Journal Paper Digests 2026 #10
- DeepProfile: An inverse fusion framework for root zone soil moisture profile estimation
- Alternatives to Equivalent Soil Mass in Monitoring, Reporting and Verification of Changes in Soil Carbon
- Organic Carbon and Texture Control Moisture Dependence of Soil Shortwave Infrared Reflectance
Organic Carbon and Texture Control Moisture Dependence of Soil Shortwave Infrared Reflectance
Moisture content affects soil reflectance in the optical domain (400–2500 nm), acting as a confounding factor in soil property prediction models. Soil reflectance needs to be simulated efficiently for varying levels of soil moisture, in order to aid soil property prediction efforts and inform physical land surface models. Here, we built on previous work that investigated how soil reflectance decreases with increasing soil moisture. We explored how the relationship between the reflectance and soil moisture content changes as a function of wavelength and soil characteristics. For this purpose, we acquired the spectra of 28 soil samples from various locations across Europe in a laboratory setting, at different levels of soil moisture. The soil reflectance-moisture relationship was found to be wavelength-dependent and best represented by decreasing exponential functions. The rates of exponential decrease, however, varied across soil samples and were normalised to isolate effects of different soil characteristics. It was found that organic carbon (OC), clay and silt content displayed a statistically significant relationship with the normalisation factor, a proxy for how quickly soil ‘darkens’ with increasing soil moisture content. A multiple linear regression model was used to describe the normalisation factor based on OC content and soil textural information. The resulting model was able to explain 67% of the variance, with OC and clay content accounting for almost 70% of the relative feature importance. Our findings call for the inclusion of OC content and textural information, especially clay content, in physical models of soil moisture-reflectance, for more efficient simulations of soil reflectance at varying levels of soil moisture, to support climate models and soil property predictions efforts based on field and remotely sensed data.
Alternatives to Equivalent Soil Mass in Monitoring, Reporting and Verification of Changes in Soil Carbon
Sequestering carbon in soils is a key action to address climate change and food security. Schemes incentivising farmers to change land management practices to sequester more carbon in soils are underpinned by soil monitoring protocols. Accurate estimation of soil organic carbon (SOC) stocks is essential for the integrity of such carbon credit schemes. Common SOC estimation methods like sampling to fixed depth are prone to errors due to changes in bulk density over time, particularly under changing management practices. Equivalent Soil Mass (ESM), utilising a reference mass rather than a reference volume for SOC estimation, arguably alleviates this. In practice, the potentially large variation in sampled core lengths (and thus masses) still introduces substantial variability into SOC estimates. This work compares four approaches to SOC estimation: (1) ESM10, based on the 10th percentile of sampled masses, currently implemented in the Australian Soil Carbon Method; (2) ESMμ, based on the mean of sampled masses, and (3) EBD10 and (4) EBDμ, built on the concept of equivalent bulk density (EBD), based on either the 10th percentile or mean of the average soil sublayer bulk densities. Variability in ESM and EBD and in resulting SOC estimates was quantified using soil cores from nine intensively sampled farms in eastern Australia. The proposed alternatives, particularly ESMμ and EBDμ offered more stable and accurate SOC estimates, reducing variance by up to 38%–86% compared to ESM10. These findings can be applied to support the evolution of improved methods of soil carbon monitoring, reporting and verification in Australia and internationally.
DeepProfile: An inverse fusion framework for root zone soil moisture profile estimation
Root zone soil moisture (RZSM) is a critical variable for understanding land–atmosphere interactions, hydrological processes, and agricultural productivity. Direct remote sensing of RZSM remains challenging due to the shallow sensing depth and the ill-posed nature of inversing a profile, with the existing global RZSM products mainly derived from model-based data assimilation. These products offer valuable information but exhibit inconsistent accuracy and disparate vertical discretizations. As no single root-zone soil moisture product is superior globally, fusing them offers a means to integrate their complementary strengths into a unified and consistent framework. In this study, a DeepProfile framework was proposed for estimating the continuous soil moisture profile throughout the top 100 cm layer of soil by integrating three widely used RZSM products; Soil Moisture Active Passive level 4 (SMAP L4), Global Land Data Assimilation System (GLDAS) version 2, and the fifth generation European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis (ERA5-land). Unlike traditional fusion methods requiring harmonized inputs, DeepProfile treats parent products as learning targets, optimizing the integral of a polynomial profile to match heterogeneous layers without enforcing identical vertical or spatiotemporal coverage. This yields a continuous analytical profile for flexible depth extraction, utilizing location-specific triple collocation weights to optimally balance product contributions. Evaluation against in-situ measurements from 2373 stations across 45 global networks demonstrated strong agreement in near-surface and intermediate layers (≤50 cm), with median RMSE values below 0.06 m3/m3 and correlation coefficients (R) exceeding 0.72. The use of SMAP near-surface soil moisture was found critical for the satisfactory results for the top 50 cm. The model also showed promising performance at deeper layers of >50 cm (R > 0.65), although accuracy declined with depth due to weaker observational constraints. The proposed DeepProfile offers a scalable and transferable solution for generating depth-resolved soil moisture estimates, with potential applications in hydrological modeling, drought monitoring, and weather forecasting.