Search

The soil's pH is the single most important indicator of the soil's quality, whether for agriculture, pollution control or environmental health and ecosystem functioning. Well documented data on soil pH are sparse for the whole of China — data for only 4700 soil profiles were available from China's Second National Soil Inventory. By combining those data, standardized for the topsoil (0–20 cm), with 17 environmental covariates at a fine resolution (3 arc-second or 90 m) we have predicted the soil's pH at that resolution, that is at more than 109 points. We did so by parallel computing over tiles, each 100 km × 100 km, with two machine learning techniques, namely Random Forest and XGBoost. The predictions for the tiles were then merged into a single map of soil pH for the whole of China. The quality of the predictions were assessed by cross-validation. The root mean squared error (RMSE) was an acceptable 0.71 pH units per point, and Lin's Concordance Correlation Coefficient was 0.84. The hybrid model revealed that climate (mean annual precipitation and mean annual temperature) and soil type were the main factors determining the soil's pH. The pH map showed acid soil mainly in southern and north-eastern China, and alkaline soil dominant in northern and western China. This map can provide a benchmark against which to evaluate the impacts of changes in land use and climate on the soil's pH, and it can guide advisors and agencies who make decisions on remediation and prevention of soil acidification, salinization and pollution by heavy metals, for which we provide examples for cadmium and mercury.

Although soils have a high potential to offset CO2 emissions through its conversion into soil organic carbon (SOC) with long turnover time, it is widely accepted that there is an upper limit of soil stable C storage, which is referred to SOC saturation. In this study we estimate SOC saturation in French topsoil (0-30cm) and subsoil (30-50cm), using the Hassink equation and calculate the additional SOC sequestration potential (SOCsp) by the difference between SOC saturation and fine fraction C on an unbiased sampling set of sites covering whole mainland France. We then map with fine resolution the geographical distribution of SOCsp over the French territory using a regression Kriging approach with environmental covariates. Results show that the controlling factors of SOCsp differ from topsoil and subsoil. The main controlling factor of SOCsp in topsoils is land use. Nearly half of forest topsoils are over-saturated with a SOCsp close to 0 (mean and standard error at 0.19±0.12) whereas cropland, vineyard and orchard soils are largely unsaturated with degrees of C saturation deficit at 36.45±0.68% and 57.10±1.64%, respectively. The determinant of C sequestration potential in subsoils is related to parent material. There is a large additional SOCsp in subsoil for all land uses with degrees of C saturation deficit between 48.52±4.83% and 68.68±0.42%. Overall the SOCsp for French soils appears to be very large (1008Mt C for topsoil and 1360Mt C for subsoil) when compared to previous total SOC stocks estimates of about 3.5Gt in French topsoil. Our results also show that overall, 176Mt C exceed C saturation in French topsoil and might thus be very sensitive to land use change.

International audience; Digital Soil Mapping and Assessment (DSMA) has progressed from challenging traditional soil science paradigms, through small scale prototyping, to large-scale implementation capturing quantitative measures of soil attributes and functions. This paper considers the future for DSMA in the context of a highly uncertain world where high-quality knowledge of soil dynamics will be important for responding to the challenges of sustainability. Irrespective of whether the need is for survival, increased productivity or broadening the services provided from land management, or simply securing the soil itself, we see DSMA as a fundamental approach and essential tool. With a broadening need and a strong foundation in the practice of DSMA now in place, the theory, tools and technology of DSMA will grow significantly. We explore expected changes in covariate data, the modelling process, the nature of base data generation and product delivery that will lead to tracking and forecasting a much wider range of soil attributes and functions at finer spatial and temporal resolutions over larger areas, particularly globally. Equally importantly, we expect the application and impact of DSMA to broaden and be used, directly and collaterally, in the analysis of land management issues in coming decades. It has the capacity to provide the background to a soil and landscape ‘digital twin’ and the consequent transformation in monitoring and forecasting the impacts of land management practices. We envision the continued growth of DSMA skills amongst soil scientists and a much broader community of practice involved in developing and utilizing DSMA products and tools. Consequently, there will be a widening and deepening role of public-private partnerships in this development and application.

