Species Distribution Modelling Using An Open Source Geospatial Software Stack

Presentation | Presented

  • Allan Hollander, Information Center For The Environment, University of California, Davis

An important task in conservation planning is mapping the distribution of rare
and threatened species. Field observations used to produce such maps are
almost always scant in number, so it is often useful to generate these maps
using modelling based on the spatial pattern of association of the observations
of the species with environmental variables such as climate or soil conditions.
Using examples from mapping species statewide in California, this presentation
will illustrate how such modelling can be carried out easily using an open
source geospatial software stack. The major tools used in these analyses are
GRASS for storage and manipulation of species observation points and
environmental data layers, StarSpan for efficient point sampling of the
environmental layers, and the software suite R for statistical modelling. In
particular, an interface (spgrass6) linking GRASS to the spatial capabilities of
R allows for rapid experimentation with a variety of statistical modelling
techniques. In R one can make changes such as using a different subset of
environmental variables, or switching from logistic regression to classification
trees, and in GRASS immediately visualize the change in the predicted
distribution maps. The presentation will also address workflow issues -- given a
burgeoning open source tool box, how does one integrate these tools to
efficiently accomplish a modelling task such as producing a large set of species
distribution maps.

Supporting Files