Sequential sampling to measure the infiltration rate within relatively homogeneous soil units

The statistical prediction techniques Trend Surface Analysis, Kriging and Co-kriging are regularly used to provide predictive single-value soil maps. Kriging and Co-kriging perform well in situations where regionally distributed variables show clear spatial structure. If the amount of variation of the variables under study is relatively small, however, no gain in precision is gained by Kriging and Co-kriging, as compared with simply averaging over an area or by applying Trend Surface Analysis. This study was carried out on an older terrace of the Allier river in Central France where attention was focused on the infiltration rate (Inf). Kriging log(Inf) values did barely improve the predictions as compared with Trend Surface Analysis, whereas deleting part of the observations did not result in serious changes of predictions. A satisfactorial first approach was obtained by regarding the observations as being independent. To determine the sample size which provides sufficient information with respect to the soil characteristic under study use can be made of the sequential t-test as is illustrated. Application of this test resulted in eight measurements which represented a saving of 70% as compared with the original standard scheme.