Map Unit Composition Research Articles

Sampling and classifier modification to DSMART for disaggregating soil polygon maps

DSMART has been widely used to disaggregate multi-component soil polygon map units into raster maps. The algorithm randomly selects (simple random sampling approach) an equal number of synthetic sample points from all map units and assigns soil classes proportionate to the map unit composition using the C5.0 algorithm. Conditional Latin Hypercube sampling (cLHS) is a maximally stratified random sampling that guarantees full coverage of multivariate distribution. cLHS has shown success in digital soil mapping as it optimizes sample point selection based on environmental covariates. Information-driven stratification and sampling may help improve the assignment of soil classes within map units. The objective of this study was to compare cLHS against simple random sampling (SRS) approach using three classifier algorithms in DSMART to disaggregate the soil great group and soil series maps of a sub-watershed in Southern Ontario, Canada, as a case study example. SRS and cLHS were applied in two methods, sampling “by polygon” and “by area”, and three classifiers selected were C5.0, random forest (RF), and k-nearest neighbor (K-NN) (a total of twelve combinations − 4 sampling approaches and 3 classifier algorithms). The original R-code of the DSMART package was modified to incorporate cLHS sampling and change classifier algorithms. For the soil great group predictions, RF performed better in combination with “by polygon” SRS and “by polygon” cLHS approaches (Kappa scores 0.57). For the soil series predictions, C5.0 and “by area” cLHS approach (Kappa score 0.50) performed well compared to other combinations (Kappa score 0.25–0.44). Overall, cLHS performed marginally better with C5.0 and RF while, SRS performed better with K-NN. Disaggregated maps using cLHS generated lower prediction uncertainty values.

Disaggregation of conventional soil map by generating multi realizations of soil class distribution (case study: Saadat Shahr plain, Iran).

Conventional soil maps generally depict information about soil spatial distribution in the framework of crisp boundaries of tessellated soil polygons. Such maps in standard soil survey procedures determine map unit composition on the basis of relative acreage occupied by individual major and minor soil components within soil map unit without addressing the specific location of each component in the polygon boundary. These limitations in addition to the sharp-edge boundaries of conventional soil maps are considered obstacles for modern land resource management. To increase detail in the polygon of conventional soil maps, we have produced a spatially disaggregated soil class map of a relatively flat agricultural plain called Saadat Shahr in South-Central Iran, using DSMART algorithm. DSMART is a known DSM-based disaggregation and harmonization algorithm that works through resampled classification trees to estimate the probability of the existence of each possible soil classes and also to prepare the maps of the most probable soil class, second most probable, and so on in raster format. The conventional soil map and 124 georeferenced profiles, as well as a set of numerical and categorical auxiliary data in 10-m resolutions in the extent of the study area utilized as the SCORPAN variables, were used as the inputs of the DSMART algorithm. A set of 78 independent sampling points generated by Latin hypercube sampling scheme were investigated and then used for validation of the DSMART raster outputs. The results indicated an improvement in disaggregated maps in the case of allocating soil components within the map units. In the generated DSMART, overall accuracy for seven soil subgroups was 68%. The best prediction obtained for Typic Xerorthents and Typic Calcixerepts, meanwhile a few classes were poorly predicted. For second most probable and third most probable maps, 17% and 0.5% of predicted soils match that observed respectively. This study revealed that DSMART as a disaggregation method can be used for enhancing existence soil map with poor descriptive data in the case of allocating soil classes in a more detailed way compared to the relevant original map.

Harmonized soil property values for broad-scale modelling (WISE30sec) with estimates of global soil carbon stocks

