Vegetation index as a parameter for identifying spatial variability zones in early stage selection t
Experimental trials may be assumed by breeders as being uniform such that significant differences observed between treatments are exclusively due to genetic effects. This is theoretically possible if spatial variability is effectively accounted for and partitioned in analyses via incomplete, complete blocking or by row-and-column designs. This may be especially important in the early stages of selection where large areas are involved. However, blocking does not always ensure control of spatial variation. Inappropriate blocking may even inflate the estimated experimental error variance, thus reducing heritability values. Spatial variability can be measured through either a priori or posteriori approach. A priori methods of quantifying spatial variability are based on identifying existing trends within a field and may then be accounted for using appropriate experimental designs. Vegetation indices, such as the normalized difference vegetation index (NDVI), obtained via reflectance measurements at the red and near-infrared (NIR) wavelengths, have been successfully used for a priori mapping field variability. A rotation crop (Crotalaria juncea L.) was considered for a priori mapping field variability for a Progeny Assessment Trial (PAT), which is the first selection stage of the Sugarcane Breeding Program of the Inter-University Network for the Sugar and Ethanol Development Sector (RIDESA) from the Federal University of Sao Carlos (UFSCar), Brazil. About 300 families were tested for cane yield in an area of 6.5 ha. The NDVI was calculated using the rotation crop spectral reflectance from an aerial platform using a multispectral camera. Following the imagery post-processing, if spatial variability was identified, correlated traits were considered and then used as covariates in the statistical genetic modeling, aiming an increase in the selection accuracy. The NDVI was able to subdivide the experimental area into three different zones of spatial variability varying from low to high yield potential. Adjusting the statistical genetic model, it was possible to identify different estimates of genetic values among the families, an expected result since the within-field variability could be handled efficiently. Therefore, an increase in the heritability of the trait under selection could be observed. The results suggest that vegetation index should be considered at early breeding trials, aiming to enhance the selection accuracy, one of the ways to accelerate the genetic gain within a breeding program. Even more promising results could be obtained using additional and complementary remotely sensed measurements such as soil electrical conductivity. Keywords: NDVI, Remote sensing, Site variability, Selection accuracy
