Several recent studies have found that the Northern Hemisphere storm tracks have undergone a secular increase in intensity during the second half of the twentieth century. However, most of these studies are based on analyses of the NCEP/NCAR reanalysis data, and some studies have suggested that eddy variance and covariance statistics, which are frequently used to define storm track intensity, may be subject to quite severe biases due to model biases and changes in the observational network and quality of observations. These biases may have introduced spurious climate signals into the data set.

In this paper, we examine changes in seasonal mean flow using 500 hPa height fields produced by the NCEP/NCAR reanalysis data, and mean sea level pressure field based on the work of Trenberth and Paolino. Seasonal mean height and MSLP fields are less affected by model biases and changes in observational network compared to variance/covariance statistics, in particular, the MSLP field is quite strongly constrained by surface observations. Thus the quality of these fields are expected to be much better than that of the eddy variance/covariance fields, especially over oceans where the amount of upper level observations has undergone significant changes over the years. In addition, the Trenberth and Paolino MSLP field is independent of the NCEP/NCAR reanalysis, thus serving as an independent check of the results.

Previous studies have shown that storm track and mean flow variations are strongly tied to each other, with storm track changes both forcing mean flow anomalies as well as being forced by changes in the mean flow. We employ a canonical correlation analysis on detrended monthly mean flow anomalies and 300 hPa eddy meridional velocity variance (which serves as a proxy for storm track activity) anomalies. The analysis is performed on data since 1979, during a period when the observational network is relatively uniform. Using the CCA relationship thus obtained, we use seasonal mean flow anomalies for the years 1949 to 1999 to "predict" the changes in storm track activity over the second half of the twentieth century. The results suggest that observed mean flow changes are consistent with a secular increase in storm track activity. However, the "predicted" magnitude of storm track intensification is less than that seen in the NCEP/NCAR reanalysis data. Possible reasons for such discrepancies will be discussed in the paper.