Skip to main content
Mathematics and Computer Science Division

Development of an Ocean Model Adjoint for Decadal Prediction

Developing an adjoint of the Parallel Ocean Program using automatic differentiation

This project will develop an adjoint of the Parallel Ocean Program (POP; version 2.0.1) using automatic differentiation (AD) techniques. We have already had success with AD on sea ice models and will use this knowledge with POP.

It is now unequivocal that the Earth’s climate system is warming. The most recent IPCC assessment concludes that the increased temperatures in the latter 20th century are very likely due to anthropogenic greenhouse gases, and continued greenhouse emissions will likely result in even larger increases during the 21st century. Even if controls could be put on greenhouse emissions immediately, there is still some climate change that will take place in the next few decades from emissions already in the system which are not easily removed.

It is crucial that we develop ways to do reliable decadal predictions of climate to inform adaptation decisions. An important part of any prediction system is the data assimilation system which initializes the model with observations. New ocean observation networks can provide the raw observations for such a system. The other component needed is an adjoint of the ocean model.