Marine Cloud Heterogeneity in Satellite Observations and a General Circulation Model

Abstract

Satellite observations from the Multiangle Imaging SpectroRadiometer (MISR) and MOD-erate Resolution Imaging Spectroradiometer (MODIS) from 2007 are used to investigate the relationships between the optical depth distribution, sub-pixel heterogeneity (Hσ), albedo and the spatial structure of the cloud field at the scales of General Circulation Models (GCMs). Cloud spatial structure is quantified with both cloud fraction and the equivalent diameter of the cloud field (CD). Solar zenith angle, cloud fraction and CD explain 85% of the variability in domain albedo, and 39% of the variance in domain mean Hσ. CD varies between regions and cloud regimes independently of cloud fraction. The large unexplained variability in Hσ indicates that differences in sub-pixel heterogeneity within regimes are as important as those between regimes. The heterogeneity of cloud optical depth is quantified using the ratio of the distribu-tion’s logarithmic and linear means (χ). Heterogeneity and the mean cloud optical depth decrease and increase as CD increases, respectively. The increase of Hσ and partially cloudy pixel fraction with CD suggests that the sensitivity of cloud optical depth to CD is underestimated. Regimes with large CD and small Hσ can be measured most accurately and compared robustly with the cloud fields simulated by GCMs. The heterogeneity parameterisation of Global Atmosphere 7.0 (GA7) is validated using these satellite observations. To reduce retrieval errors, we focus the validation on overcast domains and mid-latitude stratocumulus and frontal clouds as they have high CD. There is a strong compensating bias between the albedo of high-topped and low-topped overcast clouds. An overestimation of heterogeneity with increasing low-topped cloud fraction in overcast clouds contributes to 30% of the albedo underestimation for the low-topped clouds. We test the hypothesis that a bias in low-altitude cloud contributes to the Southern Ocean Reflected Solar Radiation (RSR) bias in GA7. The model reproduces the sign of the hemispheric asymmetry in low-cloud fraction found in the observations. However, the zonal position of the asymmetry is closer to the equator than in the observations which means that the heterogeneity bias in low-topped cloud does not directly contribute to the Southern Ocean RSR bias in GA7.