Search

The aims of this research are (i) to compare random forest (RF), boosted regression tree (BRT), and multinomial logistic regression (MnLR) models to prepare the prediction maps of soil great group and subgroup levels, (ii) determination of the most important environmental covariates influencing the production of digital soil mapping (DSM) in an arid climate, (iii) to evaluate the efficiency of spectra indices extracted from Sentinel-2A digital images and data capability of ALOS-PALSAR radar data, and (iv) investigating the effect of sub-surface genetic horizons in the modeling of different types of soil map classes distribution. The principal component analysis method was employed to select the best set from the pool of environmental covariates (n = 46) such as geomorphometric parameters (GPs), RS indices, and diagnostic soil properties (DSP). The relative importance results indicate that Gypsic (GYP) subsurface horizon, standardized height (StH), slope length (SL), and normalized different vegetation index (NDVI) had an important role in the prediction of soil classes compared to the other selected covariates. DSM methodology was used in this research by incorporating of RF model and representative soil-forming factors that can be used for preparing the maps of soil classes in low-relief areas with a similar soil-landscape relationship. Totally, in this study places a spotlight on the profound impact of sub-surface genetic horizons, shedding light on their pivotal role in accurately modeling soil class distributions. These findings not only advance our comprehension of soil variability in arid regions but also hold immense implications for the burgeoning field of pedometrics. [ABSTRACT FROM AUTHOR]

Digital mapping of soil organic carbon (SOC) is essential for visualizing the spatial distribution at different regions and scales. However, existing studies using remote sensing are limited by the low spectral resolution of multispectral data for accurate estimation of SOC, and the failure of laboratory visible and near-infrared and shortwave infrared (Vis-NIR) spectroscopy to perform pixel-level mapping. To address these limitations, this study proposes a SOC mapping framework that integrates satellite earth observations and proximal sensing spectral data to reconstruct images having a spectral resolution of 400–2500 nm and a spatial resolution of 10 m, then the clustering probability analysis was applied on the reconstructed images, to develop a high accuracy SOC prediction model and demonstrate its effectiveness across geographical and spatial scales. Specifically, a total of 324 topsoil samples were collected from Baoqing and Suiling counties in northeast China, whose laboratory Vis-NIR reflectance spectra and SOC were measured. Sentinel-2A images of bare soils averaged over three years before and after the samples collection (i.e., 2018 to 2021) were collected. A total of 264 samples from eight fields in Belgium were used to evaluate the performance of the model at the field scale. Then, two models were developed (i) a global regression model with direct prediction using random forest and (ii) a clustering probability model with Gaussian Mixture Model-random forest. Results demonstrated that (1) reconstructed images produced more accurate prediction results than the individual laboratory Vis-NIR spectroscopy and Sentinel-2A models. (2) At the regional scale, clustering probability model achieved better performance in quantification of SOC (coefficient of determination [R2] = 0.78, and root mean square error [RMSE] = 0.78 %) than global regression model (R2 = 0.65, and RMSE = 1.06 %). (3) For field scale, the reconstructed image-based clustering probability SOC prediction model was shown to be effective when using the Belgium data, providing a R2 of 0.67, and RMSE of 0.15 %. (4) Integrating remote sensing multispectral and proximal Vis-NIR information can lead to a smoother and more continuous SOC distribution within the region, and can better show its spatial heterogeneity within the field. This study highlights the advantages of integrating remote sensing with proximal sensing for improving the accuracy of SOC prediction and mapping, enabling applicability of the model across geographical and spatial scales with appreciable accuracy. [ABSTRACT FROM AUTHOR]

• A fine gridding framework for obtaining finer soil property maps for broad-scale sites. • Including more comprehensive and finer resolution Scorpan covariates to provide finer spatial variability of soil properties. • Comprehensively evaluation of the framework performance in multiple soil physicochemical properties and soil layer depths. • Concordance correlation coefficients and coefficients of determination exceed 0.9 for most soil properties. Fine-scale spatial distribution of soil physicochemical properties is crucial for soil quality management, agriculture planning and geotechnical engineering. Existing soil map databases are usually developed in national scale, potentially leading to issues of coarse resolution and restricted applicability in fine-scaled studies. For broad-scale area, conventional digital soil mapping methods are challenging due to the lack of representative soil profiles and their uneven distribution. This study addresses these challenges by developing a downscaling-based framework to map 11 soil physicochemical properties at a 30-m resolution. The methodology involves constructing regression models using soil properties derived from coarse national soil maps and soil-forming covariates. Predictions are subsequently refined using fine-resolution covariates. To capture fine-scale spatial variability across diverse landscapes, 30-m resolution remote sensing data, relief, climate, and spatial covariates were prepared. The fine-gridding framework employs tree-based machine learning models, particularly random forest (RF), to enhance prediction accuracy. Evaluation of the fine-gridded soil maps demonstrated commendable results, with high consistency in summary statistics, semi-variograms, and evaluation metrics compared to reference maps. Mass preservation of RF-predicted maps exhibit high performances with both concordance correlation coefficients (CCCs) and coefficients of determination (R2s) exceeding 0.9 across most scenarios. This study provides a robust approach for enhancing the spatial resolution of soil maps, facilitating their use in local fine-scale applications. The proposed strategy offers a valuable solution for broad-scale areas that require finer soil maps but lack sufficient qualified soil profiles. [ABSTRACT FROM AUTHOR]

• Higher model performance for granulometry than SOC% using seedbed spectra. • PLSR outperformed other algorithms for SOC% estimation, CR for granulometry. • Low contribution of seedbed spectra to the DSM of SOC% using environmental covariates. Soil organic carbon (SOC) is central to the functioning of terrestrial ecosystems, has climate mitigation potential and provides several benefits for soil health. Understanding the spatial distribution of SOC can help formulate sustainable soil management practices. Digital soil mapping (DSM) uses advanced statistical and geostatistical methods to estimate soil properties across large areas. DSM integrates climate data, topographic features, geology, legacy soil maps, land management and remote sensing data. Bare soil spectra may reflect the presence of particular soil components, making satellite derived spectra suitable predictors of SOC. Bare soil spectra derived from Sentinel-2 were used to estimate SOC concentration (SOC%) and granulometric fractions in the plough layer (0–30 cm) of agricultural parcels in northern Belgium. Thereafter, the estimation performance of SOC% was compared for three DSM models: one with bare soil spectra, one with environmental covariates (topography, granulometry and vegetation), and a combined model with bare soil spectra and environmental covariates. The estimation performance of sand, silt and clay fractions using bare soil spectra from the spring seedbed (R2: 0.53–0.74; RPD: 1.49–2.05; RPIQ: 1.52–2.39) was higher than that of SOC% (R2: 0.16; RPD: 1.08; RPIQ: 1.32). The highest estimation performance of SOC% was obtained for a DSM model including all covariates (R2: 0.28; RPD: 1.18; RPIQ: 1.44), but the contribution of spring seedbed spectra to a model containing environmental covariates was small. The results provide valuable insights for refining soil property estimation using DSM with spectral and environmental covariates. [ABSTRACT FROM AUTHOR]

Mapping the spatiotemporal distribution of soil heavy metals is a prerequisite for soil pollution prevention and control. However, current mapping methods make it difficult to balance robustness and interpretability. In this study, we proposed a spatiotemporal interpretable mapping framework for soil heavy metals and conducted a case study of soil cadmium (Cd) in cropland of the Poyang Lake region in China. The results showed that the average soil Cd concentration increased significantly (p < 0.01) from 0.13 mg/kg in 1983 to 0.19 mg/kg in 2010. Soil Cd in 1983 was mainly controlled by human activities and topography, whereas it was jointly driven by human activities, topography, and climate in 2010. In addition, the framework revealed the spatial distribution of soil Cd and their uncertainties at a 12.5-m resolution in 1983 (R 2 = 0.94) and 2010 (R 2 = 0.97). The top three covariates in 1983 were nighttime-light, PM 10 , and land surface temperature (day), whereas those in 2010 were mean annual minimum temperature, gross domestic product, and industrial enterprises, with Shapley additive explanation (SHAP) values of 0.44, 0.42, and 0.27, and 0.31, 0.11, and 0.09 mg/kg, respectively. By integrating machine learning, variable and model selection, and the SHAP module, the proposed framework achieves a balance between the robustness and interpretability in mapping the spatiotemporal distribution of soil Cd, going beyond the limitations of the classical statistical "smoothing effect" and the artificial intelligence "black box." Our study emphasizes the importance of considering time, space, models, variables, and mechanisms in digital soil mapping, which will enable more accurate and scientific predictions of soil contaminants across time and space. [Display omitted] • Developed a spatiotemporal interpretable mapping framework of soil heavy metal at a 12.5-m resolution. • Quantified the driving mechanism and spatiotemporal clustering pattern of soil Cd. • Integrated advanced digital soil mapping technologies and the explainable SHAP module. • Achieved a balance between robustness and interpretability. [ABSTRACT FROM AUTHOR]

• The selection of satellite sensor and radar system parameters greatly affected the model output. • The model was improved when more polarizations, orbital directions and frequencies were involved. • Models built using multiband radar datasets achieved a comparable accuracy to models based on optical data. • The model built by fusing SAR-optical data achieved better results than the model without SAR data. • The development of this continental-scale DSM work largely benefits from GEE. The advent of cloud computing platforms (e.g., Google Earth Engine (GEE)) and the massive amounts of optical and radar Earth Observation (EO) data hosted by these platforms present new opportunities for mapping soil pH at large scales. However, existing studies generally lack consensus on the effects of satellite and radar sensor parameters on GEE-based soil pH prediction models. In this study, we assessed the suitability of long-term radar (C-band Sentinel-1 and L-band PALSAR-1/2) and optical (Sentinel-2) EO data on GEE for the digital mapping of soil pH on a continental (Europe) scale and determined the most appropriate radar sensor parameters. Thirteen scenarios with different data configurations were simulated and combined with the 2018 LUCAS soil database and two machine learners (boosted regression trees and extreme gradient boosting) to develop soil prediction models. Results showed that the selection of modeling techniques, satellite sensors and radar system parameters largely affected the model output. Models involving a single polarization mode of PALSAR-1/2 data performed the worst (RPD = 1.24). Models based on Sentinel-1 data performed better than those built using PALSAR-1/2 data. The model performance was improved when a model involved more polarization bands, orbital directions, and band frequencies. The multiband model built using the two radar datasets achieved a comparable accuracy to the model based on optical data. Moreover, the model that fused radar-optical data achieved better results, with RPD values of 1.56 and 1.46 for the models with and without radar data, respectively; its performance was comparable to that of models built with commonly used variables (topography and climate). The analysis of importance indicated that long-term optical and radar EO data on GEE were important in our model. The modelling of soil pH at the continental scale largely benefits from GEE. The predicted maps exhibited strong spatial heterogeneity among different biogeographic regions, with similar spatial patterns under different modelling scenarios. [ABSTRACT FROM AUTHOR]

