Many soil surveys lead to important decision making in agriculture, such as delineation of areas targeted for fertilization or some remedial treatment. Such decisions are often based on critical values of the concentrations of nutrient or salt in the soil. If the estimates are less or more than specified thresholds, farmers are advised to act. But such estimates are usually affected by large uncertainty, arising from sampling, modelling and interpolation, which must be quantified to allow an evaluation of the risk involved in any decision. Geostatistics allows to assess such uncertainty through the determination of a conditional cumulative distribution function (ccdf) of the unknown attribute value. This paper considers the problem of modelling uncertainty about the value of an attribute at any unvisited location. The uncertainty is modelled through the ccdf conditional to the local information and gives the probability that the unknown is not greater than any given threshold. The paper describes a non-parametric approach to estimate the uncertainty, called "indicator kriging" (Journel, 1983), based on the interpretation of the conditional probability as the conditional expectation of an indicator random variable. A soil survey data set of a 18000 ha-area in southern Italy was used as a support for presenting a potential application of modern Geostatistics to agricultural management decision making. Accounting of "soft" information as provided by a geological and soil maps is shown to reduce the uncertainty.
Accounting for local uncertainty in agricultural management decision making
Buttafuoco Gabriele;
2000
Abstract
Many soil surveys lead to important decision making in agriculture, such as delineation of areas targeted for fertilization or some remedial treatment. Such decisions are often based on critical values of the concentrations of nutrient or salt in the soil. If the estimates are less or more than specified thresholds, farmers are advised to act. But such estimates are usually affected by large uncertainty, arising from sampling, modelling and interpolation, which must be quantified to allow an evaluation of the risk involved in any decision. Geostatistics allows to assess such uncertainty through the determination of a conditional cumulative distribution function (ccdf) of the unknown attribute value. This paper considers the problem of modelling uncertainty about the value of an attribute at any unvisited location. The uncertainty is modelled through the ccdf conditional to the local information and gives the probability that the unknown is not greater than any given threshold. The paper describes a non-parametric approach to estimate the uncertainty, called "indicator kriging" (Journel, 1983), based on the interpretation of the conditional probability as the conditional expectation of an indicator random variable. A soil survey data set of a 18000 ha-area in southern Italy was used as a support for presenting a potential application of modern Geostatistics to agricultural management decision making. Accounting of "soft" information as provided by a geological and soil maps is shown to reduce the uncertainty.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.