Search

Many soil functions and processes are controlled by the soil particle size distribution. Accordingly, nationwide geoinformation on this soil property is required to enable climate-smart and resilient land management. This study presents a new deep learning approach to simultaneously model the contents of the three particle sizes of sand, silt, and clay and their variations with depth throughout the landscape. The approach allows for the consideration of the natural soil horizon boundaries and the inclusion of the surrounding landscape context of each soil profile to investigate the soil–landscape relation. Applied to the agricultural soil landscape of Germany, the approach generated a three-dimensional continuous data product with a resolution of 100 m in geographic space and a depth resolution of 1 cm. The approach relies on a patch-wise multi-target convolutional neural network (CNN) model. Genetic algorithm optimization was applied for CNN parameter tuning. Overall, the effectiveness of the CNN algorithm in generating multidimensional, multivariate, national-scale soil data products was demonstrated. The predictive performance resulted in a median root mean square error of 17.8 mass-% for the sand, 14.4 mass-% for the silt, and 9.3 mass-% for the clay content in the top ten centimeters. This increased to 20.9, 16.5, and 11.8 mass-% at a 40 cm depth. The generated data product is the first of its kind. However, even though the potential of this deep learning approach to understand and model the complex soil–landscape relation is virtually limitless, its limitations are data driven concerning the approximation of the soil-forming factors and the available soil profile data.

Many soil functions and processes are controlled by the soil particle size distribution. Accordingly, nationwide geoinformation on this soil property is required to enable climate-smart and resilient land management. This study presents a new deep learning approach to simultaneously model the contents of the three particle sizes of sand, silt, and clay and their variations with depth throughout the landscape. The approach allows for the consideration of the natural soil horizon boundaries and the inclusion of the surrounding landscape context of each soil profile to investigate the soil–landscape relation. Applied to the agricultural soil landscape of Germany, the approach generated a three-dimensional continuous data product with a resolution of 100 m in geographic space and a depth resolution of 1 cm. The approach relies on a patch-wise multi-target convolutional neural network (CNN) model. Genetic algorithm optimization was applied for CNN parameter tuning. Overall, the effectiveness of the CNN algorithm in generating multidimensional, multivariate, national-scale soil data products was demonstrated. The predictive performance resulted in a median root mean square error of 17.8 mass-% for the sand, 14.4 mass-% for the silt, and 9.3 mass-% for the clay content in the top ten centimeters. This increased to 20.9, 16.5, and 11.8 mass-% at a 40 cm depth. The generated data product is the first of its kind. However, even though the potential of this deep learning approach to understand and model the complex soil–landscape relation is virtually limitless, its limitations are data driven concerning the approximation of the soil-forming factors and the available soil profile data.

Soil organic matter (SOM) and the ratio of soil organic carbon to total nitrogen (C/N ratio) are fundamental to the ecosystem services provided by soils. Therefore, understanding the spatial distribution and relationships between the SOM components mineral-associated organic matter (MAOM), particulate organic matter (POM), and C/N ratio is crucial. Three ensemble machine learning models were trained to obtain spatial predictions of the C/N ratio, MAOM, and POM in German agricultural topsoil (0–10 cm). Parameter optimization and model evaluation were performed using nested cross-validation. Additionally, a modification to the regressor chain was applied to capture and interpret the interactions among the C/N ratio, MAOM, and POM. The ensemble models yielded mean absolute percent errors (MAPEs) of 8.2% for the C/N ratio, 14.8% for MAOM, and 28.6% for POM. Soil type, pedo-climatic region, hydrological unit, and soilscapes were found to explain 75% of the variance in MAOM and POM, and 50% in the C/N ratio. The modified regressor chain indicated a nonlinear relationship between the C/N ratio and SOM due to the different decomposition rates of SOM as a result of variety in its nutrient quality. These spatial predictions enhance the understanding of soil properties’ distribution in Germany.