• This study estimated that 17.50 Pg C was stored in Chinese grasslands to a depth of 20 cm, and the average SOC density was 4.69 kg C m−2. • A path network showing interacting effects of climate, topography, soil characteristics, and vegetation on SOC. • The direct effect of topography on SOC was relatively weak, but it did have a significant positive impact on vegetation at a broad geographical scale. Estimations of the patterns and controls of soil organic carbon (SOC) could provide instructive insights into the potential impact of future global change on soil carbon (C). In this work, we combined GeoDetector and random forest (RF) to estimate SOC stocks in Chinese grassland ecosystems with uncertainty assessments, and identified a network of cross-correlated environmental covariates for determining SOC based on a dataset collected from 813 sampling sites (0–20 cm) collected from 2000 to 2014. We predicted that 17.50 Pg C was stored in Chinese grasslands to a depth of 20 cm and that the average SOC density was 4.69 kg C/m−2(−|−). The effectiveness of using RF to predict SOC was demonstrated by an accuracy assessment based on 10-fold cross-validation, with a ratio of performance to deviation (RPD) of 2.89. The SOC stocks in southern China were lower than those in northern China. A high SOC density was found in northeastern China and on the Qinghai–Tibet Plateau. Soil properties had the strongest direct effect on SOC. Climate was significantly negatively associated with SOC and indirectly affected SOC via its effect on soil properties. Topography had a significant direct impact on vegetation, but its direct effect on SOC was relatively weak. This study emphasizes the patterns and heterogeneity of SOC stocks as well as the relative significance of climate, vegetation, soil characteristics, topography, and their complex interrelationships in controlling SOC. These results may provide a theoretical foundation for developing sustainable management systems and calibrating C process models. [ABSTRACT FROM AUTHOR]

• All the models were able to produce SOC maps with mean values consistent with the actual mean SOC value. • Adding residual kriging to ML models yields marginal improved prediction accuracy of SOC. • VDCN yield information about age and evolution stage of soils. • Agricultural activities play an important role in the interpretation of the results. • Incorporating VDCN and CNBL is vital in future DSM for SOC in lowlands, reflecting their significant impact. Soil organic carbon (SOC) plays a crucial role in the global carbon cycle and in maintaining soil functions in the context of land use and climate change. Understanding the spatial distribution of SOC is essential for the management of agricultural land to optimize soil health and carbon storage. In this study, we investigated the spatial distribution of SOC in an agricultural lowland area of the Lombardy region, Italy, using machine learning (ML) techniques combined with residual kriging. ML models, including the artificial neural network (ANN), extreme learning machine (ELM), and random forest (RF), were trained on 120 SOC observations and eight environmental variables to predict SOC values across the study area. The performance of this ML approach was assessed using a ten-fold nested cross-validation process. The ELM and RF models showed better predictive performances based on the concordance correlation coefficient and root mean square error (RMSE), with RF slightly outperforming ELM based on the RMSE. The residuals of each iteration from the ML models were interpolated by ordinary kriging (OK) and added to the ML-based trend model in a hybrid regression-kriging approach. This approach which accounted for the spatial autocorrelation of the prediction residuals, resulting in a marginally improved prediction accuracy in the ML models. In addition, we found that vertical distance to the channel network and channel network base level are important predictor variables that should be considered in future digital soil models for SOC in lowland areas, given their importance in this study. Furthermore, this study highlights that predicted SOC values were low, particularly in Luvisols, which can be explained by the long history of agricultural land use depleting SOC due to agricultural management and loss of organic plant residues. The prediction maps depicted spatial variation and patterns of SOC in the study area. Our findings may help to refine soil management practices and contribute to improving soil health and carbon sequestration in agricultural lowland areas. [ABSTRACT FROM AUTHOR]

[Display omitted] • A targeted field sampling plan was designed using a digital soil mapping (DSM) method. • Baseline soil organic carbon (SOC) stocks were modeled with legacy and local samples. • The optimal number of local samples needed for model improvement varied by ranch size. • DSM-based SOC concentration models outperformed bulk density and SOC stock models. • Selecting calibration dataset individually at each point may improve model accuracy. Accurately quantifying high-resolution field-scale soil organic carbon (SOC) stocks is challenging yet crucial for improving site-specific land management and carbon accounting. This challenge is even greater when the study units are large heterogenous ranches. This study utilizes a digital soil mapping (DSM) approach and a U.S. legacy soil dataset, combined with soil, climate, biotic, and topographic covariate datasets, to design a targeted soil sampling plan for acquiring local samples. The resulting local samples were then used in combination with the legacy data to build optimal ranch-scale SOC stock models. We provide an example of this approach using ranch units in the western U.S. as a case study. In our approach we first applied a clustering analysis to generate spatial clusters. This was followed by adopting a conditioned Latin hypercube sampling scheme within each cluster, to generate sets of strategically selected local sampling points. The local samples required for improved estimates of SOC stocks were determined to have a sample size of 15 and 40 soil cores, within the respective 13 and 36 km2 parcels. While our modeling results for SOC concentrations at a relatively homogeneous site in eastern Montana showed a significant two-fold improvement in model fit when using individually selected calibration datasets for each point, as opposed to selecting the calibration dataset as a whole at the ranch level, the disparity between the pixel- and ranch-based models was inconsequential for the other two ranch sites in Colorado that were more spatially diverse in terms of land management and vegetation cover. Compared to SOC concentration models (R2 between 0.3 and 0.7), the performance of models for bulk density (BD) (R2 < 0.4) and SOC stocks (R2 < 0.2) were relatively poor. Strategies including utilizing a subset of covariates, incorporating broader-scale national calibration datasets, and modeling by depths did not further improve BD and SOC stock models. Future work should explore whether the addition of temporally dynamic environmental covariates can improve SOC stock estimates, and whether the DSM-supported targeted field sampling strategy and high-resolution mapping approach can be successfully applied elsewhere. [ABSTRACT FROM AUTHOR]

• Digital soil maps created for biological processes as related to soil N dynamics. • Multi-year crop inventory covariates increase predictive accuracy by 27%. • Concordance of 0.47 achieved for estimates of growing season N mineralization. • Spatial predictions of N parameters inform N fertilizer recommendations. For the international digital soil mapping (DSM) community, adequate spatial estimates of nitrogen (N) mineralization have yet to be generated. This is due, in part, to an inability to capture critical N controls at the regional and provincial scales. While the influence of climate, vegetation, and relief are accessible predictors in DSM, the effect of soil management is known for its important influence on N dynamics, but has hitherto been elusive for soil mappers. For the purpose of producing N maps to inform N fertilizer management, the intention of this study was to determine the importance of novel crop frequency layers, as a proxy for soil management, through the development of provincial scale DSMs of total nitrogen (TN), biological nitrogen availability (BNA) and the estimate of N mineralization over a growing season (GSN) as calculated from TN and BNA results. Crop frequency covariates were developed that estimated the frequency a particular crop type was planted over a 10-year period, thus capturing cropping system and tillage intensity. TN results were 27% higher using crop frequency layers and the support vector machine learner, with a Lin's concordance correlation coefficient (concordance) of 0.45. BNA predictions increased by 24% using crop frequency layers and the stochastic gradient boosting learner for a final concordance of 0.45. GSN showed the least improvement using the crop frequency layers (6%) but resulted in the highest concordance (0.47) using crop frequency layers with the stochastic gradient boosting learner. The stable N pool, represented by TN, showed climate covariates with the highest importance; whereas, the labile pool, based on BNA measures, was best predicted and controlled by organism covariates. The successful inclusion of crop frequency layers into N maps indicated that the number of times in forages and potatoes over a 10-year period was of the greatest importance. As tillage intensity is most pronounced in potatoes, and as forages contribute to increased biomass and building soil organic matter levels, the results showed that increasing the number of years in forages had a positive correlation with GSN and the stable and labile N pools. [ABSTRACT FROM AUTHOR]

[Display omitted] • Introducing a digital soil mapping method that uses a physically based model. • Water movement functions as a proxy for pedogenesis through covariate relationships. • Hydrologic models captured temporal trends of water tables and soil moisture. • Spatial-temporal patterns of water movement were related with fragic properties. • Hydrologic models add a time and depth component to bidimensional soil maps. Digital soil mapping (DSM) relies on machine-learning and geostatistics to represent soil property observations across space. DSM techniques are powerful but often empirical, being limited to the quality and density of point samples. Water dynamics are closely related to soil variability, and the physics that govern water movement are well known. Hydrological properties can hence be simulated by physical models through space and time, unveiling key characteristics about soils. We propose the use of hydrologic models to map soils across the surface (2D), depth (1D), and time (1D)–which provides a 4D approach to digital soil mapping (4DSM). The Distributed Hydrology Soil Vegetation Model (DHSVM) was applied to a watershed currently under pasture. Moisture sensors and wells were installed at different depths in the watershed on summit, sideslope and toeslope positions to validate the model. DHSVM simulations of soil moisture distribution and depth to saturation were performed during the hydrological year (October 2008-September 2009). Clusters of similar pixels based on soil moisture values were determined using Dynamic Time Warping (DTW) to align temporal data and K-means. Clustering was performed both seasonally and for the entire year. Temporal patterns simulated by DHSVM matched measurements given by moisture sensors and wells. Seasonal clusters differed from the annual cluster. Distinct clusters were observed for each season and with depth, showing that spatiotemporal soil variability is lost when statically assessing soils. Spatiotemporal clusters corroborated field observations of fragipan occurrence not explicitly spatially mapped by Soil Survey Geographic Database (SSURGO). If a connection can be made between water and soils, static and dynamic soil variability can be predicted using physically based hydrologic models. Hydrologic models can benefit soil mapping by enabling reliable 4D simulation of water dynamics, which are fundamental to soil variability and soil classification and directly relate to biological, physical and chemical soil processes not captured by typical soil sampling protocols. [ABSTRACT FROM AUTHOR]

