View article

The communication of uncertainty is not only a challenge when soil information has been produced but also in the planning stage. When planning a survey of soil properties it is necessary to make decisions about the sampling density. Sampling density determines both the quality of predictions and the cost of fieldwork. In this study, we considered four ways in which the relationship between sample density and the uncertainty of predictions can be related, based on prior information about the variability of the target quantity. These were offset correlation, prediction intervals, conditional probabilities of the interpretation errors and implicit loss functions. Offset correlation is a measure of the consistency of kriging predictions made from sample grids with the same spacing but different origins. Prediction intervals and conditional probabilities are based on the prediction distribution of the variable of interest. All four of these methods were investigated using the information on soil pH and Se concentration in grain in Malawi. They were presented to a group of stakeholders, who were asked to use them in turn to select a sampling density. Their responses were evaluated and they were then asked to rank the methods based on their effectiveness, in their experience, and in terms of finding a level of uncertainty that they were able to tolerate when deciding about a sampling grid spacing. Our results show that the approach that stakeholders favoured was offset correlation, and some approaches were not well understood (conditional probability and implicit loss function). During feedback sessions, the stakeholders highlighted that they were more familiar with …