Submitted to: Transactions of the ASAE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/1/2003
Publication Date: 5/1/2003
Citation: Han, Schufeng, Schneider, Sally M., Evans, Robert G. 2003. EVALUATING cokriging for imporving soil nutrient sampling efficiency. Transactions of the ASAE, Vol. 46(3):845-849. American Society of Agricultural Engineers ISSN 0001-2351. Interpretive Summary: New technologies allow farmers to apply different amounts of fertilizer to different parts of the field based on the expected need of the crop. Crop need is determined by analyzing soil samples from different parts of the field. More soil samples result in a more accurate map of fertilizer needs, but collecting and analyzing soil samples is expensive. This study found that using information on the soil texture (the amount of sand, silt, and clay in the soil) in different parts of the field along with information on fertilizer need reduced the number of soil samples needed to get the same accuracy for some soil nutrients. Soil texture does not change very much over the years. Fertilizer needs can change a great deal from year to year. This means that a large number of soil texture samples could be collected just one time and the information would be valid for many years. Samples for fertilizer need would have to be collected and analyzed more frequently, but not as many samples would be needed to get acceptable accuracy when used with the more extensive soil texture information. Fewer samples result in reduced cost, while maintaining the same accuracy. Accurate maps of fertilizer can then be used to apply enough fertilizer for good crop growth without excess. Excess fertlizer wastes money an can be source of environmental pollution.
Technical Abstract: The spatial variability of soil texture and soil nitrate - N, P, and K was studied in two center-pivot irrigated fields (89 ha total). Two soil texture components (clay and silt) were found to be correlated with soil nitrate - N, P, and K, and were used as auxiliary variables in the cokriging procedure to estimate soil nitrate - N, P, and K at unsampled locations. With a sampling density of 2.7 sites/ha (61 x 61 m grid) as the baseline, removal of 50% of the sampling sites resulted in a normalized mean absolute error (NMAE) of 34.5%, 22.9%, and 15.3% for soil nitrate - N, P, and K, respectively. These numbers reduced to 32.8%, 20.7%, and 12.0% when only 25% of the sampling sites were removed. The study showed that the cokriging technique provided slightly better estimates than the ordinary kriging method for soil P and K at a higher sampling density (>2.1 sites/ha). However, when a variable has a large random variation, such as the soil nitrate - N, cokriging did not provide better estimates than ordinary kriging. The results of this study provide guidelines on the selection of kriging or cokriging in improving the soil nutrient sampling efficiency.