• A method for SOM content mapping based on a multitemporal soil line was developed. • Mapping detail was 30 m with an accuracy of R2 = 0.8. • The neural network automated the selection of remote sensing data and bare soil. • The proposed approach can be applied to a wide range of soils and large areas. A new method for constructing detailed maps of the soil organic matter (SOM) distribution in the top layer of arable land has been developed and proposed. The method is based on the theory of spectral neighborhood of the soil line (SNSL) and the technology of constructing a multitemporal soil line (MSL). The method is based on the processing of big remote sensing data (BRSD) from 1984 to 2023. Filtering of BRSD and detection of bare soil surface (BSS) is carried out on the basis of neural networks. The method was implemented for BSS with an area about of 79,000 ha with a spatial resolution of 30 m in the Mtsensk district (Oryol Oblast, Russia). Verification was provided by four independent field surveys (which were carried out using three various methods) in different years (2022–2023). The regression is described by a polynomial of degree 2. The coefficient of determination (R2) of the regression was 0.8. The proposed method can be widely used for mapping of SOM in the areas of transition from leached chernozems (Luvic Chernozems) to sod-podzolic (Albic Retisol) and gray forest (Luvic Phaeozems) soils or in similar nature conditions. [ABSTRACT FROM AUTHOR]

• Novel approach integrating water erosion and tillage erosion in a Digital Soil Mapping based approach for SOC mapping. • As compared to reference sites "plateau", SOC stocks are half on eroded sites and double on depositional sites. • Tillage erosion affects SOC stocks in the top soil whereas water erosion affects SOC stocks on the entire profile. Although agricultural intensification has generally increased crop yields, it also resulted in a range of environmental issues. These include increased erosion rates and declined soil organic carbon (SOC) stocks. In order to improve our understanding on how erosion impacts the overall SOC storage capacity of croplands, this study analyses the 3D distribution of SOC as a function of water and tillage erosion in a conventionally ploughed field in the Belgian silt loam region. We present a novel methodological framework to integrate the output of an advanced erosion model as a co-variate within a Digital Soil Mapping (DSM) approach with the objective to create detailed SOC maps. More precisely, we combined (i) the Water and Tillage Erosion Model and Sediment Delivery Model (WaTEM/SEDEM), simulating spatial patterns of soil erosion and sediment deposition due to water and tillage erosions, with (ii) a SOC sampling campaign, resulting in a SOC spatial distribution model that considers both types of erosion. The results show that, as compared to plateaus, SOC stocks are nearly half as large along eroding slopes (i.e. convex slope positions affected by tillage erosion and steep slopes affected by water erosion). Yet, they are up to twice as large in areas characterized by sediment deposition (i.e. concave positions due to tillage erosion and foot slope positions due to water erosion). Our results further show that tillage erosion has a significant influence on the SOC stocks in the top 0.7 m and in particular on the top 0.4 m. The influence of water erosion is less strong but mostly significant along the entire depth profile. Overall, this work demonstrates the relevance of considering different erosion processes when aiming to predict spatial patterns of SOC. Considering the top 1 m and a WaTEM/SEDEM application at a resolution of 10 m our 3D SOC modelling approach obtained a coefficient of determination (R2) of 0.62, a relative root mean square error (RRMSE) of 30.5 % and a relative mean absolute error (RMAE) of 26.0 %. While future work may likely lead to further improvements e.g. a more detailed SOC sampling network along the foot slopes and thalwegs in order to obtain a more spatially detailed prediction of the area characterized by depositions due to water erosion, we demonstrate the great potential of existing erosion and deposition models in doing so. [ABSTRACT FROM AUTHOR]

Imbalanced datasets are one of the main challenges in digital soil mapping. For these datasets, machine learning techniques commonly overestimate the majority classes and underestimate the minority ones. In general, this generates maps with poor precision and unrealistic results. Considering these maps for land use decision-making can have dire consequences. This is the case of acid sulfate (AS) soils, a type of harmful soil that can generate serious environmental damage when drained in agricultural or forestry activities. Therefore, it is necessary to create high-precision maps to avoid environmental damage. Although most soil class datasets in nature are imbalanced, this problem has hardly been studied. One of the main objectives of this work is the evaluation of different techniques to address the problem of imbalanced datasets. The methods considered to balance the dataset are an undersampling technique, the addition of more samples, and the combination of both. For increasing the number of samples from the minority class, we develop a new technique by creating artificial samples from the quaternary geological map. The method used for the modeling is Random Forest, one of the best methods for the classification of AS soils. Balancing the dataset improves the performance of the model in all the studied cases, where the values of the metrics for both classes are above 80%. The consideration of artificial non-AS soil samples improves the prediction of the model for the AS soils. Furthermore, we create AS soil probability maps for the four balanced datasets and the imbalanced dataset. The modeled AS soil probability maps created from the balanced datasets have high precision. A detailed comparison between the maps is made. The predictions of some of these maps match between 75%–80% of the study area. In addition, the extent of the AS soils obtained in all the cases is compared with the extent of the AS soils in the conventionally produced occurrence map. The good results of this study confirm the importance of balancing the dataset to improve the prediction and classification of AS soils. • Different techniques were evaluated for addressing the problem of imbalanced datasets. • Balancing datasets improved prediction accuracy of acid sulfate soil mapping. • Highly accurate acid sulfate soil mapping from balanced datasets and random forest. • Detailed comparison of modeled acid sulfate soil probability maps in western Finland. [ABSTRACT FROM AUTHOR]

• A location correction method is proposed for widespread legacy soil profiles. • Consistency between covariates and soil profile description is enhanced. • Location error is reduced, and soil mapping results are improved significantly. • Elevation was the primary controlling factor for correcting location error. Digital soil mapping relies on statistical relationships between soil profile observations and environmental covariates at the sample locations. However, inherent limitations of legacy soil profiles, such as inaccurate georeferencing, could frequently introduce location errors into these soil profiles that affect the quality of digital soil mapping. To address this challenge, this study focuses on reducing the location error of legacy soil profiles and evaluating the resulting impact on digital soil mapping. We improved the agreement between the detailed descriptive information of legacy soil profiles and the relatively accurate environmental covariates (such as elevation, slope, and land use) to reduce the location errors of the legacy soil profiles. Quantile regression forest models were constructed to predict soil properties and their uncertainty using legacy soil profiles before and after location error correction. Our results demonstrate that for the majority of soil variables, correcting positional errors in legacy soil profiles to some extent enhances the accuracy of the digital soil mapping. The largest improvement was found for soil organic carbon at 0–5 cm depth interval, with 21 % increase of MEC. The impact of reduced location error is particularly noteworthy in regions characterized by complex terrain. In addition, the details of the predicted maps of legacy soil profiles were improved after correcting for positional errors, which better represent the spatial variation of soil properties across China. Besides, we also found that elevation was the primary controlling factor for correcting location error of legacy soil profiles. This research presents a step towards producing high-resolution and high-quality spatial soil datasets, which can provide essential support for soil management and ensure future soil security. [ABSTRACT FROM AUTHOR]

Determining the spatial distribution of soil groups accurately is crucial for managing soil resources. However, limitations persist in the mapping of soil groups using multi-textural features derived from remote sensing images. Identification of the optimal window size for multi-textural feature extraction and the most effective classification method for soil group recognition using remote sensing multi-textural features remains unresolved. In this study, we investigated soil groups in a representative area of the Jiaodong Peninsula. We extracted the mean and entropy texture parameters for various window sizes (3 × 3 to 25 × 25 in odd increments) from Landsat 8 images to determine the optimal sizes for multi-textural feature extraction. The efficacy of identifying soil groups via textural features was analyzed using maximum likelihood classification (MLC), support vector machine (SVM), artificial neural network (ANN), and random forest (RF) methods to ascertain the most suitable classification approach. The results indicate that the optimal window sizes were 19 × 19 for the mean parameter and 23 × 23 for the entropy parameter. The SVM method outperformed the MLC, ANN, and RF methods in terms of the classification accuracy. Notably, the SVM classification method reached a peak accuracy of 71.61% when combining multi-textural features with the optimal windows. This demonstrates the feasibility of different soil groups using multi-textural information from remote sensing images. These findings have notable implications in guiding digital soil mapping using multi-textural features. • The optimal window for identifying soil groups was extracted. • Multi-textural features improved the accuracy of soil recognition using remote sensing. • The optimal classification method for soil group recognition based on the texture features was determined. • The SVM method achieved a maximum accuracy of 71.61% after fusing multi-texture features. [ABSTRACT FROM AUTHOR]

Accurately mapping the spatial distribution of soil organic matter (SOM) content is critical for informed land management decisions and comprehensive climate change analyses. Remote sensing-based SOM mapping models during periods of bare soil exposure have demonstrated efficacy in various regional studies. However, integrating bare soil imagery with growing season imagery for SOM content mapping remains a complex process. We conducted a study in Youyi Farm, a representative area of black soil in Northeast China. We collected 574 soil samples (0–20 cm) with SOM content through field sampling and laboratory analysis. Additionally, cloud-free Sentinel-2 images were obtained from the Google Earth Engine (GEE) platform for both the bare soil period (April-June, October) and crop growth period (July-September) from 2019 to 2021. To assess the influence of crop growth information on SOM mapping, we incorporated remote sensing imagery during the crop growth period, considering different crop type zones (maize (Zea mays L.), soybean (Glycine max L.), and rice (Oryza sativa L.)). We conducted overall and zonal regressions using the random forest (RF) model to validate the prediction results through cross-validation. Our findings indicate that: (1) adding crop growth period images to the bare soil period images in different years can improve the accuracy of SOM mapping. For example, in the overall regression model of 2020, the highest accuracy was achieved by using the combination of May-July images, with an R2 value of 0.70 and an RMSE value of 0.71 %; (2) zonal regression by differentiating crop types can effectively improve the SOM mapping accuracy. In 2019, using zonal regression, the R2 of SOM mapping accuracy was improved by 0.02 and the RMSE was reduced by 0.03 % compared with the overall regression; (3) precipitation is an important factor affecting the accuracy of SOM prediction, and the lower the precipitation, the higher the accuracy of SOM prediction. In summary, the results of this study show that in the SOM remote sensing mapping of the black soil area, the growing period remote sensing information of different crop types should be comprehensively considered and combined with the image data of the years of lower precipitation, the accuracy of the SOM mapping can be effectively improved, which provides a new technological path and an application basis for the enhancement of the accuracy in remote sensing mapping with soil attributes. [Display omitted] • An integrated bare soil and crop growth information method was developed for SOM mapping. • Zonal regression on various crop types further improves SOM prediction accuracy. • Single-stage crop images reflect SOM variation better than multi-stage composites. • Precipitation is a key factor influencing the accuracy of SOM mapping. [ABSTRACT FROM AUTHOR]

