An error in the snow cover:
In preparation for the pre-1973 reanalysis, another error was discovered. The snow cover for the NH was obtained from a file of snow cover which starts in 1973 maintained by Don Garrett of CPC. The file was in ON84 format and had to be converted to GRIB. The NH snow records were first merged into one long file. The index array of the week number for each year was used to drive a do-loop, and a problem came is matching the FORTRAN to the file format. The concatenated file had been written as an unblocked file (-s unblocked) as required by GrADS.
Our previous experience was almost exclusively with blocked files. So instead of positioning the file at the beginning of the proper year, the file position kept pointing at January 1973, and the snow cover for 1973 was used also for the period 1974-1994. 1973, 1995 and 1996 have the correct snow cover. A preliminary evaluation of the impact of this mistake on the surface temperature follows.
Preliminary analysis suggests that the impact of using the 1973 snow cover throughout the 1974-1994 period affects the 2m temperature but that the impact is not dominant since its anomalies are driven more by the upper circulation than by the snow cover. Fig. 1, prepared by Suru Saha, shows the 2m temperature and the 2m temperature anomaly for the NCEP and the ECMWF reanalyses. The high latitude temperatures are lower for the ECMWF, which may be associated with a cold bias that has been found and recently corrected in their operations. The temperature anomalies, however, seem in rather good agreement between the two reanalyses. Figs. 2 and 3, prepared by Muthuvel Chelliah, show the 1974-1994 monthly average T2m temperature anomaly averaged over areas where the snow cover is most variable (northern US, 35N-45N, 127-165; and Asia, 35N-50N, 30E-120E). The figures show also the surface anomalies from CAMS data based on station observations, and their difference. There is a good correlation between the two estimates of the anomalies, and, most importantly, there is no noticeable jump between the agreement of reanalysis and observed anomalies before and after January 1995 (when the correct snow cover started to be used). This is more clearly shown on Fig. 4 which superimposes the observed and reanalysis temperature anomalies for 1993-1996. An examination of monthly winter temperature anomaly maps before and after January 1995 also shows a general good agreement between CAMS and reanalysis 2m temperature anomalies, and no noticeable improvement after the correct snow cover was used (not shown). We plan to run again 1979 with the correct snow to get a more complete assessment.