What sets the predictability timescales of SST and upper-ocean heat content in the Atlantic and Pacific basins?

Understanding the extent to which sea surface temperatures (SSTs) are predictable is important due to the strong impact of SSTs on climate variables, such as temperature and rainfall over adjacent landmasses. For example, low-frequency variability of North Atlantic SSTs, termed the Atlantic Multidecadal Variability, impacts temperatures over North America and Europe, precipitation over the Sahel, and the strength of Atlantic hurricanes. As a result, decadal predictability experiments have received much attention, in particular CMIP5 and CMIP6 include decadal prediction experiments. However, models differ substantially on both the magnitude and spatial pattern of predictability timescales of SST and upper-ocean heat content (UOHC).
The overarching goal of this work is to determine the processes that set predictability timescales for SST and UOHC in the Atlantic and Pacific basins and to ascertain how well these processes are represented in models. Predictability timescales will be calculated from both gridded observations and CMIP5/CMIP6 models, and the results will be compared quantitatively. In order to do this, we will calculate predictability timescales from observations and both subsampled CMIP preindustrial control integrations and CMIP historical runs, in which the observational data and model data are temporally sampled and processed (e.g., removal of anthropogenic signal) in the same way. This will enable us to isolate regions where models differ from observations.
We will address the underlying mechanisms that lead to the predictability timescales for SST and UOHC in both observations and models. In order to determine the degree to which stochastic atmospheric forcing integrated over the oceanic mixed layer can explain spatial variations on predictability timescales, we will (1) determine the extent to which spatial variations in the wintertime mixed layer depth can explain spatial variations in predictability timescales for SST and UOHC and (2) compare observed predictability timescales for SST to those predicted by an idealized red noise model with no ocean dynamics. We will isolate regions where active ocean dynamics (rather than thermodynamics) play a role in setting predictability timescales. The relevant ocean dynamics in these regions will be determined by (1) analyzing lagged correlations between time series in these regions and atmospheric forcing and ocean dynamical variables and (2) computing heat budgets in the regions of interest.
Our proposal is targeted at the first priority area of NOAA’s CVP-Decadal Climate Variability and Predictability Competition. Our proposal is directly relevant to the program objectives to investigate the mechanisms that govern variability and predictability of the ocean on interannual to decadal timescales. We will be using both long-term observational data and models (e.g., CMIP5/CMIP6) and our regional focus will be both the Atlantic and Pacific basins. A main focus of the proposal is a rigorous model-data comparison of predictability timescales. This intercomparison will enable us to assess the realism of models currently used for decadal predictions and provide model developers with both regions and processes that need to be improved in order to better predict decadal climate variability. This is directly relevant to program’s overarching goal to “…identify state, mechanisms, and sources of predictability on the interannual to decadal timescale, which will help to lead to future improvements in skillful decadal prediction systems for climate (ocean and atmosphere).” More broadly, our proposal is relevant to NOAA’s mission to help society “cope with, and adapt to, today’s variations in climate and to prepare for tomorrow’s”.