Soils play a key role in providing a range of ecosystem services. Quality-assessed soil information, with quantified uncertainty levels, is needed to address a range of global issues. Traditional mapping methods, which recognize that soil classes are “important carriers of soil information”, were used to prepare an updated harmonized dataset of derived soil properties for the world at a nominal resolution of 30 by 30arc sec (WISE30sec). The map unit composition was determined using an overlay of the Harmonized World Soil Database, with minor corrections, and the Köppen–Geiger climate zones map as categorical co-variate. Property estimates for the respective component soil units were derived using taxonomy-based transfer rules that draw on a statistical analysis of some 21,000 soil profiles. Best estimates (mean±standard deviation) for twenty soil properties were calculated for seven depth intervals (up to 2m depth or less when thinner): organic carbon content, total nitrogen, C/N ratio, pH(H2O), CECsoil, CECclay, effective CEC, total exchangeable bases (TEB), base saturation, aluminium saturation, calcium carbonate content, gypsum content, exchangeable sodium percentage (ESP), electrical conductivity, particle size distribution (content of sand, silt and clay), proportion of coarse fragments (>2mm), bulk density, and available water capacity (−33 to −1500kPa); also the dominant soil drainage class. Coefficients of variation tend to be large. WISE30sec may be used for applications at a broad scale (<1:1M) upon consideration of the underlying data lineage, generalizations, and the associated uncertainties. As an example, the database was used to calculate the global soil organic carbon (SOC) stock to 2m depth. Some 30% (607±87PgC) of this stock (2060±215PgC) is held in the Northern Circumpolar Region, which is considered most sensitive to climate change.

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

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

Experimental Order 1 soil survey of vacant urban land, Detroit, Michigan, USA

An Order 1 soil survey was made of a vacant lot<0.1ha in size using a square grid with a two meter spacing (n=90 points) to test the hypothesis that there is a mappable pattern of anthropogenic polypedons against the alternative that the distribution is random. The vacant lot was formed circa 1998 by demolition of a wood-frame home from the 1920s in an urban residential setting in Detroit, Michigan. The anthropogenic soils studied were formed on a 0–2% slope under grass vegetation from artificial fill comprised of a mixture of poorly drained native soil materials, glacial sediments and artifacts (objects of anthropogenic origin), mainly waste building materials. The results suggest that anthropogenic soils on vacant urban land are mappable, even at the scale of a single vacant lot. The soils approximated an anthrosequence, i.e. a genetically related group of profiles whose characteristics differ mainly as a result of anthropogenic activity. The anthrosequence is comprised of soil phases whose range of characteristics can be used to define anthropogenic taxa, or to characterize the map unit composition of native soil–urban land complexes found on vacant property produced by building demolition. The results also suggest that, in the absence of historic records, demolition site history can be reconstructed from the spatial distribution of artifacts and soil types. More Order 1 surveys are needed to define the characteristics of anthrosequences in other urban settings and involving different soil orders, and to develop predictive urban soil–landscape models needed to provide better interpretive information for users of urban soil survey maps.

Disaggregation of component soil series on an Ohio County soil survey map using possibilistic decision trees

Data mining from existing County soil surveys can improve the utility of maps for research, management and decision making. Aggregated soil series information in soil survey map units can be disaggregated by following a possibilistic decision tree approach to provide maps at the soil series level. The “overall map unit composition percentage” available in the soil survey tabular data can be treated as a possibility distribution of the prevalence of different soil series within all occurrences of the map unit in the survey area. A case study was conducted in Monroe County in southeastern Ohio. Three different learning approaches including C4.5 decision trees, nonspecificity based possibilistic decision trees and clustering based possibilistic decision trees were applied to the existent County survey, and the efficiency of predicting soils at the series level was assessed using an independent soil series point data set. The results showed an improvement in prediction accuracy by following the clustered possibilistic decision tree approach. The maps were useful in identifying the locations of component soil series within map unit associations, consociations, and complexes. Data mining from existent soil survey maps using the component information available in the tabular data can serve as a guide for disaggregating soil map units to create soil series maps, identifying misplacement of polygon boundaries, identifying presence of inclusions, and correcting mislabeled polygons, when updating soil surveys.

