The brightness temperature (T B p) observed by the Soil Moisture Active Passive (SMAP) satellite mission is significantly affected by the soil permittivity (ε s), surface roughness and vegetation opacity (τ p). This study assesses the impact of these factors on simulating the SMAP horizontally (p = H) and vertically (p = V) polarized T B p measurements and retrieving the liquid soil water content (θ liq) for both frozen and thawed soils in the typical Tibetan desert and meadow ecosystems. For this investigation, the zero-order approximation of the radiative transfer equations, i.e., τ-ω emission model, is configured with surface roughness and τ p parameterizations adopted by current SMAP soil moisture retrieval algorithms, and the ε s is computed with the four-phase dielectric mixing model that is applicable for both frozen and thawed soils. For the Tibetan desert site, the τ-ω emission model with above configurations underestimates year-round the SMAP T B H measurements (bias > 20 K), while T B V are underestimated during the cold season. Implementation of a new surface roughness parameterization reduces the T B H underestimation, and the improved T B p simulations lead to better θ liq retrievals produced by the single channel algorithm (SCA) using the T B V as well as T B H measurements. The remaining T B H and T B V underestimations are removed by further adopting a new ε s parameterization. For the Tibetan meadow site, the τ-ω emission model overestimates both T B H and T B V during the warm season and underestimates T B H during the cold season when the vegetation is sparse. Implementation of the new surface roughness parameterization reduces the T B H underestimation, and further the T B p overestimation is mitigated by adopting a new τ p parameterization derived from a discrete radiative transfer model previously developed and tested for the same site. The in-situ measured θ liq dynamics are better captured by corresponding retrievals for both frozen and thawed soils with implementation of the new surface roughness and τ p parameterizations, which reduces the unbiased RMSEs by more than 40%. The parameterizations developed in this study are useful to provide consistent and reasonable T B p simulations and θ liq retrievals over the Tibetan Plateau for both frozen and thawed soils based on both SMAP T B H and T B V measurements.