Figure - available from: Remote Sensing
This content is subject to copyright.
Comparison of the results from different models in BY (top) and BW (bottom). Examples of the exposed soil composite and soil types as input covariates (A,B); Respective SOC prediction for satellite and soil models (C,D); SOC predictions of the combined models (E,F). Red dots show the locations within the research area.
Source publication
Precise knowledge about the soil organic carbon (SOC) content in cropland soils is one requirement to design and execute effective climate and food policies. In digital soil mapping (DSM), machine learning algorithms are used to predict soil properties from covariates derived from traditional soil mapping, digital elevation models, land use, and Ea...
Citations
... The data are characterized by different geometric and semantic resolutions and thus by different scalespecific explanatory power (Möller & Volk, 2015). One approach to reduce or resolve the scale-specific discrepancies is to aggregate the results on small-scaled reference units (Volk et al., 2010), as shown in Fig. 4. Currently, statewide or regional Digital Soil Mapping products are also being generated for Germany (e.g., Broeg et al., 2023;Gebauer et al., 2022;Möller et al., 2022;Sakhaee et al., 2022;Zepp et al., 2021) that may act as nationwide data bases in the future. ...
Soils provide habitat, regulation and utilization functions. Therefore, Germany aims to reduce soil sealing to 30 ha day $$^{-1}$$ - 1 by 2030 and to eliminate it by 2050. About 55 ha day $$^{-1}$$ - 1 of soil are damaged (average 2018–2021), but detailed information on its soil quality is lacking. This study proposes a new approach using geo-information and remote sensing data to assess agricultural soil loss in Lower Saxony and Brandenburg. Soil quality is assessed based on erosion resistance, runoff regulation, filter functions, yield potential and the Müncheberg Soil Quality Rating from 2006 to 2015. Data from the German Soil Map at a scale of 1:200,000 (BÜK 200), climate, topography, CORINE Land Cover (CLC) and Imperviousness Layer (IMCC), both provided by the Copernicus Land Monitoring Service (CLMS), are used to generate information on soil functions, potentials and agricultural soil loss due to sealing. For the first time, soil losses under arable land are assessed spatially, quantitatively and qualitatively. An estimate of the qualitative loss of agricultural soil in Germany between 2006 and 2015 is obtained by intersecting the soil evaluation results with the quantitative soil loss according to IMCC. Between 2006 and 2015, about 73,300 ha of land were sealed in Germany, affecting about 37,000 ha of agricultural soils. This corresponds to a sealing rate of 11 ha per day for Germany. In Lower Saxony and Brandenburg, agricultural soils were sealed at a rate of 1.9 ha day $$^{-1}$$ - 1 and 0.8 ha day $$^{-1}$$ - 1 respectively, removing these soils from primary land use. In Lower Saxony, 75% of soils with moderate or better biotic yield potential have been removed from primary land use, while in Brandenburg this figure is as high as 88%. Implementing this approach can help decision-makers reassess sealed land and support Germany’s sustainable development strategy.
... As illustrated in Fig. 5, not all the local samples are located within the prediction area. For SOC models trained on SRCs, it has been shown that additional samples can improve prediction accuracy, even if they come from outside the research area (Broeg et al., 2023). ...
National soil organic carbon (SOC) maps are essential to improve greenhouse gas accounting and support climate-smart agriculture. Large-scale SOC models based on wall-to-wall soil information from remote sensing remain a challenge due to the high diversity of natural soil conditions and the difficulty of accounting for the spatial location of the soil samples. In this study, we tested if the implementation of local ensemble models (LEM) can be used to improve the SOC predictions from Landsat-based soil reflectance composites (SRC) for Germany. For this, we divided the research area into 30 times 30 km tiles and calculated local generalized linear models (GLM) based on random, nearby observations. Based on the GLMs, local SOC maps were predicted and aggre-gated using a moving window approach. The local variable importance was analyzed to identify spatial dependencies in the correlation between the SRC and SOC. For the final SOC map, a Random Forest (RF) model was trained using the aggregated local SOC predictions, the SRC, and a full set of training samples from the agricultural soil inventory. The results show that the LEM was able to improve the accuracy (R 2 = 0.68; RMSE = 5.6 g kg − 1), compared to the maps based on a single, global model (R 2 = 0.52; RMSE = 6.8 g kg − 1). The local variable importance of the spectral bands showed clear spatial patterns throughout the research area. Differences can be explained by the local soil conditions, influencing the correlation between SOC and the spectral properties. Compared to the widely adopted integration of distance covariates such as geographical coordinates, the LEM was able the reduce the spatial autocorrelation to a greater extent and to improve the prediction accuracy, especially for underrepresented SOC values. The LEM presents a new method to integrate spatial information and increase the interpretability of DSM models.
