In this paper, six semi-empirical bidirectional polarization distribution functions (BPDF) models for polarized reflectance of land surfaces (Nadal–Bréon model, Waquet model, Maignan model, Litvinov model, Diner model and Xie-Cheng model) were quantitatively intercompared using the recently released database of representative BPDFs generated from POLDER measurements over a wide range of surface types at global scale. The intercomparison technique involved two strategies: one for fitting and the other for priori modeling. Our results suggest that (1) Nadal–Bréon model and Litvinov model provide best fits to the POLDER measurements with average RMSEs equal to 0.174% and 0.173%, respectively; (2) as for priori modeling, Xie-Cheng model performs best among these models as its average RMSE is 0.249%, indicating the corresponding surface-type-based free parameters can be used for a priori model of surface polarized reflectance; (3) despite the semi-empirical models cannot estimate negative polarized reflectance at backward scattering direction, impact of negative polarized reflectance on fitting and priori modeling is negligible. The results provide a priori knowledge of the model performances over various surface types and can be applied to future researches of land optical properties and aerosol parameters retrieval over land surfaces.