Search

Peatlands, which only cover 3 to 5 percent of the global land area, can store up to twice the amount of carbon as the world's forests. Although recognised for their significant role in the global carbon cycle, discovering the global extent of peatlands and their carbon stock remains challenging. Referring to the UNEP's global peatland map, here we present PEATGRIDS, a data product containing global maps of peat thickness and carbon stock created created using the digital soil mapping approach. We compiled over 25,000 observations of peatland thickness, bulk density (BD) and carbon content (CC), globally. Using the Random Forest (RF) algorithm, we estimated peat thickness and peat BD and CC at ~1 km resolution at multiple depths (0–2 m) globally. The estimates were generated using 19 land surface covariates from digital maps and remote sensing images of land use, soil characteristics, topographical features, and climate parameters. The RF models for peat thickness were trained on 25,200 points grouped into six geographic regions. Validation of the peat thickness estimates showed a good performance, with the coefficient of determination (R2)ranging from 0.15 to 0.72. The prediction for peat BD and CC followed the same model architecture and were trained on 17,000 and 7,000 points, respectively. Overall, BD and CC models performed well and consistently across soil layers with average R2 values of 0.61 for BD and 0.48 for CC. Based on the estimated peat thickness, BD and CC, the carbon stock of global peatland was estimated to be 1,029 Pg C for peat dominated area of 6.57 million km2. PEATGRIDS is made available at https://doi.org/10.5281/zenodo.12559239 (Widyastuti et al., 2024) to support further analyses and modelling of peatlands across the globe. [ABSTRACT FROM AUTHOR]

Fire alters soil hydrologic properties leading to increased risk of catastrophic debris flows and post‐fire flooding. As a result, US federal agencies map soil burn severity (SBS) via direct soil observation and adjustment of rasters of burned area reflectance. We developed a unique application of digital soil mapping (DSM) to map SBS in the Creek Fire which burned 154,000 ha in the Sierra Nevada. We utilized 169 ground‐based observations of SBS in combination with raster proxies of soil forming factors, pre‐fire fuel conditions, and fire effects to vegetation to build a digital soil mapping model of soil burn severity (DSMSBS) using a random forest algorithm and compared the DSMSBS map to the established SBS map. The DSMSBS model had a cross‐validation accuracy of 48%. The established technique had 46% agreement between field observations and pixels. However, since the established technique is manual, it could not be compared to the DSMSBS model via cross‐validation. We produced SBS class uncertainty maps, which showed high prediction probabilities around field observations, and low probabilities away from field observations. SBS prediction probabilities could aid post‐fire assessment teams with sample prioritization. We report 107 km2 more area classified as high and moderate SBS compared to the established technique. We conclude that blending soil forming factors based mapping and vegetation burn severity mapping can improve SBS mapping. This represents a shift in SBS mapping away from validating remotely sensed reflectance imagery and toward a quantitative soil landscape model, which incorporates both fire and soils information to directly predict SBS. Core Ideas: A digital soil mapping method to map wildfire soil burn severity (DSMSBS) was developed.Soil field observations were combined with rasters of environmental covariates and fire effects.Excellent fidelity between field observations of soil burn severity (SBS) and the final DSMSBS map was reported.Direct classification of SBS improves SBS mapping compared to validation of remotely sensed burn severity.Class probabilities generated for SBS may aid post‐fire assessment. [ABSTRACT FROM AUTHOR]

Soil compaction is one of the most important and readily mitigated threats to soil health. Digital Soil Mapping (DSM) has emerged as an efficient method to provide broad-scale maps by combining soil information with environmental covariates. Until now, soil information input to DSM has been mainly composed of point-based quantitative measurements of soil properties and/or of soil type/horizon classes derived from laboratory analysis, point observations, or soil maps. In this study, we used field estimates of soil compaction to map soil behavior to compaction at a national scale. The results from a previous study enabled clustering of six different behaviors using the in situ field observations. Mapping potential responses to soil compaction is an effective land management tool for preventing future compaction. Random forest was used to make spatial predictions of soil behavior to compaction over cultivated soils of mainland France (about 210,000 km2). Modeling was performed at 90 m resolution. The map enabled us to spatially identify clusters of possible responses to compaction. Most clusters were consistent with known geographic distributions of some soil types and properties. This consistency was checked by comparing maps with both national and local-scale external sources of soil information. The best spatial predictors were available digital maps of soil properties (clay, silt, sand, organic carbon (SOC) content, and pH), some indicators of soil structural quality using SOC and clay content, and environmental covariates (T °C and relief-related covariates). Predicted maps were interpretable to support management recommendations to mitigate soil compactness at the soil–scape scale. Simple observational field data that are usually collected by soil surveyors, then stored and available in soil databases, provide valuable input data for digital mapping of soil behavior to compaction and assessment of inherent soil sensitivity to compaction. [ABSTRACT FROM AUTHOR]

Intensive agricultural practices lead to a deterioration in soil quality, causing a decline in farm productivity and quality, and disturbing the ecosystem balance in command areas. To achieve sustainable production and implement effective soil management strategies, understanding the state and spatial variability of soil quality is essential. The study aims to enhance the understanding of soil quality variability and provide actionable insights for sustainable soil management. In this regard, principal component analysis (PCA) and digital soil mapping were used to assess and map the spatial variability of the soil quality index (SQI) in the Cauvery command area, Mandya district, Karnataka, India. A total of 145 georeferenced soil samples were drawn at 0–15 cm depth and analyzed for physico-chemical properties. PCA was used to reduce the dataset into a minimum dataset as eight important soil indicators and to determine relative weightage factors, which were used for assessing SQI with linear and non-linear scoring methods. For spatial assessment of SQI, the random forest algorithm with environmental covariates was used to map eight soil indicators selected in the minimum dataset. The soil property maps were subjected to linear and non-linear scoring, followed by multiplying with corresponding weightage factors and summation to produce SQI maps. Results reveal that values of SQI calculated using linear scoring, range from 0.10 to 0.64, with a mean of 0.39, while non-linear scoring exhibits a wider range of 0.12 to 0.78 and a mean of 0.48. With a slight higher sensitivity index of 6.5, non-linear scoring proved to be the better scoring method compared to linear scoring. Spatial assessment shows that the R2 and LCC between the calculated and predicted SQI were higher for non-linear scoring (0.66 and 0.66) compared to linear scoring (0.60 and 0.65). The SQI maps reveal high spatial variability with more than 40 percent of soils classified as moderate-to-low index. The soils with low SQI were distributed in eastern parts, whereas western parts exhibited high-to-very-high soil quality. To achieve production goals and improve soil quality in the eastern region, sustainable soil and crop management strategies must be developed, and their effects on soil quality should be assessed. [ABSTRACT FROM AUTHOR]

Synthesizing bare soil pictures in regions with complex vegetation is challenging, which hinders the accuracy of predicting soil organic carbon (SOC) in specific areas. An SOC prediction model was developed in this study by integrating the convolutional neural network and long and short-term memory network (CNN-LSTM) algorithms, taking into consideration soil-forming factors such as climate, vegetation, and topography in Hainan. Compared with common algorithmic models (random forest, CNN, LSTM), the SOC prediction model based on the CNN-LSTM algorithm achieved high accuracy (R2 = 0.69, RMSE = 6.06 g kg−1, RPIQ = 1.96). The model predicted that the SOC content ranged from 5.49 to 36.68 g kg−1, with Hainan in the central and southern parts of the region with high SOC values and the surrounding areas with low SOC values, and that the SOC was roughly distributed as follows: high in the mountainous areas and low in the flat areas. Among the four models, CNN-LSTM outperformed LSTM, CNN, and random forest models in terms of R2 accuracy by 11.3%, 23.2%, and 53.3%, respectively. The CNN-LSTM model demonstrates its applicability in predicting SOC content and shows great potential in complex areas where obtaining sample data is challenging and where SOC is influenced by multiple interacting factors. Furthermore, it shows significant potential for advancing the broader field of digital soil mapping. [ABSTRACT FROM AUTHOR]

