SSS10.3

Since the establishment of Digital Soil Mapping (DSM) as a research field, the main focus has been on implementing new methods to improve the predictive performance of soil maps. However, considerably less effort has been invested in investigating the best way to communicate the quality of soil mapping products with users. This is essential for soil maps to be adopted by a broader community, future research guidance and most importantly, to ensure that they are used correctly. We introduce a high-resolution 3D soil modelling and mapping platform for the Netherlands (BIS-3D) using a quantile regression forest (QRF) for spatial interpolation approach that includes an assessment of the map quality using GlobalSoilMap (GSM) accuracy thresholds. Our objectives are twofold: a) providing accurate and high-resolution (25m) soil pH, soil organic carbon, and soil texture (clay, silt, and sand) maps over 3D space including prediction uncertainty; and b) providing an intuitive way to communicate accuracy of soil maps for users by means of accuracy thresholds. In this work, the first outputs of the modelling and mapping platform BIS-3D are being presented.

QRF models were trained and validated, yielding average predictions for each target location and depth as well as the 90% prediction interval. Predicted soil maps were evaluated using an independent validation data set based on a stratified random sampling design covering the entire Netherlands (1151 locations). Furthermore, at every validation location, predictions were assessed as A, AA or AAA quality using the GSM specifications.

First results for soil pH (KCl) using 15887 soil observations between depths 0-2 m and 180 covariates reveal a mean square error skill score (SS_mse) = 0.88, RMSE = 0.49 and bias = 0.01 for out of bag predictions. Model evaluation using the independent validation set resulted in SS_mse = 0.66, RMSE = 0.81 and bias = 0.12 across all depths. Prediction accuracy was highest for depths between 0-15 cm (SS_mse = 0.66, RMSE = 0.76) and 60-100 cm (SSmse = 0.69, RMSE = 0.78) and lowest for 100-200 cm (SSmse = 0.61, RMSE = 0.86). The soil measurement (observation) was within the 90% prediction interval of model predictions in 83% of the cases, indicating that QRF is slightly over-optimistic in quantifying the prediction uncertainty. 61% of predictions that were independently validated over all depths were within the highest GSM accuracy threshold (AAA = +/- 0.5 pH), 23% were AA (+/- 1.0 pH), 9% were A (+/- 1.5 pH) and the remaining 7% were below A. A categorical physical geography map was the most important covariate, although other covariates associated with relief, geomorphology, land use and temperature were also effective. However, such variable importance measurements are merely indications and should be handled with care. The BIS-3D can easily be extended for predicting additional soil properties and it may provide a basis for decision makers to easily assess to what extent and in which areas soil maps can be used for their applications.

How to cite: Helfenstein, A., Mulder, V. L., Heuvelink, G. B. M., and Okx, J.: BIS-3D: high resolution 3D soil maps for the Netherlands using accuracy thresholds, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-7836, https://doi.org/10.5194/egusphere-egu21-7836, 2021.

‘Strategic objective 1’ of the United Nations Convention to Combat Desertification (UNCCD) aims to improve conditions of affected ecosystems, combat desertification/land degradation, promote sustainable land management, and contribute to land degradation neutrality. The indicator ‘Proportion of land that is degraded over total land area’ (SO1) is compiled from three sub-indicators: ‘Trends in land cover’ (SO1-1), ‘Trends in land productivity or functioning of the land’ (SO1-2), ‘Trends in carbon stocks above and below ground’ (SO1-3).

Soil organic carbon (SOC) stock can be adopted as the metric of SO1-3, until globally accepted methods for estimating the total terrestrial system carbon stocks will be elaborated. SOC can be considered as one of the most important properties of soil, which shows not just spatial but temporal variability. According to our previous results in the topic, UNCCD default data of SOC stock for Hungary is strongly recommended to be replaced with country specific estimation of SOC stock.

SOC stock maps were compiled in the framework of DOSoReMI.hu (Digital, Optimized, Soil Related Maps and Information in Hungary) initiative, predicted by proper digital soil mapping (DSM) method. Reference soil data were derived from a countrywide monitoring system. The selection of environmental covariates was based on the SCORPAN model. The elaborated SOC stock mapping methodology have two components: (1) point support modelling, where SOC stock is computed at the level of soil profile, and (2) spatial modelling (quantile regression forest), where spatial prediction and uncertainty quantification are carried out using the computed SOC stock values.

