Revealing bias characteristics of cloud diurnal variation to aid climate model tuning and improvement

The cloud fraction diurnal variation (CDV) regulates the Earth system’s radiative budget and balance, influencing atmospheric variables such as temperature and humidity, as well as physical processes like precipitation and tropical cyclones. However, significant simulation biases of CDV exist in climate models. To date, most model evaluations have focused on the daily mean cloud fraction (CFR), while the CDV has received less attention.

Research led by Guoxing Chen, a research scientist at Fudan University, together with his master’s student Hongtao Yang and colleagues, selected a global climate model, FGOALS-f3-L, to reveal the bias characteristics of CDV in this model. Their study was recently published in Atmospheric and Oceanic Science Letters under the title “Bias characteristics of cloud diurnal variation in the FGOALS-f3-L model.”

By comparing the cloud fraction output from the FGOALS-f3-L model with observational datasets from ISCCP and CERES, the study analyzed the model biases in both CFR and CDV, and quantitatively assessed the contributions of high-, mid-, and low-level cloud biases to the total CDV bias.

The results indicate that the daytime low-level cloud fraction is severely underestimated, mostly contributing to the CDV bias to total cloud fraction. Mid- and high-level clouds exhibit opposing biases, which partially offset the contribution of low-level cloud biases. Additionally, the study found that biases in CDV have a significant impact on the model’s simulation of shortwave cloud radiative effects—an impact that can reach or exceed half of that caused by biases in CFR.

“The diurnal variation of cloud fraction plays an important role in simulating both cloud characteristics and shortwave cloud radiative effects in climate models and deserves more attention,” says Dr. Chen. “Revealing the bias characteristics of cloud diurnal variation can provide targeted guidance for climate model developers to tune and improve the model simulations.”

Next, the team will introduce variables such as cloud optical thickness and albedo to quantitatively isolate the contribution of CDV to biases in simulated shortwave cloud radiative effects, and will further analyze the bias-contribution characteristics among different factors.

More information:
Hongtao Yang et al, Bias characteristics of cloud diurnal variation in the FGOALS-f3-L model, Atmospheric and Oceanic Science Letters (2025). DOI: 10.1016/j.aosl.2025.100636