Using machine learning to quantify historical changes in ocean heat content

University Of California-Santa Barbara, Santa Barbara CA

Investigators

Abstract

This proposal will estimate how much the global ocean has warmed over the past half century and look at the spatial and temporal patterns of changes in ocean heat content. The project will use a machine learning approach to combine historical data such that errors and biases are minimized. An exciting aspect of the project is that it will also estimate heat content for the deep, abyssal ocean (deeper than 2000m). Ocean heat content is an important indicator for how much excess heat the Earth system is accumulating and is thus important for improving understanding and prediction of climate change. The project will involve students, including providing internships for students from Historically Black Colleges and Universities. This project will use ensemble Artificial Neural Networks (EANN) to estimate the total ocean heat content over the past fifty years. The use of EANN machine learning methods will reduce systematic biases in the historical temperature data sets and yield an improved historical data set with error estimates. The project will then also look at spatial and temporal patterns of ocean warming. A novel aspect of the project is that it will include estimates of OHC for the abyssal ocean deeper than 2000m. The project has strong potential for broader impacts by providing a state-of-the-art estimate of ocean warming which could be used to constrain ocean climate models. The project also broadens the participation of underrepresented minority students through internships. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.