Missouri Botanical Garden

Post-Doc, Center for Conservation and Sustainable Development

Other Affiliations: Louisiana State University, Biological Sciences

Louisiana State University

Thesis Title: Environmental, Stochastic and Historical Determinants of Taxonomic and Phylogenetic Gradients. The Use of Computer Simulation Models

About

Describing and explaining patterns of spatial variation in species richness at broad geographic extents has been a major focus of research for ecologists and evolutionary biologist since the work of Alexander Von Humbolt in the early 1800’s (Brown and Sax 2004). During these 200 years of research, a large number of explanations have been proposed to account for patterns of spatial variation across broad geographical areas (e.g. Currie 1991, Colwell and Lees 2000, Wiens et al. 2006, Araujo et al. 2008). With time, these possible explanations have been accumulating; in 1966 the main recognized mechanisms were only six (Pianka 1966), but they grew to about 30 by 2003 (Willig et al. 2003). Although there has been significant progress in our understanding of the mechanisms that give rise to spatial gradients in diversity, it has been very difficult to reach an agreement about which of these hypothesized processes are indeed responsible for generating and maintaining species richness gradients. From my point of view, there are at least three reasons why we have not been able to come up with a well accepted model to explain broad patterns of species richness variation.

A first major source of difficulty is the almost complete dependence on observational studies to test macroecological hypotheses. In vivo, in vitro and field experiments have played a major role in our understanding of biological processes from cells to communities. However, performing traditional experiments at the spatial, temporal or taxonomic scales that are meaningful to study geographic species richness gradients is virtually impossible (Brown 1995). Additionally, how results from small scale experimentation apply to such broad scales is not clear, and frequently not appropriate (Kerr et al. 2007). However, we can take advantage of computer simulations to perform in silico manipulations and experiments of virtual systems that can then be compared with empirical data (Peck 2004). Computer simulations are playing a major role in understanding a number of fundamental processes in many branches of science. Within ecology, simulations have also played an important role and are increasingly used. Surprisingly however, computer simulations and simulation experiments have been mostly ignored by macroecologists, and only recent work has been made to take advantage of this technology in the study of macroecological and biogeographic problems (e.g. Bokma et al. 2001, Stevens et al. 2003, Brayard et al. 2005, Davies et al. 2005, Rangel et al. 2007). Computer simulation will likely play a major role in the future study of diversity gradients (Gotelli 2008).
Additionally, the use of appropriate null models will greatly add rigor and push forward our understanding of species richness gradients (Gotelli and Graves 1996). Recent ecological research has shown that there are well known spatially explicit processes that are capable of creating structure in ecological systems without the need to incorporate additional ecological complexity (Colwell and Lees 2000, Hubbell 2001, Bahn and McGill 2007, Arita and Vazquez-Dominguez 2008, Bahn et al. 2008). In general, these models are based on well known principles like speciation, extinction, dispersal, or demographic stochasticity, and typically treat all species or individuals as equivalent. Relatively simple models like these should be thoroughly studied to understand the scope of their explanatory power. Only then, these models can be used as null hypotheses to investigate more complex models that incorporate additional mechanisms. However, most macroecological studies still use simple and naïve null hypotheses of traditional statistical analysis to test hypotheses.

Finally, many of the hypothesized mechanisms can produce similar species richness gradients; frequently, the exact relationship between predictors and richness is not specified by the hypotheses, but instead parameters to provide the best fit are calculated from the data. Consequently, the match of a variable or group of variables that represent a particular mechanism with empirical richness is a fundamental test, but it is also a weak test. Stronger predictions are possible. Testing a model parameterized with one data set to explain other data sets is a possibility (e.g. Francis and Currie 2003, Kalmar and Currie 2007). Also developing hypotheses with a priori expectations about parameter values are extremely helpful; an example of such a hypothesis is the metabolic theory which predicts not only the direction of the relationship between richness and temperature, but also the slope of the relationship (Allen and Gillooly 2006, Hawkins et al. 2007, Gotelli 2008). Finally, mechanisms that drive diversity gradients should be able to account for species richness variation, but also for other patterns that can arise from the same mechanisms. This is the principle of pattern-oriented modeling (Grimm et al. 2005, Rangel et al. 2007), and can be used to provide stronger tests of macroecological hypotheses.

I will build my dissertation around these three shortcomings and will take opportunity to use its various chapters to address them. In this dissertation, I will use computer simulations based on fundamental evolutionary and biogeographic processes to test hypotheses about determinants of species richness gradients across continents. The first chapter uses a case study to investigate how simple stochastic simulation models can explain species richness gradients. The second chapter expands on these simulations to quantify the species-environment relationship expected by stochastic diversification. Chapter three uses data from various independent species richness gradients to test whether there is really evidence for environmental determinants of species richness taking into account appropriate null models defined by chapter two. Chapter four contrasts simulation models where niche conservatism and environmental variables play different roles in determining richness gradients. Finally, chapter five addresses whether stochastic and environmentally driven simulations can reproduce other macroecological patterns, particularly phylogenetic structure.
If one is interested in building a model of the construction of species richness gradients, one needs to define the proximal factors that determine the number of species present in a particular location. Just like population size is determined by the number of individuals that are born, die, leave from or arrive at a local population, the number of species present in a given place depends on equivalent evolutionary and biogeographic processes:

S=speciation-extinction+colonization-emigration

In this context, species are added to a given region by speciation events that occur within the region and by dispersal of species from surrounding regions during range expansion or range movement; species are removed from a region when a species goes extinct from that region by range contraction or fragmentation, or from emigration caused by the movement of the distribution of species elsewhere. Hence, at broad geographic extents, patterns of variation in species richness depend on spatially explicit processes of diversification (speciation and extinction) and distribution (range movement and range expansion/contraction) of species within a clade. The work of the macroecologist is to understand how environmental, historical and stochastic forces control these proximal processes and consequently control species richness gradients.

Contact Information