This research focuses on understanding the spatial variation of Soil Organic Matter (SOM) and pH levels in the North of Morocco. The study employs a comprehensive approach to enhance predictive modelling, incorporating the Boruta algorithm for effective environmental covariates selection and optimizing model parameters through hyperparameter optimization. Utilizing a Random Forest (RF) model with remote sensing indices and topographic features, the research predicts SOM and pH to identify key contributors to their spatial variability. SOM prediction saw significant success, with a notable correlation to remote sensing indices such as the RVI, NDVI, and TNDVI. These indices, indicative of vegetation health and productivity, emerged as primary influencers of SOM. In comparison, the influence of topographic features like elevation, slope, and aspect was found to be less significant. Conversely, predicting pH was challenging due to the minimal spatial variability within the dataset. Addressing this limitation could involve dataset expansion or alternative models for low-correlated data handling. Despite the RF model's limited efficacy in pH prediction, an observable correlation between SOM and pH was identified, consistent with prior research. Areas with higher SOM exhibited lower pH values, indicating relative soil acidification from organic matter decomposition. The study's RF model demonstrated potential in SOM prediction using remote sensing indices, but enhancing pH prediction is essential. Future research may explore dataset expansion, diverse sampling, or testing alternative predictive models for better performance with low-correlated datasets. The study offers valuable insights for advanced predictive model development and enriches understanding of soil management practices. [ABSTRACT FROM AUTHOR]

Large-scale digital soil maps are essential for rational and sustainable land management as well as accurate fertilizer application. This study focuses on digital mapping of soil properties, namely soil organic carbon (SOC), pH, nitrogen (N), phosphorus (P), and potassium (K) in Chernozem topsoil and subsoil. The study was conducted on two arable fields in the Cis-Ural forest-steppe zone of the Republic of Bashkortostan (Russia). The random forest algorithm in combination with terrain attributes and Sentinel-2A satellite data was applied for spatial prediction of soil properties. The root-mean-square error (RMSE) and coefficient of determination (R2) were used to determine the model performance. According to the Pearson correlation, a significant positive relationship between SOC and N content was found at all sites and depths (R = 0.76–0.92). A cross-validation revealed that SOC (R2 = 0.22–0.62, RMSE = 0.35–0.89%) and N (R2 = 0.16–0.60, RMSE = 21.11–36.6 mg kg−1) were best predicted among other properties at all depths using remote sensing data, whereas the performance of predictive models decreased with depth. However, a relationship between the content of some soil properties and their spatial distribution at study depths was observed, which can be used as an additional explanatory variable. We suppose that digital mapping of soil properties at the arable field scale should not be limited to topographic and remote sensing variables. Based on this information, the use of auxiliary variables, such as collocated soil information in combination with relief and remote sensing data can be effective in more accurately estimating the spatial distribution of properties across arable fields at different depths. Overall, this study provides valuable insights into spatial modelling of the vertical distribution of soil properties, highlighting the significance of remote sensing data at the arable field scale. The findings can be valuable for land managers, agronomists, and policymakers seeking sustainable land management practices and efficient fertilizer application, as well as for developing further mapping procedures for arable fields. [ABSTRACT FROM AUTHOR]

Soil particle size fractions (PSFs) are important properties for understanding the physical and chemical processes in soil systems. Knowledge about the distribution of soil PSFs is critical for sustainable soil management. Although log-ratio transformations have been widely applied to soil PSFs prediction, the statistical distribution of original data and the transformed data given by log-ratio transformations is different, resulting in biased estimates of soil PSFs. Therefore, multivariate random forest (MRF) was utilized for the simultaneous prediction of soil PSFs, as it is able to capture dependencies and internal relations among the three components. Specifically, 243 soil samples collected across the Loess Plateau were used. Meanwhile, Landsat data, terrain attributes, and climatic variables were employed as environmental variables for spatial prediction of soil PSFs. The results depicted that MRF gave satisfactory soil PSF prediction performance, where the R2 values were 0.62, 0.53, and 0.73 for sand, silt, and clay, respectively. Among the environmental variables, nighttime land surface temperature (LST_N) presented the highest importance in predicting soil PSFs in the Loess Plateau, China. Maps of soil PSFs and texture were generated at a 30 m resolution, which can be utilized as alternative data for soil erosion management and ecosystem conservation. [ABSTRACT FROM AUTHOR]

Digital soil maps are paramount for supporting environmental process analysis, planning for the conservation of ecosystems, and sustainable agriculture. The availability of dense time series of surface reflectance data provides valuable information for digital soil mapping (DSM). A detailed soil survey, along with a stack of Landsat 8 SR data and a rainfall time series, were analyzed to evaluate the influence of soil on the temporal patterns of vegetation greenness, assessed using the normalized difference vegetation index (NDVI). Based on these relationships, imagery depicting land surface phenology (LSP) metrics and other soil-forming factors proxies were evaluated as environmental covariates for DSM. The random forest algorithm was applied as a predictive model to relate soils and environmental covariates. The study focused on four soils typical of tropical conditions under pasture cover. Soil parent material and topography covariates were found to be similarly important to LSP metrics, especially those LSP images related to the seasonal availability of water to plants, registering significant contributions to the random forest model. Stronger effects of rainfall seasonality on LSP were observed for the Red Latosol (Ferralsol). The results of this study demonstrate that the addition of temporal variability of vegetation greenness can be used to assess soil subsurface processes and assist in DSM. [ABSTRACT FROM AUTHOR]

Soil organic carbon (SOC) is an important soil characteristic as well as a way how to mitigate climate change. Information on its content and spatial distribution is thus crucial. Digital soil mapping (DSM) is a suitable way to evaluate spatial distribution of soil properties thanks to its ability to obtain accurate information about soil. This research aims to apply machine learning algorithms using various environmental covariates to generate digital SOC maps for mineral topsoils in the Liberec and Domažlice districts, located in the Czech Republic. The soil class, land cover, and geology maps as well as terrain covariates extracted from the digital elevation model and remote sensing data were used as covariates in modelling. The spatial distribution of SOC was predicted based on its relationships with covariates using random forest (RF), cubist, and quantile random forest (QRF) models. Results of the RF model showed that land cover (vegetation) and elevation were the most important environmental variables in the SOC prediction in both districts. The RF had better efficiency and accuracy than the cubist and QRF to predict SOC in both districts. The greatest R² value (0.63) was observed in the Domažlice district using the RF model. However, cubist and QRF showed appropriate performance in both districts, too. [ABSTRACT FROM AUTHOR]

Digital soil maps find application in numerous fields, making their accuracy a crucial factor. Mapping soil properties in homogeneous landscapes where the soil surface is concealed, as in forests, presents a complex challenge. In this study, we evaluated the spatial distribution of soil organic carbon stocks (SOCstock) under forest vegetation using three methods: regression kriging (RK), random forest (RF), and RF combined with ordinary kriging of residuals (RFOK) in combination with Sentinel-2A satellite data. We also compared their accuracies and identified key influencing factors. We determined that SOCstock ranged from 0.6 to 10.9 kg/m2 with an average value of 4.9 kg/m2. Among the modelling approaches, we found that the RFOK exhibited the highest accuracy (RMSE = 1.58 kg/m2, NSE = 0.33), while the RK demonstrated a lack of spatial correlation of residuals, rendering this method inapplicable. An analysis of variable importance revealed that the SWIR B12 band of the Sentinel-2A satellite contributed the most to RFOK predictions. We concluded that the RFOK hybrid approach outperformed the others, potentially serving as a foundation for digital soil mapping under similar environmental conditions. Therefore, it is essential to consider spatial correlations when mapping soil properties in ecosystems that are inaccessible for capturing the spectral response of the soil surface. [ABSTRACT FROM AUTHOR]

Digital soil mapping relies on relating soils to a particular set of covariates, which capture inherent soil spatial variation. In digital mapping of soil classes, the most commonly used covariates are topographic attributes, RS attributes, and maps, including geology, geomorphology, and land use; in contrast, the subsurface soil characteristics are usually ignored. Therefore, we investigate the possibility of using soil diagnostic characteristics as covariates in a mountainous landscape as the main aim of this study. Conventional covariates (CC) and a combination of soil subsurface covariates with conventional covariates (SCC) were used as covariates, and random forest (RF), Multinomial Logistic Regression (LR), and C5.0 Decision Trees (C5) were used as different machine learning algorithms in digital mapping of soil family classes. Based on the results, the RF model with the SCC dataset had the best performance (KC = 0.85, OA = 90). In all three models, adding soil covariates to the sets of covariates increased the model performance. Soil covariates, slope, and aspect were selected as the principal auxiliary variables in describing the distribution of soil family classes. [ABSTRACT FROM AUTHOR]

