UC Berkeley

A major goal of landscape genomics is to understand how spatiotemporal variability in complex environments influences evolutionary dynamics, and consequently geographic patterns of genomic diversity, in natural populations. Recent computational advances enable this spatiotemporal complexity to be described and analyzed in unprecedented detail. One such advance is the improvement of forward-time landscape genomic simulation, allowing arbitrarily complex evolutionary scenarios that mimic real-world systems to be created and studied in silico. In chapter 1, I present Geonomics, a new, user-friendly Python package for performing complex, spatially explicit, landscape genomic simulations on changing landscapes and with full spatial pedigrees. I describe the structure and function of Geonomics in detail, show that its results are consistent with expectations for a variety of validation tests based on classical population genetics, then demonstrate its utility and flexibility with example scenarios featuring polygenic selection, selection on multiple traits, and non-stationary environmental change on realistic landscapes. Taken together, these tests and demonstrations establish Geonomics as a robust platform for population genomic simulations that incorporate complex spatiotemporal dynamics.

In chapter 2, I apply the software developed in chapter 1 to one of the major areas of theoretical and applied interest in landscape genomics: the evolutionary consequences of climate change. I model climate change realistically, as the decoupling of historical environmental gradients that generates novel multivariate selective environments. I then simulate evolutionary responses to that climate change event across a range of genomic architectures, defined as the full factorial crossing of discretized levels of three key architectural components: the number of genes per trait (polygenicity), the recombination rate between neighboring genes (linkage), and the number of distinct genotypes generating identical phenotypes (genotypic redundancy). I use the results to test a series of hypotheses about the influence of polygenicity, linkage, and redundancy on gene flow, maladaptation, and demographic decline. Results show that a commonly assumed mechanism of evolutionary rescue, adaptive gene flow from populations whose current climates approximate future projection, can be less effective than in situ adaptation under some architectures, likely because of maladaptive introgression caused by the decoupling of environmental gradients. I also find that high polygenicity aggravates maladaptation and demographic decline, a concerning result given the likely polygenic nature of many climate-adapted traits, but that higher genotypic redundancy increases adaptive capacity across all scenarios, adding to the growing recognition of its importance. Overall, this chapter shows that genomic architecture, though it is often ignored, can exert large influence over the effectiveness and relative magnitudes of adaptive gene flow and in situ adaptation in a spatially distributed population subjected to climate change.

Another major computational advance facilitating the study of spatiotemporal evolutionary dynamics is the advent of massive and distributed geocomputation. In chapter 3, I use this tool set to study, in unprecedented detail, global geographic variability in the seasonality of terrestrial plant productivity – i.e., land surface phenology (LSP). Not only does the geography of LSP convey critical information about environmental controls on plant function and carbon cycling, but it has important implications for evolutionary biogeography: spatial asynchrony in LSP can indicate the potential for spatial asynchrony in reproductive phenology, and thus for increased genetic isolation and divergence between conspecific populations. Thus, whereas chapter 2 provides an example of the spatial nature of an evolving system influencing its temporal dynamics, this chapter provides an example of the less-appreciated inverse situation: the potential for a system’s temporal complexity to influence its evolutionary dynamics. Despite its importance, LSP research lacks mapping methodologies that can characterize the full diversity of terrestrial phenologies, and LSP asynchrony mapping is even less developed. Here, I develop a multivariate, generalized, and robustly-validated LSP mapping methodology, based on simple harmonic regression, then apply it to a 10-year, 0.05° dataset of MODIS near-infrared reflectance of vegetation (NIRV, a proxy of plant productivity). This produces a global LSP map that reveals surprising diversity, including both regional patterns of heterogeneity that are corroborated by prior research and intercontinental patterns of convergence that recapitulate major bioclimatic and biogeographic gradients. Next, I calculate and present a global map of LSP asynchrony, and use machine learning to explore regional variability in its potential climatic and physiographic drivers. I describe LSP asynchrony hotspots in the world’s five Mediterranean climate regions, where asynchrony appears driven by precipitation asynchrony and spatial variability in vegetation structure, and in tropical montane regions, where minimum temperature asynchrony and precipitation asynchrony appear to be interacting drivers. Lastly, I use an ensemble of regressions within global high-asynchrony regions to demonstrate that phenological asynchrony between climatically similar sites is most frequent at lower latitudes, supporting the notion that phenological asynchrony is most likely to cause allochrony and consequent evolutionary divergence in the tropics.