Soil hydraulic properties: A simple and practical approach to estimate the number of samples

Abstract

There have been a number of studies dealing with soil hydraulic properties. Yet, there is apoor discussion on the number of samples necessary to represent such variables that usuallyvary orders of magnitude in space. In the present paper, we examine the adequate number ofsamples for two soil saturated hydraulic conductivity (Ksat) data sets: (1) normaldistribution (a 40 year-old pasture) and (2) non-normal distribution (primary forest). Toassess the adequate number of samples in each case, we used for normal distribution, anstatistical criterion of standard deviation lower than 5% compared to a high sampling effort(n = 25) as an indicative of a proper representation of Ksat variability. In the case of nonnormal distribution, we used the same criterion but using median absolute deviation (a nonparametric statistics). Both data sets were available in Salemi et al. (2013) and were Ksatmeasured at 0.15 m soil depth for medium-textured inceptisols in São Paulo State, Brazil. Foreach data set, we simulated 10 ‘new’ samplings in which we calculated mean and standarddeviation from sample 1 to 25 (for normal data) and median and median absolute deviation(for non-normal data). We found that, on average, at least 17 to 22 samples had to becollected to meet the adopted criterion for normal data whereas 20 to 25 had to be collectedfor non-normal data. Such numbers of samples exceed those used in a number of papers.Additional examples of this method with a light modification are given to establish number ofsamples in new study areas as well as to estimate number of samples when comparing two(or more) land-uses. Simple and practical procedures like those presented here couldestimate the number of samples that adequately represents soil hydraulic propertiesvariability.