Digital Soil Mapping Research Articles

Proximal and remote sensing – what makes the best farm digital soil maps?

Context Digital soil maps (DSM) across large areas have an inability to capture soil variation at within-fields despite being at fine spatial resolutions. In addition, creating field-extent soil maps is relatively rare, largely due to cost. Aims To overcome these limitations by creating soil maps across multiple fields/farms and assessing the value of different remote sensing (RS) and on-the-go proximal (PS) datasets to do this. Methods The value of different RS and on-the-go PS data was tested individually, and in combination for mapping three different topsoil and subsoil properties (organic carbon, clay, and pH) for three cropping farms across Australia using DSM techniques. Key results Using both PS and RS data layers created the best predictions. Using RS data only generally led to better predictions than PS data only, likely because soil variation is driven by a number of factors, and there is a larger suite of RS variables that represent these. Despite this, PS gamma radiometrics potassium was the most widely used variable in the PS and RS scenario. The RS variables based on satellite imagery (NDVI and bare earth) were important predictors for many models, demonstrating that imagery of crops and bare soil represent variation in soil well. Conclusions The results demonstrate the value of combining both PS and RS data layers together to map agronomically important topsoil and subsoil properties at fine spatial resolutions across diverse cropping farms. Implications Growers that invest in implementing this could then use these products to inform important decisions regarding management of soil and crops.

Digital Soil Mapping (DSM) Using a GIS-Based RF Machine Learning Model: The Case of Strandzha Mountains (Thrace Peninsula, Türkiye)

Digital Soil Mapping (DSM) Using a GIS-Based RF Machine Learning Model: The Case of Strandzha Mountains (Thrace Peninsula, Türkiye)

Space-time mapping of soil organic carbon stock and its local drivers: Potential for use in carbon accounting

The measurement and tracking of changes in soil organic carbon (SOC) stocks and its underpinning drivers are important in greenhouse gas accounting and carbon trading schemes. While repeated direct measurement of a location over time is the ‘gold standard’ for SOC stock change estimation, this is currently unfeasible at scale due to sample collection and analysis costs. Here, we used a data-driven, space–time digital soil mapping modelling approach (ST-DSM) coupled with an interpretive machine learning (IML) technique to both predict and quantify the relative contributions of covariates to annual SOC stocks and their uncertainty at 0–30 cm depth and at a 90 m spatial resolution for Australia between 1990 and 2018. A quantile regression forest model was calibrated using data from a datacube comprising of SOC stock values collected over the prediction years and a suite of dynamic and static environmental covariates. The evaluation of the ST-DSM model on a test set was satisfactory. Long-term annual mean (Mg C/ha), and total (Pg C) topsoil SOC stock were 37.6 (12.3 – 101.1, 90 % PI) and 28.3 (9.25 – 83.1, 90 % PI) respectively. The mean annual rate of SOC stock change over the Australian continent was 0.11 (±0.22) Mg C/ha yr−1 and the total net change was 23.5 Tg C yr−1, indicating Australia’s soils were a net sink of SOC over the 29-year period. However, SOC stocks declined in the montane ecoregion, a great concern given its fragility and high SOC stock. The long-term mean fraction of photosynthetic vegetation was the most important individual local contributor to SOC stock prediction in many parts of Australia. The aggregated effect of covariates based on the soil forming factors indicates vegetation as the principal local driver of SOC stock across most of the Australian continent, followed by climate. The evaluation of our results against independent, temporally paired field data and estimates from a nationally accepted SOC process-based model yielded promising results. Our approach demonstrates the use of readily available and inexpensive covariates to estimate SOC stock changes, as well as the underlying drivers over a larger area but at a spatial resolution suitable for decision making. This approach presents a new monitoring and verification possibility for carbon projects and national carbon accounting especially with the growing number of existing profile observations in soil databases.

Spatial Evaluation of Land and Soil Properties with Geography Information Systems (GIS): The Case Study from Meriç District of Thrace Region in Türkiye

This research was carried out within the scope of spatial evaluation of the land and soil properties of Edirne-Meriç district, located in the Thrace region (Türkiye) by using GIS. Arc GIS 10.3.1 software was used in the classification of soil and land features. Digital soil maps (1/25.000 scale) were used to determine soil and land properties. Alos Palsar (12.5m) satellite images were used to determine land elevation and aspect distributions. As a result of the research, it was seen that the soil class with the largest area in terms of large soil groups in Meriç district is lime-free brown forest soils (261.2 km2). III. It was determined that class lands (153.7 km2) cover the largest area. In general, it was observed that the soil depth was greater than 150 cm (261.9 km2). It was determined that 23.3 km2 of the Meriç district lands were exposed to severe water erosion. It was determined that the study area consists of lands with a steep slope of 12-20% (126.7 km2). It has been observed that the height distribution of Meriç district lands varies between 4.7-120.5 m. It was defined that the majority of the lands were in the southwestern direction group. It is thought that the soil and land information database created as a result of the research will make significant contributions to researchers and the public, institutions and organizations.

