Habitat Suitability Modeling with Maxent

In this lab, you'll be using the software package Maxent to model a species potential habitat.

Learning Outcomes

• Learn to use Maxent for habitat suitability modeling
• Evaluate models using AUC values

The MaxEnt Tutorial

The main web page for MaxEnt is at: Maxent software for modeling species niches and distributions

If you scroll down, you'll see a link to a tutorial for MaxEnt. This is an excellent tutorial that uses data available on the web site. Go through the tutorial. You don't have to worry too much about the "Output Formats" as the logistic is the most commonly used and you can use it for this lab.

The sections "Replication" and "Regularization" are important. You'll want to try different values for Regularization to minimize over-fitting of data. Don't be shy about increasing the values up to 20 or 50 to see the effect and then finding an appropriate value for your model based on a balance between "fit" and "complexity".

The sections on "SWD" and "Projecting" will show you how to use MaxEnt to create a predicted model for another scenario, such as climate change. This section is not required for this lab but could be exciting to come back to.

Definitely open the "lambdas" file that will be in the output folder and take a look at the number of parameters. You can decifer the lambas file yourself using the information in the ENTTools paper or use the "MaxEnt Lambas File" option in the "Utilities" menu in BlueSpray. For most biological phenomenon, if you have more than about 20 parameters, I believe the data is being over fit. This will vary based on the species and the number of predictors.