SSS10.4
Presentations: Fri, 27 May | Room G1
Soil texture is a key information in precision agriculture for improving soil knowledge and crop performances. A precise mapping of its variability is thereby imperative for rationally planning cultivations and targeting interventions. Unlike direct soil texture measurements that are punctual, destructive, and time-consuming, remote sensing surveys can give widespread, non-invasive, and fast indirect evidence of clay, silt, and sand content. In this study we investigate the performance of Airborne Gamma Ray Spectroscopy (AGRS) for discriminating different texture classes in the ternary diagram of soil texture.
The Mezzano valley (Ferrara, Italy), a 180 km2 rural area reclaimed in the last century, represents an extraordinary benchmark for validating our method. This area, for which a public soil texture map at 1:50000 scale and a spatial resolution of 500 m is available, was scanned by an AGRS system mounted on a dedicated aircraft. The aircraft flew over the study area in a grid-like path of ~500 m spacing, collecting 1469 geolocalized spectra. The K and Th punctual measurements were spatially interpolated by Ordinary Kriging to elaborate K and Th maps with the identical spatial resolution of the soil texture map. Simple and multiple linear correlations, as well as a non‑linear Machine Learning algorithm, were then performed between gamma and soil texture data.
The obtained results by a simple linear regression analysis highlight a moderate positive (negative) correlation between clay (sand) content and K and Th abundances. Multiple linear regressions show a similar trend, with the limitation that the calculated clay, silt, and sand values populate the soil texture ternary diagram in a straight line. Finally, we demonstrate that the most accurate reconstruction of soil texture values is obtained by a non-linear fitting based on the Machine Learning algorithm.
How to cite: Maino, A., Alberi, M., Anceschi, E., Chiarelli, E., Cicala, L., Colonna, T., De Cesare, M., Guastaldi, E., Lopane, N., Mantovani, F., Martini, N., Montuschi, M., Piccioli, S., Raptis, K. G. C., Russo, A., Semenza, F., and Strati, V.: Mapping soil texture with airborne gamma ray spectroscopy, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-361, https://doi.org/10.5194/egusphere-egu22-361, 2022.
The evaluation of shale gas in place by adsorption and free gas is critical for future shale gas reserves development. Laboratory experiments suggest that approximately 50% of the stored shale gas is adsorbed onto the kerogen. Molecular simulation researches suggest methane adsorption capacity increases with the increasing maturity of kerogen. Understanding how thermal maturity controls methane adsorption in kerogen is crucial for predicting shale gas resources. Although porosity and chemical surface functionalities (sorption sites) are the main difference between kerogens with different maturity, the study of their impact on methane adsorption via both simulation and experiment methods is in the preliminary stage.
The comparison of molecular simulations on Type II kerogen matrix with slit models and laboratory experiments on isolated kerogens are carried out to illustrate their impact on methane adsorption in kerogen with different maturity, and reveal the predominant controlling factor. Grand Canonical Monte Carlo (GCMC) and molecular dynamic (MD) simulations are applied to obtain simulation results, including the micropore texture, methane adsorption capacity, and adsorption behavior. Laboratory experiments, high-pressure methane adsorption, low-pressure gas sorption, scanning electron microscopy, and Transmission Electron Microscope, are carried out on isolated kerogens for verifying and comparing with simulation results.
The results indicate micropore volume (Vmicro) and equilibrium methane adsorption amount (Qm) of isolated kerogens (10-75 mm3/g TOC, and 21.3-75.8 mg/g TOC) are comparable with simulated overmature (KIID) matrix and slit kerogens (19-261 mm3/g TOC, and 36.5-148 mg/g TOC). The higher results from the simulation are due to the pore interconnectivity is not considered in simulated kerogens. Both experiment and simulation suggest Type I (a) isotherms are contributed by small micropores, and Type I (b) isotherms are contributed by larger micropores. The methane adsorption capacity of the kerogen matrix increases with increased maturity and decreases with increased temperature. A positive correlation between Vmicro and Qm is observed with R2>0.96. The relative number density and relative coordination number of methane around functional groups at 25 and 100 °C from molecular dynamic (MD) simulation show methane only have selectivity with few functional groups at very low pressure (<1.6 bar at 25 °C), and the affinity becomes close and weaker at higher pressure. Moreover, similar adsorption heats (23.2, 23.1, 23.5, 22.8 KJ/mol) of methane with different maturity kerogens are observed, showing the interactions between methane and different kerogens are close. Therefore, the impact of functional groups on the methane adsorption capacity is minimal, especially in high-pressure conditions, and micropore is regarded as the key control for methane adsorption.
How to cite: Li, W., Stevens, L., Zheng, D., and Snape, C.: A molecular simulation and laboratory characterization study of micropore and sorption sites impact on methane adsorption in kerogen, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-758, https://doi.org/10.5194/egusphere-egu22-758, 2022.
SOC prediction from remote sensing is often hindered by disturbing factors at the soil surface, such as photosynthetic active and non-photosynthetic active vegetation, variation in soil moisture or surface roughness. With the increasing amount of freely available satellite data, recent studies have focused on stabilizing the soil reflectance by building reflectance composites using time series of images. Even if SOC predictions from composite images are promising, it is still not well established if the resulting composite spectra mirror the reflectance fingerprint of the optimal conditions to predict topsoil properties (i.e. a smooth, dry and bare soil).
We have collected 303 photos of soil surfaces in the Belgium loam belt where five main classes of surface conditions were distinguished: smooth seeded soils, soil crusts, vegetation, moist soils and soils covered by crop residues. Reflectance spectra were then extracted from the Sentinel-2 images coinciding with the date of the photos. The Normalized Burn Ratio (NBR2) was calculated to characterize the soil surface, and a threshold of NBR2 < 0.05 was found to be able to separate wet soils and soils covered by crop residues from dry bare soils. Additionally, we found that normalizing the spectra (i.e. dividing the reflectance of each band by the mean reflectance of all spectral bands) allows for cancelling the albedo shift between soil crusts and smooth soils in seed-bed conditions. We then built the exposed soil composite from Sentinel-2 imagery (covering the spring periods of 2016-2021), and used the reflectance information to predict SOC content by means of a Partial Least Square Regression Model (PLSR) with 10-fold cross-validation. The uncertainty of the models (expressed as q0.05+q0.95/q0.50) was assessed via bootstrapping, where each model was repeated 100 times with a slightly different calibration dataset. The cross validation of the model gave satisfactory results (R² = 0.49 ± 0.10, RMSE = 3.4 ± 0.6 g C kg-1 and RPD = 1.4 ± 0.2). The resulting SOC prediction maps show that (1) the uncertainty of prediction decreases when the number of scenes per pixel increases, and reaches a minimum when more than six scenes per pixel are used (median uncertainty of all pixels is 28% of predicted SOC value) and (2) the uncertainty of prediction diminishes if SOC predictions are aggregated per field (median uncertainty of fields is 22% of predicted value). The results of a validation against an independent data set showed a median difference of 0.5 g C kg-1 ± 2.8 g C kg-1 SOC between the measured and predicted SOC contents at field scale. Overall, this compositing method shows both realistic SOC patterns at the field scale and regional patterns corresponding to the ones reported in the literature.
How to cite: Dvorakova, K., Heiden, U., Pepers, K., Staats, G., van Os, G., Ferber, F., and van Wesemael, B.: Improving SOC predictions from Sentinel-2 soil composites by assessing surface conditions and uncertainties, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-1618, https://doi.org/10.5194/egusphere-egu22-1618, 2022.
