Sampling and Geostatistics for Precision Agriculture

Keywords: variogram, kriging, site-specific management, ancillary data, standardised variogram, management zones.

Precise or site-specific management of soil and crop properties depends on knowing how they vary from place to place. This often depends on obtaining information by sampling. The data from the samples are then used to express the spatial variation usually in the form of digital maps that can either be displayed as such or used by computer programs to manage irrigation or fertiliser applications, for example. Soil and crop sampling and analysis are expensive and consequently this often results in too few samples to describe the variation accurately. The aim of this paper is to illustrate the importance of sampling, the limitations of sampling that is too sparse and to suggest how sparse data can be used effectively. The variogram of geostatistics can be used in several ways to determine a suitable sampling interval. When there is no prior information on the scale of variation and the variable is unlikely to be strongly correlated with available ancillary data, a nested survey and analysis can provide a first approximation to the variogram to describe the approximate spatial scale. The spatial variation in ancillary data, such as aerial photographs, electrical conductivity or elevation, is often related to soil and crop properties. If so, variograms of ancillary data can indicate the likely scale of variation in the soil or crop. Existing variograms of soil or crop properties can be used to determine an optimal sampling interval based on the kriging prediction errors, or an interval of less than half the variogram range can be used to ensure a spatially dependent sample. If there are too few data to compute the variogram in the conventional way it could be estimated by residual maximum likelihood (REML), or standardised variograms from ancillary data can be used to krige soil or crop data from a small, but spatially dependent sample. Finally, management zones might be created from ancillary data and sampling targeted to the zones to provide estimates of the soil or crop variables within them.

M A Oliver, Soil Research Centre, Department of Geography and Environmental Science, The University of Reading, Whiteknights, Reading RG6 6DW, UK.

R Kerry, Department of Geography, Brigham Young University, 690 SWKT, Provo, UT, 84602, USA.

32 pages, 13 figures, 7 tables, 31 references.