Mapping Soil Test Phosphorus and Potassium for Variable‐Rate Fertilizer Application

Abstract

Variable‐rate fertilizer (VRF) application requires knowledge of the spatial variability of soil test P and K within a field. The objectives of the study were to: (i) evaluate cell (area) vs. point soil sampling, both on a grid basis, and (ii) compare methods for mapping location specific soil test data. Soils in two central Wisconsin fields were sampled with two methods; study site soils included two alfisols and one entisol. The grid‐cell method involved dividing fields into 318‐ft square cells and compositing soil cores to give one sample per cell. The grid‐point method involved soil sampling at grid intersection points spaced on a 106‐ft square grid. Soil test P and K maps were constructed with nine mapping methods, including: field average, field median, CELL A (five soil cores), CELL B (72 soil cores), Delaunay triangulation, polynomial trend surface, inverse distance squared gridding, point kriging, and block kriging. Mapping accuracy was determined with map overlay comparisons where the Delaunay triangulation contour maps (106‐ft data) served as the control maps. On average, the CELL A and CELL B methods improved mapping accuracy by 14 and 33 percentage points over the field average. Even with the CELL B method, on average, 38% of the two fields would receive an incorrect application of fertilizer, indicating that cell mapping methods are not acceptable for making VRF management maps. Grid‐point sampling improved soil test P mapping accuracy by 20 percentage points over the CELL A method, even with point sampling in a 318‐ft grid. Mathematical procedures used to create a 53‐ft grid of data, from field sample data, did not improve map accuracy. Soil samples should be collected on a triangular or unaligned systematic grid. Sample spacing will depend on field variability, but probably should not exceed 300 ft. Research Question Traditionally, farm fields are soil sampled to determine the field average soil test P and K. One rate of fertilizer is recommended and applied, a rate that generally results in over‐ or under‐fertilization of portions of the field. Computer‐controlled fertilizer application equipment allows fertilizer rates or blends to be changed over short distances. Variable‐rate fertilizer (VRF) application requires accurate maps of the spatial variability of soil fertility levels. The objectives of this study were to: (i) evaluate cell (area) vs. point soil sampling, both on a grid basis, and (ii) compare methods for mapping location specific soil test data. The comparisons are limited to soil test P and K data. Literature Summary The profitability of VRF practices depends on the accuracy of soil test maps used to make fertilizer rate maps. Various techniques and sources of data are being used to create nutrient management maps, including soil surveys, infrared photographs, satellite imagery, and a variety of grid‐ and cell‐sampling schemes. Results from two studies show that sampling by soil type is not acceptable for mapping soil test levels. Another study reports that a 200‐ft soil sampling grid was adequate for developing soil test level maps, but that significant detail was lost at a 400‐ft soil sampling interval. Soil sampling areas larger than 99 by 99 ft were found to be inadequate for mapping soil test levels in 40‐acre fields. Methods for mapping point‐specific soil test data have not been evaluated. Study Description The spatial variability of soil test P and K was determined for two fields in central Wisconsin. Soil test P and K maps were made using 19 combinations of soil sampling and mapping methods. Delaunay contour maps based on the 106‐ft grid‐point P and K data were treated as control maps. All other soil test P and K maps were overlaid on the control maps to determine the accuracy of the mapping methods. Details of the study sites, soil sampling, and mapping methods are as follows: Soil Trinrud site—Zurich silt loam Kohel site—Rosholt sandy loam and Plainfield loamy sand Soil Sampling Grid‐cell sampling (318 by 318 ft) CELL A—composite of five soil cores collected on a diagonal CELL B—average of nine grid‐point soil samples in a cell Grid‐point sampling on a 106‐, 212‐, and 318‐ft grid Field average—average of 106‐ft sample point data Field median—median of 106‐ft sample point data Gridding and Contouring Methods Delaunay triangulation, inverse distance squared, polynomial trend surface, point kriging, and block, kriging Applied Question Should VRF application be based on soil test maps from grid‐cell soil sampling? Variable‐rate fertilizer application based on grid‐cell soil test maps can be expected to correctly fertilize, on average, about 60% of a field. This is only slightly better than a single rate fertilizer application based on the median value of multiple soil samples collected within a field. In this study, when only five soil cores were composited to represent a grid‐cell, VRF application correctly fertilized less than 50% of the field. Grid‐cell sampling is important for determining the field average or median soil test levels for single rate fertilizer applications, but should not be used for VRF programs. Did soil sampling at grid‐points and methods to enhance the point data improve the accuracy of soil test maps? Soil test P and K maps were more accurate when they were developed with grid‐point data than grid‐cell data. Grid‐point sampling on a 318‐ft grid resulted in a 20‐percentage point improvement in mapping accuracy for soil test P than with the CELL A (five soil cores per sample) method, with the same investment in soil sampling and testing. Mathematical methods that create a finer grid of data did not improve map accuracy. These methods are not a substitute for collecting more soil samples. Map accuracy decreased as sampling density decreased from the 106‐ to the 318‐ft grid. Based on the two study fields, sampling at a spacing greater than 106‐ft will result in a 30% or greater loss in mapping accuracy. Recommendations What sampling and mapping procedure should be used to develop soil test maps for application of VRF technologies? Soil test maps for VRF application technologies should be constructed from grid‐point soil sampling data. We recommend a two‐step approach to grid‐point sampling in order to avoid unnecessary sampling and testing costs. Fields that have shown soil test P and K levels in nonresponsive categories and have consistently received applications of nutrients meeting or exceeding crop removal, should be grid‐point sampled on a 300‐ft grid. If areas of optimum or lower soil test values are identified, additional samples should be collected to create a finer grid and then be combined with the initial data to create soil test or fertilizer management maps. Fields that have tested in responsive categories and that have received applications of nutrients meeting crop removal should be sampled on a grid no larger than 200 ft. Again, if optimum or lower soil test areas are identified, additional samples should be collected to create a finer grid or soil test data. Soil samples should probably be collected on a triangular or unaligned systematic grid.