treeline. The pine mistletoe was monitored along a transect at its upper range limit, and displayed an upward shift, seemingly caused by climate warming (Dobbertin et al. 2005).
Treeline and ecotones
In addition to herbaceous and understorey species mentioned in the previous paragraph, the advance of tree species in mountain areas have been also often studied. Since detection of the altitudinal tree limit is relatively easy, in comparison of the mapping of herbaceous species distribution, and because the mechanisms underlying the determination of treeline have been of interest for a long time (Broll & Keplin 2005), altitudinal tree limit has been used as a study model of forest ecosystem response to climate warming. However, because the position of the treeline was shown to be controlled as much by anthropogenic pressure as it is by climate (Bader & Ruijten 2008), effect of climate warming on its advance has been debated. The consensus is that current advance of treeline has been mostly triggered by pasture abandonment above treeline, but that climate warming also favoured tree regeneration and growth (Penuelas & Boada 2003; Cairns & Moen 2004; Motta et al. 2006; Bolli et al. 2007; Cairns et al. 2007; Gehrig-Fasel et al. 2007; Batllori & Gutiérrez 2008). However, treeline advance was not expected when it was already at its potential highest limit, i.e. its climatic limit (Ninot et al. 2008). Only few authors proposed climate change as the only cause of treeline advance (Kullman 2002; Sanz-Elorza et al. 2003; Beckage et al. 2008; Wieser et al. 2009),
Retraction at lower limits of species distribution
Whereas cold temperatures seem to represent a real constraint on the development of species at their northern or upper distribution range, warm temperatures seem not to be the decisive factor at southern or lower limits (Vetaas 2002). Other factors such as habitat replacement by species shifting at their colder limit, habitat fragmentation or other anthropogenic disturbances could influence the species response (contraction, extinction) to temperature increase at their lower/warmer limits, making its interpretation difficult. Thus, partly because of the scattered species presence at lower margins compared to upper margins and, consequently, partly because their more difficult detection, very few studies were undergone on plant species at their lower limits. As expected, a decline of species at their lower limits was observed, either based on their reproductive performances (García-Camacho & Escudero 2009) on their abundance, frequency or growth (Klanderud & Birks 2003; Lesica & McCune 2004; Mitzunaga et al. 2005; Pauli et al. 2007; Erschbamer et al. 2009) preceding a probable extinction.
Shift of the species distribution optimum
For the moment, few studies have focused on the whole range of species distribution of mountain plants, because not enough complete datasets were available at two different dates. This kind of method should detect as well extinction or migration at each range limit, as a change in species frequency at the lower or upper elevations within the original range. Thus, optimum or mean elevation of a species is an interesting information because it integrates the presence of the species over its whole distribution range. Only three studies reported changes in species elevation using such an approach. The first one reported an upward shift of mean elevation of 316 species in the southern French Alps (Dupouey et al. 1998). Another one described a significant mean shift of 171 species optimum in French mountain forests (Lenoir et al. 2008), and others showed the increase of the averaged cover-weighted mean elevation of ten widely distributed species along a transect in Californian mountains (Kelly & Goulden 2008). All these three studies attributed the observed shift to climate warming.
Changes in community composition
When it is not possible to consider the whole range of species, for example when the number of plots is not large enough, it can be still interesting to study changes at the level of the entire plant community. Communities integrate ecological parameters of the environment. Parallel to the development of phytosociology discipline in the seventies, different authors have proposed indicator values of plant species for different factors of the environment, for the species pool of temperate central Europe (Ellenberg 1974) and Switzerland (Landolt 1977). These indicator values were used as such in other countries of Europe, or were completed or modified (e.g. for Italy by Pignatti et al. 2001). More recently, indicators were constructed for French forest species based on a coupled statistical analysis of floristic and environmental data from the Ecoplant database (Gégout et al. 2005). For example, such indicator values can be averaged at the community level, and their change over time monitored, without a precise knowledge of the movements of each species belonging to the community. Such an approach can be used when the number of plots resampled is low, or when the total area sampled is small. Based on such calculations at the community level, Keller et al. (2000) showed that climate warming had provoked a change in community composition towards a more thermophilous vegetation.
