Novel Tools for Adjusting Spatial Variability in the Early Sugarcane Breeding Stage

Technologies that allow for the identification of spatial variability in field trials have great potential to accelerate the progress of plant breeding. Therefore, we aimed to evaluate a digital soil mapping approach and a high-density soil sampling procedure as tools for identifying and adjusting spatial dependence in the early sugarcane breeding stage. Two experiments were conducted in regions with different soil classifications. To investigate spatial variability structure, a high-density soil sampling procedure was performed in a regular grid for physical and chemical properties. Soil apparent electrical conductivity (ECa) was measured in both experimental areas with an EM38-MK2® sensor. In addition, principal component analysis (PCA) was employed to reduce the dimensionality of the physical and chemical soil data set. The spatial dependency index for each principal component (PC) was determined based on the semivariogram coefficients. After conducting the PCA and obtaining the different thematic maps, we determined the exact position and location of each experimental plot within the field. For each experiment, tons of cane per hectare (TCH) data were obtained and analyzed using different mixed linear models, i.e., independent (ID) and separable first-order autoregressive (AR1 x AR1) models. For both models, forward selection methodology was used to incorporate information from the PCA and the ECa sensor. The PCA based on high-density soil sampling data captured part of the total variability in the data for Experimental Area 1 and proved to be an efficient index to be incorporated as a covariate in the statistical model, reducing the experimental error (CVe). When incorporated into the different statistical models, the ECa information increased the accuracy of selection of the experimental clones, with an emphasis on the AR1 x AR1 model. Therefore, this type of geotechnology can be widely used in various plant breeding study areas, e.g., experimentation phases, genomic selection and genotype x environment interaction studies.