We analyzed how SOC stock changed between 1998 and 2016.  Nationwide SOC stock predictions were compiled for the years 1998, 2010, 2013, and 2016. For the intermediate years, we do not recommend to calculate SOC stock values, because we have no information on the dynamics of change in the intervening years. Based on the 1998 SOC stock prediction, we compiled a SOC stock map for 2018, using only land use conversion factors, according to the default data conversion values.

According to the elaborated scheme during the respective period, significant changes cannot be detected, only tendentious SOC stock changes appear. Based on our results, we recommend to use spatially predicted layers for all years when data are available, rather than calculating SOC stock change based on land use conversion factors.

Acknowledgment: Our research was supported by the Hungarian National Research, Development and Innovation Office (NKFIH; K-131820) and by the Premium Postdoctoral Scholarship of the Hungarian Academy of Sciences (PREMIUM-2019-390) (Gábor Szatmári).

How to cite: Laborczi, A., Szatmári, G., Mészáros, J., Koós, S., Pirkó, B., and Pásztor, L.: Supporting land degradation neutrality assessment by soil organic carbon stock mapping in Hungary, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-4832, https://doi.org/10.5194/egusphere-egu21-4832, 2021.

The estimation of the soil organic carbon (SOC) content plays an important role for carbon sequestration in the context of climate change and soil degradation. Reflectance spectroscopy has proven to be promising technique for SOC quantification in the laboratory and increasingly from air and spaceborne platforms, where hyperspectral imagery provides great potential for mapping SOC on larger scales.

The PRISMA (PRecursore IperSpettrale della Missione Applicativa) is an earth-observation satellite with a medium spatial resolution hyperspectral radiometer onboard, developed and maintained by the Italian Space Agency.

The Pan-European Land Use/ Land Cover Area Frame Survey (LUCAS) topsoil database contains soil physical, chemical and spectral data for most European countries. Based on the LUCAS points located in Hungary, a synthetized spectral dataset was created and matched to the spectral characteristic of PRISMA sensor, later used for building up machine learning based models (random forest, artificial neural network). SOC levels for the sample area was predicted using generated models and mainly PRISMA imagery.

Our sample imagery data was generated from five consecutive, cloud-free PRISMA images covering 4500 km² in the central part of the Great Plain in Hungary, which is one of the most important agricultural areas of the country, used mainly for crops on arable lands. The images were recorded in 2020 February when most croplands are not covered by vegetation therefore our tests were implemented on bare soils.

We tested the prediction accuracy of hyperspectral imagery data supplemented by various environmental datasets as additional predictor variables in four scenarios: (i) using solely hyperspectral imagery data (ii) spectral imagery data, elevation and its derived parameters (e.g. slope, aspect, topographic wetness index etc.) (iii) spectral imagery data and land-use information and (iv) all aforementioned data in fusion.

For validation two types of datasets were used: (i) measured data at the observation sites of the Hungarian Soil Information and Monitoring System and (ii) the recently compiled national SOC maps., which provides a suitable and formerly tested spatial representation of the carbon stock of the Hungarian soils.

Acknowledgment: Our research was supported by the Cooperative Doctoral Programme for Doctoral Scholarships (1015642) and by the OTKA thematic research projects K-131820 and K-124290 of the Hungarian National Research, Development and Innovation Office and by the Scholarship of Human Resource Supporter (NTP-NFTÖ-20-B-0022). Our project carried out using PRISMA Products, © of the Italian Space Agency (ASI), delivered under an ASI License to use.

How to cite: Kovács, Z. A., Mészáros, J., Árvai, M., Laborczi, A., Szatmári, G., László, P., and Pásztor, L.: Testing PRISMA hyperspectral satellite imagery in predicting soil carbon content based on synthetized LUCAS spectral data, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-15450, https://doi.org/10.5194/egusphere-egu21-15450, 2021.

