Search

[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]

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]

• 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] • Spatial distribution of eight heavy metals (HMs) were studied in arid harrats soils. • Landsat 8/9 OLI single-monthly multi-year synthetic indexes were used for HMs mapping. • Stepwise multiple regression (SMLR) and random forest (RF) coupled 52 covariates used for HMs modelling. • The SMLR had better performance than RF: R 2 = 0.38–0.54;17–40, NRMSE = 18–26 %; 19–27 %, respectively. • The studied HMs derived from lithologic origins, thus considered background levels. Characterizing the background content and source(s) of heavy metals (HMs) is important for the evaluation of potential pollution in soils. Samples were collected from the topsoil layers (0.3 m) of 19 soil pedons in the harrats arid region in Saudi Arabia, and the concentrations of Co, Cr, Cu, Fe, Mn, Ni, Pb, and Zn were analyzed. Stepwise multiple linear regression (SMLR) and random forest (RF) were used to model and digitally map the studied HMs concentrations and establish their background levels using fifty-two environmental covariates (ECOVs) extracted from time series Landsat 8/9 OLI and digital elevation models. The values of the studied HMs ranged from 0.95 to 22.2 mg kg−1 (Co), 0.8 to 58.8 mg kg−1 (Cr), 1.65 to 34 mg kg−1 (Cu), 2,206 to 25,970 mg kg−1 (Fe), 33.5 to 788 mg kg−1 (Mn), 1.8 to 134.7 mg kg−1 (Ni), 2.25 to 12.5 mg kg−1 (Pb), and 2.1 to 55.9 mg kg−1 (Zn) with a coefficient of variation that ranged between 27.3 % (Pb) and 63.7 % (Ni). These are all very low compared to results from anthropogenically affected areas near the border of the harrats region, precisely mining (i.e., Mahad AD'Dahab) and industrial (i.e., Riyadh and Jubail) areas as well as to the common average ranges for these HMs in soil. The model's predictions and available observations were compared with statistics from 5-fold cross-validation with 3 repetitions. Findings revealed that mean R 2 varied between 0.38 (Zn) and 0.54 (Cu) with normalized root mean square errors (NRMSEs) between 18.53 % (Zn) and 26.03 % (Cr) for SMLR, which better than RF that yielded mean R 2 that ranged from 0.17 (Ni) to 0.40 (Cu) with NRMSEs between 19.15 % (Co) and 27.76 % (Mn). The studied HMs are considered background concentrations in the area, and therefore are important for any future environmental pollution/monitoring studies. Our study demonstrated the capacity of SMLR to use available ECOVs to predict HMs and generate background levels at a large scale in specific locations. Additionally, this study investigated the relationships between the factors affecting HMs accumulation in natural surface soil in arid regions. The methodology proposed was effective for describing HMs spatial variability and can be easily implemented in other ecosystems. [ABSTRACT FROM AUTHOR]

• SOC content in multiple soil horizons of the Heihe River Basin was mapped. • EL gave satisfactory estimation result with high accuracy and low uncertainty. • MAP showed the highest contribution to the estimation of SOC content. • SOC content of farmland in the A-horizon was higher than that in the O-horizon. Estimating the spatial distribution of soil organic carbon (SOC) content is essential for evaluating ecosystem functions and establishing global climate change models. Most of the previous SOC mapping studies have focused on estimating SOC content at fixed depths. However, soils at the same depth are not always belonging to the same soil horizon. Therefore, the characteristics of soil horizons should be considered. In this study, the SOC content of multiple soil horizons in the middle and upper reaches of the Heihe River Basin was estimated using multivariate factors and ensemble learning (EL). The results show that EL obtains the best SOC content estimation performance, where the R2 values are 0.58, 0.57 and 0.69 for the O-horizon, A-horizon, and B-horizon, respectively. Environmental factors show a high degree of contribution to estimate SOC content, and mean annual precipitation (MAP) presents the highest contribution amongst all the factors. For the middle and upper reaches of the Heihe River Basin, the SOC content is high in the southwest and low in the northeast. Amongst the multiple soil horizons, the O-horizon usually showed higher SOC content than the A-horizon and the B-horizon. But the SOC content of the A-horizon is higher than that of the O-horizon for farmland. From 2012 to 2013, the SOC content of forestland and farmland increased in the O-horizon and A-horizon. The application of EL improves the mapping performance of SOC content in different soil horizons, and the involvement of multivariable factors reveals the distribution of SOC content in the Heihe River Basin. This study creates maps of SOC content in different soil horizons and analyses the effect of land covers on SOC content, which provides insights for ecological conservation and regional sustainable development. [ABSTRACT FROM AUTHOR]

• The effect of sampling densities on accuracy of prediction model was explored. • Geostatistical models without environmental variables have higher prediction accuracy in plain areas when the sampling size is large. • Machine learning models had an advantage of prediction accuracy when sampling densities were below 5.31 samples/103 km2. • Quantitatively identified the driving factors of the spatial distribution of soil sand content in eastern China. Accurate prediction of the spatial distribution of soil sand content is a pre-requisite for land use management, soil quality evaluation and erosion control, as it determines the transport and movement of soil water, fertilizer, air and heat. Digital soil mapping (DSM) is extensively employed for predicting soil properties. However, practical research is required to address the challenge of selecting an optimal prediction model that is both cost-effective and accurate at a specific sampling density. In this study, topsoil samples were collected from 2,848 sampling points in the eastern plains of China (107,200 km2). The performance of different prediction models for mapping soil sand content was compared at 12 levels of sampling density. Moreover, the geographical detector, a statistical method used to assess the spatial stratified heterogeneity of variables, was adopted to determine the major drivers of spatial variation in soil sand content. The results indicated that climate factors are the major drivers of the spatial variability in soil sand content. For the 100% sample size (26.57 samples/103 km2), the geostatistical models that did not depend on environmental variables (ordinary kriging, sequential Gaussian simulation) performed best, followed by the machine learning models (random forest, cubist and support vector machine) and the geostatistical model with environmental variables (co-kriging). Sampling density had a considerable impact on model accuracy, and the advantages of machine learning models became apparent when sampling densities were below 20% (5.31 samples/103 km2). Therefore, the best combination of prediction model and sampling density should be selected to obtain maps of soil sand content economically and accurately. This study provides a valuable reference for the selection of prediction methods in the practical application of DSM. [ABSTRACT FROM AUTHOR]

[Display omitted] • Soil chemical ripening was identified following Urmia Lake desiccation. • The spatial distribution of geochemical indices was successfully mapped using satellite imageries. • The geochemical indices demonstrate the transition from geogenesis to pedogenesis. • Ba/Sr emerged as the best index for detecting soil chemical ripening. • Digital mapping of the ripening process assists in interpreting pedogenesis. Urmia Lake in Iran has been gradually drying up since the 2000 s, revealing soil that was once underwater. There is limited research on how this drying process affects the distribution of soil chemical ripening in the lakebed sediments. To investigate this, 131 surface soil samples were collected in July 2017 from six zones based on their distance from the lake's shoreline. Elemental concentrations in the samples were measured using a portable X-ray fluorescence (XRF) spectrometer. Five geochemical indices (GCIs) were calculated: Titanium/Zirconium, Barium/Strontium, Silicon/Aluminium, Silicon/R 2 O 3 , and the Darmody index. A remote sensing image (Landsat-8 OLI) from the same sampling period was used as environmental covariates to explain the spatial distribution of GCIs. Random forests (RF) and cubist tree models were employed to model and map these GCIs. The bootstrapping method was used to assess the uncertainty of the maps. Results indicated that the RF model performed better than the cubist model based on the concordance values in the validation dataset. Remotely sensed clay minerals ratio and normalized burn ratio were identified as the most significant covariates in making predictions. The study found that the distribution of all geochemical indices was directly related to their distance from the shoreline. Ba/Sr emerged as the best index for detecting soil chemical ripening in the study area. The index displayed the most significant differences between the drying zones. This research demonstrated that combining satellite imagery with GCIs using the digital soil mapping (DSM) technique can effectively detect soil ripening progression. [ABSTRACT FROM AUTHOR]

