Overall Conclusions: Too Hot to Handle?

Throughout the course of this blog, the emphasis has changed somewhat. To begin with I introduced the general concepts around species, their migrationary capacity and their response to predicted changes in climate. Migration rates were examined, including concepts such as Reid’s paradox, which describes post glacial species migration rates of up to 1000 m per year. Spatial variability of future climate change led to theories of favourable microclimates and possible refugia.

I have shown how climate change has already started to effect species, such as the Turdus migratorius and Marmota flaviventris due to changes in phenology in the Colardo rockies. Other species such as tree populations in North America were analyzed in the context of their range shift due to climate change.

From there, I decided to link past migrations from glacial refugia to possible future modelled migrations. Southern glacial refugia in the mediterranean peninsulas of Iberia, Italy and the Balkans were discussed. Methods of locating glacial refugia such as phylogeographic techniques and reconstructing post glacial colonization through macrofossils and pollen records have been demonstrated. This information was used to infer whether migration rates described in Reid’s paradox could be feasile. I concluded that the existence of cryptic northern glacial refugia was the most likely explanation.

Having shown evidence from past climate change and post glacial recolonization, I reviewed current vegetation modelling techniques. Bioclimate models were found to be the most commonly used, however several criticisms were apparent. These were mainly aimed at their simplicity, as they did not take into account: land use, inter species interaction and evolutionary change and other such parameters. Dynamic global vegetation models are the next generation of model from which future climate impacts on ecosystems should be drawn. An example of how DGVMs may be used was shown, although significant further research into modelling processes needs to be undertaken.

From this blog I hope to have shown the possible effects of climate change on species, and how it will force them to migrate to higher latitudes and altitudes. If i were to have continued this blog I would have focussed on theories and findings as they are published to further my understanding of the topic at the forefront of the field. I have found this experience to be interesting and exciting and would like to thank Dr. Anson Mackay for giving us this unique way of exhibiting our knowledge as well as contributing to academic debate.

Application of a Dynamic Global Vegetation Model

Today’s post will look at the future impacts of Climate Change on the earths ecosystems from predictions from dynamic global vegetation models (DGVMs). A paper by Gonzalez et al. (2010) attempts to model future changes in global biomes by comparing observed changes in 20th century climate with modelled predictions of 21st century vegetation changes. Three global climate models (GCMs) were used in cooperation with the MC1 DGVM.

Results from the projected climate data (Figure 2) and projected vegetation changes (Figure 3) are shown. The GCMs project widespread temperature increase and precipitation changes by 2100. This includes global average temperature increases of 2.4-4 degrees. Average precipitation increases at rates of 0.03-0.04 mm, but becomes increasingly spatially variable. The model also predicts that wildfire frequency will increase in around 1/3 of the global area.

Results from global vegetation modelling follows observed patterns of global biomes. MC1 projections show potential extensive changes under 2071-2100 scenarios. Temperate mixed forest shows the highest areas of potential change, with desert showing the lowest. Gonzalez et al. (2010) believe between one-tenth and one-half of the earths biomes may be highly to very highly vulnerable to change.

Vegetation projections suggest potential latitudinal biome shifts of up to 400 km. Temperate mixed forest shows high vulnerability due to projected loss of coniferous species leading to conversion to temperate broadleaf forest. Tropical ecosystems show low vulnerability to change due to their high temperature tolerance combined with projections of increased precipitation around the equator (Malhi et al. 2008). MC1 data is congruent with data from other DGVMs which agree on shift of boreal forest into tundra at high latitudes and some forest lost in the Amazon.

As for effects, a large proportion of the world’s population live in areas in high vulnerability of potential biome changes. Biome change may alter ecosystem services, such as wood for timber, grass species preferred for grazing and water retention capacity of watersheds for human consumption.

