An L-band radiometry for retrieving surface soil moisture (SM) is challenged by the confounding influence of surface roughness scattering and vegetation attenuation. This study presents a novel retrieval framework that uses Global Navigation Satellite System-Reflectometry (GNSS-R) coherence time metrics from SM Active-Passive-Reflectometry (SMAP-R) data to characterize global surface roughness and vegetation parameters. By analyzing the decay of signal-to-noise ratio (SNR) across varying coherent integration times, we estimate the temporal coherence time ( tau c ) and derive surface roughness ( sigma ) without ancillary topographic data. A polarization mixing parameter ( Pmix ), derived from the degree of polarization (DoP), is related to the polarization decoupling factor ( Q ) used in SMAP models. These variables feed into a tau - omega radiative transfer model to retrieve vegetation optical depth (VOD) ( tau ) and single-scattering albedo ( omega ) directly from observations. The resulting global maps of sigma , Q,tau , and omega show enhanced spatial structure and biome sensitivity. They can constrain the empirical SMAP static ancillary inputs used now for constraining the inversion of L-band brightness temperatures for SM. To evaluate the impact of the new vegetation parameters, we apply a neural network trained on SMAP data to retrieve SM using both the original and derived VOD inputs. Differences in SM are spatially coherent and physically consistent, with lower values in high-biomass regions and improved retrieval in arid zones. This work demonstrates that GNSS-R coherence metrics can be integrated with the L-band radiometry to produce observation-driven land surface parameters. The approach reduces dependence on empirical or static climatological inputs and supports improved SM retrievals from passive microwave missions.