Lab 9 - Model Ensemble

USING AN ENSEMBLE OF MODELS

 

Thus far we have learned how to use an Information Theoretic Approach to model selection, learned some techniques for deriving metrics at the landscape scale, and generated inductive Species Distribution Models using three separate algorithms: Logistic Regression, Random Forests, and MaxEnt. I hope that you have found these techniques useful. But the question remains, "What do we do with these models?, Which model is the best? And How do I choose a model result?" I don't have an exact answer, but during this lab we will briefly explore some of the ways in which one might combine and present output from multiple SDM's.

This lab is done entirely within Program R. Download this file, SDM_Compare.r,  and we will work from it.

 

 

HOMEWORK ASSIGNMENT

 

For this assignment I would like for you to provide a written document providing your professional opinion. The document should be written in a 'White Paper' format (i.e. professional within-department memo) and can be between one and two pages in length. Please address the following items as if you were providing guidance to a group of your peers who were tasked with determining the best approach to answer these wildlife management questions. Please use the Pariette Cactus as our species of consideration. Recall that Pariette Cactus is listed under the Endangered Species Act and that its habitat happens to be located where intense Oil & Gas Development is occurring. This is not a spatial question, per se, but you may find it useful to know that for each well permit, Pariette Cactus cannot be within 300 feet of the disturbed area:

  • Given your experience with the three algorithms (Logistic Regression, Random Forest and MaxEnt):
    • Which algorithm would you choose to create a species distribution model? Why? (i.e. inference, prediction, etc.)
    • If our goal was to delineate areas for a particular species distribution which would trigger an automatic Section 7 consultation and/or mandatory monitoring for species occurrence, how/where would you define that boundary?
    • What other, if any, considerations should be taken into account for creating the management strategy based upon your species distribution model(s)?

      This assignment is due in my email inbox by 5pm on April 3, 2014.