A RELATIVE-FLUX-CORRECTION SCHEME FOR ANALYZING 3-DIMENSIONAL DATA OFA TILE-DRAINED AGRICULTURAL PLOT

The use of simple geostatistical tools is often constrained by data tr end (nonstationarity) to characterize the spatial variability of soil properties in the subsurface environment influenced by any site-specif ic feature(s). Adaptive approaches, such as site-specific robust-resis tant schemes, median polishing, trend analysis, etc., are thus used to preprocess the spatial data before analyzing for their spatial struct ures. Soil water nitrate-nitrogen (NO3-N) concentration (mg l(-1)) and soil moisture content (cm) data collected jointly from 175 sites arra nged on a 5 x 7 x 5 three-dimensional (3-D) grid network of 7.6 m x 7. 6 m x 0.3 m spacings in a tile-drained agricultural plot were analyzed for their three-dimensional spatial distribution and for possible cor egionalization. We propose a physical process-based correction scheme to preprocess the nonstationary spatial data of soil NO3-N concentrati on and soil moisture content. Using the subsurface-drain flow phenomen on, we developed a relative-Darcy-flux-based correction scheme to remo ve any tile drainage-induced nonstationarity in the spatial data of so il NO3-N concentration and soil moisture content prior to conducting t he spatial analysis in the 3-D soil volume. 3-D composite semivariogra ms of relative-flux-corrected NO3-N concentration and relative-flux-co rrected moisture content showed anisotropic linear structures in three principal directions. Linear models characterized by steep slopes wer e found in the directions perpendicular to tile line as opposed to nug get models found in the direction parallel to the tile line. Good spat ial correlation between the relative-flux-corrected NO3-N concentratio n and relative-flux-corrected soil moisture content and their anisotro pic linear semivariograms produced anisotropic linear cross semivariog rams in 3-D. The 3-D composite cross semivariogram will be useful in p redicting the more expensive variable, (relative-flux-corrected) soil water NO3-N concentration, at unsampled locations in the soil profile with a cheaper surrogate, the measured (relative-flux-corrected) soil moisture content. (C) 1997 Elsevier Science B.V.