Carbon sequestration in soils under agricultural use can contribute to climate change mitigation. Spatial–temporal soil organic carbon (SOC) monitoring requires more efficient data acquisition. This study aims to evaluate the potential of spectral on-the-go proximal measurements to serve these needs. The study was conducted as a long-term field experiment. SOC values ranged between 14 and 25 g kg−1 due to different fertilization treatments. Partial least squares regression models were built based on the spectral laboratory and field data collected with two spectrometers (site-specific and on-the-go). Correction of the field data based on the laboratory data was done by testing linear transformation, piecewise direct standardization, and external parameter orthogonalization (EPO). Different preprocessing methods were applied to extract the best possible information content from the sensor signal. The models were then thoroughly interpreted concerning spectral wavelength importance using regression coefficients and variable importance in projection scores. The detailed wavelength importance analysis disclosed the challenge of using soil spectroscopy for SOC monitoring. The use of different spectrometers under varying soil conditions revealed shifts in wavelength importance. Still, our findings on the use of on-the-go spectroscopy for spatial–temporal SOC monitoring are promising.

Effective monitoring of the soil organic carbon (SOC) content at the field scale is crucial for supporting sustainable agricultural practices. This study evaluates the utility of multispectral data acquired by an unmanned aerial vehicles (UAV) during bare soil conditions for predicting the SOC content of a long-term experimental field site (LTE) in Saxony-Anhalt, Germany. Our methodology involves constructing predictive models using multiple algorithms (CUBIST, MARS, linear regression) and applying image correction techniques to enhance prediction accuracy by mitigating the influence of confounding factors such as crop residuals. Among the tested models, the CUBIST algorithm, combined with a pixel selection strategy employing a 2 m radius and stratified image correction, demonstrates the most promising results, achieving an R-squared value of 0.54 and an RMSE of 1.9 g kg−1. Spatial distribution maps generated by this optimized model effectively depict the impact of organic fertilization on the SOC content, although the clarity of these patterns varies depending on the image processing method and algorithm used. Our findings highlight the potential of utilizing UAV-derived multispectral data for SOC monitoring at the LTE scale. However, further research is warranted to assess the generalizability of this approach to agricultural fields with lower SOC variability.

Agricultural soils serve as crucial storage sites for soil organic carbon (SOC). Their appropriate management is pivotal for mitigating climate change. Continuous monitoring is imperative to evaluate spatial and temporal changes in SOC within agricultural fields. In-field datasets of Vis-NIR soil spectra were collected on a long-term experimental site using an on-the-go spectrophotometer. Data processing for continuous SOC prediction involves a two-step modeling approach. In Step 1, a partial least square (PLSR) regression model is trained to establish a relationship between the SOC content and spectral information, including spectral preprocessing. In Step 2, the predicted SOC content obtained from the PLSR models is interpolated using ordinary kriging. Among the tested spectral preprocessing techniques and semivariogram models, Savitzky–Golay and the Gap-Segment derivative preprocessing along with a Gaussian semivariogram model, yielded the best performance resulting in a root mean square error of 1.24 and 1.26 g kg−1. A striping effect due to the transect-based data collection was addressed by testing the effectiveness of extending the spatial separation distance, employing data aggregation, and defining the distribution based on treatment plots using block kriging. Overall, the results highlight the high potential of on-the-go spectral Vis-NIR data for field-scale spatial-temporal monitoring of SOC.

The national-scale evaluation and modelling of the impact of agricultural management and cli-mate change on soils, crop growth, and the environment require soil information at a spatial res-olution addressing individual agricultural fields. This manuscript presents a data science ap-proach which agglomerates the soil parameter space into a limited number of functional soil pro-cess units (SPUs) which may be used to run agricultural process models. In fact, two unsupervised classification methods were developed to generate a multivariate 3D data product consisting of SPUs, each being defined by a multivariate parameter distribution along the depth profile from 0 to 100 cm. The two methods account for differences in variable types and distributions and in-volve genetic algorithm optimization to identify those SPUs with the lowest internal variability and maximum inter-unit difference with regards to both, their soil characteristics and landscape setting. The high potential of the methods was demonstrated by applying them to the agricultural German soil landscape. The resulting data product consists of twenty SPUs. It has a 100 m raster resolution in the 2D mapping space, and its resolution along the depth profile is 1 cm. It includes the soil properties texture, stone content, bulk density, hydromorphic properties, total organic carbon content, and pH.

