Spatial variability of water related soil physical properties.

Abstract

A study of soil variability was performed on an 87 hectare area within a uniform mapping unit--Pima Clay loam-- at The University of Arizona Branch Experiment Station at Marana. The primary interest was with respect to soilwater parameters. From 36 sites selected by an unbalanced three-stage nested design, 180 core samples were collected at 30 cm depth intervals to 150 cm. In addition, 500 bulk samples were taken at the 60 cm depth on an equally spaced grid over a secondary sampling area of 96 by 76 meters. The measured parameters showed different patterns of spatial variation. For example, to estimate means within 10% for the 30 cm depth 5, 51, and 1,011 samples would be needed for bulk density, the porosity index, and the saturated hydraulic conductivity, respectively (using the 0.05 level of significance). Most of the other estimates for number of samples required were in a range of 50 to 100. Most coefficients of variation were between 10 and 50% with bulk density lower and saturated hydraulic conductivity higher. Variance components for the three stages obtained from the analysis of variance revealed that the variation among fields was smaller than within fields and sections for the majority of the measured parameters, their average relative contribution to the total variance being 25, 44, and 31%, respectively. Values of 15-bar moisture retention corresponding to the 500 bulk samples showed a frequency distribution close to the normal with a slight tendency toward skewness. Values of bulk density were normally distributed at each depth and on the combined 180 samples. The highly skewed distributed values of the saturated hydraulic conductivity proved to be normally distributed after a logarithmic transformation. The porosity index showed a nonconsistent distribution pattern at the different depths and a moderately skewed frequency distribution for the composite 180 samples. Close relationships were found between bulk density and per cent sand and silt. A highly significant correlation (significant at th 0.01 level) between 15-bar water retention and clay content existed. Values of the logarithm of the hydraulic conductivity showed a high degree of correlation with values of per cent pores drained at 50 millibars (correlation coefficients of high absolute values and significant at the 0.01 level). Particle size distribution exhibited a decrease of silt and clay and a corresponding increase of sand with depth. The average percentages of sand, silt, and clay at 30 cm depth were 23.3, 41.2, and 35.3, and at the 150 cm depth were 39.7, 35.6, and 24.7, respectively. A similar trend was revealed for bulk density which ranged from 1.42 at 30 cm depth to 1.57 g/cm³ at the 150 cm depth. Soil moisture release curves for each depth showed similar general shapes. The "porosity index" describing the moisture release curve in the low pressure range varied from 3.58 at 30 cm depth to 5.79 at the 150 cm depth. Mean values of the saturated hydraulic conductivity also tended to increase with depth--1.71 and 7.03 cm/hr at 30 and 150 cm depth, respectively. Comparison between the sampling scheme used and three-stage balanced designs revealed that at least two alternatives would be more effective in decreasing the variance of the mean, but they do not provide any degrees of freedom for the third stage. An apparent compromise was found to exist between the scheme used and the optimum unbalanced designs selected for efficient estimation of variance components for the majority of the measured parameters.