Understanding the intrinsic predictability limit of the mid-latitude atmosphere, defined as the inherent limit on forecasting under nearly perfect conditions, is crucial to numerical weather prediction. Based on 20-day perturbed ensemble forecasts using a global convection-permitting model at the European Centre for Medium-Range Weather Forecasts [ECMWF; 9-km operational model; viz., the Integrated Forecast System (IFS)], this study revisits the intrinsic predictability limit and examines its dependence on variables, altitudes, and scales in midlatitudes. In particular, this study addresses the underexplored predictability of gravity wave momentum fluxes (GWMFs). The intrinsic predictability is quantified via the Loss Predictability Index (LPI), defined as the ratio of error energy (ensemble variance) to reference energy (the ensemble average of the energy over the members), offering a quantitative metric that complements the visual estimation methods employed by Zhang et al. (2019).
From the LPI perspective, a physical-space analysis reveals that the useful prediction skill (LPI = 60%) of most variables is about two weeks in the troposphere, which is the classical predictability time, while the predictability limit (LPI = 90%) of most variables approaches or even exceeds 20 days in the same layer. In a further scale‑dependent analysis with 500‑hPa total energy as an example, the predictability limit is about two weeks at the synoptic scale (zonal wavelengths ~800 km < λ ≤ ~10000 km), while it exceeds 20 days at the planetary scale (λ > ~10000 km). This indicates that only when planetary-scale perturbations are filtered out does the classical two-week predictability limit hold. Furthermore, the underexplored predictability limit of GWMFs becomes longer with increasing scale, consistent with that of well-documented forecast variables. Therefore, the predictability limit of GWMFs is much longer than that of the mesoscale wind perturbations that form GWMFs (~50 km < λ ≤ ~800 km). This is likely because gravity wave packets are much larger in scale than the winds induced by mesoscale gravity waves. Overall, these findings not only deepen the understanding of intrinsic predictability but also highlight the practical value of such global convection-permitting models.

Reference:
Li, Kelin, Junhong Wei, Manshi Weng, Y. Qiang Sun, and Xubin Zhang, 2026: Revisiting the Atmospheric Intrinsic Predictability in Midlatitudes: A Variable-, Altitude-, and Scale-Dependent Error Growth Study from the Loss Predictability Index Perspective. (manuscript in preparation)
 

内在可预报性,集合预报,全球对流解析模型,重力波动量通量,尺度依赖性