I. Velicogna} (CIRES and Dept. of Physics, Univ of Colorado, Boulder, CO 80309, USA; phone: 1-303-4925141; fax: 1-303-4927935; e-mail: isabella at colorado.edu)

J. Wahr (CIRES and Dept. of Physics, Univ of Colorado, Boulder, CO 80309, USA; fax: 1-303-4927935; e-mail: wahr@colorado.edu)

Many satellite geodetic techniques, including (among others) GPS geodesy, satellite altimetry and the new generation of gravity satellite missions such as GRACE, use the atmospheric pressure field to correct for the atmospheric contribution to the observables. Different methods have different requirements for accuracy and temporal and spatial sampling, but each would benefit greatly from improved constraint of pressure. To examine the quality of available pressure data, we have compared sea level pressure from NCEP/NCAR's (National Centers for Envi ronmental Prediction/National Center for Atmospheric Research) global Reanalysis model and the ECMWF (European Center for Medium-Range Weather Forecasts) model with sea level pressure measurements in the United States and the Mediterranean region. Observed pressure data are themselves subject to transcription and other errors which must be addressed; we used a variance analysis to simply remove significant outliers in the pressure measurement time series. Comparison indicates significant discrepancies between the models and observations, with L$_2$-norm residuals that can be larger than 1 mbar. Average errors appear to be correlated with regions of high topographic relief, possibly relating to poor resolution of altitude effects by the coarse grids of the models (2.5 degree grid spacing for the NCEP/NCAR reanalysis model; 1.125 degrees for the ECMWF model). We show that the L$_2$-norm residuals between models and observations are generally around 1-1.5 mbar, except in areas of high topographic relief where the discrepancy can be bigger. A 1 mbar error in the pressure correction would make uncertainties in pressure the limiting error source for using GRACE in hydrological applications at scales greater than 200 km. We are investigating (using bootstrapping techniques) the spatial distribution of surface barometer observing stations that would be required to improve pressure field constraint to the accuracy required for satellite geodetic corrections.