Sampling procedure simulating on-the-go sensing for soil nutrients

The emergence of a new sensor technology based on the use of ion-selective membranes provides an increasing number of opportunities for on-the-go field measurements of soil nutrients and soil pH. In the future, on-the-go sensing should provide a cost-effective monitoring of heterogeneous soils with high sampling resolution. It is suitable for site-specific management because it can be focused on the spatial representativity of observation. This study evaluates the on-the-go-sensing sampling design by comparing it with a standard approach to soil sampling for soil pH and the base nutrients P, K, and Mg under local field conditions in Germany. Soil samples were taken in two test sites at a resolution and in a manner as if they were sampled with an on-the-go sensing system and were compared with soil samples taken at a coarser resolution and with standard methods. In general, a higher variability was observed among the on-the-go samples due to their smaller sample support. The finer sampling resolution of the on-the-go design improved field-scale semivariogram-analysis results, identifying the spatial structures for soil pH, P, and Mg clearly. In addition, kriged maps of these soil parameters had predominantly higher estimation accuracies. However, the on-the-go samples were strongly influenced by the small-scale variability of K in one of the test sites. This variability increased the kriging standard deviation for K by 50% compared with standard sampling design. Despite of this problem, the on-the-go-sensing sampling design revealed field-scale spatial variability for base nutrient status more accurately. Except for K, the mean absolute error of fertilizer-application maps was reduced when using the on-the-go sample design in comparison with the standard sample design (Ca: 210/268 kg ha(-1), P: 2.85/6.75 kg ha(-1), K: 13.7/6.0 kg ha(-1), Mg 5.7/6.8 kg ha(-1)). This will reduce over-and underfertilization using variable-rate fertilizer-application systems. In the future, it will be of interest if real on-the-go soil-sensor measurements exhibit the same variability behavior addressed here or if results will differ substantially.