• DSM of SOC using different remote sensing data was reviewed. • Majority of studies were conducted on mixed land-use types followed by cropland. • Machine learning techniques were used more than (geo)statistical ones. • Terrain variables as predictors had the largest contribution to map SOC. • Data acquired by satellite sensors was employed more than the other types. Soil organic carbon (SOC) has attracted a lot of attention in the soil science community. Freely available remote sensing data combined with advanced digital soil mapping (DSM) techniques has led to a better understanding and management of SOC. This paper has considered the published literature with a focus on digital mapping of SOC using remote sensing data within 2010 to 2023 intervals. The objective was to consider all the important aspects of SOC prediction and mapping, including different land-use types, DSM algorithms, environmental variables, and remote sensing data sources. According to this review conducted on the 217 papers, cropland was the most popular type of land use. Regarding the DSM algorithms, random forest (RF) appeared in the largest number of studies. The terrain and spectral variables derived from the digital elevation model (DEM) and remote sensing images, were the highest demanding among all those used as input predictors. In addition, satellite platforms provided the largest portion of the remote sensing data used for the calibration of DSM models. This review provides quantitative insight into recent trends of SOC digital mapping using remote sensing technology while suggesting some directions for future development of the topic. [ABSTRACT FROM AUTHOR]

• Soil properties predicted using SEM in arid region of Iran. • SEM allows including expert knowledge in modelling soil properties. • SEM yields interpretable models but lacks predictive power. • SEM provides insights into soil-environment relationships. • Strong surface–subsurface soil relationships found. Predicting soil properties and evaluating their functions along with their related driving factors is useful for providing useful geographical information for soil management, which is especially important in arid and semi-arid regions. This study investigates the use of a structural equation modeling (SEM) approach for assessing the effects of soil forming factors through environ-mental proxies on three key soil properties, namely soil organic carbon (SOC), calcium carbonate equivalent (CCE) and clay content (clay) in an arid and semi-arid region of Iran. Using a set of 259 soil profiles collected over years 2016–2020 in the Qazvin plain, the cause-effect relationships were estimated between these soil properties and nine environmental factors derived from a digital elevation model and from satellite images. Focusing on two main horizons A and B, it was shown that normalized difference vegetation index, midslope position, elevation, multi-resolution valley bottom flatness, and saga wetness index are impacting these soil properties. Inside each horizon, the effect of CCE and clay on SOC was also evidenced, but to an extent that depends on the horizon. For each soil property, we were able to clearly identify the relationships between the two horizons. Although our SEM approach proved to be useful for identifying and estimating the cause-effect relationships, it failed to provide a good predictive model as required for a relevant digital soil mapping of these soil properties. However, as the SEM approach allows combining soil science knowledge inside a model that accounts for soil forming factors, external factors, and soil system at the same time, it permits an investigation of their potential cause-effect relationships in a rich theoretical framework. The SEM methodology is thus potentially useful for soil scientists that are studying various soil properties in other parts of the world, even if it cannot be advocated as an efficient digital soil mapping method in general. [ABSTRACT FROM AUTHOR]

• DSOLMap is a new high-resolution global digital soil property map for the SWAT + model. • Soil data play a crucial role in the hydrological cycle and its modelling. • An impact on streamflow performance using different global soil maps was observed. • The use of DSOLMap allows for an improved daily flow simulation. • DSOLMap can be useful for hydrological modelling in regions with limited soil data. This research paper addresses the ongoing challenge of developing fine-resolution global digital soil property maps for hydrological modelling applications. Hydrological models are essential for understanding watershed dynamics and the impact of human activities on water resources. Soil data, which plays a crucial role in the hydrological cycle, is a requisite model input. Global digital soil property maps usually have coarse spatial resolutions, adding considerable uncertainty to hydrological models despite calibration efforts. To address this issue, a new global digital soil property map with 250 m spatial resolution, known as Digital Soil Open Land Map (DSOLMap), was developed and evaluated in this study. The DSOLMap has a finer spatial resolution than existing global soil maps and a more detailed soil profile divided into six soil horizons. This new high-resolution global digital soil property map was tailored to the SWAT + model format. SWAT + is the latest released version of the Soil and Water Assessment Tool (SWAT), one of the most comprehensive hydrological models, and is widely used worldwide. A hydrological evaluation was conducted with the DSOLMap and its results were compared to two other global soil databases using the SWAT + model in a basin located in the north of Spain. The findings showed that using more detailed, finer-resolution soil data, such as those that the DSOLMap offers, improved the hydrological performance of the SWAT + model on a daily scale before and after calibration and validation procedures. The DSOLMap represents a global step forward in hydrological modelling, notably for regions with scarce or unavailable soil information. This new digital soil property map can help decision-makers address global challenges related to water resources and environmental issues through hydrological modelling. [ABSTRACT FROM AUTHOR]

• We evaluated the potential of using a combination of multitemporal bare soil and crop growth images for SOM prediction. • Remote sensing images during the bare soil period are more suitable for SOM mapping. • The accuracy of SOM mapping can be improved by combining images with appropriate periods. • The use of appropriate multitemporal crop growth images combined with environmental covariates has great potential in SOM mapping. Mapping of soil organic matter (SOM) in cultivated land is one of the important aspects of digital soil mapping, and its results are of great significance for agricultural precision management and carbon cycle assessment. This study takes Youyi Farm, a typical black soil area in Northeast China, as the research area and utilizes all Landsat-8 images covering the study area from April to October during 2014–2022. After masking clouds, all images were synthesized monthly. According to the local crop phenology, the period from April to October was divided into the bare soil period (April to June), peak crop growth period (July to August), and late crop growth period (September to October). Based on the random forest regression algorithm, differences in the accuracy of SOM mapping using synthetic images from different periods were evaluated, and the impact of adding environmental covariates on the SOM mapping accuracy was analyzed. The results showed that (1) when using a single-temporal synthesized image for SOM mapping, the order of accuracy was bare soil period > peak crop growth period > late crop growth period, with the synthesized image in May exhibiting the highest accuracy, with an RMSE of 0.979 %; (2) when using a multitemporal image combination for SOM mapping, the combination of optimal months (April, May, June) in the same periods can obtain the best SOM mapping accuracy, with an RMSE of 0.919 %; and (3) adding environmental covariates can effectively improve the accuracy of SOM mapping, especially when using the growing season remote sensing images, RMSE decreased from 1.136 % to 0.909 %. This study expands the applicable areas and conditions of SOM remote sensing mapping. [ABSTRACT FROM AUTHOR]

Abstract A better understanding of the spatial pattern of soil organic matter (SOM) is important for scientific soil management. As multisource secondary data become increasingly cheap and readily available, numerous methods have been established to incorporate secondary information; however, these methods exhibit limitations under certain conditions due to their relatively strict requirements on secondary data. In this study, we tried to integrate sampled soil data and secondary data more effectively within the framework of Bayesian maximum entropy (BME). Specifically, multiple linear regression (MLR) and geographically weighted regression (GWR) were run 100 times based on environmental covariates such as terrain indices, vegetation indices and categorical variables obtained from soil maps. Then, the 95% confidence interval was derived from the multiple prediction values at each of the soft data points. For comparison, some conventional techniques, including ordinary kriging (OK), regression kriging (RK) and geographically weighted regression kriging (GWRK), were also applied. The results showed that BME exhibited a prediction accuracy comparable to that of OK and maintained the prediction uncertainty at a low level, while other studied methods (MLR, GWR, RK and GWRK) were all significantly inferior to BME and OK. The proposed methodology in this study represents a promising scenario for the digital soil mapping, especially when the relationships between the target soil attributes and various secondary information are not strong or residuals of trend models show insignificant spatial autocorrelation. Highlights • Prediction confidence intervals of regression models can be used as soft data. • Incorporation of localized soft data optimizes estimation accuracy and uncertainty. • Bayesian maximum entropy with localized soft data is feasible in soil modeling. [ABSTRACT FROM AUTHOR]