Topographic variables derived from digital elevation models (DEM) are the most commonly used predictor in soil organic carbon (SOC) prediction. The predictability of these variates for the spatial heterogenity of SOC is influenced by the size of DEM resolution. This paper aimed to investigate the resolution effect of the topography at multiple DEM resolutions and select the most important predictors and the optimal resolution for SOC prediction in a large-scale and morphologically complex area. The study area, covering 15 600 km2, was located in a mountainous province in China. A total of 3901 soil samples and three machine learning algorithms including random forest, gradient boosted regression tree, and artificial neural network were used to construct the SOC predictive models at resolutions of 30, 90, 150, 210, 270, 330, and 390 m. Topographic factors were derived from the DEM with an original resolution of 30 m and were subsequently resampled to varying resolutions by using the bilinear interpolation algorithm. The importance of each variate in each predictive model was computed for the most important predictor selection, and the determination coefficient (R2) and Root Mean Square Error (RMSE) of each predictive model were computed for the accuracy evaluation. Comparative analysis on point-based topographic representation, the accuracy of predictive models, and the variable importance was conducted among different resolutions to investigate the resolution effect on the SOC prediction and to select the optimal resolution for this study area. The results showed that topography was the key factor influencing the SOC distribution in this large-scale area due to the region's mountainous nature, and the relative contribution of the topography to predict SOC distribution varied with DEM resolutions. Elevation was the most important topographic factor at all DEM resolutions. Accuracies of predictive models built on the three machine learning algorithms were all dependent on DEM resolution, indicating that DEM resolution had an important effect on SOC prediction. Overall, the random forest predictive models outperformed the other two algorithms, and its most accurate result was obtained at the 210-m resolution with an R2 of 0.25 and an RMSE of 4.77 g kg–1. In conclusion, we found that finer resolutions did not necessarily produce more accurate SOC predictions even in a large-scale and morphologically complex area. Therefore, investigating the resolution effect is suggested to select an optimal resolution for the SOC prediction in other regions characterized by similar geomorphological conditions as this study area. [ABSTRACT FROM AUTHOR]

Many regions of Iran lack digital map of soil properties. The Chahardowli plain in western Iran is one of these areas. Due to the importance of sand, silt, and clay components, having quantitative and continuous data on abrupt changes in these two properties in this area is very critical. Therefore, to study sand, silt, and clay, samples were taken at depths of 0-5, 5-15, 15-30, 30-60, and 60-100 cm, according to GlobalSoilMap. Finally, 145 samples were collected from 30 soil profiles. The significant covariates were selected by Random Forest Recursive Feature Elimination (RF-RFE). Relationships between these characteristics and environmental predictors were modeled using random forest (RF), decision tree (DT), and multiple linear regression (MLR) models. The accuracy and precision of the models used for all three particles showed that the RF model had the most accurate prediction with R² and RMSE of 0.82 and 2.34 for clay, 0.80 and 3.87 for sand, and 0.85 and 2.89 for silt, respectively. In this study, terrain-based variables had a greater impact on improving accuracy than remote-sensing variables. The current study showed that even with limited information, digital mapping of sand, silt, and clay particles under GlobalSoilMap and the use of environmental factors can provide acceptable results. [ABSTRACT FROM AUTHOR]

Attempt to evaluate soil fertility was and still is one of the most challenging public importance. Soil nutrients are the key factors in soil fertility. For this reason, when constructing soil fertility potential, many researchers prefer to investigate soil nutrient status or use and assessment of qualitative research methods. Quantifying soil fertility is challenging since various factors such as numerous physical and chemical characteristics of soil might affect it. The proper selection of factors that may more accurately describe soil fertility is another issue. So, in this study, we developed a regional soil fertility index (SFI) based on different soil nutrients for quantifying soil fertility. After receiving fertility, a comparative study of machine learning techniques was carried out to construct its distribution map, using digital soil mapping (DSM). The spatial distribution of the SFI map showed that 55% of the studied area had poor fertility, 27.25% had moderately fertile soils, and only a tiny area had fertile soils. The results indicated that heavy soil texture and high calcium carbonate content were the most limiting factor and phosphorus and zinc were the most limiting nutrients across the studied area. Comparing machine learning techniques yielded the finding that the Random forest model has the best performance for predicting SFI (R²= 0.86) compared with the Decision tree (R²= 0.53) and Multi-linear regression (R²= 0.35). Therefore, specific soil fertility management practices and training farmers on the proper use of soil fertility management practices are recommended. [ABSTRACT FROM AUTHOR]

The world's largest cities are characterized by environmental pollution, which endangers human health and biological species. Therefore, environmental monitoring and mapping of pollution levels are important tasks. To this end, we applied regression kriging (RK) and random forest (RF) for digital mapping of potentially toxic elements (PTEs) on the territory of Saint Petersburg (Russia). We predicted concentrations of As, Cd, Cu, Hg, Ni, Pb and Zn PTEs based on 87 sites. The covariates were represented by remote sensing, terrain and anthropogenic variables. The results showed that the elements were characterized by high variability and their content varied between 1 and 37 mg/kg for As, 0.1 and 2.7 for Cd, 4.7 and 674 for Cu, 0.005 and 3.6 for Hg, 2.8 and 66.8 for Ni, 3.4 and 858.8 for Pb, 15.4 and 1306.5 for Zn. We found that RK was the best performing model for Cu, Hg, Ni, Pb and Zn, with the lowest mean absolute error (MAE), root mean squared error (RMSE), and highest coefficient of determination (R2) and Nash–Sutcliffe model efficiency coefficient (NSE) values. RF method was better for predicting As content, while the performance of the models for Cd was the same. Nevertheless, R2 values remained below 0.23, signifying a difficulty in effectively modeling PTEs within urbanized regions. The study provides insight into the benefits of combining regression analysis and kriging, and highlights the importance of considering multiple modeling approaches in spatial prediction of PTEs. [ABSTRACT FROM AUTHOR]

Readily available environmental covariates in current digital soil mapping usually do not indicate the spatial differences between deep soil attributes. This, to a large extent, leads to a decrease in the accuracy of 3D soil mapping with depth, which seriously affects the quality of soil information generated. This study tested the hypothesis that spatialized laboratory soil spectral information can be used as environmental covariates to improve the accuracy of 3D soil attribute mapping and proposed a new type of environmental covariable. In the first step, with soil-forming environmental covariates and independent soil profiles, laboratory vis-NIR spectral data of soil samples resampled into six bands in Anhui province, China, were spatially interpolated to generate spatial distributions of soil spectral measurements at multiple depths. In the second step, we constructed three sets of covariates using the laboratory soil spectral distribution maps at multiple depths: conventional soil-forming variables (C), conventional soil-forming variables plus satellite remote sensing wavebands (C+SRS) and conventional soil-forming variables plus spatialized laboratory soil spectral information (C+LSS). In the third step, we used the three sets of environmental covariates to develop random forest models for predicting soil attributes (pH; CEC, cation exchange capacity; Silt; SOC, soil organic carbon; TP, total phosphorus) at multiple depths. We compared the 3D soil mapping accuracies between these three sets of covariates based on another dataset of 132 soil profiles (collected in the 1980s). The results show that the use of spatialized laboratory soil spectral information as additional environmental covariates has a 50% improvement in prediction accuracy compared with that of only conventional covariates, and a 30% improvement in prediction accuracy compared with that of the satellite remote sensing wavebands as additional covariates. This indicates that spatialized laboratory soil spectral information can improve the accuracy of 3D digital soil mapping. [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]

The North China agro–pastoral zone is a large, ecologically fragile zone in the arid and semi-arid regions. Quantitative remote sensing inversion of soil organic carbon (SOC) in this region can facilitate understanding of the current status of degraded land restoration and provide data support for carbon cycling research in the region. Deep learning (DNN) for SOC inversion has been W.a hot topic over the past decade, but there have been few studies at the regional scale in the arid and semi-arid zones. In this study, a DNN model with five hidden layers and five skip connections was established using 644 spatially distributed SOC samples and Landsat 8 OLI imagery. The model was compared with the random forest algorithm in terms of generalization ability. The main conclusions were as follows: 1. The DNN algorithm can establish a high-precision SOC inversion model (R2 = 0.52, RMSE = 0.7), with 90% of errors concentrated in the range of −2.5 to 2.5 kg·C/m2; 2. the Boruta variable-screening algorithm can effectively improve the model accuracy of the random forest algorithm, but due to the DNN's better ability to mine hidden information in the data, the improvement effect on the DNN model accuracy is limited; 3. the SOC samples in arid and semi-arid areas are highly positively skewed, with a significant impact on the modeling accuracy of DNN, and conversion is required to obtain a model with better generalization ability; and 4. in arid and semi-arid regions, SOC has a weak correlation with vegetation indices but a stronger correlation with temperature, elevation, and aridity. This study established a reliable deep learning model for SOC density in a large arid and semi-arid region, providing a reference and framework for the establishment of SOC inversion models in other regions. [ABSTRACT FROM AUTHOR]

