Monitoring and modelling soil organic carbon (SOC) in space and time can help us to better understand soil carbon dynamics and is of key importance to support climate change research and policy. ...Although machine learning (ML) has attracted a lot of attention in the digital soil mapping (DSM) community for its powerful ability to learn from data and predict soil properties, such as SOC, it is better at capturing soil spatial variation than soil temporal dynamics. By contrast, process-oriented (PO) models benefit from mechanistic knowledge to express physiochemical and biological processes that govern SOC temporal changes. Therefore, integrating PO and ML models seems a promising means to represent physically plausible SOC dynamics while retaining the spatial prediction accuracy of ML models. In this study, a hybrid modelling framework was developed and tested for predicting topsoil SOC stock in space and time for a regional cropland area located in eastern China. In essence, the hybrid model uses predictions of the PO model in unsampled years as additional training data of the ML model, with a weighting parameter assigned to balance the importance of SOC values from the PO model and real measurements. The results indicated that temporal trends of SOC stock modelled by PO and ML models were largely different, while they were notably similar between the PO and hybrid models. Cross-validation showed that the hybrid model had the best performance (RMSE = 0.29 kg m−2), with a 19 % improvement compared with the ML model. We conclude that the proposed hybrid framework not only enhances space-time soil carbon mapping in terms of prediction accuracy and physical plausibility, it also provides insights for soil management and policy decisions in the face of future climate change and intensified human activities.
Display omitted
•Clarifies the advantages and disadvantages of process-oriented (PO) models and machine learning (ML) models in space-time soil carbon modelling•Proposes a general framework (POML method) for integrating PO and ML models•Evaluates and compares POML model with single PO and ML model in spatial patterns, temporal trends and prediction accuracies.•Discusses applicability and future horizons on space-time modelling of soil carbon using POML framework
Despite the increased usage of global soil property maps, a proper review of the maps rarely takes place. This study aims to explore the options for such a review with an application for the S-World ...global soil property database. Global soil organic carbon (SOC) and clay content maps from S-World were studied at two spatial resolutions in three steps. First, a comparative analysis with an ensemble of seven datasets derived from five other global soil databases was done. Second, a validation of S-World was done with independent soil observations from the WoSIS soil profile database. Third, a methodological evaluation of S-world took place by looking at the variation of soil properties per soil type and short distance variability. In the comparative analysis, S-World and the ensemble of other maps show similar spatial patterns. However, the ensemble locally shows large discrepancies (e.g., in boreal regions where typically SOC contents are high and the sampling density is low). Overall, the results show that S-World is not deviating strongly from the model ensemble (91% of the area falls within a 1.5% SOC range in the topsoil). The validation with the WoSIS database showed that S-World was able to capture a large part of the variation (with, e.g., a root mean square difference of 1.7% for SOC in the topsoil and a mean difference of 1.2%). Finally, the methodological evaluation revealed that estimates of the ranges of soil properties for the different soil types can be improved by using the larger WoSIS database. It is concluded that the review through the comparison, validation, and evaluation provides a good overview of the strengths and the weaknesses of S-World. The three approaches to review the database each provide specific insights regarding the quality of the database. Specific evaluation criteria for an application will determine whether S-World is a suitable soil database for use in global environmental studies.
Soils are at the crossroads of many existential issues that humanity is currently facing. Soils are a finite resource that is under threat, mainly due to human pressure. There is an urgent need to ...map and monitor them at field, regional, and global scales in order to improve their management and prevent their degradation. This remains a challenge due to the high and often complex spatial variability inherent to soils. Over the last four decades, major research efforts in the field of pedometrics have led to the development of methods allowing to capture the complex nature of soils. As a result, digital soil mapping (DSM) approaches have been developed for quantifying soils in space and time. DSM and monitoring have become operational thanks to the harmonization of soil databases, advances in spatial modeling and machine learning, and the increasing availability of spatiotemporal covariates, including the exponential increase in freely available remote sensing (RS) data. The latter boosted research in DSM, allowing the mapping of soils at high resolution and assessing the changes through time. We present a review of the main contributions and developments of French (inter)national research, which has a long history in both RS and DSM. Thanks to the French SPOT satellite constellation that started in the early 1980s, the French RS and soil research communities have pioneered DSM using remote sensing. This review describes the data, tools, and methods using RS imagery to support the spatial predictions of a wide range of soil properties and discusses their pros and cons. The review demonstrates that RS data are frequently used in soil mapping (i) by considering them as a substitute for analytical measurements, or (ii) by considering them as covariates related to the controlling factors of soil formation and evolution. It further highlights the great potential of RS imagery to improve DSM, and provides an overview of the main challenges and prospects related to digital soil mapping and future sensors. This opens up broad prospects for the use of RS for DSM and natural resource monitoring.