Abstract Soil mappers have traditionally relied on tacit knowledge and qualitative assessment of soil-landscape relationships to obtain the physiographic context necessary to predict soil distribution and spatial patterns. This assessment implicitly utilizes a nested hierarchical approach based on differences in the phenomenon scale of soil forming factors where climate, landscape, parent material, and topography are examined in sequence to create a model of soil-landscape relationships. Our objective was to predict parent material distribution using expert knowledge paired with quantitative digital terrain attributes as part of a nested approach to digital soil mapping. The study took place at an 890-hectare research farm in Logan County, Arkansas, which is part of the Arkansas River Valley. Two major groups of parent material are identifiable in the Arkansas River Valley: residual sandstone and shale on erosional uplands, and silty/clayey pedisediment in depositional areas. A 5-m digital elevation model was used to derive thirteen terrain attributes for the study site. Three of the terrain attributes, namely topographic position index, multi-resolution valley bottom flatness, and vertical distance to channel network, were utilized as part of a rule-based approach to model parent material distribution based on preliminary reconnaissance and expert knowledge of the area. The model was validated by sampling 20 locations using a conditioned Latin hypercube sampling (cLHS) design to evaluate the prediction accuracy. Seventy-five percent of cLHS samples were accurately predicted to be residuum or pedisediment. The resulting map also had 90% agreement with the National Cooperative Soil Survey map; however, the digital map was able to provide more spatially explicit information especially on inclusions. Incorporating parent material distribution as part of a nested hierarchical approach to digital soil mapping aids in constraining and predicting soil properties, enables a more straightforward examination of physiographic context and can ultimately lead to more accurate digital soil maps for land management. Highlights • A rule-based model was capable of predicting parent material distribution. • Overall mapping accuracy of parent material was 75%. • There was 90% agreement of parent material compared with a traditional soil map. [ABSTRACT FROM AUTHOR]

Abstract Soil legacy data is ubiquitous and usually contains routine soil analysis information. In Iran, like most places, legacy soil data constitutes genetic horizon soil information recorded from excavated soil profiles. Describing and sampling from each genetic horizon is assumed to be heterogeneous from site to site. Digital soil mapping (DSM) using observed data is valuable because it provides a means to exploit the available information together with leveraging commonly available information by way of environmental covariates. It creates a much more detailed view of soil at the landscape scale. The purpose of this paper is to model and map the spatial distribution of nitrogen, phosphorous and boron at four standardized depths: 0–15, 15–30, 30–60, 60–100 cm, in an area of 7300 ha in the north west of Iran, and compare different model types. To circumvent the issue of heterogeneous soil depth observations from site to site, mass-preserving soil depth function splines were used to harmonise the soil profile observed data to the aforementioned standard depths. This facilitated the spatial modelling of each of the target variables for each standard depth with the aim of creating digital soil maps. Twenty-three covariates were extracted from a publically available digital elevation model (DEM) as well as freely available Landsat 8 ETM+ imagery. The DEM-derivative covariates used in this study were divided into three main categories: i) Morphometry; ii) hydrology; and iii) lighting visibility. Both Random Forest and Cubist were assessed as candidate models for predicting each target variable. The results showed that Cubist was the most accurate method. Terrain attributes play an important role in estimating N, P, and B, while optical images do not have significant role. The most important findings of this paper in terms of environmental hazards are that the inundated regions in the west part of the study area are susceptible to boron contamination, providing future guidance for remediation. Highlights • Digital Soil Mapping was used to assess vertical distribution of soil nutrients. • DEM derivatives and Landsat imagery indices relate to soil nutrients. • Cubist models outperformed Random Forest for Nitrogen, Phosphorous and Boron. • Soil model uncertainty is greatest at lower soil depths. [ABSTRACT FROM AUTHOR]

Abstract The soil organic carbon (SOC) and total nitrogen (TN) of topsoil could be greatly affected by the conversion of grassland to cropland (CGC) due to vegetation harvesting and soil erosion. However, few studies have evaluated the effects of these changes in the subsoil and distinguished its main controls at the regional scale. This paper investigated the total changes in the SOC stock (SOC S) and TN stock (TN S) of soils from 0 to 30 cm and 0–100 cm in depth over the past three decades in the Songnen Plain, China, a typical Mollisol region. The CGC resulted in a moderate loss of topsoil C and N but an increase in subsoil C and N, and the soil mapping results showed that approximately 7.47 × 1011 kg C and 1.51 × 1011 kg N were accumulated during the past three decades. Prediction uncertainty was assessed by 100 model runs with random jackknife partitions, and high uncertainty was found in the areas with rapidly changing SOC S and TN S levels. Our findings indicated that subsoil should be considered in the estimation of the SOC S and TN S at the regional scale. The SOC S and TN S of the CGC areas were anticipated to increase after long-term cultivation. In view of ongoing cropland expansion, up-to-date land use and soil type data are vital for selecting monitoring sites and understanding long-term soil evolution at the regional scale. Highlights • Conversion of grassland to cropland led to a moderate loss of topsoil C and N, but an increase in C and N in subsoil. • Subsoil should be considered for the estimation of SOC and TN stocks at the regional scale. • The SOC and TN stocks of recent cultivated soils were anticipated to increase after a long term cultivation. [ABSTRACT FROM AUTHOR]

Quantitative estimation for soil erosion is necessary for protection of the environment, and to improve agricultural productivity. However, due to the large area, sparse and limited data in Tibet, soil erosion there is still poorly quantified. Here, we improved the factors of the Revised Universal Soil Loss Equation (RUSLE) and calculated water erosion in Tibet. Rainfall erosivity ( R ) was calculated with the 0.25°CPC Morphing technique (CMORPH) data and subsequently downscaled to 1-km spatial resolution using artificial neural network (ANN) based on environmental covariates; slope length and steepness ( LS ) was estimated using the 3 arc sec Shuttle Rader Topography Mission (SRTM) digital elevation model (DEM); cover management ( C ) and control practice ( P ) were assigned based on land cover and protection measurements; and soil erodibility ( K ) was calculated using the Environmental Policy Integrated Climate model (EPIC) with inputs of the contents of sand, silt, clay and organic carbon in soil samples from Tibet. We used the data-mining algorithm to model the K factor and the spatially referenced variables to generate a K factor map. The obtained factors were then used to calculate soil loss in Tibet at1-km resolution. Our study estimated the annual water erosion at5.43 t ha −1  y −1 in Tibet, about5.44 × 10 8  t of soil lost yearly. The erosion rate increased from northwest to southeast, with most serious erosion occurring in the humid rain forest area in the southeast of Tibet. Our estimates of erosion area were generally consistent with previous national estimates. The largest differences were in the humid zone, Hengduan Mountain, and Yarlung Zangbo River basin, which are characterized by complex terrain and climate. Because of the applications of the best available data, we supply better, quantitative, finer spatial resolution estimates than previous studies. Our study is valuable for assessment of soil erosion in other data-scarce area suffering from soil loss by water erosion. [ABSTRACT FROM AUTHOR]