International audience; The density of soil observations is a major determinant of digital soil mapping (DSM) prediction accuracy. In this study, we investigated the effect of soil sampling density on the performance of DSM to predict topsoil particle-size distribution in the Mayenne region of France. We tested two prediction algorithms, namely ordinary kriging (OK) and quantile random forest (QRF). The study area is a region of ~5000 km2 with the highest density of field soil observations in France (1 profile per 0.64 km2). The number of training sites was progressively reduced (from n = 7500 to n = 400, corresponding to 1 profile per 0.7 km2 to 1 profile per 13 km2) to simulate the different density of observations. For OK and QRF, we tested random subsampling for splitting the data into training and testing datasets using k-fold cross validation. For QRF we also tested conditioned Latin hypercube sampling based on the point coordinates or the covariates. The results indicated that, with increasing density of observations, OK performed as well or even better than QRF, depending on the particle-size fraction. For silt prediction, OK was systematically better than QRF. However, the prediction intervals were much larger for OK than for QRF, and OK did not seem to estimate uncertainty correctly. Overall, the performance indicators increased with the density of observations with a threshold at about 1 profile per 2 km2 which suggests that the main limitation of DSM prediction accuracy using QRF is the amount of data collected in the field, not the type of calibration sampling strategy. Future DSM activities should focus on gathering more field observations.

International audience; This work presents the first high-resolution map of soil organic carbon (SOC) in mainland France, including soils below 30 cm. The research was performed within the framework of GlobalSoilMap (GSM). SOC predictions for different depth layers (0–5 cm, 5–15 cm, 15–30 cm, 30–60 cm, 60–100 cm and > 100 cm) were made at 90 and 500 m resolution for mainland France, along with their upper and lower confidence intervals. The maps were developed using data mining and an elaborate cross-validation scheme. The 90 m maps were compared to 500 m resolution GlobalSoilMap maps and the SoilGrids1km (SG1km) product. The latter is a global model for predicting soil properties for the same depth layers, at 1 km resolution. At 90 and 500 m resolution, the predicted SOC content was unbiased and showed good agreement with the measured SOC, despite the poor model diagnostics and decrease of performance with depth. It was found that the subsoil (> 30 cm) carbon pool for France contributes 49% to the total soil carbon stock. The use of coarser resolution prediction grids resulted in smoother spatial patterns and wider confidence intervals; however, it did not bias the estimated carbon stocks. Applying GlobalSoilMap specifications to France, using a large soil dataset and all the exhaustive spatially available data outperformed SG1km predictions. The overall spatial patterns of the SG1km SOC content were found to be very similar to the GlobalSoilMap maps. However, the SG1km overestimated the SOC content and carbon stocks (> 75% for the total carbon stock, and 100% for the stocks below 30 cm) and showed a similar spatial distribution over the different soil depth layers. The main reason for the overestimation was that the local data used in SG1km was rather small (56 samples) and not representative in terms of SOC content or represented soil types; the profiles had far higher SOC content and this may have propagated in the modelled vertical profile and the kriging part of the residuals. Improvements for SG1km may entail the use of a representative national subsample from large national soil databases. Furthermore, a bottom-up approach such as GlobalSoilMap may be more favourable when considering prediction accuracies, data privacy policies and local acceptance of generated products.

International audience; It has long been acknowledged that the soil spatial samplings used as inputs to DSM models are strong drivers – and often limiting factors – of the performances of such models. However, few studies have focused on evaluating this impact and identifying the related spatial sampling characteristics. In this study, a numerical experiment was conducted on this topic using the pseudo values of topsoil clay content obtained from an airborne Visible Near InfraRed-Short Wave InfraRed (Vis-NIR-SWIR) hyperspectral image in the Cap Bon region (Tunisia) as the source of the spatial sampling. Twelve thousand DSM models were built by running a Random Forest algorithm from soil spatial sampling of different sizes and average spacings (from 200 m to 2000 m) and different spatial distributions (from clustered to regularly distributed), aiming to mimic the various situations encountered when handling legacy data. These DSM models were evaluated with regard to both their prediction performances and their ability to estimate their overall and local uncertainties. Three evaluation methods were applied: a model-based one, a classical model-free one using 25% of the sites removed from the initial soil data, and a reference one using a set of 100,000 independent sites selected by stratified random sampling over the entire region. The results showed that: 1) While, as expected, the performances of the DSM models increased when the spacing of the sample increased, this increase was diminished for the smallest spacing as soon as 50% of the spatially structured variance was captured by the sampling, 2) Sampling that provided complete and even distributions in the geographical space and had as great spread of the target soil property as possible increased the DSM performances, while complete and even sampling distributions in the covariate space had less impacts, 3) Systematic underestimations of the overall uncertainty of DSM models were observed, that were all the more important that the sparse samplings poorly covered the real distribution of the target soil property and that the dense sampling were unevenly distributed in the geographical space, 4) The local uncertainties were underestimated for sparse sampling and over-estimated for dense sampling while being sensitive to the same sampling characteristics as overall uncertainty. Such finding have practical outcomes on sampling strategies and DSM model evaluation that are discussed.

