Fan et al. (2025) An Improved Neural Network‐Based Scale‐Adaptive Cloud Fraction Scheme: Incorporation of Atmospheric Stability

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Journal of Geophysical Research Atmospheres
• Year: 2025
• Date: 2025-11-14
• Authors: Q. Fan, Guoxing Chen, Hongtao Yang, Wei‐Chyung Wang, Yanhong Gao, Guangtao Dong
• DOI: 10.1029/2025jd044619

Research Groups

Not explicitly stated in the abstract.

Short Summary

This study enhances a neural network-based cloud fraction scheme for general circulation models by incorporating atmospheric stability and predicting cloud volume fraction. The improved scheme significantly boosts prediction accuracy and scale adaptability for cloud area fraction, especially for low-level clouds, and effectively predicts cloud volume fraction.

Objective

• To improve a neural network-based scale-adaptive cloud fraction scheme by considering atmospheric stability as an input and introducing cloud volume fraction as an additional output, aiming to enhance prediction accuracy and scale adaptability for cloud representation in general circulation models.

Study Configuration

• Spatial Scale: Global/regional, focusing on cloud spatial distribution, vertical structure, low-level clouds, the Southeast Pacific, and the Pacific Cross-Section Intercomparison transect.
• Temporal Scale: Seasonal variation.

Methodology and Data

• Models used: Neural network-based scale-adaptive cloud fraction scheme (updated version), FGOALS-f3-L General Circulation Model (GCM).
• Data sources: CloudSat observations, simulation data from the FGOALS-f3-L GCM.

Main Results

• Including atmospheric stability significantly enhances both prediction accuracy and scale adaptability for cloud area fraction compared to the original scheme.
• Root-mean-square errors (RMSEs) for liquid, mixed-phase, and ice clouds are reduced by 15.5%, 11.6%, and 6.67%, respectively, on the test data set.
• Predictions of cloud area-fraction vertical structure and spatial distribution show RMSE reductions of 33% and 17%, respectively.
• Improvements are most pronounced for low-level clouds, particularly in simulating the seasonal variation of stratocumulus over the Southeast Pacific and cloud regime transitions along the Pacific Cross-Section Intercomparison transect.
• The new scheme performs well in predicting cloud volume fraction, showing accuracy and scale adaptability comparable to that for area fraction.
• Offline evaluations using FGOALS-f3-L GCM simulation data confirm the superiority of the updated scheme.

Contributions

• Development of an improved neural network-based scale-adaptive cloud fraction scheme that incorporates atmospheric stability and predicts both cloud area and volume fractions.
• Demonstrated significant enhancements in prediction accuracy and scale adaptability for cloud fraction, particularly for low-level clouds and specific cloud regimes.
• Provides a potential pathway to improve cloud fraction representation and associated climate effects in general circulation models.

Funding

Not explicitly stated in the abstract.

Citation

@article{Fan2025Improved,
  author = {Fan, Q. and Chen, Guoxing and Yang, Hongtao and Wang, Wei‐Chyung and Gao, Yanhong and Dong, Guangtao},
  title = {An Improved Neural Network‐Based Scale‐Adaptive Cloud Fraction Scheme: Incorporation of Atmospheric Stability},
  journal = {Journal of Geophysical Research Atmospheres},
  year = {2025},
  doi = {10.1029/2025jd044619},
  url = {https://doi.org/10.1029/2025jd044619}
}