• A spatio-temporal prediction framework was developed to map SOC. • The inter-decadal maps of China SOC were updated to the 2010s. • Effects of climatic factors for SOC variability are location- and scale- specific. • The soils in China function as carbon sinks at national scale from 1980s to 2010s. Spatiotemporal variability in soil organic carbon (SOC) plays key roles in ecosystem stability and terrestrial carbon cycling. However, the inter-decadal characteristics of SOC variability and its climate drivers in China from the 1980s to 2010s remain unclear. In this study, a modeling framework for combining spatiotemporal prediction models was proposed. Within this model, soil samples from each of the two periods were combined and calibrated to obtain three models, and their prediction outputs were selectively combined to obtain SOC maps for the different inter-decadal periods. A total of 8473 soil samples from three periods (1980s, 2000s and 2010s) were used for model validation. The results suggested that the SOC prediction accuracy in the top 0–20 cm of soil (SOC 020) was better than that of the first meter of soil (SOC 0100), and the predicted spatial patterns shared characteristics with findings from previous studies. Predictions for the 1990s exhibited relatively high uncertainty, especially in the Tibetan Plateau, Tien Shan, and Northeast China. Both individual and coupled climatic factors exerted scale- and location-specific effects on the SOC. Coupled climatic factors accounted for more than 65% of the SOC variability in both soil layers. The regional SOC in North and Northeast China exhibited inter-decadal decline. Soil in China functioned as carbon sinks from the 1980s to 2010s at a national scale, with SOC 020 increasing from 30.40 Pg to 33.97 Pg and SOC 0100 from 79.29 Pg to 82.26 Pg. Overall, our results provide reliable baseline data for optimizing carbon cycle models of terrestrial ecosystems, and to guide future carbon sequestration policies. [ABSTRACT FROM AUTHOR]

[Display omitted] • Soil organic carbon (SOC) stocks predicted using machine learning algorithms. • Spatial resolutions of environmental variables examined on national and global data. • Source of data (national or global) defines the importance of environmental variables. • Independent reference data showed a significant decrease in prediction accuracy. • DSM should be re-evaluated by local data when spatial data differ from the original data. We evaluated how the spatial resolution of environmental variables (n = 47) altered their ability to predict soil organic carbon (SOC) stocks (0–30 cm depth) using training data from Gridded Soil Survey Geographic-gSSURGO and SoilGrids databases. Training and validation subsamples (1,629) were selected using a conditioned Latin hypercube sampling (cLHS) design based on environmental variables in Vermont, U.S. The predictive relationships between environmental variables and SOC stock (t C ha−1) were developed using machine learning algorithms. The algorithms were trained (70 %) and evaluated (30 %) using a random subset of database subsamples, respectively, with an additional evaluation step using local, independent SOC reference data (n = 272). The Random Forest (RF) algorithm outperformed other algorithms at all spatial resolutions in estimating SOC stocks. As spatial resolution increased, model performance with the gSSURGO database increased (R2 = 0.33–0.62 and RMSE = 42.42–34.92), while no such trend was observed for the SoilGrids database. The best SOC stock model prediction using the SoilGrids database was achieved with a 10 m resolution (R2 = 0.54 and RMSE = 4.67). Evaluation of modeled results using the external, or independent, reference data showed a significant decrease compared to the internal validation in prediction accuracy (R2 = 0.11–0.14 for gSSURGO and, R2 = −0.19 for SoilGrids). The gSSURGO database showed that soil maps (including suborders, drainage classes, temperature, and moisture) and geology/landform maps had a greater influence than other environmental variables at all spatial resolution scales. In contrast, climatic- and DEM-related variables were more significant for the SoilGrids database. Our study suggested that the origin of the SOC stock database and the sampling scheme largely affects the importance of environmental variables assigned in the machine learning algorithm. Our results confirmed that the variable and data sources, model type, and combination of environmental variables significantly influenced prediction accuracy. In conclusion, DSM products should be re-evaluated with local references when used for spatial extents that are different from those for which they were initially designed. [ABSTRACT FROM AUTHOR]

• The challenges current approaches facing is discussed. • Using urban functional types to reflect the effect of human activities on urban heavy metal accumulation. • Combing the advantages of receptor models and GeoDetector to select key environmental covariates and tracing pollution sources. • Extending the transferability of the models is urgent to promote the application of the models. Heavy metal pollution in urban soil is one of the global environmental problems, and monitoring its spatial distribution can provide basic data for assessing and managing the ecological and environmental problems. Due to the strong human activities in the urban space and the high heterogeneity and fragmentation of the landscape, it is of great significance to explore how to characterize the diversity in human activities within the city and to establish a paradigm for mapping spatial distribution of heavy metal in urban soils. This paper reviewed the theory of soil-landscape model, the data sources for extracting environmental variables, and the geo-statistical methods for establishing regression models. Furthermore, receptor models for tracing heavy metals were discussed. We suggested that using multi-source heterogeneous geographic data (geological data, soil type data, meteorological data, terrain data, time series Landsat images) to extract environmental covariates. In particular, integrating social perception data (point of interest, Tencent user data) and remote sensing image data to extract the urban functional district types that reflect differences in human activities. GeoDetector is recommended to select key environmental covariates that spatially coupling with urban soil heavy metals, and combined with receptor models to trace the sources of heavy metals. Machine learning algorithm is a good choice to construct the estimation models by cooperating with key environmental variables, and to verify the general transferability of the models by using the local weighted and 'spiking' modeling strategies. The newly introduced methods and ideas may be expected to provide technical methods for the monitoring of heavy metal pollution in urban soils, and to further expand the application of digital soil mapping theories and methods in urban soil mapping. [ABSTRACT FROM AUTHOR]

Precise mapping of the spatial distribution of total soil nitrogen (TSN) is essential for soil resource management, agronomic sustainability, and nitrogen sequestration potential. Present study used three random forest (RF) and three multiple linear stepwise regressions (MLSR) models to map the distribution of topsoil (0–20 cm) TSN content in Lushun City in China's northeastern Liaoning Province. A total of 12 covariates (including topography, climate, and remote sensing images) and 115 soil samples were employed. An independent validation set of 23 samples was used to verify the performance of the model based on mean absolute prediction error (MAE), root mean square error (RMSE), coefficient of determination ( R 2 ), and Lin's concordance correlation coefficient (LCCC). Accuracy assessments showed that the RF model, combined with all environmental variables, had the best prediction performance. The second best was produced by the use of remote sensing alone. The third was the model that used only topographic and climatic variables alone. Remote sensing was not significantly inferior to the use of the model with all variables and can be used to build a realistic model. In the model with all covariates, the distribution of TSN is mainly explained by remote sensing, followed by topography variables, and climatic variables; their relative importance (RI) was 48%, 36%, and 16%, respectively. The results of remote sensing on the robust dependence of TSN should provide a link to dense natural vegetation in this forested environment. Additionally, remote sensing and its derived environmental variables should be used as the main predictors when mapping TSN in forested areas and other areas with similar vegetation coverage. [ABSTRACT FROM AUTHOR]

Multi-scale geospatial and absolute variation of surface and near-surface soil physical and chemical properties can be mapped and quantified by coupling digital soil mapping techniques with high resolution remote sensing products. The goal of this research was to advance data-driven digital soil mapping techniques by developing an approach that can integrate multi-scale digital surface topography and reflectance-derived remote sensing products, and characterize multi-scale soil-landscape relations of Quaternary alluvial and eolian deposits. The study area spanned the arid landscape encompassed by the Barry M. Goldwater Range West (BMGRW), which is administered by the Marine Corps Air Station Yuma, in southwestern Arizona, USA. An iterative principal component analysis (iPCA) was implemented for LiDAR elevation- and Landsat ETM + -derived soil predictors, termed environmental covariates. Principal components that characterize >95% of covariate space variability were then integrated and classified using an ISODATA (Iterative Self-Organizing Data) unsupervised technique. The classified map was further segmented into polygons based on a region growing algorithm, yielding multi-scale maps of soil-landscape relations that were compared with maps of soil landforms identified from aerial photographs, satellite images and field observation. The approach identified and mapped the spatial variability of soil-landscape relationships in alluvial and eolian deposits and illustrated the applicability of coupling covariate selection and integration by iPCA, ISODATA classification of integrated data layers, and image segmentation for effective spatial prediction of soil–landscape characteristics. The approach developed here is data-driven, applicable for multi-scale mapping, allows incorporation of a wide variety of covariates, and maps spatially homogenous soil-landscape units that are necessary for hydrologic models, land and ecosystem management decisions, and hazard assessment. [ABSTRACT FROM AUTHOR]

