An Eulerian chemical model is used to assess the relative importance of a variety of processes associated with producing high surface ozone episodes during selected periods of the North American Research Strategy for Tropospheric Ozone (NARSTO) 1995 field campaign over the northeastern United States. A comparison of the observed and predicted hourly surface ozone mixing ratios showed that the model qualitatively reproduced the observed ozone trends over the northeastern United States. The model had a mean hourly bias of ~15 ppb with a wide range of skill compared to individual observing stations clustered around one node. The simulated mass budget tendency terms are compared for days with low ozone values immediately followed by days with high values. The later days showed observed and simulated ozone mixing ratios aloft to be of order twice that found on preceding days, although the associated chemical mix appeared to have relatively little potential for the subsequent generation of new ozone. Under conditions of shallow mixing over urban regions, simulated surface ozone production rates were negative (a net loss) throughout much of the day with convective mixing bringing newly produced ozone from aloft to the surface. It is noted that surface ozone levels appeared to be relatively insensitive to mixing layer growth rates. ¿ 2000 American Geophysical Union