Puget Sound Acidification and Impacts (PSAI)

PSAI is an in-progress tool that adjusts LiveOcean biogeochemical (LiveOcean-BGC) model output to align more closely with observations. It uses machine learning methods trained on model-observation differences, also called model residuals or model-data misfits, to dynamically correct model bias (overall misfit).

The tool requires geographic coordinates, water column depth, date, and LiveOcean-BGC output for salinity, temperature, and, optionally, other seawater properties.

This software is under development and has benefitted from extensive input and guidance from the LiveOcean modeling team.

Repository Organization

This repository is organized into six primary folders:

1. Observations: Contains compiled observations; information about the original data sources, preprocessing scripts and methods, quality assurance and quality control procedures, a methods manual for data handling, and both original and processed final datasets, including source and collection information and quality control flags.

2. Model_Observation_Pairing: Contains methods for pairing observations with model output, calculating model-data misfits, and creating the final paired dataset.

3. Algorithm_Development: Contains code for training machine learning models for bias adjustments, along with methods and validation procedures used.

4. Assessments: Contains code and documentation for evaluating the final model using withheld datasets.

5. PSAI: Contains the final trained model for adjusting LiveOcean-BGC output.

6. Misc: Containes other miscellaneous related code and documentation.