Traditional soil maps have helped us to better understand soil, to form our concepts and to teach and transfer our ideas about it, and so they have been used for many purposes. Although, soil maps are available in many countries, there is a need for them to be updated because they are often deficient in that their spatial delineations and their descriptions are subjective and lack assessments of uncertainty. Updating them is a priority for federal soil surveys worldwide as well as for research, teaching and communication. New data from sensors and quantitative ‘digital’ methods provide us with the tools to do so. Here, we present an approach to update large scale, national soil maps with data derived from a combination of traditional soil profile classifications, classifications made with visible–near infrared (vis–NIR) spectroscopy, and digital soil class mapping (DSM). Our results present an update of the Australian Soil Classification (ASC) orders map. The overall error rate of the DSM model, tested on an independent validation set, was 55.6%, and a few of the orders were poorly classified. We discuss the possible reasons for these errors, but argue that compared to the previous ASC maps, our classification was derived objectively, using currently best available data sets and methods, the classification model was interpretable in terms of the factors of soil formation, the modelling produced a 1 × 1 km resolution soil map with estimates of spatial uncertainty for each soil order and our map has no artefacts at state and territory borders. [ABSTRACT FROM AUTHOR]

Mapping soil nutrients can help smallholder farmers identify soil nutrient status and implement site-specific soil management schemes. In the past, Digital Soil Mapping has seldom been utilized to guide soil nutrient management in smallholder farm settings in South India. The objective of this research was to analyze the spatial resolution effects of different remote sensing images on soil total nitrogen (TN) prediction models in two smallholder villages, Kothapally and Masuti in South India. Regression kriging (RK) was used to characterize the spatial pattern of TN in the topsoil (0–15 cm) by incorporating spectral indices with different spatial resolutions. The results suggested that soil moisture, vegetation, and soil crusts can contribute to the conservation of soil TN in both study areas. Soil prediction models with different spatial resolutions showed a similar spatial pattern of soil TN. The results also demonstrated that the effect of very fine spatial remote sensing spectral data inputs does not always lead to an increase of soil prediction model performance. A RapidEye-based (5 m) soil TN prediction model had lower prediction accuracy than a Landsat 8-based (30 m) soil TN prediction model in Masuti. WorldView-2/GeoEye-1/Pleiades-1A-based (2 m) soil TN prediction models had the highest prediction accuracy in both study areas. The spectral indices based on new bands of WorldView-2 such as coastal, yellow, red edge, and new near infrared bands had relatively strong correlations with soil TN. The utilization of Very High Spatial resolution images such as WorldView-2 in Digital Soil Mapping could improve soil model performance and spatial characterization. Remote sensing-based soil prediction models have high potential to be widely applied in smallholder farm settings. [ABSTRACT FROM AUTHOR]

Nowadays there is an increasing effort for raising global awareness for the importance of soils to ensure food security, to improve agricultural and environmental planning and monitoring, and to establish effective and sustainable land management policies to counteract soil degradation. This study presents the INFOSOLO legacy database as the first effort to develop a soil information system in Portugal, suitable to compile soil data produced in the country, and to support stakeholders and land managers in decision-making. The database currently includes soil data from a set of 9934 horizons/layers studied in 3461 soil profiles across the country between 1966 and 2014. Data was extracted from scattered soil surveys, research projects, and academic studies carried out by public Portuguese and other European institutions, with a series of validation tests and harmonization procedures being implemented in order to access and improve the quality of the data. The importance of the INFOSOLO legacy dataset was discussed and exemplified with the analysis of the spatial and temporal distribution of selected soil properties, namely the organic carbon content, pH, and cation exchange capacity. For these properties, 1 km grid maps were also developed for the topsoil horizons/layers in Portugal using different spatial modelling approaches. To highlight the importance of using INFOSOLO for soil characterization, the EU-wide soil database LUCAS was used to compare both datasets in terms of data distributions, spatial continuity and accuracy of modelling outputs. The comparison also included the digital soil maps provided by the SoilGrids product for Portugal. The comparison results highlighted specific areas in the country for which INFOSOLO is capable to deliver accurate but also more reliable predictions when compared with the LUCAS and SoilGrids results. Thus, INFOSOLO provides the basis for improving soil information in the country and for raising national awareness of the importance of soil resources to the country's development. [ABSTRACT FROM AUTHOR]

• It was assumed that soils covered by forest can be considered as saturated in SOC. • Cubist-based PTF was developed for predicting and mapping saturated SOC content. • Environmental conditions are also important factors in predicting SOC saturation. • Significant SOC deficit was revealed on large part of Hungary with high variability. • Highest SOC deficits occur in areas with medium to high SOC content. Stakeholders and policymakers have been becoming more and more interested not just in the potential organic carbon (SOC) saturation level of soils but also in spatially explicit information on the degree of SOC deficit, which can support future policy and sustainable management strategies, and carbon sequestration-associated spatial planning. Thus the objective of our study was to develop a cubist-based pedotransfer function (PTF) for predicting and mapping the saturated SOC content of the topsoils (0–30 cm) in Hungary and then compare the resulting map with the actual SOC map to determine and assess the degree of SOC deficit. It was assumed that topsoils covered by permanent forests can be practically considered as saturated in SOC. Using the monitoring points of the Hungarian Soil Information and Monitoring System located in forests as reference soil profiles, we developed a cubist-based PTF. The transparent model structure provided by cubist allowed to show that not just the physicochemical properties of soils (e.g., texture, and pH) but also environmental conditions, such as topography (e.g., slope, altitude, and topographical position) and climate (e.g., long-term mean annual temperature, and evaporation), characterizing landscape are important factors in predicting the level of SOC saturation. Our results also pointed out that there is SOC deficit on large part of the country (∼80%) showing high spatial variability. It was also revealed that the most considerable potential for additional SOC sequestration can be found related to soils with medium to high actual SOC content. [ABSTRACT FROM AUTHOR]

• Digital mapping of soil clay types was performed using easily accessible covariates. • Multiple linear regression and random forest algorithms were compared. • Random forest resulted in higher prediction performance. • Topography and soil properties mainly affected palygorskite and kaolinite variation. • Soil properties were considered responsible for the spatial distribution of illite. Understanding the abundance variability of clay minerals, as fundamental soil components, will help the users to improve land management and address concerns over climate change and soil fertility. Therefore, this investigation aimed to model the abundance and spatial distribution of clay types, including palygorskite, illite, and kaolinite, and identify the most significant variables affecting their variability using a digital soil mapping (DSM) approach in Darab district, southern Iran. Multiple Linear Regression (MLR) and Random Forest (RF) techniques were applied to link clay types and environmental attributes that were obtained from a Landsat-8 operational land imager (OLI) and digital elevation model (DEM). A ten-fold cross-validation approach was applied to calibrate and validate the models, and 50 bootstrap models were used to quantify the prediction uncertainty. The models accuracy was defined by the coefficient of determination (R2), root mean square error (RMSE), and the ratio of performance to the interquartile range (RPIQ). Findings denoted that the RF model better predicts the abundance and variability of clay minerals in the study area (R2 = 0.56, 0.47, and 0.48, RMSE = 5.3, 1.91 and 0.63 % and RPIQ = 2.82, 3.28 and 2.62 for palygorskite, illite and kaolinite, respectively). Based on the feature selection analysis, topographic covariates and soil properties determined palygorskite and kaolinite content variations, while for illite, only soil properties could explain the spatial distribution. Besides, the RF produced a lower uncertainty for palygorskite compared to the other clay types. The present research can provide new insight into the spatial variability of clay minerals in arid and semi-arid regions of Iran that could be extended to other similar environments. Moreover, the results showed that the easily available environmental variables could provide reliable predictions. However, other environmental covariates, such as XRF analysis, Vis-NIR, and MIR spectroscopy, are also recommended as input variables for further studies. [ABSTRACT FROM AUTHOR]