Soil properties could be assessed with reflectance spectroscopy (soil spectroscopy, SS) in vis-NIR region (400-2500 nm) through absorption features found in soil spectra. A high spectral resolution (up to 1 nm) drives to high dimensional and multicollinear data. This issue is usually addressed prior to modelling with feature extraction methods such as Principal Component Analysis, or embedded methods such as Partial Least Squares Regression (PLSR). Feature Selection (FS) wrapper methods are promising dimensionality reduction approaches barely used in SS. The objective of this study was two-fold: i) evaluate the performance of FS wrapper methods built from Random Forest (RF) algorithm to predict soil organic matter (SOM), clay and carbonates using laboratory spectroscopy, ii) test the performance of FS methods for dimensionality reduction in SS. The reflectance of 100 soil samples from Sierra de las Nieves National Park (Spain), was measured under laboratory conditions using an ASD FieldSpec Pro JR. A spectral preprocessing method, Continumm Removal (CR), was applied to raw spectra. The RF wrapper considered two different feature searching approaches: Sequential Forward Selection (SFS) and Sequential Flotant Forward Selection (SFFS). The performance of RF with FS (RF-FS) was compared to that of Partial Least Squares Regression (PLSR) and RF (without FS). Models were evaluated with R-squared, root mean squared error (RMSE) and ratio of prediction to deviation (RPD).
RF-FS models outperformed PLSR and RF models for the three SAP. RF-FS best models had a RPD of 2.19 for SOM, 1.64 for carbonates and 1.52 for clay, whereas PLSR models had RPD values of 1.59, 1.22 and 1.3, and RF 1.38, 1.23 and 1.23 for SOM, carbonates, and clay, respectively. Therefore, FS was useful in obtaining models with improved accuracy by reducing redundant features and avoiding multicollinearity (Hughes effect). The application of FS wrapper methods reduced the number of features in the RF-FS models to less than 1% of the starting features. Features were selected across all spectra from SOM and clay, and around 900, 1900 and 2350 nm for carbonates. This research, thus, shows an alternative to different feature extraction approaches for modelling soil properties based on FS methods and machine learning.
How to cite: Canero, F. M., Cardenas-Martinez, A., Aragones, D., and Rodriguez-Galiano, V.: An assessment of Random Forest wrappers for selecting important features of spectroscopy data in the modelling of soil properties, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2284, https://doi.org/10.5194/egusphere-egu22-2284, 2022.
Digital soil mapping approaches predict soil properties based on the relationship between soil observations and related environmental covariates using machine learning models. In this research, we applied deep neural networks to predict the spatial distribution of soil properties in Germany using 1976 soil observations and 170 environmental covariates which are derived from several sources (e.g., remote sensing data). However, a major problem with using deep neural networks is that the exact contribution of environmental covariates in the overall result is unknown. To address this issue and improve the interpretability of deep neural networks, several model-agnostic interpretation tools (i.e., post hoc analyses and techniques) are used to understand previously trained "black-box models" or their predictions. For example, a permutation feature importance technique ranked remote sensing images as the most important predictors to explain the spatial variability of soil organic carbon in the study area. This is the first study to use deep neural networks with explainable algorithms to explore and visualize the spatial distribution of soil properties in Germany.
Keywords: explainable machine learning; deep neural networks; soil properties; Germany
How to cite: Taghizadeh-Mehrjardi, R. and Scholten, T.: Explainable deep neural networks for exploring spatial variability of soil properties in Germany, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2445, https://doi.org/10.5194/egusphere-egu22-2445, 2022.
Artificial neural networks (ANN), which are mainly used in pattern and image recognition, have now found a wide range of applications. In recent years, different variants of ANN have also been increasingly used in the geosciences. They have proven to be a useful tool for complex questions that also involve a large amount of data. In their basic form, however, deep-learning algorithms do not provide interpretable predictive uncertainty. In the geosciences in particular, they have been used more as black-box models that require interpretation by an expert or do not allow for specific interpretation. Therefore, we implement in our explorative study on soil classification a Bayesian deep learning approach (i.e. a method to add uncertainty to deep networks) known as last layer Laplace approximation. This is a technique that can be applied as a post-hoc "add-on" without destroying the otherwise good performance of deep classifiers.
Our target soil type variable provides us with a large amount of information about soil processes and properties, which is a great advantage since it would take a lot of time and money to collect all this information individually. At the same time, soil maps are in high demand by authorities, construction companies or farmers. In our study area around Tübingen in southern Germany, there are 39 different soil types, determined according to the German soil systematics, which we consider individually for the prediction, but also combine into superordinate categories with similar properties, which is possible at low computational cost under the Laplace approximation. In addition to the underlying soil map, remotely sensed variables such as satellite imagery, a digital elevation model and its derivatives, and climate data are used as input to the model, which is designed to learn the relationship between these and the soil type. As a test case, we then explicitly include the Swabian Jura as a prediction region for the environment. This region is characterised by very different soil types due to its extremely different development and the resulting geology, climate and terrain.
Our goal is then to enrich soil type maps with a structured uncertainty, which is estimated to be high in the area of the Swabian Jura. This will help to better understand the causality of machine learning models in soil science and their transferability to regions other than the training and validation area.
How to cite: Rau, K., Gläßle, T., Hennig, P., and Scholten, T.: Spatial prediction of soil type maps with Neural Networks including quantification of model uncertainty, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2766, https://doi.org/10.5194/egusphere-egu22-2766, 2022.
Accurate soil organic carbon content estimation is critical as a proxy for carbon sequestration, and as one of the indicators for soil health. Here, we collected 497 soil samples during 2015 to 2019, as well as five environmental covariates (organic carbon (OC) input, normalized difference vegetation index (NDVI), elevation, clay and precipitation) at a resolution of 30 m, aggregated these to represent agricultural fields and then compiled a soil organic carbon (SOC) content map for the agricultural region of Wallonia using Gradient Boosting Machine. We calculated OC input from both main crops and cover crops for each individual field. As the cover crops do not occur in the agricultural census, we identified cover crops based on long time-series of NDVI values obtained from the Google Earth Engine platform. The quality of the predictions was assessed by independent validation and we obtained an R2 of 0.77. The Empirical Mode Decomposition indicated that OC input and NDVI were the domain factors at field scale, whereas the remainder of the covariates determined the distribution of SOC at the scale of the entire Walloon region. The SOC map showed an overall northwest to southeast trend i.e. an increase in SOC contents up to the Sambre-Meuse valley followed by a decrease further to the South. The map shows both regional trends in SOC and effects of differences in land use and/or management (including crop rotation and frequency of cover crops) between individual fields. The field-scale map can be used as a benchmark and reference to farmers and agencies in monitoring SOC content changes and optimizing decisions for sustainable land use.
How to cite: Zhou, Y., Chartin, C., Van Oost, K., and van Wesemael, B.: High-resolution soil organic carbon mapping at the field scale in Southern Belgium (Wallonia) , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-3258, https://doi.org/10.5194/egusphere-egu22-3258, 2022.
The intensive use of soil and the non-adoption of optimal management practices leads to the loss of soil organic carbon (SOC) from soil. SOC accumulates in the atmosphere in the form of CO2, thus affecting the global temperature. Numerous studies have been carried out in the monitoring of SOC in exposed croplands at global and regional scales, demonstrating the potential of remote sensing to estimate SOC amongst the disturbance effects encountered on the Earth Observation monitoring that affect the prediction of soil properties, soil moisture ranks within the most important. The current study is driven by the need to eliminate the influence of ambient factors and evaluate the efficiency of multitemporal analysis by leveraging numerical simulations of soil moisture data as an auxiliary variable along with Sentinel-2’s reflectance values. Multi-year high-resolution coupled atmospheric-soil numerical simulations were utilized from BEYOND/NOA’s operational implementation of Weather Research and Forecasting Model (WRF-ARW) on a 2-km grid spacing configuration over Greece. Spectral data were extracted using Sentinel 2 multitemporal imagery (February to December 2020) at the sampling points of the European topsoil dataset provided by the Land Use / Coverage Framework Survey (LUCAS) 2009 and 2015 in Greek croplands with the support of Google Earth Engine, totaling 643 sampling points. After that, bare soil masking was performed, using as a limiting factor the values between 0 and 0.25 to NDVI, NBR2 < 0.08 and the difference between B3 and B2, resulting in 180 sampling points which had exposed bare soil at any given time in the aforementioned period. The SOC prediction was performed using Sentinel 2 multitemporal bands together with soil moisture. Datasets were randomly separated in calibration (75%) and validation samples (25%). Cubist regression algorithm was applied to train predictive models in three separate modeling modes: multitemporal Sentinel-2 bands averages (S2mean), multitemporal Sentinel-2 bands (S2multitemporal) and multitemporal Sentinel-2 bands and soil moisture (S2+M). Model performance to the multitemporal modes (S2 and S2+M) was measured by averaging the predicted values for each sampling point. Mode S2+M achieved the best accuracy among the modes, reaching an R2 of 0.68, RMSE of 9.19 and RPIQ of 1.21 , while the S2multitemporal mode had a R2 of 0.62, RMSE of 9.91 and RPIQ of 1.12 and the S2mean with R2 of 0.31, RMSE of 12.58 and RPIQ of 0.87. The modes with multitemporal data proved to be more powerful for SOC prediction than the mode with average spectral values, due to the large amount of spectral information for each sample. The use of NWP-derived soil moisture as an auxiliary variable improved the performance of SOC estimation, due to the direct influence of soil moisture on SOC rates. Therefore, this study indicates that multitemporal Sentinel-2 imagery and NWP-derived soil moisture information can improve the accuracy of SOC prediction. Further investigation is currently focused upon including additional soil-climate variables as well as test different combinations of thresholds in bare soil masking towards a better performance in the prediction of this soil property.