Analysing the Lund-Potsdam-Jena Model (LPJ)

Todays post will be to do with analysing one of the dynamic global vegetation models (DGVM) previously discussed. The one chosen is the LPJ model discussed in Sitch et al. (2003). It is one of many fully integrated DGVMs that have been developed following the criticism of bioclimate envelope models. The model describes vegetation in terms of fractional coverage of a grid cell, taking into account different plant functional types (PFTs) (Table 1).

Each PFT is then assigned bioclimatic limits (Table 2) which determines whether the species can survive/regenerate under the climatic conditions of the particular grid cell.

LPJ addresses the limitations of bioclimate envelope models be including representation of vegetation structure, dynamics, competition between PFTs and soil biogeography into the model. Figure 1 shows the model logic.

The model can predict changes on different spatial scales. My post will focus on those on the global scale. Figure 2 shows a simulated map of potential natural vegetation for the modern climate. It accurately shows the boreal evergreen forests in Canada and northern Eurasia, the boreal deciduous forests in Siberia, and the transition into temperate ecosystems of north America, western Europe and China. LPJ simulates the transition from savanna into evergreen rainforests near the equator, as well as northern tundra, and grasslands in drier areas.

Figure 2. LPJ predictions of PFT distribution

LPJ has been shown to be able to reproduce relatively accurate models of current global ecosystems. This proves the model’s effectivness and provides some reliability for its application to future modelling. Looking forward, attempts should be made to use this information to examine possible effects of climate on species on a global scale. Having shown an example of a DGVM and how they work and my next post will show how one can be applied to the modelling of a certain area, with hopefully some results.

Vegetation Modelling - Looking to the Future

Having looked back at how species responded to the changing climates of Northern Hemisphere glaciation and subsequent warming it is now important to look forward to how this may help us model how species will react to future climate change. Historical records are important as they can provide the background for predictions for the future. As shown in a previous post, ‘The Bioclimate Envelope Model - outdated and simple?’, the bioclimate envelope model is still one of the most commonly used models, although it lacks the complexity it simulates. Looking forward, Dynamic Global Vegetation Models (DGVMs) will provide greater capacity for accurate predictions of global shifts of species. 

Some challenges that these models face include (Neilson et al. 2005): 

• ‘Accurately estimating the importance of long-distance dispersal in the establishment and proliferation of species within new vegetation communities.’
• ‘Aggregating information from individual species into the category of plant functional types (PFTs).’
• ‘Incorporating temporal and spatial heterogeneity within large spatial grid cells into the modeling of migration.’

One factor that has to be taken into account is how the warming will affect the ecosystem alteration. Ecosystems tend to change species composition by in situ conversion, where the subdominant species replace the dominant ones (due to different tolerances) or via migration of species from the local region. Adaptability of ecosystems to rapid changes appears to be related to the diversity of species traits that are available to assume old and new functions (Loreau et al. 2002). The more adaptive an ecosystem is the less likely there will be sudden changes in ecosystem function. An example of this can be seen in the southwestern United States with the replacement of the regionally dominant pinyon pine (Pinus remota) with another species (Pinus edulis) during the glacial to interglacial transition (Lanner & Wan Devender 1998). The shift of dominance may have been large but the ecosystem retained its function during the climatic change.

DGVMs are conceived to merge vegetation distribution and ecosystem process models (Cramer et al. 2001). Recent advances in the capabilities of DGVMs have been impressive and include dynamic simulations of vegetation distributions for national and global assessments of climate change (Cox et al. 2000). There have been difficulties with spatial and temporal scales, seeing as current DGVMs often operate with a grid cell size of approximately 50 km on a side, much larger than local dispersal distances. Also, many of the most important processes (e.g., photosynthesis, flowering, and seed set) are also faster than the model time step, which is typically daily to monthly, and must therefore be scaled up. However there will always be a threshold beyond which certain parameters can be simulated. 

