Assessment of subseasonal forecast skill of tropical–extratropical cloud bands over the four subtropical convergence zones

Abstract

Tropical–extratropical cloud bands act as a primary source of moisture and precipitation for the subtropics, yet their representation and predictability on subseasonal time‐scales remain challenging. This study evaluates the subseasonal predictive skill of these cloud bands up to four weeks within the European Centre for Medium‐Range Weather Forecasts (ECMWF) reforecasts from 2003 to 2022 across the four main subtropical convergence zones. To conduct this assessment, a combination of spatial, deterministic, and probabilistic metrics is applied to the detected cloud bands, the outgoing long‐wave radiation from which cloud bands are identified, and the large‐scale circulation that drives them. We find that, while exact day‐to‐day predictions are limited to approximately one week, a substantial portion of forecast error stems from temporal displacement. Accounting for these timing uncertainties via a time‐windowed approach yields a marked improvement of skill scores, revealing that the model resolves the spatial structure of large‐scale drivers successfully well into the subseasonal range, although with a systematic bias toward premature onset. Despite this, the reforecasts exhibit a loss of variance at mesoscales, leading to overly smooth fields and an underestimation of extreme convective anomalies. Furthermore, the ability of skilful large‐scale circulation forecasts to translate into improved cloud‐band prediction exhibits marked regional contrasts. The South Indian Ocean maintains the highest baseline spatial skill for up to two weeks, whereas the dynamical linkage driving predictability is strongest in the Meiyu–Baiu frontal zone across all subseasonal lead times (1–28 days). In the Southern Hemisphere basins, particularly the South Pacific, this dynamical linkage degrades at extended lead times, as the model defaults to predictable but convectively suppressed states. These findings highlight that representing large‐scale circulation accurately is not sufficient; advancing the parametrization of mesoscale convective organization remains essential to leverage subseasonal predictability fully.