Recent observations within the substorm recovery phase postmidnight diffuse aurora reveal phenomena that are not well understood. Strong neutral winds are seen within the postmidnight diffuse aurora in measurements from the series of Atmospheric Response in Aurora (ARIA) campaigns, which show a great deal of vertical structure and large vertical shears. The ARIA observations also consistently show a peak in the wind magnitude between around 110 and 120 km, which has not been well explained, with a strong shear below the peak that is sometimes found to be unstable, with a Richardson number < 0.25. A three-dimensional (3-D) high-resolution model developed recently at UCLA has been used to perform simulations to try to understand the poorly explained phenomena found in the ARIA measurements. Model simulations show that large zonal and meridional winds can be produced in the region of the postmidnight diffuse aurora using auroral forcing parameters within the range of observations. A wind maximum is produced in the 110--120 km altitude range, within the same height range as the maxima found in the ARIA observations, using auroral forcing alone. When sufficiently high vertical resolution is used for auroral forcing parameters within the range of observations, the region below the peak may be unstable, with a Richardson number < 0.25. When simulations are performed with auroral forcing based on measurements from the ARIA I campaign, and background winds and tides are introduced from the Coupled Thermosphere--Ionosphere Plasmasphere (CTIP) model <Fuller-Rowell et al., 1987; Bailey et al., 1997>, the basic features of the vertical structure of the ARIA I wind measurements are reproduced, although the magnitude of the peak around 115 km altitude is somewhat smaller than that observed. The 115 km peak is likely produced by a combination of auroral and tidal forcing processes. The vertical structure of the winds is sensitive to the phases of the tidal modes. Improvements in the ability of the model to simulate the observations from the ARIA campaign may be due to the use of the 3-D rather than a 2-D high-resolution model.