Stomatal conductance (gs) is a key variable in Earth system models as it regulates the transfer of carbon and water between the terrestrial biosphere and the lower atmosphere. Various approaches have been developed that aim for a simple representation of stomatal regulation applicable at the global scale. These models differ, among others, in their response to atmospheric humidity, which induces stomatal closure in a dry atmosphere. In this study, we compared the widely used empirical Ball-Berry and Leuning stomatal conductance models to an alternative empirical approach, an optimization-based approach, and a semimechanistic hydraulic model. We evaluated these models using evapotranspiration (ET) and gross primary productivity (GPP) observations derived from eddy covariance measurements at 56 sites across multiple biomes and climatic conditions. The different models were embedded in the land surface model JSBACH. Differences in performance across plant functional types or climatic conditions were small, partly owing to the large variations in the observational data. The models yielded comparable results at low to moderate atmospheric drought but diverged under dry atmospheric conditions, where models with a low sensitivity to air humidity tended to overestimate gs. The Ball-Berry model gave the best fit to the data for most biomes and climatic conditions, but all evaluated approaches have proven adequate for use in land surface models. Our findings further encourage future efforts toward a vegetation-type-specific parameterization of gs to improve the modeling of coupled terrestrial carbon and water dynamics