• Remote sensing data and machine learning helped to assessing soil depth. • Terrain attributes were identified as important variables in predicting soil depth. • The RF model performed better than the SVM to predict soil depth. • Evaluation of Thermal region in images from different seasons was crucial to detect soil depth. • Field traditional soil depth maps validated the automatic ones. Soil depth is one of the most critical factors which impact on culture productivity and makes difficult appropriate management decisions. However, assessing this parameter is also the most challenging tasks in the agronomic field. The objective of this work was to predict the spatial distribution of soil depth from space techniques (remote sensing, RS) and machine learning. A total of 292 sites were allocated (based on the toposequence approach) and drilled (0–2 m depth) at three different locations in Brazil. Based on these, in-situ traditional depth maps (denominated field-map) were elaborated for validation. Afterwards, we created a strategy to achieve these different depths by RS (RS) approach. Landsat 8 OLI bands, Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI) and emissivity in dry and rainy seasons as well as terrain attributes were applied to predict soil depth. For this purpose, the most important covariates were selected using Recursive Feature Selection (RFE) based on Random Forest (RF) and Support Vector Machine (SVM). Afterwards, the application of RF and SVM by selected covariates were compared based on tenfold cross validation for each location. The best model was selected based on R2, RMSE and MAE, accuracy and bootstrapping approach and uncertainties. Terrain attributes were important to discriminate soil depth. Although, LST and NDVI also presented important contribution to this task. Different seasons implies on water and plant dynamics in deep and shallow soils. This impacted on NDVI and LST as detected by RS. Thus, the method brings more variables to infer soil depth. The RF model performed better than SVM to predict soil depth with an average of 0.77 R2. The accuracy between a digital soil mapping and a field-map reached 0.58 to 0.81 indicating an important result considering the difficulty of the objective This may help pedologists and farmers as well as water and plants environmental monitoring. [ABSTRACT FROM AUTHOR]

There is an increasing concern on how human activities could affect the ecosystems of the Galapagos Islands, a hot-spot for biodiversity recognized worldwide. Despite the high number of studies related to the ecological relationships between species in these islands, almost no research has been conducted on the description of their soils, a basic component of the ecosystems of Galapagos. In 1962, Belgian researchers from Ghent University organized a geo-pedological mission in order to collect detailed information on soil diversity on the Santa Cruz Island. Here we use this legacy data in a Digital Soil Mapping framework in order to quantify the spatial distribution of Soil Organic Carbon (SOC) stocks in this area. Our results indicate that the soils of Santa Cruz store about 706 Gg SOC in the upper 10 cm. SOC accumulation is mainly driven by climatic factors, which are highly influenced by both altitude and the direction of predominant winds. An increase in the amount of rainfall, as predicted by climate change scenarios, will result in an overall increase of the SOC stocks and likely modify the vegetation species composition within the different bioclimatic strata of the islands. A SOC monitoring program, based on spectroscopic analyses could be used to determine temporal variations in SOC stocks in a quick and cost-effective manner while minimizing human disturbances in this area. [ABSTRACT FROM AUTHOR]

Mountain soils play an essential role in ecosystem management. Assessment of land potentials can provide detailed spatial information particularly concerning nutrient availability. Spatial distributions of topsoil carbon, nitrogen and available phosphorus in mountain regions were identified using supervised learning methods, and a functional landscape analysis was performed in order to determine the spatial soil fertility pattern for the Soyang Lake watershed in South Korea. Specific research aims were (1) to identify important predictors; (2) to develop digital soil maps; (3) to assess land potentials using digital soil maps. Soil profiles and samples were collected by conditioned Latin Hypercube Sampling considering operational field constraints such as accessibility and no-go areas contaminated by landmines as well as budget limitations. Terrain parameters and different vegetation indices were derived for the covariates. We compared a generalized additive model (GAM) to random forest (RF) and support vector regression (SVR). For the predictor selection, we used the recursive feature elimination (RFE). A land potential assessment for soil nutrients was conducted using trimmed k-mean cluster analysis. Results suggested that vegetation indices have powerful abilities to predict soil nutrients. Using selected predictors via RFE improved prediction results. RF showed the best performance. Cluster analysis identified four land potential classes: fertile, medium and low fertile with an additional class dominated by high phosphorus and low carbon and nitrogen contents due to human impact. This study provides an effective approach to map land potentials for mountain ecosystem management. [ABSTRACT FROM AUTHOR]

Digital soil mapping (DSM) products represent estimates of spatially distributed soil properties. These estimations comprise an element of uncertainty that is not evenly distributed over the area covered by DSM. If we quantify the uncertainty spatially explicit, this information can be used to improve the quality of DSM by optimizing the sampling design. This study follows a DSM approach using a Random Forest regression model, legacy soil samples, and terrain covariates to estimate topsoil silt and clay contents in a small catchment of 4.2 km 2 in the Three Gorges Reservoir Area, Central China. We aim (i) to introduce a method to derive spatial uncertainty, and (ii) to improve the initial DSM approaches by additional sampling that is guided by the spatial uncertainty. The proposed uncertainty measure is based on multiple realizations of individual and randomized decision tree models. We used the spatial uncertainty of the initial DSM approaches to stratify the study area and thereby to identify potential sampling areas of high uncertainties. Further, we tested how precisely available legacy samples cover the variability of the covariates within each potential sampling area to define the final sampling area and to apply a purposive sampling design. For the final Random Forest model calibration, we combined the legacy sample set with the additional samples. This uncertainty-driven DSM refinement was evaluated by comparing it to a second approach. In this second approach, the additional samples were replaced by a random sample set of the same size, obtained from the entire study area. For the comparative analysis, external, bootstrap-, and cross-validation was applied. The DSM approach using the uncertainty-driven refinement performed best. The averaged spatial uncertainty was reduced by 31% for silt and by 27% for clay compared to the initial DSM approach. Using external validation, the accuracy increased by the same proportions, while showing an overall accuracy of R 2 = 0.59 for silt and R 2 = 0.56 for clay. [ABSTRACT FROM AUTHOR]

Digital soil mapping (DSM) relies on statistical relationships between soil properties and covariates (e.g. terrain attributes, land use class, geology) which may not explain a large proportion of measured soil properties variability. This uncertainty combined with low spatial resolution of existing maps makes it difficult to monitor soils. Diffuse reflectance infrared Fourier transform mid-infrared spectroscopy (midDRIFTS) with partial least square regression (PLSR) may offer an alternative source of high quality data for improved DSM. Previously validated midDRIFTS-PLSR models were used to predict soil total carbon (TC), total inorganic carbon (TIC), total organic carbon (TOC) and texture of 1170 samples from contrasting agroecological regions (~ 3600 km 2 ), Kraichgau (K) and Swabian Alb (SA), Southwest Germany. MidDRIFTS-PLSR predictions were integrated with geostatistics for soil property maps (200 m resolution). An average of 93% of the total soil samples were predicted within the confidence intervals of the midDRIFTS-PLSR models for the respective properties. Ordinary kriging (OK) resulted in maps for TC, TIC, TOC and texture with a root mean square standardized error (RMSSE) ~ 1. Soil organic matter (SOM) (DRIFTS_SOM) and texture (DRIFTS_texture) maps developed in the current study were of higher spatial resolution than previously existing maps. The DRIFTS_SOM and DRIFTS_texture in the both regions showed considerable differences when compared to existing maps. The DRIFTS_SOM in K and SA regions had overlaps of 45 and 69% with the 1:200,000 existing map. While the DRIFTS_texture in K had an overlap of 92% with the 1:1,000,000 existing map, the overlap in SA region was only 11%. We conclude that traditional DSM with covariates can be improved via higher sampling density which is made possible using midDRIFTS-PLSR. Incorporation of mid-infrared spectral data with both remote sensing and other environmental data would be a further application to cope with uncertainty associated to both spectroscopic and spatial modeling. [ABSTRACT FROM AUTHOR]