Soil organic carbon (SOC) is the largest reservoir of organic carbon in the terrestrial biosphere and is the main constituent of soil organic matter, which underpins key soil functions such as storage and filtration of water, and nutrient cycling. SOC concentrations are controlled by several dynamic variables, ranging from micro-scale properties like particle aggregation, to larger-scale drivers such as climate and land cover. Hence, soils are vulnerable to climate change and human disturbances, with implications for ecosystem services such as agriculture and global warming mitigation. Recent decades have seen greater efforts to monitor SOC dynamics, such as the UKCEH Countryside Survey, and to predict concentrations of SOC where we have no measurements, using geostatistics or machine learning approaches. Yet, there is still much to be understood about what controls spatial patterns of SOC, and how effectively different modelling approaches can capture this. Here, we compare predictions by nine maps of the spatial distribution of topsoil SOC in Great Britain. We found broad similarities in SOC concentrations predicted by all maps, which each showed right-skewed distributions with similar median values (43 to 97 g kg^-1). The greatest differences between maps occur at higher latitudes and are reflected in the upper ends of the SOC distributions. While the maps generally exhibit a sharp rise in SOC concentrations with increasing latitude from ~54^oN, values predicted by the ISRIC-2017 and FAO-GSOC maps show weaker increases with increasing latitude, and peak at lower values of 332 g kg^-1 and 354 g kg^-1, respectively. We demonstrate that most of the maps, regardless of the modelling approach taken or the underlying data used, produced similar estimates of SOC concentration, including broad spatial patterns. This work will form the basis of more detailed future assessments of the sensitivity of SOC mapping to analytical methods versus the data used to drive these methods, and will be used to assess the importance of using stratified random field survey approaches for generating more accurate predictions of areas that cannot be sampled. Exploration of why and where different and coincident SOC predictions occur between maps should shed light on the utility of different modelling techniques and machine-learning meta-analyses of driving variables currently used to map SOC. Understanding how SOC predictions differ across all current national scale GB maps is a first step in improving modelling and assessment of SOC stock and change.

How to cite: Feeney, C., Cosby, J., Robinson, D., Thomas, A., and Emmett, B.: A comparison of soil organic carbon concentration maps of Great Britain, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-4700, https://doi.org/10.5194/egusphere-egu21-4700, 2021.

Parent material is an essential soil property, whose mapping is a challenging task. Usually, large scale geological maps are used if they are available. However, in many cases, especially in medium and large scale mapping, such source data are too old or not existing at all. In this project have been looking for a solution for this problem. Our aim is to create a new, large scale, lithological map of parent material in an old mining region.

The study area is the Dorogi Basin in northern central Hungary. It is known for coal mining, which ended in 2003 after more than two centuries. The latest large scale (1:10,000) geological map series from this area was made in the 1960’s, in the “golden age” of mining.

Google Earth Engine was selected as main GIS platform, using mainly open source data and programs for mapping. We have used data originating from Earth Observation as ancillary information (e.g. satellite images, SRTM) and machine learning techniques to spatially predict parent material. The satellite images were used to calculate several geological indices, which can be used as indicators of chemical composition. We examined the use of multiple satellite platform (Sentinel-2, Landsat 8, ASTER) as it has different geological indices. The existing geological maps were used for training in the classification concerning the lithological composition.To predict the parent materials we have used random forest, using geomorphometric features and geological indices as predictors. The newly compiled map was validated by comparing it with the old one.

Acknowledgment: Our research was supported by the Hungarian National Research,Development and Innovation Office (NKFIH; K-131820) and from the part of G.A. financial support was provided from the NRDI Fund of Hungary, Thematic Excellence Programme no. TKP2020-NKA-06 (National Challenges Subprogramme) funding scheme.

How to cite: Takáts, T., Mészáros, J., Albert, G., and Pásztor, L.: Mapping parent material using data originating from Earth Observation as ancillary information, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-4910, https://doi.org/10.5194/egusphere-egu21-4910, 2021.

Soil maps are an important source of data in monitoring natural resources and land use planning. However, in many countries, soil maps were prepared at a reconnaissance level. This detail is not enough for land use planning. Soil texture is one of the most important soil physical properties that affect water holding capacity, nutrient availability, and crop growth. The spatial distribution of soil texture at a high resolution is essential for crop planning and management. Digital soil mapping is the method of spatial data generation with the advantages of current technologies. It supplies fast, accurate, and reproducible results.