Numerous applications in agriculture, climate, ecology, hydrology, and the environment are severely constrained by the lack of detailed information on soil texture. The purpose of this study was to predict soil particle-size fractions (PSF) in the Ri-Bhoi district of Meghalaya state, India, using a random forest model (RF). For the modeling of soil particle-size fractions, we employed 95 soil profiles (456 depth-wise layers) gathered from a recent national land resource inventory as well as currently accessible environmental variables. Sand, silt, and clay content were predicted using the Random Forest model at varied depths of 0–5, 5–15, 30–60, 60–100, and 100–200 cm. Our results showed the R2 for sand was found to be 0.30 (0–5 cm), 0.28 (5–15 cm), and 0.21 (15–30 cm). For the sand, silt, and clay fractions, respectively, the concordance correlation coefficient (CCC) was found to be greater in the 0–30 cm, 0–60 cm, and 0–15 cm depths. When there is a reasonably close monitoring of the coverage probability with a confidence level along the 1:1 line, prediction interval coverage probability (PICP) gives a decent indicator of what to anticipate. The most crucial variables for the prediction of sand and silt were channel network base level (CNBL) and LS-Factor, whereas Min Temperature of Coldest Month (°C) (BIO6) was discovered for clay prediction. For all three soil texture fractions, the range between the 5% lower and 95% higher prediction bounds was large, indicating that the existing spatial predictions may be improved. The maps of soil texture were significantly more precise, and they accurately depicted the spatial variations of particle-size fractions. Additionally, there is still a need to investigate novel methodologies for extensive digital soil mapping, which will be very advantageous for many international initiatives. [ABSTRACT FROM AUTHOR]

Inorganic carbon is the largest source of carbon in terrestrial surface, particularly in arid and semiarid regions, including the Chahardowli Plain in western Iran. Inorganic carbon plays an equal or greater role than organic soil carbon in these areas, although less attention has been paid in quantifying their variability. The objective of this study was to model and map calcium carbonate equivalent (CCE) presenting inorganic carbon in soil using machine learning and digital soil mapping techniques. Chahardowli Plain in foothills of the Zagros Mountains in the southeast of Kurdistan Province in Iran was taken as a case study area. CCE was measured at 0–5, 5–15, 15–30, 30–60, and 60–100 cm depths following GloalSoilMap.net project specifications. A total of 145 samples were collected from 30 soil profiles using the conditional Latin hypercube (cLHS) method of sampling. Relationships between CCE and environmental predictors were modeled using random forest (RF) and decision tree (DT) models. In general, the RF model performed slightly superior than the DT model. The mean value of CCE increased with soil depth, from 3.5% (0–5 cm) to 63.8% (30–60 cm). Remote sensing (RS) variables and terrestrial variables were equally important. The importance of RS variables was higher at the surface than terrestrial variables, and vice versa. The most significant variables were Channel Network Base Level (CNBL) variable and Difference Vegetation Index (DVI) with the same variable importance value (21.1%). In areas affected by river activities, the use of the CNBL and vertical distance to channel networks (VDCN) as variables in digital soil mapping (DSM) could increase the accuracy of soil property prediction maps. The VDCN played a principal role in soil distribution in the study area by affecting the rate of discharge and, thus, erosion and sedimentation. A high percentage of carbonate in parts of the region could exacerbate nutrient deficiencies for most crops and provide information for sustainably managing agricultural activity. [ABSTRACT FROM AUTHOR]

Understanding the spatial variation of soil properties is essential for monitoring land capabilities as well as the sustainable management of soil resources. The aim of this study was to predict digital soil properties mapping using 23 environmental variables, i.e., terrain attributes and remote sensing (RS) indices, across 1500 km2 of Mashhad plain lands. To achieve this purpose, a total of 180 soil samples (0–10 cm) were taken. The random forest (RF) model combined with ordinary kriging (OK), as well as regression kriging (RK), were applied to relate environmental variables and the studied soil properties. The results revealed that RF-OK was the best model with R2 and RMSE for silt (0.89% and 0.10%), followed by calcium carbonate equivalent (0.88% and 3.30%), clay (0.87% and 2.26%), soil organic carbon (0.86% and 0.24%), sand (0.84% and 4.21%), and pH (0.82% and 5.42%). The RS covariates, including band 5 (B5), modified soil-adjusted vegetation index (MSAVI), difference vegetation index (DVI), band 2 (B2), carbonate rock index (CRI2), gypsum index (GI), and enhanced vegetation index (EVI), and terrain attributes, including topographic wetness index (TWI) and elevation (EL), and topographic position index (TPI), were the most important variables in modeling different soil properties. RF-OK showed the prediction and uncertainty maps related to high precision and low standard deviation in most study areas, which indicate low overfitting and overtraining in modeling processes. In general, the RF-OK model, with low cost and high accuracy, can be applicable to use for predicting different soil properties, as well as spatial information acquired from an effort to maps to managing agriculture in areas at different conditions. Finally, this method can be applied to other regions of similar properties and for similar purposes. [ABSTRACT FROM AUTHOR]

The problem of salinization/spreading of saline soils is becoming more urgent in many regions of the world, especially in context of climate change. The monitoring of salt-affected soils' properties is a necessary procedure in land management and irrigation planning and is aimed to obtain high crop harvest and reduce degradation processes. In this work, a machine learning method was applied for modeling of the spatial distribution of topsoil (0–20 cm) properties—in particular: soil organic carbon (SOC), pH, and salt content (dry residue). A random forest (RF) machine learning approach was used in combination with environmental variables to predict soil properties in a semi-arid area (Trans-Ural steppe zone). Soil, salinity, and texture maps; topography attributes; and remote sensing data (RSD) were used as predictors. The coefficient of determination (R2) and the root mean square error (RMSE) were used to estimate the performance of the RF model. The cross-validation result showed that the RF model achieved an R2 of 0.59 and an RMSE of 0.68 for SOM; 0.36 and 0.65, respectively, for soil pH; and 0.78 and 1.21, respectively for dry residue prediction. The SOC content ranged from 0.8 to 2.8%, with an average value of 1.9%; soil pH ranged from 5.9 to 8.4, with an average of 7.2; dry residue varied greatly from 0.04 to 16.8%, with an average value of 1.3%. A variable importance analysis indicated that remote sensing variables (salinity indices and NDVI) were dominant in the spatial prediction of soil parameters. The importance of RSD for evaluating saline soils and their properties is explained by their absorption characteristics/reflectivity in the visible and near-infrared spectra. Solonchak soils are distinguished by a salt crust on the land surface and, as a result, reduced SOC contents and vegetation biomass. However, the change in saline and non-saline soils over a short distance with mosaic structure of soil cover requires high-resolution RSD or aerial images obtained from unmanned aerial vehicle/drones for successful digital mapping of soil parameters. The presented results provide an effective method to estimate soil properties in saline landscapes for further land management/reclamation planning of degraded soils in arid and semi-arid regions. [ABSTRACT FROM AUTHOR]

Accurate understanding of spatial distribution and variability of soil total nitrogen (TN) is critical for the site-specific nitrogen management. Based on 4337 newly obtained soil observations and 33 covariates, this study applied the random forest (RF) algorithm and modified regression kriging (RF combined with residual kriging: RFK, hereafter) model to spatially predict and map topsoil TN content in agricultural areas of Henan Province, central China. According to the RFK prediction, topsoil TN content ranged from 0.52 to 1.81 g kg−1, and the farmland with the topsoil TN contents of 1.00–1.23 g kg−1 and 0.80–1.23 g kg−1 accounted for 48.2% and 81.2% of the total farmland area, respectively. Spatially, the topsoil TN in the study area was generally higher in the west and lower in the east. By using the Boruta variable selection algorithm, soil organic matter (SOM) and available potassium contents in topsoil, nitrogen deposition, average annual precipitation, livestock discharges, and topsoil pH were identified as the main factors driving the spatial distribution and variation of soil TN in the study area. The RF and RFK models used showed the expected performance and achieved acceptable TN prediction accuracy. In comparison, RFK performed slightly better than the RF model. The R2 and RMSE achieved by the RFK model were improved by 4.5% and 4.5%, respectively, compared with that by the RF model. However, the results suggest that RFK was inferior to the RF model in quantifying prediction uncertainty and thus may have a slight disadvantage in model reliability. [ABSTRACT FROM AUTHOR]

