The mixing height is an important parameter for many boundary layer applications such as air quality, but is difficult to parameterize accurately because of the many physical processes that contribute to its evolution. Prognostic equations used to model the mixing height are generally derived from a parameterization of the TKE budget equation either averaged over the entire mixing layer or applied at the top of the mixing layer. The equations proposed by different investigators vary in complexity and applicability based on which terms in the TKE budget have been neglected and how the remaining terms are parameterized. In addition, a host of techniques exists for analyzing profiles of wind and scalars to extract the mixing height. Two commonly used techniques are the parcel method and the bulk Richardson number method. In the last two decades, remote sensors have been developed that are useful for detecting the height of the mixing layer. Among these are the 915-MHz boundary layer radar wind profilers. In this paper we compare the mixing height evolutions observed by wind profilers to the mixing heights predicted by several mixing height prognostic and diagnostic routines used, for example, in mesoscale numerical models and the meteorological preprocessors to dispersion models. To evaluate the different routines, measurement of one or more of the following parameters is required: temperature and wind profiles, surface sensible heat and momentum fluxes, and lapse rate of potential temperature above the capping inversion. We therefore are restricted to observational data sets that include rawinsondes and measurements of the turbulent fluxes. We are further required to focus on daytime periods when the mixing height is above the minimum detectable range of the wind profilers (150–200 m AGL)