Characterising Uncertainty in Vegetation Optical Depth Retrievals Using the Land Parameter Retrieval Model

Microwave radiometers have been used to monitor the Earth’s land and oceans since the late 1970s, beginning with sensors such as the Scanning Multichannel Microwave Radiometer (SMMR). Due to the physical properties of the atmosphere, specifically the high transmissivity of atmospheric windows in much of the microwave spectrum, microwave radiation propagates with minimal attenuation, enabling observations through clouds and light precipitation. Depending on the frequency of observation, these sensors are largely unaffected by atmospheric and illumination conditions and can acquire measurements both day and night. Therefore, missions such as NASA’s Advanced Microwave Scanning Radiometer 2 (AMSR2) and Soil Moisture Active Passive (SMAP) provide frequent global mapping (typically every 2–3 days), however, retrievals are not always consistent, as other factors such as frozen soils and radio‑frequency interference can influence the measurements. A fundamental measurement collected by passive microwave sensors (radiometers) is brightness temperature, which serves as the primary input for retrieving parameters related to soil and vegetation water content, yielding products such as soil moisture and Vegetation Optical Depth (VOD). Passive sensors rely on naturally emitted microwave radiation from the Earth system and do not illuminate the surface, in contrast to active sensors (radars). VOD is not a directly measurable physical property, but a model-based parameter primarily estimated using remotely sensed data. By quantifying canopy opacity, VOD offers a critical proxy for Vegetation Water Content (VWC) and above-ground biomass. While high-frequency bands (e.g., C-, X-, and Ku-bands) interact primarily with leaves and small branches to reflect upper canopy VWC, longer wavelengths (e.g., L-band) penetrate deeper to interact with trunks and woody structure. This multi-band capability allows for a comprehensive assessment of ecosystem hydraulic status, drought impact, and given sufficient spatio-temporal coverage ecosystem resilience. However, while soil moisture is an established Essential Climate Variable with defined GCOS measurement uncertainty target for surface soil moisture (<0.08 m³m^-3, k=2) VOD lacks standardised guidance on uncertainty targets. This absence represents a critical gap in both product specifications and the scientific literature, limiting confidence in VOD interpretations and constraining its reliability as an indicator of vegetation water content in long‑term climate studies. Addressing this gap is therefore central to advancing the use of VOD in climate monitoring frameworks. This study explores the uncertainties associated with VOD retrievals within the Land Parameter Retrieval Model (LPRM), a widely used forward radiative transfer model. Utilising dual-polarised brightness temperature data from AMSR2 and performing Monte Carlo sensitivity analysis, we characterise how uncertainties in the model input parameters propagate through the VOD retrieval process. The research outlines a preliminary traceability diagram, identifying the sensitivity of the LPRM algorithm across different frequency bands and land cover types. By estimating uncertainty magnitudes under various scenarios, this work provides a framework for improving the reliability of VOD and VWC estimates, facilitating their integration into eco-hydrological models and early warning systems for vegetation stress.