The estimation of Soil Organic Carbon (SOC) from optical image spectroscopy typically relies on the availability of bare-soil conditions, which are increasingly rare due to the widespread adoption of conservation agriculture practices. This study evaluates alternative strategies for SOC prediction under limited bare-soil availability by comparing four methodological approaches based on Sentinel-2 imagery and related products: (i) bare-soil multispectral composites, (ii) vegetation indices, (iii) AlphaEarth Satellite Embeddings, and (iv) a hybrid geostatistical-machine learning model (KpR-Cubist). These methods were tested across three cropland regions with contrasting pedoclimatic conditions: Italy, France, and Taiwan. The evaluation relied on >1800 topsoil samples collected between 2020 and 2024. Results show that bare-soil availability varies significantly by region, with cloud cover and vegetation/farm management being the main limiting factors. Models using Satellite Embed-dings consistently achieved the highest predictive accuracy (RPIQ up to 2.24) , outperforming conventional bare-soil composite and vegetation-based models. Incorporating spatial coordinates further improved model performance, revealing strong spatial autocorrelation in SOC distribution. The hybrid kriging-Cubist approach achieved comparable accuracy to the embedding-based models, confirming the value of integrating spatial dependence into data-driven frameworks. Overall, the study demonstrates that deep-learning-derived satellite embeddings models provide effective alternatives for SOC estimation in croplands where bare-soil imagery is increasingly unavailable due to sustainable soil management practices.

Monitoring soil organic carbon (SOC) dynamics is crucial for sustainable agriculture, soil health, and climate change mitigation. In this study, we assessed 5-year SOC changes in conventional (CO) and conservation agriculture (CA) fields across France and Italy using a combination of Sentinel-2 satellite imagery and machine learning (ML) models. A total of 922 topsoil samples (0-20 cm) were collected to build a training dataset representing the main soil types and climatic regions. Satellite-derived spectral reflectance, obtained through a pixelwise temporal mosaicking approach (median and 90th percentile composites), was paired with laboratory-measured SOC to train soil-type-specific Cubist models, which were validated both internally and against the independent LUCAS dataset. The models demonstrated high predictive performance (normalized Root Mean Square error = 0.18-0.21, Ratio of Performance to InterQuantile range approximate to 2.0) and were then applied to 48 CA fields where conservation practices had been in place for at least 5 years, allowing the mapping of SOC values before and after the implementation of these practices. To provide a benchmark, SOC mapping was also conducted for 48 CO fields, covering the same time window (2018-2023). Results indicate that CA practices led to significant SOC increases in most fields, particularly in Cambisols and Luvisols, with an average Delta SOC of +0.95 g kg-1 over 5 years, whereas CO fields showed negligible changes. While Italy and France differed in baseline SOC levels, CA fields in both countries exhibited SOC gains tied to higher SOC: clay ratio and improved soil health, demonstrating the tangible benefits of crop rotation, permanent soil cover, and reduced tillage. This study demonstrates that combining Sentinel-2 imagery with validated ML models enables robust monitoring of SOC dynamics across contrasting management systems, soil types, and climatic conditions. The approach provides a scalable and cost-effective tool to evaluate agricultural practices and to guide strategies for soil health and carbon sequestration.

Soil spectroscopy is increasingly used to provide accurate and affordable prediction of various soil properties. However, variability in instrument characteristics and operational protocols still hinders the integration and harmonization of Visible-Near InfraRed-Short Wave InfraRed (VNIR-SWIR) Soil Spectral Libraries (SSLs). This study examines the use of 99% pure silica sand, Lucky Bay (LB), as an Internal Soil Standard (ISS), to reduce spectral variability and correct the systematic errors across laboratories. A ring trial test using 60 samples across 11 laboratories was conducted to assess the effect of the ISS correction on spectral consistency and model performance for Soil Organic Carbon (SOC) and clay content predictions. Spectral correction using the ISS reduced the Standard Deviation (SD) in reflectance by 13-70%, effectively reducing dissimilarity across instruments. Partial Least Squares Regression (PLSR) modeling using mean spectra of all instruments showed that a standard protocol resulted in prediction accuracy (Ratio of Performance to InterQuartile range (RPIQ) = 1.50) comparable to a frequently used reference instrument (Foss XDS, RPIQ = 1.57). While models built on individual datasets performed well, combining non-corrected spectra reduced prediction performance (RPIQ) by 11%, which was decreased to 8% after ISS correction. Therefore, we concluded that this approach is particularly useful for increasing interoperability of soil spectral datasets acquired across different instruments and laboratory conditions, which is key in the development of large-scale SSLs. Furthermore, reflectance variation at 1700 nm of the ISS was found as a practical Quality Assurance and Quality Control (QA/QC) indicator, offering a real-time baseline for evaluating instrument performance.

