Traditional Soil Survey Research Articles

Soil loss estimation using RUSLE model: Comparison of conventional and digital soil data at watershed scale in central Iran

Sustainable agriculture and eco-environment conservation face constant threats from soil erosion in mountainous regions. This study aimed to predict soil loss in the mountainous watershed by applying the Revise Universal Equation Soil Loss (RUSLE) and Sediment Delivery Ratio (SDR) models and to evaluate the efficacy of the RUSLE model through the observed data in a hydrometric station. Comparing a polygonal, and a digital soil map (prepared by the Kriging method) for deriving the K factor to estimate soil loss was another objective of the recent research. In this study, two scenarios were examined for the calculation of spatial variation of the K-factor; scenario I comprised soil data derived from a legacy soil map with eight soil units and the data from their representative soil profiles (i.e., traditional soil survey), and scenario II comprised K-factor derived from 100 studied sites using kriging technique (i.e., digital soil survey). The results of RUSLE modeling showed that there was no significant difference between scenario I (7.97 t/(ha. yr)) and scenario II (7.93 t/(ha. yr)) for predicting soil loss with the RUSLE model. This finding confirms that the RUSLE model with limited and legacy data can provide a reliable prediction of soil loss in the given watershed. Moreover, long and short-term periods were used to estimate soil loss, while in the long-term period, estimated soil loss had higher accordance with actual soil loss. Sediment delivery ratio was calculated using models of Vanoni, Boyce, USDA-SCS, and Slope-based models. The results indicated that the Slope-based model had the highest fitness (R2=0.946, ME=0.27, and RMSE=0.275 for scenario I and R2=0.9443, ME=0.26 and RMSE=0.40 for scenario II), with observed sedimentation rate at the hydrometric station at the outlet of the watershed. Overall, predicted soil loss severity map using the traditional soil map by RUSLE model could provide trustworthy information for decision makers and governors at the given and similar watersheds in semiarid regions.

Prediction of soil organic carbon in soil profiles based on visible–near-infrared hyperspectral imaging spectroscopy

Complete and coherent soil profile information is important for studying soil morphological characteristics, vertical variation patterns of soil physicochemical properties, and identifying soil types. However, traditional soil survey methods use a limited number of discrete sample points at different depths as a data source to explore the correlation between soil depth and soil properties. Due to the variability of soil properties at vertical depth, there is a strong uncertainty in the conclusions drawn from the limited sample points. With the development of near-Earth remote sensing technology, the near-Earth hyperspectral imaging system can acquire hyperspectral images of soil profiles and characterize the fine texture differences and spectral variations of soil with continuous spectral curves and images, which is useful for inversion of soil properties. Soil organic carbon (SOC) is one of the important physicochemical properties to identify the genetic horizon, therefore, this paper explored the ability of visible–near-infrared (Vis-NIR) hyperspectral imaging spectroscopy to predict SOC in soil monoliths. The dataset contained five soil monoliths retrieved from the upper Yellow River, China. A random forest (RF) model was developed for predicting organic carbon, and the organic carbon content in the genetic horizon was used to establish a depth function to obtain the continuous content variation with the depth. These differences with and without depth function modeling, and the effect of three equal-spacing sampling schemes (1, 5, and 10 cm) on the depth function and the accuracy differences in five spectral statistics of the maximum (max), mean, median, mode, and minimum (min) within the image region of interest (ROI) were compared. The results indicated that (1) the depth function under the 1-cm sampling scheme predicted organic carbon the best, (2) the mode of the spectra within the ROI could yield a comparable or even a higher accuracy than that obtained with the mean, and (3) imaging spectroscopy could be used to visualize the distribution of organic carbon on soil profiles with relatively uniform morphology. These results indicated that imaging spectroscopy could be used for predicting organic carbon in undisturbed soil profiles.

