KRIGING ON HIGHLY SKEWED DATA FOR DTPA-EXTRACTABLE SOIL ZN WITH AUXILIARY INFORMATION FOR PH AND ORGANIC CARBON

Authors: WU, J - CORNELL UNIVERSITY; Norvell, Wendell; Welch, Ross

Submitted to: Geoderma
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/1/2005
Publication Date: 1/4/2006
Citation: Wu, J., Norvell, W.A., Welch, R.M. 2006. Kriging on highly skewed data for dtpa-extractable soil zn with auxiliary information for ph and organic carbon. Geoderma. 134:187-199.

Interpretive Summary: Knowledge of distribution of crop-available trace elements in soils is limited by the sparseness of data from known locations and high variability. The lack of data can be overcome in part by relating the availability of a trace element to other soil properties that may be better known. Variability in the data can be managed with the use of mathematical transformations. We studied the distribution of the essential micronutrient zinc in northern North Dakota, a highly productive agricultural region. Zinc extracted from soil by the chelating agent DTPA [Zn(DTPA)]was used as a measure of zinc available to crops. Soil organic carbon and soil pH were used as auxiliary variables to improve predictions. Predictions using geostatistics were made for Zn(DTPA) with or without information for organic matter, pH or both. Four methods of transformation were used. The combination of variable choice and transformation methods yielded 16 different combinations to be tested. We found that information on soil organic carbon or pH always improved predictions. Transformation of the data also improved predictions, especially for soils with low concentrations of available zinc which farmers may wish to fertilize. The best predictions were obtained with a geostatistical method called multi-gaussian ordinary cokriging.

Technical Abstract: Knowledge of distribution of crop-available trace elements in soils is limited by the sparseness of geo-referenced data and the inherent variability of the more-labile forms of these elements. Cokriging with auxiliary variables can sometimes improve estimates for a less densely sampled primary variable, while some skewed or erratic data can often be made more suitable for geostatistical modeling by appropriate transformation. Geo-referenced data from northern North Dakota for Zn extracted from soil by the chelating agent DTPA (diethylenetriaminepentaacetic acid), i.e., Zn(DTPA), were used to assess the benefits of data transformation and cokriging for predicting this measure of available soil Zn at unsampled locations. Soil organic carbon and pH were used as auxiliary variables for cokriging. Data for Zn(DTPA), OC, and pH were available for 587 locations. The statistical distribution of the data for Zn(DTPA) was approximately log-normal, whereas the data for OC and pH were approximately normally distributed. Three transformations of the skewed data for Zn(DTPA) were compared: computing logarithms, standardized rank-ordering, and assignment of normal scores. Predictions for Zn(DTPA) were compared for ordinary (co)kriging, lognormal simple (co)kriging, rank-order (co)kriging, and multi-Gaussian (normal score) ordinary (co)kriging. For comparisons, the Zn(DTPA) data were partitioned into a training set of 293 sites and a testing set of 294 sites according to a stratified randomized approach. The Zn(DTPA) from the testing set were reserved for testing the predictions based on the training set. Co-kriging on Zn(DTPA) using OC or pH as auxiliary variables was consistently more effective than kriging on Zn(DTPA) alone. Co-kriging with OC and pH together provided additional improvement. Data transformation generally improved kriged estimates, especially for low concentrations of Zn(DTPA), < 0.5 mg kg-1, which are considered indicative of soils containing inadequate Zn for optimal crop growth. Multi-Gaussian ordinary cokriging provided slightly, but consistently, better estimates for Zn(DTPA) than did log-normal simple cokriging or rank-ordered ordinary cokriging.