Search

Monitoring vegetation moisture conditions is paramount to better understand and assess drought impacts on vegetation, enhance crop yield predictions, and improve ecosystem models. Passive microwave remote sensing allows retrievals of the vegetation optical depth (VOD; [unitless]), which is directly proportional to the vegetation water content (VWC; in units of water mass per unit area [kg/m2]). However, VWC is largely dependent on the dry biomass and structure imprints on the VOD signal. Previously, statistical models have been used to isolate the water component from the biomass and structure components. Physically-based approaches have not yet been proposed for this goal. In this study, we present a multi-sensor semi-physical approach to retrieve the vegetation moisture from the VOD and express it as Live Fuel Moisture Content (LFMC [%]; the percentage of water mass per dry biomass unit). The study is performed in the western United States for the period April 2015 – December 2018. There, in situ LFMC samples are available for assessment. We rely on a VOD model based on vegetation height data from GEDI/Sentinel-2 and radar backscatter from Sentinel-1, which account for the biomass and structure components. Vegetation moisture is retrieved at L-, X- and Ku-bands by minimizing the difference between the modeled VOD and the VOD estimates from SMAP (L-band) and AMSR-2 (X- and Ku-band) satellites. Results show that the LFMC retrievals are independent of canopy height, land cover, and radar backscatter, demonstrating the capability of the proposed algorithm to separate water dynamics from the biomass/structure component in VOD. LFMC estimates at X- and Ku-bands reproduce well the expected spatio-temporal dynamics of in situ LFMC. Results show good agreement with in situ at a regional scale, with Pearson's correlations (r) between in situ LFMC samples and LFMC estimates of 0.64 (Ku-band), 0.60 (X-band) and 0.47 (L-band). Similar results are obtained independently for shrub and forest sites at X- and Ku-bands. In most comparisons between in situ and estimated LFMC, biases are below 10% of the dynamic range of LFMC. Performance at L-band is limited by the fact that this frequency senses the full vertical extent of the canopy, while in situ samples are taken only from top of canopy leaves to which X- and Ku-bands are much more sensitive. More insight will be needed for grasslands (r = 0.44 at X-band) using time-dynamic canopy height data. Furthermore, a pixel-scale assessment is conducted, showing a good agreement in most sites (r > 0.6). The proposed method can be tailored to exploit the synergies of past (e.g. , AMSR-E), current (e.g. , AMSR-2) and future satellite sensors such as CIMR and ROSE-L for global vegetation moisture mapping at different canopy layers. [Display omitted] • A multi-sensor semi-physical algorithm is applied to retrieve vegetation moisture. • The water component in vegetation optical depth is disentangled from dry biomass. • Vegetation moisture retrievals at X- and Ku-bands capture the in situ dynamics. • L-band has poorer performance due to lack of in situ samples in wood canopy layers. [ABSTRACT FROM AUTHOR]

Seasonal snow-derived water is a critical component of the water supply in the mountains and downstream regions, and the accurate characterization of available water in the form of snow-water-equivalent (SWE), peak SWE, and snowmelt onset are essential inputs for water management efforts. Arising from recent advancements in artificial intelligence (AI) and machine learning (ML), we introduce a large-scale ML SWE model leveraging publicly available data sources and open-source software. The model demonstrates the application of a limited feature space in a relatively simple ML architecture without the need for process-based formulations to effectively estimate spatially continuous SWE at a daily temporal resolution. Beginning with in situ SWE measurements (i.e., SNOTEL), lidar-derived terrain features, and temporal variables, we employ localized feature engineering and optimization via gradient-boosting decision trees to identify regionally unique drivers of snowpack dynamics and use the optimal features to train regionally independent artificial neural networks to estimate regional SWE at a 1 km spatial resolution. The model results yield respectable skill in reconstructed 1 km gridded SWE magnitudes in a hindcast simulation of the 2019 water year that is independent of the training and testing data. Comparing model estimates to over 6200 observations, the model demonstrates a weighted RMSE of 15.4 cm, Kling-Gupta Efficiency metric of 0.86, and a percent bias of 0.71% across 23 snow-influenced regions in the western U.S. The model simulation produces peak SWE estimates within 10 cm for twenty of the twenty-three regions, demonstrating capability in effectively capturing regional snow accumulation processes. The demonstration of low-error ML workflows capable of providing near-real-time, spatially continuous SWE estimates at a high spatial resolution provides proof-of-concept and a foundation to effectively update snow state variables that drive water supply forecasts in snow-dominated regions. • A new Machine Learning-based SWE estimation model for the Western United States. • Utilization of open-source tools and up-to-date in-situ observations to estimate SWE. • Reconstructed SWE magnitudes exhibit respectable skill in a hindcast simulation. • Output may serve as a supplemental state-of-snowpack input to hydrologic models. [ABSTRACT FROM AUTHOR]