The present study was conducted to compare generalized linear model (GLM), random forest (RF), and Cubist to produce available phosphorus (AP) and potassium (AK) maps and to identify the covariates that control mineral distribution in Lorestan Province, Iran. To this end, the locations for collecting 173 soil samples were determined through the conditioned Latin hypercube sampling (cLHS) method, at four different land-uses (orchards, paddy fields, agricultural, and abandoned fields). The performance of the models was assessed by coefficient of determination (R2), root-mean-square error (RMSE), and mean absolute error (MAE) indices. The results showed that the RF model fitted better than GLM and Cubist models and could explain 40 and 57% of AP and AK distribution, respectively. The R2, RMSE, and MAE of the RF model were 0.4, 2.81, and 2.43 for predicting AP and equal to 0.57, 143.77, and 116.61 for predicting AK, respectively. The most important predictors selected by the RF model were valley depth and soil-adjusted vegetation index (SAVI) for AP and AK, respectively. The maps showed higher AP and AK content in apricot orchards compared to other land-uses. No difference was observed between AP and AK content on paddy fields, agricultural, and abandoned areas. The higher AP and AK contents were related to orchard management practices, such as failure to dispose of plant residuals and fertilizer consumption. It can be concluded that the orchards (by increasing soil quality) was the best land-use in line with sustainable management for the study area. However, generalizing the results needs more detailed research. [ABSTRACT FROM AUTHOR]

Soil pH affects food production, pollution control and ecosystem services. Mapping soil pH levels, therefore, provides policymakers with crucial information for developing sustainable soil use and management policies. In this study, we used the LUCAS 2015 TOPSOIL data to map soil pH at a European level. We used random forest kriging (RFK) to build a predictive model of spatial variability of soil pH, as well as random forest (RF) without co‐kriging and boosted regression trees (BRT) modelling techniques. Model accuracy was evaluated using a ten‐fold cross‐validation procedure. While we found that all methods accurately predicted soil pH, the accuracy of the RFK method was best with regression performance metrics of: R2 = 0.81 for pH (H2O) and pH (CaCl2); RMSE = 0.59 for pH (H2O) and RMSE = 0.61 in pH (CaCl2); MAE = 0.41 for pH (H2O) and MAE = 0.43 in pH (CaCl2). Dominant explanatory variables in the RF and BRT modelling were topography and remote sensing variables, respectively. The generated maps broadly depicted similar spatial patterns of soil pH, with an increasing gradient of soil pH from north to south Europe, with the highest values mainly concentrated along the Mediterranean coast. The mapping could provide spatial reference for soil pH assessment and dynamic monitoring. [ABSTRACT FROM AUTHOR]

Pedometricians have spent a lot of effort on mapping soil types and basic soil properties. However, end‐users typically need a more elaborate soil quality index for land management. Soil quality indices are typically derived from multiple individual soil properties by evaluating whether specific criteria are met. If this is based on individually mapped soil properties, then an important consequence is that cross‐correlations between soil properties are ignored. This makes it impossible to quantify the uncertainties associated with the mapped indices. The objective of this study was to map a soil potential multifunctionality index for agriculture (Agri‐SPMI) over a 12,125 km2 study region located along the French Mediterranean coast to help urban planners preserve soils of the highest quality. The index considered the ability of soils to fulfil four functions under five land use scenarios. A binary map represented each soil function fulfilment for a given scenario. The final soil quality index map was the sum of the 20 binary maps. A regression cokriging model was developed to map the basic soil properties first individually from legacy soil data and spatial soil covariates using a random forest algorithm, and next, interpolate the residuals using cokriging and the linear model of coregionalisation. The mapping uncertainties of soil properties were propagated by calculating the soil quality index over 300 stochastic simulations of soil properties derived from the linear models of coregionalisation. Results showed a poor prediction accuracy of the quality index, mainly because some soil properties were poorly predicted (notably available water capacity and coarse fragments) and used in combination with extreme thresholds that defined land suitability. Overall, the uncertainty was correctly quantified because the stochastic simulations reproduced the width of the observed distribution well, but the shapes of the distributions differed considerably from those of the observations. We envisage some ways for improvement, such as creating probability maps instead of the mean from simulations, and changing the prediction support from point to area. Highlights: The study mapped a soil quality index (Agri‐SPMI) to help preserve soils of highest quality along the French Mediterranean coast.The Agri‐SPMI considered the ability of soils to fulfill four functions under five land use scenarios, derived from multiple individual soil properties.A regression cokriging model was developed to map the basic soil properties, followed by interpolating the residuals using cokriging and the linear model of coregionalisation.The study accurately quantified mapping uncertainties using stochastic simulations, but the soil quality index prediction accuracy was poor, with suggestions for improvement. [ABSTRACT FROM AUTHOR]

Spatial information on land and soil resources are critical towards addressing land degradation for ensuring sustainable soil and crop management. To address these needs, digital soil mapping techniques have emerged as an efficient and low-cost solution. Although digital soil mapping has typically leveraged geospatial environmental variables (e.g. remote sensing), the application and integration of spectroscopic data with those environmental variables remain limited. Hence, this study combines visible and nearinfrared (Vis-NIR) spectroscopy, remote sensing, and topographic data and applies random forests, hybridized with particle swarm optimization algorithm (RF + PSO), to predict the spatial variability of soil clay content, electrical conductivity (EC), and calcium carbonate equivalent (CCE) for 370km² of agricultural land in western Iran. Using a conditioned Latin hypercube approach, 220 soil samples at the 0–20 cm depth increment were acquired throughout the study area. Three sets of environmental covariates were tested: Scenario A (Vis-NIR spectroscopy data), Scenario B (environmental data), and Scenario C (Vis-NIR spectroscopy þ environmental data). According to the 10-fold crossvalidation procedure with 100 replications, the RF + PSO model showed an acceptable level accuracy for all scenarios, although the accuracy of the RF + PSO model using the Scenario C data was higher than all other scenarios: the Lin’s Concordance Correlation Coefficient values were 0.77, 0.83, and 0.74 for the clay contents, EC, and CCE, respectively. The results demonstrated that the combination of Vis-NIR spectroscopic data and commonly available environmental covariates provided the best input data for the hybridized model and enhanced its performance. [ABSTRACT FROM AUTHOR]

Soil texture is an important property that controls the mobility of the water and nutrients in soil. This study examined the capability of machine learning (ML) models in estimating soil texture fractions using different combinations of remotely sensed data from Sentinel-1 (S1), Sentinel-2 (S2), and terrain-derived covariates (TDC) across two contrasting agroecological regions in Southwest Germany, Kraichgau and the Swabian Alb. Importantly, we tested the predictive power of three different ML models: the random forest (RF), the support vector machine (SVM), and extreme gradient boosting (XGB) coupled with the remote sensing data covariates. As expected, ML model performance was not consistent regarding the input covariates, soil texture fractions, and study regions. For example, in the Swabian Alb, the SVM model performed the best for the sand content with S2 + TDC (RMSE = 3.63%, R2 = 0.42), and XGB best predicted the clay content with S1 + S2 + TDC (RMSE = 6.84%, R2 = 0.64). In Kraichgau, the best models for sand (RMSE = 7.54%, R2 = 0.79) and clay contents (RMSE = 6.14%, R2 = 0.48) were obtained using XGB and SVM, respectively. Moreover, the results indicated that TDC were critical in estimating soil texture fractions, especially in Kraichgau, which indicated that topography plays an important role in defining the spatial distribution of soil properties. In contrast, the contribution of remote sensing data better predicted the silt and clay content in the Swabian Alb. The transferability of a region-specific model to the other region was low as indicated by poor predictive performance. The resulting soil-texture-fraction maps could be a significant source of information for efficient land resource management and environmental monitoring. Nonetheless, further research to evaluate the added value of the Sentinel imagery and to better analyze the spatial transferability of machine learning models is highly recommended. [ABSTRACT FROM AUTHOR]

Mapping the spatial distribution of soil organic carbon (SOC) in lands covered by tropical forests is important to understand the relationship and dynamics of SOC in this type of ecosystem. In this study, the Random Forest (RF) algorithm was used to map SOC stocks of topsoil (0–15 cm) in forest lands of the Dominican Republic. The methodology was developed using geospatial datasets available in the Google Earth Engine (GEE) platform combined with a set of 268 soil samples. Twenty environmental covariates were analyzed, including climate, topography, and vegetation. The results indicate that Model A (combining all 20 covariates) was only marginally better than Model B (combining topographic and climatic covariates), and Model C (only combining multispectral remote sensing data derived from Landsat 8 OLI images). Model A and Model B yielded SOC mean values of 110.35 and 110.87 Mg C ha−1, respectively. Model A reported the lowest prediction error and uncertainty with an R2 of 0.83, an RMSE of 35.02 Mg C ha−1. There was a strong dependence of SOC stocks on multispectral remote sensing data. Therefore, multispectral remote sensing proved accurate to map SOC stocks in forest ecosystems in the region. [ABSTRACT FROM AUTHOR]

