Effective monitoring of soil and vegetation properties on a global scale is essential for better understanding climate changes, hydrological dynamics, and ecological processes. Passive microwave remote sensing at C-band radio frequency, with long observation period and relatively high penetration capability, has been widely used to retrieve soil moisture (SM) and vegetation optical depth (C-VOD). The retrieval process is generally achieved by inversion of the ti-w radiative transfer model, which depends on crucial parameters such as effective scattering albedo (w) and soil roughness (HR) R ) for accurate retrievals. Current SM/C-VOD retrieval algorithms, such as the Land Parameter Retrieval Model (LPRM), predominantly rely on globally-constant w and HR R values, ignoring the inherent sensitivity of those parameters to varying soil conditions and vegetation types. To evaluate the impact of w and HR R variables on SM and C-VOD retrievals and to improve their accuracy, this study proposed and evaluated a novel retrieval approach from AMSR2 C-band observations during 2017-2020 using the C-band Microwave Emission of the Biosphere (C-MEB) model. We evaluated two new retrieval algorithms, considering either a globally-constant calibration or a land cover-based calibration of w and HR. R . As a benchmark for the calibration, we optimized the values of w and HR R by evaluating the retrieved SM against in situ measurements from the International Soil Moisture Network (ISMN) and OzNet hydrological monitoring networks. The main originality compared to previous algorithms is that i) it includes a comprehensive calibration exploring the optimal values of w and HR, R , applicable globally or tailored to specific land cover; ii) field SM measurements were leveraged to constrain the calibrated value of w and HR. R . For the globally-constant calibration, the optimal values of w = 0.05 and HR R = 0.1 were found to yield the best results. For the land cover-based calibration, an inverse relationship between w/HR R and canopy height was observed, with w ranging from 0.04 to 0.06 and HR R ranging from 0.1 to 0.7 for heights between 0 and 30 m. The algorithm employing a land cover-based calibration (INRAE Bordeaux 2, IB2) exhibited better performance than the one utilizing a globally-constant calibration (INRAE Bordeaux 1, IB1) in evaluating retrieved SM against in situ measurements, as well as in evaluating C-VOD vs various vegetation variables including aboveground biomass (AGB), tree cover, canopy height and several optical vegetation indices. Comparison with LPRM suggested that our IB2 C-VOD retrievals present improved performances in terms of both spatial and temporal results with all considered vegetation variables (spatial correlation (R) between various vegetation variables and C-VOD of 0.76-0.83 for IB2 vs 0.69-0.79 for LPRM), and exhibited lower saturation effects when compared with AGB. In addition, the IB2 SM produced lower root mean squared error (RMSE) (0.147 vs 0.217 m3/m3), 3 /m 3 ), bias (-0.03 vs 0.09 m3/m3), 3 /m 3 ), and ubRMSE (0.066 vs 0.067 m3/m3) 3 /m 3 ) when compared with in situ measurements, although it showed a lower R compared to LPRM SM.