How to cite: Coblinski, J., Bartsotas, N., Tziolas, N., Tsakiridis, N., Kontoes, C., and Zalidis, G.: Prediction of soil organic carbon content using multitemporal Sentinel-2 imagery data and NWP-derived soil moisture over Greek croplands, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-3873, https://doi.org/10.5194/egusphere-egu22-3873, 2022.
Relevant soil information at different scales would greatly help addressing many of the Sustainable Development Goals. Digital Soil Mapping is an established methodology to create maps of soil properties at different resolutions and extents. Many projects across the globe have provided information on primary soil properties, such as soil textural fractions, soil organic carbon content, cation exchange capacity and soil pH. For environmental modelling and assessment, maps of complex soil properties are also important. These can be defined as properties that cannot be measured directly in the laboratory but are derived from primary soil properties, for instance by simple calculations, pedotransfer functions or more advanced spatial analyses. Examples are available water capacity, soil carbon density and stocks, as well as soil erodibility. There are two main approaches to map complex properties: 1) “model first, interpolate later”, where the complex property is first calculated at point locations where the primary properties are known and then mapped; and 2) “interpolate first, model later”, where the complex property is calculated from maps of the primary properties contributing to it.
We present and discuss these two approaches for global applications using legacy data with a non-uniformspatial distribution of observations and the SoilGrids workflow. We compare the results for available water capacity of the 0 to 100 cm depth interval and soil carbon densities for six depth layers. Both properties were derived from a combination of simple calculations for point locations where the input soil properties were available and pedotransfer functions for other point locations where basic soil properties were available. There were substantial differences between the “model first, interpolate later” and “interpolate first, model later” approaches, both in point-wise evaluation metrics and in landscape patterns.
How to cite: Poggio, L., Kempen, B., Turek, M.-E., Batjes, N. H., Genova, G., de Sousa, L. M., Rossiter, D., and Heuvelink, G.: Mapping of complex soil properties at global scale , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4311, https://doi.org/10.5194/egusphere-egu22-4311, 2022.
Parent material is an essential soil property, whose mapping is a challenging task, since parent material – landscape models are even less established quantitatively, then those used in traditional soil mapping, due to more difficult and indirect cognizability of the deeper layers from surface. For the compilation of a reliable parent material map with quantified accuracy a digital mapping method was elaborated.
- Disaggregation of legacy geology map by RF modelling
- Spatially predicting the reliability of the disaggregated map
- Spatial identification of less reliable, stable predictions
- Elaboration of a sampling design
- Field work, collecting spatially non-exhaustive field observation (visual)
- Interpretation of the newly collected data
- Testing the improvement in the performance of the digital parent material map by involving increasing number of ground truth data
With the use of remotely sensed data and machine learning a new, large scale parent material was complied in an old mining region of Hungary. Different scale existing geological maps were used for training and for testing the classification concerning the lithological composition. To predict the parent material we applied various machine learning methods (Random Forest, Support Vector Machine and Conditional Neural network) using data originating from Earth Observation as ancillary information. Satellite imagery data, was used both in form of native spectral bands and derived spectral indices. Various derivatives of SRTM provided morphological auxiliary data. Digital soil property maps were also introduced into the modelling process. Finally 63 predictors were applied.
We examined the importance of each variable and we found that the data of the morphometric variables (e.g. MRVBF, elevation, slope) and some soil particle size fractions (i.e. clay, silt, sand) are the most important ones, compared to the rest of the tested spectral variables.
The resulting classified maps were validated several ways:
- after the first results, we run some analysis on the predicted map to examine its overall accuracy, and it equals to 0.77;
- we checked the difference of the predicted and the original maps;
- we also examined the number of predicted unique value of each pixel and the percentage of the most frequently predicted value.
In the next step field work was organized for the collection of spatially non-exhaustive field observation. A sampling design was elaborated based the evaluation results and taking into consideration the fact that there quite a few outcrops in the area which could help our work. After interpreting the newly collected data the improvement in the performance of the digital parent material map are being tested by involving increasing number of ground truth data.
Our paper will present the most recent results.
Acknowledgment: Our research was supported by the Hungarian National Research, Development and Innovation Office (NKFIH; K-131820)
How to cite: Takáts, T., Mészáros, J., Albert, G., Kovács, Z. A., and Pásztor, L.: Evaluation and improvement of the predictivity of a digital parent material map, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4822, https://doi.org/10.5194/egusphere-egu22-4822, 2022.
Digital Soil Mapping (DSM) is an established methodology to create maps of soil properties at different resolutions and extents. It establishes a statistical relationship between the measured values at point observations and environmental covariates selected to describe the soil forming factors and to explain the spatial variability of the soil properties. These relationships are then used to map the target soil properties across the area of interest. In this example, we used 1423 measurements on soil organic carbon and pH for the 0-20 cm soil layer from a Mongolian soil survey. This survey was organised within the framework of the “National Program to Combat Desertification” to determine the primary soil quality indicators for desertification assessment in Mongolia and it was conducted based on the state network of the Meteorological and Environmental Research Agency starting in 2012. The samples are collected from 1500 monitoring points every 5 years. We used data from the monitoring round between 2012 and 2015 We used two sets of covariates for modelling predictive relationships. The first is the set used in SoilGrids at 250 m resolution with over 400 layers available of which about 180 were used for the modelling, after de-correlation. The second is a reduced set of about 40 covariates at 100 m resolution derived mainly from Sentinel (1 and 2) images, ERA5 for climate data and ALOS for morphological information. In this study we will compare the results of the two models, with both point-wise evaluation matrices and assessment of spatial patterns. In the evaluation the expertise of local partners will also be used.
How to cite: Genova, G., Poggio, L., Kempen, B., Baatar, U.-O., Sarangerel, E., and Ganbold, U.: DSM with covariates at different resolution for Mongolia , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5101, https://doi.org/10.5194/egusphere-egu22-5101, 2022.
