Machine learning disentangles bias causes of shortwave cloud radiative effect in a climate model

Aug 4, 2026

Upload by Levi

Large bias exists in shortwave cloud radiative effect (SWCRE) in general circulation models, mainly driven by the joint representation of cloud fraction and cloud water contents. An effective machine-learning approach is developed to separate the individual bias contributions of key cloud parameters controlling SWCRE. A random-forest surrogate for SWCRE is trained using observations and FGOALS-f3-L simulations, using CFR, CSC, LWP, IWP, TOA clear-sky solar flux, and solar zenith angle. The surrogate achieves R² > 0.96 and quantifies parameter-wise SWCRE bias via partial radiation perturbation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Abstract

Introduction

Cloud and cloud radiative effects in climate models

SWCRE biases and their regional dependence

Challenges in quantitatively disentangling bias causes

Prior methods: radiation kernel and partial radiation perturbation

Proposed method and analysis framework

Machine learning disentangles bias causes of shortwave cloud radiative effect in a climate model Hongtao Yanga, Guoxing Chena,b,*, Wei-Chyung Wangc, Qing Baod, and Jiandong Lid aDepartment of Atmospheric and Oceanic Sciences, Institute of Atmospheric Sciences, and CMA-FDU Joint Laboratory of Marine Meteorology, Fudan University, Shanghai, China b Shanghai Frontier Science Center of Atmosphere-Ocean Interaction, Fudan University, Shanghai, China cAtmospheric Sciences Research Center, University at Albany, State University of New York, Albany, NY, USA d State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China Corresponding author: Guoxing Chen ([chenguoxing@fudan.edu.cn](chenguoxing@fudan.edu.cn)) Abstract Large bias exists in shortwave cloud radiative effect (SWCRE) of general circulation models (GCMs), attributed mainly to the combined effect of cloud fraction and water contents, whose representations in models remain challenging. Here we show an effective machine-learning approach to dissect the individual bias of relevant cloud parameters determining SWCRE. A surrogate model for calculating SWCRE was developed based on random forest using observations and FGOALS-f3-L simulation data of cloud fraction (CFR), cloud-solar concurrence ratio (CSC), cloud liquid and ice water paths (LWP and IWP), TOA upward clear-sky solar flux (SUC), and solar zenith angle. The model, which achieves high determination coefficient > 0.96 in the validation phase, was then used to quantify SWCRE bias associated with these parameters following the partial radiation perturbation method. The global-mean SWCRE bias (in W m-2) is contributed by CFR (+5.11), LWP (-6.58), IWP (-1.67), and CSC (+4.38), while SUC plays a minor role; the large CSC contribution highlights the importance of cloud diurnal variation. Regionally, the relative importance varies according to climate regimes. In Tropics, overestimated LWP and IWP exist over lands, while oceans exhibit underestimated CFR and CSC. In contrast, the extratropical lands and oceans have, respectively, too-small CSC and the 'too few, too bright' low-level clouds. We thus suggest that machine learning, in addition for developing GCM physical parameterizations, can also be utilized for diagnosing and understanding complex cloud-climate interactions. Introduction Clouds play an important role in the earth climate system (e.g., refs. 1) . They can reflect solar shortwave radiation back into space while trapping the longwave radiation emitted by the surface and the atmosphere. These cloud radiative effects (CREs) significantly alter the surface radiation budget and balance, affecting both the weather migration (e.g., refs. 2–3) and the climate change (e.g., refs. 4–6) . However, the accurate simulation of clouds and CREs in climate models remains challenging (7) due to inadequate understanding of certain physical processes (e.g., sub-grid physics, ice nuclei; 8) and over-simplified physical parameterizations (e.g., refs. 9), casting clouds as the critical source of uncertainties in studies on climate change and climate modeling (10) . Current climate models have marked biases in both shortwave and longwave cloud radiative effects (SWCRE and LWCRE), wherein the SWCRE bias dominates (11–13) . Specifically, the models tend to simulate too weak SWCRE over the southeast Pacific (14), the Southern Ocean (15), and East Asia (16–18), allowing too much solar radiation reaching the surface. As these regions are featured with remarkable different climate regimes, the bias causes are also region dependent. For example, biases over the southeast Pacific could be partially attributed to the model deficiency in representing effects of anthropogenic aerosols on the stratocumulus microphysical properties (e.g., refs. 19–20); biases over the Southern Ocean are believed to be connected wi

FAQs

What is the main source of bias in shortwave cloud radiative effect in climate models?

The dominant bias arises from the combined representation of cloud fraction and cloud water contents, whose modeled representations remain challenging.

How does the study quantify the individual contribution of cloud parameters to SWCRE bias?

It builds a random-forest surrogate model to emulate SWCRE and then applies the partial radiation perturbation method to compute parameter-wise bias contributions.

Which cloud parameters contribute most to the global-mean SWCRE bias?

The global-mean SWCRE bias is mainly contributed by CFR (+5.11), LWP (-6.58), IWP (-1.67), and CSC (+4.38), while SUC plays a minor role.

Aug 4, 2026

Upload by Levi