Statistical structure functions are used to evaluate sounding data from the 6–7 March day of the 1982 AVE/VAS Ground Truth Field Experiment. Functional analyses are performed for five observation times starting at 1200 GMT 6 March and ending at 0000 GMT 7 March, and for the composite 12 h period. Data consist of mesoscale soundings from a special ground truth rawinsonde network and VAS-derived soundings from both a physical algorithm and a regression technique. The standard parameters of temperature, geopotential height, and mixing ratio are evaluated at the 850, 700, 500, 300 and 200 mb levels. Integrated parameters of thickness and precipitable water also are investigated.
Using structure function analyses, estimates of root-mean-square (rms) data uncertainty are obtained for the three data sources. Then, VAS soundings from the physical retrieval scheme are compared with those from the regression technique. Results indicate that both schemes have similar error characteristics and capabilities for determining gradients of mesoscale temperature and geopotential height. Signal-to-noise ratios for these parameters were quite favorable and greater than those of mixing ratio. Finally, sounding retrievals are evaluated against those from the ground-truth rawinsonde network. These results show that the VAS data generally describe weaker gradients than observed with the radiosondes. A notable exception is physically-derived mixing ratio at 850 mb.