Which phenomena influencing species distribution could interact with climate warming
Often, in previously quoted studies, vegetation not only reacted to climate warming, but also to some other direct or indirect anthropogenic disturbances. In some cases, the response to climate change has been considered as being independent from the response to these other anthropogenic disturbances (e.g. eutrophication). But, such other anthropogenic disturbances could also reinforced the impact of climate change: habitat fragmentation at the southern margins of species associated to climate warming favoured species loss (Lavergne et al. 2005; Lavergne et al. 2006), whereas pasture abandonment triggered the upward shift of subalpine or woody species favoured by temperature increase (Vittoz et al. 2009).
The species shifts already observed could be also the consequence of a slow re-colonisation process triggered by temperature increase since the end of the Little Ice age, i.e. around 1850 (Kammer et al. 2007), rising the question of inertia of species response.
WHERE AND HOW TO DETECT VEGETATION CHANGES AND SPECIES MIGRATION? GENERAL METHODOLOGY
The study of vegetation changes under anthropogenic disturbances implies to have repeated inventories over years at one’s disposal, or at least to find historical monitoring data for reinventory. These ancient studies should provide information on species presence together with either an accurate geolocalisation or a detailed ecological description of the environment (elevation, exposition, soil type, pH, habitat…). The record of species in each plot should be as exhaustive as possible.
Authors of ancient studies did not always take care of the exact localisation of the plots they studied because their goal was often a static description of the vegetation in a region, or the study of an ecological question at one point in time, without any plan to come back later. Reinventories on permanent plots can only be based on fine localised plots, either on several plots with restricted area, or on fewer plots but with extended area. However, there are methods which can be used in case of resampling studies of non permanent plots. In this latter case, instead of studying changes in the appearance, disappearance or abundance of species at one or several locations exactly relocated and resampled, one can build a model of the relationship between species presence or abundance and environmental factors as elevation or aspect, and study the changes in this model over time. Such a method no longer needs permanent plots, but is instead relying on an appropriate global sampling scheme of vegetation at different dates. This method can be applied to the study of single species niche as well as to community changes, as long as environmental factors have been recorded together with vegetation data.
Another advantage of resampling in non-permanent plots is that it can more easily avoid the problem of intrinsic stand ageing between each inventory observed on permanent plot. When resampling a limited set of permanent plots, the first factor that will influence vegetation dynamics is the ageing and maturation of the stands at each plot (canopy closure, tree ageing, litter accumulation, progressive soil acidification…). It is usually difficult to separate such a maturation effect from other impacts. In non-permanent plot designs, one can rebuild a representative sample of the area under study at each date of sampling, not affected by stand ageing by an appropriate selection of plots according to their degree of maturation.
In resampling studies, two different cases should be distinguished: those where the initial sampling campaign was designed in order to be redone later, and those where it was not planned to do any later resampling. The first case includes the monitoring networks launched in France and Europe during the last decades. Such networks usually provide homogeneous vegetation data over space and time. In these networks, the sampling is often designed in order to be representative of the vegetation in the area under study at a given date. In comparison with studies not initially designed to be resampled, monitoring networks most often cover larger areas (region- or nation-wide) whereas the former are restricted to a selected area. In monitoring networks, planned for repeated inventories, protocols of vegetation recording are most often available, and the quality of vegetation censuses is usually better, as well as their exhaustiveness.
Which part of species range to observe?
Species range limits
Historically, studies have mostly focused on species’ shifts at the upper limits, and the lower limits for animals mainly.
An expansion at the upper limit is expected under effect of climate warming. Hence, migration should be put in evidence by recording an arrival of species above their previous upper limit. It could be determined by recording new species arriving into a delimited area, for example a mountain summit. The shift is measured as the difference with the last recorded upper elevation known.
At lower limits, a retraction of the range is expected under climate warming. Method consists in a careful search and check of species presence at delimited localities known as the previous lower limit of the species, and a record of the today lower elevation of the species.
These methods usually focus on very restricted areas because the exhaustive search for upper or lower limits can be very long. It raises the question of the sampling intensity. Indeed, the extinction or arrival of a species at a locality could be the result of a species move within its distribution range, without any true change in upper or lower limits. Such a change will be confused with a change in elevation if the sampling intensity is too low.