Prediction of Soil Properties Using Quantile Regression Forest Machine Learning Algorithm – A Case Study of Salem and Rasipuram Block, Tamil Nadu, India

Digital soil mapping is a growing technology for mapping soil properties instead of conventional soil mapping. Especially for what are all countries have large geographical areas and human, not accessible areas. Compared to conventional soil mapping it is cost-wise less and more accurate. At the world level, globalsoilmap.net has taken the initiative for creating digital soil maps. In India like countries very much needed for digital soil mapping, is essential for agricultural planning, and decision-makers decide on it. This study predicted the soil properties such as sand, silt, clay, pH, and OC using the Quantile Regression Forest machine learning algorithm also provides uncertainty. The main aim of this study was to predict the soil properties in the top two depth intervals such as surface and subsurface. For achieving this goal, 56 soil samples were collected across the study area, and many environmental covariates were used for that such as DEM derivatives, satellite imagery, and Climatic Data. This study, using 56 soil samples data taken from the traditional soil survey, is a limited number of soil samples this tried to achieve a higher accuracy result using QRF.

Soilscapes of New Zealand: Pedologic diversity as organised along environmental gradients

Soilscapes are a regionalisation of the landscape into a suite of contiguous segments with an homogeneous set of soil forming factors. The concept has been present in the soil science literature since the late 1950s, proving to be a useful framework for generalising traditional soil survey maps. This concept is inherently multi-scale, and current approaches generally build either on existing soil maps, or on cluster analysis of environmental proxies of soil forming factors. In this study, in parallel with the concept of beta-diversity in numerical ecology, we built a soilscape regionalisation for New Zealand based on detecting the turnover in soil types along different environmental gradients. It was implemented using Gradient Forest, a method that analyses the frequency of species (in our case, soil type) turnover alongside such gradients, which are then ranked relative to their ecological (in our case, pedological) relevance. Based on this analysis, the environmental covariates were transformed to amplify their most important parts in terms of soil type turnover. The transformed variables were then segmented into contiguous regions, at three different levels of scale. Finally, hierarchical clustering was used to produce discrete soilscapes, at different scales, and for the entire country. This approach was developed using pseudo-observations derived from a national scale coverage of soil types, and validated using an independent set of actual soil profile observations. Validation results reveal strong agreement between the dissimilarity quantified between soilscapes and the taxonomic distance between actual soil profile observations (R = 0.61, P < 0.001). We expect the resulting layers from this soilscape delineation exercise to be used for a diverse suite of pedometrics applications, from stratifying soil sampling schemes, to optimising the soil survey boundaries, and for digital soil mapping.

Comparison of Models for Spatial Distribution and Prediction of Cadmium in Subtropical Forest Soils, Guangdong, China

Cadmium (Cd) is a toxic metal and found in various soils, including forest soils. The great spatial heterogeneity in soil Cd makes it difficult to determine its distribution. Both traditional soil surveys and spatial modeling have been used to study the natural distribution of Cd. However, traditional methods are highly labor-intensive and expensive, while modeling is often encumbered by the need to select the proper predictors. In this study, based on intensive soil sampling (385 soil pits plus 64 verification soil pits) in subtropical forests in Yunfu, Guangdong, China, we examined the impacting factors and the possibility of combining existing soil information with digital elevation model (DEM)-derived variables to predict the Cd concentration at different soil depths along the landscape. A well-developed artificial neural network model (ANN), multi-variate analysis, and principal component analysis were used and compared using the same dataset. The results show that soil Cd concentration varied with soil depth and was affected by the top 0–20 cm soil properties, such as soil sand or clay content, and some DEM-related variables (e.g., slope and vertical slope position, varying with depth). The vertical variability in Cd content was found to be correlated with metal contents (e.g., Cu, Zn, Pb, Ni) and Cd contents in the layer immediately above. The selection of candidate predictors differed among different prediction models. The ANN models showed acceptable accuracy (around 30% of predictions have a relative error of less than 10%) and could be used to assess the large-scale Cd impact on environmental quality in the context of intensifying industrialization and climate change, particularly for ecosystem management in this region or other regions with similar conditions.