The European Green Deal emphasizes the importance of healthy soils for our planet and society. In order to monitor soil health, modelling soil organic matter (SOM) in space and over time is necessary to assess changes in the fertility of agricultural soils, combat climate change and maintain ecosystem services. In digital soil mapping, most recent statistical modelling approaches have used time series of remote sensing data, which became available from the early 1980s onwards, together with soil observations to make predictions in space and over time. While this has clear advantages, it does not provide spatially explicit, explanatory information for time periods before the 1980s, even though observations in soil databases are often from beforehand. In this study, we modelled changes in SOM in 3D space over 65 years on a national scale in the Netherlands. We used SOM observations from 345 000 locations from 0 to 2 m depth between 1953 and 2018. The covariates were comprised of proxies of soil forming factors either considered to be static (e.g. relief, parent material) or dynamic over these 65 years. As dynamic covariates, we used indices of digitized historic (1960 – 1980) and more recent (1986 – 2018) land use maps. These dynamic covariates were chosen for two reasons. Firstly, land use and land cover change are the main drivers of SOM change over the time period of several decades. This is especially true in the Netherlands, where the anthropogenic influence on soils has been tremendous. Approximately 82 % of the land surface are agricultural, urban or infrastructure areas, while 15 % consists of (managed) peatlands and up to 20 % has been reclaimed from the sea. Secondly, by including carefully mapped historic land use, we were able to take advantage of a longer time series of soil data to make space-time predictions of SOM over a longer time period. Predictions were made using the quantile regression forest (QRF) algorithm, whereby sampling depth and year were included during calibration. SOM predictions were validated in two ways: a) over the 65-year period using a 10-fold cross-validation and b) specifically for 1998 and 2018, where designed-based statistical inference was possible using a probability sample. We computed the mean error (ME), root mean squared error (RMSE), model efficiency coefficient (MEC) as accuracy metrics and the prediction interval coverage probability (PICP) as an evaluation of the prediction uncertainty. Results showed that spatial patterns were realistic and properly reproduced but that prediction of temporal dynamics was more challenging. This research is also of interest for spatio-temporal soil modelling in other regions of the world that have soil data from the early and mid-19th century and historical land use and land cover data.
How to cite: Helfenstein, A., Mulder, V. L., Heuvelink, G. B. M., and Hack-ten Broeke, M. J. D.: Machine learning in four dimensions for mapping soil organic matter changes between 1953-2018 at 25m resolution in the Netherlands, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5624, https://doi.org/10.5194/egusphere-egu22-5624, 2022.
Urban soils have a high capacity to accumulate C, whereas urban soil-C stocks are exposed to multiple direct or indirectly anthropogenic effects and therefore very variable and dynamic. The intracity variability of soil-C stocks is affected by functional and historical zoning, land management and mesoclimatic anomalies (e.g., urban heat island). The intercity variability is likely explained by the difference in regional climate conditions. The current research aimed to analyze variability in urban soil C stocks by a comparative analysis of six settlements following the climatic gradient in Central European Russia.
In 2019-2021 urban soil-C stocks were observed in the residential areas of six settlements representing different bioclimatic zones: Murmansk (68N, 33E; forest-tundra), Apatity (67N, 33E; north taiga), Moscow (55N, 37E; south taiga), Pushchino (54N, 37E; mixed and deciduous forests), Kursk (51N, 36E; forest-steppe) and Rostov-on-Don (47N, 39E; dry steppe). In each settlement, total 50 locations were selected following a random stratified scheme and mixed soil samples from the depths 0-10, 10-20, 30-30 and 30-50 сm were collected in each location. Soil organic (SOC) and inorganic (SIC) C stocks (all the depths) as well as microbial (basal) respiration and half-life time (only 0-10 cm) were analyzed in the collected samples. The intracity variation was investigated and mapped by digital soil mapping techniques linking field data to conventional (i.e., vegetation, relief and parent materials) and urban-specific (i.e., historical zoning and distances to infrastructures) covariates. In result, spatial variability and profile distribution of SOC and SIC were analyzed.
Total C stocks ranged from 15 kgC m-2 in Pushchino to over 30 kgC m-2 in Rostov-on-Don and Kursk. The highest contribution of subsoil (below 30 cm) layers was shown for the south settlements, where urban soils were often formed on top of the buried Chernozems, whereas in the polar climate the role of subsoil in total urban C stocks was much smaller. The outcomes confirmed that for topsoils urban C stocks were higher than in the natural zonal soils for the northern sites (Moscow and Apatity), whereas the opposite was shown for the settlements to the south from Pushchino. Half-life time of organic matter decreased from almost 30 years in Murmansk to less than 5 years in Rostov-on-Don. This illustrates law resistance of urban soil-C stocks to microbial decomposition under warm climate conditions. The contribution of organic and inorganic C to the total C stocks also clearly followed climate gradient and SIC share in Rostov-on-Don was 7 times higher than in Murmansk and Apatity. Historical zoning and land-cover were the major predictors of urban soil-C intracity variability with high C stocks in the historical centers and under trees and shrubs compared to the recently developed lawns. Land cover mainly explained variability in topsoil C stocks, where subsoil C stocks were more dependent on the urban history.
Acknowledgments The field and laboratory analysis of C stocks was performed with the support of Russian Foundation for Basic Research project № 19-29-05187. Spatial analysis and modeling was supported by Russian Science Foundation project № 19-77-30012.
How to cite: Vasenev, V., Pan, X., Gorbov, S., Korneykova, M., Slukovskaya, M., Sarzhanov, D., and Dolgikh, A.: Intra- and intercity variability of urban soil-C stocks along a climatic gradient, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5971, https://doi.org/10.5194/egusphere-egu22-5971, 2022.
Clustering is still an active method of soil sampling. The defined clusters not only can direct soil sampling for digital soil mapping, but also can serve as management zones for variable application of inputs based on soil sample analysis. This study compares a nonspatial and spatial fuzzy clustering approach of management zones delineation enabling sampling design for both site specific and zonal fertilization. An actual yield potential and a bare soil composite computed from Sentinel-2 imagery featuring soil texture serve as covariates for clustering analysis ensuring a high interpretability of results. The minimum area of clusters and the number of sampling locations is defined by a user. The optimum number of constrained clusters is selected for every field based on silhouette index calculation ensuring the variable soil sampling density according to the variability of field conditions. Soil sampling locations were selected using shortest weighted distances to centroids of clusters or randomly. The results showed that nonspatial fuzzy clustering worked only for relatively homogenous fields with low number of clusters. For highly heterogeneous fields, the formed clusters were not spatially compact, because no spatial weight matrix was used. However, setting the minimum cluster size equal or greater than 1ha significantly improved the clusters compactness for heterogeneous fields even using nonspatial approach. The application of spatial approach improved the cluster compactness of heterogenous fields regardless to cluster size which makes this approach more universal.
The research has been supported by the project no. QK21010247 "Management optimization of unbalanced fields by means of digital soil mapping and soil moisture changes monitoring in order to stabilize the achievable yield" funding by Ministry of Agriculture.
How to cite: Minařík, R., Žížala, D., Skála, J., Kraus, M., and Lukas, V.: Soil sampling for variable-rate fertilization using spatial clustering, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-8289, https://doi.org/10.5194/egusphere-egu22-8289, 2022.
Spatially distributed soil information as input for hydrological models has the potential to improve the representation and physical realism of spatio-temporal hydrological processes. Since spatially distributed soil information is often not available, lumped parameters are frequently used in hydrological models to describe soil functions. However, especially the modeling of hydrological processes in the vadose zone – and consequently groundwater recharge – requires information on soil hydraulic properties. The main objective of this study is the prediction of future groundwater recharge rates for the extent of Austria under changing climate conditions. To reach this goal, we use Machine Learning (ML) based soil hydraulic maps as a basis for the parameterization of the COntinuous SEmi-distributed RunOff model (COSERO).
For the spatial prediction of the soil parameters, XGBoost, a boosting ML-algorithm, was trained with soil hydraulic maps of the federal state of Lower Austria and available environmental raster datasets (e.g. climate data, digital elevation model, landcover etc.). Based on the Austrian wide available environmental covariates, the trained XGBoost model was then used to predict relevant soil hydraulic properties for the whole area of Austria (approx. 83 900 km²) at a target resolution of 1 x 1 km².
For our hydrological model set-up, we rescale the predicted soil hydraulic properties into the model parameter range and domain. After parameter optimization, i.e. in our case scaling the mean and thereby keeping the spatial patterns of the parameters, the conceptual rainfall-runoff model COSERO simulates spatially distributed discharge for the study area. We compare our model results to simulations of a model version using lumped soil parameters to assess the differences in the spatial distribution of groundwater recharge rates. Additionally, we analyze the quality of discharge simulations depending on the respective parameterization of the model. Overall, the results show an increased performance when using distributed soil hydraulic properties.
