In this talk I will discuss (i) recent improvements in seasonal climate predictions and their underlying dynamical mechanisms, and (ii) a framework for quantifying the dynamical drivers of projected future changes in climate variability. Both are founded in Hasselmann’s seminal contributions to our understanding of climate variability.

Regarding the first point, it is well known that the El Niño-Southern Oscillation (ENSO) provides most of the skill for global seasonal climate forecasts, yet, predicting it more than a year in advance remains a major challenge. Although Artificial Intelligence (AI) offers promising advancements in ENSO forecasting, linking the forecast skill of the AI model to specific physical processes is typically not possible. To provide improved explainable predictions, we developed a low-order model that combines a deterministic representation of ENSO (i.e., Jin’s recharge oscillator model) that is coupled to Hasselmann-type stochastic models representing the other climate modes in the global oceans. This new “eXtended Recharge Oscillator” (XRO) model incorporates the two different sources of predictability that Jin and Hasselmann respectively conceptualized. Key characteristics of ENSO dynamics, including ENSO’s seasonal synchronization, skewness, persistence, as well as ENSO’s seasonally modulated interactions with other climate modes are well captured by the XRO model. Importantly, it demonstrates high hindcast accuracy, providing skillful ENSO forecasts up to 17-19 months lead-time, comparable to the most skillful AI ENSO forecast models. Sensitivity experiments indicate that accounting accurately for climate-mode interactions significantly improves the long-lead time ENSO forecast skill. This improvement primarily originates from initial condition memory effects in the Indian Ocean and, to a lesser extent, the extratropical Pacific and Atlantic Oceans. These findings suggest that improved initialization of the global oceans in concert with improved representation of ENSO and its teleconnections in climate models that are used for seasonal forecasts are crucial for extending the lead time of skillful ENSO forecasts.

Regarding the second point, future changes to climate variability in response to anthropogenic warming beyond specific modes such as ENSO have so far not been well-characterized. We find that future changes to sea surface temperature (SST) variability (and correspondingly marine heatwave intensity) are often spatially heterogeneous in climate model projections. To understand the dynamical drivers of these patterns, we developed an extension of the original Hasselmann model, which allowed us to quantify the effect of changes to three drivers on SST variability: ocean "memory" (the SST damping timescale), ENSO teleconnections, and stochastic noise forcing. Here I will present an example application of this framework that allows us to understand the projected changes of SST variability in the North Pacific, with direct application to marine heatwave intensity changes.