Diffuse reflectance spectroscopy and digital soil mapping for assessing soil-associated off-road vehicle mobility risk

Soil attributes such as granulometric fractions and Atterberg limits (LL: liquid limit, PL: plastic limit, and PI: plasticity index) are needed to assess off-road vehicle mobility (OVM) risks. Parameters describing these attributes are generally measured in soil samples collected from a few locations through cumbersome laboratory methods. Although diffuse reflectance spectroscopy (DRS) can rapidly yield estimates for soil attributes in samples collected from specific locations and digital soil mapping (DSM) can transform such discrete measurements into spatially-continuous inference systems, these two technologies are rarely used for assessing OVM risks. In this study, we combined the DRS and DSM approaches for deriving spatially-continuous estimates for the key vehicle mobility parameters (gravel, sand, and fine particles; Cu: coefficients of uniformity; Cc: coefficient of curvature; LL; and PI) and classified soils using the Unified Soil Classification System (USCS). A total of 204 soil samples were collected from the north-eastern Himalayan state of Sikkim for measuring these parameters along with spectral reflectance over the visible and near-infrared region. Results of the chemometric models in the DRS approach showed that the USCS parameters may be estimated with the coefficient of determination (R2) values as high as 0.72. The fine (<2 mm diameter) fraction spectra provided the best estimates for the Atterberg limits while a combination of spectra collected from fine and coarse (>2 mm diameter) fractions was effective in estimating other granulometric fractions except for sand, which was best estimated using the coarse fraction spectra. With the DSM approach allowing effective mapping of these parameters, a spatially-continuous framework to quantify soil-associated OVM risks was developed for Sikkim for the first time.

Geospatial prediction of total soil carbon in European agricultural land based on deep learning

Accurate geospatial prediction of soil parameters provides a basis for large-scale digital soil mapping, making efficient use of the expensive and time-consuming process of field soil sampling. To date, few studies have used deep learning for geospatial prediction of soil parameters, but there is evidence that it may provide higher accuracy compared to machine learning methods. To address this research gap, this study proposed a deep neural network (DNN) for geospatial prediction of total soil carbon (TC) in European agricultural land and compared it with the eight most commonly used machine learning methods based on studies indexed in the Web of Science Core Collection. A total of 6209 preprocessed soil samples from the Geochemical mapping of agricultural and grazing land soil (GEMAS) dataset in heterogeneous agricultural areas covering 4,899,602 km2 in Europe were used. Prediction was performed based on 96 environmental covariates from climate and remote sensing sources, with extensive comprehensive hyperparameter tuning for all evaluated methods. DNN outperformed all evaluated machine learning methods (R2 = 0.663, RMSE = 9.595, MAE = 5.565), followed by Quantile Random Forest (QRF) (R2 = 0.635, RMSE = 25.993, MAE = 22.081). The ability of DNN to accurately predict small TC values and thus produce relatively low absolute residuals was a major reason for the higher prediction accuracy compared to machine learning methods. Climate parameters were the main factors in the achieved prediction accuracy, with 23 of the 25 environmental covariates with the highest variable importance being climate or land surface temperature parameters. These results demonstrate the superiority of DNN over machine learning methods for TC prediction, while highlighting the need for more recent soil sampling to assess the impact of climate change on TC content in European agricultural land.