In summary, this study demonstrates the importance of considering the variability of soil information in a hydrological model framework and evaluates the suitability of implementing digital soil mapping products in groundwater recharge modeling.
How to cite: Zeitfogel, H., Herrnegger, M., Feigl, M., and Schulz, K.: Groundwater recharge modeling – the importance of distributed soil information in hydrological models , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-8526, https://doi.org/10.5194/egusphere-egu22-8526, 2022.
Sustainable agricultural landscape management needs reliable and accurate soil maps and updated geospatial soil information. The traditional process of soil surveying is time-consuming and limited in terms of accuracy and spatial distribution. This problem can be partly overcome by Geographical Information Systems (GIS) and the application of digital soil mapping (DSM) approaches. The DSM analyse the relations between soil properties and environmental variables derived from the Digital Elevation Models (DEM) as well as from remotely-sensed information. Moreover, DSM uses these relations and hence, allows for the regionalization of point observations of soil properties. Several machine-learning methods are used today for DSM. The main goal of this study is the evaluation of different supervised machine-learning techniques for mapping several topsoil properties in an agricultural lowland area of Lombardy region, Italy, and interpreting the modelled relationships. The methods analysed are Random Forest, Gradient Boosting Machine, Support Vector Machine and Generalized Additive Model. We applied the models to predict different correlated soil properties such as the soil organic carbon (SOC), texture (sand, silt, clay content) and topsoil depth. Cross validation performances of these models were determined, and diagnostic tools for the post-hoc interpretation of these black-box models were applied to assess their interpretability as well as similarities and differences in the modelled relationships, which reflect each model’s abilities and biases. An important challenge is the interpretation of the effects of highly correlated predictors, which is achieved using a transformation-based post-hoc interpretation technique. The study helps to identify the best-performing predictive model for lowland area and to understand the robustness of the applied models. The selected models will be used to provide valuable information for facilitating a sustainable land use in an area with a unique soil water cycle as well as for assessing how future climate and socioeconomic changes may influence water content, soil pollution dynamics and food security.
How to cite: Adeniyi, O. D., Brenning, A., and Maerker, M.: Assessing machine-learning algorithms for digital soil mapping in an agricultural lowland area: a case study of Lombardy region., EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-8560, https://doi.org/10.5194/egusphere-egu22-8560, 2022.
Passive and active electromagnetic methods such as L-Band radiometry and radar reflectometry have great potential to provide, from ground or as remote applications, soil moisture maps on local or regional scales. To calculate values of absolute water content from the measured dielectric constant, calibration data is necessary. This is usually acquired in the lab by weighing soil samples. The required drying lasts several days to weeks depending on the soil type. A large number of samples distributed over the entire investigation area is desirable in order to increase the accuracy of the derived moisture maps. This, however, requires increased effort and thus higher costs.
We suggest the use of nuclear magnetic resonance (NMR) to gather this kind of calibration data. NMR measures the water content in porous media directly by stimulating the proton spins of the water molecules. The amplitude of the received response signal is linearly correlated with the number of protons in the sensitive volume of the device, i.e., with the amount of water in it: zero is measured when water is absent, while 100% water corresponds to maximum signal amplitude. In contrast to conventional laboratory NMR, the single-sided NMR technology enables mobile tools that are easy to handle in the field. Absolute soil moisture data is collected just by placing the sensor at the size of a shoebox or suitcase on the ground and the result for a single spot is available after a few minutes instead of days when taking and drying samples.
We successfully tested the single-sided NMR technology at one of our L-Band radiometry test sites predominated by clayey loam. In addition to the quad-bike based passive areal L-Band data acquisition, pointwise single-sided NMR measurements were performed on a profile with the length of 600 m at 10 m spacing. The sensitive NMR volume was adjusted to a depth of 1 cm. A total of 10 samples were taken for verification and analysed in the lab. The absolute water contents provided by NMR excellently agree with those of the samples. Moreover, the NMR profile results are also in good agreement with the L-Band measurements on the same profile. Future investigations will focus on the feasibility of the single-sided NMR method for other soil types and on the interpretation of the NMR relaxation behavior, which allows estimating the water-filled pore size distribution. In addition to the water content, this additional information is useful to estimate water mobility and storage capacity in the topsoil.
How to cite: Costabel, S., Ibs-von Seht, M., and Jove, R.: Mobile single-sided NMR technology as a calibrating tool for areal soil moisture mapping, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-8765, https://doi.org/10.5194/egusphere-egu22-8765, 2022.
The study of urban soils of large cities is complicated due to their heterogeneity and continued reconstruction. Large territorial coverage and administrative prohibitions in some areas led to the difficulty of conducting full-fledged field sampling campaigns and results in inaccuracy. Еxpress methods of chemical elements’ content analysis using portable XRF devices allows to quickly assess the pollution level, minimizing the most complicating factors of research. However, the results obtained using a pXRF analyzer require adjustment of the instrument readings as they can be affected by a set of factors such as humidity, sample heterogeneity, and inter-element interference. Modern models of pXRF analyzers allow automatic correction of instrument readings through the correction factors stored in the device's memory.
Our research focused on the development of such correction factors for the Olympus Vanta, one of the most common pXRF analyzers available today. Urban soils are characterized by high heterogeneity both in terms of potentially toxic metal (PTM) content, particle size distribution, and the proportion of organic matter in the soil. Overall 85 soil samples from three sites in the Moscow megalopolis with different levels of PTM pollution were collected for the device validation: the Repin’s square in the city-center (high level), the RUDN University campus (medium level), and the urban forest in Moscow Timiryazev Agricultural Academy (low level).
Soil samples were collected from 0-10 cm depth, analyzed for moisture content and bulk density, dried, ground, sieved through a sieve with a 2 mm mesh diameter and analyzed by Olympus Vanta C device. Exposure time was 90 sec in the "Soil" mode. The ICP-OES measurements were taken by EPA 6010B. The carbon content was determined by Vario TOC Select (Elementar). Soil pHwater was determined by the potentiometric method. Further, all samples were divided into groups based on different particle size distributions: sand, loam, peat, and their mixtures. Finally, the samples were grouped by the PTM concentrations. International indices (IPI, PINemerow, and PERI) were used to assess the accuracy of complex soil pollution. The correction factors were calculated for five PTMs (Cu, Ni, Zn, Pb, Cd).
For sand, the pXRF-measured concentration corresponded to the ICP-OES result with the conversion factor K=1. The surplus of pXRF readings for samples with peat domination was 1.5-2, but the addition of mineral substrates (sand and loam) to the peat mixtures decreased the coefficient to 1.1-1.4. Among studied PTM, copper and lead had the most stable conversion factors, while other elements had different factors in different intervals of concentrations. However, for all studied elements, the pXRF-readings were unreliable at concentrations less than 5-10 ppm. The pollution indices calculated based on pXRF and ICP-OES data differed but in most cases corresponded to an equal level of contamination. Overall, the Olympus Vanta C portable XRF-analyzer is a promising device for the assessing and mapping of PTM pollution in highly heterogenic urban soils, but pXRF readings of samples with low PTM concentrations and high organic matter content require correction.
The study was supported by the Russian Science Foundation Project #19-77-300-12.
How to cite: Romzaykina, O., Slukovskaya, M., Vasenev, V., Paltseva, A., Sarzhanov, D., and Losev, A.: Assessing urban soils’ pollution in Moscow megalopolis by portable X-ray fluorescence analyzer, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9180, https://doi.org/10.5194/egusphere-egu22-9180, 2022.