Sustainable agriculture and eco-environment conservation face constant threats from soil erosion in mountainous regions. This study aimed to predict soil loss in the mountainous watershed by applying the Revise Universal Equation Soil Loss (RUSLE) and Sediment Delivery Ratio (SDR) models and to evaluate the efficacy of the RUSLE model through the observed data in a hydrometric station. Comparing a polygonal, and a digital soil map (prepared by the Kriging method) for deriving the K factor to estimate soil loss was another objective of the recent research. In this study, two scenarios were examined for the calculation of spatial variation of the K -factor; scenario I comprised soil data derived from a legacy soil map with eight soil units and the data from their representative soil profiles (i.e., traditional soil survey), and scenario II comprised K -factor derived from 100 studied sites using kriging technique (i.e., digital soil survey). The results of RUSLE modeling showed that there was no significant difference between scenario I (7.97 t/(ha. yr)) and scenario II (7.93 t/(ha. yr)) for predicting soil loss with the RUSLE model. This finding confirms that the RUSLE model with limited and legacy data can provide a reliable prediction of soil loss in the given watershed. Moreover, long and short-term periods were used to estimate soil loss, while in the long-term period, estimated soil loss had higher accordance with actual soil loss. Sediment delivery ratio was calculated using models of Vanoni, Boyce, USDA-SCS, and Slope-based models. The results indicated that the Slope-based model had the highest fitness (R2=0.946, ME=0.27, and RMSE=0.275 for scenario I and R2=0.9443, ME=0.26 and RMSE=0.40 for scenario II), with observed sedimentation rate at the hydrometric station at the outlet of the watershed. Overall, predicted soil loss severity map using the traditional soil map by RUSLE model could provide trustworthy information for decision makers and governors at the given and similar watersheds in semiarid regions. • The efficacy of conventional and digital maps were assessed for deriving K factor. • No significant difference was between two scenarios for prediction soil loss using RSULE model. • Legacy data can provide a reliable prediction of soil loss in the given watershed. • The results indicated the Slope-based model had the highest performance for calculation SDR. [ABSTRACT FROM AUTHOR]

The historical conversion of forests to rainfed agricultural lands in the semi-arid forest ecosystems is one of the primary sources of human-induced, greenhouse gas emission and causes of soil organic carbon (SOC) loss. This study aims to predict SOC contents as an extremely crucial factor in soil formation and fertility in the topsoil of traditionally managed semiarid oak forests. The research investigates the complex relationship between SOC and multiple soil and environmental factors using machine learning (ML) techniques. A total of 175 soil samples were taken from the topsoil (0–30 cm). In total, 59 soil-environmental covariates were acquired from various sources, including soil property maps, derivatives of digital elevation model, remote sensing data, climatic model data, and geological data—all of which characterized the environmental variables. To generate the predictive models, Random Forest (RF) and k -nearest neighbors (k-NN) models were trained; additionally, those models were combined into an ensemble model (k-NN-RF) for predicting SOC. In addition, we utilized a bootstrapping method to quantify uncertainties for SOC predictions. To understand the soil-environmental relationships, post-hoc model assessment was conducted using variable importance analysis, demonstrating that the main predictors included canopy cover percentage, terrain surface texture, calcium carbonate equivalent, and midslope position. External validation of ML performance revealed that RF had the highest accuracy, achieving a Lin's concordance correlation coefficient (CCC) of 0.74, followed by k-NN-RF with a CCC of 0.63. The uncertainty analysis demonstrated consistently low uncertainty across the entire study area. However, areas with complex canopy cover exhibited a higher potential for uncertainty compared to areas with more uniform canopy cover. Results indicated that the maximum SOC content was found in the forest and pasture lands. These findings underscore the significance of canopy cover in predicting SOC, which serves as a crucial indicator for forest ecosystem management and land use change. Therefore, this research provides valuable insights for developing management strategies that consider the interplay between canopy cover and soil organic carbon in semi-arid ecosystems. • Forest canopy cover variations, as land use legacy were the most important predictor of SOC. • Random Forest model was effective in predicting SOC. • Partial dependence plot (PDP) analysis increases the interoperability of the Random Forest model. • Uncertainty analysis enhances our understanding on SOC variation in different canopy covers. [ABSTRACT FROM AUTHOR]

To identify microbial resources for dehalogenation, develop effective remediation strategies, and reveal their significance in "One Health", it is crucial to understand the occurrence, distribution, and drivers of soil dehalogenation functional traits and taxonomy groups at a broad scale, which is currently not well understood. To address the gaps, we characterized the biogeography of both dehalogenation traits and taxa assigned to six dehalogenation pathways, by metagenomic profiling global 4821 soils from eight habitats, based on a manually curated dehalogenase database (DhgaseDB). We found dehalogenation genes and microbes assigned to different pathways are everywhere, but varied consistently across habitats. The similarity of dehalogention traits and taxa composition declines with geographic distance, and that patterns are strongly correlated with geo-environmental factors. We identified anthropogenic organohalide pesticide inputs as the most influential factor on dehalogenation gene abundance, while soil properties, particularly pH, exert a larger impact on dehalogenation taxa diversity. Ultimately, we generated predictive maps of soil dehalogenation gene abundance and taxa diversity for the first time, highlighting the microbial dehalogenation hotpots in East Asia, Australia, Southern Africa, and coastal regions. Collectively, our study highlights the significant role of various microbial dehalogenation processes in organohalide biotransformation and environmental microecology, providing the necessary methodological basis for a deeper comprehension of the underlying mechanisms, thereby contributing to the advancement of tailored strategies for organohalide remediation. • DhgaseDB was established for fast and accurate analysis of dehalogenation genes and taxa. • Dehalogenation genes and microbes are prevalent in global soils, but varied consistently across habitats. • Pesticide inputs are key to gene abundance, while soil pH significantly affects microbial diversity. • The first predictive maps of soil dehalogenation gene abundance and taxa diversity were generated. [ABSTRACT FROM AUTHOR]

Soil organic matter (SOM) is crucial for karst ecosystems, affecting cropland health, climate change mitigation, and rocky desertification control. However, there are limited research on cropland SOM prediction in karst areas with complex topography and diverse microclimates. Here, we compared the performance of four machine learning algorithms—random forest (RF), support vector regression (SVR), multilayer perceptron regression (MLP), and gradient boosting regression trees (GBRT)—for predicting cropland SOM in a typical karst landform area in 2019. Our results indicated that the GBRT model achieved the highest prediction accuracy with an R² of 0.69, MAE of 2.19 g/kg, RMSE of 3.37 g/kg, and LCCC of 0.82. Using the GBRT model and spatial data on climate, topography, and remote sensing, we predicted SOM for each 30 m × 30 m grid cell. The analysis revealed higher SOM content in the northeastern and southwestern regions and lower content in the central area, ranging from 13.95 to 47.81 g/kg, with an average of 27.16 g/kg. Lime soil had the highest SOM content, while purple soil had the lowest. Paddy fields showed significantly higher SOM than dry land. Over the past 40 years, SOM content has slightly increased, while its spatial distribution has remained stable. [Display omitted] • GBRT algorithm could effectively predict SOM content in the typical karst cropland. • Climate, topography and remote sensing vegetation indices were key variables for SOM prediction in karst cropland. • In karst cropland, lime soil exhibited the highest SOM content, while purple soil had the lowest. • Over the past 40 years, the mean SOM content has shown a slight upward trend since the national soil survey in 1980s. [ABSTRACT FROM AUTHOR]

This study investigates the complexity of spatial soil modelling, particularly focusing on the challenge of variable vertical support in traditional soil data collection. Traditional soil sampling, described in terms of horizons, often fails to accurately pinpoint the specific depths for specific soil properties. This gap is significant, as depth-specific data is crucial for a thorough understanding of soil formation processes and for assessing potential environmental impacts. In digital soil mapping (DSM), the prevalent reliance on standardised depth intervals and mass-preserving spline functions for data resampling results in a modelling approach that tends to disregard depth-related details, thereby introducing potential uncertainties. To address these limitations, this work explores how effectively Gaussian process regression (GPR) can model soil in 3D. This technique comprises two key components: a mean function and a semivariogram-like kernel. Unlike conventional methods that make a single prediction, GPR provides detailed probability distributions. This capability allows for the quantification of prediction uncertainty at various points, offering insights for decision-making and risk assessment purposes. Moreover, GPR has the capability to make volume or block estimates and assess associated uncertainties. Enabling volume-based predictions enriches the range of strategies available for land management. In this research, we employ GPR as a novel 3D soil modelling technique and compare its performance with traditional spline-based methods. The comparison is conducted through a case study on a farm in northern New South Wales, Australia, focusing on the 3D mapping of soil pH and electrical conductivity (EC). Our results demonstrate the ability of GPR to estimate soil properties across various volumes, utilising data from multiple vertical supports, thereby offering a more versatile approach for soil modelling in diverse spatial contexts. • Enable 3D digital soil mapping. • Account for multiple vertical support. • Produce both point and volume/block predictions. • Quantify the probability of exceedance a certain value. • Introduce an open-source GPR package. [ABSTRACT FROM AUTHOR]

• Maximum value selection in certain areas is effective for preprocessing. • SOC map did not work as the covariate in whole area but low-site-density areas. • Soil thickness linked to the tephra distribution and the microtopography. Soil thickness is a crucial parameter for local, regional and globalscale assessments, including those related to carbon cycles, water retention capacity, and landslide risk in the context of biogeochemical, soil physical, and slope stability models. This study had two primary objectives: 1) to predict maps depicting the probability of the A-horizon thickness exceeding 15 cm (AT15) and the A and B-horizon thickness exceeding 75 cm (soil thickness, ST75), across the expansive mountainous regions of Japan. 2) To evaluate the enhancement of model performance through the preselection of the maximum value and implementation of enhanced modeling using gridded modern soil survey data to utilize geographically clustered data. Initially, the study modeled the SOC, bulk density, and coarse fraction proportion using data from a modern soil survey of 0–30 cm. Subsequently, we employed a random forest model for AT15 and ST75 using a dataset of geographically clustered legacy soil profiles as training data, and topographic, vegetation, climatic, and volcanic attributes, and the predicted soil organic carbon and related properties as explanatory variables. We reduced the number of legacy soil profiles in the densely sampled areas by selecting the maximum soil thickness before modeling. The study assessed the final model performance at the national scale, with the accuracy and F1 scores reaching 0.691 and 0.754 for AT15 and 0.600 and 0.511 for ST75, respectively. Maximum value selection proved effective for modeling ST75 but not AT15. Although an enhanced model incorporating data from legacy and modern surveys did not improve accuracy at the national level, it did improve recall in sparsely sampled areas, for ST75. The resulting spatial pattern of the predicted ST75 effectively captured variations in soil thickness associated with topography, geology, and tephra deposition. However, ST75, representing pedological soil thickness, tended to be strongly underestimated compared to geotechnical soil thickness, including the C-horizon, especially in areas affected by tephra deposition. This study represents a significant contribution as it delivers Japan's first national-scale soil thickness map and an approach for utilizing clustered legacy data. It is expected to find practical applications in estimating water storage capacity for forest management and other relevant purposes. [ABSTRACT FROM AUTHOR]

