Evaluation of International Satellite Cloud Climatology Project (ISCCP) D2 cloud amount changes and their connections to large-scale dynamics

Abstract

The International Satellite Cloud Climatology Project (ISCCP) D2 dataset exhibits a 2.6% per decade decrease in the global all-cloud cloud amount from July 1983 through September 2001. This result is consistent with other recent findings that provide evidence that the cloud amount has decreased on a decadal-scale. Such changes in cloud amount should have an obvious impact on the climate system through changes in heating and the radiation budget of the atmosphere. However, the changes evident in the ISCCP data seem too large to be accepted without question. Because these data are used as a verification tool for the global climate modeling community, it is imperative that the nature of these changes are better understood and verified for similarities with other data sources. Otherwise, climate studies might be comparing their results with faulty information. This study represents an attempt to characterize and verify the ISCCP D2 cloud amount changes. One possible reason why the ISCCP D2 trend might be too large is the presence of artifacts in the data related to changes in the number of geosynchronous satellites in orbit. This leads to changes in the viewing angle for each pixel in the dataset and explains roughly one-third of the trend in the global cloud amount. In order to account for this phenomenon, this study focuses on the region from 90°E to 180° and 30°N to 30°S where the satellite coverage has been relatively constant. It is shown that the slope of the cloud amount change in this region is still very large. This leaves open the possibility that there is other contamination in the ISCCP data, and calls into question the validity of the large cloud amount trend. Several steps are taken to examine the nature of the cloud amount changes in this region. First of all, the changes in the ISCCP cloud amount data are characterized by three criteria: where and when the changes are occurring and the types of clouds expressing them. These patterns are examined for features that appear physically reasonable. These patterns can then be checked against patterns obtained from the NOAA Interpolated OLR and PATMOS-A cloud amount datasets. These data, from sensors mounted on polar-orbiting satellites, do not experience the viewing-angle problem of ISCCP but should still corroborate evidence of real cloud amount changes. The most unique aspect of this study is the use of reanalysis data to look for signals of climate change that are related to changes in the ISCCP cloud amount data. The average ISCCP all-cloud cloud amount for the region of interest is regressed onto wind fields, geopotential height fields, divergence fields, and other data that represent how the climate has changed over the span of the ISCCP dataset. Maps of regression coefficients represent how those fields change in response to a unit increase in cloud amount. These patterns help to identify atmospheric phenomena that are connected with variations in cloud amount in the region of interest. Furthermore, the true cloud amount trend in the region of interest can be diagnosed by making time series of how well the regression maps project onto reanalysis fields at each time step. These "proxy cloud time series" represent how the true cloud amount must be changing to effect the observable changes in the reanalysis data. Both results provide a unique way to discover whether the ISCCP D2 cloud amount changes are also evident in other data sources. It is shown that the cloud amount changes evident in the ISCCP D2 dataset are indicative of changes in the intensity and location of convection associated with the Inter-Tropical Convergence Zone (ITCZ). The spatial patterns of these changes are somewhat consistent with the NOAA Interpolated OLR and PATMOS-A cloud amount datasets. However the trends in the regionally averaged time series of these data are not significantly different from zero. This supports the conclusion that the ISCCP trend is too large. Using data from the NCEPNCAR reanalysis and the ERA-40 reanalysis, it is shown that the changes in the ISCCP D2 cloud amount time series in the region of interest are highly correlated with changes in the Walker-Hadley circulation. The patterns of these changes are consistent with the redistribution of convection indicated by each of the satellite datasets, and appear to be associated with ENSO since they are also consistent with the results of Bjerknes (1969). The reanalysis data also provide independent confirmation that the actual cloud amount in the region of interest is likely not changing in a statistically significant way during the period spanned by the ISCCP D2 dataset. Therefore, while the variability of cloud amount due to ENS0 is evidently captured by the ISCCP D2 dataset, the long-term trend in the ISCCP cloud amount is likely not physically realistic.