In this study, a soil texture map with 30 m spatial resolution was produced for an alluvial plain covering an area of approximately 10,000 ha. In the study, 11 Topographic Environmental Variables obtained from NASA's ASTER Global Digital Elevation model were used. Another input parameters were clay, silt, and sand values determined for 91 soil samples obtained through field studies.

R Core Environment (3.6.1) and related packages were used for environmental variable extraction, modeling, and spatial mapping. For model building, 70 % of data was used and the rest of the data was used for validation. Random Forest Algorithm offers interpretability for pedological information extraction by determining the importance of environmental variables in digital soil mapping. Random Forest Algorithm is preferred because of working in small data sets, harmoniously. The most important topographic environmental variables for clay were elevation, aspect, and slope. For sand, it was the elevation, aspect, and topographic wetness index. And for silt, it was the elevation, slope length, and planform curvature. Root Mean Square Error (RMSE), was used as a model performance measure. In the train data, R² values for clay, sand and silt were 0.84, 0.75, 0.85 and RMSE values were 5.23 %, 3.03 %, 5.48 % respectively. In the test data, R² and RMSE values were 0.26, 0.11, 0.10 and 11.8 %, 6.74 %, 13.71 % respectively.

There are high differences between RMSE values of training and test data sets. This event may be caused by the small sample size and to be discussed subject in different studies. High resolution (30 m) data of clay, silt, and sand contents can be useful for hydrological studies and for the preparation of land use plans. Digital soil maps can guide policymakers in creating site-specific land management plans. As well as it can be used for monitoring soil fertility and providing ecosystem services. This study revealed important results regarding the use of digital soil mapping in practice with its analytical and statistical accuracy.

How to cite: Kaya, F. and Başayiğit, L.: Digital Mapping of Soil Particle Size Distribution in an Alluvial Plain Using the Random Forest Algorithm, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-7828, https://doi.org/10.5194/egusphere-egu21-7828, 2021.

Open peatlands are a recognizable feature in the boreal landscape that are commonly mapped from aerial photographs. However, wet soils also occur on tree covered peatlands and in the riparian zones of forest streams and surrounding lakes. Comparisons between field data and available maps show that only 36 % of wet soils in the boreal landscape are marked on maps, making them difficult to manage. Wet soils have lower bearing capacity than dry soils and are more susceptible to soil disturbance from land-use management with heavy machinery. Topographical modelling of wet area indices has been suggested as a solution to this problem and high-resolution digital elevation models (DEM) derived from airborne LiDAR are becoming accessible in many countries. However, most of these topographical methods relies on the user to define appropriate threshold values in order to define wet areas. Soil textures, topography and climatic differences make any application difficult on a large scale. This complex landscape variability can be captured by utilizing machine learners that uses automated data mining methods to discover patterns in large data sets. By using soil moisture data from 20 000 field plots from the National Forest Inventory of Sweden, we combined information from 24 indices and ancillary environmental features using a machine learning known as extreme gradient boosting. Extreme gradient boosting used the field data to learn how to classify soil moisture and delivered high performance compared to many traditional single algorithm methods. With this method we mapped soil moisture at 2 m spatial resolution across the Swedish forest landscape in five days using a workstation with 32 cores. This new map captured 79 % (kappa 0.69) of all wet soils compared to only 36 % (kappa 0.39) captured by current maps. In addition to capture open wetlands this new map also capture riparian zones and previously unmapped cryptic wetlands underneath the forest canopy. The new maps can, for example, be used to plan hydrologically adapted buffer zones, suggest machine free zones near streams and lakes in order to prevent rutting from forestry machines to reduce sediment, mercury and nutrient loads to downstream streams, lakes and sea.

How to cite: Lidberg, W., Larson, J., Paul, S., Laudon, H., and Ågren, A.: Using machine learning to generate high-resolution soil wetness maps for planning forest management, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-2578, https://doi.org/10.5194/egusphere-egu21-2578, 2021.

