Soil spatial variability: Areal interpolations of physical and chemical parameters.

Abstract

Four fields of 117 ha area located at the University of Arizona's Maricopa Agricultural Center were selected for this study. Two soil series, the Casa Grande sandy clay loam and Trix clay loam occur. Surface samples (0-25 cm) were collected on a 98 m interval and 3 rows providing 47 sites per field. Sites were classified either as surveying (32) or testing (15) in each of the four fields. Additional samples at 25-50, 50-75, 75-100, and 100-125 cm were obtained with duplicate surface undisturbed cores at 5 sites per field. Soil parameters include bulk density, saturated hydraulic conductivity, moisture retention, particle size analysis, pH, EC, soluble cations, SAR, and ESP. A quantification of the spatial interdependence of samples was developed based on the variogram of soil parameters. A linear model was best fitted to the clay, EC, Ca²⁺, Mg²⁺, Na⁺, SAR and ESP, and a spherical model to the sand, silt, pH, and K⁺ observed variograms. A comparison of variograms obtained conventionally and with the robust estimation of Cressie and Hawkins (1980) for sand and Ca²⁺ were performed with a fixed couples number per class and with a fixed class size. Additionally, a negative log-likelihood function along with cross-validation criteria were used with the jackknifing method to validate and determine variogram parameters. Three interpolation techniques have been compared for estimating 11 soil properties at the test sites. The techniques include Arithmetic Mean, Inversely Weighted Average, and Kriging with various numbers of neighbor estimates. Using 4 point estimates resulted in nearly identical results, but the 8 point estimates gave more contrast for results among the alternative techniques. Jackknifing was used with 4, 8, 15, 25 neighbors for estimating 188 points of sand and Ca²⁺ with the three techniques. Sand showed a definite advantage of Kriging by lowering the Mean Square Error with increasing neighbor number. The simple interpolator Arithmetic Mean was comparable and sometimes even better than the other techniques. Kriging, the most complex technique, was not the absolute best interpolator over all situations as perhaps expected. The spatial dependence for the 11 soil variables was studied by preparing contour maps by punctual Kriging. Sand and Ca²⁺ were also mapped by block Kriging estimates.