Abstract
An objective upper level analysis procedure, which attempts to optimize smaller scale features, is described. It makes use of quantities derived for an array of overlapping triangles formed from adjacent radiosonde stations; these values and the direct observations are then smoothed to an array of grid points.
Tests of the procedure, applied to analytically prescribed wave functions, suggest that atmospheric features with characteristic wavelength scales ≲1700 km cannot be adequately described by data from the existing radiosonde network over the United States. Some implications of this limitation, as regards short-range forecasting, are discussed.
