My Account: Log In | Join | Renew
Table of Contents
Select All Chapters
Smart systems and geographical information system (GIS) techniques allow for the display of decisions that are humanistic; however, in soil survey, little effort has been dedicated to incorporating uncertainty into soil interpretations or to quantifying the risk assumed by using such information. Review of the literature and current research indicate that efforts have been made in several disciplines that may be applicable to development of a risk-based approach to soil interpretations. Such methodology can be made available in a user friendly personal computer (PC) environment and has the potential for an increased use of soil survey information.
Soil Survey is a comprehensive program to inventory and interpret soil resources. It is based on working models that explain the relationships between sets of soils properties and landscape features. Scientists identify, describe, classify, map, and interpret soil landscapes. Map polygons are rarely homogeneous due to small map scales relative to variability on the ground and to the conventions used to define and classify soils. Particular attention is given to soils that are limiting for most uses compared with the dominant soils in a map unit. Emphasis on sustainable productivity and ecosystem stability has increased interest in the reliability of standard survey products and services to facilitate the transfer of soil-related technology. Most land use decisions involve interpretations of data, such as map unit response information, rather than the data itself. The capability to consistently identify and delineate areas of a map unit gives rise to precision. The capability to correctly identify, locate, and describe included components, especially the use-limiting ones, is associated with accuracy. It is not possible to state with certainty the reliability of soil survey information because the quality and utility of a product or service is what the customer believes it is. The provision of information and its acceptance each has a risk. The risk of the consumer is that of using information that has an unacceptable probability of providing information that is accurate and relevant for a given decision. Quality of data and interpretations can be enhanced with additional documentation of the relationships underlying the mapping and interpretation processes. Estimates of accuracy and precision of soil map units accompanied by levels of confidence may be useful starting points to minimize consumer risks and improve reliability.
Soil survey information (SSI), such as soil survey reports and soil databases, is increasingly used by individuals lacking the background to make informed judgments of the quality or applicability of the SSI. To reduce misuse of SSI, a uniform, National Cooperative Soil Survey sanctioned system to assess and rate the quality of SSI is needed. This chapter presents a first approach to the development of the logic and elements of such a system. The system rates both qualitative and quantitative SSI so that both kinds of ratings can be equated in decision-making. Ratings are encoded in a format that facilitates computer storage and retrieval. A fuzzy set logic technique is used to illustrate the use of expert opinion to assign a SSI quality class rating within the system.
These comments are derived from more than 10 years experience in progressive soil survey and more than 25 years as a soils consultant. These experiences have verified that the maps and data contained within soil survey reports are accurate, especially for more extensive interpretations. Accuracy of the soil survey map and resultant interpretations decrease with decrease in sample size and with increase in geologic and geomorphic complexity of the subject area. Risk of improper site assessment increases with decrease in accuracy of the soil survey, with decrease in sample size and with increase in divergence from the modal concept of each interpretive group (soil series or phase). When several interpretative groups are used to characterize a small site, the risk of the combined site characterization being faulty increases rapidly. The recurring lesson is that to maintain the highest level of accuracy of on-site investigations, classifications and evaluations must be carried out by an experienced pedologist.
STATSGO maps can be very useful complements to research on environmental applications of soil survey information provided that one fully understands and appreciates both the way soil groupings were formed to construct the legend and the structure of the associated attribute files. Data needs associated with the legend are for more precise soil composition information for individual delineations, and for more detailed soil survey information in areas of unsurveyed or poorly understood soil-landscape distributions. Data needs associated with the attribute files occur primarily because of the need for data that are not currently in the database. Organic C data for subsurface horizons and yield data for a larger number of crops are two such needs. These data needs notwithstanding, STATSGO can be used to show generalized distributions of hydric soils, areas of potential and actual pesticide use, and areas where the risk of pesticide transmission to groundwater may be particularly high. Users are cautioned that small scale maps such as STATSGO do entail some loss of information and should not be used as substitutes for more detailed, site-specific information.
Risk is recognised as one of five pillars in the evaluation of sustainable land management, on a par with increasing productivity, natural resource conservation, economic viability, and social acceptability; however, most land resource assessments do not evaluate crop production risk, and they do not include estimates of spatial and temporal variability. This chapter describes some of the dimensions of risk, and explores the use of long-term yield records and crop growth simulation for estimating risk. Assessments of crop production risk must be added to land resource evaluations to estimate requirements for farm safety net programs, and also to evaluate sustainability in the future.
Four types of soil data, each with a characteristic reliability, are proposed for point observations and areas of land. Distinctions are based on procedures by which data are obtained: (i) measurement; (ii) estimation by experts; (iii) calculation by pedotransfer-functions, and (iv) same, by simulation models. Data are either static, requiring a single measurement, or dynamic, requiring measurements over a period of time. Three case studies illustrate: (i) interpolation of a static point measurement with multiple indicator kriging allowing an expression of risk of exceeding critical values; (ii) interpolation based on a time series of nitrate fluxes, calculated with a deterministic simulation model. Reliability was difficult to define because input data consisted of measured and estimated parameters with unknown reliability, and (iii) expression of data reliability in an uncertainty analysis by using rotated random scan and Monte Carlo techniques in the context of simulating water fluxes in a clay soil. Results reflect data reliability and use of this procedure is therefore recommended.