Aggregation Strategies for SSURGO Data: Effects on SWAT Soil Inputs and Hydrologic Outputs

The USDA-NRCS Soil Survey Geographic (SSURGO) dataset provides the most comprehensive, detailed soil data coverage across the United States. Correct usage of these data within a hydrologic model depends on assumptions that define how soil property data are aggregated. To reduce data intensity and improve model efficiency, most hydrologic modeling studies using SSURGO assume that soil property data are adequately grouped into some notion of a “soil type” which is represented by the map unit, denoted as map unit key (MUkey), within SSURGO. However, the map unit is not intended for this purpose as continuity in map unit design or composition is not guaranteed between adjacent surveys of different vintages. This causes problems when several survey areas are used together, because similar soils are assigned a different map unit across the boundaries of soil survey maps. We present a methodology for aggregating soil data among multiple soil survey areas according to soil taxonomic information available in SSURGO. Results indicate that the aggregation method provides an acceptable representation of soil parameter values and distributions while eliminating the reliance on an arbitrary map unit for soil type identification. The results of the hydrologic modeling using the Soil and Water Assessment Tool (SWAT) in the San Joaquin River Watershed indicate that the commonly used aggregation method and the newly developed method satisfactorily estimated soil and surface hydrologic processes as compared to using a non-aggregated soil dataset. For the soil hydrologic processes, the SWAT model output from our aggregation method accurately estimated soil water content (mean difference compared to the non-aggregated soil dataset output of −4 mm for western San Joaquin River Watershed subbasins and 15 mm for the eastern San Joaquin River watershed subbasins) and lateral flow (3 mm for the western subbasins and 0.2 mm for the eastern subbasins) as compared to using a non-aggregated soil dataset. For the surface hydrologic processes, the SWAT model under predicted surface runoff (−0.5 mm for the western subbasins and −0.1 mm for the eastern subbasins) and sediment yield (−0.02 t/ha for the western subbasins and −9×10−4 t/ha for the eastern subbasins) as compared to using a non-aggregated soil dataset. While some variations were statistically significant, the differences were numerically small. The results show that soil taxonomy provides a robust framework for grouping soils.

Consequences of soil map unit uncertainty on environmental risk assessment

Soil maps are being applied in modelling of environmental risk at a regional scale. Analysis of soil maps shows that there can be considerable uncertainty in map unit composition with consequent spatial variability in soil properties within map units. Uncertainty of map unit composition arises from a number of factors including map labels that do not reflect the mix of soil types present and imprecise location of map unit boundaries. Usually little, if any, information is supplied as to the extent and magnitude of these uncertainties. In this paper we develop datasets to quantify soil property variability for map units on the Canterbury Plains and use expert knowledge to quantify uncertainty in map unit composition for a small case-study area. We then apply Monte Carlo sampling to the estimated range of soil properties to input soil data into the GLEAMS simulation model to assess the effect of soil variability on nitrate leaching. The simulation results show that there is considerable uncertainty in leaching from data derived from current soil map units. Even simple map units (map units characterised by a single soil type) have a wide range of potential leaching related to variation in soil profiles within the soil type. When map-unit inclusions were accounted for, variation in leaching risk can increase by 10–30%. Compound map units encompass the range of variability of the named component soils and there is a smaller increase in leaching variability in these units (compared with simple map units) when inclusions are added to the analysis. We discuss the implications of this study to the use of current soil survey data and make recommendations for future soil survey practice. In particular, greater effort is needed on description of the composition and reliability of soil map units.

A Comparison of Various Methods of Gathering Documentation for a Soil Survey

Simulated transects and random point methods of different sample styles and sizes were tested against the known data results of two map unit investigations, each with a large number of sample observations (three delineations each with a total of 309 observations per map unit). Statistics for epipedon thickness and epipedon thickness class placement were calculated for each method. The statistical mean of each method was compared to the mean and selected arbitrary lower and upper limits for the known dataset. Most methods tested approximated the data derived from the larger known dataset for both epipedon thickness and epipedon thickness class. The implication of this study is that most methods of documentation for map unit composition or estimating soil property measurements will approximate the larger known dataset. This is true whether it is a random or systematic method, and regardless of number of observations per delineation, as long as approximately the same number of observations are gathered per map unit.

