Zeng et al. (2025) A Neural Network Parametrization of Volumetric Cloud Fraction Profiles Using Satellite Observations and MERRA‐2 Reanalysis Meteorological Data

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

Identification

• Journal: Journal of Advances in Modeling Earth Systems
• Year: 2025
• Date: 2025-11-27
• Authors: Shan Zeng, Kuan‐Man Xu, Yongxiang Hu, Seiji Kato, Seung‐Hee Ham
• DOI: 10.1029/2025ms004959

Research Groups

[Information not available in the provided abstract.]

Short Summary

This study develops a deep machine learning (DML) physical parameterization for volumetric cloud fraction (VCF) using satellite lidar-radar measurements and reanalysis data. The DML model, particularly an LSTM network, effectively captures cloud physical processes, outperforming MERRA-2 reanalysis in representing various cloud types and improving VCF histograms across different spatial and temporal scales.

Objective

• To develop a deep machine learning (DML) physical parameterization of volumetric cloud fraction (VCF) using satellite lidar-radar measurements and collocated meteorological variables from reanalysis data.
• To evaluate the DML model's ability to capture underlying cloud physical processes, its performance against existing reanalysis data (MERRA-2), and its representation of cloud distributions and interannual variations.
• To identify the dominant meteorological factors influencing VCF through sensitivity analysis.

Study Configuration

• Spatial Scale: Global, 3-dimensional grid volumes, covering tropical, subtropical, and midlatitude storm-track regions.
• Temporal Scale: Monthly mean seasonal variations, interannual variations (including El Niño-Southern Oscillation - ENSO).

Methodology and Data

• Models used: Deep Machine Learning (DML), specifically a sequence-to-sequence Long Short-Term Memory (LSTM) neural network with a customized loss function.
• Data sources: Satellite lidar-radar measurements for observed VCF profiles, MERRA-2 reanalysis data for collocated meteorological variables.

Main Results

• The DML (LSTM) network effectively captures underlying cloud physical processes.
• DML predictions outperform MERRA-2 reanalysis in representing low-level clouds in tropical and subtropical regions.
• DML predictions outperform MERRA-2 reanalysis in representing low- and middle-level clouds over midlatitude storm-track regions.
• DML significantly improves VCF histograms compared to MERRA-2.
• Improvements are evident in vertical distributions of zonally, meridionally, and globally averaged VCFs, geographic distributions of low-, middle-, and high-level clouds, and seasonal variations in monthly mean VCF.
• DML predictions effectively capture El Niño-Southern Oscillation (ENSO) and other interannual variations.
• Sensitivity analysis reveals that relative humidity (RH) is the dominant factor influencing globally averaged VCF at low and middle altitudes, followed by temperature.
• At higher altitudes, temperature becomes the primary driver of VCF, primarily through its effect on RH.
• Increases in pressure vertical velocity (ω) are associated with minor decreases in VCF, with an effect less significant than RH and temperature.

Contributions

• Development of a novel deep machine learning parameterization for volumetric cloud fraction (VCF) using satellite lidar-radar observations and reanalysis data.
• Demonstrated superior performance of the DML model over a state-of-the-art reanalysis product (MERRA-2) in representing various cloud types, distributions, and histograms globally.
• Enhanced capability to capture interannual climate variability, including ENSO, in cloud fraction predictions.
• Provided quantitative insights into the primary meteorological drivers (relative humidity and temperature) of VCF variations across different altitudes through sensitivity analysis.

Funding

[Information not available in the provided abstract.]

Citation

@article{Zeng2025Neural,
  author = {Zeng, Shan and Xu, Kuan‐Man and Hu, Yongxiang and Kato, Seiji and Ham, Seung‐Hee},
  title = {A Neural Network Parametrization of Volumetric Cloud Fraction Profiles Using Satellite Observations and MERRA‐2 Reanalysis Meteorological Data},
  journal = {Journal of Advances in Modeling Earth Systems},
  year = {2025},
  doi = {10.1029/2025ms004959},
  url = {https://doi.org/10.1029/2025ms004959}
}