Abstract: The bare surface backscattering coefficient is a key foundation for quantitative remote sensing modeling and surface parameter inversion in microwave radar. However, due to factors such as the high cost of field measurements, limited sample sizes, and complex observation conditions, data-driven modeling that relies solely on measured data is prone to overfitting and suffers from severely limited generalization capabilities. Meanwhile, physical models, which are based on theoretical assumptions and parameter simplifications, struggle to effectively eliminate systematic biases between the models and actual observations. To address these issues, this paper proposes a modeling method for bare surface backscattering coefficients based on simulation-to-measurement knowledge transfer. The method is implemented in two stages. First, a physical model is used to generate a large-scale simulated dataset across a wide range of surface parameters (such as roughness and soil moisture). This dataset serves as a source-domain pre-training benchmark model, enabling it to fully learn the nonlinear mapping relationship between surface physical quantities and scattering responses. Subsequently, a limited number of real field measurement samples are introduced as target domain data. By employing a fine-tuning strategy, the method facilitates knowledge transfer from theoretical simulation to actual measurements. Experimental results reveal that the proposed method significantly improves prediction accuracy and robustness under small-sample conditions, providing reliable technical support for high-precision, rapid modeling of bare surface scattering and subsequent remote sensing inversion.