Prospects for climate predictability on decadal timescales are thought to primarily reside in the ocean or through external forcing

This Project Grant award from the National Oceanic and Atmospheric Administration (NOAA) under the Climate and Atmospheric Research program (CFDA 11.431) will support a $415,615 research project at North Carolina State University (NC State) to investigate the mechanisms driving multi-year sea surface temperature anomaly (SSTA) predictability. The key objectives are to: 1) assess the role of low cloud-SST feedback strength in model differences in SSTA persistence, 2) determine the influence of ENSO on SSTA persistence, and 3) assess the combined effects of cloud feedback and ENSO on SSTA predictability. This research aims to improve the mechanistic understanding of how clouds and ENSO contribute to climate predictability on multi-year timescales, with implications for operational SSTA prediction models, data-driven forecasting, and climate model studies. The project will train a graduate student in climate modeling, model-observation comparison, and climate dynamics analysis.

However, prospects for predictability on multi-year timescales, such as the persistence of sea surface temperature anomalies (ssta) into year two (YR2), are not well understood due to a lack of understanding of mechanisms which may provide potential predictability. This lack of mechanistic understanding hinders the ability to assess whether phenomena such as multi-year marine heat waves or impacts from multi-year ssta like terrestrial drought might feasibly be predictable. A metric for potential predictability of ssta from observations indicates that ssta may be predictable into YR2 in the northeast (NE) and southeast (SE) subtropical pacific. These regions encompass the well-known thermodynamically coupled pacific meridional modes and the NE pacific has recently been a region of pronounced and persistent marine heat waves. This supports the idea that multi-year predictability does exist in these regions and is not primarily connected to external forcing but rather to internal climate processes. In particular, the local low cloud-SST feedback strength and enso appear to be key drivers of these multi-year ssta persistence signals. Further, models with a weaker cloud-SST feedback than observations tend to underestimate the persistence of the ssta into YR2, whereas models with too strong cloud-SST feedback overestimate the predictability into YR2. To investigate these mechanisms driving ssta predictability into YR2, we propose the following: objective 1: we will assess the importance of low cloud-SST feedback strength in inter-model differences in NE and SE subtropical pacific ssta persistence into YR2. This will be done through analysis of both climate models and observed estimates. Objective 2: we will determine the role of enso on NE and SE subtropical pacific ssta persistence into YR2 through comparison across a process-based climate model hierarchy that varies only in the characteristics of enso variability in the tropical pacific. Objective 3: we will determine the combined influence of enso and low cloud feedback strength on NE and SE subtropical pacific ssta persistence into YR2. This will be done by comparing persistence from a model hierarchy with a strong versus weak cloud feedback, with and without enso. The proposed research will fill a GAP in the mechanistic understanding of clouds as a potential source of climate predictability into YR2. The targeted model hierarchies will improve understanding of how enso impacts multi-year predictability of sstas and how cloud feedback may compound this effect. Our research is closely aligned to the CVP call to further process-level understanding of the role of the ocean, land, atmosphere, and/or cryosphere on multi-year to decadal timescales. Through this work, a graduate student will be trained to run and analyze climate models, to compare climate models to observations, and to understand climate dynamics and predictability. Overall, the proposed work will reveal important implications for models that are used 1) operationally for ssta predictions, 2) to train data driven prediction models, and 3) to perform perfect model predictability studies on multi-year timescales.