High-Accuracy Mapping of Soil Parent Material Types in Hilly Areas at the County Scale Using Machine Learning Algorithms

Conventional maps of soil parent material (SPM) types obtained by field survey and manual mapping or predictions from other map data have limited accuracy. Digital soil mapping of SPM types necessitates accurate acquisition of SPM distribution information, which is still a challenge in hilly areas. This study developed a high-accuracy method for SPM identification in hilly areas at the county scale. Based on geographic information system technology, seven feature variables were extracted from the geological map, geomorphic map, digital elevation model, and remote sensing image data of Shanggao County, Jiangxi Province, China. Different feature combination schemes were designed to develop SPM identification models based on random forest (RF), support vector machine (SVM), and maximum likelihood classification (MLC) algorithms. The best SPM identification results were obtained from the RF algorithm using the combination of geological type, geomorphic type, elevation, and slope. Confusion matrices were constructed based on a field survey of 586 validation samples, and the results were evaluated in terms of overall accuracy, precision, recall, F1 score, and Kappa coefficient. The overall accuracy and Kappa coefficient of the results from the optimal RF model were 83.11% and 0.79, respectively, which were 26.11% and 0.31 higher than those of the conventional map, respectively. Its precision and recall for various SPM types were greater than 75%. A comprehensive comparison of the accuracy, uncertainty, and plotting performance of the SPM recognition results reveals that the RF algorithm outperforms the SVM algorithm and the MLC algorithm. Geological type was the largest contributor to SPM identification, followed by geomorphic type, elevation, and slope. The importance of different feature variables varied for distinct SPM types. The accuracy of SPM identification was not improved by selecting more feature variables, such as land use type, normalised difference vegetation index, and topographic wetness index. This study demonstrates the feasibility of high-accuracy county-level SPM mapping in hilly areas based on the RF algorithm using geological type, geomorphic type, elevation, and slope as feature variables. As hilly areas have typical topographic features and SPM types, the proposed method of SPM mapping can be useful for application in other similar areas. There are a few limitations in this study with regard to data quality and resolution, feature variable selection, classification algorithm generalisation, and study area representativeness, which may affect the outcomes and need to be solved.

Evaluating the extrapolation potential of random forest digital soil mapping

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.

Enhancing the accuracy of digital soil mapping using the surface and subsurface soil characteristics as continuous diagnostic layers.

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.

SOLA INCOGNITA: UNSOLVED PEDOGENETIC CHALLENGES

For the anniversary of the Faculty of Soil Science of Moscow State University, the author examines the issues that have remained unresolved in the field of classical genetic soil science over the past half century, and offers a list of tasks that need to be solved in the near future. It is noted that against the background of decreased interest in pedogenetic research, problems are accumulating in related fields that require a clear understanding of the genesis of soils: soil classification, digital soil mapping, paleopedology and soil assessment. One of the unsolved issues remains the estimation of the rates of soilforming processes, without which it is dificult both to reconstruct the evolution of soils in the past and to predict the behavior of soil systems in the future. The mechanisms of textural differentiation have not been deciphered both in temperate regions and in the tropics; in many cases, soils clearly differentiated by the content of the clay fraction are adjacent to soils with a profile homogeneous in texture. The mechanisms of deep illuviation of humus are unknown in the absence of signs of spodic or natric properties. There is no clear understanding of the mechanisms of formation of compacted subsurface horizons of soils (fragipan) and horizons cemented with opal (duripan). It is proposed to create a bank of unresolved soilgenetic issues that require analysis and solution

Synergistic use of proximally sensed and time series remotely sensed imagery to map soil sodicity

Soil sodicity can cause crusting in the topsoil which impedes water infiltration, and erosion in the subsoil owing to instability. These can impose constraints to sugarcane growth in the form of reduced plant emergence and root extension. To improve soil capability, a map of topsoil (0–0.15 m) exchangeable sodium percentage (ESP) can be used to determine the amount of gypsum required to ameliorate soil condition. Herein, a digital soil mapping (DSM) approach was used to map top- and subsoil (0.3–0.45 m) ESP. To do this, we first examined prediction agreement of a calibration model between digital data and ESP, first comparing individual use of remotely sensed time series imagery (i.e., Sentinel-2 and Landsat-8) or proximal sensor (i.e., electromagnetic [EM] induction and gamma-ray [γ-ray]) data. To account for the large amount of remote data we combined them using principal component analysis (PCA). We also investigated combining digital data using a backward feature elimination – BFE), along with minimum calibration sample size (n = 150–10) determined. Prediction agreement (Lin's concordance correlation coefficient – LCCC) and accuracy (ratio of performance to deviation – RPD) were assessed using an independent validation dataset (n = 32). To predict topsoil ESP, PCs from Landsat-8 provided excellent (LCCC = 0.90) agreement and very good (RPD = 2.26) accuracy, followed by Sentintel-2 (0.86, 1.75), γ-ray (0.83, 1.84) and EM (0.78, 1.78), with subsoil ESP also excellent (0.90) and very good (2.51) using Landsat-8, however; Sentintel-2 outperformed the others with even better results (0.95, 3.24), followed by EM (0.78, 1.76) and γ-ray (0.73, 1.54). Moreover, slightly better results were achieved when EM, γ-ray, and PCs of Sentinel-2 were combined, with top- and subsoil ESP agreement (0.93 and 0.94, respectively) and accuracy (2.53 and 3.14, respectively) excellent. A minimum of 40 sampling sites were required (1.2 sites ha−1) for the good prediction agreement and accuracy of top- (0.85, 2.13) and subsoil (0.84, 1.90) ESP. Based on the best results, DSM of ESP was made to recommend gypsum application rates to ameliorate sodic soils according to the Six-Easy-Steps Management Guidelines. The northern fields required variable rates (5, 7.5, and 10 t ha−1) of gypsum, whereas the southern fields would not require gypsum.

