This project will combine large flow-cytometry datasets with novel machine learning models to reveal the geographical distribution of phytoplankton and show how the environment shapes these patterns. Neural network methods for flow-cytometry data analysis will be applied to data from over 100 cruises across the Pacific and Atlantic Oceans. The project will develop computationally efficient mixture of neural network models, a generative model framework for changepoint detection, and spatially dependent convolutional neural networks. These methods will make oceanographic data analysis more automatic and efficient while also allowing for model-based rediscovery of ocean provinces as well as predictive mapping of ocean microbe populations and traits. The proposed methodology will advance AI and statistics, data science, and oceanography while also being useful across a broad range of disciplines that deal with complex high-dimensional dependent data such as environmental science, ecology, agriculture, epidemiology, and econometrics. The methodology will also be useful for various data science industries that handle high-dimensional mixture data or flow cytometry. Public-use software packages will be created. The project will develop computationally efficient neural network models that automatically classify cell level data with environmental covariates. This will streamline the analysis and reveal biological responses to changing environments. Generative neural networks will b