The soil–environmental relationship identified and standardised over the years has expedited the growth of digital soil-mapping techniques; hence, various machine learning algorithms are involved in predicting soil attributes. Therefore, comparing the different machine learning algorithms is essential to provide insights into the performance of the different algorithms in predicting soil information for Indian landscapes. In this study, we compared a suite of six machine learning algorithms to predict quantitative (Cubist, decision tree, k-NN, multiple linear regression, random forest, support vector regression) and qualitative (C5.0, k-NN, multinomial logistic regression, naïve Bayes, random forest, support vector machine) soil information separately at a regional level. The soil information, including the quantitative (pH, OC, and CEC) and qualitative (order, suborder, and great group) attributes, were extracted from the legacy soil maps using stratified random sampling procedures. A total of 4479 soil observations sampled were non-spatially partitioned and intersected with 39 environmental covariate parameters. The predicted maps depicted the complex soil–environmental relationships for the study area at a 30 m spatial resolution. The comparison was facilitated based on the evaluation metrics derived from the test datasets and visual interpretations of the predicted maps. Permutation feature importance analysis was utilised as the model-agnostic interpretation tool to determine the contribution of the covariate parameters to the model's calibration. The R2 values for the pH, OC, and CEC ranged from 0.19 to 0.38; 0.04 to 0.13; and 0.14 to 0.40, whereas the RMSE values ranged from 0.75 to 0.86; 0.25 to 0.26; and 8.84 to 10.49, respectively. Irrespective of the algorithms, the overall accuracy percentages for the soil order, suborder, and great group class ranged from 31 to 67; 26 to 65; and 27 to 65, respectively. The tree-based ensemble random forest and rule-based tree models' (Cubist and C5.0) algorithms efficiently predicted the soil properties spatially. However, the efficiency of the other models can be substantially increased by advocating additional parameterisation measures. The range and scale of the quantitative soil attributes, in addition to the sampling frequency and design, greatly influenced the model's output. The comprehensive comparison of the algorithms can be utilised to support model selection and mapping at a varied scale. The derived digital soil maps will help farmers and policy makers to adopt precision information for making decisions at the farm level leading to productivity enhancements through the optimal use of nutrients and the sustainability of the agricultural ecosystem, ensuring food security. [ABSTRACT FROM AUTHOR]

The knowledge about the spatial distribution of soil organic carbon stock (SOCS) helps in sustainable land-use management and ecosystem functioning. No such study has been attempted in the complex topography and land use of Himalayas, which is associated with great spatial heterogeneity and uncertainties. Therefore, in this study digital soil mapping (DSM) was used to predict and evaluate the spatial distribution of SOCS using advanced geostatistical methods and a machine learning algorithm in the Himalayan region of Jammu and Kashmir, India. Eighty-three soil samples were collected across different land uses. Auxiliary variables (spectral indices and topographic parameters) derived from satellite data were used as predictors. Geostatistical methods—ordinary kriging (OK) and regression kriging (RK)—and a machine learning method—random forest (RF)—were used for assessing the spatial distribution and variability of SOCS with inter-comparison of models for their prediction performance. The best fit model validation criteria used were coefficient of determination (R2) and root mean square error (RMSE) with resulting maps validated by cross-validation. The SOCS concentration varied from 1.12 Mg/ha to 70.60 Mg/ha. The semivariogram analysis of OK and RK indicated moderate spatial dependence. RF (RMSE = 8.21) performed better than OK (RMSE = 15.60) and RK (RMSE = 17.73) while OK performed better than RK. Therefore, it may be concluded that RF provides better estimation and spatial variability of SOCS; however, further selection and choice of auxiliary variables and higher soil sampling density could improve the accuracy of RK prediction. [ABSTRACT FROM AUTHOR]

Despite the increasing adoption of proximal sensors worldwide, rare works have coupled proximal with remotely sensed data to spatially predict soil properties. This study evaluated the contribution of proximal and remotely sensed data to predict soil texture and available contents of micronutrients using portable X-ray fluorescence (pXRF) spectrometry, magnetic susceptibility (MS), and terrain attributes (TA) via random forest algorithm. Samples were collected in Brazil from soils with high, moderate, and low weathering degrees (Oxisols, Ultisols, Inceptisols, respectively), and analyzed by pXRF and MS and for texture and available micronutrients. Seventeen TA were generated from a digital elevation model of 12.5 m spatial resolution. Predictions were made via: (i) TA; (ii) TA + pXRF; (iii) TA + MS; (iv) TA + MS + pXRF; (v) MS + pXRF; and (vi) pXRF; and validated via root mean square error (RMSE) and coefficient of determination (R2). The best predictions were achieved by: pXRF dataset alone for available Cu (R² = 0.80) and clay (R2 = 0.67) content; MS + pXRF dataset for available Fe (R2 = 0.68) and sand (R2 = 0.69) content; TA + pXRF + MS dataset for available Mn (R2 = 0.87) content. PXRF data were key to the best predictions. Soil property maps created from these predictions supported the adoption of sustainable soil management practices. [ABSTRACT FROM AUTHOR]

Soil provides a key interface between the atmosphere and the lithosphere and plays an important role in food production, ecosystem services, and biodiversity. Recently, demands for applying machine learning (ML) methods to improve the knowledge and understanding of soil behavior have increased. While real-world datasets are inherently imbalanced, ML models overestimate the majority classes and underestimate the minority ones. The aim of this study was to investigate the effects of imbalance in training data on the performance of a random forest model (RF). The original dataset (imbalanced) included 6100 soil texture data from the surface layer of agricultural fields in northern Iran. A synthetic resampling approach using the synthetic minority oversampling technique (SMOTE) was employed to make a balanced dataset from the original data. Bioclimatic and remotely sensed data, distance, and terrain attributes were used as environmental covariates to model and map soil textural classes. Results showed that based on mean minimal depth (MMD), when imbalanced data was used, distance and annual mean precipitation were important, but when balanced data were employed, terrain attributes and remotely sensed data played a key role in predicting soil texture. Balanced data also improved the accuracies from 44% to 59% and 0.30 to 0.52 with regard to the overall accuracy and kappa values, respectively. Similar increasing trends were observed for the recall and F-scores. It is concluded that, in modeling soil texture classes using RF models through a digital soil mapping approach, data should be balanced before modeling. [ABSTRACT FROM AUTHOR]

Classification of fields into management units based on soil variability and fertility is important for spatial crop planning. The present study was conducted in Chukanagallu subwatershed (97 km2), Koppal district of Northern Karnataka Plateau, India to map the soil fertility management units and to analyse the suitability of soil for different crops. Random forest regression and classification algorithms were used to map the differentiating characteristics of soil series (soil depth, coarse fragments and soil colour), physicochemical properties (pH, EC and OC) and fertility parameters (P2O5, K2O, S, Fe, Mn, Zn, Cu, B). Random forest model performed well for the prediction of fertility parameters (R2 = 44–73%) and physicochemical properties (R2 = 39–83%) compared to soil depth and coarse fragments (R2 = 17–18%). Predicted soil fertility parameters and physicochemical properties were used for the delineation of different homogenous fertility management units. Soil series characteristics and fertility parameters were also evaluated using a multi-criteria approach for suitability of soil for cotton, groundnut and rice cultivation and the results showed that major area of subwatershed is moderately suitable for the cultivation of cotton, rice and groundnut. The management units derived from DSM approach were symmetrical in production potential and requires similar management aspects which are useful for appropriate planning of management strategies such as crop selections and nutrient management to achieve sustainable production. [ABSTRACT FROM AUTHOR]