The national-scale evaluation and modelling of the impact of agricultural management and cli-mate change on soils, crop growth, and the environment require soil information at a spatial res-olution addressing individual agricultural fields. This manuscript presents a data science ap-proach which agglomerates the soil parameter space into a limited number of functional soil pro-cess units (SPUs) which may be used to run agricultural process models. In fact, two unsupervised classification methods were developed to generate a multivariate 3D data product consisting of SPUs, each being defined by a multivariate parameter distribution along the depth profile from 0 to 100 cm. The two methods account for differences in variable types and distributions and in-volve genetic algorithm optimization to identify those SPUs with the lowest internal variability and maximum inter-unit difference with regards to both, their soil characteristics and landscape setting. The high potential of the methods was demonstrated by applying them to the agricultural German soil landscape. The resulting data product consists of twenty SPUs. It has a 100 m raster resolution in the 2D mapping space, and its resolution along the depth profile is 1 cm. It includes the soil properties texture, stone content, bulk density, hydromorphic properties, total organic carbon content, and pH.

As the largest terrestrial carbon pool, soil organic carbon (SOC) has the potential to influence and mitigate climate change; thus, SOC monitoring is of high importance in the frameworks of various international treaties. Therefore, high-resolution SOC maps are required. Machine learning (ML) offers new opportunities to develop these maps due to its ability to data mine large datasets. The aim of this study was to apply three algorithms commonly used in digital soil mapping – random forest (RF), boosted regression trees (BRT), and support vector machine for regression (SVR) – on the first German agricultural soil inventory to model the agricultural topsoil (0–30 cm) SOC content and develop a two-model approach to address the high variability in SOC in German agricultural soils. Model performance is often limited by the size and quality of the soil dataset available for calibration and validation. Therefore, the impact of enlarging the training dataset was tested by including data from the European Land Use/Cover Area frame Survey for agricultural sites in Germany. Nested cross-validation was implemented for model evaluation and parameter tuning. Grid search and the differential evolution algorithm were also applied to ensure that each algorithm was appropriately tuned. The SOC content of the German agricultural soil inventory was highly variable, ranging from 4 to 480 g kg−1. However, only 4 % of all soils contained more than 87 g kg−1 SOC and were considered organic or degraded organic soils. The results showed that SVR produced the best performance, with a root-mean-square error (RMSE) of 32 g kg−1 when the algorithms were trained on the full dataset. However, the average RMSE of all algorithms decreased by 34 % when mineral and organic soils were modelled separately, with the best result from SVR presenting an RMSE of 21 g kg−1. The model performance was enhanced by up to 1 % for mineral soils and by up to 2 % for organic soils. Despite the ability of machine lea

High-resolution soil maps are urgently needed by land managers and researchers for a variety of applications. Digital soil mapping (DSM) allows us to regionalize soil properties by relating them to environmental covariates with the help of an empirical model. In this study, a legacy soil dataset was used to train a machine learning algorithm in order to predict the particle size distribution within the catchment of the Bode River in Saxony-Anhalt (Germany). The random forest ensemble learning method was used to predict soil texture based on environmental covariates originating from a digital elevation model, land cover data and geologic maps. We studied the usefulness of clustering applications in addressing various aspects of the DSM procedure. To improve areal representativity of the legacy soil data in terms of spatial variability, the environmental covariates were used to cluster the landscape of the study area into spatial units for stratified random sampling. Different sampling strategies were used to create balanced training data and were evaluated on their ability to improve model performance. Clustering applications were also involved in feature selection and stratified cross-validation. Under the best-performing sampling strategy, the resulting models achieved an R2 of 0.29 to 0.50 in topsoils and 0.16–0.32 in deeper soil layers. Overall, clustering applications appear to be a versatile tool to be employed at various steps of the DSM procedure. Beyond their successful application, further application fields in DSM were identified. One of them is to find adequate means to include expert knowledge.