[Display omitted] • A 250 m resolution black soil distribution map of China was developed. • Random forest classifier can accurately identify black soil at the national scale. • Daily mean land surface temperature and vegetation greatly affect black soil distribution. • A large area of unused black soil with reclamation potential in China. Black soil (BS) is an important soil resource. However, there is a lack of accurate information on BS spatial distribution, which affects the sustainable use and protection of BS. In this study, we linked BS distribution and machine learning. Random forest (RF) classifier and recent soil survey data were used to produce a high–precision BS distribution map in China. We analyzed environmental control factors of BS distribution. The results showed that the RF classifier method can well estimate the spatial distribution of BS at the national scale with an overall accuracy of 0.63 ∼ 0.91. The hot spots of BS distribution were mainly located in the Songnen Plain, Sanjiang Plain, the eastern side of the Greater Khingan Mountains and the western side of Changbai Mountains in Northeast China, the northeastern foot of Qilian Mountains in Qinghai–Tibet Plateau and the northern foot of Tianshan Mountains in Northwest China. Nearly 45 % BS area was covered by cropland. The daily mean land surface temperature and vegetation coverage had important controling effects on the BS distribution. Excluding the alpine mountains, BS area with potential for conversion to cropland can increase by 18.12 % (3.27 × 104 km2). The mapped BS distribution of China can serve as a benchmark for BS resource monitoring and play an important role in its assessment and protection. [ABSTRACT FROM AUTHOR]

• Soil elemental composition is key to infer other soil attributes. • DSM and remote sensing can take punctual XRF information to maps. • We mapped the main chemical elements using punctual XRF information and DSM. • XRF allied with DSM can be a new and low-cost way to large areas soil mapping. • Spatial XRF information can be used to access soil variations in the landscape. Spatial soil information has significantly contributed to public policy and planning. The use of X-ray fluorescence spectrometry (XRF) in soil science, primarily in laboratory settings or for isolated point analyses, has been well documented. This study broadens the application of XRF data by integrating it with remote sensing data for a comprehensive understanding of soil patterns and processes. Focusing on a 2574 km2 area in Brazil, we utilized a lab-based XRF instrument to measure the total concentrations of key chemical elements in the soil, such as Al, Si, Ti, and Fe. These geochemical data were then spatialized using the digital soil mapping (DSM) framework. We employed a synthetic bare soil image and elevation data as covariates in the analysis. The spatial accuracy assessment yielded R2 values ranging from 0.65 to 0.81 for Fe, 0.62 to 0.78 for Ti, 0.52 to 0.63 for Si, and 0.48 to 0.63 for Al. The digital soil maps of geochemical elements aligned well with existing pedological, and geological maps as evaluated using multiple correspondence analysis. The integration of spatialized XRF data with DSM and remote sensing techniques shows significant promise in assessing soil variations across landscapes based on their chemical composition. [ABSTRACT FROM AUTHOR]

[Display omitted] • Estimating sampling fieldwork time by mapping operational constraints is possible. • MOOS reduces fieldwork time compared to other methods. • It allows a practitioner to choose among a variety of sampling designs. Sampling for Digital Soil Mapping is an expensive and time-constrained operation. It is crucial to consider these limitations in practical situations, particularly when dealing with large-scale areas that are remote and poorly accessible. To address this issue, several authors have proposed methods based on cost constraints optimization to reduce the travel time between sampling sites. These methods focused on optimizing the access cost associated to each sample site, but have not explicitly addressed field work time required for the whole sampling campaign. Hence, an estimation of fieldwork time is of great interest to assists soil surveyors in efficiently planning and executing optimized field surveys. The goal of this study is to propose, implement and test a new method named Multi-Objective Operational Sampling (MOOS), to minimize sampling route time, while ensuring that sample representativeness of the area is maintained. It offers multiple optimal sampling designs, allowing practitioners to select the most suitable option based on their desired sample quality and available time resources. The proposed sampling method is derived from conditioned Latin Hypercube sampling (cLHS) that optimizes both total field work time (travel time and on-site sampling time) and sample representativeness of the study area (cLHS objective function). The use of a multi-objective optimization algorithm (NSGA II) provides a variety of optimal sampling designs with varying sample size. The sampling route time computation is based on an access cost map derived from remote sensing images and expert annotation data. A least-cost algorithm is used to create a time matrix allowing precise evaluation of the time required to connect each pair of sites and thus determine an optimal path. The proposed method has been implemented and tested on sampling for pH H 2 O mapping within a 651 points kilometric grid in the northern part of Saudi Arabia, where soil analyses were conducted over a 1,069 km2 area. MOOS method was compared to two other common approaches: classical cLHS and cLHS incorporating access cost. The performance of each method was assessed with the cross-validated RMSE and sampling route time in days. Results show that the MOOS method outperforms the two others in terms of sampling route time, especially with increasing sample size, gaining up to 1 day of work for the presented case study. It still ensures a relevant map accuracy and sample representativeness when compared to the two methods. This approach yields promising outcomes for field sampling in digital soil mapping. By simultaneously optimizing both sample representativeness and cost constraints, it holds potential as a valuable decision support tool for soil surveyors facing sampling designs in poorly accessible areas. [ABSTRACT FROM AUTHOR]

• Effective digital soil mapping can thrive with minimal covariates. • Principal components may not outperform variance inflation factor in modeling. • Quantile Regression Post-Processing assesses uncertainty regardless of model. • Variance inflation factor and recursive feature elimination ensure model parsimony. A common practice in digital soil mapping (DSM) is to incorporate many environmental covariates into a machine-learning algorithm to predict the spatial patterns of soil attributes. Variance inflation factor (VIF), principal component analysis (PCA), and recursive feature elimination (RFE) are three statistical methods that can be used to reduce the number of covariates. This study aims 1) to compare VIF and PCA approaches; 2) to identify an approach to determine the minimum number of covariates in DSM to ensure model parsimony using RFE after using VIF; and 3) to examine methods to interpret the impact of covariates on the variability of the predicted soil properties. The study area was the province of British Columbia (BC), Canada. This study used legacy data for four soil properties to make digital soil maps: soil organic carbon (SOC%), pH, clay%, and coarse fragment (CF%). Seven models were made for each soil property to determine the influence on validation results by using a different number of covariates produced by various methods on validation results. The results showed that the number of covariates could be reduced from 70 to 4 to 12 with only a little or no difference in concordance correlation coefficient (CCC) validation results. The CCC results of pH models using 70 and 7 covariates were both 0.74, and for other soil properties, this difference was negligible. The validation results obtained from PCA models showed that the performance of PCA in reducing the number of covariates was not as effective as when using VIF. Moreover, this study showed that covariates related to precipitation were the most important for modeling SOC%, soil pH, and clay%. Topographic covariates were the most influential covariates for modeling soil CF%. This study emphasizes the potential benefits of combining various data reduction methods to achieve optimal outcomes and generate the most parsimonious and interpretable models. [ABSTRACT FROM AUTHOR]

• Samples from different depths exhibit varying levels of importance for 3D mapping. • Surface soils contribute to significant variation in prediction accuracy. • Reducing subsurface soils consistently reduce prediction accuracy. • Taking more lateral profiles is more important than vertical samples. Sampling density and depth play crucial roles in three-dimensional (3D) soil modeling and prediction, particularly in digital soil mapping (DSM). However, previous studies have yielded inconsistent and even contradictory results regarding impacts of sampling density and depth on the accuracy of 3D DSM. Hence, this study aimed to evaluate the impacts of sampling depth, vertical sampling density and lateral profile density on 3D soil mapping accuracy based on a case study of soil organic carbon (SOC). To achieve this, a comprehensive analysis was conducted based on 511 soil samples collected from 111 profiles in a local hilly area spanning 5.52 km2 in China. A 3D regression geostatistical approach was employed for the analysis. The results revealed that samples taken from different depth intervals exhibited varying degrees of importance in relation to prediction accuracy. Notably, a reduced number of surface soil samples (0–0.3 m) led to significant fluctuations in the accuracy of predictions for the entire soil profile. Furthermore, a lower number of subsurface soil samples (0.3–0.6 m) also diminished the overall prediction accuracy, while the influence of deeper soil samples (0.6–1.2 m) on the overall accuracy was relatively less pronounced. Reducing the number of profiles for calibration led to significantly worse and more variable prediction accuracy, compared to reducing vertical samples. For 3D mapping, this study recommends prioritizing the collection of additional lateral profiles, specifically focusing on surface and subsurface soils, especially for properties that display vertical variation similar to SOC. [ABSTRACT FROM AUTHOR]