10 pages; International audience; A better understanding of the links between dissolved organic matter and biogeochemical processes in soil could help in evaluating global soil dynamics. To assess the effects of land cover and parental material on soil biogeochemistry, we studied 120 soil samples collected from various ecosystems in Burgundy, France. The potential solubility and aromaticity of dissolved organic matter was characterised by pressurised hot-water extraction of organic carbon (PH-WEOC). Soil physico-chemical characteristics (pH, texture, soil carbon and nitrogen) were measured, as was the δ13C signature both in soils and in PH-WEOC. We also determined bacterial and fungal abundance and the genetic structure of bacterial communities. Our results show that the potential solubility of soil organic carbon is correlated to carbon and clay content in the soil. The aromaticity of PH-WEOC and its δ13C signature reflect differences in the decomposition pathways of soil organic matter and in the production of water-extractable organic compounds, in relation to land cover. The genetic structure of bacterial communities is related to soil texture and pH, and to PH-WEOC, revealing that water-extractable organic matter is closely related to the dynamics of bacterial communities. This comprehensive study, at the regional scale, thus provides better definition of the relationships between water-extractable organic matter and soil biogeochemical properties.

Reconnaissance soil maps at 1:250,000 scale are the most detailed source of soil information for large parts of France. For many environmental applications, however, the level of detail and accuracy of these maps is insufficient. Funds are lacking to refine and update these maps by traditional soil survey. In this study we investigated the merit of digital soil mapping to refine and improve the 1:250,000 reconnaissance soil map of a 1580 km2 area in Haute-Normandie, France. The soil map was produced in 1988 and distinguishes nine soil class units. The approach taken was to predict soil class from a large number of environmental covariates using regression techniques. The covariates used include DEM derivatives, geology and land cover maps. Because very few soil point observations were available within the area, we calibrated the regression model by sampling the soil map on a grid. We calibrated three models: classification tree (CT), multinomial logistic regression (MLR) and random forests (RF), and used these models to predict the nine soil classes across the study area. The new and original maps were validated with field data from 123 locations selected with a stratified simple random sampling design. For MLR, the estimate of the overall purity was 65.9%, while that of the reconnaissance map was 55.5%. The difference between the purity estimates of these maps was statistically significant (p = 0.014). The significant improvement over the existing soil map is remarkable because the regression model was calibrated with the existing soil map and uses no additional soil observations.

National-scale soil monitoring networks are required to identify where the soil quality is threatened and evaluate the effectiveness of any remediation efforts. In particular, trace elements (TEs) should be monitored to assess human exposure to potentially toxic elements and to ensure that crops take up sufficient quantities of elements that have essential functions in the human body. However the spatial variation of TEs is often highly complex since hot-spots at point sources are superimposed upon more regular patterns produced by diffuse pollution and natural processes. Geostatistical methods used to analyse national-scale soil monitoring networks must be general enough to accommodate such behaviour but simple enough to be computed in a reasonable time for large datasets. We show how a simplified version of a recently developed robust geostatistical algorithm to map the underlying variation of six TEs (Cr, Cu, Pb, Ni, Th and Zn) across France using observations from the French National Soil Monitoring Network (Réseau de Mesures de la Qualité des Sols). Cross-validation results suggest that these TEs cannot be modelled by non-robust methods but that the simplified robust approach is sufficient. Differences between the maps of different elements and the abundance of TEs from different parent materials are evident. Large concentrations of Cr, Ni and Zn occur in soils on Jurassic rocks whereas Pb and Th concentrations are large in soils on crystalline rocks. Volcanic parent material leads to large concentrations of Cr, Cu, Ni and Zn but small concentrations of Pb and Th. Diffuse pollution of certain elements (mainly Pb, and to a lesser extent Zn) is evident in industrial regions in the north and the north-east of France and close to Paris. The pattern of outlying values is indicative of local anthropogenic processes such as industrial pollution in the north of France and close to Paris, and application of Cu on vineyards and of geological anomalies such as large concentrations of some TEs in the south of the Massif Central Mountains. Future phases of the RMQS will describe the spatial and temporal trends of the concentrations of these TEs.