Natural variability of soils makes it difficult to obtain values that adequately represent the soil properties within a certain area. The establishment of a minimum number of observations necessary to represent a soil property (with an acceptable error) is fundamental for the reliability of experimental results. In the present paper, the conventional t-test method has been compared with the bootstrap method, in order to estimate the number of observations needed to calculate the parameters that characterize the relation between hydraulic conductivity and soil water content, determined by the instantaneous profile method. A drainage experiment was realized in a red-yellow Oxisol (Hapludox) in Piracicaba, State of São Paulo, Brazil, on an experimental plot with 45 observation points in a straight line with a distance of one meter between points. Soil moisture was measured by TDR and the matric potential by tensiometers during three weeks of water redistribution. After data-processing, the parameters of the K(è) function presented a non-normal distribution and outliers had to be eliminated in order to normalize the data set for the t-test application. Therefore, the conventional method should only be used if the elimination of outliers is supposedly reasonable. Both analysis methods indicated the need of a high number of repetitions, reaffirming that hydraulic conductivity functions based on few repetitions should not be extrapolated to larger areas.
hydraulic conductivity; variability; bootstrap