Vegetation Optical Depth (VOD) is a microwave-derived indicator linked to vegetation water content (VWC) and above-ground biomass. In this study, we improve VOD estimates by optimising key parameters of the Land Parameter Retrieval Model (LPRM) through minimising residuals between modelled brightness temperatures and those observed by AMSR2 at 10.7 GHz. Guided by a global sensitivity analysis that reveals strong param eter interactions, we jointly optimise surface roughness, effective temperature, and single scattering albedo at each location. We evaluate two optimisation schemes, which differ only in the effective-temperature input: a Microwave scenario, which uses AMSR2 Ka-band observations, and an ERA5 scenario, which uses ERA5-Land skin temperature. Both optimisation strategies reduce model residuals and strengthen the agreement between VOD and MODIS leaf area index (LAI) seasonal cycles, with the largest gains in forests. On seasonal-cycle agreement with MODIS LAI, the ERA5 scenario delivers the larger improvement relative to the Microwave scenario. Performance diverges for anomalies: VOD from the ERA5 scenario degrades in arid regions but retains gains in forests, whereas the Microwave optimisation maintains or improves anomaly agreement more broadly. Consistent with this, the com parison against in situ Normalised Microwave Reflectance Index, a VWC proxy, shows higher correlations for the Microwave scenario, while ERA5 fails to track day-to-day VWC variability. Both approaches resolve known boreal seasonality issues. Finally, improvements in VOD do not necessarily translate into better soil moisture estimates (SM), confirming known VOD-SM trade-offs.