Relationships between parent material and soil are not well understood and generally only reported in qualitative form. We present a classification of parent material for pedologic purposes, which includes twelve lithology classes based on mineralogical and chemical composition. The relationships of these lithology classes with six key soil properties (soil organic carbon, pH, cation exchange capacity, sum of bases, total P and clay %) were examined in a case study over New South Wales, Australia. We used multiple linear regression, Random Forest and Cubist tree models based on a soil dataset of over 3200 points. Semi-quantitative estimates are derived of change in these soil properties with a change in lithology class, and an associated silica index , for example, a 22% relative decrease in soil organic carbon with each 10% rise in silica, broadly equivalent to a change from shale to granite, assuming other factors remain constant. Parent material covariates are essential for the effective modelling and mapping of soil properties. Widely available lithology data have the potential for greater use in digital soil modelling and mapping (DSMM) programs. We compared the performance of the classified lithology data with other continuous, geophysical parent material covariates such as gamma radiometrics in digital soil models and maps over NSW. The lithology covariate was demonstrated to exert the greatest influence on all six soil properties, coming well ahead of all geophysical parent material and other environmental covariates. Validation statistics demonstrated strong improvement in both model and map quality when the lithology covariate was included. For example, Lin's concordance for the Cubist sum of bases model rose from 0.46 with no parent material covariates to 0.58 with the continuous geophysical covariates to a high of 0.77 when lithology was also used. The improvement was typically slightly less marked in the final digital maps than for the calibration models, probably due to the lower reliability of the lithology grid derived from broad scale polygonal geological and soil data. A process is suggested for the application of lithology data into DSMM programs. Despite the potential drawbacks of using polygonal data, properly organised categorical lithology data can be a strong covariate to complement other continuous geophysical data sources in DSMM programs, particularly where reliable and fine scale geological and soil data are available. [ABSTRACT FROM AUTHOR]

The beginning of the 20th century through the 1970s were good times for soil science in the USA, with relatively strong funding and overall growth in the profession. However, the soil science discipline in the USA hit hard times in the 1980s and 1990s. Federal funding for soil survey work began to decline as did student numbers in university programs and membership in the Soil Science Society of America (SSSA). Despite this, there were still many positive advances within soil science in the USA during these two decades. There was an increased use of geophysical instrumentation, remote sensing, geographic information systems (GIS), and global positioning systems (GPS), and research began in digital soil mapping, all of which lead to better understanding of the spatial distribution and variability of soils. Many NRCS soil products were put online, making them widely available to the general public, the use of soil knowledge was expanded into new areas such as archaeology and environmental work, and historic connections to geology were re-established. While expansion into new areas required soil science to evolve as a field, separating the discipline to an extent from its agricultural roots, it also helped reinvigorate the discipline. As we move through the early parts of the 21st century, student numbers are increasing in university soil science programs and membership in SSSA is at an all-time high. Digital soil mapping is being incorporated into the National Cooperative Soil Survey, and the impact of humans on the soil system is being fully recognized. The importance of soils is being recognized by events such as the United Nations declaration of 2015 as the “International Year of Soils”. The expansion of soils into new areas and widening recognition of the importance of soils gives the field hope for a bright future in the USA. [ABSTRACT FROM AUTHOR]

• BSSIs are more suitable for digital soil mapping in bare soil areas than VIs. • DI, RI, NDI, RNDI and DRI are the best BSSIs for SOCD prediction. • Airborne hyperspectral images are important data for SOCD mapping. • DNN model offers a beneficial method for high-precision digital soil mapping. Soil organic carbon density (SOCD) is an important parameter of agricultural soils and is useful for the improvement of environment and agricultural production. Proximal and remote sensing techniques are effective methods for digital soil mapping of SOCD. The current study used three types of spectral data, including laboratory proximal spectra, airborne hyperspectral and Sentinel 2 multispectral images, to predict SOCD in an agricultural land. Bare soil spectral indices (BSSIs) were developed to predict SOCD and compared with the published vegetation indices (VIs). With 45 soil samples, the partial least square regression (PLSR), back propagation neural network (BPNN) and deep neural network (DNN) prediction models were established to map SOCD. The results showed that the laboratory spectra (R2 = 0.70–0.80) had the best performance of SOCD prediction, followed by the airborne (R2 = 0.43–0.81) and Sentinel 2 (R2 = 0.14–0.57) spectra. The SOCD maps derived from airborne and Sentinel 2 images had similar spatial distribution trends. The BSSIs (R2 = 0.24–0.81) obtained higher accuracy than the VIs (R2 = 0.14–0.74) in SOCD prediction. Moreover, DNN model was the best for digital soil mapping of SOCD among three prediction models. This study offered an effective approach for mapping SOCD in bare soil areas. [ABSTRACT FROM AUTHOR]

[Display omitted] • Ensemble machine learning increseased accuracy of NPK balance models. • NPK stocks, soil erosion, MAP, DEM are very important predictors of NPK models. • The R2 of NPK balance predictions were 62%, 58%, and 58%, respectively. • Mean annual balance of NPK per hectare was −33.2 kg N, +2.70 kg P, and −74.6 kg K. Soil nutrient depletion is one of the major causes of high yield gaps and nutrient deficiencies in East Africa highlands, including Rwanda. This research sought to determine the current soil nutrient balance and its spatial variation in 10 Rwandan agro-ecological zones. Soil nitrogen (N), phosphorus (P) and potassium (K) depletion in croplands were calculated using data from 455 field trials of the Optimizing Fertilizer Recommendations in Africa (OFRA) project in Rwanda. Calculated soil nutrient balances (NPK) and 15 environmental covariates were used to calibrate soil nutrient depletion models using ensemble machine learning (EML) and 10-fold cross-validation. In the 2019–2020 growing season, annual N and K depletions were 33.6 kg N ha−1 yr−1 and 71.0 kg K ha−1 yr−1, with a positive P balance of 2.30 kg P ha−1 yr−1. High soil nutrient uptake and high soil nutrient loss due to erosion and leaching were two main causes of NPK depletion. Spatial variations of NPK balance were influenced by soil nutrient stocks, soil erosion, elevation, rainfall, soil texture, and soil bulk density. The 10-fold cross-validation showed that coefficients of determination (R2) of NPK models were 62%, 58%, and 58%, respectively. Compared to single models, ensemble machine learning improved NPK model accuracy up to 5%. Our research revealed that soil nutrient depletion was highest in the northwest and lowest in the southeast of the study area. We conclude that increasing soil nutrient inputs without reducing soil nutrient loss due to soil degradation will not decrease soil nutrient depletion in Rwanda and ensemble machine learning outperforms single models in predicting soil nutrient balance. The solution to reduce high soil nutrient depletion in all agro-ecological zones of Rwanda would be to prioritize soil and water conservation measures and increase soil nutrient inputs. [ABSTRACT FROM AUTHOR]

Predicting soil property distribution from a catena in the digital environment has been explored by many researchers with only slightly better than modest results. In this study, the landform recognition algorithm “geomorphons” in the GRASS GIS environment was explored to determine if this landscape model could improve predictions of soil properties. For 74 borings on the Wabash glacial moraine in Wells County, Indiana, measurements were made for: A horizon thickness, depth to chroma 2 features, effervescence, dense glacial till, carbonate concretions, and autochthonous platy structure. A digital elevation model (DEM) generated from light detection and ranging (LiDAR) data was used for the study site. The geomorphons algorithm was used to generate 10 original landforms: “flat”, “footslope”, “summit”, “ridge”, “shoulder”, “spur”, “slope”, “hollow”, “valley”, and “depression” that were aggregated to new landforms coinciding with slope positions: “toeslope”, “footslope”, “backslope”, “shoulder”, “summit”, and “depression” recognized by soil surveyors. Linear Discriminant Analysis (LDA) and Multinomial Logistics Regression Analysis (MLR) were used to aggregate the measured soil properties into the landform groups. The aggregation of geomorphons groups improved the MRL predictions to 83% accuracy. Also, the aggregation of geomorphons to five landforms to predict soil property distribution on the landscape gave promising results for the low-relief and relatively flat area of northeast Indiana. To test if the true mean value of each soil property for each landform was reliable for generalizing population characteristics, relative standard error (RSE) was calculated as a proportion of standard error to population mean from a bootstrap estimation. The range of RSE values for all soil properties and landforms was between ~ 0.7% and ~ 19%. Since the estimates of the measured soil properties all have RSE values of less than 25%, they can be considered reliable. [ABSTRACT FROM AUTHOR]