Integration of multispectral and hyperspectral data to map magnetic susceptibility and soil attributes at depth: A novel framework

The understanding of attributes and magnetic susceptibility (χ) at soil surface, mainly subsurface, is crucial due to their role to identify climate changes, soil degradation, soil classification systems, soil fertility, and pedogenesis. The integration of proximal sensing (PS) and remote sensing (RS) data sources could increase the efficiency of Digital Soil Mapping. Nevertheless, products of this integration need to be evaluated in hybrid, stochastic, and deterministic models to predict soil attributes and χ at surface and subsurface. This study investigates the PS and RS integration by applying four deterministic (e.g. Bayesian Regularised Neural Network, Generalised Linear Model, Random Forest and Cubist) and hybrid models (e.g. Regression Kriging of residuals of the best-fitted model) to create a new environmental variable, the Best Synthetic Soil Image (BSSI), at three soil depths (e.g. 0 – 20, 40 – 60 and 80 – 100 cm) that quantitatively represent the soil spectral signature. We also used the BSSI in a comparison with bare soil surface (e.g. SYSI - Synthetic Soil Image) to predict soil attributes and χ by performing the deterministic and hybrid models. We hypothesize that the BSSI, which integrates PS and RS data, enhances soil modelling predictions at subsurface by selecting the best model approach. The BSSI demonstrated original and valuable contribution to increase the predictive model power at deeper layers, while SYSI was effective at upper layers. The PS and RS integration helped to identify the main soil patterns horizontally and vertically, which traditional soil surveys have not been capable of representing.

Lithological indicators of discontinuities in mountain soils rich in calcium carbonate in the Polish Carpathians

Six soils located within the Polish Carpathians, developed on calcium carbonate–rich sedimentary parent materials and representing various reference groups, were investigated in order to detect the lithic discontinuity. We propose using a multidirectional approach to assess the lithic discontinuity in these soils, one that includes grain size distribution, geochemical composition, heavy mineral content and micromorphology, supported by a traditional soil survey. A further aim of this process was to identify the possible admixture of allochthonous material of aeolian origin. The studied soils presented lithic discontinuities mostly at the contact of underlying calcium carbonate–rich coarsegrained slope deposits with the overlaying colluvium layer having a lower content of rock fragments. The significant changes in grain size distribution, especially in the silt and sand content, as well as high Uniformity Values and partially, high Lithological Discontinuity Index values, confirmed the occurrence of a lithic discontinuity in all studied soils. High heterogeneity in the soil profiles was also confirmed by the distribution of the major oxides; however, their distribution did not clearly indicate the lithic discontinuity. The most visible distinctions were noted from CaO content, which resulted from the deposition of carbonate-free materials (aeolian silts) and their mixing with the calcium carbonate–rich parent material. Furthermore, the analysis of heavy mineral content confirmed the allochthonous origin of the upper (and in some cases also the middle) parts of all profiles, which was manifested by the presence of highly weathering-resistant minerals such as zircon, epidote and various types of garnets. The micromorphological features of some of the studied soils showed distinctiveness within the soil profile, manifested by changes in b-fabric pattern, the occurrence and distribution of secondary carbonate and the coarse and fine coarse and fine ratio. Based on the high content of silt within the upper and middle parts of the soils, the content of Hf and Zr, as well as the higher content of weathering-resistant minerals, admixture of aeolian silt could be considered in some of the studied soils, yet with weak character. However, the dominance of minerals typical for metamorphic and igneous rocks suggested that the supply of aeolian silt was associated with loess covers rather than local sedimentary material.

