The NCEP/NCAR 50-year reanalysis data set has been widely used by the climate community to assess decadal time scale global change and climate variations. For such purposes, it is crucial that the data set does not contain spurious secular trends or biases. Such biases could be introduced into the reanalysis data due to many reasons, among them changes in data quality, quantity, and coverage, as well as the availability of new data types. While there have been suggestions that the quality of the reanalysis is better during the 1990s compared to the 1950s and 1960s, there have not been any clear quantitative measure of how the data quality has changed over the years.

In this study, we will make use of collocated observations, first to assess whether there have been changes in the quality of observations. Of particular interest to us is the quality of the reanalysis in the mid-latitude upper troposphere. Several previous studies have compared collocated aircraft observations from short periods of time to quantify the observational error associated with aircraft observations. In this study, we are extending such comparisons all the way from the late 1960s to the end of the 20th century to examine whether there have been changes in the observational error statistics. Preliminary results show that there has been a slight increase in observational accuracy. The observations are then compared to reanalysis data, and the results suggest that the differences between the observations and reanalysis during the late 60s and early 70s are much bigger than the degradation of the observations themselves, suggesting that the reanalysis data error is much larger during those periods. Quantitative results will be presented at the conference.

We will be extending our analyses to rawinsonde and surface ship observations. The ECMWF 40-year reanalysis will also be assessed if it becomes available later this year.