Potassium (K) deficiency in wine grapes results in reduced vine growth, premature leaf drop, and yield and color loss. K can be fixed in the interlayer of clay minerals in a process called K fixation, which leads to high spatial variability in soil K. In the Lodi American Viticulture Area (AVA) management of winegrapes is complicated by a mix of K fixing soils and non-K fixing soils. Here, we leverage a digital soil mapping (DSM) framework to identify the spatial distribution of K fixation and availability, to disentangle the complexity of K management in the region. Soil samples (n = 107) were collected, analyzed for the K fixation index, K availability and cation exchange capacity (CEC), and aggregated into two depths (0–30 cm and 30–100 cm). Soil samples were intersected with remotely sensed proxies for the soil forming factors and existing soil survey data and used to train a "super learner" ensemble or combination of base models, including random forest (RF), extreme gradient boosting (XGB) and cubist. Base models were combined via model averaging (each model weighted by its R2) or model stacking (linear combination of base models via OLS regression), and model performance was compared. We generated mapped uncertainties from a super learning framework by utilizing bootstrapped realizations of each base model and weighting each bootstrapped base model map via the β-coefficients generated in the ensemble fitting step. Bootstrapped maps of the super learner were utilized to generate upper and lower 90% prediction limits. For the K fixation index at the 0–30 cm depth, RF outperformed other models (R2 = 0.42), whereas a linear combination of all base models performed best in the 30–100 cm depth (R2 = 0.41). Results improved for K availability in the 0–30 cm horizons (R2 = 0.48) and the 30–100 cm horizons (R2 = 0.46). Overall, predictions for CEC were superior to both the K fixation index and K availability at both depths (0–30 cm; R2 = 0.71) and (30–100 cm; R2 = 0.51). We conclude that K fixation and availability can be predictively mapped with marginal success, while CEC is amenable to a DSM framework. CEC is tied more to soil genesis and formation, while K fixation and availability are affected by K fertilization. Finally, we compared the DSM K fixation index map to an existing soil landscape model K fixation index map, to facilitate discussion of the connection between pedogenic state factors, soil forming processes and soil properties in soil mapping. Results inform the pedogenic foundations of DSM, as well as global efforts to utilize DSM to map K fixing soils and manage K-based crop interventions. [ABSTRACT FROM AUTHOR]

• The FOD technique refines soil spectral features, improving SOC mapping. • 3D spectral indices correlate more strongly with SOC compared to 2D indices. • Integrating low-order FOD and 3D indices maximizes SOC prediction accuracy. • ZY1-02D hyperspectral satellite demonstrated advanced soil property prediction. Monitoring soil organic carbon (SOC) content is crucial for climate change mitigation and sustaining ecological balance. Despite the unparalleled advantages of hyperspectral data in capturing nuanced variations in soil properties through its high spectral resolution, effectively extracting useful features from numerous bands via spectral processing techniques remains a formidable challenge. This study proposes an integrated approach combining fractional-order derivative (FOD) technique and optimal band combination algorithm using ZY1-02D satellite hyperspectral data to estimate SOC in Northeast China's Black soil region. Three modeling strategies were compared: (1) FOD-transformed reflectance (FOD spectra), (2) FOD spectra with traditional 2D spectral indices (FOD + 2D SI), and (3) FOD spectra with new 3D spectral indices (FOD + 3D SI). These strategies were implemented using the random forest model with the aim of the optimal SOC prediction. Results showed that the application of FOD technique for spectral transformation effectively addressed the challenges posed by overlapping peaks and baseline drift inherent in the original spectral reflectance. Additionally, FOD transformation enhanced subtle soil spectral features and yielded more pronounced spectral variations with increasing fractional order, as compared to the original spectral data and conventional integer-order derivatives (i.e., first and second-order derivatives). However, as the FOD order continued to increase beyond 1.4, the spectral curve exhibited amplified noise and distortion, thereby adversely impacting subsequent model development. The 3D spectral indices correlate more robustly with SOC than 2D indices. The model that combines 0.6-order FOD and 3D spectral indices achieved the best accuracy (R2 = 0.66, RMSE = 2.99 g/kg and MAE = 2.42 g/kg), significantly outperforming the models built by 0.6-order FOD spectra (R2 = 0.48, RMSE = 3.65 g kg−1, and MAE = 2.93 g kg−1) and 0.8-order FOD + 2D SI modeling strategy (R2 = 0.55, RMSE = 3.54 g kg−1, and MAE = 2.85 g kg−1). These findings indicated that FOD and 3D spectral indices exhibit superior synergistic performance in SOC prediction, demonstrating their feasibility and providing valuable insights for large-scale soil property prediction and mapping using satellite hyperspectral data. [ABSTRACT FROM AUTHOR]

Soil biological properties (SBPSs) are crucial for soil health and ecosystem functioning. Mapping these properties can provide insights into soil ecosystems and help develop management and conservation strategies. The objective of this study was to map five important indicators of SBPSs, namely urease (Ur), acid phosphatase (ACP), alkaline phosphatase enzyme (ALP) activity, metabolic quotient (qCO 2), and soil microbial biomass to soil organic carbon ratio (MBC: SOC) using machine learning algorithms (MLA) and to identify the most important driving factors of these properties. The study was conducted in Honam sub-basin, Lorestan, Iran, on an area of 14,000 ha. 90 surface soil samples (0–30 cm) were collected and analyzed in the laboratory. 60 soil and environmental covariates, including soil, topography, and remote sensing data (RS), were used to represent influencing factors. The most relevant covariates were selected based on correlation and recursive feature elimination (RFE) analysis using two state-of-the-art MLAs, Random Forest (RF) and Extreme Gradient Boosting (XGBoost) trees, were used to predict SBPSs based on the selected covariates. Relative importance analysis and partial dependence plots were used to evaluate the effects of different factors on the prediction of SBPSs. The results showed that 28 soil and environmental covariates were selected as relevant factors. Both RF and XGBoost models performed well and achieved acceptable accuracy (as measured by R2 value). XGBoost outperformed RF on qCO 2 (R2 = 0.61), ALP (R2 = 0.73) and Ur (R2 = 0.75), while RF performed better on ACP (R2 = 0.65) and the MBC: SOC (R2 = 0.81). Relative importance analysis showed that ACP and Ur were most strongly associated with the soil variables at 79 % and 42 %, respectively. ALP and qCO 2 were strongly correlated with topographic attributes at 56 % and 52 %, respectively. MBS: SOC was about 40 % associated with the RS indices. The influence of each SBPS varied depending on the specific types of covariates. These results highlight the potential of targeted strategies for effective management and protection of soil ecosystem leading to improved soil health and ecosystem function. Further research should focus on developing methods that encompass a broader range of indices that influence soil microbial activity in drylands. [ABSTRACT FROM AUTHOR]

Heterogeneity in soil fertility conditions impacts fertiliser use efficiency in smallholder cropping systems in sub-Saharan Africa. A study was performed to generate insights in nutrient limitations for cassava (Manihot esculenta Crantz.). We conducted 627 nutrient omission trials over three years in South East (SEN) and South West Nigeria (SWN), and in the Southern (TSZ) and Lake Zone of Tanzania (TLZ) to quantify variation in root yield responses to N, P and K, and relate these to digital soil maps and weather information. Mean fresh root yields were 30, 21, 13 and 15 Mg ha-1 with an application of 150–40-180 kg N-P-K ha-1 and 20, 16, 11 and 14 Mg ha-1 without nutrient addition in the four study areas, respectively. Root yield response to nutrients was largest in SEN, with mean root yield reductions of 5.7, 3.3 and 2.7 Mg ha-1 due to omission of N, P and K, respectively. Differences in yield and yield response to fertiliser between study areas were governed by rainfall conditions, which were most favourable in SEN, and least favourable in the TLZ. Within study areas, large spatial variation was observed, while temporal variation was limited. Spatial variation in yield response was largest in SEN with standard deviations (sd) of 4.2 Mg ha-1 for K, 3.2 Mg ha-1 for P and 3.0 Mg ha-1 for N, opposite to the order of the mean yield responses. More than 75% of the variation in root yield response was observed at < 10 km scale, and close to 50% at < 1 km scale. Large variation (sd = 1.8 – 2.5 Mg ha-1) in response to N was also observed in all other study areas, and to K in SWN (sd = 2.4 Mg ha-1) and in the TSZ (sd = 2.1 Mg ha-1). Responses to N and P were weakly but significantly correlated to organic C, total N and clay contents, while response to K was correlated to extractable cations and clay content. Random forest models explained 16 – 59% of the variation in nutrient responses within study areas using digital soil maps and weather information. Our results confirm that responses to fertiliser nutrients vary in smallholder systems at a very local scale, and patterns vary between study areas and between individual nutrients. Digital soil maps and weather information can explain some of this variation and could support the development of site-specific recommendations at an appropriate scale. However, digital soil map data are unlikely sufficient to provide reliable advice at farm or plot level. Further research is needed to evaluate the capability of other factors, particularly local indicators of soil fertility, crop productivity and crop management intensity to complement digital soil maps in the development of field-specific fertiliser advice. • Cassava root yield was evaluated in 627 nutrient omission trials in 4 study areas. • Rainfall governed overall differences in fertiliser responses between study areas. • Nutrient responses within study areas varied substantially at < 1 km distances. • Digital soil map parameters explained roughly 40% of this short-range variation. • Digital soil maps can enable developing site-specific cassava fertiliser advice. [ABSTRACT FROM AUTHOR]

• Soil reflectance composites were generated with Sentinel-2 L2A data. • A multi-channel CNN learning framework for soil organic carbon prediction is proposed. • A dual input deep learning architecture improves the predictions. • The framework produced results that can be interpreted. The Copernicus Sentinel-2 multispectral imagery data may be aggregated to extract large-scale, bare soil, reflectance composites, which enable soil mapping applications. In this paper, this approach was tested in the German federal state of Bavaria, to provide estimations for soil organic carbon (SOC). Different temporal ranges were considered for the generation of the composites, including multi-annual and seasonal ranges. A novel multi-channel convolutional neural network (CNN) is proposed. By leveraging the advantages of deep learning techniques, it utilizes complementary information from different spectral pre-treatment techniques. The SOC predictions indicated little dissimilarity amongst the different composites, with the best performance attained for the six-year composite containing only spring months (RMSE = 12.03 g C · kg−1, R2 = 0.64, RPIQ = 0.89). It has been demonstrated that these outcomes outperform other well-known machine learning techniques. An ablation analysis was accordingly performed to evaluate the interplay of the CNN's different components to disentangle the advantages of each aspect of the proposed framework. Finally, a DUal inPut deep LearnIng architecture, named DUPLICITE, is proposed, which concatenates deep spectral features derived from the CNN mentioned earlier, as well as topographical and environmental covariates through an artificial neural network (ANN) to exploit their complementarity. The proposed approach was demonstrated to provide an improvement in the overall prediction performance (RMSE = 11.60 gC · kg−1, R2 = 0.67, RPIQ = 0.92). [ABSTRACT FROM AUTHOR]