... It should also be mentioned that the effects of soil conditions, such as soil moisture or roughness, could affect the generation of the bare-soil composite maps and, therefore, the accuracy of the prediction models [75,76], making it more difficult to map soil indicators (e.g., SOC) at large scales [77]. By using the synergy of radar and optical data, e.g., Sentinel-1 data and Sentinel-2 data [78], or by using hyperspectral data [79,80], this influence could be eliminated [81]. The incorporation of climatic data may also assist in the exclusion of time periods that correspond to precipitation events. ...
There is a growing realization among policymakers that in order to pave the way for the development of evidence-based conservation recommendations for policy, it is essential to improve the capacity for soil-health monitoring by adopting multidimensional and integrated approaches. However, the existing ready-to-use maps are characterized mainly by a coarse spatial resolution (>200 m) and information that is not up to date, making their use insufficient for the EU’s policy requirements, such as the common agricultural policy. This work, by utilizing the Soil Data Cube, which is a self-hosted custom tool, provides yearly estimations of soil thematic maps (e.g., exposed soil, soil organic carbon, clay content) covering all the agricultural area in Lithuania. The pipeline exploits various Earth observation data such as a time series of Sentinel-2 satellite imagery (2018–2022), the LUCAS (Land Use/Cover Area Frame Statistical Survey) topsoil database, the European Integrated Administration and Control System (IACS) and artificial intelligence (AI) architectures to improve the prediction accuracy as well as the spatial resolution (10 m), enabling discrimination at the parcel level. Five different prediction models were tested with the convolutional neural network (CNN) model to achieve the best accuracy for both targeted indicators (SOC and clay) related to the R2 metric (0.51 for SOC and 0.57 for clay). The model predictions supported by the prediction uncertainties based on the PIR formula (average PIR 0.48 for SOC and 0.61 for clay) provide valuable information on the model’s interpretation and stability. The model application and the final predictions of the soil indicators were carried out based on national bare-soil-reflectance composite layers, generated by employing a pixel-based composite approach to the overlaid annual bare-soil maps and by using a combination of a series of vegetation indices such as NDVI, NBR2, and SCL. The findings of this work provide new insights for the generation of soil thematic maps on a large scale, leading to more efficient and sustainable soil management, supporting policymakers and the agri-food private sector.
... As illustrated in Fig. 5, not all the local samples are located within the prediction area. For SOC models trained on SRCs, it has been shown that additional samples can improve prediction accuracy, even if they come from outside the research area (Broeg et al., 2023). ...
... As illustrated inFigure 5, not all the local samples are located within the prediction area. For SOC 286 models trained on SRCs, it has been shown that additional samples can improve prediction accuracy, even if they 287 come from outside the research area(Broeg et al., 2023). ...
... Using NDVI thresholds, it is impossible to single out the BSS without capturing three or four adjacent spectral areas. At the same time, in several studies, BSS was distinguished precisely by the NDVI thresholds [8,36,37]. It remains to be assumed that the colleagues do not have problems with the detection of areas with crop residues, clouds, fires, and waterlogging. ...
... Of the seven works [6,8,9,30,[36][37][38], three used VI [6,9,30]. In the work of 2021 [7], we showed that it is possible to apply VI to correct soil maps, but this is not enough. ...
... In several works [30,37,45,53], the application of multitemporal RSD series in the topic of agriculture and landscape classification is studied, which brings together the directions of our research. Of particular interest is the approach of applying multitemporal series of RSD with the use of the BSS spectral characteristics [37,45]. ...
