Upper fronts are associated with strong horizontal gradients of both temperature and wind speed on a scale that is not well resolved by rawinsonde observations. Even so, mesoscale data assimilation systems are capable of ingesting observations from a variety of sources and depicting such features. This study examines upper fronts that occurred over the continental U.S. during March-April 1996 with the objective of verifying the performance of the NORAPS (Navy Operational Regional Atmospheric Prediction System) data assimilation system using ACARS (ARINC Communications, Addressing, and Reporting System) aircraft observations. ACARS observations are taken every 5 to 8 minutes during level flight, which yields a horizontal resolution along the flight track of less than 100 km and so can resolve the approximately 200 km width scale for upper fronts. The ACARS temperature observations are not currently used in the data assimilation system and so present an independent set of observations. Thirty distinct upper fronts (duration greater than 12 h and temperature gradient greater than 2 deg C/100 km) were identified and tracked from the NORAPS analyses during the period of the study. In general, the analyzed temperature gradient was weaker than that observed in the ACARS data. The latter depicted a temperature gradient of 8 deg C/100 km for two cases, whereas the analyzed gradient did not exceed 6 deg C/100 km. Most upper fronts (47%) attained maximum intensity when located in the base of the upper level trough, although 33% (13%) did so just downstream (upstream) of the trough line. Most of the useable aircraft tracks were near 200-300 mb, therefore the portion of the upper front above the tropopause was examined in greater detail than the portion below the tropopause, although the latter would be expected to contain stronger temperatui->h(0g(2gd4i((*"