Estimation of Map Unit Composition from Transect Data

AbstractWe are concerned with the documentation of taxonomic variability within soil survey map units, and how that variability is expressed statistically. Twelve randomly selected transects were used to sample soil properties and classify pedons within a fluvial map unit of the Boone County, Missouri, soil survey. Map unit taxonomic variability was high, with 28 taxonomic family classes. Twenty‐five percent of the sampled soils were in the named taxonomic class of the map unit, but most were similar inclusions. Interpretive purity of the sample was 83%. Variances about this sample proportion were calculated using three methods: (i) simple random sampling, (ii) cluster sampling, and (iii) two‐stage random sampling. The simple random sampling formula provided the smallest variance, compared with the cluster formula (55% larger) or the two‐stage formula (72% larger). However, the interpretation of transect sampling as simple random sampling is questioned. The two‐stage random sampling method seemed most conceptually correct, but is computationally intensive and requires questionable assumptions about the population size. Confidence intervals based on small sample sizes drawn from binomial populations are highly suspect, and perhaps should be referred to as “approximate bounds”. It is important to carefully consider the assumptions and effects of any statistical technique chosen to analyze map unit composition.

Use of Logistic Regression and GIS Modeling to Predict Groundwater Vulnerability to Pesticides

AbstractSoils are considered important components of many pesticide contamination models and are frequently the direct or indirect targets of pesticides applied during agricultural activities. Soil texture is commonly referenced on pesticide labels as an important factor in the selection and application of pesticides and in identifying target areas that are vulnerable to leaching. In general, no guidelines exist for the common interpretation of generic soil texture terms found on pesticide labels, for example, coarse or coarse‐textured soils. In the present study, a significant logistic regression model (P = 0.017) was developed that is based on the soil particle‐size class composition of sections containing wells sampled for DBCP (1,3‐dibromochloropropane). Particle‐size class is a concept used in Soil Taxonomy to describe soil family texture and is a component of soil family names. The model contains terms for the sandy and fine particle‐size classes. The model was validated using data obtained from sources independent of those used to develop the model. Records in the California Soils Map Unit Inventory database that describes the soil map unit composition of &gt;65 000 sections (1600 m2 or 1 mi2) were used to generate probability scores for &gt;15 000 sections located in the San Joaquin Valley, CA. A geographic information system, an information management technique that is becoming an accepted tool for a wide range of regulatory agencies, was used to generate visual images of the probability scores. A map was developed that depicts four distinct probability classes of the DBCP‐contamination status of section‐sized areas and their distribution within the study area.

Geographic Patterns of Soil Water‐Holding Capacity in the Contiguous United States

AbstractThe geographic patterns of soil water‐holding capacity are important for studying the response of vegetation and water supply to climate change. This study was conducted to evaluate which dataset to use for national‐scale water balance modeling. Spatial databases of soil properties for the contiguous USA were derived from the USDA national soil database (NATSGO) and the United Nations soil map of the world. The NATSGO data generally resulted in finer textures, greater rock fragment content, and more shallow soils. The organic matter content in the surface was similar, but the NATSGO database did not contain sufficient information to estimate organic matter content of the lower soil. The Rawls model was used to estimate soil water‐holding capacity of the 0‐ to 50‐cm depth from particle‐size distribution, organic matter content, rock fragment content, and soil depth data. The Saxton model was used to estimate the water‐holding capacity of the 50‐ to 150‐cm depth interval, because it didn't require organic matter data. The water‐holding capacity in the surface layer based on the NATSGO data was larger than that based on the soil map of the world despite greater rock fragment contents and shallower soil depths. There were few similarities in the water‐holding capacities of the 50‐ to 150‐cm depth layer because of large differences in rock fragment content and soil depth. The NATSGO dataset is recommended because the map unit composition was based on a statistical framework with a large sample size and it better characterizes rock fragment content and soil depth.

A method for identifying soil catenas and determining map unit composition used in a reconnaissance soil survey in Tanzania

A method of reconnaissance soil survey is described in which land units are delineated by photo interpretation and soils are investigated along traverses. The start points of traverses are randomly located along the existing road network within each land unit. Each traverse is oriented to cross an interfluve and is systematically sampled at intervals defined by the distance from the crest line to the margins of the valley floor on either side. Proportions of different soil types within each land unit can be calculated, with standard deviations if required, and typical soil catenas recorded. The method proved difficult to use in areas without clearly defined interfluves and where the geology and landform are complex. The authors conclude that, despite limitations, the method is rapid and is especially suited to terrain where soil distribution can be characterised by catenas. The method can be implemented by a number of survey parties to provide quantitative estimates of the proportions of soils within mapping units; this is of great benefit to land use planners.

