A method for estimating mixing length in the prediction of turbulent kinetic energy by the Navy's Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS) has been documented (Walters and Miller 1999, Walters and Miller 2001). The sensitivity of the method to initial conditions and to model vertical and horizontal resolution will be presented for conditions of low and moderate upper-tropospheric turbulence which occurred on 18 April 2000 over Albuquerque, New Mexico, with the approach of a cut-off low pressure system. The variability of the COAMPS turbulent kinetic energy solution for (1) cold-start initializations using NOGAPS or ETA as first guess fields, (2) a varying number of allowable first guess model vertical levels, and (3) different COAMPS horizontal and vertical resolutions will be compared to observations of integrated optical turbulence. Turbulence in COAMPS is generated using a modified form of the Mellor and Yamada (1982) 2.5 closure scheme for parameterizing mixing length with the Bougeault and Lacarrere (1989) scheme for computing optical turbulence.