Geostatistical analysis of soil moisture variability in grassland

This study sought to identify patterns of spatial variations in volumetric soil moisture content (θ) of a portion of the Tsukuba grasslands of Japan. The θ of the topsoil (0–0.15 m profile) of a square, 100 ha plot of grassland was sampled on a roughly rectangular grid, at the rate one sample per hectare. Spatial maps of θ were developed using ordinary kriging methods. The use of a kriging estimator in concert with a Gaussian semivariogram, the measurement error model, and exact interpolation within the grassland were deemed appropriate for the data collected. Estimated autocorrelation values begin at roughly 0.577 and gradually decreased towards zero when the distance between points reached roughly 750 m. Negative but non-significant correlations was noted for inter-sampling point distances ranging from 750 to 800 m. As the distance further increased, the autocorrelation value became alternatively positive and negative, suggesting that the soil and ecosystem system exhibit periodic variations through natural influences. Based on idealized spherical variograms, θ values tended to be dependent for inter-sampling point distances of less than 750 m, but independent for distances beyond 800 m.