The distribution of trace elements in soils is complex and reflects the geochemistry of the original geological substrates modified by variety of environmental and human-induced changes of soil environment. An effective use of geological information within digital soil mapping and geochemical mapping over large require a degree of class aggregation into several broad nominal (or ordinal) classes. Nevertheless, there are several potential weaknesses of the reclassification of lithological information - lithological variation within geological units, variation in composition of individual lithological types, and inadequate description of lithology in geological map. Hence, we tested how the predictive geochemical mapping using environmental correlation will be sensitive under various complement scenarios using aggregated geological substrates and additional numeric covariates that partially represent parent material such as subsoil texture, land gravity data (gravity survey Bouguer anomaly) and other geophysical spatial data (airborne magnetic and gamma radiometric surveys). To compare various scenarios, we have used lithological classification in combinations with other numerical substrate-wise covariates in pragmatic predictive geochemical models using quantile regression forest over contrast area (approximately 11 000 km2) in the Czech Republic. Thee independent geochemical datasets for soil trace elements after the acid digestion procedure were used to train and validate the predictive models. Lithology-wise covariates were iteratively combined with the joint set of other readily available covariates representing topography, land use, remotely sensed surface characterisation (using a cloudless bare soil composite assembled from Sentinel 2) and depositional inputs of trace elements into soil to compare the prediction of topsoil concentrations of trace elements under various research strategies for parametrisation of lithological information. The results enabled to select optimal covariates suite for lithology parametrisation for the complex nation-wide model for topsoil contents of trace elements.
The research has been supported by the Technology Agency of the Czech Republic under the research project No. SS03010364.
How to cite: Skála, J., Žížala, D., and Minařík, R.: Influence of parameterization strategy for parent material effects in predictive mapping of topsoil geochemistry , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9243, https://doi.org/10.5194/egusphere-egu22-9243, 2022.
EJP SOIL, which is a European Joint Programme Cofund on Agricultural Soil Management, has posed questions concerning the application of various soil observation datasets to account, monitor and map agricultural soil carbon, fertility and degradation. One of the most exciting issue is, how the continental LUCAS Topsoil dataset and national soil observation/monitoring databases could be harmonized and/or should complement or improve each other to produce Europe wide spatial soil information to support European contribution towards international reporting on soils, and the accuracy of European agri-environmental policies. Methods are being searched by a broad international team. In our paper we present a countrywide case study for the comparison of (i) the representativity of the Hungarian Soil Information and Monitoring System (SIMS) versus LUCAS Topsoil dataset and (ii) some map products, which were modelled by the usage of the two reference data sources.
The difference between national monitoring systems and LUCAS Topsoil dataset is mainly due to (i) the different measurement methods applied to determine soil properties, and (ii) the different sampling strategy, both in terms of sampling location and sampling depth. While transformation (using unit conversions, mass-preserving splines to derive soil properties for similar soil depth, pedotransfer functions for methods conversion, etc.) of the SIMS soil data into the units, methods and soil depth used in the LUCAS dataset could be carried out more or less straightforwardly, the spatial representativity, which strongly affects the performance of any digital soil map based on the given observation dataset, is a challenging feature, which could and should be checked.
From a statistical point of view, a sample is said to be representative if it reflects the characteristics of the population the best. First we analysed whether the two sets of soil data represent the same population applying two approaches: by the comparison of (i) empirical cumulative distribution function and (ii) the mean values aggregated for different land use categories of the selected soil properties coming from the two system. Although based on the Kolmogorov-Smirnov test we could conclude that the Hungarian subset of the LUCAS Topsoil dataset and the data of the Hungarian Soil Information and Monitoring System come from the same population in the case of particle size distribution, pH, organic carbon and carbonate content, the land use based comparison did not give satisfying results.
In the next round the countrywide spatial predictivity of the two datasets have been tested. Primary soil property map pairs have been compiled using the same ancillary datasets and digital mapping methods but the two different observation datasets. The two map products for the same property have been compared by both global measures and cell-to-cell statistics. In addition to pairwise comparison of basic statistical features (histograms, scatter plots), we have examined the spatial distribution of the differences. In our presentation our findings and experiences will be discussed.
Acknowledgment: Our research was supported by the Hungarian National Research, Development and Innovation Office (NKFIH; K-131820), European Union’s Horizon 2020 Research and Innovation programme under grant agreement No. 862695.
How to cite: Pásztor, L., Bakacsi, Z., Szatmári, G., Kassai, P., Szabó, B., Laborczi, A., Kocsis, M., and Benő, A.: A comparative study of the Hungarian Soil Monitoring System and LUCAS Topsoil dataset and their countrywide spatial predictivity, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9669, https://doi.org/10.5194/egusphere-egu22-9669, 2022.
Hungarian Soil Spatial Data Infrastructure has been recently renewed in the frame of DOSoReMI.hu initiative. Soil property, soil type and functional soil maps were compiled. The set of the applied digital soil mapping techniques has been gradually broadened incorporating and eventually integrating geostatistical, machine learning and GIS tools and very recently spatially non-exhaustive ancillary observations, which has been also hypothesized to be successfully utilizable within DSM framework. (i) Vast, digitally processed legacy soil data, (ii) a spectrum library compiled by the measurements of 6600 soil samples with countrywide origin, (iii) and the results of a nationwide citizen science campaign targeted to collect proxy data on soil health were involved.
Soil property maps have been compiled partly according to international specifications (GlobalSoilMap.net, GSOC, GSASmap), partly to fulfill specific demands on the final products. Secondary (derived) soil features were also predicted. (i) Soil hydraulic properties were mapped applying generalized pedotransfer functions; (ii) spatial assessment of certain provisioning and regulating soil functions was carried out by the involvement of soil property maps in digital process/crop models. The nationwide, thematic digital soil maps compiled in the frame and spin-off of our research is utilized in a number of ways, for the support of national activities (LDN, SDGs, ESS assessment). A new soil portal was also elaborated for publishing of the created DSM products together with the result of their accuracy assesment.
Our paper will present
- the new approaches for the population and extension of DOSoReMI.hu, and
- various national functional applications of DOSoReMI.hu.
Acknowledgement: Our research has been supported by the Hungarian National Research, Development and Innovation Office (NRDI; Grant No: K 131820).
How to cite: Laborczi, A., Szatmári, G., Mészáros, J., Takács, K., Takáts, T., Árvai, M., Kovács, Z., Szabó, B., and Pásztor, L.: Population, extension and some functional applications of DOSoReMI.hu, the renewed Hungarian Soil Spatial Data Infrastructure, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9727, https://doi.org/10.5194/egusphere-egu22-9727, 2022.
The interest in soil analyses with visible and near infrared spectroscopy (Vis-NIRS, ~350-2500nm) has increased rapidly (Nocita et al., 2015; Gholizadeh et al., 2013, Stenberg et al., 2010) due to simple use of the technique and its fastness compared to wet and dry chemistry. Vis-NIR soil spectroscopy has been identified as one of the proximal sensor techniques with most information about organic matter and clay minerals (Gholizadeh et al., 2013) that is very interesting for the agricultural community looking for ways to rapidly assess how management practices affect soil organic carbon stocks, for example. Organic molecules and functional groups in the organic matter absorb strongly in the Vis-NIR range and therefore relate to organic carbon (Stenberg et al., 2010).
In order to interpret, validate and calibrate the Vis-NIR spectra, reference data analysed with wet and dry chemistry is needed. Here, we will present the development and use of the Austrian soil spectral library that currently consists of around 600 agricultural soil samples. The soil spectral library has been built up from representative agricultural air-dried soil samples from farmers and agricultural long-term experiments that have first been analysed with wet and dry chemistry for soil organic matter characteristics including total organic carbon (TOC), labile carbon, total nitrogen and potentially mineralisable nitrogen, among other soil fertility characteristics. The soil spectral library is continuously being extended by more representative agricultural soil samples from farmers and long-term experiments. Its harmonisation is being carried out under the EJP SOIL ProbeField project on a European scale as well as under the global GLOSOLAN initiative on soil spectroscopy cooperation. This is to ensure generic, robust and well performing models that could be used in a simple and fast manner on local, regional and national scales in Austria, as well as to be connected to larger geographical and soil type coverage on a European and global scales through ProbeField and GLOSOLAN networks.
ProbeField is part of EJP SOIL (EU, H2020, grant agreement No 862695)
Gholizadeh, A., Borůvka, L., Saberioon, M., and Vašát, R.: Visible, Near-Infrared, and Mid-Infrared Spectroscopy Applications for Soil Assessment with Emphasis on Soil Organic Matter Content and Quality: State-of-the-Art and Key Issues, Applied Spectroscopy, 67, 1349-1362, 10.1366/13-07288, 2013.
