Analysis of site-specific N-fertilization on-farm trials in cereals under assump-tion of spatial covariance structures

Abstract

The application of real-time sensors provides the possi- bility of adapting nitrogen fertilization to requirements in heterogeneous cereal crops site-specifically. Efficient and environmentally sound fertilization strategies can be de- veloped by the use of on-farm trials as a contribution to sensor-based farming. This paper assesses what agro- nomic effects can be found using the mechanical sensor "Crop-Meter" for nitrogen fertilization. On-farm trials were conducted to compare the application of uniform ni- trogen rates with sensor-based site-specifically applied rates. As an example, an approach is demonstrated taking the crop heterogeneity into account as a factor of interest in the statistical analysis, whereas heterogeneity is often treated as undesired noise factor. To include the hetero- geneity, yield data were classified by magnitude of the above-ground plant mass in order to create comparable conditions. Each class was analyzed by assuming mixed models taking into account spatial co-variance structures, which provided substantially better model fits than the classic ANOVA model. The trial considered showed that the reduction of nitrogen fertilizer in site-specific appli- cation (mean: 22 kg ha–1 N) caused no statistically sig- nificant yield differences by comparison with uniform application. Therefore, the spatial variation of nitrogen fertilizers seems a useful contribution to reducing direct cost of crop production and to decreasing environmental impacts.