• The mean K was 0.035 t ha h ha−1 MJ−1 mm−1 in China and varies with soil and water conservation zones. • The RF model has high accuracy in predicting K, with R2 = 0.49, RMSE = 0.0077 t ha h ha−1 MJ−1 mm−1. • The machine learning method represented more spatial variation detail. • Topographical and climatic variables were the primary variables affecting K. Soil erodibility (K) is the intrinsic susceptibility of a soil to water erosion. Currently, its detailed and accurate spatial distribution information especially over large areas is urgently required for national and regional soil erosion assessment and soil conservation decision making. This study combined pedotransfer function with digital soil mapping techniques to develop a high-resolution map of soil erodibility across China. The First, based on a recent national soil survey, we adopted the erosion-productivity impact calculator (EPIC) to calculate soil erodibility values at 4710 soil sampling points. Then, with the caclulated values of points, we used five techniques including polygon linking (PL), ordinary kriging (OK), Cubist, extreme gradient boosting (XGBoost), and random forests (RF) to generate spatial distribution of soil erodibility. The three latter machine learning techniques modeled the quantitative relationships between soil erodibility and a set of environmental covariates. The results showed that machine learning methods exhibited much more spatial details than the PL and OK did. Among the five techniques the RF achieved the highest accuracy with R2 of 0.49 and RMSE of 0.0077 t ha h ha−1 MJ−1 mm−1 based on 10-fold cross-validation. Spatial uncertainty analysis of the RF predictions showed that high uncerntainty occurred in northwestern China and low uncertainty in the center and southeast. We found that topographical and climatic variables are major environmental factors indirectly controlling spatial variation of soil erodibility while the soil particle composition and SOC contents directly influence the variation. [ABSTRACT FROM AUTHOR]

• National soil organic carbon maps for cropland based on soil reflectance composite. • Local ensemble model that accounts for spatial dependencies in predictions. • Local variable importance of the spectral bands is attributed to soil properties. • New approach to reduce spatial autocorrelation and increase model interpretability. • Enhancing digital soil mapping in regions with heterogeneous soil conditions. National soil organic carbon (SOC) maps are essential to improve greenhouse gas accounting and support climate-smart agriculture. Large-scale SOC models based on wall-to-wall soil information from remote sensing remain a challenge due to the high diversity of natural soil conditions and the difficulty of accounting for the spatial location of the soil samples. In this study, we tested if the implementation of local ensemble models (LEM) can be used to improve the SOC predictions from Landsat-based soil reflectance composites (SRC) for Germany. For this, we divided the research area into 30 times 30 km tiles and calculated local generalized linear models (GLM) based on random, nearby observations. Based on the GLMs, local SOC maps were predicted and aggregated using a moving window approach. The local variable importance was analyzed to identify spatial dependencies in the correlation between the SRC and SOC. For the final SOC map, a Random Forest (RF) model was trained using the aggregated local SOC predictions, the SRC, and a full set of training samples from the agricultural soil inventory. The results show that the LEM was able to improve the accuracy (R2 = 0.68; RMSE = 5.6 g kg−1), compared to the maps based on a single, global model (R2 = 0.52; RMSE = 6.8 g kg−1). The local variable importance of the spectral bands showed clear spatial patterns throughout the research area. Differences can be explained by the local soil conditions, influencing the correlation between SOC and the spectral properties. Compared to the widely adopted integration of distance covariates such as geographical coordinates, the LEM was able the reduce the spatial autocorrelation to a greater extent and to improve the prediction accuracy, especially for underrepresented SOC values. The LEM presents a new method to integrate spatial information and increase the interpretability of DSM models. [ABSTRACT FROM AUTHOR]

[Display omitted] Soil spatial information has traditionally been presented as polygon maps at coarse scales. Solving global and local issues, including food security, water regulation, land degradation, and climate change requires higher quality, more consistent and detailed soil information. Accurate prediction of soil variation over large and complex areas with limited samples remains a challenge, which is especially significant for China due to its vast land area which contains the most diverse soil landscapes in the world. Here, we integrated predictive soil mapping paradigm with adaptive depth function fitting, state-of-the-art ensemble machine learning and high-resolution soil-forming environment characterization in a high-performance parallel computing environment to generate 90-m resolution national gridded maps of nine soil properties (pH, organic carbon, nitrogen, phosphorus, potassium, cation exchange capacity, bulk density, coarse fragments, and thickness) at multiple depths across China. This was based on approximately 5000 representative soil profiles collected in a recent national soil survey and a suite of detailed covariates to characterize soil-forming environments. The predictive accuracy ranged from very good to moderate (Model Efficiency Coefficients from 0.71 to 0.36) at 0–5 cm. The predictive accuracy for most soil properties declined with depth. Compared with previous soil maps, we achieved significantly more detailed and accurate predictions which could well represent soil variations across the territory and are a significant contribution to the GlobalSoilMap.net project. The relative importance of soil-forming factors in the predictions varied by specific soil property and depth, suggesting the complexity and non-stationarity of comprehensive multi-factor interactions in the process of soil development. [ABSTRACT FROM AUTHOR]

The area of Burdur Lake gradually decreased from 211 to 122 km2 between 1975 and 2021. In this study, we explored the effects of time, environmental factors, and land uses on the spatial distribution of soil chemical properties (SOC, CaCO 3 , EC, and pH). In total 240 soil samples were taken (0–20 cm depth) at a 400 × 400 m grid sampling in the former lakebed, pasture, and agricultural fields representing a chronosequence. We found a considerable temporal and topographic trend in SOC, CaCO 3 , EC, pH, and vegetation density. The SOC, EC, and pH were higher in younger soils. Land use significantly affected soil properties. Natural vegetation species mainly grew in the pasture but only sparsely in the former lakebed areas. Salt-affected and alkaline soils and exposure time affected the density and spatial distribution of vegetation in the former lakebed and pasture areas. We tested different models to spatially predict soil properties and found that random forest had higher prediction accuracy for SOC (r2 = 0.50) and CaCO 3 (r2 = 0.36), whereas the cubist model had higher prediction accuracy for EC (r2 = 0.60) and pH (r2 = 0.38). We conclude that the vegetation density and distribution of SOC, CaCO 3 , EC, and pH were controlled by topography, time, and land use. The time and space-for-time substitution approach can be used to study soil chronosequence and biosequence in areas with shrinking lakes. [ABSTRACT FROM AUTHOR]

• Multi-sensor comparison for SOM mapping. • Landsat-8 is currently the best choice for SOM mapping. • Temporal resolution crucial for accurate SOM prediction. • Similar spatial trends in different SOM products. Using remote sensing images to predict the spatial distribution of SOM content is a classic digital soil mapping problem. The purpose of this study is to compare the SOM mapping performance of Landsat 8 OLI, Landsat 7 ETM +, Sentinel-2 MSI and MODIS. We take the dry land in the black soil area of Northeast China as the study area and 791 soil samples were collected in the field to directly measure the SOM content, which serves as the evaluation dataset. We used four images to synthesize soil images for May over multi-year periods. May was identified as the best time of year to image the bare soil in Northeast China. All images used for the median synthesis consisted of cloud masked datasets acquired in May. Then, the image band, spectral index were used as inputs, and the SOM mapping results derived from the different satellite images was evaluated using the random forest regression algorithm. In addition, we use simulation methods to evaluate the impact of temporal, spatial, and spectral differences on SOM mapping. The results show that (1) Landsat-8 is currently the best choice for SOM mapping, Sentinel-2 images have the most potential for future SOM mapping, Landsat-7 and MODIS images can provide support for determining the historical spatial distribution of SOM content; (2) improving the temporal, spatial, and spectral resolution of images can improve the accuracy of SOM prediction, and temporal resolution is the most important because higher temporal resolution makes the selection of image pixels more conducive to accurate SOM prediction, thereby reducing the impact of the field environment; and (3) there is little difference in the spatial distribution trends and average values identified from the SOM products obtained from the four satellites, and can be generalized in some application scenarios. [ABSTRACT FROM AUTHOR]

[Display omitted] • We produced a high resolution (30 m) SOC map in cropland of Southern China. • Our produced SOC map has high accuracy with R2 = 0.73 and RMSE = 2.91 g kg−1. • The map accuracy was enhanced by incorporating soil management information. • Soil properties and soil management measures are most relevant to SOC variation in cropland. An accurate and fine map of soil organic carbon (SOC) plays a vital role in understanding the global carbon cycle and achieving soil carbon sequestration potential. Although global and national maps of SOC are already available with various spatial resolutions, limited sample size, and relative coarse resolution hinder its accuracy and application in local regions. Here, we collected 13,424 soil samples and information of 45 environmental covariates from cropland in Jiangxi Province, Southern China during 2012 and 2013. Then, the optimal covariates were selected by a recursive feature elimination algorithm for mapping SOC. After that, we deployed the random forest (RF) to produce a fine map (30 m) of SOC in the cropland of Jiangxi Province and 100 times bootstrapping was performed to calculate the prediction uncertainty. Finally, we determined the impacts of various covariates on SOC variability using RF and partial least squares structural equation modeling. Our results showed that compared with the predictive model without soil management information, introducing soil management information improved the predictability of SOC with an increase in R2 by 7.35 % (0.73 vs 0.68) and decrease in RMSE by 7.03 % (2.91 vs 3.13 g kg−1). Our results well estimated the uncertainty of the predicted result with a PICP of 0.91 for a 90 % prediction interval. Soil properties and soil management activities make the largest contribution for modelling SOC. Specifically, the total nitrogen content, straw return amount, total potassium content, and multi-resolution valley bottom flatness were found as the most important factors for mapping SOC in the cropland of our study area. Overall, this study deepened our knowledge of the variation of SOC and also emphasizes that incorporating soil management information could help us to achieve more accurate predictions of SOC. [ABSTRACT FROM AUTHOR]