Soil health cluster analysis based on national monitoring of soil indicators

AbstractA major challenge in soil science is to monitor and understand the state and change of soils at a national scale to inform decision making and policy. To address this, there is a need to identify key parameters for soil health and function and determine how they relate to other parameters, including traditional soil surveys. Here we present a national‐scale dataset of topsoil sampled as part of a wider agri‐environment monitoring scheme in Wales, UK. Over 1,350 topsoils (0–15 cm) were sampled across a very wide range of habitats and a range of physical, chemical and biological soil quality indicators were measured. We show consistent differences in soil physicochemical properties across habitat types, with carbon decreasing and pH increasing across the habitat productivity gradient from bogs through woodlands and grasslands to arable systems. The soils within our dataset are largely within the limits identified as important for supporting habitat function, with the exception of excessive phosphate levels in mesotrophic grassland. Cluster detection methods identified four soil functional classes based on measured topsoil properties, which were more related to habitat type than the genesis‐based soil classification from soil maps. These soil functional classes can be interpreted as phenoforms within the soil genoforms found by traditional soil classification. This shows the importance of land‐use management in determining the soil health and functional capacity of soils. Our work provides an account of the current state of soil health in Wales, its relationship to soil function and a baseline for future monitoring to track changes against agri‐environment and other policy targets.HighlightsWe measured soil physicochemical properties in ~1,350 sites in a variety of temperate habitatsThere was a strong gradient in soil carbon and pH, with other variables being correlated with theseMesotrophic grassland sites had phosphate levels above the identified limit for good functioningSoil classes from topsoil properties were more related to land use than soil map classifications

Digital soil mapping for fire prediction and management in rangelands

BackgroundSoil properties have important effects on fire occurrence and spread, but soils are often overlooked in fire prediction models. Quantifying soil−fire linkages is limited by information in conventional soil maps, but digital soil mapping products (e.g., detailed soil property maps) could improve both wildfire prediction models and post-fire management decisions.ResultsOf our estimated 3.7 Mkm2 of rangeland in the continental US and Alaska, an average of 38 000 km2 burned per year between 2008 and 2017. To highlight the role of soils in fire ecology, we present 1) a conceptual framework explaining why soil information can be useful for fire models, 2) a comprehensive suite of literature examples that used soil property information in traditional soil survey for predicting wildfire, and 3) specific examples of how more detailed soil information can be applied for pre- and post-fire decisions.ConclusionsDigital soil mapping can improve fire prediction models and inform post-fire management decisions.

3D spectral analysis in the VNIR–SWIR spectral region as a tool for soil classification

Visible, near-infrared and shortwave-infrared (VNIR–SWIR) spectroscopy has proven to be an efficient, rapid and low-cost method for soil spectral analysis that can improve on the results obtained from today's traditional methods of conducting soil surveys. Nonetheless, this tool is used mostly in the laboratory and at surface level. The main objective of this paper is to develop a new optical method for characterizing soil profiles, towards improving the efficiency and accuracy of the traditional soil survey. We used airborne hyperspectral data from the AisaFENIX sensor for surface classification and ASD spectral measurements of soil samples for subsurface analysis. A total of 643 soil samples were extracted from 48 cores, each core representing a soil profile. All samples were air-dried, crushed and sieved, and then analyzed by ASD spectrometer under laboratory conditions. Clay content was also measured to provide additional information. The 3D spectral data were analyzed using SAM algorithm, spectral gradient (m), k-means clustering and gley horizon parameter (G) to classify soils and distinguish between soil horizons in each core. The results suggest that these parameters can provide satisfactory results both from laboratory measurements and hyperspectral remote sensing data (R2=0.81 for clay content and R2=0.78 for gleying conditions) in order to distinguish between the soil horizons using 3D spectral information. Moreover, the method is satisfactory for obtaining soil types from 3D spectral sensing as well as evaluating the catena development and other spatial soil distributions.