To understand, model and predict landscape evolution, ecosystem services and hydrological processes the availability of detailed observation-based soil data is extremely valuable. Estonia has a national digitized soil map based on decades of Soviet era field mapping. It maps more than 750 000 soil units throughout Estonia at a scale of 1:10 000 - with 75% of mapped units smaller than 4.0 ha. However, due to the way it was recorded the data is not immediately useful for numerical modelling. We synthesized the EstSoil-EH dataset - more than 20 eco-hydrological variables on soil, topography and land use for Estonia as numerical and categorical values - using data fusion and machine learning.
As additional feature information we used a 5m DEM, the Estonian Topographic Database, and EU-SoilHydroGrids layers. For each soil unit, we analysed type, texture, and layer information from the originally recorded composite text-based soil information, which contains the actual texture class, classifiers for rock content, peat soils, distinct compositional layers, and their depths. Subsequently, we derived soil layering, clay, silt, and sand contents and coarse fragments of the soil layers. In addition, we aggregated and predicted physical variables related to water and carbon cycles (bulk density, hydraulic conductivity, organic carbon content, available water capacity). We validated our modelled data and achieved satisfying degrees of agreement depending on the variables type.

How to cite: Kmoch, A., Kanal, A., Astover, A., Kull, A., Virro, H., Helm, A., Pärtel, M., Ostonen, I., and Uuemaa, E.: EstSoil-EH - Developing a high-resolution eco-hydrological modelling parameters dataset for Estonia, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-857, https://doi.org/10.5194/egusphere-egu21-857, 2021.

Soil water content is a key property for modelling the water balance in hydrological, eco-hydrological and agro-hydrological models. Currently available global maps of soil water retention are mostly based on pedotransfer functions applied to maps of other basic soil properties. We developed global maps of the volumetric water content at 10, 33 and 1500 kPa by direct mapping based on soil water content data derived from the WoSIS Soil Profile Database and covariates describing vegetation, terrain morphology, climate, geology and hydrology using the SoilGrids workflow. The preparation of the input soil data consisted of the verification of available volumetric water content data and conversion of gravimetric to volumetric data using measured and estimated bulk density. In total we had 9609, 41082 and 49224 soil water content observations at 10, 33 and 1500 kPa, respectively, and prepared around 200 covariates as candidate predictors. After covariates selection, model tuning and cross-validation and final model fitting for 3D spatial prediction, results were presented for the globe with uncertainty estimation. The results were also compared to other available global maps of water retention to evaluate differences between direct mapping against other types of approaches. Directly developing global maps of soil water content, with associated uncertainty, is a novel approach for this type of properties, and contributes to improving global soil data availability and quality.

How to cite: Turek, M. E., Heuvelink, G., Batjes, N., and Poggio, L.: Global mapping of volumetric water content at 10, 33 and 1500 kPa using the WoSIS global database, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-15548, https://doi.org/10.5194/egusphere-egu21-15548, 2021.

The response of soils to different human forcings may vary among soil classes (in magnitude and direction of change) depending on their resistance and resilience. We propose a modelling framework for mapping soil-class specific references (i.e., genosoils) and their variants (i.e., phenosoils) that can be used for assessing changes in soil condition due to land use change and management practices. The methodology consists of a first step that creates groups characterized by homogeneous soil-forming factors for a given reference time, under the hypothesis that these groups represent soil classes resulting from multimillennial natural pedogenesis and historic anthropedogenesis (i.e., soil formation processes modified by human activities) (i.e., pedogenons). In this study we applied the methodology to New South Wales (Australia) at the time of the European settlement, because from 1788 onwards the intensification of land use may have accelerated the rate of change of soil properties. A thousand pedogenon classes were generated applying k-means clustering to a set of quantitative state variables that represent the soil-forming factors at the time of the European settlement. Hierarchical clustering was applied to the centroids of the pedogenon classes for assessing their similarities and organization. In a second step, information on native vegetation extent, status (cleared or intact), and current land use was combined for creating a categorical map distinguishing areas with different expected degree of human-induced soil change. The combination of both maps resulted in 5448 subclasses, ranging from remnant genosoils (located in protected areas of intact native vegetation), genosoils II, cleared, grazing and cropping phenosoils. For each pedogenon there was at least a 90-m grid cell classified as a remnant genosoil. The median of the proportion of the pedogenon of origin preserved as a remnant genosoil was 5.3%. Phenosoils grazing and cropping occupied larger areas, with mean values of 73 km² and 153 km² respectively. Finally, we tested differences in topsoil pH, as proxy for soil condition, by genosoil and phenosoil classes using legacy soil data accessed with the Soil Data Federator from the Terrestrial Ecosystem Research Network. A gls model indicated that the effects of pedogenon, genosoil/phenosoil and their interaction were statistically significant (p < 0.001). Paired mean comparisons suggested that mean pH did not differ between remnant genosoils and genosoil II, but the mean pH of both genosoil classes differed from phenosoils. Estimated pH means did not differ between phenosoil classes, although it followed the trend remnant genosoil < genosoil II < phenosoil cleared < phenosoil grazing < phenosoil cropping. The proposed methodology has several potential applications, including soil security and soil change assessment, and designing soil monitoring surveys.

