Search

Different turbulent entrainment‐mixing mechanisms between clouds and environment are essential to cloud‐related processes; however, accurate representation of entrainment‐mixing in weather/climate models still poses a challenge. This study exploits the use of machine learning (ML) to address this challenge. Four ML (Light Gradient Boosting Machine [LGB], eXtreme Gradient Boosting, Random Forest, and Support Vector Regression) are examined and compared. It is found that LGB performs best, and thus is selected to understand the impact of entrainment‐mixing on microphysics using simulation data from Explicit Mixing Parcel Model. Compared with traditional parameterizations, the trained LGB provides more accurate microphysical properties (number concentration and cloud droplet spectral dispersion). The partial dependences of predicted microphysics on features exhibit a strong alignment with physical mechanisms and expectations, as determined by the interpreting method, thus overcoming the limitations of the "black box" scheme. The underlying mechanisms are that the smaller number concentration and larger spectral dispersion correspond to more inhomogeneous entrainment‐mixing. Specifically, number concentration after entrainment‐mixing is positively correlated with adiabatic number concentration and liquid water content affected by entrainment‐mixing, and inversely correlated with adiabatic volume mean radius. Spectral dispersion after entrainment‐mixing is negatively correlated with liquid water content affected by entrainment‐mixing, turbulent dissipation rate and relative humidity of entrained air. Sensitivity analysis further suggests that number concentration is mainly determined by cloud microphysical properties whereas spectral dispersion is influenced by both cloud microphysical properties and environmental variables. The results indicate that the LGB scheme has the potential to enhance the representation of entrainment‐mixing in weather/climate models. Plain Language Summary: Entrainment‐mixing processes occurring between clouds and environmental air have significant effects on cloud‐climate feedback, precipitation, and radiative transfer. Accurately representing these processes has been challenging with previously proposed parameterizations. Machine learning (ML) excels at identifying complex nonlinear relationships and avoids the inherent limitations of conventional parameterizations. Thus, we explore the use of ML to acquire the number concentration and cloud droplet spectral dispersion affected by entrainment‐mixing processes. Simulation data from Explicit Mixing Parcel Model are employed to train, validate, and test the four MLs, including Light Gradient Boosting Machine (LGB), eXtreme Gradient Boosting, Random Forest and Support Vector Regression. After evaluation, the LGB is shown to obtain the most accurate microphysics among the four ML schemes and traditional parameterizations. Additionally, the interpreting method for peeking inside the "black box" reveals that a smaller number concentration and a larger spectral dispersion are indicative of the more inhomogeneous entrainment‐mixing. The relative correlations between predicted microphysics and various features align with our expectations and the underlying physical principles. Sensitivity tests further confirm that incorporating more features produces a more robust and efficient prediction. Overall, this study affirms the reliability and applicability of ML to develop/replace subgrid parameterizations in actual weather/climate models. Key Points: Machine learning schemes are trained to predict cloud microphysical properties affected by turbulent entrainment‐mixing processesThe proposed Light Gradient Boosting Machine provides more accurate microphysical properties compared with traditional parameterization schemesThe partial dependencies of microphysics on features prove a robust alignment with physical mechanisms through the interpreting method [ABSTRACT FROM AUTHOR]

The first rainy season (April–May–June) extreme precipitation events over South China tend to frequently occur in specific hotspots, resulting in a great threat to local society and economy. Abundant moisture supply is essential to precipitation formation. In this study, we employ the FLEXPART model to investigate the geographical moisture sources for extreme precipitation events in three hotspots of Guangdong from west to east (termed as Area‐A, B, and C). During both the periods before and after the onset of the South China Sea Summer Monsoon (hereafter SCSSM), the moisture contributions of ocean regions are larger than that of land regions for all the hotspots, but the main moisture source regions are different. Before the onset of SCSSM, the main source regions are South China Sea and Southeastern Asia, which contribute 36.50 and 31.79% to Area‐A, 35.85 and 27.83% to Area‐B, while 42.27 and 23.81% to Area‐C, respectively. After the onset of SCSSM, the contribution of moisture from Indian Ocean increases a lot. Thus, the main source regions are Indian Ocean, Southeastern Asia and South China Sea, which contribute 31.52, 24.58 and 18.13% to Area‐A, 29.21, 21.82 and 14.99% to Area‐B, 26.62, 20.70%, and 28.19% to Area‐C, respectively. Among the three hotspots, the contribution of ocean regions for Area‐C is the largest. Southeastern Asia is the main land contributor before and after the onset of SCSSM for all the hotspots. This study implies the important role of both ocean and land moisture sources on extreme precipitation variations over South China. [ABSTRACT FROM AUTHOR]