Neighborhood Size of Training Data Influences Soil Map Disaggregation

Core Ideas Focal summaries of covariate data around sampling points affect model performance. Support vector machine and random forest approaches produced the best results. A 150‐m neighborhood emerged as the best model, albeit with a general soil map. Multiscale covariate data reflect realistic patterns of soil–landscape features. Soil class mapping relies on the ability of sample locations to represent portions of the landscape with similar soil types; however, most digital soil mapping (DSM) approaches intersect sample locations with one raster pixel per covariate layer regardless of pixel size. This approach does not take the variability of covariate information adjacent to the training data into account. The objective here was to disaggregate a soil map in a semiarid Arizona rangeland (78,569 ha) by exploring different neighborhood sizes for extracting covariate data to points. Eight machine learning algorithms were compared to assess the influence of summarizing covariate data in 0‐, 15‐, 30‐, 60‐, 90‐, 120‐, 150‐, and 180‐m circular neighborhoods and a multiscale model. Κ values of all models ranged between 0.24 and 0.44 and increased with neighborhood size up to 150 m. Support vector machine and random forest algorithms performed best across all scales. The radial support vector machine model using a 150‐m neighborhood had the highest Κ and produced a more generalized map compared with the best multiscale model (random forest), which resulted in a mix of general and detailed soil features. Evaluating a range of neighborhood sizes for aggregating covariate data provides a method of accounting for multiscale processes that are important for predicting soil patterns without modifying the pixel size of the final maps. Incorporating concepts from traditional soil surveys with DSM approaches can strengthen ties between them and optimize the extraction of landscape information for predicting soil properties.

Прогресивний підхід до традиційних великомасштабних обстежень ґрунтів

Прогресивний підхід до традиційних великомасштабних обстежень ґрунтів

Analysis of Digital Elevation Model and LNDSAT Data Using Geographic Information System for Soil Mapping in Urban Areas

This study applies digital analysis methods of topographic data derived from digital elevation models (DEMs) and Landsat remotely sensed spectral data using GIS tools to evaluate the quality and limitations of the morphometric parameters (terrain attributes: TAs). This aims to check its suitability for digital soil mapping (DSM) and survey in urban areas at the target scale 1:50,000. This scale represents the standard scale level for compiling soil inventories within all German states. The study is conducted on an urban area of 112.68 km2 in the southwest part of the state of Berlin in Germany. These relief units are the basis for determining the soil mapping units at the scale of 1:50,000. The generated preliminary soil map was compared to soil maps made using traditional soil survey methods. For the mainly natural soils, the equivalence area is 94.91%, and for the anthropogenic soils, the equivalence area is 95.34%. The proposed methodology is adequate for preliminary mapping of soil units based on the digital derivation of TAs. Landsat scenes are spatially explicit, physical representations of environmental covariates on the land surface. The free DEM-ASTER in combination with Landsat OLI images is found to be the appropriate model to represent the terrain surface and derive the TAs for environmental modeling and fitting of derivation the relief units and their topography features. However, the 30 m spatial resolution and the fairly coarse spectral resolution of DEMs and Landsat images limit their utility for digital soil mapping at this scale in urban areas with little topographic variation.

Validating a Digital Soil Map with Corn Yield Data for Precision Agriculture Decision Support

Capturing the variability in soil‐landscape properties is a challenge for grain producers attempting to integrate spatial information into the decision process of precision agriculture (PA). Digital soil maps (DSMs) use traditional soil survey information and can be the basis for PA subfield delineation (e.g., management zones). However, public soil survey maps provide only general descriptions of soil‐landscape features. Therefore, improved DSMs are needed that use high‐resolution data that more precisely model soil‐landscape characteristics. Additionally, reliable methods are needed to validate DSM products for PA. The objective of this study was to validate with corn (Zea mays L.) yield data the performance of a new DSM product, termed Environmental Response Unit (ERU), compared with the USDA Soil Survey Geographic (SSURGO) soil map. The ERU was developed by integrating SSURGO information with high‐resolution elevation data. For validation, corn yield maps were collected and corrected for common data collection errors from 409 fields across Indiana, Iowa, Minnesota, and Nebraska in 2010 to 2012. Reductions in the area‐weighted variance (Rv) of corn yield for ERU and SSURGO were calculated relative to the whole‐field variance. The average Rv across all site‐years for SSURGO and ERU was 16 and 25%, respectively, which equated to a 57% higher median yield variance reduction with ERU over SSURGO. This variance reduction technique showed the potential of ERU as an improved model better representing soil‐landscape properties that impact corn yield. This research also has application potential for determining the success of a DSM for identifying management zones in PA. The variance reduction metric (Rv) provided a direct comparison between DSM models. The greatest reduction in corn yield variance was exhibited by the ERU DSM model. Physiographic derivatives from high‐precision data sets improved DSM functionality. Yield data helped outline interrelationships among various soil properties.