How to cite: Roman Dobarco, M., McBratney, A., Minasny, B., and Malone, B.: Digital pedogenon mapping as basis for assessing changes in soil condition, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-10460, https://doi.org/10.5194/egusphere-egu21-10460, 2021.

Due to high rate of nutrient removal by cotton plants, the productive cotton-growing soils of Australia is becoming depleted of exchangeable (exch.) cations. For long-term development, data on exch. calcium (Ca), magnesium (Mg), potassium (K) and sodium (Na) throughout the soil profile is required. However, traditional laboratory analysis is tedious. The visible-near-infrared (Vis-NIR) spectroscopy is an alternative; whereby, spectral libraries are built which couple soil data and Vis-NIR spectra using models. While various models have been used to predict exch. cations, their performance was seldom systematically compared. Moreover, most previous studies have focused on prediction of topsoil (0–0.3 m) exch. cations while the effects of depth on applicability of topsoil spectral libraries are rarely investigated. Our first aim was to determine which model (i.e. partial least squares regression (PLSR), Cubist, random forest (RF), or support vector machine regression (SVMR)) produces the best prediction of topsoil exch. Ca, Mg, K and Na. The second aim was to evaluate if the best topsoil model can be used to predict subsurface (0.3–0.6 m) and subsoil (0.9–1.2 m) cations. The third aim was to explore the effect of spiking on the prediction in subsurface and subsoil. The fourth aim was to see if combining all depths to build a profile spectral library improved prediction. Based on independent validation, PLSR was superior for topsoil exch. cations prediction, while Cubist outperformed PLSR in some cases when spiking was applied, and the profile spectral library was considered. Topsoil PLSR could be applied to predict exch. Ca and Mg in the subsurface and subsoil, while spiking improved prediction. Moreover, a profile spectral library achieved equivalent results with when topsoil samples coupled with spiking were considered. We, therefore, recommended to predict exch. Ca and Mg throughout the profile using topsoil spectral library coupled with spiking approach.

How to cite: Zhao, D., Arshad, M., Wang, J., and Triantafilis, J.: Soil exchangeable cations estimation using Vis-NIR spectroscopy in different depths: Effects of multiple calibration models and spiking, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-13575, https://doi.org/10.5194/egusphere-egu21-13575, 2021.

Due to certain socio-economic processes and technical pressure, the number of potential data sources targeting the Earth’s surface increases rapidly as well as the data generated by them. Soil mapping heavily relied on these changes in the paradigm shift, which took place in the population and interpretation of spatial soil information in the last decade. In digital soil mapping practice, auxiliary, environmental co-variables, which are related to soil forming factors and processes, have been taken into account in spatially exhaustive form. However, the potential hidden in spatially non-exhaustive (most frequently point-like), ancillary information – originating from observations also targeting the soil mantle – is far from being exploited. In their thematic features, accuracy and reliability they are inferior to primary field and/or laboratory measurements collected directly, but they are generated in more facile, cheaper way, in greater volume, with denser temporal and spatial coverage and characteristically they are available in significantly easier form. Data sequences of various installed field sensors, data collections by proximal sensing techniques, information supply by farmers and land managers as well as citizen science are considered as possible information sources. Essentially, the (soft) data supplied by them don’t provide spatially exhaustive coverage, neither direct pedological reference, nevertheless they are hypothesized to be utilized as auxiliary information within DSM framework. In a recently started project we started to investigate, (i) in which way and with what efficiency these ancillary information originating from different secondary sources can be applied, furthermore (ii) in what manner their application influences (hopefully improves) the results, accuracy and reliability of goal-specific spatial predictions. The elaborated digital mapping procedures, which are based on (i) large amount of data with differing quality and (ii) integrated geostatistical and data mining methods can be absorbed in various earth and environmental science applications.