Accurate prediction of digital soil maps allows for the evaluation of larger areas with respect to the design of efficient land management plans at the regional scale. Nowadays, there is an increasing demand for high spatial resolution‐gridded soil data for crop planning and management because it saves time and costs. One of the most essential soil physical properties affecting water holding capacity, nutrient availability and crop growth is soil texture. It exhibits a high spatial variability, but accurate maps for larger scales are lacking. The aim of this research was to produce gridded maps of soil texture fractions (clay, silt and sand) using regression‐based approaches and to establish soil texture classes using classification‐based techniques for the semi‐arid Piedmont plain of Iran. To this end, a digital elevation model and derived topographic indices, vegetation and soil‐based indices generated from 4‐year timeseries of remote sensing products of Landsat 8 OLI were used as covariates. The decision tree (linear) and its improved version of random forest (nonlinear) algorithms were used for both the regression and classification analysis. For both algorithms, the topography‐based indices and remotely sensed products were the most important predictors for the soil particle fractions. For the estimation of the different textural classes with multiple algorithms, we recorded a moderate overall accuracy rate of 54% and a Kappa coefficient of 17% for the validation datasets. It was observed that the nonlinear classification method of the random forest was more effective, and this was also the case for the regression modelling. In general, the random forest algorithm produced a more useful gridded map to help to design regional management plans based on soil properties. [ABSTRACT FROM AUTHOR]

Soil salinization and sodification, the primary causes of land degradation and desertification in arid and semi-arid regions, demand effective monitoring for sustainable land management. This study explores the utility of partial least square (PLS) latent variables (LVs) derived from visible and near-infrared (Vis-NIR) spectroscopy, combined with remote sensing (RS) and auxiliary variables, to predict electrical conductivity (EC) and sodium absorption ratio (SAR) in northern Xinjiang, China. Using 90 soil samples from the Karamay district, machine learning models (Random Forest, Support Vector Regression, Cubist) were tested in four scenarios. Modeling results showed that RS and Land use alone were unreliable predictors, but the addition of topographic attributes significantly improved the prediction accuracy for both EC and SAR. The incorporation of PLS LVs derived from Vis-NIR spectroscopy led to the highest performance by the Random Forest model for EC (CCC = 0.83, R2 = 0.80, nRMSE = 0.48, RPD = 2.12) and SAR (CCC = 0.78, R2 = 0.74, nRMSE = 0.58, RPD = 2.25). The variable importance analysis identified PLS LVs, certain topographic attributes (e.g., valley depth, elevation, channel network base level, diffuse insolation), and specific RS data (i.e., polarization index of VV + VH) as the most influential predictors in the study area. This study affirms the efficiency of Vis-NIR data for digital soil mapping, offering a cost-effective solution. In conclusion, the integration of proximal soil sensing techniques and highly relevant topographic attributes with the RF model has the potential to yield a reliable spatial model for mapping soil EC and SAR. This integrated approach allows for the delineation of hazardous zones, which in turn enables the consideration of best management practices and contributes to the reduction of the risk of degradation in salt-affected and sodicity-affected soils. • The effectiveness of a variety of auxiliary variables was assessed in predicting EC and SAR. • PLS latent variables (LVs) derived from Vis-NIR led to the highest performance for predicting EC and SAR. • Among the various ML models, Random Forest (RF) demonstrated the highest efficiency. [ABSTRACT FROM AUTHOR]

Soil texture data are the basic input parameters for many Earth System Models. As the largest middle–low altitude permafrost regions on the planet, the land surface processes on the Qinghai–Tibet Plateau can affect regional and even global water and energy cycles. However, the spatial distribution of soil texture data on the plateau is largely unavailable due to the difficulty of obtaining field data. Based on collection data from field surveys and environmental factors, we predicted the spatial distribution of clay, silt, and sand contents at a 1 km resolution, from 0–5, 5–15, 15–30, 30–60, 60–100, and 100–200 cm soil depth layers. The random forest models were constructed to predict the soil texture according to the relationships between environmental factors and soil texture data. The results showed that the soil particles of the QTP are dominated by sand, which accounts for more than 70% of the total particles. As for the spatial distribution, silt and clay contents are high in the southeast plateau, and low values of silt and clay mainly appeared in the northwest plateau. Climate and NDVI values are the most important factors that affect the spatial distribution of soil texture on the QTP. The results of this study provide the soil texture data at different depths for the whole plateau at a spatial resolution of 1 km, and the dataset can be used as an input parameter for many Earth System Models. [ABSTRACT FROM AUTHOR]

Agricultural practices affect arbuscular mycorrhizal fungal (AMF) hyphae growth and glomalin production, which is a recalcitrant carbon (C) source in soil. Since the spatial distribution of glomalin is an interesting issue for agronomists in terms of carbon sequestration, digital maps are a cost-free and useful approach. For this study, a set of 120 soil samples was collected from an experimental area of 310 km2 from the Sarab region of Iran. Soil total glomalin (TG) and easily extractable glomalin (EEG) were determined via ELISA using the monoclonal antibody 32B11. Soil organic carbon (OC) was also measured. The ratios of TG/OC and EEG/OC as the glomalin–C quotes of OC were calculated. A total of 17 terrain-related attributes were also derived from the digital elevation model (DEM) and used as static environmental covariates in digital soil mapping (DSM) using three predictive models, including multiple linear regression (MLR), random forests (RF), and Cubist (CU). The major findings were as follows: (a) DSM facilitated the interpretation of recalcitrant C source variation; (b) RF outperformed MLR and CU as models in predicting and mapping the spatial distribution of glomalin using available covariates; (c) the best accuracy in predictions was for EEG, followed by EEG/OC, TG, and TG/OC. [ABSTRACT FROM AUTHOR]

Recent increase in the frequency and spatial extent of wildfires motivates the quick recognition of the affected soil properties over large areas. Digital Soil Mapping is a valuable approach to map soil attributes based on remote sensing and field observations. We predicted the spatial distribution of post-fire litter composition in a 40,600 ha basin burned on the 2017 wildfire of Chile. Remotely sensed data of topography, vegetation structure and spectral indices (SI) were used as predictors of random forest (RF) models. Litter sampled in 60 hillslopes after the fire provided training and validation data. Predictors selected by the Variable Selection Using Random Forests (VSURF) algorithm resulted in models for litter composition with acceptable accuracy (coefficient of determination, R2 = 0.51–0.64, Normalized Root Mean Square Error, NRMSE = 16.9–22.1, percentage bias, pbias = −0.35%-0.5%). Modelled litter parameters decrease in concentration respect to the degree of burn severity, and the pre-fire biomass. Because pre-fire vegetation was conditioned by land cover and by a previous (2 years old) wildfire event, our results highlight the cumulative effect of severe wildfires in the depletion of litter composition. [ABSTRACT FROM AUTHOR]

Forest soils have a high potential to store carbon and thus mitigate climate change. The information on spatial distribution of soil organic carbon (SOC) stocks is thus very important. This study aims to analyse the importance of environmental predictors for forest SOC stock prediction at the regional and national scale in the Czech Republic. A big database of forest soil data for more than 7 000 sites was compiled from several surveys. SOC stocks were calculated from SOC content and bulk density for the topsoil mineral layer 0–30 cm. Spatial prediction models were developed separately for individual natural forest areas and for four subsets with different altitude range, using random forest method. The importance of environmental predictors in the models strongly differs between regions and altitudes. At lower altitudes, forest edaphic series and soil classes are strong predictors, while at higher altitudes the predictors related to topography become more important. The importance of soil classes depends on the pedodiversity level and on the difference in SOC stock between the classes. The contribution of forest types as predictors is limited when one (mostly coniferous) type dominates. Better prediction results can be obtained in smaller, but consistent regions, like some natural forest areas. [ABSTRACT FROM AUTHOR]

Soil organic matter (SOM) plays pivotal roles in characterizing dryland structure and function; however, remotely sensed spatially-detailed SOM mapping in these regions remains a challenge. Various digital soil mapping approaches based on either single-period remote sensing or spectral indices in other ecosystems usually produce inaccurate, poorly constrained estimates of dryland SOM. Here, a framework for spatially-detailed SOM mapping was proposed based on cross-wavelet transform (XWT) that exploits ecologically meaningful features from intra-annual fractional vegetation and soil-related endmember records. In this framework, paired green vegetation (GV) and soil-related endmembers (i.e., dark surface (DA), saline land (SA), sand land (SL)) sequences were adopted to extract 30 XWT features in temporally and spatially continuous domains of cross-wavelet spectrum. We then selected representative features as exploratory covariates for SOM mapping, integrated with four state-of-the-art machine learning approaches, i.e., ridge regression (RR), least squares-support vector machines (LS-SVM), random forests (RF), and gradient boosted regression trees (GBRT). The results reported that SOM maps from 13 coupled filtered XWT features and four machine learning approaches were consistent with soil-landscape knowledge, as evidenced by a spatially-detailed gradient from oasis to barren. This framework also presented more accurate and reliable results than arithmetically averaged features of intra-annual endmembers and existing datasets. Among the four approaches, both RF and GBRT were more appropriate in the XWT-based framework, showing superior accuracy, robustness, and lower uncertainty. The XWT synthetically characterized soil fertility from the consecutive structure of intra-annual vegetation and soil-related endmember sequences. Therefore, the proposed framework improved the understanding of SOM and land degradation neutrality, potentially leading to more sustainable management of dryland systems. [ABSTRACT FROM AUTHOR]