Soil Organic Carbon (SOC) plays a critical role in the global carbon cycle. Accurately estimating SOC in cultivated lands is essential for assessing their carbon sequestration potential, overall soil quality, and for developing adaptive agricultural management strategies, serving as an indicator of soil health and degradation. Traditional methods for SOC mapping are usually time-consuming, however, the rapid advancement of remote sensing technology offers a novel approach to SOC estimation. This research describes a new method for SOC mapping, evaluating the current capabilities of remote sensing technologies, focusing on multispectral data. The highly frequent revisit time Sentinel-2 data underwent extensive preprocessing, first, to isolate bare soil areas. Then, using ground truth data and remote sensing data, a dataset was created and used to train machine learning regression algorithms. Various regression methods were tested. Our results indicate that Sentinel-2 data can be effectively used for SOC estimation and that the proposed method works adequately. These findings highlight the potential of remote sensing data for SOC mapping and emphasize the need for further research in this field.

Multispectral imaging satellites such as Sentinel-2 are considered a possible tool to assist in the mapping of soil organic carbon (SOC) using images of bare soil. However, the reported results are variable. The measured reflectance of the soil surface is not only related to SOC but also to several other environmental and edaphic factors. Soil texture is one such factor that strongly affects soil reflectance. Depending on the spatial correlation with SOC, the influence of soil texture may improve or hinder the estimation of SOC from spectral data. This study aimed to investigate these influences using local models at 34 sites in different pedo-climatic zones across 10 European countries. The study sites were individual agricultural fields or a few fields in close proximity. For each site, local models to predict SOC and the clay particle size fraction were developed using the Sentinel-2 temporal mosaics of bare soil images. Overall, predicting SOC and clay was difficult, and prediction performances with a ratio of performance to deviation (RPD) > 1.5 were observed at 8 and 12 of the 34 sites for SOC and clay, respectively. A general relationship between SOC prediction performance and the correlation of SOC and clay in soil was evident but explained only a small part of the large variability we observed in SOC prediction performance across the sites. Adding information on soil texture as additional predictors improved SOC prediction on average, but the additional benefit varied strongly between the sites. The average relative importance of the different Sentinel-2 bands in the SOC and clay models indicated that spectral information in the red and far-red regions of the visible spectrum was more important for SOC prediction than for clay prediction. The opposite was true for the region around 2200 nm, which was more important in the clay models.

Soil organic matter (SOM) is a key factor in sustaining soil fertility, sequestering greenhouse gases and reducing soil erosion, in this regard, an accurate estimation and monitoring of the SOM content is crucial for sustainable land management and climate change mitigation strategies. In recent years, there has been a growing consciousness of the need to better understand the dynamics of SOM across different farm management in time and space. In this context, the main objective of the study is to improve understanding regarding the relationship between SOM and the main farming systems adopted in Italy by taking spatial correlation into account. For this purpose, a large dataset consisting of topsoil SOM values (0 - 20 cm) and environmental and farming information was collected in 597 locations (145 fields and 62 farms) representative of the whole agricultural area of Po Valley in Italy. This sizable dataset was analyzed by a novel geospatial analysis using a de-clustering approach in combination with polygon kriging for detecting and understanding the SOM variability over the different fields characterized by irregular shapes and different farming systems. The results provided clear evidences of the spatial correlation between SOM, farming systems and soil types. Higher SOM contents were detected in Cambisols (3.11 %) and in field managed according conservation agriculture practices (3.22 %) as compared to other farming systems. Moreover the inclusion of fodder crops in the rotation and the use of no-tillage are two of the most effective practices for increasing and preserving SOM according to our findings. Spatial information, such those provided in this study, could facilitate the delineation of tailored solutions for each European Member State for targeting future actions related to carbon farming, and offering crucial insights to support advancements in agriculture for enhancing soil fertility and health and for fostering sustainable agricultural practices.

