New insights into adjustment for spatial dependence of soil attributes and use of aerial images in the initial selection stage of sugarcane families

One of the challenges of sugarcane breeding programs is the initial selection of genotypes aimed at developing varieties. Following the hybridization of genitors and family production, the process generally consists of selecting the best families and individuals within those families. Furthermore, the traditional selection model based on experimental designs assumes that the experimental field is well-prepared and homogeneous within blocks. If the allocation of blocks is incorrect, the ordering of families will be compromised, directly impacting the selection process and the program's success. Researchers are looking for quick, non-destructive alternatives to contribute to a less biased family selection process. These alternatives include modeling and statistical analysis or alternative data collection through images by unmanned aerial vehicles. This work proposes adjusting the ranking of sugarcane families by incorporating soil attributes in the statistical model and evaluating how some vegetation indices (VI) derived from the visible spectrum are associated with sugarcane yield (TSH). The experiment consisted of 60 families in a randomized complete block design with four replications. We also collected 36 soil samples and aerial images. The construction of the field map based on soil attributes and geostatistics indicated that the spatial position of the experimental blocks was incorrect. The correction implemented in this work allowed the ranking of families without the influence of the systematic variation in soil attributes. Additionally, the VI used showed a linear association with TSH, indicating the possibility of using aerial images to select or discard families in the initial stages of a breeding program.