Deep Neural Network Based Scale‐Adaptive Cloud Vertical Overlap Parameterization

Xiaowei Wu, Feng Zhang, Hua Zhang, Haixia Xiao, Guoxing Chen

Geophysical Research Letters · 18 February 2026 · China Meteorological Administration (CMA), Fudan University (FuXi)

In one sentence. A scale-adaptive deep neural network parameterization of cloud vertical overlap, usable in climate models, keeps accuracy above 0.9 across resolutions and beats traditional schemes.

Abstract

Abstract Cloud vertical overlap has important implications for cloud radiative effects in climate models. Due to its spatiotemporal complexity, it is difficult to simulate cloud overlap in climate models. Decorrelation length () is crucial for depicting the vertical overlap of clouds in climate models. This study introduces a scale‐adaptive parameterization scheme, which can be integrated into climate models. The study considered three atmospheric statistics and dynamic factors: latitude, wind vertical shear, and horizontal grid scale, to develop an parameterization scheme by deep neural network (DNN). For different resolutions, the simulation accuracy of this parameterization scheme can basically be maintained at above 0.9, which significantly surpasses that of other traditional overlap schemes. The parameterization scheme can be seamlessly integrated into the climate model and has the potential to improve the simulation accuracy of cloud vertical overlap, thereby enhancing the model's capability to accurately depict cloud radiation effects.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.