Soil salinity is considered one of the biggest constraints to crop production, particularly in arid and semi-arid regions affected by recurrent and long periods of drought, where high salinity levels severely impact plant stress and consequently agricultural production. Climate change accelerates soil salinization, driven by factors such as soil conditions, land use/land cover changes, and water deficits, over extensive spatial and temporal scales. Continuous monitoring of areas at risk of salinization plays a critical role in supporting effective land management and enhancing agricultural production. For these purposes, this work aims to propose a spatiotemporal method for monitoring soil salinization using spectral indices derived from Earth observation data. The proposed approach was tested in the Zaghouan Region in northeastern Tunisia, a region where soils are characterized by alarming levels of salinization. To address this concern, remote sensing techniques were applied for the analysis of satellite imagery generated from Landsat 5, Landsat 8, and Landsat 9 missions. A comprehensive field survey complemented this approach, involving the collection of 229 geo-referenced soil samples. These samples were representative of distinct soil salinity classes, including non-saline, slightly saline, moderately saline, strongly saline, and very strongly saline soils. Soil salinity modeling using Landsat-8 OLI data revealed that the SI-5 index provided the most accurate predictions, with an R2 of 0.67 and an RMSE of 0.12 dS/m. By 2023, 42.3% of the study area was classified as strongly or very strongly saline, indicating a significant increase in salinity over time. This rise in salinity corresponds to notable land use and land cover (LULC) changes, as 55.9% of the study area experienced LULC shifts between 2000 and 2023. A decline in vegetation cover coincided with increasing salinity, showing an inverse relationship between these factors. Additionally, the results highlight the complex interplay among these variables demonstrating that soil salinity levels are significantly impacted by climate change indicators, with a negative correlation between precipitation and salinity (r = −0.85, p < 0.001). Recognizing the interconnections between soil salinity, LULC changes, and climate variables is essential for developing comprehensive strategies, such as targeted irrigation practices and land suitability assessments. Earth observation and remote sensing play a critical role in enabling more sustainable and effective soil management in response to both human activities and climate-induced changes.

Visible–near-infrared–shortwave-infrared (VNIR–SWIR) spectroscopy is one of the most promising sensing techniques to meet ever-growing demands for soil information and data. To ensure the successful application of this technique in the field, efficient methods for tackling detrimental moisture effects on soil spectra are critical. In this paper, mathematical techniques for reducing or removing the effects of soil moisture content (SMC) from spectra are reviewed. The reviewed techniques encompass the most common spectral pre-processing and algorithms, as well as less frequently reported methods including approaches within the remote sensing domain. Examples of studies describing their effectiveness in the search for calibration model improvement are provided. Moreover, the advantages and disadvantages of the different techniques are summarized. Future research including further studies on a wider range of soil types, in-field conditions, and systematic experiments considering several SMC levels to enable the definition of threshold values for the effectiveness of the discussed methods is recommended.

This work explores the use of multitemporal hyperspectral PRISMA data to estimate clay, sand, organic carbon (SOC) and calcium carbonates. A pixel-wise approach was carried out obtaining composite bare soil data from: (1) median spectra; (2) spectra at the minimum value of the soil surface moisture index S2WI; (3) at the maximum value of the bare soil index; (4) the 90th quantile; and (5) the 10th quantile. These data were used for the calibration of machine learning regression algorithms. The minimum of the S2WI was the extraction method, and support vector regression was the algorithm, in the best performing calibrated models. Clay was estimated with the best performance, with a RMSE of 5.25% and an R2 of 0.89. SOC provided good results, with a RMSE of 0.84 g/kg and R2 of 0.86.