DGVMs should be the primary tool for decision makers to evaluate the question of safe rates of climate change, and impacts on carbon and other biosphere feedbacks to the atmosphere. However, migration must also be incorporated into the models. The entire process of migration is lacking in observational data for use in the vegetation models. Data required includes: the processes from flowering and seed set to short and long distance dispersal, establishment and growth, and then to completion of the cycle. Data needs include: quantification of initial establishment and early growth in the face of competition and other biotic interactions. 


Incorporating an expanded concept of PFTs in DGVMs including migration and in situ conversion will be important to model future global ecosystem and species shifts. I have looked at the negative aspects of DVGMs in this post, even through they do represent the most accurate modelling technique for future use. In the next post I will show an example of how these models can be used.

Final Conclusions on Glacial Refugia

The post will attempt to give a well rounded review of the debate over cryptic glacial refugia in the Northern Hemisphere.

The consensus is that glacial refugia existed in the mountains of the mediterranean peninsulas in Iberia, Italy and the Balkans. This is where temperate tree species are thought to have persisted during full glaciations. However, there is another theory that cryptic glacial refugia existed in parts of northern Europe. For example the tree pollen records in northern Europe in the early postglacial period suggest huge migrations rates of trees of up to 1500m per year if they colonized from southern refugia. The theory of cryptic refugia involves some tree populations being able to survive in favourable microclimates during glaciations where the regional climate would not normally be suitable (Stewart & Lister 2001). How the population manage to survive are not fully understood, and many scientists (Tzedakis et al. 2003) put the presence of the early post glacial pollen to isolated long distance dispersal events.

Examples of cryptic populations can be seen today in certain areas in the current climate. In the central Sahara a total of 223 individuals of the conifer, Cupressus dupreziana, managed to persists in ravines, where rainfall averages at 30 mm per annum (subject to large variability) (Abdoun & Beddiaf 2002). This is similar in the way in which certain temperate tree species may have been able to survive in northern Europe during glaciations.

Techniques used to identify refugia are: pollen analysis, macrofossils and molecular techniques. Pollen analysis can be used to show the presence of species in an area. For species with low production and that are non wind pollinated, presence of the species can be concluded with confidence, however with wind pollinated species the pollen could have been produced outside the regional area. The records are quite vast and are an important tool in assessing species presence.

Macrofossils are the most precise way to identify species. Macrofossils have been recovered from paleolithic sites in northern Spain, dated between 17,000 and 15,000 years BP revealing a diverse assemblage of temperate and mediterranean species. This suggests a glacial refugium was present. Although this suggests the prescence of tree species, it does not make clear the type of vegetation (e.g. forest, woodland, steppe, steppe-tundra). Willis et al. (2000) studied macroscopic charcoal particles in glacial sediments in Hungry. They concluded that there were ‘oases’ in central Europe where temperate flora and fauna could survive the glacial conditions.

Molecular techniques involve the phylogeographic techniques discussed in a previous post. They involve studying, amongst others, species diversity as identifier of refugia. See ‘Glacial refugia - are they being too cryptic’ for a refresher on phylogeographic techniques. An example of how it is used is the prescence of Scot’s pine, Pinus sylvestris, in Scotland. High nuclear genetic diversity in scottish populations of P. sylvestris suggest a cryptic glacial refugium (Sinclair et al. 1999) during the last glaciation.

The evidence suggests that plant species persisted during glaciations in cryptic glacial refugia. While some academics still believe post glacial recolonization took place by long distance dispersal from southern refugia, I believe the cyptic glacial refugia hypothesis to be a better fit. 

The hypothesis manages to explain the huge migration rates had been inferred from colonization from southern refugia, which were highly unrealistic. This breakthrough has impact on future modelling of vegetation change due to climate change. After understanding how glacial refugia exists and the debates over post glacial colonization, the blog will move on to looking at the future and how vegetation modelling will be important for assessing climate impacts on species migration.