Multi-property digital soil mapping at 30-m spatial resolution down to 1 m using extreme gradient boosting tree model and environmental covariates

Information on the variability of soil properties including clay, calcium carbonate equivalent (CCE) and acidity (pH) in space and depth play major role in sustainable agricultural development but scarce in many regions in the world, specifically in developing countries. Digital soil mapping technique has shown strong success in the prediction of soil properties both in space and depth from their relationship with environmental covariates. The aim of the study was to digitally map soil properties including clay, CCE and pH at multiple depths and identify the dominant factors controlling their variability using Minadoab region of Iran as a case study area due to its significance for agricultural production and economic development in the country. A total of 104 soil profiles were randomly selected and described in field during 15–30 July 2017, and 386 samples were collected from each genetic horizon with variable depths. A continuous spline function was fitted to the horizon data and properties were estimated at five standardized depths following GlobalSoilMap project. The soil layers were then grouped into three classes: i) H1: top (0–5 cm); ii) H2–H4: middle (5–15, 15–30 and 30–60 cm); and iii) H5: bottom (60–100 cm) for easy interpretation. A total of 17 static environmental covariates were derived from terrain-related attributes. Additionally, matching with the soil sampling date, a total of eight band ratios in addition to six individual bands were derived from Landsat-8 OLI (Product ID: LC08_L1TP_168034_20170721_20170728_01_T1) for the study area. Finally, all covariates were categorized in nine groups representing various factors controlling soil variations. The digital maps associated with uncertainties were prepared using extreme gradient boosting (XGBoost) Tree model and bootstrapping method, respectively. Overall, the highest accuracy using XGBoost Tree model was found for CCE (R2 = 0.73, on average), followed by pH (R2 = 0.72, on average) and clay (R2 = 0.70, on average). The highest amount of clay was observed for the bottom layer, followed by middle and top layers across the study area. H1 to H3 layers were subjected to tillage operations over the time and may have contributed to the redistribution of clays. A similar trend was also observed for CCE and pH as the location of the study area is underlain by massive limestone. The uncertainties helped quantify over- and under-estimation of the predictions and the factors with the highest variable importance was identified as the dominant factor. The information generated for the study area provides soil variability information that is critical for developing management strategies for sustainable production in the area.

Delineation of high‐resolution soil carbon management zones using digital soil mapping: A step towards mitigating climate change in the Northeastern Himalayas, India

AbstractDelineation of carbon management zones (CMZs) by capturing geospatial distribution of soil organic carbon (SOC) stock down the profile is an effective strategy for precision agriculture and climate change mitigation. Satellite (Landsat OLI 8), terrain (SRTM 30 m DEM) and bioclimatic (WorldClim dataset) factors were used as covariables in this digital soil mapping approach. Depth harmonization using the quadratic spline method (equal‐area) was carried out prior to quantile regression forest (QRF) algorithm‐based modelling to estimate SOC stock at six standard soil depths (0–5, 5–15, 15–30, 30–60, 60–100 and 100–200 cm). Soil depth and SOC stock for the whole soil profile were used for the delineation of CMZs using fuzzy k‐means clustering. The predicted SOC stock, varied from 14.68 to 42.35 Mg ha−1 in the top layer (0–5 cm depth), while 17.91 to 36.88, 14.15 to 34.70, 12.55 to 35.59, 10.30 to 28.52 and 7.26 to 20.16 Mg ha−1 in the depths of 5–15, 15–30, 30–60, 60–100 and 100–200 cm, respectively. The QRF algorithm performed well in predicting SOC stock with high R2, which ranged from .67 to .83 for all the soil depths. To delineate three CMZs, modified partitioning entropy and the fuzzy performance index were used. In CMZ2, there was a significant increase in SOC stock, followed by CMZ1 and CMZ3. This zone (CMZ2) was located in the central region of the study area and was mostly covered by dense forest and perennial plantations (rubber). The CMZs provided the necessary foundation for the development of site‐specific carbon management techniques that can enhance ecosystem service and meet climate change mitigation goals.