The use of remote sensing data methods is affordable for the mapping of soil properties of the plowed layer over croplands. Carried out in the framework of the ongoing STEROPES project of the European Joint H2020 Program SOIL, this work is focused on the feasibility of Sentinel-2 based approaches for the high resolution mapping of topsoil clay and organic carbon (SOC) contents at the within-farm or within-field scales, for cropland sites of contrasted climates and soil types across the Northern hemisphere. Four pixelwise temporal mosaicking methods, using a two years-Sentinel-2 time series and several spectral indices (NDVI, NBR2, BSI, S2WI), were developed and compared for i) pure bare soil condition (maxBSI), ii) driest soil condition (minS2WI), iii) average bare soil condition (Median) and iv) dry soil conditions excluding extreme reflectance values (R90). Three spectral modeling approaches, using the Sentinel-2 bands of the output temporal mosaics as covariates, were tested and compared: (i) Quantile Regression Forest (QRF) algorithm; (ii) QRF adding longitude and latitude as covariates (QRFxy); (iii) a hybrid approach, Linear Mixed Effect Model (LMEM), that includes spatial autocorrelation of the soil properties. We tested pairs of mosaic and spectral approaches on ten sites in Turkiye, Italy, Lithuania, and USA where soil samples were collected and SOC and clay content were measured in the lab. The average RPIQ of the best performances among the test sites was 2.50 both for SOC (RMSE = 0.15%) and clay (RMSE = 3.3%). Both accuracy level and uncertainty were mainly influenced by site characteristics of cloud frequency, soil types and management. Generally, the models including a spatial component (QRFxy and LMEM) were the best performing, while the best spatial mosaicking approaches mostly were Median and R90. The most frequent optimal combination of mosaicking and model type was Median or R90 and QRFxy for SOC, and R90 and LMEM for clay estimation.

Proximal and remote sensing methods are currently investigated in order to meet the growing demand for fast and cheap analysis of soils. For this reason, also at European level, several funds are employed in order to optimize the procedure for obtaining the best results. In this context, in the framework of the European Joint Programme EJP SOIL, the project ProbeField (https://ejpsoil.eu/soil-research/probefield) is funded to develop quick and simple methods to analyse soil properties directly in the field, namely, without the need to bring soil to the laboratory, overcoming the obstacles of the data acquisition at field condition. Among the available proximal sensors, visible and near infrared spectroscopy has the greatest potential, especially for soil organic carbon and clay minerals that have specific spectral signatures in the visible and near infrared spectral region. ProbeField’s ultimate goal is developing a protocol for the best practices for soil mapping directly in the field based on proximal sensors.

The accurate, timely, and non-destructive estimation of maize total-above ground biomass (TAB) and theoretical biochemical methane potential (TBMP) under different phenological stages is a substantial part of agricultural remote sensing. The assimilation of UAV and machine learning (ML) data may be successfully applied in predicting maize TAB and TBMP; however, in the Nordic-Baltic region, these technologies are not fully exploited. Therefore, in this study, during the maize growing period, we tracked unmanned aerial vehicle (UAV) based multispectral bands (blue, red, green, red edge, and infrared) at the main phenological stages. In the next step, we calculated UAV-based vegetation indices, which were combined with field measurements and different ML models, including generalized linear, random forest, as well as support vector machines. The results showed that the best ML predictions were obtained during the maize blister (R2)–Dough (R4) growth period when the prediction models managed to explain 88–95% of TAB and 88–97% TBMP variation. However, for the practical usage of farmers, the earliest suitable timing for adequate TAB and TBMP prediction in the Nordic-Baltic area is stage V7–V10. We conclude that UAV techniques in combination with ML models were successfully applied for maize TAB and TBMP estimation, but similar research should be continued for further improvements.

The PRISMA satellite is equipped with an advanced hyperspectral Earth observation technology capable of improving the accuracy of quantitative estimation of bio-geophysical variables in various Earth Science Applications and in particular for soil science. The purpose of this research was to evaluate the ability of the PRISMA hyperspectral imager to estimate topsoil properties (i.e., organic carbon, clay, sand, silt), in comparison with current satellite multispectral sensors. To investigate this expectation, a test was carried out using topsoil data collected in Italy following two approaches. Firstly, PRISMA, Sentinel-2 and Landsat 8 spectral simulated datasets were obtained from the spectral resampling of a laboratory soil library. Subsequently, bare soil reflectance data were obtained from two experimental areas in Italy, using real satellites images, at dates close to each other. The estimation models of soil properties were calibrated employing both Partial Least Square Regression and Cubist Regression algorithms. The results of the study revealed that the best accuracies in retrieving topsoil properties were obtained by PRISMA data, using both laboratory and real datasets. Indeed, the resampled spectra of the hyperspectral imager provided the best Ratio of Performance to Inter-Quartile distance (RPIQ) for clay (4.87), sand (3.80), and organic carbon (2.59) estimation, for the spectral soil library datasets. For the bare soil reflectance obtained from real satellite imagery, a higher level of prediction accuracy was obtained from PRISMA data, with RPIQ ± SE values of 2.32 ± 0.07 for clay, 3.85 ± 0.19 for silt, and 3.51 ± 0.16 for soil organic carbon. The results for the PRISMA hyperspectral satellite imagery with the Cubist Regression provided the best performance in the prediction of silt, sand, clay and SOC. The same variables were better estimated using PLSR models in the case of the resampled hyperspectral data. The statistical accuracy in the retrieval of SOC from real and resampled PRISMA data revealed the potential of the actual hyperspectral satellite. The results supported the expected good ability of the PRISMA imager to estimate topsoil properties.