Using Ground‐penetrating Radar to Study Changes in Soil Map Unit Composition in Selected Histosols

AbstractIntensive cultivation practices and subsidence of Histosols in Florida's Everglades Agricultural Area (EAA) has resulted in the need to reevaluate soils information gathered and published in the early 1970s as part of the National Cooperative Soil Survey Program. This study was initiated to determine the present range, thickness, distribution, and spatial variability of selected Histosol map units in the EAA using ground‐penetrating radar (GPR). Four consociations were selected for study. The thickness of the organic materials has decreased significantly when compared to the thickness in the published soil survey report. When compared with the GPR data, the proportions of areas mapped in the 1970s that have thick organic layers (&gt;90 cm) had decreased with a corresponding increase of areas with thin organic layers (&lt;90 cm). This shift represents approximately a 35% increase in lithic subgroups. Also, many areas previously mapped as consociations need to be recorrelated as complexes; only the Okeechobee muck map unit would remain a discrete consociation. Using the GPR increased sampling speed and enabled a continuous profile of organic thickness over bedrock to be obtained.

Estimating Map Unit Composition

Soil Survey HorizonsVolume 25, Issue 2 p. 21-25 Article Estimating Map Unit Composition R. J. Bigler, R. J. Bigler Soil survey party leader Soil Conservation Service, Devils Lake, ND, 58301Search for more papers by this authorK. J. Liudahl, K. J. Liudahl soil scientist Soil Conservation Service, Devils Lake, ND, 58301Search for more papers by this author R. J. Bigler, R. J. Bigler Soil survey party leader Soil Conservation Service, Devils Lake, ND, 58301Search for more papers by this authorK. J. Liudahl, K. J. Liudahl soil scientist Soil Conservation Service, Devils Lake, ND, 58301Search for more papers by this author First published: Summer 1984 https://doi.org/10.2136/sh1984.2.0021Citations: 2Read the full textAboutPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat Citing Literature Volume25, Issue2Summer 1984Pages 21-25 RelatedInformation

Determination of Map Unit Composition by a Random Selection of Transects

AbstractA procedure for random sampling of mapping units and a statistical method for analyzing field data have been evaluated in five Alabama soil surveys. Taxonomic composition, consistency of delineations, and minimum number of transects needed for mapping unit characterization were determined by routine field examination, investigation of randomly selected transects, and statistical analysis of field data. The method requires that soil delineations be identified by landscape features with minimum traversing and observations and that potential transects (transects that may be selected for study) be located while mapping. Each delineation is represented by at least one potential transect; potential transects should traverse areas that represent the concept of a mapping unit. These transects make up the total available population of a mapping unit. From this population, randomly selected transects are sampled for detailed study. A minimum of 10 point observations is made on each. This procedure has shown that many mapping units in a survey area can be characterized at 80% confidence level by determining soil occurrence along less than 10 transects. In some cases it may be necessary to group soil series into similar interpretive units to reduce transect numbers and to present data that can be used.

Map Unit Composition Assessment Using Drainage Classes Defined by Landsat Data

AbstractSoil survey map units are designed such that the dominant soil represents the major proportion of the unit. At times, soil mapping delineations do not adequately represent conditions as stated in the map unit descriptions. Digital analysis of Landsat multispectral scanner (MSS) data provides a means of accurately delineating and quantifying soil map unit composition.Digital analysis of Landsat MSS data collected on 9 June 1973 was used to prepare a spectral map for a 430‐ha area in Clinton County, Indiana. Sixteen spectral classes were defined, representing 12 soil and 4 vegetation classes. The 12 spectral soil classes correlated with soil drainage characteristics and were grouped into four drainage classes. The four spectral vegetation classes were grouped for convenience into one vegetation class.The soil delineations produced using these groupings were compared to a conventionally prepared soil map. Three map units were investigated in detail: (i) Mahalasville silty clay loam, poorly drained; (ii) Reesville silt loam, 0 to 2% slopes, somewhat poorly drained; and (iii) Xenia silt loam, 2 to 6% slopes, eroded, moderately well drained.Results indicate that soil drainage characteristics can be identified. Correlation of drainage characteristics with soil series allows for the composition of soil map units to be accurately ascertained.