During the 1998 Leonid meteor shower, multi-instrument observations of persistent meteor trains were made from the Starfire Optical Range on Kirtland Air Force Base, New Mexico, and from a secondary site in nearby Placitas, New Mexico. The University of Illinois Na resonance lidar measured the Na density and temperature in the trains, while various cameras captured images and videos of the trains, some of which were observed to persist for more than 30 min. The Na density measurements allow the contribution of Na airglow to the observed train luminescence to be quantified for the first time. To do this, persistent train luminescence is numerically modeled. Cylindrical symmetry is assumed, and observed values of the Na density, temperature, and diffusivity are used. It is found that the expected Na luminosity is consistent with narrowband CCD all-sky camera observations, but that these emissions can contribute only a small fraction of the total light observed in a 0.5--1 &mgr; bandwidth. Other potential luminosity sources are examined, in particular, light resulting from the possible excitation of monoxides of meteoric metals (particularly FeO) and O2(b1&Sgr;g+) during reactions between atmospheric oxygen species and meteoric metals. It is found that the total luminosity of these combined processes falls somewhat short of explaining the observed brightness, and thus additional luminosity sources still are needed. In addition, the brightness distribution, the so-called hollow cylinder effect, remains unexplained. ¿ 2001 American Geophysical Union