More important is the problem of frequency increase because of an increasing exhaustiveness along the successive inventories (Archaux et al. 2006, Archaux et al. 2008, Vittoz & Guisan 2007). It is a very general observation that the number of species observed on a given area increase along time due to an improvement of observers exhaustiveness, either because they are better trained, or because they use previous lists of recorded species in the field when recording the new list. This is especially relevant for designs that were not planned to be resampled from the beginning.
One advantage of the method is that only few data are necessary, such as a list of localities for lower or upper limits of species, or a record of species found on a mountain summit. Such kind of data is available in a large number of ancient publications.
The position of species optima along the elevation gradient can shift due to a homogeneous upslope movement of the whole species range, or an extinction or colonisation at one of the range limits. Simple models can be used to estimate the position of this optimum, such as the classical Gaussian logit model. This implies that the species displays a continuous and not-fragmented distribution in order to fit the model on the observed species frequency. Hence it is necessary to work with sufficiently frequent and not scattered species. The calculated optimum is usually not sensitive to a homogeneous increase of species frequency along the entire studied gradient. Thus, the position of the optimum does not depend on the exhaustiveness of the plant censuses, which is an interesting property of this indicator.
However, a lot of presence/absence observations are needed for each species in order to fit the distribution model, implying the availability of rather large data sets.
When not enough data are available to model the distribution of species according to elevation, it is still possible to test if vegetation has moved. One can use external information available for each species on its ecological requirements (e.g. life traits, Ellenberg’s or Landolt’s indicator values) and integrate this information at the whole plant community level, by averaging or other procedures. For example, temperature indicator values are available for most of the plant species of the Alps. Using these indicator values averaged at each plot, it is possible to test if vegetation has shifted towards a more thermophilous or cryophilous state between two inventories.
Such indicator values are not very sensitive to the presence or absence of a given species, nor sensitive to the plant census exhaustiveness. Thus, they are not much influenced by the quality of the vegetation census. Because they use a priori information, calibrated in other, previous, studies, such indicator values can be calculated without a precise knowledge of the distribution of species in the area under study, contrary to the modeling of the niche optimum, which needs to gather presence/absence data in a large number of plots.
When data are resampled in permanent plots, the simple comparison of these community indicators between successive dates provides information about the shift in vegetation. In case of non-permanent plots, shifts of vegetation must be identified through the study of the shifts in the relationship between the indicator values and elevation. For example, we will study in such cases (non-permanent plots) the relationship between Landolt’s mean indicator value for temperature and elevation and the shift in this relationship between dates. A shift in this relationship implies that vegetation has moved.
Different spatial scales
Depending on the extent of the studied area, the significance of the results will differ. Working on small areas is, of course, easier but the results cannot be generalized. Methods based on the detection of changes in the species lower and upper limits, and methods based on the resampling of permanent plots can be used in small areas, because they do not require large datasets. Methods modeling the whole distribution range of a species, such as those used in non-permanent networks, require many plots along the studied environmental gradients and consequently larger study areas.
Different temporal scales
There is a trade-off between the time interval over which vegetation changes can be observed and the quality and reliability of the data available. Historical data are often old enough to observe vegetation changes on a long time scale, until a century in Europe. But the quality of old data is questionable, with unclear methodology of vegetation census, low accuracy of plots location, low exhaustiveness. At the opposite, homogeneous and high quality vegetation databases built from current monitoring networks only offer a short time period of study, because they were established only relatively recently. In this latter case, the observed response of species to environmental changes could be weak due to a too short period of observation.
Response of plant species to climate changes could depend on the habitat type they belong to (forest or meadows, subalpine or alpine belts…). Moreover, environmental changes other than climate change could have more marked effects in some habitats, relatively to others. Some habitats are prone to specific bias in the detection of long-term changes in vegetation composition. For example, studies based on forest permanent plots are prone to the stand ageing bias (see above). On the other hand, reinventories above the treeline are strongly influenced by pasture abandonment. It is important to avoid or control such confounding factors when studying effects of other environmental changes such as climate change. For example, reinventory in forests could avoid the stand ageing bias by using non permanent plots, carefully selected in order to get an equal average stand age at each sampled date. When working in the alpine belts, the choice of the study area should be made far above the treeline.