For most of the arable land in Russia (132–137 million ha), the dominant and accurate soil information is stored in the form of map archives on paper without coordinate reference. The last traditional soil map(s) (TSM, TSMs) were created over 30 years ago. Traditional and/or archival soil map(s) (ASM, ASMs) are outdated in terms of storage formats, dates, and methods of production. The technology of constructing a multitemporal soil line (MSL) makes it possible to update ASMs and TSMs based on the processing of big remote-sensing data (RSD). To construct an MSL, the spectral characteristics of the bare soil surface (BSS) are used. The BSS on RSD is distinguished within the framework of the conceptual apparatus of the spectral neighborhood of the soil line. The filtering of big RSD is based on deep machine learning. In the course of the work, a vector georeferenced version of the ASM and an updated soil map were created based on the coefficient “C” of the MSL. The maps were verified based on field surveys (76 soil pits). The updated map is called the map of soil interpretation of the coefficient “C” (SIC “C”). The SIC “C” map has a more detailed legend compared to the ASM (7 sections/chapters instead of 5), greater accuracy (smaller errors of the first and second kind), and potential suitability for calculating soil organic matter/carbon (SOM/SOC) reserves (soil types/areals in the SIC “C” map are statistically significant are divided according to the thickness of the organomineral horizon and the content of SOM in the plowed layer). When updating, a systematic underestimation of the numbers of contours and areas of soils with manifestations of negative/degradation soil processes (slitization and erosion) on the TSM was established. In the process of updating, all three shortcomings of the ASMs/TSMs (archaic storage, dates, and methods of creation) were eliminated. The SIC “C” map is digital (thematic raster), modern, and created based on big data processing methods. For the first time, the actualization of the soil map was carried out based on the MSL characteristics (coefficient “C”).
... The transfer learning approach has been demonstrated to be applicable in pedometrics, especially in studies on the prediction of soil properties by creating and using spectral reflectance libraries [45][46][47]. By integrating the transfer learning approach, relevant DSM studies were conducted, such as the parent material [48], organic carbon at the local scale [49], USDA Soil Taxonomy at the sub-group level [50], USDA Soil Taxonomy at the soil great group level [51], soil organic carbon in cropland soils [52], and soil particle fractions [53]. ...
A comprehensive understanding of soil salinity distribution in arid regions is essential for making informed decisions regarding agricultural suitability, water resource management, and land use planning. A methodology was developed to identify soil salinity in Sudan by utilizing optical and radar-based satellite data as well as variables obtained from digital elevation models that are known to indicate variations in soil salinity. The methodology includes the transfer of models to areas where similar conditions prevail. A geographically coordinated database was established, incorporating a variety of environmental variables based on Google Earth Engine (GEE) and Electrical Conductivity (EC) measurements from the saturation extract of soil samples collected at three different depths (0–30, 30–60, and 60–90 cm). Thereafter, Multinomial Logistic Regression (MNLR) and Gradient Boosting Algorithm (GBM), were utilized to spatially classify the salinity levels in the region. To determine the applicability of the model trained at the reference site to the target area, a Multivariate Environmental Similarity Surface (MESS) analysis was conducted. The producer’s accuracy, user’s accuracy, and Tau index parameters were used to evaluate the model’s accuracy, and spatial confusion indices were computed to assess uncertainty. At different soil depths, Tau index values for the reference area ranged from 0.38 to 0.77, whereas values for target area samples ranged from 0.66 to 0.88, decreasing as the depth increased. Clay normalized ratio (CLNR), Salinity Index 1, and SAR data were important variables in the modeling. It was found that the subsoils in the middle and northwest regions of both the reference and target areas had a higher salinity level compared to the topsoil. This study highlighted the effectiveness of model transfer as a means of identifying and evaluating the management of regions facing significant salinity-related challenges. This approach can be instrumental in identifying alternative areas suitable for agricultural activities at a regional level.
... MLR models of soil properties (Al + ½ Fe extracted by AO, bulk density, and phosphate retention) and raster covariate data (see Figure 4) were determined for the 16 sampled pedons and for a combined set of the 16 sampled pedons and the 18 NRCS pedons using the Best Subsets function called "regsubsets" (Lumley based on Fortran code by Alan Miller, 2020). A similar approach was taken by Broeg et al. (2023) when comparing model effectiveness by training their soil property models with a dataset from one German state, validating the model with a dataset from another German state, and then training the model with a mixed-dataset and applying the model to both states. The variance inflation factor (VIF) was then determined using the "vif" function (Fox & Weisberg, 2019) to quantify the extent of correlation between one predictor and the others in a MLR model. ...
Maui is one of five Hawaiian Islands affected by orographic climate effect, exhibiting a massive precipitation gradient across western Haleakalā. However, high variability of volcanic ash deposits as a parent material across the study area complicates the ability to isolate the influence of climate on soil formation. Little is documented about the spatial extent of ash deposition, frequency and intensity of volcanic ejecta events, and composition of ash. Therefore, andic soils, which contain short range order (SRO) aluminosilicates and iron oxides that result in unique soil chemical and physical properties, are challenging to map. Using environmental and andic property data from 16 pedons sampled in the study area—bulk density, phosphate retention, and aluminum plus ½ iron extracted by ammonium oxalate—we applied multiple linear regression to create spatial prediction models of these three soil properties. Predicted soil properties were then used to classify andic soils. The mean prediction for an independent set of pedons showed a soil classification accuracy of 50% in the study area for Andisols (data to 60 cm), andic intergrades (data to 75 cm), and non‐andic soils. Soil property predictions using depth‐weighted average data to 1 m increased soil classification user accuracy of Andisols to 87.5%, andic‐intergrades to 100%, and non‐andic soils to 83.3%. Whether a soil exhibits andic soil properties within 60 or 75 cm is irrelevant when considering prior or current presence of ash in a soil. Accounting for all available pedon data with depth proves most important when attempting to predict andic properties and classes.