Permafrost Carbon: What soil science has to say about its distribution and fate

Permafrost Carbon: What soil science has to say about its distribution and fate

Investigating geostatistical methods to model within-field yield variability of cranberries for potential management zones

Cranberry harvesting methods give only one yield value per field making characterization of within-field variation, the usual first step in precision farming, difficult. Time-consuming berry count yield and fruit rot estimations are the best “ground truth” indication of yield variation within fields. Correlations and coincidence of binary classifications based on less expensive methods such as enhanced vegetation index (EVI) from imagery, and area to point (AtoP) kriging of useable, poor quality and trash yields were compared with this “ground truth”. In general AtoP kriged values gave higher correlations and kappa statistic values with berry counts and fruit rot than EVI. Geostatistical disaggregation of per field yield totals using AtoP kriging with EVI as an external drift (AtoPKED) was also investigated. Factorial kriging was used to separate the several scales of variation in “ground truth” and EVI data and determine which ones were most spatially coherent/manageable and which related best to the AtoP kriged data. The spatial trend component of pre-harvest berry counts and AtoP kriging of yields both gave a good initial definition of spatially coherent, relatively permanent management zones. They were related to topography and depth of water table in the soil which are key factors governing cranberry yield. AtoP kriging or AtoPKED are recommended for defining management zones as they are less expensive than berry counts. The value of AtoP kriging to precision farmers for other crops to map soils at the farm scale with some imagery and just one bulked soil sample per field or use nutrient levels associated with each polygon of traditional soil survey maps is discussed in the conclusions.

A Pedological Tale IV: Conversations with Turtle, Bear, and Eagle

A Pedological Tale IV: Conversations with Turtle, Bear, and Eagle

Tihonenko D. G., Degtiarov V. V., Gorin M. O., Veremeenko S. I., Furman V. M., Gavva D. V. Cartography of soils / Ed. D. G. Tihonenko, M. O. Gorin. – Kharkiv : Maidan, 2014. – 494 p.

In the textbook, the difficult issues of cartography of soils are considered as an independent section of Soil Science. The proposed program of the discipline "Cartography of soils" provides the detailed research by future specialists – soil scientists and agrochemists of theory and practice of soil mapping and preparation of cartograms, serving as cartographic basis of rational land use, which is a basic mean of production. The reviewed textbook is powerful on the content, but from the point of view of a scientist and a producer, the volume makes a service and performs another, pocket function, becomes a reliable reference book that is needed very much in-process. A positive factor is almost exhaustive information about the modern electronic-ecological and space methods of researches. A specialist must learn and get acquainted with the newest methods, as in the near time a modern productive equipment of the remote methods, undoubtedly, will be basic in cartographic work. The collective, complex, well justified experience of application the methods of remote sensing of the Earth and GIS technologies for soil mapping. Much attention is paid to the traditional (classic) soil surveys. Without this knowledge, it is difficult to present the successful mastering of new methods of mapping. In theoretical terms, it is important that the soil is considered by the authors as a component of biogeocenosis, ecosystem, geosystem. This is consistent with the opinion the great majority of the classics of the modern theory of ecosystems. After each section of the textbook the checklists and the tasks for independent work are offered. The tutorial "Cartography of soils" is based on the textbook of the same name, prepared under the editorship of Professor D. G. Tihonenko by the group of authors: D. G. Tihonenko, N. A. Gorin, S. V. Rybalko, V. I. Sydorenko, A. A. Remynsky, A. A. Georgy, M. A. Shchukovsky, A. I. Gorb, A. F. Saenko, S. A. Baluk, L. L. Velichko, M. V. Lesnoy, V. I. Filon, I. M. Belaya, V. P. Poluyanov, A. I. Fateev, the editor and originator – N. A. Gorin. When preparing the textbook the authors have considered the experience of the Lviv-Odessa school of soil mapping, which is presented in the textbook "Cartography of soil cover" and "Factors of soil formation" (S. P. Poznyak, E. N. Kraseha, M. G.Kot), the experience of Kherson State Agrarian University – "Geoinformational systems in the agricultural sphere" (V. V. Morozov and A. V. Morozov and colleagues), the experience of the National Scientific Centre "A. N. Sokolovsky Institute of Soil Science and Agricultural Chemistry" of NAAS of Ukraine – "The problems and prospects of remote sensing of soils of Ukraine” (S. G. Truskavetsky), the experience of the Kharkiv school (Karazin Kharkiv National University) of cartographicalproviding with ecological researches and nature conservation (V. A. Peresadko), the experience of Zaporizhzhya filial brunch “The Institute of soil conservation of Ukraine”, agrochemical cartography (A. A. Syzonenko), etc. The disadvantages include a surprisingly meager print run of the textbook, which comes under a signature stamp given by MES of Ukraine, in 300 copies, which in any case does not provide its needs at the time when the cartographic work started again throughout the territory of Ukraine.

