Teaching in Yellowstone
David W. Roberts

Department of Ecology

Montana State University
droberts@montana.edu


I am a vegetation ecologist interested in:
  • vegetation theory

  • simulation modeling

  • multivariate analysis of ecological communities

Teaching

As I currently serve as Department Head, my teaching assignment is somewhat reduced. My primary teaching assignment is one undergraduate-level course and two alternating graduate-level courses:

Software

I am the author of several R packages for ecological analysis.

These packages are available at the Comprehensive R Archive Network here

Recent and Current Research Projects

Grand Teton National Park Vegetation Data Management System

As part of an effort to develop a simulation model of the future vegetation of Grand Teton National Park (see just below) I have undertaken a systematic revision of the basic data on vegetation of the Park (and nearby National Forests) using a combination of the GRASS GIS system, the PostgreSQL database system, and R statistical software. The base data consists of 745 detailed vegetation sample plots, and 1122 accuracy assessment plots with reduced detail on vegetation composition. The vegetation map itself consists of more than 49000 polygons of 52 specific map codes.

PostgreSQL is an extraordinary open-source relational database that links directly to GRASS and R, so that data are held in a single location, with spatial operations managed by GRASS, statistical calculations managed by R, and full database query capability in PostgreSQL. The current SQL schema for the project is available here

The integration of the three software components allows me to use the best tool for each task while managing the data as a single entity. GRASS provides an effective integration of raster and vector data, and is connected directly to the PostgreSQL database for point and polygon attributes. While R is capable of working with spatial data I find the GRASS data storage superior for large raster data sets, and the GRASS geographic and imaging routines simpler for routine use. While PostgreSQL can be spatiallly enabled into PostGIS, by storing the polygon spatial data in GRASS rather than PostgreSQL the database tables are directly readable and writable in R, which I use for any aspects of quantitative analysis.

The GRASS GIS system allows comprehensive script-based management of spatial data and imagery.

Simulating the Future of Grand Teton National Park

I am developing a GIS-based simulation model of the future vegetation of Grand Teton National Park. The project grows out of a previous Park-directed effort map the vegetation of the Park (see here). The base data are described just above

The model is a state-and-transition model written in FORTRAN interfaced to the GIS for both ARC and GRASS. Each polygon follows a type-specific successional trajectory subject to type- and time-specific disturbance regimes. Climate change is implemented as a change in base environment, which in turn can change the underlying vegetation type.

Solid lines (with numbers greater than one) represent successional development in years; dotted lines (with numbers less than one) represent disturbance events in probability of occurrence in a given year.

Work is still in progress and interim reports will be posted here.

Terrestrial Ecosystem Unit Inventory of the Beartooth Mountains

The Beartooth project is what the US Forest Service calls a "Terrestrial Ecosystem Unit Inventory." This is an integrated potential vegetation/soil classification that employs sample plots with full vascular plant species abundances and full soil pits. Our work is on the Custer National Forest section of the Beartooth Mountains, and includes 511 sample plots collected through a collaboration of the Custer National Forest and the LabDSV at Montana State University. We have recently the final ecological type classification and mapping. The final report is available here.