Diversity Estimators for Locating Glacial Refugia

This next post is going to focus on how species diversity is assessed in order to infer areas of glacial refugia. A common assumption is that glacial refugia harbor higher levels of genetic diversity than areas of subsequent colonization. This is the basis on which many glacial refugia are located. However Comps et al. (2001) believe this to be too simplistic, as the way in which genetic diversity is estimated can influence the results. To give a brief background of Comps et al. (2001), they studied European beech (Fagus sylvatica) across Europe. As F. sylvatica is wind pollinated a detailed pollen record is available in which the genetic data can be compared against. In addition it was chosen as its closest relative is Fagus orientalis so there are no problems of interspecies introgression.

The question is what pattern of diversity is expected from recolonization of a refugium? Colonizations of new areas normally consist of populations of a few individuals therefore showing a small sample of the genetic diversity of the source population. How much of a reduction is dependent of diversity estimators. For the new population, allelic richness (which is the number of alleles per population) would be expected to be lower than genetic diversity (which is the probability that two alleles sampled at random from a population are different). In F. sylvatica, allelic richness was lower in populations in Northern Europe, consistent with loss of rare alleles due to recolonization. 

However a reduction in genetic diversity was not found. Comps et al. (2001) found a negative correlation between allelic richness and genetic diversity. This meant areas with a decreased allelic richness displayed an increased genetic diversity. Since the majority of papers only calculate genetic diversity, academics may be misled in the identification of glacial refugia.

Personally I take the findings of Comps et al. (2001) with a pinch of salt. I have used this example not to show that a wrong estimator is being used, just that other estimators are available and might give slightly different results. I believe that the increase in genetic diversity when allelic richness decreased was a conclusion unique to the species in question and general conclusions should not be taken from this case study. It does bring an extra addition to the glacial refugia debate and I believe that diversity estimators should be taken into account when discussing methods of glacial refugia identification.

Glacial Refugia - A Refresher and Possible Explanation of Some of the Flaws

As previously stated, this blog will now focus on two main aspects; glacial refugia and species migration models. To kickstart this, a refresher to glacial refugia is required. During glaciations, temperatures plummet and aridity increases. Biologists often hypothesize the fate of temperate fauna and flora during these periods. Refugia describes the areas in which these species inhabit during the full glacial conditions, and often determine the current patterns of biodiversity we see in temperate species. The precise locations and impact on present day distribution is often under debate and new information is constantly being found that enhances our understanding.

Willis & Whittaker (2000) propose that the function of glacial refugia in temperate latitudes is to protect species diversity, while in tropical zones its primary function is to encourage speciation. It has also been proposed that glaciation can complete speciation events inaugurated in earlier geological time. An example of this can be seen in the formation of three new species of black-throated warblers in North America during the last glacial period. Studies of the Amazon basin have led to questioning of the glacial refugia hypotheses. Increasing aridity would lead to the wide scale change to savanna from lowland tropical rainforest. Under the hypothesis this would lead to small refugial belts in the mountains where the conditions would be wetter. However study of pollen records from the Amazon basin have shown that the area was not replaced with savanna and the forests have dominated throughout the glacial period. It has been suggested that the effectiveness of cold-stage aridity to roll back the rainforest has been greatly overestimated. Study of genetic diversity of the lowland canopy tree Poulsenai armata in Central America has shown a greater within-population diversity than would be consistent with postglacial expansion from a distant refugia.

Another twist in the Amazon-refugia debate is the distribution of endemic species within lowland Amazonia. It is hypothesized by Nores (1999) that this patterning of endemic species is the product of sea level rises of over 100 metres during the interglacials of the Quaternary and late Tertiary periods. This led to the formation of two large islands and several smaller archipelagos encouraging speciation through geographic isolation. It is important to always question the current paradigm as this is the way in which scientific discoveries are found. What is generally accepted as true is not necessarily so, only the best possible explanation of the current evidence, and I feel it is important to question areas that do not fit in with this paradigm, in this case the example of Central American forests and Amazonian species diversity.

The debate over glacial refugia is still in its infancy and will be delved deeper in subsequent posts of this blog. The next post will focus on how species diversity is assessed and what impact this has on determining glacial refugia.