Habitat Modeling - Overview
Overview - Habitat Modeling
Habitat modeling, also referred to as “species distribution modeling”, “niche modeling”, “habitat suitability modeling”, and other names, is the process of mapping where a species (or other spatially determined element) is likely to occur. GIS is a central component in this process, as you can imagine, and so we focus here on the role GIS plays. We also, necessarily, cover a bit of the statistical analysis involves. This is to give proper context to the GIS analyses we perform and what assumptions or limitations they incur or free us from in performing habitat classification. It is important to note, however, that you should not be content with the level of statistics covered here; rather, if you continue to pursue habitat modeling (or whatever you want to call it), you should invest more time understanding the nuances of the various approaches so that you don’t misuse them.
In this section, our lectures cover the statistical approaches to habitat modeling as well as model evaluation. The specific learning objectives are outlined below.
Learning Objectives
| Topic | Learning Objectives |
|---|---|
| §3.1 - Section Intro [PPT] [Recording] |
• Explain what habitat modeling is and how recent technological advances have improved habitat modeling. •Describe the dangers of making habitat modeling easy to execute |
| §3.2 - Modelling Approaches [PPT] [Recording] |
• Describe the utility of habitat modeling in conservation & land management • Explain the term “Hutchinsonian Niche” in the context of describing habitat • Explain the concepts: “Ecological Model”, “Data Model”, “Statistical Model” • Describe the difference between “geographic space” and “parameter space” • Describe the different kinds of observation data: “habitat”, “non-habitat”, and “available habitat” • Describe how statistical logic responds to different kinds of observation data • Contrast “generative” (aka “rule-based”) vs “discriminative” (aka “statistical”) models • List the benefits and drawbacks of the following generative approaches: - Envelopes - Mahalanobis D2 • List the benefits and drawbacks of the following statistical approaches: - Discriminant Functions Analysis (DFA). - Generative Linear Models (GLMs) & extensions - Classification and Regression Tree Models (CART) & extensions - Maximum Entropy (MaxEnt) • Articulate the ecological implications and uses of statistical models |
| §3.3 - Model Evaluation [PPT] [Recording] |
• Explain what is means to “validate a model” • Describe what data are needed to validate a model • Explain how different model types require different types of validation • Contrast the terms “reliable” and “discriminant” in the context of model performance • Explain how accuracy, rationale, and interpretability are important in model outputs • Compute a confusion matrix using validation data and your model • Identify the false positives and false negatives in your confusion matrix • Compute Accuracy and the Kappa Statistic from your confusion matrix • Explain what the Kappa Statistic shows that Accuracy doesn’t • Explain sensitivity and specificity in terms of a model’s discrimination performance • Use the ROC approach to determine the optimal probability threshold • Explain instances were you might deviate from the optimal probability threshold • Use misclassified data to better understand your analysis |