Soil texture is an essential and extremely variable physical property that strongly influences many other soil properties that are highly relevant for agricultural production, e.g., fertility and water retention capacity. In plain areas, terrain properties derived from a digital elevation model are not effective for digital soil mapping, and the variation in the properties of such areas remains a challenge. In this regard, remote sensing can facilitate the mapping of soil properties. The purpose of this study was to evaluate the efficiency of using of data obtained from the Thematic Mapper (TM) sensor of Landsat 5 for digital soil mapping in a semi-arid region, based on multiple linear regression (MLR) and a random forest model (RFM). To this end, 399 samples of the soil surface layer (0–20 cm) were used to predict the sand, silt and clay contents, using the bands 1, 2, 3, 4, 5 and 7, the Normalized Difference Vegetation Index (NDVI), the grain size index (GSI), and the relationships between bands 3 and 2, bands 3 and 7, and bands 5 and 7 (clay index) of the Landsat 5 TM sensor as covariates. Among these covariates, only band 1 (b1), the relationship between bands 5 and 7 (b5/b7) for sand, silt and clay, and band 4 (b4) for silt were not significantly correlated according to Pearson's correlation analysis. The validation of the models showed that the properties were best estimated using the RFM, which explained 63% and 56% of the spatial variability of sand and clay, respectively, whereas the MLR explained 30% of the spatial variation of silt. An analysis of the relevance of the variables predicted by the RFM showed that the covariates b3/b7, b5, NDVI and b2 explained most of the variability of the considered properties. The RFM proved to be more advantageous than the MLR with respect to insensitivity to overfitting and the presence of noise in the data. In addition, the RFM produced more realistic distribution maps of the soil properties than did the MLR, taking into account that the estimated values of the soil attributes were in the same range as the calibration data, while the MLR model estimates were out of the range of the calibration data. [ABSTRACT FROM AUTHOR]

[Display omitted] • GPR data can be used for indirect mapping of the spatial structure of forest soils. • Forest soils under diverse conditions show differences in GPR cross-sections. • Soil properties are described by applying GPR signal attributes analysis. • Soil organic carbon influences the GPR signal amplitude. This article focused on studying the landscape-scale spatial variation of boreal forest soils in data-scarce areas using ground-penetrating radar (GPR). An important goal of the research was to discuss correlations between attributes of the soil profile under boreal forest and GPR data. We detected the soil horizons of Entic Podzols and determined their attributes based on a GPR transect with a 400 MHz antenna unit. GPR cross-section showed the soil thickness variation over the transect being 18.8 cm to 44.7 cm. The mean depth for the soil's lower boundary from sounding was 23.1 cm versus the probing of 24 cm. The fieldwork research has revealed GPR patterns for different ecological settings: typical forest, paludified forest, and felling area. We investigated the GPR applicability for delineating organic layers in soils and considered quantifying soil organic carbon content (SOC) relying on GPR attributes. To achieve this purpose, we have surveyed Albic Podzols with a 1,700 MHz antenna unit considering vegetation cover. The resolution of GPR was ±1 cm, which made to determine relatively thin horizons. GPR reflected signal amplitudes demonstrated sensitivity to SOC variations in the organic layers. The obtained relationship between SOC and GPR attributes indicates the possibility of determining the carbon storage variations in situ. However, the problem is to substantiate the mechanism of the SOC influence on the signature of the GPR traces. The findings prove that GPR is a tool for predicting the state and scope of the transformation of sandy forest soils. GPR cross-sections can be the basement for a novel approach for forest soil mapping on a landscape scale. In prospect, digital soil mapping can include a sequence of methods: remote sensing – GPR – manual probing. Admittedly, alignments between soil characteristics and GPR attributes are needed to identify the capabilities and limitations of the method. [ABSTRACT FROM AUTHOR]

Many areas of developing countries are covered by legacy soil surveys, which, however are hardly used, as they are not available in digital form, used outdated standards, and have unknown quality. There have been very few attempts to rescue and renew these surveys, nor are there established criteria for the evaluation of their quality. We therefore decided to test the applicability of the Cornell Adequacy Criteria (CAC) to assess the quality of several renewed soil surveys in or near the Limpopo National Park, Mozambique (centroid: 23° 18′ 55.57″ S, 31° 55′ 16.24″ E), using the concepts of digital soil mapping. The quality was assessed for mapping and monitoring soil organic carbon (SOC), in terms of geodetic control, positional accuracy, map scale, and texture and adequacy of map legend. Metadata was attached to the renewed maps. SOC stocks were estimated qualitatively based on the description of the map units and quantitatively by the measure-and-multiply approach from legacy laboratory measurements. The positional accuracy of georegistration was 13 to 45% of the square root of a Minimum Legible Area (MLA). Point and area-class layers could be created with high positional accuracy. However the index of maximum reduction was high, indicating that the original publication scale could be reduced. Map unit definitions and overall information content of the surveys were adequate. Integration of remotely sensed optical imagery and digital elevation models could be used to derive accurate contours, against which the positional accuracy of contour-based map borders was assessed. Less than 30% of their lengths were within a distance equal to the square root of MLA. These sources could not be used to evaluate internal map borders, due to the subdued topography and major land-use changes since the original survey. Qualitative estimates of SOC are between low and medium, consistent with other studies in this area. The CAC proved to be a useful framework for determining the fitness for use of legacy surveys. [ABSTRACT FROM AUTHOR]

• Generally mapping accuracy increases with sample sizes. • The general trend of mapping accuracy is often disrupted by randomness in sampling. • Variation of mapping accuracy is collectively affected by many mapping elements. • Relationship of mapping accuracy with mapping elements could be unique in an area. The accuracy of digital soil mapping (DSM) is related to many elements of a soil survey. However, most past studies commonly focus on one or two elements of DSM related to sample sizes and modeling methods, ignoring the effect of the soil-forming environment. For instance, different landform types can cause striking soil variations. As a result, studies often present different or even contradictory relationships between sample sizes and mapping accuracy. This study evaluated the accuracy of DSM in predicting soil organic carbon content due to the interaction of sample sizes, modelling methods and two landform types (i.e., hill-mountains and plains), based on 1861 soil samples from an existing soil survey conducted in the Guangdong province of China (about 180 000 km2). The study evaluated six modeling methods, including multiple linear regression, ordinary kriging, regression kriging, quantile regression forest, geographical weighting regression and radial basis function. The sample sizes varied from 100 to 1300 with an interval of 100. Results show that generally mapping accuracy increased with the sample sizes. However, this relationship was often disrupted by randomness in sampling procedure. Further, the different modeling methods or different landform types could lead to further variation, while the general relationship did not change. Thus, this study demonstrates that variation of mapping accuracy was collectively affected by many elements of DSM: sample sizes along with randomness in sampling, modelling methods and landform types, and differed between different accuracy metrics. The collective effect is the reason for different or contradictory relationships of mapping accuracy reported in the literature, where only one or two elements were investigated. For the collective effect, the relationship of mapping accuracy with elements of DSM could be unique in an area. [ABSTRACT FROM AUTHOR]