Improving Prediction of Marine Low Clouds Using Cloud Droplet Number Concentration in a Convolutional Neural Network

Abstract

Marine low clouds play a crucial role in cooling the climate, but accurately predicting them remains challenging due to their highly non‐linear response to various factors. Previous studies usually overlook the effects of cloud droplet number concentration (Nd) and the non‐local information of the target grids. To address these challenges, we introduce a convolutional neural network model (CNNMet‐Nd) that uses both local and non‐local information and includes Nd as a cloud‐controlling factor to enhance the predictive ability of daily cloud cover, albedo, and cloud radiative effects (CRE) for global marine low clouds. CNNMet‐Nd demonstrates superior performance, explaining over 70% of the variance in these three cloud variables for scenes of 1° × 1°, a notable improvement over past efforts. CNNMet‐Nd also accurately replicates geographical patterns of trends in marine low clouds from 2003 to 2022. In contrast, a similar model without Nd (CNNMet) struggles to predict long‐term trends in cloud properties effectively. Permutation importance analysis further highlights the critical role of Nd in CNNMet‐N's predictive success. Further comparisons with an artificial neural network (ANNMet‐Nd) model, which uses the same inputs but without considering spatial dependence, show CNNMet‐Nd's superior performance with R2 values for cloud cover, albedo, and CRE being 0.16, 0.12, and 0.18 higher, respectively. This highlights the importance of incorporating non‐local information, at least on a daily scale, into low cloud predictions to enhance climate model parameterizations.