Wesley Ebisuzaki, Masao Kanamitsu
Climate Prediction Center, NCEP
Washington, DC
Jerry Potter, Michael Fiorino
PCMDI, Lawrence Livermore National Laboratory
Livermore, CA
NCEP/DOE AMIP-II Reanalysis (Reanalysis-2) is a follow-on project to the NCEP/NCAR Reanalysis Project (N/N Reanalysis). In Reanalysis-2, global analyses are made using an updated forecast model, updated data assimilation system, improved diagnostic outputs and fixes for the known processing problems of N/N Reanalysis.
Reanalysis-2 should be considered an updated N/N Reanalysis and not a next-generation reanalysis. Although Reanalysis-2 has some significant improvements, a next-generation reanalysis would have much higher resolution, assimilation of rainfall and radiances, a much improved forecast model, (better) use of SSM/I data and be based on better theoretical techniques such as 4D-variational assimilation.
Preliminary results from Reanalysis-2 have been encouraging. The fixes to the human processing errors have made major changes to some of the fields and changes to the system itself have lead to other significant improvements.
One unique feature of Reanalysis-2 is the treatment of soil moisture. N/N Reanalysis forced the soil moisture by using the model precipitation and a 60-day nudging term to climatology. Compared to Illinois observations. the annual cycle in N/N Reanalysis was too strong. Perhaps the soil model in Reanalysis was incompatible with the model used to compute the soil-moisture climatology.
In experiments the nudging term was removed which resulted in unrealistically dry regions in South America. This is suggestive of a a positive feedback between the model precipitation and the soil moisture. In order to avoid the positive feedback amplifying the precipitation bias, Reanalysis-2 uses observed precipitation to correct the errors in the model precipitation.
The observed precipitation (pentad means based on gauge and satellite estimates from Xie-Arkin) was used to correct the errors in the model precipitation when the soil was not frozen. In cases of no model runoff, the error in the model precipitation during the previous pentad is added to the upper soil layer uniformly in time. For example if the the model precipitation was low by 1 mm/day then 1 mm/day was added top soil layer during the next pentad.
The soil-moisture correction scheme is modified in cases of (model) runoff. Errors in the model precipitation are assumed to affect the runoff rather than the soil moisture. The only exception is when the observed precipitation is less than the water entering the top soil layer. In this case, soil moisture is removed by an amount of ((model precip. - runoff) - obs. precip.).
The evaluations of Reanalysis-2 are still preliminary. However, the short-wave fluxes show a significant improvement (Fig. 1). Overall the new convective scheme and humidity diffusion have improved the precipitation especially in the summer over the South-East United States and over the polar regions (Fig. 2). Fixing of the snow mask leads to significant 2-m temperature differences where the mask was in error (Fig. 3) With the new soil-moisture correction scheme, the soil moisture now has a much stronger interannual variability (Fig. 4) (data from J. Huang).
fixes the known processing errors in the N/N Reanalysis as well as uses an improved forecast model and data assimilation system. Beside providing, in our opinion, a better reanalysis; Reanalysis-2 can used with N/N Reanalysis to understand the sensitivity of the analysis to changes in the model parameterizations. Such sensitivity is one factor affecting the accuracy of the various reanalysis efforts. More information and analyses can be obtained from reanalysis2 home page.
