First Vegetation Signatures from a Multi-Constellation, Multi-Frequency, and Polarimetric GNSS-Transmissometry and GNSS-Reflectometry Experiment over the Amazon Rainforest

L-band Vegetation Optical Depth (VOD) retrievals in tropical forest almost always use the τ–ω model, which parameterizes canopy scattering with an effective single-scattering albedo and usually neglects higher-order scattering. This work reports nine months of continuous observations from a four-channel ground-based GNSS-Transmissometry/Reflectometry (GNSS-T/GNSS-R) instrument on the 45 m walk-up tower at the LBA K67 site, Tapajós National Forest, Brazil. Four antennas feed two dual-channel receivers for the direct, reflected (LHCP and RHCP), and canopy-transmitted (RHCP) signals. Three direct observational diagnostics are consistent with a canopy volume-scattering contribution. The transmitted opacity increases with incidence angle but recovers only part of the expected path-length dependence, shows no dependence on the LiDAR-derived canopy height, and the circular polarization ratio departs from the angular decline of a coherent reflection. A joint retrieval recovers the apparent volume-scattering amplitude Gvω0 as a positive residual of the measured return over the modeled coherent two-way ground path. Gvω0 is an effective scattering residual, not the τ–ω single-scattering albedo ω. For GPS L1 it decreases from 0.85[0.59,1.19] to 0.41[0.28,0.57] through the dry-to-wet transition. This seasonal magnitude is an internal-consistency result from the measurement chain, without independent validation. It depends on two conservative choices. A smooth-surface soil model maximizes the subtracted coherent ground return. A two-term decomposition omits the canopy–ground interaction paths.