Nocita, M., Stevens, A., van Wesemael, B., Aitkenhead, M., Bachmann, M., Barthès, B., Ben Dor, E., Brown, D. J., Clairotte, M., Csorba, A., Dardenne, P., Demattê, J. A. M., Genot, V., Guerrero, C., Knadel, M., Montanarella, L., Noon, C., Ramirez-Lopez, L., Robertson, J., Sakai, H., Soriano-Disla, J. M., Shepherd, K. D., Stenberg, B., Towett, E. K., Vargas, R., and Wetterlind, J.: Chapter Four - Soil Spectroscopy: An Alternative to Wet Chemistry for Soil Monitoring, in: Advances in Agronomy, edited by: Sparks, D. L., Academic Press, 139-159, https://doi.org/10.1016/bs.agron.2015.02.002, 2015.
Stenberg, B., Viscarra Rossel, R. A., Mouazen, A. M., and Wetterlind, J.: Chapter Five - Visible and Near Infrared Spectroscopy in Soil Science, in: Advances in Agronomy, edited by: Sparks, D. L., Academic Press, 163-215, https://doi.org/10.1016/S0065-2113(10)07005-7, 2010.
How to cite: Sandén, T., Lippl, M., Reiter, E., Dersch, G., Spiegel, H., and Baumgarten, A.: Austrian Soil Spectral Library for future soil fertility assessments, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9803, https://doi.org/10.5194/egusphere-egu22-9803, 2022.
Soil spectroscopy in the mid-infrared (MIR) allows the fast and cost-effective derivation of multiple physical and chemical soil properties, e.g., soil organic carbon (SOC) and soil texture, from a single reflectance spectrum. The recent development of extensive soil spectral libraries and field-portable handheld FTIR spectrometers have opened up new opportunities for the widespread application of soil reflectance spectroscopy in the geo- and environmental sciences. Compared to laboratory measurements on pre-treated soil material, field recordings of MIR spectra are impacted by in situ environmental conditions that modify and degrade the measured reflectance signal, most prominently variations in soil moisture and particle size across samples. These conditions prevent leveraging available MIR soil spectral libraries to build predictive models of soil properties directly.
We evaluated the capacity of the External Parameter Orthogonalization (EPO) algorithm to compensate for moisture and particle size-induced effects on MIR reflectance spectra recorded in the field to generate laboratory-equivalent spectra from the in-situ data, which would allow calibrations of predictive soil property models from soil spectral library data to be transferred to field-recorded spectra. An archive of 230 soils collected across five soil regions in Germany covering a broad range of parent materials, soil texture classes and organic carbon contents was used to evaluate the approach. For each soil sample, MIR reflectance spectra had been acquired both in the field, i.e., measured in situ on the soil surface, and in the laboratory on pre-treated (sieved and ground) soil material. Field spectra were corrected for environmental effects by EPO and used to predict SOC and soil texture with predictive models developed on the laboratory spectra.
Analysis of the EPO-transformed spectra showed that the algorithm could compensate for some of the significant environmental effects present in the field data, e.g., non-linear baseline shifts and large-scale water absorption features, effectively reducing variation across the soil samples that is not linked to the physical and chemical soil properties of interest. EPO-transformation of the spectra further allowed a robust transfer of calibrations developed on laboratory spectra of pre-treated soils to the field spectra. Predictive accuracies for SOC and soil texture were lower than for pure laboratory applications but generally in line with models developed with an extensive regional calibration sample directly on the field MIR spectra.
The correction of field MIR spectra with the EPO algorithm thus represents a promising approach to integrating existing soil spectral libraries into the development of predictive soil property models for in-situ MIR reflectance spectra as it would allow the development of predictive models without requiring a large number of additional regional calibration samples for field application of MIR soil spectroscopy.
How to cite: Hutengs, C., Seidel, M., and Vohland, M.: Compensation of moisture and particle size effects on soil mid-infrared (MIR) reflectance spectra collected in the field with External Parameter Orthogonalization, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10092, https://doi.org/10.5194/egusphere-egu22-10092, 2022.
Salinization in arid or semiarid regions with water-logging limits cropland yield, threatening food security. The highest level of farmland salinization, that is, abandoned salinized farmland, is a trade-off between inadequate drainage facilities and sustainable farming. The evolution of abandoned salinized farmlands is closely related to the development of cropping systems. However, detecting abandoned salinized farmland using time-series remote sensing data has not been investigated well by previous studies. In this study, a novel approach was proposed to detect the dynamics of abandoned salinized farmland using time series multi-spectral and thermal imagery. Thirty-two years of temporal Landsat imagery (from 1988 to 2019) was used to assess the evolution of salinization in Hetao, a two-thousand-year-old irrigation district in northern China. As intermediate variables of the proposed method, the crop-specific planting area was retrieved via its unique temporal vegetation index (VI) pattern, in which the shape-model-fitting technology and the k-means cluster algorithm were used. The desert area was stripped from the clustered non-vegetative area using its distinct features in the thermal band. Subsequently, the abandoned salinized farmland was distinguished from the urban area by threshold-based saline index (SI). In addition, a regression model between electrical conductance (EC) and SI was established, and the spatial saline degree was evaluated by the SI map in uncropped and unfrozen seasons. The results show that the cropland has constantly been expanding in recent decades (from 4.7*105 ha to 7.1*105 ha), while the planting area of maize and sunflower has grown and the area of wheat has decreased. Significant desalinization progress was observed in Hetao, where both the area of salt-affected land (salt-free area increased approximately 4*105 ha) and the abandoned salinized farmland decreased (reduced from 0.45 *105 ha to 0.19 *105 ha). This could be mainly attributed to three reasons: the popularization of water-saving irrigation technology, the construction of artificial drainage facilities, and a shift in cropping patterns. The decrease in irrigation and the increase in drainage have deepened the groundwater table in Hetao, which weakens the salt collection capacity of the abandoned salinized farmland. The results demonstrated the promising possibility of reutilizing abandoned salinized farmland via a leaching campaign where the groundwater table is sufficiently deep to stop salinization.
How to cite: Zhao, L.: Assessing the long-term evolution of abandoned salinized farmland via temporal remote sensing data, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10809, https://doi.org/10.5194/egusphere-egu22-10809, 2022.
Mid-infrared spectroscopy is an efficient technique for soil carbon analysis. Efforts to measure and monitor carbon through mid-infrared spectroscopy require the development of soil spectral libraries. These libraries are used for the construction of calibration models which relate analyte values to spectra. The optimization of these models is an important process for the accurate and resource-efficient estimation of soil carbon. This study demonstrates the effect on model performance of subsetting a soil spectral library for soil organic carbon estimation. Various subsetting criteria were tested across different landscapes in the United States, and results are presented in the context of the development of new soil spectral libraries.
How to cite: Dorantes, M., Fuentes, B., and Miller, D.: Optimization of calibration models for soil carbon estimation using mid-infrared spectroscopy, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11036, https://doi.org/10.5194/egusphere-egu22-11036, 2022.
Precision Agriculture (PA) applied on a widespread basis can be a building block for reduced ecosystem degradation without compromising food security. One problem of farmers in the implementation of PA applications is the lack of high spatial resolution soil information.
The EU-funded research project ‘pH-BB: Precision liming in Brandenburg’ aims at developing innovative nutrient management strategies based on proximal soil sensing data. For this study, the pH-BB project provided us the data of the 361 soil samples, which were taken at an 800-hectare farm in Brandenburg close to Frankfurt (Oder) and analyzed in the laboratory for the texture fractions clay, silt, and sand. We used the Google Earth Engine to process remote sensing earth observation (EO) data of 1474 Sentinel-1 (S1) SAR and 85 cloud free Sentinel-2 (S2) scenes available at the study site during the period 2016-03-01 - 2021-11-09. Vegetation and soil indices were computed with optical S2 data and backscatter in VV and VH polarization were extracted from the S1 datasets. To derive long-term persistent patterns in the EO data, simple statistical parameters such as coefficient of variation, standard deviation, maximum pixel value, etc. were calculated along the temporal domain of the EO data. Together with calculated terrain attributes, 24 covariate grids were finally available for model building. Reference samples (rs) were randomly divided into a training dataset (70% of rs) and a validation dataset (30% of rs). Pixel values of the covariate datasets at the sampling locations were added to the rs datasets.
