Summary Spatial variations of soil moisture content, textural components and yield components of two rainfed crops (vetch-oat and durum wheat) are studied by using classical statistical and geostatistical tools. The measurements were made at 81 nodes of a15×10m grid superimposed on a 1.21 ha plot selected in a 13.5 ha-field. The variability of each variable is characterized by a coefficient of variation ranging from 11% to more than 100% and a range varying from a distance smaller than the sampling interval to about 80 m. The concepts of dispersion and estimation variances are used to estimate the accuracy of the calculated mean values and to derive the relation between the dimensions of the sampled plot and the number of measurements needed to estimate average values with a given prespecified error. The results show that the total relative errors affecting the mean values calculated from 81 observations are in the range between 3 and 26% according to the variables, the major contribution arising from the finite character of the sampled plot. They also suggest that for the case where variables have spatial structure, decreasing the error on mean values cannot be achieved without increasing the size of the sampling surface.
Read full abstract