Program Results

Introduction to the event

Is the Western North Pacific Subtropical High (WNPSH) predictable on seasonal timescales?
This long-standing question remains at the heart of climate research. In our study, we revisit this challenge and propose new approaches to address it. We find that the predictability of the WNPSH is highly state-dependent—influenced not only by ENSO and Indian Ocean–Pacific SST conditions, but also by decadal variability (e.g., differences between the early and late 20th century). These dependencies pose significant hurdles for dynamical–statistical models such as multivariate regression or machine learning. While machine learning can help overcome some limitations, uncertainties in historical data (e.g., from the 1940s) may remain unresolved.

We also uncover asymmetry in ensemble sensitivity—that is, how forecast outcomes respond to input conditions—across different WNPSH states and ENSO phases. These insights highlight both the complexity and the opportunities in advancing seasonal prediction of the subtropical high.

The ensemble sensitivity of WNPSH to SSTA and NINO3.4 indices (y-axis is forecast lead time in month, x-axis is the input variables’ month). Three features are observed (1) Shorter lead time has stronger sensitivity. (2) Transition from El Nino to La Nina leads to stronger WNPSH and (3) SSTA has shorter memory in WNPSH than Nino 3.4.