Cloud type variability derived from MODIS, CALIPSO, and CloudSat over tropics and its impact on surface irradiances

Earlier studies indicate that the strength and width of Hadley circulation changes depending upon ENSO phase. During an El Nino phase, the Hadley circulation is stronger and narrower than the circulation during a La Nina phase. A study by Loeb et al. (2014) shows that interannual variations in high-level clouds associated with circulation over the descending and ascending branches of the Hadley circulation is the primary driver of Top-of-atmosphere, surface, and atmospheric radiation variability with circulation. Loeb et al. use MODIS-derived clouds, which might influence their result in two ways. First, the cloud type based on cloud height determined from MODIS observations has an uncertainty where, for example, undetected thin cirrus overlapping with lower-level clouds can increase the occurrence of mid-level clouds. Second, the variability of the low-level cloud fraction might be caused by the variability of higher-level cloud fractions. Even if the low-level cloud fraction is not altered by the ENSO phase, changing the higher-level cloud fraction might be interpreted as a low-level cloud fraction change. In addition, the low-level cloud fraction change can be simply masked by overlapping high-level clouds.

In this study, we use CALIPSO and CloudSat data to evaluate the MODIS-derived cloud and investigate how active-sensor derived clouds vary with ENSO phase. Preliminary results show that the difference between the CALIPSO/CloudSat- and MODIS-derived mean cloud fraction for high-, mid-, and low-level clouds exposed to space are (MODIS – CALIPSO/CloudSat), -0.05, 0.06, and -0.08 over the ascending branch, -0.02, 0.05, and -0.09 over the northern hemisphere descending branch, and 0.00, 0.08, and -0.13 over the southern hemisphere descending branch. This result agrees with earlier studies showing larger mid-level cloud amounts derived from passive sensors compared with that derived from active sensors. The cloud fraction of high-, mid-, and low-level clouds that are not exposed to space are, 0.15, 0.09, and 0.20, over the ascending branch, 0.05, 0.04, and 0.12, over the northern hemisphere descending branch, and 0.05, 0.04, and 0.14 over the southern hemisphere descending branch. These account for 24%, 63%, and 52% of the total high-, mid-, and low-level clouds over the ascending branch, 16%, 45%, and 35% over the northern hemisphere descending branch, and 17%, 45%, and 29% over the southern hemisphere descending branch.

Despite missing overlapping clouds in MODIS-derived cloud data, when deseasonalized cloud fraction anomalies are computed, the correlation coefficient of high-, mid-, and low-level cloud fraction anomalies derived from MODIS and CALIPSO/CloudSat are, respectively, 0.49, 0.45, and 0.87 over the ascending branch, 0.72, 0.69, and 0.74 over the northern hemisphere descending branch, and 0.42, 0.37, and 0.87 over the southern hemisphere descending branch. The standard deviation of deseasonalized cloud fraction anomalies derived from MODIS is, however, smaller than the standard deviation of cloud fraction anomalies derived from CALIPSO/CloudSat for all cloud types over all branches.

The differences in the cloud fraction in turn affect surface irradiances computed with MODIS-derived cloud properties. In this study, we will further investigate how the cloud fraction difference affects surface irradiance computations.

Reference Loeb, N. G., D. Rutan, S. Kato, and W. Wang, 2014: Observing interannual variations in Hadley circulation atmospheric diabatic heating and circulation strength, submitted to J. Climate.