Site-specific spatially continuous soil texture data is required for many purposes such as the simulation of carbon dynamics, the estimation of drought impact on agriculture, or the modeling of water erosion rates. At large scales, there are often only conventional polygon-based soil texture maps, which are hardly reproducible, contain abrupt changes at polygon borders, and therefore are not suitable for most quantitative applications. Digital soil mapping methods can provide the required soil texture information in form of reproducible site-specific predictions with associated uncertainties.Machine learning models were trained in a nested cross-validation approach to predict the spatial distribution of the topsoil (0 to 30 cm) clay, silt, and sand contents in 100 m resolution. The differential evolution algorithm was applied to optimize the model parameters. High-quality nation-wide soil texture data of 2991 soil profiles was obtained from the first German agricultural soil inventory. We tested an iterative approach by training models on predictor datasets of increasing size, which contained up to 50 variables.The best results were achieved when training the models on the complete predictor dataset. They explained about 59% of the variance in clay, 75% of the variance in silt, and 77% of the variance in sand content. The RMSE values ranged between approximately 8.2 wt.% (clay), 11.8 wt.% (silt), and 15.0 wt.% (sand). Due to their high performance, models were able to predict the spatial texture distribution. They captured the high importance of the soil forming factors parent material and relief. Our results demonstrate the high predictive power of machine learning in predicting soil texture at large scales. The iterative approach enhanced model interpretability. It revealed that the incorporated soil maps partly substituted the relief and parent material predictors. Overall, the spatially continuous soil texture predictions provide valuable input for many quantita

Societal demands on soil functionality in agricultural soil-landscapes are confronted with yield losses and environmental impact. Soil functional information at national scale is required to address these challenges. On behalf of the well-known theory that soils and their site-specific characteristics are the product of the interaction of the soil-forming factors, pedometricians seek to model the soil-landscape relationship using machine learning. Following the rationale that similarity in soils is reflected by similarity in landscape characteristics, we defined soil functional types (SFTs) which were projected into space by machine learning. Each SFT is described by a multivariate soil parameter distribution along its depth profile. SFTs were derived by employing multivariate similarity analysis on the dataset of the Agricultural Soil Inventory. Soil profiles were compared on behalf of differing sets of soil properties considering the top 100 and 200 cm, respectively. Various depth weighting coefficients were tested to attribute topsoil properties higher importance. Support vector machine (SVM) models were then trained employing optimization with a distributed multiple-population hybrid Genetic algorithm for parameter tuning. Model training, tuning, and evaluation were implemented in a nested k-fold cross-validation approach to avoid overfitting. With regards to the SFTs, organic soils were differentiated from mineral soils of various particle size distributions being partly influenced by waterlogging and groundwater. Further SFTs reflect soils with a depth limitation within the top 100 cm and high stone content. Altogether, with SVM predictive model accuracies between 0.7 and 0.9, the agricultural soil-landscape of Germany was represented with eight SFTs. Soil functionality with regards to the soil’s capacity to store plant-available water and soil organic carbon is well characterized. Four additional soil functions are described to a certain extent. An extension

From a theoretical point of view, result-based agri-environmental payments are clearly preferable to action-based payments. However, they suffer from two major practical disadvantages: costs of measuring the results and payment uncertainty for the participating farmers. In this paper, we propose an alternative design to overcome these two disadvantages by means of modelling (instead of measuring) the results. We describe the concept of agri-environmental payments by modelled results (PAMR), including a hypothetical example of payments for the protection and enhancement of soil functions. We offer a comprehensive discussion of the relative advantages and disadvantages of PAMR, showing that it not only unites most of the different advantages of result-based and action-based schemes, but also adds two new advantages: the potential to address trade-offs among multiple policy objectives and management for long-term environmental effects. We argue that PAMR would be a valuable addition to the agri-environmental policy toolbox in the EU and beyond, while also reflecting recent advancements in agri-environmental modelling.

Machine-learning algorithms are good at computing non-linear problems and fitting complex composite functions, which makes them an adequate tool for addressing multiple environmental research questions. One important application is the development of pedotransfer functions (PTFs). This study aims to develop water retention PTFs for two remote tropical mountain regions with rather different soil landscapes: (1) those dominated by peat soils and soils under volcanic influence with high organic matter contents and (2) those dominated by tropical mineral soils. Two tuning procedures were compared to fit boosted regression tree models: (1) tuning with grid search, which is the standard approach in pedometrics; and (2) tuning with differential evolution optimization. A nested cross-validation approach was applied to generate robust models. The area-specific PTFs developed outperform other more general PTFs. Furthermore, the first PTF for typical soils of Páramo landscapes (Ecuador), i.e., organic soils under volcanic influence, is presented. Overall, the results confirmed the differential evolution algorithm's high potential for tuning machine-learning models. While models based on tuning with grid search roughly predicted the response variables' mean for both areas, models applying the differential evolution algorithm for parameter tuning explained up to 25 times more of the response variables' variance.