PACE SDM TEAMS

Our project is to develop a basic workflow for Species Distribution Modeling (SDMs) and Boosted Regression models using PACE data as predictors. The goal is to collaborate on getting the basic steps programmed using 3 projects in 3 different regions as motivation.

The basic approach is the following:

graph LR
  A[Observation data - dataframe w time, lat, lon] --> G{matchup obs to xarray}
  B[Predictor data - xarray] --> G
  G --> C[dataframe w obs data, abund or 0/1, and predictor variables]
  C --> D{XGBoost - boosted regression tree - for modeling}
  D --> E([Visualize - make maps of suitability])
  D --> F([Feature Importance])

Loading

Projects

• Chesepeake Bay oxygen: Oxygen front
• Gulf of Maine squids: Shelf Science Squidsters
• Canary Current upwelling region sardines: FishTok Guinea

Collaborators

List all participants on the project. Here is a good space to share your personal goals for the hackweek and things you can help with.

Name	Team	GitHub	Things I bring	Affiliation
Artem Dzhulai	NES	adzhulai	Python, Git, general oceanorgaphy	Univ of Rhode Island
Dante Horemans	Chesepeake	dantehoremans	R, Python, Fortran, ROMs, ML/AI, estuarine/coastal ecology	Virginia Institute of Marine Science
Eli Holmes	Floater	eeholmes	Project Helper; Git; Python; SDMs	NOAA; Univ of Wash; OceanHackWeek
Frederic Bonou	Canary Current	FREDERICBONOU		National Univ of Sciences, Technology, Engineering and Mathematics, Benin
Jiang	Chesepeake
Jing Tan	Chesepeake	jit079	Python, Git, Remote-sensing data, ML/AI	Scripps Institute of Oceanography
Natalie McCourt	NES	spacenatalie	Python, Git, netCDF	UMBC
Sajna Hussain	NES	Python, fisheries, SDMs	Oregon State Univ, CICOES
Haley Synan	NES	hsynan	Python, ML, general oceanography	NOAA Fisheries/IBSS
Jamon Jordan	NES	justjamon	Python, seascapes, SDMs	Oregon State Univ, CEOAS

Data and Methods

Response Data

We need observations of presence/absence, abundance, or level (oxygen) with date/time, lat, lon. We need a large number of points distributed across a variety of regions in our study regions.

Explanatory Data

Plan to explore PACE products.

• Light penetration. We'd need to compute.

Proposed methods/tools

XBoost

Additional resources or background reading

Optional: links to manuscripts or technical documents providing background information, context, or other relevant information.

Tasks

• Sunday - organize and get on GitHub
• Tuesday - get our observation data together
• Wednesday - get our response data together
• Thursday - get XBoost working
• Friday - put presentation together

Project Results

Documentation

Build

mkdocs gh-deploy

First time make

pip install mkdocs mkdocstrings[python] mkdocs-material mkdocs-jupyter
pip install -e .
# mkdocs new .
mkdocs gh-deploy

I had to downgrade

pip install "pydantic<2"

Add a function

• add to existing file in src or add new file
• add to docs/reference.md
• if added new file, edit src/__init__.py

Add an example notebook

• add to existing file in docs/examples
• add to docs/index.md
• add to mkdocs.yml