Soil Spatial Variability as Influenced by Landscape Position and Soil Sampling Strategy

Understanding field nutrient variability may improve our ability to increase agronomic profitability and reduce the adverse effects of agriculture on the environment. Increased profitability can be attained by reducing fertilizer expenditures in areas that have been over-fertilized or by increasing yields in areas of the field that may have been underfertilized. A principle component in predicting fertilizer requirements is the assessment of the ability of the soil to supply nutrients. The objectives of this thesis were to evaluate the influence of landscape position on soil nutrient variability and to develop soil sampling protocols for precision farming. A 65 ha field located in eastern South Dakota was used for this study. Soil samples from 30 by 30 m grid were collected from the Oto 15-cm depth. Soil samples were analyzed for total N, o15 N, total C, o13 C, N03 -N, Olsen P, K, Zn, and pH. Semivariances, semivariograms, skewness, and kurtosis were calculated. The individual soil sample sites were characterized as belonging to toeslope, footslope, lower backslope, upper backslope, shoulder, and summit. The semivariances increased with distance to a maximum value and then deceased with further increasing in distance. The fact that the semivariogram for elevation had a similar shape to that of the individual nutrients showed that topography influenced the soil nutrient distributions. The soil nutrients generally had the lowest concentrations in convex areas and the highest concentrations in concave areas. Grain yields were lowest at toeslope and highest at lower backslope landscape positions. The influence of landscape position on nutrient concentration and grain yield was the direct result of drainage. The soil series followed the topographic changes. The grid soil sampling requirement was estimated by evaluating the influence of grid spacing on the semivariograms. Semivariograms based on 60 m and 90 m grid spacing were calculated for the different start points. There were 4 different start points for the 60 m grid and 9 different start points for the 90 m grid. At the 60 m grid spacing, total N, total C, and pH had similar semivariograms for the different start points, while the semivariograms for the different start points were not similar for Olsen P, K, and Zn. At the 90 m grid spacing, only soil pH had similar shapes of semivariograms for the 9 different start points. These results showed that the optimum grid distance was dependent on landscape position, with a 90 m grid being adequate for pH, a 60 m grid being adequate for total N, total C, and pH, and at least a 30 m grid being required for Olsen P, K, and Zn. Based on this analysis, over 600 samples would be required for Olsen P, K, and Zn recommendations in a 65 ha field. Many farmers would be unwilling to collect this number of samples, and therefore, an alternative sampling procedure that sampled the field according to landscape position was evaluated. A computer resampling procedure of the individual landscape positions indicated that the number of samples required were influenced by soil nutrient and that between 10 to 15 sampling sites were required from each landscape position in order for N03-N, Olsen P, K, and Zn estimates to be within 20% of the true value 80 % of the time. These results showed that less samples were required for sampling by landscape position than grid sampling. Nutrient information from both sampling procedures can be used for developing site specific fertilizer recommendations, which should reduce the impact of agriculture on the environment and increase profitability