Ocean Machine LEarning Toolkit (OMLET)

Originally developed during OceanHackWeek 2018 at the University of Washington, OMLET was a machine learning tool, designed in python, aimed at predicting dissolved oxygen and nutrients (phosphate, nitrate & nitrite, and silicate) from temperature, salinity, and pressure values obtained from both sensor and in-situ water measurements. OMLET utilized time series data from the Station ALOHA during the Hawaii Ocean Time-series program (HOT, 1988-2014), in addition to King County Pugent Sound data to create two machine learning models. The models were then given Line P (Vancouver Island to Ocean Station Papa) data to predict oxygen and nutrient values for Station Papa and then compared to actual values.

In order to predict these parameters, OMLET used a variety of models including Support Vector Regression (SVR), Random Forest Regression (RFR), and Artificial Neural Networks (ANN). Python libraries including Pandas, xArray, and Scikit-Learn were the foundation for creating these models.

See the full project on GitHub here.