• Digital Soil Mapping of oxalate-extractable iron and aluminium contents. • Predictions at 25 m spatial resolution across six soil depth intervals. • Prediction uncertainty using a quantile regression forest algorithm. • Map quality assessed with design-based statistical inference. • Maps aid in balancing crop production, water quality and carbon sequestration. Amorphous iron- and aluminium-(hydr)oxides are key soil properties in controlling the dynamics of phosphorus availability and carbon storage. These oxides affect the potential of soils to retain phosphorus and carbon, thus affecting ecosystem services such as crop production, water quality and carbon sequestration. In this study, we spatially predicted oxalate-extractable Fe and Al (Fe OX , Al OX) contents in the Netherlands at 25 m resolution across six soil depth layers between 0 and 200 cm and quantified the associated prediction uncertainty using quantile regression forest. For model training and validation, geo-referenced data of Fe OX and Al OX contents were used including 12,110 wet-chemical observations and 102,393 NIR spectroscopy observations. Over 150 spatial covariates were selected that provide information about soil typology, climate, soil organisms, land use, relief, parent material and space (sampling depth and oblique coordinates). Map quality was assessed by comparing predictions with observations using an independent data set of 4841 soil samples from agricultural fields. Soil sample locations were selected by stratified random sampling, allowing us to assess map quality using design-based statistical inference. Map quality was evaluated using the metrics Model Efficiency Coefficient (MEC), Root Mean Square Error (RMSE) and Mean Error (ME). Map quality differed, depending on the target variable and soil depth, with MEC ranging from 0.19 to 0.80, RMSE from 13.5 to 56.9 mmol kg−1 and ME from −6.8 to 6.8 mmol kg−1. Overall, map quality was better for topsoil than for subsoil and better for Al OX contents than for Fe OX contents. Prediction uncertainty quality was evaluated by calculating the Prediction Interval Coverage Probability of the 90 per cent Prediction Interval, which were close to 0.90 in all cases and slightly below 0.90 for Al OX. Thus, prediction uncertainties were generally reliable, though for Al OX contents uncertainty was slightly underpredicted. The maps are a valuable tool for site-specific manure and fertiliser management strategies aiming to balance crop production, water quality and carbon sequestration in agriculture. [ABSTRACT FROM AUTHOR]

• This study quantified global soil C research trends and the proportional focus on SOC and SIC pools. • Soil inorganic C is an overlooked pool with only 4% share in global soil C research. • Approximately 40% of the soil C research was dedicated to climate change. • Soil C research is concentrated in few countries with 70% research contributed by 9 countries. • It is imperative to expand research focus on SIC to gain a comprehensive understanding of global C cycle. Soils are a major player in the global carbon (C) cycle and climate change by functioning as a sink or a source of atmospheric carbon dioxide (CO 2). The largest terrestrial C reservoir in soils comprises two main pools: organic (SOC) and inorganic C (SIC), each having distinct fates and functions but with a large disparity in global research attention. This study quantified global soil C research trends and the proportional focus on SOC and SIC pools based on a bibliometric analysis and raise the importance of SIC pools fully underrepresented in research, applications, and modeling. Studies on soil C pools started in 1905 and has produced over 47,000 publications (>1.7 million citations). Although the global C stocks down to 2 m depth are nearly the same for SOC and SIC, the research has dominantly examined SOC (>96 % of publications and citations) with a minimal share on SIC (<4%). Approximately 40 % of the soil C research was related to climate change. Despite poor coverage and publications, the climate change-related research impact (citations per document) of SIC studies was higher than that of SOC. Mineral associated organic carbon, machine learning, soil health, and biochar were the recent top trend topics for SOC research (2020–2023), whereas digital soil mapping, soil properties, soil acidification, and calcite were recent top trend topics for SIC. SOC research was contributed by 151 countries compared to 88 for SIC. As assessed by publications, soil C research was mainly concentrated in a few countries, with only 9 countries accounting for 70 % of the research. China and the USA were the major producers (45 %), collaborators (37 %), and funders of soil C research. SIC is a long-lived soil C pool with a turnover rate (leaching and recrystallization) of more than 1000 years in natural ecosystems, but intensive agricultural practices have accelerated SIC losses, making SIC an important player in global C cycle and climate change. The lack of attention and investment towards SIC research could jeopardize the ongoing efforts to mitigate climate change impacts to meet the 1.5–2.0 °C targets under the Paris Climate Agreement of 2015. This bibliographic study calls to expand the research focus on SIC and including SIC fluxes in C budgets and models, without which the representation of the global C cycle is incomplete. [ABSTRACT FROM AUTHOR]

There is increasing interest in the application of digital soil mapping (DSM) projections to infer changes in soil carbon across both space and time. This approach relies on the assumption that the spatially modeled soil carbon-environment relationship can be transferred over time. However, this assumption is rarely tested due to a lack of temporally independent validation data. This paper assesses this assumption by developing models of topsoil organic carbon stocks (SOCS) with a deep learning algorithm and data covering mainland China pertaining to the 1980s and 2010s. The temporal prediction performance of models capturing a specific period was assessed by evaluating their performance in the prediction of data during another period. The results revealed that the prediction accuracy of temporal modeling decreased, as indicated by the coefficient of determination, and was lower than that of spatial modeling. The lower prediction accuracy obtained with the DSM-projection approach may result from differences in the magnitudes of influential variables across periods. We found that different levels of environmental similarity and model projection sensitivity to dynamic variables may cause discrepancies in forecast and hindcast accuracy. Our results demonstrate that the prediction error in temporal modeling is related to the degree of environmental similarity between periods. Our findings generally support the implementation of the DSM-projection approach in soil carbon change modeling. However, caution should be exercised, as there exists much uncertainty regarding the projection of spatial models over time. • DSM projection may facilitate a more precise estimation of large-scale soil carbon changes. • Discrepancies in the environment between periods reduce the temporal predictive skill. • The variability in the model performance is related to the magnitude of environmental influences. [ABSTRACT FROM AUTHOR]

• We spatially predicted soil chemical attributes in Antarctica with Random Forest. • We evaluated the performance through accuracy and uncertainty issues. • Bases Sum, pH, and Na values were higher in Antarctic Peninsula and TOC, H + Al, and P were higher in Maritime Antarctica. • Climate, relief, and vegetation influenced the attributes distribution at different scales. • We created novel soil chemical maps for important regions of the Antarctic continent. Digital Soil Mapping includes quantitative estimations of soil attributes in areas with little or no soil information, being important in remote landscapes such as the Antarctic ice-free areas. Our objective was to predict the spatial distribution of major soil chemical attributes in two important Antarctic Peninsula regions; as well as to identify the main environmental drivers of these attributes. For this, we compiled one of the largest soil databases of Antarctica and applied the machine learning algorithm Random Forest to predict seven soil chemical attributes. We also used covariates selection and partial dependence analysis to better understand the relationships of the attributes with the environmental covariates. Bases sum was the attribute which presented the highest prediction performance, whereas Na and P predictions presented the lowest. Both accuracy and uncertainty indicators showed the difficulties of Random Forest in handling natural outliers. The soil attributes distribution was related to the mean annual temperature and annual precipitation, multispectral bands and indexes related to the vegetation response, rookeries distance expressing the birds' activity, and topographic attributes. The attributes showed a climatic gradient, with higher values of bases sum, pH, and Na in Northern Antarctic Peninsula, and higher total organic carbon, H + Al and P in Maritime Antarctic. This shows the climate covariates influence the soil variability in a macroscale, whereas terrain predictors control the soil at local scales. This study showed the Digital Soil Mapping potential to surpass the limitations of conventional mapping and indicated the feasibility in obtaining interpretable predictions which can be directly associated with the soil forming factors. Finally, the data generated can be used as reference to monitor the impacts of the climate changes on the soils of Antarctica. [ABSTRACT FROM AUTHOR]

[Display omitted] • A finest map (30 m) of soil pH in cropland of Southern China was produced. • The performance of different variable selection methods was compared. • The performance of different spatial predictive algorithms on mapping soil pH was evaluated. • We analyzed the application of our produced pH map on soil heavy metal pollution. • The best applicable scenarios of RFE and FRFS were validated and analyzed. Soil pH is one of the critical indicators of soil quality. A fine resolution map of soil pH is urgently required to address practical issues of agricultural production, environmental protection, and ecosystem functioning, which often fall short of meeting the demands for local applications. To fill this gap, we used data from an extensive survey of 13,424 surface soil samples (0–0.2 m) across the cropland of Jiangxi Province in Southern China. Using digital soil mapping techniques with 46 covariates, we produced a 30 m resolution soil map of topsoil pH in cropland of Southern China. We integrate different variable selection algorithms and machine learning methods. Our results indicate the Random Forest with covariates selected by recursive feature selection had the best performance for mapping soil pH with r of 0.583 and RMSE of 0.41. The prediction interval coverage probability for our prediction was 0.92, indicating a low estimated prediction uncertainty. Climate was identified as the most critical variable for predicting soil pH with a contribution of 37.42 %, followed by soil properties (29.09 %), soil management (21.86 %), parent material (6.22 %), and biota (5.39 %) factors. The mean topsoil pH in the cropland of Jiangxi Province was 5.21, indicating a great pressure of soil acidification in this region. The high soil pH values were mainly distributed in the Northern, Western, and Eastern parts of the survey region while low soil pH values were majorly located in the central part. Compared with past surveys in 1980 s, there was no significant soil pH change in the surveyed region. Our soil pH map can provide important implications and guidance decisions on soil heavy metal pollution remediation, precision agriculture, and prevention of soil acidification. [ABSTRACT FROM AUTHOR]

• Locally enhanced methods produced more accurate and finer-resolution maps. • Top-down digital soil maps were less accurate than conventional soil maps. • Locally enhanced digital soil maps were more accurate than any other map at 0–30 cm. • Top-down digital soil maps had relatively low variance and coarse resolution. • Local maps benefit from specialized knowledge, finer resolution, and less complexity. This study presents a regional digital soil mapping (DSM) product that used a locally enhanced method in support of a bottom-up approach to create spatial soil predictions that were more accurate than one of the most accurate and detailed conventional soil mapping (CSM) products in the world, the Soil Survey Geographic (SSURGO) map from the United States Department of Agriculture, Natural Resources Conservation Service. Prior to this work, DSM products had yet to definitively outperform CSM products, except in locations where existing CSM maps were based on low investment (e.g., generalized national soil maps). While DSM has had established advantages, such as quantitative property prediction and cost efficiency, remaining challenges for widespread adoption include obtaining accuracy and practicality that exceed CSM-based maps in areas with histories of strong soil survey programs. In comparison with SSURGO, multiple top-down DSM products (SoilGrids 2.0, POLARIS, and Soil Properties and Class 100 m Grids of the USA) were evaluated along with the locally enhanced DSM product called LE-DSM (locally enhanced digital soil map). These maps were compared for their predictions of clay, silt, sand, and organic matter content at standard depth intervals to 200 cm. SSURGO outperformed the national and global top-down DSM products for 75 % of the soil properties and depths. In contrast, LE-DSM demonstrated appreciably higher accuracy than SSURGO 46 % of the time. These results indicate that bottom-up DSM products are needed for superseding the accuracy of CSM products, which themselves are constructed from a locally enhanced bottom-up approach. [ABSTRACT FROM AUTHOR]