A0707
Title: Machine learning and spatiotemporal statistics in the ocean: Fusing data sources and making nonlinear predictions
Authors: Adam Sykulski - Imperial College London (United Kingdom) [presenting]
Abstract: The ocean is observed through a variety of means, including from satellites (remotely) and from instruments deployed in the water (in-situ). Sometimes, these measurements agree, albeit with different observational noise and sampling characteristics, but sometimes, they measure fundamentally different but related quantities. Fusing these datasets, therefore, poses a significant statistical challenge to avoid biases and poor predictions. Two examples of how machine learning and spatiotemporal statistics can fuse heterogeneous ocean data in a nonlinear sense to make better predictions are presented. The first example predicts ocean surface velocities using satellite data only but trained using in-situ instruments called drifters, where a novel probabilistic prediction framework is developed using multivariate natural gradient boosting. The second example fuses satellite and in-situ data to predict the abundance of Antarctic Krill in the Southern Ocean, where a key innovation is to use spectral analysis tools to transform the datasets for use into one single predictive model.