Acknowledgment: Our research was supported by the Hungarian National Research, Development and Innovation Office (NKFIH; K-131820) and Gábor Szatmári by the Premium Postdoctoral Scholarship of the Hungarian Academy of Sciences (PREMIUM-2019-390).

How to cite: Pásztor, L., Szatmári, G., Laborczi, A., Mészáros, J., Takáts, T., Kovács, Z., Árvai, M., László, P., Koós, S., and Benő, A.: Application of information originating from spatially non-exhaustive ancillary observations in digital soil mapping, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-5116, https://doi.org/10.5194/egusphere-egu21-5116, 2021.

High quality global soil maps are crucial to face several challenges such as reducing soil erosion, climate change adaptation and mitigation, ensuring food and water security, and biodiversity conservation planning. To obtain accurate and robust soil properties maps, research and development are necessary to identify the most appropriate prediction models and to develop efficient and robust workflows. A few recent studies used Artificial Neural Networks (ANN) in Digital Soil Mapping, in some cases improving the accuracy of the predicted maps compared to other methods like Random Forest (RF). In this study we tested different ANN architectures on a global top-soil dataset of ca. 110 000 samples, comparing the results for the different architectures with the more traditional approach of RF. The target variables considered are pH, Soil Organic Carbon, Sand, Silt, and Clay. We selected 40 environmental covariates from a pool of over 400 to represent the most important soil forming factors. We tried simpler architectures (single input – single target) using point observations for one target variable with corresponding raster cell values for spatially explicit environmental covariates. We also used more complex architectures (multi input - multi target) incorporating contextual information surrounding an observation (convolutional) and with multiple target variables. Preliminary results show that increasing the number of hidden layers in the neural network does not significantly influence the results, while changing the type of architecture can play a bigger role in the overall accuracy of the model. The overall prediction accuracy of the ANN was comparable with the RF model. We conclude that ANN are a promising, relatively new, approach for Global Digital Soil Mapping and that further research is needed to improve performance.

How to cite: Genova, G., de Sousa, L., Mimmo, T., Borruso, L., and Poggio, L.: Global soil mapping with Neural Networks, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-15379, https://doi.org/10.5194/egusphere-egu21-15379, 2021.

Cropmarks are a major factor in the effectiveness of traditional aerial archaeology. The positive and negative features shown up by cropmarks are the role of the different cultivated plants and the importance of precipitation and other elements of the physical environment. In co-operation with the experts of the Eötvös Loránd University a new research was initiated to compare the pedological features of cropmark plots (CMP) and non-cropmark plots (nCMP) in order to identify demonstrable differences between them. For this purpose, the spatial soil information on primary soil properties provided by DOSoReMI.hu was employed. To compensate for the inherent vagueness of spatial predictions, together with the fact that the definition of CMPs and nCMPs is somewhat indefinite, the comparisons were carried out using data-driven, statistical approaches. In the first round three pilot areas were investigated, where Chernozem and Meadow type soils proved to be correlated with the formation of cropmarks. Kolmogorov-Smirnov tests and Random Forest models showed a different relative predominance of pedological variables in each study area. The geomorphological differences between the study areas explain these variations satisfactorily. In the next round, the identified relationships between cropmarking and soil features are planned to be utilized in the spatial inference of soil properties, where crop-marking sites will represent a unique, spatially non-exhaustive auxiliary information.

