The impact of "chi-corrected" winds upon the moisture budget over North America determined from the NCEP-NCAR reanalysis wind and humidity fields is examined. Previous researchers have suggested that notable differences which exist in the divergent wind field between reanalyses, even in well-observed regions such as North America, can negatively impact estimates of the moisture budget. That the error in the analyzed wind fields is predominantly in the divergent component of the wind, and not in the rotational component, is consistent with the fact that the large-scale vorticity analyses produced at different data centers are in much better agreement than the corresponding divergence analyses. One way to correct the analyzed divergence is by constraining the winds to minimize imbalances in both the mass and vorticity budgets, thus enforcing dynamical consistency upon the divergent circulation. This approach, known as the "chi-problem" (Sardeshmukh 1993), has been applied to the four-times daily NCEP-NCAR reanalysis for the years 1948-2001. The correction is applied at all 28 sigma levels at a spectral truncation of T21. This results in a greatly diminished vorticity budget imbalance at all levels. In particular, we find that over North America the chi-correction acts to slightly lower the altitude of the summertime low-level jet (LLJ), consistent with profiler data showing the LLJ centered about 50 mb lower than in reanalysis data (Higgins et al 1997).

The chi-correction is a conservative adjustment to the winds, well within the observational margin of error. Yet it has a pronounced impact upon the vertically integrated moisture flux convergence and hence estimates of precipitation minus evaporation (P-E) throughout the year. The chi-corrected winds alter estimates of P-E not only over the Mississippi Basin but also in the summer monsoon region and along the California coast. Over the GCIP region, changes in P-E are predominantly in the spring and summer. During the warm season, a slightly stronger chi-corrected LLJ results in a larger estimate of P-E (that is, a smaller estimate of evaporation), with an increase of about 0.25 mm/day for the monthly mean climatologies (averaged during 1973-1998) of P-E during the warm season months, and an increase in the annual mean of P-E from 0.28 mm/day to 0.39 mm/day. Greater changes occur for isolated events. For example, P-E over Iowa during July 1993, the peak of the 1993 flood, is increased by almost a factor of two.