A random forest machine learning algorithm was applied to the training dataset to train two individual models for the alr-transformed target variables silt and sand using the covariates. The developed models were then applied to the gridded datasets to predict maps for the alr-transformed target variables silt and sand. The final maps of all 3 texture fractions clay, silt, and sand were computed by back-transforming the predicted alr silt and sand grids.
In the derived models EO covariates showed the highest level of importance. Comparison of the prediction results with the validation data set showed that the spatial distribution, of the clay, silt, and sand fractions was predicted, with a root mean square error (rmse) of 6.2, 5.3, and 9.7 mass-%, respectively. A classification of the predicted maps according to the German KA5 scheme showed that especially in the sand dominated soil classes the prediction errors were lower, whereas they increased in the loamy soil classes (dependent on the clay content).
With a rmse of 5.3 - 9.7 mass-%, the performance of the approach shows good potential for surface soil texture assessments even at high resolution or for global applications or as an initial guess for soil mapping with high resolution proximal soil sensing devices.
How to cite: Meyer, S., Rühlmann, J., Bönecke, E., Kramer, E., and Marzahn, P.: Prediction of soil texture using optical and microwave earth observation data in a random forest approach, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11333, https://doi.org/10.5194/egusphere-egu22-11333, 2022.
The study goal was the preliminary mapping of Fe2O3, Nb and TiO2 contents to support of the classification of outcropping materials (focused on laterite types), in “Morro dos Seis Lagos”, in Brazilian Amazon. The methodological procedures were based on machine learning tools, gathering Sentinel-2 MSI and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) sensor data, numerical terrain models (all with 20 m spatial resolution), and geochemical legacy dataset from the Geological Survey of Brazil (CPRM). The input geochemical dataset was subdivided in training dataset (341 samples) and validation dataset (85 samples) to apply the Random Forest (RF), in 60 loops of iteration, and the model’s performance were evaluated through the average values of the metrics (R2, RMSE and MAE). Subsequently, the resulting average maps were combined using cluster analysis (k-means) via unsupervised classification, performed at the R environment through the Vegan package, where the number of resulting classes was optimized taking into account the “Simple Structure Index” (SSI) criterion. Different cluster grouping was tested considering classes number (6 to 10) and interactions (0 to 8000), and the resulting classes (zones) were contrasted with available geological map. The results showed the best performance in modeling via the Random Forest (RF) model associated with Recursive Feat Elimination (RFE) for the elements Nb (R2=0.08, RMSE = 0.86, MAE = 0.66), TiO2 (R2=0.14, RMSE = 3.90, MAE= 2.56) and Fe2O3 (R2=0.23, RMSE =19.77, MAE = 14.12). Based on the results obtained via preliminary cluster analysis, the best optimization was achieved grouping in 9 classes, according to SSI criterion. The results showed agreement when compared to the classes of the geological map available for the area, but with better detailing of the laterite facies. The conclusions of the preliminary study pointed out that advances regarding the scale detail provided better understanding the behavior of the variability in laterite and talus deposits with the support of machine-learning tools and covariates from remote sensing data. However, improvement in the cluster classification can be achieve by adding other geochemical compounds and testing different predictive models.
How to cite: Rodrigues, N., Silva, J. C. L. D., Silva, R. P. M. D., Pinheiro, H. S. K., and Carvalho Junior, W.: MAPPING OF Fe2O3 , Nb and TiO2 , AS A SUPPORT TO CLASSIFY OUTCROPPING MATERIALS IN” MORRO DOS SEIS LAGOS” CARBONATITE COMPLEX, BRAZILIAN AMAZON, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12326, https://doi.org/10.5194/egusphere-egu22-12326, 2022.
Soil coverage of Czechia has been well-mapped thanks to Complex Soil Survey (CSS) during the 1970s. Despite its considerable details, it is not capturing many of the non-negligible phenomena, and it is not as accurate as it could be as if it were performed today with today technology. This study deals with soil mosaics visible from aerial photographs on agricultural lands, which are not affected in the CSS maps, in terms of determination and classification based on the processes that create these mosaics. For this purpose, fifty localities in Czechia were selected. Attention is paid to the way of development, places of occurrence and the shape of the mosaic itself. One of the main goals is to use only freely available data such as DMR, aerial photographs and soil maps. That is because we want this analysis to be easily reusable by other pedologists in other places. In this study, we propose a classification of different soil mosaics.
The proposed classification consists of three groups, containing seven categories defined by the primary causes of occurrence. Not all sites are formed only by one degradation process. The mosaic is often composed of several degradation processes, and the resulting shape is its sum. The main factors are changes in the soil properties, soil organic matter content, and different ability in water retention (differences in soil moisture). The resulting classification can be used for a) further continuation of soil mapping (where it could serve as an aid in selecting suitable sites for soil probes), b) soil sample gathering (to decide where to gather soil samples to get the most representative soil sample of the area), c) precise determination of polygons with similar soil properties, d) better planning in precision agriculture, e) more realistic estimation of the K factor in the RUSLE equation, to raise the accuracy of estimation of soil erosion.
Furthermore, the relationship between the shape of the mosaic and the processes that create the mosaic was analysed. Because it's evident that similar processes cause similar patterns, we choose a Box-Counting method to calculate its fractal dimension. Results of this analysis didn't show any correlation, and it wasn't possible to determine the type of mosaic only according to its value of the fractal dimension. The machine learning approach would be probably more suitable for this problem. Database of shapes, divided into the categories according to the classification above. Automatical analysis of pictures of the mosaics and their sorting according to their highest similarity with one of the categories. The possibility of using more detailed satellite images for soil reflectance values in the case of soil mosaic classification could also be very beneficial in the future. For capturing the soil mosaic by satellite images, high-quality images are necessary. Such images are not freely available yet, and that's why we didn't use them in this study.
How to cite: Kocum, J., Šefrna, L., and Vlček, L.: Soil moisture: coauthor of the soil mosaic patterns, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12486, https://doi.org/10.5194/egusphere-egu22-12486, 2022.
Modeling climatic conditions and knowing about them helps us to improve ecosystem management. Climate classifications generally have been produced using stations' data, and because satellite data did not have a proper temporal period, they could not be applied as a tool for climate classification. The aim of this study was a qualitative assessment of the fitness of satellite data as covariates of an agro-climate classification. To define agroclimate classes in Iran land, temperature and precipitation were selected as the main climatological parameters in agriculture. Using data collected from 3825 synoptic, climatological, rain gauge, and evaporation stations from 2002 to 2016, an agroclimatic map was produced with a resolution of 5 km which is divided into 24 agroclimatic classes. Comparison between resulted agroclimatic classes and some remote sensed agricultural related variables including mean_yearly_NDVI-TVDI, average actual evapotranspiration (m/yr), evapotranspiration (m/yr) and average soil moisture (m3/m3), showed a very sharp visual accordance. The accordance was very clear specially in the case of TVDI which had a greater resolution of 1 km x 1 km. The results showed that satellite data can be a useful candidate (as meaningful auxiliary variables) for agroclimate classifiers. moreover, in situ based classifications can be beneficial as a tool of satellite data classification and interpretation. Another point is that, the greater the similarity between satellite data and agroclimate classified raster resolutions, the better the conditions for comparing and evaluating performance.
How to cite: Asadi Oskouei, E., Delsouz Khaki, B., and Lopez-Baeza, E.: Agroclimatological Classification for Iran Land for Earth Observation Purposes, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12761, https://doi.org/10.5194/egusphere-egu22-12761, 2022.
