SPEED Lab Research Projects

The driving general questions being addressed by the SPEED Lab include:

  1. How does external heterogeneity in the environment affect population dynamics and the spread of infectious disease?
    • How does this change as the spatial scale of the heterogeneity is varied?
    • How does this change as the temporal scale of the heterogeneity is varied?
  2. How do the above results change as the spatial scale of population interactions vary? Local interactions are used in the baseline models, but some long-distance interactions are also sometimes included.
Many variations of models are explored in pursuit of the above questions, e.g. interactions between different pure dispersal strategies or between mixed strategies; populations on landscapes with short-term and long-term disturbances occuring on two different spatial scales, and so on.

Our models have a number of binary divisions:

  1. Computational versus mathematical
  2. Population ecological versus epidemiological
  3. Lattice-based versus household-based
  4. Discrete-time versus continuous-time
Our projects have covered both branches of the above splits, to varying extents.

Ongoing projects

Descriptions of projects will be added here later.

Local dispersal on fragmented / clustered heterogeneous landscapes

The image below shows a population on a landscape map. The map is divided into small sites (pixels in the image), each representing a small area of land. Each site can be in one of three states (colors): occupied by the species of interest (red), empty suitable habitat (white), or empty unsuitable habitat (black). Unsuitable sites represent habitat the population cannot survive in (e.g. wrong soil type for a plant, or human-disturbed/developed habitat, etc.).

locally-dispersing population on
a heterogeneous landscape with gradient of habitat clustering

In the map, 50% of the sites are suitable, and 50% are unsuitable, across the entire landscape. Local clustering of the suitable sites varies from 0.3 at the left edge to 0.97 at the right edge. Local clustering is measured by imagining that you are standing on a suitable site, and asking what is the probability that a neighboring site is also suitable.

When the landscape is more fragmented (on the left), the population cannot persist. As the habitat is more clustered (on the right), the population survives at fairly high density, even though the habitat loss is the same across the entire landscape.

If you click on the image above to open the magnified version, you can also see that in the right half of the landscape, there are small patches of white (suitable) sites which are empty. The population has gone extinct on these small patches, and cannot recolonize them because in this model, the population reproduces using only local dispersal. Whenever an occupied site reproduces, it drops its offspring on a neighboring site.

Publications 7, 8, 14, and 15 on my Publications page describe models of populations on heterogeneous landscapes similar to the one shown above.