Refining a reconnaissance soil map by calibrating regression models with data from the same map (Normandy, France)

Reconnaissance soil maps at 1:250,000 scale are the most detailed source of soil information for large parts of France. For many environmental applications, however, the level of detail and accuracy of these maps is insufficient. Funds are lacking to refine and update these maps by traditional soil survey. In this study we investigated the merit of digital soil mapping to refine and improve the 1:250,000 reconnaissance soil map of a 1580km2 area in Haute-Normandie, France. The soil map was produced in 1988 and distinguishes nine soil class units. The approach taken was to predict soil class from a large number of environmental covariates using regression techniques. The covariates used include DEM derivatives, geology and land cover maps. Because very few soil point observations were available within the area, we calibrated the regression model by sampling the soil map on a grid. We calibrated three models: classification tree (CT), multinomial logistic regression (MLR) and random forests (RF), and used these models to predict the nine soil classes across the study area. The new and original maps were validated with field data from 123 locations selected with a stratified simple random sampling design. For MLR, the estimate of the overall purity was 65.9%, while that of the reconnaissance map was 55.5%. The difference between the purity estimates of these maps was statistically significant (p=0.014). The significant improvement over the existing soil map is remarkable because the regression model was calibrated with the existing soil map and uses no additional soil observations.

Digital Mapping of Soil Classes Using Decision Tree and Auxiliary Data in the Ardakan Region, Iran

Digital soil mapping (DSM) involves acquisition of field soil observations and matching them with environmental variables that can explain the distribution of soils. The harmonization of these data sets, through computer-based methods, are increasingly being found to be as reliable as traditional soil mapping practices, but without the prohibitive costs. Therefore, the present research developed decision tree models for spatial prediction of soil classes in a 720 km2 area located in an arid region of central Iran, where traditional soil survey methods are difficult to undertake. Using the conditioned Latin hypercube sampling method, the locations of 187 soil profiles were selected, which were then described, sampled, analyzed, and allocated to six Great Groups according to the USDA Soil Taxonomy system. Auxiliary data representing the soil forming factors were derived from a digital elevation model (DEM), Landsat 7 ETM+ images, and a map of geomorphology. The accuracy of the decision tree models was evaluated using overall, user, and producer accuracy based on an independent validation data set. Our results showed some auxiliary variables had more influence on the prediction of soil classes which included: topographic wetness index, geomorphological map, multiresolution index of valley bottom flatness, elevation, and principal components of Landsat 7 ETM+ images. Furthermore, the results have confirmed the DSM model successfully predicted Great Groups with overall accuracy up to 67.5%. Our results suggest that the developed methodology could be used to predict soil classes in the arid region of Iran.