In forests of Western Europe, which are nearly all managed, the species composition of the tree stratum is mainly the historical result of the choices made by forest managers. Due to this strong and direct anthropogenic control, it is a biased indicator of environmental changes. Thus, that is the reason why we focused mainly on the understorey vegetation.
Description of the thesis chapters
The focus of this thesis is on the response of mountain plant species to climate warming over the Alps (figure 3). Three different aspects are highlighted, that were rarely studied so far: (i) dynamics of lower limits, (ii) changes in species optimum and (iii) changes in indicator values at the community level. Two of the studies are based on large scale sampling schemes. In these two large-scale studies, a new methodology of analysis is applied that is adapted for samples collected in non-permanent plots.
In this study, particular attention is paid to the detection and separation of climate change effects from other anthropogenic changes, especially land-use change, which is one of the main drivers of environmental changes in the montane belt of southern Europe.
Table of contents :
I. General Introduction
A. What are the changes in the environment that promote changes in vegetation?
1. Eutrophication and acidification due to atmospheric deposition
2. Species loss from habitat fragmentation
3. Forest colonisation after agricultural abandonment and its consequences on vegetation
5. Land use legacy
6. Forest management and forest dynamics
B. Climatic change
1. Expected responses of species and biomes
2. Particular case of mountains
C. Knowledge on observed vegetation shifts related to climate warming
1. Observed species shifts in mountain areas
a. Extension at the upper limit of species distribution
b. Treeline and ecotones
c. Retraction at lower limits of species distribution
d. Shift of the species distribution optimum
e. Changes in community composition
2. Which phenomena influencing species distribution could interact with climate warming
D. Where and how to detect vegetation changes and species migration? General methodology
1. Different approaches
a. Different methodologies
b. Which part of species range to observe?
c. Different spatial scales
d. Different temporal scales
e. Different habitats
2. Description of the thesis chapters
II. Shift of mountain vegetation in forests of the Southern Alps: climate or land-use change?
1. Objectives of the study
C. Material and Methods
1. Study area
2. Vegetation dataset
3. Data analyses
a. Estimation of species shifts
b. Species shifts and plant traits
c. Species shifts and successional dynamics
1. Estimation of species shifts
2. Species shifts and plant traits
3. Species shifts and successional dynamics
1. Causes of the observed shift
2. Forest closing, ageing and maturation
3. Lag between vegetation shift and climate warming
H. Annex of the chapter two
III. Long term changes in plant communities of the Maurienne valley, French Alps
B. Material and Methods
1. Study area
2. Sampling protocol
a. Historical data
3. Data analyses
a. Species selection
b. Vegetation indicators
c. Plot ordination
d. Testing for changes in vegetation between the two inventories
e. Development of forest structure from 1985 to 2000
1. Abiotic models of vegetation indicators
a. Ellenberg indicator values
b. Landolt indicator values
c. Consistency of vegetation indicators: correlation between Ellenberg and Landolt values
d. Ordination results
2. Vegetation changes between the two inventories
a. Shift in mean Ellenberg and Landolt indicator values
b. Shift in the position along the correspondence analysis axes
c. Changes in stand dendrometric characteristics
1. Response to climate change
2. Changes in soil characteristics and nitrogen atmospheric deposition
3. Forest development and structure
4. Sampling bias
G. Annex of the chapter three
IV. Plant species’ range shifts in mountainous areas – all uphill from here?
C. Material & Methods
1. Shifts at the upper range margin
2. Shifts at the lower range margin
3. Comparing range shifts at the lower and upper limit of the same species
1. Shifts at the upper range margins of species reaching the summit areas
2. Shifts of the upper range margins along the upper slopes of Piz Languard (3000 – 3262 m a.s.l.)
3. Shifts at the lower range margin
4. Shifts at opposite range margins of the same species
V. One century of vegetation change on Isla Persa, a nunatak in the Bernina massif in the Swiss Alps
C. Study area
3 D. Methods
1. Field plant inventories
1. Data reliability
2. Biodiversity increase and climate change
3. Biological traits
4. Importance of the biodiversity increase
H. Annex of the chapter five
IV. General Discussion
A. What conclusions can be drawn by connecting the different studies results?
1. A spatially varied response to climate change
2. Responses to land-use change and other anthropogenic pressures
3. Advantages and drawbacks of different methods: sampling bias, spatial scale and sampling intensity
B. General conclusions