Uncertainty assessment of spatial soil information

A - Papers appearing in refereed journals

Heuvelink, G. B. M and Webster, R. 2023. Uncertainty assessment of spatial soil information. Encyclopedia of soils in the environment. pp. 1-13. https://doi.org/1016/B978-0-12-822974-3.00174-9

AuthorsHeuvelink, G. B. M and Webster, R.

Uncertainty is present in our daily lives. It affects our decisions on what to do. The weather forecast might tell us that there is a 60% chance that it will rain: we take umbrellas. If it says that the chance of rain is only 10% we might decide to leave our umbrellas at home and risk getting wet. More seriously, farmers want to know the likelihood of disease in their crops and the deficiencies in plant
nutrients in the soil. These are matters that affect profit and loss of farm business. Agencies responsible for public health and environmental protection need to weigh the risk of doing nothing in the face of uncertain threats against the cost of acting unnecessarily to counter them when the threats are almost non-existent. There are many examples of decision making problems involving uncertain soil information. They include the remediation of polluted soil, the prevention of soil erosion, and the mitigation of pesticide leaching. They are practical matters, not purely academic exercises in statistics. All measurements of soil properties (and other environmental variables) contain error in the sense that they depart from the true values. That error arises from imperfections in the analytical instruments, from the people who use them and from errors that occur during the processing of the recorded data to make them suitable for storage in information databases. Short-range spatial variation
is another source of error, given that soil samples are never returned to where they were taken and sampling locations have
positional error. Soil taken from location s and analysed in the laboratory might differ substantially from the soil at location s + h, even if |h| is as small as a few decimeters. Composite soil sampling can diminish these differences, but some error inevitably persists because even such a composite is still only a sample of all the soil at that site. All this means that we can never be sure about the true state of the soil: we, the producers and users of soil information, are to some extent uncertain. Uncertainty tends to increase when measurements of basic soil properties are used to obtain derived ones via pedotransfer functions or mechanistic models of dynamic soil processes, for example. Interpolation from measurements to create maps of soil properties adds to the errors of measurement and so too increases uncertainties. We must conclude that considerable uncertainty is often associated with the information that is stored in soil databases and presented in various forms, including maps. This does not mean that the information is of no value; uncertainty is not the same as ignorance. In many cases we do know a great deal about the soil, but we must also acknowledge that the information is not perfect.
Some numerical expression of the uncertainty is important because it is needed to determine whether the information is
sufficiently accurate for the purpose that a user has in mind. Soil data of too poor a quality might lead to flawed decisions with serious undesirable consequences, both economic and environmental. For instance, the European legislation on the use of pesticides in agriculture depends crucially on the leaching potential of these substances to the ground- and surface-water, which in turn depends importantly on soil properties. In these circumstances users should be aware of the quality of the soil information so that they can be sure that it is sufficiently reliable for their purposes. Ideally they should account for the uncertainty of the information when making their decisions. This chapter (i) provides a statistical definition of uncertainty in soil information; (ii) extends this definition to uncertainty in spatial soil information; (iii) reviews methods that are used to quantify uncertainty in soil information, while paying attention to different sources of uncertainty; (iv) shows how uncertainty in soil information propagates through subsequent analyses; and (v) explains how uncertainty information can be used in decision making. It focuses on the quantification of uncertainty of soil properties that are measured and recorded on continuous scales: properties such as pH, particle-size distribution, and soil organic matter content. The chapter also addresses uncertainty of categorical variables, such as soil type and diagnostic properties recorded as present or absent, i.e. binary variables. It begins with defining uncertainty in a single soil measurement

Year of Publication2023
JournalEncyclopedia of soils in the environment
Journal citationpp. 1-13
Digital Object Identifier (DOI)https://doi.org/1016/B978-0-12-822974-3.00174-9
PubMed IDElsevier
Open accessPublished as non-open access
FunderBiotechnology and Biological Sciences Research Council
Funder project or codeS2N - Soil to Nutrition [ISPG]
Publisher's version
Output statusPublished
Publication dates
OnlineMar 2023

Permalink - https://repository.rothamsted.ac.uk/item/98vz7/uncertainty-assessment-of-spatial-soil-information

87 total views
34 total downloads
5 views this month
4 downloads this month
Download files as zip