Vertebrate Modeling








Vertebrate Database and Model Review

Accepting external review of vertebrate database and spatial models.

Reviewer's will need Microsoft Access and an internet connection in order to access the Reviewer Database (see below). The database is a tool for reviewers to assess species ' models by examining individual model parameters. The interface has been developed to guide reviewers through the process and summarize responses using the Bayesian framework. The database and a user manual are available below. Please read through the Reviewer Manual and review the contents of this webpage to familiarize yourself with the SE-GAP dataset. Hardcopy materials are available, if needed.

If you are interested in hosting a reviewer workshop at your organization, please contact us to arrange.


Reviewer Resources:

Draft Species Maps:

Vertebrate Modeling Database Amphibian Species
Reviewer Manual Avian Species
Ancillary Data Metadata Mammalian Species
Vertebrate Modeling: Process & Review Reptilian Species


For More Information:

  Contact: Steve Williams   Contact: Matt Rubino  
  Vertebrate Mapping Coordinator   Biologist, GIS and Database Specialist  
  220 David Clark Labs   214 David Clark Labs  
  Biodiversity and Spatial Information Center   Biodiversity and Spatial Information Center  
  Department of Biology, NCSU   Department of Biology, NCSU  
  Raleigh, NC 27695-7617   Raleigh, NC 27695-7617  
  919-513-7413   919-513-7280  
  steve_williams@ncsu.edu   matt_rubino@ncsu.edu  


Overview of Review Process

The purpose of model review by external experts is both to inform the process with which models are developed and potentially revised, and to provide users confidence that species models are accurate and useable within the scale and context they are intended. That being said, the review process itself needs to be kept relatively simple and concise while at the same time assessing major model components and overall performance.

We have chosen to conduct model review in a Bayesian Belief Network Framework so model development and review are more transparent. This allows both modelers and users to understand which aspects of the model are stronger or weaker than others and quantifies the assessment for comparison and usability. Initially, model developers ranked model components a priori (Figure 1). This internal appraisal can then be compare to outside assessment.


Figure 1. A Priori Rankings Generated by Modelers.



Ranking Model Components

Reviewers are asked to respond to six questions ranking their level of agreement on a scale of 1 to 5:

  1. The delineated geographic range is an accurate representation of the known range extent of the species (Figure 2).
    We are defining the known range as the geographical area that is regularly used by a species. It does not include transient or migratory sightings.

  2. Figure 2. Species' Range Map.

  3. The parameters used to model the predicted distribution of the the species accurately represent its habitat requirements in the Southeastern United States (Figure 3).
    In otherwords, do you agree that the selection of map units and ancillary data parameters are the best 'formula' for a species habitat?

  4. Figure 3. Selection of habitat model parameters.

  5. The spatial data adequately represent the species' habitat requirements (Figure 4).

  6. Figure 4. Spatial data examples.

  7. The mapped predicted distribution adequately represents the distribution of the species' habitat within the identified range extent (Figure 5).

  8. Figure 5. Species' Predicted Distribution Map.

  9. The published literature adequately documents the breeding habitat requirements for the species in the Southeastern United States.
  10. I am an expert in the natural history of the species.

Combining and Evaluating Scores

Within the Bayesian framework, reviewer responses to the first five questions are weighted by the reveiwer's self-confidence to produce a weighted composite score for a given model (Figure 6).

Figure 6. Bayesian Belief Network Model Score Weighting.
 



Return to Top

Last updated: October 14, 2008