ABSTRACT
In Chapter 1, I summarized the history and value of geostatistics in precision agriculture, and in Chapter 4 Marchant et al. considered the accuracy of geostatistical predictions for precision management decisions. Sampling plays an important role in the accuracy of such predictions, but this seldom receives sufficient attention. For precision farming surveys in the United Kingdom and many other countries, sampling is often at about one sample per hectare or even more sparse (Godwin and Miller, 2003) because of the costs involved in both obtaining and processing the samples. This approach to sampling, however, takes no account of either the spatial scale of variation or of how many sampling points might be needed for further analyses. As a result many of the maps created for precision agriculture are based on too few samples to provide accurate maps by interpolation, and are not suitable for site-specific management. Geostatistics can provide information about the scale of variation through the variogram and provide predictions at unsampled places by kriging.
