|Wu, Jay - CORNELL UNIVERSITY|
|Hopkins, D - NORTH DAKOTA STATE UNIV|
|Smith, D - U.S. GEOLOGICAL SURVEY|
|Ulmer, M - USDA-NRCS|
Submitted to: Soil Science Society of America Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: September 19, 2002
Publication Date: May 1, 2003
Citation: WU, J., NORVELL, W.A., HOPKINS, D.G., SMITH, D.B., ULMER, M.G., WELCH, R.M. IMPROVED PREDICTION AND MAPPING OF SOIL COPPER BY KRIGING WITH AUXILIARY DATA FOR CATION-EXCHANGE CAPACITY. SOIL SCIENCE SOCIETY OF AMERICA JOURNAL. 2003. v. 67. p. 919-927. Interpretive Summary: Knowledge of the distribution of copper and other trace elements in soils is important for agricultural and environmental reasons. Copper is one of the essential elements, required in small amounts by both plants and animals. Copper may also have a beneficial effect on some crops when present at levels somewhat higher than those meeting minimal plant requirements, for example, by enhancing resistance to disease or other stresses. On the other hand, elevated levels of soil Cu are highly toxic to plant roots, and pollution of soils with Cu-rich wastes can be a serious environmental problem. Unfortunately, actual measurements of copper or other trace elements in soils are rarely available in sufficient abundance or over large enough areas to permit the creation of accurate maps at the scale of states, much less at a scale suited for use at the level of county, township, or farm operations. In contrast to the sparsity of information about trace elements, information is more widely available for major soil characteristics, such as cation exchange capacity, organic matter, pH and texture. We tested several methods of geostatistics which allow the use of information about a secondary characteristic to improve the mapping of a variable of primary interest. In particular, we used information about cation exchange capacity to improve the accuracy of maps of copper in soils of northern North Dakota. Quantitative comparisons of predictions at known sites showed that inclusion of cation exchange capacity data improved accuracy of soil copper predictions by about 40%. The best of the methods tested, standardized cokriging, was used to create a map of soil copper for 18 counties in northern North Dakota.
Technical Abstract: Knowledge of the distribution of copper and other trace elements in soils is important for agricultural & environmental reasons. Measurements of trace elements in soils are rarely available in sufficient abundance or over large enough areas to permit mapping at the scale of counties, townships, or farm operations. In contrast, information is more widely available for major soil characteristics, such as CEC, organic matter, pH texture. We tested three geostatistical methods which use information about a secondary characteristic to improve the mapping of a variable of primary interest. Standardized cokriging, traditional cokriging, & ordinary kriging combined with regression were compared to ordinary kriging alone as predictors of soil Cu. Soil CEC was chosen as the secondary variable. Data available from several sources for total Cu or CEC in soils of northern North Dakota were used to test predictions by the four kriging methods and to prepare maps. Quantitative predictions of soil Cu were tested by partitioning 619 sites with complete data into a training set, having data for Cu and CEC, and a testing set which used only CEC to predict Cu, but reserved the measured Cu values for testing predictions. All methods utilizing CEC data improved predictions substantially in comparison to ordinary kriging which used only data for soil Cu. Standardized cokriging provided the most successful predictions, reducing the mean absolute error by about 40% in comparison to ordinary kriging which did not benefit from information about CEC. Maps for soil Cu, prepared from all available data for 1062 sites, were quite similar for the three kriging methods which used data for both Cu and CEC, but substantially different from the map derived from ordinary kriging which used data for Cu only.