Soil organic carbon stock retrieval from Sentinel-2A using a hybrid approach.

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. Weconcluded 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.

Evaluation of digital soil mapping projection in soil organic carbon change modeling

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

Digital soil characteristics mapping for aiding site-specific management practices in the West Nile Delta, Egypt

Soil survey system is the key word for sustainable soil use and management. This study aimed to generate a digital soil mapping (DSM) using various spatial resolutions of freely available digital elevation models (DEMs) i.e., (30 m ALOS “Advanced Land Observing Satellite”, 12.5 m ALOS, 30 m SRTM “Shuttle Radar Topography Mission”, and 90 m SRTM) to assess the spatial variability of soil physio-chemical characteristics in El-Beheira Province, Western Nile Delta. A total of 196 geo-referenced soil profiles were collected from agricultural fields, and the soil's contents of total nitrogen (TN), potassium (K), available phosphorous (P), calcium (Ca), sulfur (S), magnesium (Mg), zinc (Zn), copper (Cu), manganese (Mn), and boron (B) were analyzed. The study revealed that the soil's contents of TN, P, S, B, and Cu were low, and Mg was suspected to be higher than K, leading to an Mg-induced K shortage. The study also found wide-ranging spatial variability structures of inverse distance weight (IDW) interpolated maps for several soil physio-chemical characteristics. The DSM further revealed that poor status was present in 85.4% of TN, 91.7% of P, 76.9% of S, and 87% of both B and Cu. The conclusion is that soil variability, as demonstrated in DSM, outweighs uniform treatment, and this discovery will assist extension workers, scientists, and decision-makers in implementing site-specific nutrient management strategies. Validation of the DSM is recommended for crop and variety-specific nutrient treatment rates. Understanding the geographical variance in the soil is critical for long-term nutrient management and higher output.

The Montes del Plata Soil Spatial Inference System

Soil is one of the prior components of a forest site, hence determining its final productivity. Effective soil management requires an understanding of soil distribution patterns within the landscape. The ability to accurately predict and map soil properties over Montes del Plata (MDP) landholdings can improve silvicultural management decisions that can result in higher and sustainable forest productivity. Today's decisions are based mainly on interpreting conventional soil maps (discrete semi-detailed) produced from local empirical experts' knowledge. Creation of conventional soil maps is a high time-consumable non-parametrizable process which result does not represent soil heterogeneous diversity, both in location and depth. In this work, we applied Digital Soil Mapping (DSM) techniques to produce continuous soil maps showing different soil-predicted property values with a concomitant measured uncertainty grade over 1,200 hectares. We were able to produce continuous soil properties maps with an average precision accuracy of 60% by developing a DSM protocol in three representative farms of MDP landholdings, in areas to be forested with Eucalyptus on its second cycles. A set of 53 environmental covariates was developed and harmonized at 30 m2 of spatial resolution. A field sampling strategy was defined based on the Latin Hypercube sampling method, and 180 soil profiles were collected. In laboratory conditions, we perform textural and chemical analyses on 345 soil samples from A and B horizons. Vertical harmonization was applied, and properties were estimated at three standardized depths (0-30, 30-60, and 60-100). A Random Forests Model (RMF) is fitted to predict texture, organic matter, organic carbon, pH_H2O, pH_KCl, N, P, K, Ca, Mg, Al, Na, cation exchange capacity, base saturation, and profile soil depth. Our results demonstrate that DSM techniques can be successfully applied to representative MDP's landholding soils. We anticipate our preliminary work as the starting point for using DSM techniques that result in acceptable soil continuous maps in less time. However, the developed protocol performance must be tested in similar soil conditions, and an approach must be undertaken to look for a lesser sampling density. These results, in addition to focusing on only those soil parameters linked to operative silvicultural decisions (viz. P, K, Ca, and soil depth), may lead to lower final associated costs.

