Home - PNHP/Regional_SDM GitHub Wiki

Welcome to the Aquatic Regional_SDM wiki!

Modeling Background and Approach

The modeling approach presented here is the aquatic analogue to the methods developed by the NY, PA, VA, and FL Natural Heritage Programs for distribution modeling of aquatic (lotic) species using Random Forests. Compared to our terrestrial approach which is based on 30m pixels, we use the "stream reach" as the equivalent modeling unit. Therefore, if a source feature intersects with (is assigned to) four reaches, then your “polygon” equals those four contiguous reaches.

Overview of steps

Here is the big-picture view of the steps we'll be using to process and run our data. Preprocessing steps that largely take place outside of the R environment include:

1. prepping the species data: PreppingSpeciesData
2. prepping the environmental data: PreppingEnvironmentData

The model is run within Random Forests in R, using the follow steps:

1. create sampling points in polygon location data: process_polys.
2. attribute points with environmental variables: attribute_points.
3. create random forest model: run_model.
4. run a predict of model for study area: predict_suitability.
5. QC, check for range exclusion areas:
6. run and export validation statistics:

Additional useful information:

• Setting up and using git for this project is outlined here git_setup.
• For a discussion of file and folder organization, go to the organize_files page.
• For pre-modeling assumptions, go to the pre_model page.