Soil texture is the most important soil physical property that determines water holding capacity, nutrient availability and crop growth. Spatial distribution of soil texture at a higher spatial resolution at regional and national level is essential for crop planning and management. In the present study, we mapped the soil textural classes over 16.2 M ha area of Andhra Pradesh state, India, at 250 m spatial resolution up to 2 m depth using the digital soil mapping approach. A total of 2,272 profile observations were used for the prediction of soil textural classes using the Random Forest (RF) classification algorithm. To estimate textural classes at six standard soil depth intervals (0–5, 5–15, 15–30, 30–60, 60–100 and 100–200 cm), we used continuous depth function of texture distribution using average sand, silt and clay content of different textural classes. Depth‐wise spline outputs were then transformed into textural classes as per USDA textural classification. Sixteen environmental variables including Landsat‐8 data, digital elevation model attributes and climatic variables were used for modelling. For model building, 75% of data was used and 25% of data was used for validation. Overall classification accuracy index and kappa index were calculated for validation data sets using 100 RF models. We recorded overall accuracy of 50%–65% and kappa index of 35%–47% for various depths. We found that equal‐area quadratic splines of average sand, silt and clay are useful for soil profile depth harmonization of soil textural classes and random forest classification algorithm is a promising tool for spatial prediction of texture classes at the regional level. The present high‐resolution (250 m) maps of soil texture classes are useful for different hydrological studies and preparation of proper land‐use plans. [ABSTRACT FROM AUTHOR]

The terrain covariate is one type of key environmental covariate used in digital soil mapping (DSM). Because the selection of proper terrain covariates relies largely on users' DSM knowledge and application context, automatic selection of terrain covariates is valuable for DSM users (especially non‐experts). Case‐based reasoning provides a promising solution to this demand. Recently, two case‐based reasoning strategies have been proposed (i.e., the covariate‐classification strategy and the most‐similar‐case strategy). However, there is no comparison of their performance, as well as that of the implemented methods based on the same or different strategies. This study fills this knowledge gap through a comparison experiment on the two representative methods for each strategy, respectively. Experiments adopted a DSM case base including 191 cases, which involve 38 terrain covariates. Results from a cross‐validation and a practical DSM application showed that the random forests method adopting the covariate‐classification strategy performed best. [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]

Spatial soil information is essential for informed decision-making in a wide range of fields. Digital soil mapping (DSM) using machine learning algorithms has become a popular approach for generating soil maps. DSM capitalises on the relation between environmental variables (i.e., features) and a soil property of interest. It typically needs a training dataset that covers the feature space well. Mapping in areas where there are no training data is challenging, because extrapolation in geographic space often induces extrapolation in feature space and can seriously deteriorate prediction accuracy. The objective of this study was to analyse the extrapolation effects of random forest DSM models by predicting topsoil properties (OC, clay, and pH) in four African countries using soil data from the ISRIC Africa Soil Profiles database. The study was conducted in eight experiments whereby soil data from one or three countries were used to predict in the other countries. We calculated similarities between donor and recipient areas using four measures, including soil type similarity, homosoil, dissimilarity index by area of applicability (AOA), and quantile regression forest (QRF) prediction interval width. The aim was to determine the level of agreement between these four measures and identify the method that had the strongest agreement with common validation metrics. The results indicated a positive correlation between soil type similarity, homosoil and dissimilarity index by AOA. Surprisingly, we observed a negative correlation between dissimilarity index by AOA and QRF prediction interval width. Although the cross-validation results for the trained models were acceptable, the extrapolation results were unsatisfactory, highlighting the risk of extrapolation. Using soil data from three countries instead of one increased the similarities for all measures, but it had a limited effect on improving extrapolation. Also, none of the measures had a strong correlation with the validation metrics. This was particularly disappointing for AOA and QRF, which we had expected to be strong indicators of extrapolation prediction performance. Results showed that homosoil and soil type methods had the strongest correlation with validation metrics. The results for this case study revealed limitations of using AOA and QRF as measures of extrapolation effects, highlighting the importance of not relying on these methods blindly. Further research and more case studies are needed to address the effects of extrapolation of DSM models. • Spatial extrapolation of DSM models substantially deteriorates prediction accuracy. • Weak correlation was found between the AOA dissimilarity index and DSM prediction accuracy. • Weak correlation was found between QRF prediction interval width and DSM prediction accuracy. • AOA dissimilarity index and QRF prediction interval were poorly correlated. • Soil type similarity and homosoil were better indicators of DSM extrapolation potential. [ABSTRACT FROM AUTHOR]

Decision tree algorithms, such as random forest, have become a widely adapted method for mapping soil properties in geographic space. However, implementing explicit spatial trends into these algorithms has proven problematic. Using x and y coordinates as covariates gives orthogonal artifacts in the maps, and alternative methods using distances as covariates can be inflexible and difficult to interpret. We propose instead the use of coordinates along several axes tilted at oblique angles to provide an easily interpretable method for obtaining a realistic prediction surface. We test the method on four spatial datasets and compare it to similar methods. The results show that the method provides accuracies better than or on par with the most reliable alternative methods, namely kriging and distance-based covariates. Furthermore, the proposed method is highly flexible, scalable and easily interpretable. This makes it a promising tool for mapping soil properties with complex spatial variation. [ABSTRACT FROM AUTHOR]

Digital soil mapping applies soil attributes, Remote sensing and Geomorphometrics indices to estimate soil types and properties at unobserved locations. This study carried out in order to comparison two data mining algorithms such as Random Forest (RF) and Boosting Regression tree (BRT) and two features selection principal component analysis (PCA) and variance inflation factor (VIF) for predicting soil taxonomy class at great group and subgroup levels. A total of 61 soil profile observation based on stratified random determined and digged in area with approximately 16660 hectares.19 RS indices and geomorphometrics covariates derivated from Landsate-8 imagery and DEM with 30 meters' resolution in ERDAS IMAGINE 2014 and SAGA GIS version 7.0 software's. Also to run four Data mining algorithms scenarios (PCA-RF, VIF-RF, PCA-BRT, VIF-BRT) from "Randomforest" and "C.5" packages were used in R studio software. 80% and 20% from soil profiles were applied for calibrating and validating. The results showed that in PCA and VIF approaches, eight covariates such as (Relative slope position, diffuse insolation, modified catchment, normalized height, RVI, Standard height, TWI, Valley depth) and six covariates (NDVI, DVI, Catchment area, DEM, Salinity index, Standard height) were selected. The validation results based on overall accuracy and kappa index for scenarios at great group level indicated that 88,93,62, 54 and 75,83,51,45 percentages and for subgroup level had 70, 77, 54, 47 and 60, 71, 43, 37 percentages, respectively. Generally, VIF-RF had accuracy rather than from other scenarios at two categorical level in this study area. [ABSTRACT FROM AUTHOR]

DSM has been used to meet the demand for soil information over large areas in more detailed resolution consuming less time. Soil-forming factors proxies (environmental covariates) are crucial to increase accuracy and provide new insights about the soil-landscape. DSM of the southern region of Minas Gerais state (52,982 km2) was generated. Soil profiles that comprise the legacy of decades of pedological surveys were gathered and associated with new (paleoclimatic, organic carbon, and airborne gamma-ray spectrometry) or less explored (digital terrain and climatic variables) environmental covariates under tropical conditions. An iterative routine of soil modeling, accuracy, uncertainty (entropy and MESS) assessment, and discussions concerning soil legacy was performed to equate pattern recognition with knowledge discovery from the random forest algorithm. The soil map was generated with 89% accuracy. The probability maps of the occurrence of the soil classes allowed soil-landscape insights. Less than 0.5% of the areas showed negative MESS, indicating an adequate representation of the soil samples concerning the environmental covariates. The most important environmental covariates were geophysical airborne gamma-ray spectrometry (K and eTh), distance from drainage channels, and paleoclimate (precipitation estimated from 22,000 years ago) consistent with the polygenetic soils. • Knowledge-discover routine reveals relevant environmental covariates and mapped soil. • A novel soil map for a large region was accurately generated from random forest. • Soil profiles represent well the set of environmental covariates proposed. • Geophysics, relief, and paleoclimate were the most important environmental covariates. • Paleoclimate was most important than the current climate in mapping soils. [ABSTRACT FROM AUTHOR]