In this work we aimed at developing a conceptual framework in which we improve our understanding of the controlling factors for soil organic carbon (SOC) over vast areas at different depths. We ...postulated that variability in SOC levels may be better explained by modeling SOC within soil‐landscape systems (SLSs). The study was performed in mainland France, and explanatory SOC models were developed for the sampled topsoil (0–30 cm) and subsoil (>30 cm), using both directed and undirected data‐mining techniques. With this study we demonstrated that there is a shift in controlling factors both in space and depth which were mainly related to (1) typical SLS characteristics and (2) human‐induced changes to SLSs. The controlling factors in relation to depth alter from predominantly biotic to more abiotic with increasing depth. Especially, water availability, soil texture, and physical protection control deeper stored SOC. In SLSs with similar SOC levels, different combinations of physical protection, the input of organic matter, and climatic conditions largely determined the SOC level. The SLS approach provided the means to partition the data into data sets that were having homogenous conditions with respect to this combination of controlling factors. This information may provide important information on the carbon storage and sequestration potential of a soil.
Key Points
SOC drivers alter from mainly biotic to abiotic with increasing soil depth
SOC storage depends on physical protection in addition to climate and OC input
Soil‐landscape systems explain large‐extent SOC dynamics in space and depth
Abstract For restoring soil health and mitigating climate change, information of soil organic matter is needed across space, depth and time. Here we developed a statistical modelling platform in ...three-dimensional space and time as a new paradigm for soil organic matter monitoring. Based on 869 094 soil organic matter observations from 339,231 point locations and the novel use of environmental covariates variable in three-dimensional space and time, we predicted soil organic matter and its uncertainty annually at 25 m resolution between 0–2 m depth from 1953–2022 in the Netherlands. We predicted soil organic matter decreases of more than 25% in peatlands and 0.1–0.3% in cropland mineral soils, but increases between 10–25% on reclaimed land due to land subsidence. Our analysis quantifies the substantial variations of soil organic matter in space, depth, and time, highlighting the inadequacy of evaluating soil organic matter dynamics at point scale or static mapping at a single depth for policymaking.
Site-specific estimation of lime requirement requires high-resolution maps of soil organic carbon (SOC), clay and pH. These maps can be generated with digital soil mapping models fitted on covariates ...observed by proximal soil sensors. However, the quality of the derived maps depends on the applied methodology. We assessed the effects of (i) training sample size (5–100); (ii) sampling design (simple random sampling (SRS), conditioned Latin hypercube sampling (cLHS) and k-means sampling (KM)); and (iii) prediction model (multiple linear regression (MLR) and random forest (RF)) on the prediction performance for the above mentioned three soil properties. The case study is based on conditional geostatistical simulations using 250 soil samples from a 51 ha field in Eastern Germany. Lin’s concordance correlation coefficient (CCC) and root-mean-square error (RMSE) were used to evaluate model performances. Results show that with increasing training sample sizes, relative improvements of RMSE and CCC decreased exponentially. We found the lowest median RMSE values with 100 training observations i.e., 1.73%, 0.21% and 0.3 for clay, SOC and pH, respectively. However, already with a sample size of 10, models of moderate quality (CCC > 0.65) were obtained for all three soil properties. cLHS and KM performed significantly better than SRS. MLR showed lower median RMSE values than RF for SOC and pH for smaller sample sizes, but RF outperformed MLR if at least 25–30 or 75–100 soil samples were used for SOC or pH, respectively. For clay, the median RMSE was lower with RF, regardless of sample size.
A new article type: The 'Data Article' Rossiter, David G.; Dungait, Jennifer A. J.; Mulder, Vera L. ...