Towards a cost-effective framework for estimating soil nitrogen pools using pedotransfer functions and machine learning

Globally, the strategic use of nitrogen (N) is important in optimizing economic returns and reducing soil nitrogen losses to the environment. Incorporating reliable estimates of nitrogen (N) mineralized over a growing season (GSN) into N-fertilizer rate prescriptions is critical, but may often lack a direct measurement. For this purpose, Pedotransfer functions (PTFs) of total nitrogen (TN) – representing the stable pool from which N is mineralized and biological nitrogen availability (BNA) – representing the labile pool of N mineralization were used to estimate GSN. GSN was calculated based on TN and BNA results from a soil health database (SHD), which also includes a suite of related soil health parameters (n = 2222). Using a process of recursive feature elimination (RFE) and cost-benefit feature elimination (CBFE), the best predictors of TN, BNA, and GSN were identified using a suite of machine learners (MLs) and regression analysis. For TN, RFE revealed that BNA, active carbon (AC), sand (Sa), and soil organic matter (OM) were the best predictors yielding a Lin’s concordance correlation coefficient (CCC) of 0.80 and a reduction in theoretical cost of 41 % compared to the control. CBFE resulted in AC, soil respiration (SR), clay, Sa, and OM as the most cost-effective predictors of TN with a CCC of 0.79 and a theoretical cost savings 49 % below the cost of using all appropriate soil health parameters in the SHD. With respect to BNA, the best predictors from RFE were aggregate stability (AS), AC, SR, and TN with a CCC of 0.78 and a theoretical cost reduction of 23 %. CBFE retained AC, SR, S, TN, OM and pH as predictors of BNA with a CCC of 0.78 and reduction of 29 % in theoretical cost. Finally, GSN results from RFE identified AS, AC, SR, OM and pH as the best predictors with a 0.82 CCC and 17 % reduction in theoretical cost. CBFE, on the other hand, identified AC, SR, sand, OM, and pH as the most cost-efficient predictors while maintaining a CCC of 0.82 and theoretical cost reduction of 29 %. Of the MLs used for pattern recognition (i.e., cubist, random forest, support vector machine, and stochastic gradient boosting), cubist model outperformed the others for the majority of iterations of the RFE and CBFE processes. The cost-effective framework, and the N-related PTFs developed in this study will greatly enhance our ability to predict of soil N-pool dynamics and the ability to incorporate GSN estimates into N-fertilizer recommendations for producers worldwide. Improvements in predictive strength could be achieved by incorporating climate and soil management practices into PTF development. Another area for improvement and future study would include addition of spatial and landscape variability related to N-measures via digital soil mapping applications.

National-scale mapping of soil organic carbon stock in France: New insights and lessons learned by direct and indirect approaches

Soil organic carbon (SOC) plays a crucial role in soil health and global carbon cycling, therefore accurate estimates of its spatial distribution are important for managing soil health and mitigating global climate change. Digital soil mapping shows its potential to provide accurate and high-resolution spatial distribution of SOC across scales. To convert SOC content to SOC density (SOCD), two inference trajectories exist for predicting SOCD in digital soil mapping: the direct approach (calculate-then-model) and indirect approach (model-then-calculate). However, there is a lack of comprehensive exploration regarding the differences in their performance in SOCD estimates, particularly in regions characterized by diverse pedoclimatic conditions. To bridge this knowledge gap, we evaluated the two approaches based on model performance of SOCD in France. Using 916 topsoils (0−20 ​cm) from the LUCAS Soil 2018 and 24 environmental covariates, random forest model and forward recursive feature selection were used to build the spatial predictive models of SOCD using direct and indirect approaches. The results show that, using random forest model and full covariates, both approaches show moderate performance (R2 ​= ​0.28−0.32). By utilizing forward recursive feature selection model, the number of predictors was reduced from 24 to 9, enhancing model performance for direct approach (R2 of 0.35), with no improvement for indirect approach (R2 of 0.28). The mean SOCD of the French topsoil was 5.29 and 6.14 ​kg ​m−2 by direct and indirect approaches, resulting in SOC stock of 2.8 and 3.3 ​Pg, respectively. We found that the direct approach clearly underestimated the high SOCD (>9 ​kg ​m−2), while the indirect approach performed much better for high SOCD. Our findings serve as a valuable reference for SOCD mapping, thereby providing a scientific basis for maintaining soil health.

Modelling and prediction of major soil chemical properties with Random Forest: Machine learning as tool to understand soil-environment relationships in Antarctica

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.