... Considering technical advancements in cost-effective smart sampling techniques for assessing soil sampling density in heterogeneous areas, careful consideration of appropriate sampling strategies is imperative prior to undertaking DSM [81]. The methodology can be advanced by checking the digital covariate values (for example, temperature, solar radiation, eleva-tion, and NDVI) in the soil sampling points provided as training data with the covariate values throughout the study area, especially in large areas (such as our study area), with techniques such as Multivariate Environmental Similarity Surfaces (MESS) [82][83][84] before modeling. In future research, it is recommended to use multispectral remote sensing data with higher spatial resolution [23] as well as various ancillary data such as synthetic aperture radar (SAR), especially to represent vegetation. ...
Cation exchange capacity (CEC) is a soil property that significantly determines nutrient availability and effectiveness of fertilizer applied in lands under different managements. CEC's accurate and high-resolution spatial information is needed for the sustainability of agricultural management on farms in the Nagaland state (northeast India) which are fragmented and intertwined with the forest ecosystem. The current study applied the digital soil mapping (DSM) methodology, based on the CEC values determined in soil samples obtained from 305 points in the region, which is mountainous and difficult to access. Firstly, digital auxiliary data were obtained from three open-access sources, including indices generated from the time series Landsat 8 OLI satellite, topographic variables derived from a digital elevation model (DEM), and the WorldClim dataset. Furthermore, the CEC values and the auxiliary were used data to model Lasso regression (LR), stochastic gradient boosting (GBM), support vector regression (SVR), random forest (RF), and K-nearest neighbors (KNN) machine learning (ML) algorithms were systematically compared in the R-Core Environment Program. Model performance were evaluated with the square root mean error (RMSE), determination coefficient (R 2), and mean absolute error (MAE) of 10-fold cross-validation (CV). The lowest RMSE was obtained by the RF algorithm with 4.12 cmol c kg −1 , while the others were in the following order: SVR (4.27 cmol c kg −1) <KNN (4.45 cmol c kg −1) <LR (4.67 cmol c kg −1) <GBM (5.07 cmol c kg −1). In particular, WorldClim-based climate covariates such as annual mean temperature (BIO-1), annual precipitation (BIO-12), elevation, and solar radiation were the most important variables in all algorithms. High uncertainty (SD) values have been found in areas with low soil sampling density and this finding is to be considered in future soil surveys.
... Soil surfaces on agricultural land are subject to greater human interference than other soil surfaces [15] and are thus an important source of dust emissions from wind erosion. The loss of organic carbon from agricultural soils is a huge challenge for coping with the greenhouse effect and represents the loss of basic land materials [29]. ...
Wind erosion can cause high dust emissions from agricultural land and can lead to a significant loss of carbon and nutrients from the soil. The carbon balance of farmland soil is an integral part of the carbon cycle, especially under the current drive to develop carbon-neutral practices. However, the amount of global carbon lost due to the wind erosion of farmland is unknown. In this study, global farmland dust emissions were estimated from a dust emission inventory (0.1° × 0.1°, daily) built using the improved Community Multiscale Air Quality Modeling System–FENGSHA (CMAQ-FENGSHA), and global farmland organic carbon losses were estimated by combining this with global soil organic carbon concentration data. The average global annual dust emissions from agricultural land from 2017 to 2021 were 1.75 × 109 g/s. Global dust emissions from agricultural land are concentrated in the UK, Ukraine, and Russia in Europe; in southern Canada and the central US in North America; in the area around Buenos Aires, the capital of Argentina, in South America; and in northeast China in Asia. The global average annual organic carbon loss from agricultural land was 2970 Gg for 2017–2021. The spatial distribution of emissions is roughly consistent with that of dust emissions, which are mainly concentrated in the world’s four major black soil regions. These estimates of dust and organic carbon losses from agricultural land are essential references that can inform the global responses to the carbon cycle, dust emissions, and black soil conservation.
