JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 101, NO. D3, PAGES 7291-7303, MARCH 20, 1996 Impact of analysis uncertainty upon regional atmospheric moisture flux Muyin Wang 1and Jan Paegle Department of Meteorology, Universityof Utah, Salt Lake City Abstract. Horizontal fluxes of atmospheric water vapor are studied for summer months during 1989 and 1992 overNorth and South Americaand arebased on analyses from European Centre for Medium Range Weather Forecasts, U.S. National Meteorological Center,andUnited Kingdom Meteorological Office. The calculations are performed over20øx 20ø box-shaped midlatitude domains located to the east of the Rocky Mountains in North America andto the eastof the Andes Mountains in South America. The fluxesare determined from operational center gridded analyses of wind andmoisture. Differences in the monthlymeanmoisture flux divergence determined from these analyses are as largeas 7 cm/month precipitable water equivalent over South America and3 cm/month overNorth America. Gridded analyses at higherspatial andtemporal resolution exhibitbetteragreement in the moisture budget study. However, significant discrepancies of the moisture flux divergence computed from different gridded analyses still exist. The conclusion is morepessimistic thanRasrnusson's [ 1968] estimate based on station data. Further analysis reveals thatthemostsignificant sources of errorresult frommodelsurface elevation fields,gaps in the dataarchive, anduncertainties in the wind and specific humidityanalyses. Uncertainties in the wind analyses arethe mostimportant problem. The low-level jets, in particular, are substantially different in the different dataarchives.Part of the reason for this may be dueto the way the different analysis models parameterize physical processes affecting low-level jets. The results support the inference thatthe noise/signal ratio of the moisture budget may be improved more rapidly by providing better wind observations andanalyses thanby providing better moisture data. 1. Introduction Water vapor is a relatively minor constituent of the Earth's atmosphere but a major factor in atmospheric energetics, radiation, andtransport and conversion of latent heat. Its variabilityand anomalies determine drought and flood episodes, and consequently, modulate basicelements of the regional climate. Although the importance of atmospheric water vapor is well recognized, its temporal and spatial variability has been systematically explored only in relatively restricted regions. Many earlierstudies of the waterbudget were based on subsets of the approximately 800 stations compiled within the Massachusetts Instituteof Technology (MIT) General Circulation Data Library underthe direction of V. P. Starr. These dataconsist of radiosonde soundings analyzed at 50- mbar intervals over the northern hemisphere and tropical portions of the southern hemisphere for the period May 195 8 to April 1963. Rosen et al. [1979] used this data set to investigate yearly averages of verticallyintegrated precipitable water, and zonal and meridional atmospheric water vapor 1 Now at Atmospheric Science Program, Department of Physics, Dalhousie University,Halifax, Nova Scotia, Canada Copyright 1996 by the American Geophysical Union. Papernumber95JD02896. 0148-0227/96/95JD-02896505.00 transportfor a 5 year period and discovered substantial interannual variations. Rosen and Omolayo[ 1981 ] studied the flux of water vapor across the coastlines of the northern hemisphere for composite seasons of the station data. A series of investigations by Rasmusson [1966a, b, 1967, 1968] utilized the same datasetbut focused uponthe water transportover North America. Rasmusson [1967] investigated atmospheric water vaporfluxes in more detail over North America, and confirmed the importance of the diurnal cycle and emphasized the relevance of the southerly low-level jet located east of the Rocky Mountains for the moisture balance. Rasmusson [1968] points out that a complete hydrological cycle is difficult to define because evapotranspiration and changes in ground storage of water are largely unknown,yet he was able to obtain estimates for evaporation-precipitation (E-P) based upon the horizontal flux divergenceof the atmospheric moisture using the 5-year data set of the MIT General Circulation Data Library. It is shown that the diurnal variations of the flux divergence are on the sameorder of magnitude as the computed mean flux divergence. Integrationof the flux divergence fields over sufficiently large regions reduced the random components of the error [Rasmusson, 1966b], and good results can be obtained for areas about 20 x 105 km 2 and larger. Horizontally integrated mapsof atmospheric flux divergence were used as proxies for E-P together with streamflow data to compute soil storage over southern Canada and the United States. Application of this analysis to the Central Plains and eastern United States led to an estimate of the horizontalflux divergence of water vapor that is accurate 7291