European journal of soil science,
May–June 2022, 2022-05-00, 20220501, Volume:
73, Issue:
3
Journal Article
The precision agriculture scientific field employs increasingly innovative techniques to optimize inputs, maximize profitability, and reduce environmental impacts. Therefore, obtaining a high number ...of soil samples to make precision agriculture feasible is challenging. This data bottleneck has been overcome by identifying sub-regions based on data obtained through proximal soil sensing equipment. These data can be combined with freely available remote sensing data to create more accurate maps of soil properties. Furthermore, these maps can be optimally aggregated and interpreted for soil heterogeneity through management zones. Thus, this work aimed to create and combine soil management zones from proximal soil sensing and remote sensing data. To this end, data on electrical conductivity and magnetic susceptibility, both apparent, were measured using the EM38-MK2 proximal soil sensor and the contents of the thorium and uranium elements, both equivalent, via the Medusa MS1200 proximal soil sensor for a 72-ha grain-producing area in São Paulo, Brazil. The proximal soil sensing attributes were mapped using ordinary kriging (OK). Maps were also made using kriging with external drift (KED), and the proximal soil sensor attributes data, combined with remote sensing data, such as Landsat-8, Aster, and Sentinel-2 images, in addition to 10 terrain covariables derived from the digital elevation model Alos Palsar. As a result, three management zone maps were produced via the k-means clustering algorithm: using data from proximal sensors (OK), proximal sensors combined with remote sensors (KED), and remote sensors. Seventy-two samples (0–10 cm in depth) were collected and analyzed in a laboratory (1 sample per hectare) for concentrations of clay, calcium, organic carbon, and magnesium to assess the capacity of the management zone maps created using analysis of variance. All zones created using the three data groups could distinguish the different treatment areas. The three data sources used to map management zones produced similar map zones, but the zone map using a combination of proximal and remote data did not show an improvement in defining the management zones, and using only remote sensing data lowered the significance levels of differentiating each zone compared to the OK and KED maps. In summary, this study not only underscores the global applicability of proximal and remote sensing techniques in precision agriculture but also sheds light on the nuances of their integration. The study’s findings affirm the efficacy of these advanced technologies in addressing the challenges posed by soil heterogeneity, paving the way for more nuanced and site-specific agricultural practices worldwide.
In response to the growing societal awareness of the critical role of healthy soils, there has been an increasing demand for accurate and high-resolution soil information to inform national policies ...and support sustainable land management decisions. Despite advancements in digital soil mapping and initiatives like GlobalSoilMap, quantifying soil variability and its uncertainty across space, depth and time remains a challenge. Therefore, maps of key soil properties are often still missing on a national scale, which is also the case in the Netherlands. To meet this challenge and fill this data gap, we introduce BIS-4D, a high-resolution soil modeling and mapping platform for the Netherlands. BIS-4D delivers maps of soil texture (clay, silt and sand content), bulk density, pH, total nitrogen, oxalate-extractable phosphorus, cation exchange capacity and their uncertainties at 25 m resolution between 0 and 2 m depth in 3D space. Additionally, it provides maps of soil organic matter and its uncertainty in 3D space and time between 1953 and 2023 at the same resolution and depth range. The statistical model uses machine learning informed by soil observations amounting to between 3815 and 855 950, depending on the soil property, and 366 environmental covariates. We assess the accuracy of mean and median predictions using design-based statistical inference of a probability sample and location-grouped 10-fold cross validation (CV) and prediction uncertainty using the prediction interval coverage probability.
Soil compaction is a severe threat to agricultural productivity, as it can lead to yield losses ranging from 5% to 40%. Quantification of the state of compaction can help farmers and land managers to ...determine the optimal management to avoid these losses. Bulk density is often used as an indicator for compaction. It is a costly and time-consuming measurement, making it less suitable for farmers and land managers. Alternatively, measurements of penetration resistance can be used. These measurements are cheaper and quicker but are prone to uncertainty due to the existence of a wide array of thresholds. Classifications using either measurement may provide different outcomes when used in the same location, as they approximate soil compaction using different mechanisms. In this research, we assessed the level of agreement between soil compaction classifications using bulk density and penetration resistance for an agricultural field in Flevoland, the Netherlands. Additionally, we assessed the possible financial implications of misclassification. Balanced accuracy results indicate that most thresholds from the literature show around 70% agreement between both methods, with a maximum level of agreement of 76% at 1.8 and 1.9 MPa. The expected cost of misclassification shows a dip between 1.0 and 3.0 MPa, with an effect of crop value on the shape of the cost function. Although these results are specific to our study area, we believe they show that there is a substantial effect of the choice of measurement on the outcome of soil compaction studies.