Mangrove forests provide vital ecosystem services. The increasing threats to mangrove forest extent and fragmentation can be monitored from space. Accurate spatially explicit quantification of key vegetation characteristics of mangroves, such as leaf area index (LAI), would further advance our monitoring efforts to assess ecosystem health and functioning. Here, we investigated the potential of radiative transfer models (RTM), combined with active learning (AL), to estimate LAI from Sentinel-2 spectral reflectance in the mangrove-dominated region of Ngoc Hien, Vietnam. We validated the retrieval of LAI estimates against in-situ measurements based on hemispherical photography and compared against red-edge NDVI and the Sentinel Application Platform (SNAP) biophysical processor. Our results highlight the performance of physics-based machine learning using Gaussian processes regression (GPR) in combination with AL for the estimation of mangrove LAI. Our AL-driven hybrid GPR model substantially outperformed SNAP (R² = 0.77 and 0.44 respectively) as well as the red-edge NDVI approach. Comparing two canopy RTMs, the highest accuracy was achieved by PROSAIL (RMSE = 0.13 m².m⁻², NRMSE = 9.57%, MAE = 0.1 m².m⁻²). The successful retrieval of mangrove LAI from Sentinel-2 can overcome extensive reliance on scarce in-situ measurements for training seen in other approaches and present a more scalable applicability by relying on the universal principles of physics in combination with uncertainty estimates. AL-based GPR models using RTM simulations allow us to adapt the genericity of RTMs to the peculiarities of distinct ecosystems such as mangrove forests with limited ancillary data. These findings bode potential for retrieving a wider range of vegetation variables to quantify large-scale mangrove ecosystem dynamics in space and time.

Ground‐Penetrating Radar (GPR) has recently become a powerful geophysical technique to characterize soil moisture at the field scale. We developed a data assimilation scheme to simultaneously estimate the vertical soil moisture profile and hydraulic parameters from time‐lapse GPR measurements. The assimilation scheme includes a soil hydrodynamic model to simulate the soil moisture dynamics, a full‐wave electromagnetic wave propagation model, and petrophysical relationship to link the state variable with the GPR data and a maximum likelihood ensemble assimilation algorithm. The hydraulic parameters are estimated jointly with the soil moisture using a state augmentation technique. The approach allows for the direct assimilation of GPR data, thus maximizing the use of the information. The proposed approach was validated by numerical experiments assuming wrong initial conditions and hydraulic parameters. The synthetic soil moisture profiles were generated by the Hydrus‐1D model, which then were used by the electromagnetic model and petrophysical relationship to create “observed” GPR data. The results show that the data assimilation significantly improves the accuracy of the hydrodynamic model prediction. Compared with the surface soil moisture assimilation, the GPR data assimilation better estimates the soil moisture profile and hydraulic parameters. The results also show that the estimated soil moisture profile in the loamy sand and silt soils converge to the “true” state more rapidly than in the clay one. Of the three unknown parameters of the Mualem‐van Genuchten model, the estimation of n is more accurate than that of α and K_s. The approach shows a great promise to use GPR measurements for the soil moisture profile and hydraulic parameter estimation at the field scale.

We coupled dielectric mixing models with a full-wave ground-penetrating-radar (GPR) model to estimate the soil water content by inversion. Two mixing models were taken into account in this study, namely, a power law model and the Wang and Schmugge model. With this combination, we could account for the frequency dependence of the dielectric permittivity and apparent conductivity in the inverse algorithm and directly estimate the soil water content without using an empirical petrophysical formula or a priori knowledge on soil porosity. The approach was validated by a series of experiments with sandy soil in controlled laboratory conditions. The results showed that the performance of our approach is better than the common approach, which assumes a linear dependence of apparent conductivity on frequency and uses Topp’s equation to transform permittivity to water content. GPR data were perfectly reproduced in the time and frequency domains, leading to very accurate water-content estimates with an average absolute error of less than 0.013 cm3/cm3. However, the accuracy was reduced as the water content increased. Sensitivity analysis indicated that the Green’s function was most sensitive to the water content and sand-layer thickness but much less so with DC conductivity. The results also revealed that as the frequency increased, although the permittivity was nearly constant, the apparent electrical conductivity and the attenuation increased remarkably, especially for wet sands due to dielectric losses. The successful validation of the proposed approach opens a promising avenue of development to use dielectric mixing models for soil-moisture mapping from GPR measurements.

Major components of hydrologic and elemental cycles reside underground, where their complex dynamics and linkages to surface waters are obscure. We delineated seasonal subsurface flow and transport dynamics along a hillslope in the Rocky Mountains (USA), where precipitation occurs primarily as winter snow and drainage discharges into the East River, a tributary of the Gunnison River. Hydraulic and geochemical measurements down to 10 m below ground surface supported application of transmissivity feedback of snowmelt to describe subsurface flow and transport through three zones: soil, weathering shale, and saturated fractured shale. Groundwater flow is predicted to depths of at least 176 m, although a shallower limit exists if hillslope‐scale hydraulic conductivities are higher than our local measurements. Snowmelt during the high snowpack water year 2017 sustained flow along the weathering zone and downslope within the soil, while negligible downslope flow occurred along the soil during the low snowpack water year 2018. We introduce subsurface concentration‐discharge (C‐Q) relations for explaining hillslope contributions to C‐Q observed in rivers and demonstrate their calculations based on transmissivity fluxes and measured pore water specific conductance and dissolved organic carbon. The specific conductance data show that major ions in the hillslope pore waters, primarily from the weathering and fractured shale, are about six times more concentrated than in the river, indicating hillslope solute loads are disproportionately high, while flow from this site and similar regions are relatively smaller. This methodology is applicable in different representative environments within snow‐dominated watersheds for linking their subsurface exports to surface waters.