Intra-field heterogeneity of soil properties, such as soil organic carbon (SOC), nitrogen (N), phosphorous (P), exchangeable cations, pH, or soil texture, is a function of complex interactions between biological factors, physical factors, and historic agricultural management. Mapping the crop growth and final yield heterogeneity and quantifying their link with soil properties can contribute to an optimization of amendment/fertilizer application and crop yield in a management variable zones (MVZ) approach. To this end, we studied a field of 17 ha consisting of four former fields that were merged in early 2017 and cropped with winter wheat in 2018. Historical management practices data were collected. The topsoil characteristics were analyzed by grid-based sampling and kriged to create maps. We tested the capacity of a multispectral MicaSense® RedEdge-MTM camera sensor embedded on an unmanned aerial vehicle (UAV) to map in-season growth of winter wheat. Relating several vegetation indices (VIs) to the plant area index (PAI) measured in the field highlighted the red-edge NDVI (RENDVI) as the most suitable to follow the crop growth throughout the growing season. The georeferenced final grain yield of the winter wheat was measured by a combine harvester. The spatial patterns in RENDVI at three phenological stages were mapped and analyzed together with the yield map. For each of these images a conditional inference forest (CI-forest) algorithm was used to identify the soil properties significantly influencing these spatial patterns. Historical management practices of the four former fields have induced significant heterogeneity in soil properties and crop growth. The spatial patterns of RENDVI are rather constant over time and their Spearman rank correlation with yield is similar along the growing season (r ≃ 0.7). Soil properties explain between 87% (mid-March) to 78% (mid-May) of the variance in RENDVI throughout the growing season, as well as 66% of the variance in yield. The pH and exchangeable K are the most significant factors explaining from 15 to 26% of the variance in crop growth. The methodology proposed in this paper to quantify the importance of soil parameters based on the CI-forest algorithm can contribute to a better management of amendment/fertilizer inputs by stressing the most important parameters to take into consideration for site-specific management. We also showed that heterogeneity induced by the soil properties can be described by a crop map early in the season and that this crop map can be used to optimize soil sampling and thus amendment/fertilizer management.

There is a need to update soil maps and monitor soil organic carbon (SOC) in the upper horizons or plough layer for enabling decision support and land management, while complying with several policies, especially those favoring soil carbon storage. This review paper is dedicated to the satellite-based spectral approaches for SOC assessment that have been achieved from several satellite sensors, study scales and geographical contexts in the past decade. Most approaches relying on pure spectral models have been carried out since 2019 and have dealt with temperate croplands in Europe, China and North America at the scale of small regions, of some hundreds of km2: dry combustion and wet oxidation were the analytical determination methods used for 50% and 35% of the satellite-derived SOC studies, for which measured topsoil SOC contents mainly referred to mineral soils, typically cambisols and luvisols and to a lesser extent, regosols, leptosols, stagnosols and chernozems, with annual cropping systems with a SOC value of ~15 g·kg−1 and a range of 30 g·kg−1 in median. Most satellite-derived SOC spectral prediction models used limited preprocessing and were based on bare soil pixel retrieval after Normalized Difference Vegetation Index (NDVI) thresholding. About one third of these models used partial least squares regression (PLSR), while another third used random forest (RF), and the remaining included machine learning methods such as support vector machine (SVM). We did not find any studies either on deep learning methods or on all-performance evaluations and uncertainty analysis of spatial model predictions. Nevertheless, the literature examined here identifies satellite-based spectral information, especially derived under bare soil conditions, as an interesting approach that deserves further investigations. Future research includes considering the simultaneous analysis of imagery acquired at several dates i.e., temporal mosaicking, testing the influence of possible disturbing factors and mitigating their effects fusing mixed models incorporating non-spectral ancillary information.