How to cite: Árvai, M., Czajlik, Z., Mészáros, J., Nagy, B., and Pásztor, L.: Cropmarks used in aerial archaeology as special spatial indicators of soil features potentially applicable in soil mapping, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-5204, https://doi.org/10.5194/egusphere-egu21-5204, 2021.

Recently, FAO and Global Soil Partnership (GSP) launched the Global Map of Salt-affected Soils (GSSmap) international initiative, which pursued a country-driven approach and aimed to update the global and country-level information on salt-affected soils (SAS). The objective of our study is to present how Hungary contributed to this international initiative by preparing its own SAS maps according to the GSSmap specifications. For this purpose, we used not just a combination of advanced machine learning and multivariate geostatistical techniques for predicting the spatial distribution of the selected SAS indicators (i.e., pH, electrical conductivity and exchangeable sodium percentage) for the topsoil (0–30 cm) and for the subsoil (30–100 cm), but also a number of image indices exploiting a huge amount of relevant information contained in Sentinel-2 satellite images. The importance plots of random forests showed that in addition to climatic, geomorphometric parameters and legacy soil information, image indices were the most important covariates. The performance of spatial modelling of SAS indicators was checked by 10-fold cross validation showing that the accuracy of the SAS maps was acceptable. By this study and by the resulting maps of it, we not just contributed to GSSmap, but also renewed the SAS mapping methodology in Hungary, where we paid special attention to modelling and quantifying the prediction uncertainty that had not been quantified or even taken into consideration earlier.

Acknowledgment: Our research was supported by the Hungarian National Research, Development and Innovation Office (NKFIH; K-131820 and K-124290) and by the Premium Postdoctoral Scholarship of the Hungarian Academy of Sciences (PREMIUM-2019-390) (Gábor Szatmári).

How to cite: Szatmári, G., Bakacsi, Z., Laborczi, A., Petrik, O., Pataki, R., Tóth, T., and Pásztor, L.: Elaborating Hungarian segment of the Global Map of Salt-Affected Soils (GSSmap), EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-902, https://doi.org/10.5194/egusphere-egu21-902, 2021.

Parent material is widely recognised as an important factor for soil formation. Thus, quantitative information on the lithogenetic, geochemical, and physical characteristics of the subsolum geological substrates (SSGS) are essential input parameters for digital soil mapping (DSM). Forming the interface between bedrock – the domain of geologists, and soil – the domain of soil scientists, spatial information on SSGS is however scarce. Recognising these shortcomings, a novel geochemical-physical classification system for subsolum geological substrates has been developed, in order to support DSM at a regional scale. The units of the classification system reflect the properties of the SSGS also considering multilayering structure of quaternary deposits. The basis for the classification are mineral component groups, namely dolomite, calcite, and felsic, mafic, and clay minerals. In order to test the relevance of SSGS for the prediction of spatially continuous physical and chemical soil properties, Generalized Additive Models (GAMs) were applied to the forested area of Tyrol, Austria. The plant-available water storage capacity, as a physical soil property, was predicted with r² = 0.56. The Ellenberg´s mean soil reaction indicator value for vegetation turned out to be a suitable proxy for soil pH value and was predicted with r² = 0.75. Topography and associated morphometric terrain features are formative characteristics of mountain areas and, due to its various effects on redistribution processes as well as on water and energy budget of forest sites, are considered as the most essential soil forming factors. Thus, variables derived from digital terrain models, which are available in high spatial resolution, are assumed to be one of the most important predictors for digital soil mapping. In our study we could show however, that SSGS information is the most relevant predictor for both investigated soil properties. In the plant-available water storage capacity model, the predictor variables related to SSGS account for around 76% of the variance explained. Accordingly, a special focus should be placed on the predictive relevance of parent material and the frequently unlocked potential of quantitative geological substrate information. Thus, the newly developed subsolum geological substrate information could stimulate further developments in digital soil mapping, especially in mountain environments.

How to cite: Simon, A., Wilhelmy, M., Klosterhuber, R., Geitner, C., and Katzensteiner, K.: On the role of parent material for predictive mapping of soil properties in mountain forests, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-9374, https://doi.org/10.5194/egusphere-egu21-9374, 2021.