R Gazaffi1,2, IK Nunes1, GC Jacobassi1, RG Chapola1,2, MS Carneiro1,2 and HP Hoffmann1,2
1Sugarcane Breeding Program of RIDESA/UFSCar, Brazil; rgazaffi@ufscar.br
2Federal University of São Carlos, Araras, SP, Brazil
Sugarcane breeders are always looking for new variables to improve the efficiency of identifying superior genotypes. Multispectral sensors onboard unmanned aerial vehicles (UAVs) can provide a non-destructive and faster phenotyping strategy where the information can be summarized as vegetative indices. This work aimed to understand the association between vegetative indices obtained from a sugarcane breeding population. The trial comprised 148 genotypes planted in March 2023 under a randomized complete-block design, with two replications. A Phantom 4 multispectral drone was used for phenotyping monthly from September 2023 to February 2024. The Agisoft Metashape and FieldImageR were used to obtain the Orthomosaic and the red, blue, and green wavelengths and 19 vegetative indices (BGI, BI, CIG, CIV, CIRE, DVI, EVI, GLI, GNDVI, HUE, NDRE, NDVI, NGRDI, PSRI, RVI, SCI, SI, TVI, VARI). These 22 variables were submitted to a mixed model analysis to predict the variance components and the genotypic means. Cluster analysis was then applied to the adjusted means using the Euclidean distance and the UPGMA method. Three consistent groups were observed: i) 11 variables (CIG, CIRE, DVI, GLI, GNDVI, NDRE, NDVI, NGRDI, RVI, TVI, VARI); ii) 5 variables (BGI, HUE, PSRI, SCI, SI); iii) 4 variables (Red, Blue, Green wavelengths, and BI). The EVI and CVI did not represent genetic variance. Over the months, BGI changed from group ii to iii in two months (November and December). SI turns into a single group in October and January. The closest association was between VARI and NGRDI verified for all the analyses. Despite the large number of vegetative indices, breeders can select those with a high association.