Environmental risk is defined using fuzzy logic and applied to a soil contamination problem. All the five elements of environmental risk analysis (exposure, resistance, failure event, consequence of failure, and perception) have uncertainties; however, these uncertainties may not always be expressed with probabilistic methodology. Fuzzy logic is an alternative procedure to characterize them. In this chapter the first three elements are considered. Specifically, both exposure and resistance are represented by fuzzy numbers, and consequence and perception are not considered. Failure is calculated for both series and parallel systems. The example considers soil contamination caused by failure of a landfill. The average areal level of exposure is estimated by fuzzy geostatistics and obtained as fuzzy numbers for three hazardous compounds. The allowable or threshold values of these compounds also are fuzzy numbers. Then a fuzzy reliability measure is calculated by considering the case as a series system. The fuzzy reliability measure can be used to compare alternative clean-up actions or to perform a full-scale risk analysis when the consequence and perception of failure are also considered.
Estimates of risk of degrading groundwater quality by agricultural pesticides are useful for selecting and managing pesticides. Leaching estimates incorporate soil properties, pesticide sorption, degradation, toxicity, application rate, water movement, and thus weather at the site of interest. Since future weather is not known, pesticide leaching cannot be predicted with certainty. By simulating leaching for many weather sequences typical of the site of interest, the probability of exceeding a critical concentration of pesticide in the groundwater can be estimated. In this chapter, we consider the probability of exceeding the U.S. Environmental Protection Agency (USEPA) health advisory level in the groundwater as the risk posed by that pesticide. We analyze the impact of soil variability within four map units upon the estimated risk to assess the uncertainty in that risk. Three estimators are considered: the risk for the map unit is equal to (i) the average of the risks for the different pedons within the unit, (ii) the maximum risk associated with any pedon in the unit, and (iii) the risk associated with the average concentration of the pesticide under the map unit. Results are presented for aldicarb, bromacil, diuron, fenamiphos, and simazine. The average risk for different pedons tends to be less than the risk associated with the average concentration. Maximum risks were often close to 1.0. Risk estimates for the major soil in the map unit often differed greatly from estimates based on all soils within that unit. An analysis of estimated risk and its uncertainty resulting from the use of different numbers of soil samples produced results that were highly dependent upon the risk estimator and the soil-chemical combination. Uncertainty is greatest for small sample sizes. The maximum risk and the risk associated with the average concentration are usually underestimated with small numbers of samples. One hundred or more samples may be needed to obtain 90% confidence of calculating the risk within 0.1 of the true risk. These results indicate that thematic maps of risk are highly dependent upon the estimator and the number of samples used. Such maps must be prepared and interpreted with care. The approach presented can be used to estimate risk and its uncertainty for management units that incorporate several map units.
The U.S. Environmental Protection Agency's (USEPA) Office of Pesticide Programs (OPP) is charged under FIFRA with registration or reregistration of all pesticides used in the USA. This requires characterizing the risk posed by pesticides to aquatic life. Risk to an aquatic organism is assessed by comparing the duration and concentration of pesticide exposure with the known toxicological effects of the chemical. For most chemicals the expected concentrations in the environment (EECs) and the duration of exposure must be estimated through use of environmental fate and transport computer modeling. Most models require single-valued inputs to represent soils, cropping, management and chemical parameters. These parameters, however, are often not known with a high degree of certainty or are known to be highly variable and are therefore difficult to represent in a modeling exercise. The OPP is developing a four tiered modeling system designed to represent the spatial and temporal variability in this very complex agro-eco system. Ability to represent the variability in the system will allow OPP to characterize the risk on a probabilistic basis. The system will combine modeling and field testing. Modeling will be employed to interpolate between field derived values while field results will serve as a reality check on the modeling. Higher tiered modeling, when required, will use an input-output file management device known as MUSCRAT (MUltiple SCenario Risk Assessment Tool), which will model the fate and transport of a chemical on multiple sites during the same run and compile cross-site statistics from the output.
Methods of nonparametric geostatistics are reviewed and applied to the quantification of uncertainty associated with the use of soil properties and interpretations. Correspondence between soil mapping unit properties and field measured values is enhanced by georeferencing field sampling locations. Results suggest that sequential indicator simulation procedure can be used to reproduce spatial patterns of continuity and to introduce a measure of uncertainty into soil survey data. The model of uncertainty is perceived as a potential for loss associated with a prediction of an unknown value. Multiple outcomes of conditionally simulated values quantify the model of uncertainty, while continuity of soil water flow properties in space is deduced from the quantile maps and the probability of occurrence maps.