East Asian Heat Extremes: Predictability Potential Outstrips Seasonal Forecast Skill

Seasonal forecasting systems can carry usable information about large-scale circulation variability over the western North Pacific, yet that information often fails to translate into accurate forecasts of East Asia’s summer hottest-day extremes, according to a study published in Climate Dynamics. The research, led by Zhaolu Hou and colleagues, examined summer extreme maximum temperature, known as TXx, across East Asia during 1993 to 2023, combining ERA5 reanalysis data with four seasonal forecast systems operated through the Copernicus Climate Change Service.

A central issue motivated the analysis: seasonal prediction of regional temperature extremes may derive partly from trends and low-frequency variability, and weak realized forecast skill does not necessarily mean the atmosphere itself has a short predictability limit. To disentangle these questions, the team evaluated realized skill using anomaly correlation coefficients and mean square skill scores, both on raw and detrended data. They then estimated intrinsic predictability independently, using a nonlinear analogue-based local Lyapunov exponent method, which measures how quickly forecast errors would grow even in an ideal prediction system.

The results reveal a sharp gap between these two measures. When trends were removed, the raw skill of TXx forecasts dropped markedly, leaving only weak interannual skill that varied by region, by forecast initialization month, and across the four forecast systems. Notably, positive correlations between forecasts and observations did not consistently translate into positive mean square skill scores. In contrast, ERA5 yielded an intrinsic predictability limit of roughly three months across the target regions, providing an empirical seasonal-scale error-growth benchmark for interpreting the weak realized skill.

The study assessed how the forecast systems responded to different phases of the western North Pacific subtropical high, using WNPSH-conditioned response and pathway diagnostics. The C3S systems did retain information on continuous subtropical high variability, which offers a conditional forecast opportunity. However, that opportunity redistributed skill across regions and circulation states rather than producing a uniform increase.

The redistribution was strikingly uneven. During strong phases of the subtropical high, an increase in forecasting skill over northern East Asia was largely confined to the ECMWF system, whereas reduced skill over central and southern China showed broader sign consistency across models but limited support. Crucially, neither skill change was accompanied by a comparable change in intrinsic predictability, indicating that the circulation state alters how forecast systems realize their potential rather than how predictable the extremes fundamentally are.

The multi-model ensemble added another layer of nuance. It reproduced the TXx response during weak subtropical high phases far more faithfully than during strong phases, with pattern correlations of 0.69 versus only 0.24. This asymmetry suggests the ensemble handles one circulation regime considerably better than the other.

Digging into the origins of the regional errors, the researchers identified amplitude damping, inter-model cancellation, and circulation-thermal mismatches in the ECMWF diagnostic as contributing mechanisms. Each erodes the link between large-scale circulation signals and the regional temperature extremes.

The broader message is that limitations in seasonal prediction of East Asian heat extremes are phase-dependent, region-dependent, and model-dependent. The intrinsic three-month predictability limit shows the system itself offers a seasonal window of opportunity, but the forecast systems’ limited fidelity in translating subtropical high signals into regional thermal responses prevents them from fully capitalizing on it. Weak realized skill reflects translation failures rather than an absence of predictable information, though closing that gap remains an open scientific challenge.

Subject of Research: Seasonal predictability and realized forecast skill of East Asian summer extreme maximum temperature in relation to the western North Pacific subtropical high

Article Title: Realized forecast skill and intrinsic predictability of East Asian summer TXx: WNPSH-conditioned opportunities and pathway fidelity

Article References: Hou, Z., Li, J., Li, C., Diao, Y., Tang, X., Hu, R., & Zhen, S. (2026). Realized forecast skill and intrinsic predictability of East Asian summer TXx: WNPSH-conditioned opportunities and pathway fidelity. Climate Dynamics, 64(11), Article 473. https://doi.org/10.1007/s00382-026-08438-z

Image Credits: AI Generated

DOI: 10.1007/s00382-026-08438-z

Keywords: seasonal forecasting, temperature extremes, East Asia, western North Pacific subtropical high, TXx, intrinsic predictability, Lyapunov exponent, ERA5, Copernicus Climate Change Service, multi-model ensemble, Climate Dynamics, Realized