Print version ISSN 0103-9016
MORAES, S.O.; LIBARDI, P.L.; REICHARDT, K. and BACCHI, O.O.S.. Heterogeneity of experimental points of soil-water retention curves. Sci. agric. (Piracicaba, Braz.) [online]. 1993, vol.50, n.3, pp. 393-403. ISSN 0103-9016. http://dx.doi.org/10.1590/S0103-90161993000300011.
In an area of "Terra Roxa Estruturada Latossólica" (Rhodic Kanhapludalf), in Piracicaba, SP, Brazil (20° 42' 30" S, 47° 38' 00" W, 576 m), 250 undisturbed soil samples were collected at 25 cm soil depth, according to a regular grid of spacing of 5 m, resulting a network of 25 Unes and 10 columns. These samples were used to determine 250 soil water retention curves each one with eigth experimental points, using Haines funnels (tensions of 5xl02, 1x103, 6x103 and 1x104 Pa) and Richards pressure chambers (pressures of 3x104, 8x104,3x105 and 1x106 Pa), totalizing two thousand values. Position measurements (mode, median and arithmetic mean), variability (total amplitude, interquartil amplitude, standard deviation, coefficient of variation, assimetry, kurtosis and confident limits around the mean) and number of samples to estimate the mean of the soil water content at a specific probability level, were used with the following objectives: a) to verify how close to the normal distribution are the values of soil water content for the different considered tensions and hence, to investigate what is the best position measurement; b) to quantify the variability in each considered tension, identifying the most problematic in the study of soil-water retention and also to analyse the measurement sensibility through the calculation of the necessary number of samples to estimate the mean (assuming a spatial independence of the samples). From the analysis of the obtained results, it could be concluded that the soil-water content values corresponding to tensions of 5x102 and 1x103 Pa showed very skewed distributions, so that care should be taken in using the arithmetic mean as a position measurement. The neglection of problem-samples, based on the physical analyses of results, allowed a better aproximation to the normal distributions for these tensions, indicating that a complete soil-water retention curve should always be prefered in the place of only two or three points of restricted interest as it is normally done. Soil-water contents corresponding to the several considered tensions showed low coefficients of variation (< 10 %) as compared to other soil physical measurements.
Keywords : soil-water retention curve; variability; descriptive statistics; Rhodic Kanhapludalf.