Changes in U.S. Total Cloud Cover from Ground-Based Weather Observations- Real or Artificial?

Because satellite radiance-derived cloud cover may suffer from inhomogeneities due to orbital drifts, changes in instruments, and changes in viewing angles, cloud data from human observers remain an important source of climate change information. In the U.S., this record has been disrupted by the transition to automated observing systems starting in 1992. To minimize this discontinuity, we examine records from NWS stations where human observers have augmented the automated observations, and from U.S. military stations that did not begin automated observations until the mid-2000s. Using visual and statistical examination of the frequency distributions of total cloud cover values in the Integrated Surface Database (ISD) at NCDC, we find many discontinuities in U.S. total cloud data. Some of these do not result in noticeable changes in monthly mean total cloud, and some may be avoided by using subsets of the available data. However, the adoption of the METAR reporting system in 1996 coincides with significant increases in total cloud at the majority of stations we examined, changes that are not likely to be real. The mean increase in total cloud at this time for the U.S. stations in this study was 2-3% of total sky cover. Previous reports of significant upward trends in U.S. total cloud between 1973 and this decade based on military station data are probably biased by this change. Using a preliminary version of a new adjusted dataset of total cloud cover for U.S. stations, we will reassess trends for recent time periods and compare them to results from satellite datasets and related climate variables.