The ESRL website will be unavailable for 24 hours starting

__Friday, March 6th at 5:00pm MT__due to building maintenance.## NCEP Reanalysis Atlas

Select output type for calculation details. Note some of the definitions have been slightly updated.

**Total Monthly Means**- Monthly means are calculated by averaging the 4 times daily data for January 1968 to December 1996 over all years (Feb 29 is not used) for each month.
**Individual Monthly Means**- Monthly means are calculated by averaging the 4 times daily data for January 1948 to pressent for each month.
**Monthly Transient Flux Means**- Monthly transient fluxes are calculated by averaging the 4 times daily data for January 1979 to December 1995 over all years (Feb 29 is not used) for each of the twelve months the quantity x'y' where x' is the difference from the total monthly mean X and y' is the difference from the total monthly mean Y. Each month is averaged for 17 full years.
**Interannual standard deviation**- Values for each month are calculated by summing over the squared difference between each years' monthly mean and the total monthly mean for each gridpoint.
**Total standard deviation**- Values for each month are calculated by summing the squared difference between the value averaged for one day and the monthly mean for that year for all years.
**3-10 Day filtered monthly means**- For single variables, a Lanczos filter with 101 weights is applied to each level of the the 1979-1995 data (with the longterm monthly mean removed for each month). For products, a filter is applied separately to each variable and then the product of the two fields is calculated. The ends of the data are not used so not all months have the same amount of years used for the means.
**3-10 Day filtered standard deviation**- For single variables, a Lanczos filter with 101 weights is applied to each level of the the 1979-1995 data (with the longterm monthly mean removed for each month). For products, a filter is applied separately to each variable and then the product of the two fields is calculated. The ends of the data are not used so not all months have the same amount of years used for the means.
**Diurnal Cycle by Month**- Values for each of the 4 times (0z,6z,12z,18z) are averaged over all years for each month for each separate level. A total mean is calculated for each month and the anomalies from these means for the 0z,6z,12z and 18z means are stored. Four months are presented in the atlas; January, April, July and October.
**EOF's of monthly anomalies for the Northern Hemisphere**- EOF's for each month were calculated from 17 years of anomaly data. Data for the Northern Hemisphere ONLY was used and a cosine weighting was applied to correct for the decrease in longitude spacing with latitude. EOF's were calculated separately for each level. What is plotted is the correlation between the principal component time series and the data at each grid point for the entire globe. The correlations were "corrected" to match the sign of the 1000mb level so that vertical plots could be made. However, there is no guarantee that the first EOF of one level will match the first EOF's of the other levels. In most cases, though, the correlation is quite high; especially in winter.