Non-parametric variogram to analyze spatial genetic structure in field trials

BRUNO, C.; MACCHIAVELLI, R.; BALZARINI, M.

Stuttgart

Simposio; International Symposium Agricultural Field Trials -Today and Tomorrow; 2007

Species dispersal studies provide valuable information in crop improvement and management. The study of species distribution patterns in a fine scale provide usefull information in field experiments related to cultivar experimentation of genetically modified crops, and plague, pollen, or seed dispersal in field. The amount of dispersal in plant populations can be indirectly estimated from the observed spatial patterns. Non-random spatial patterns induce spatial variability, which results in autocorrelation among observations. Spatial autocorrelation is defined as the dependence of a variable with itself. Observations are autocorrelated if there is a systematic pattern in their spatial dispersion. Positive spatial autocorrelation (i.e. nearby observations tend to be more similar than distant ones) is assumed to result from any kind of spatial process. The spatial process is often described in terms of the maximum distance to which spatial structure extends. The study of spatial population genetic structure in a microgeographic scale (the degree of relatedness between individuals separated by distance) provides a better understanding of population structure and trial management. Spatial methods for spatial data analysis can be applied to multivariate genetic data to infer species population structure from DNA samples taken at random on the field experiment. Modern genetic techniques have increased significantly the number of marker alleles and loci containing genetic information, which can be used to describe spatial genetic patterns from genetic distances. Spatial structure is rarely consistent across loci or sites, and it is generally weak. Therefore, the analysis of each allele or locus separately may not be sensitive enough to describe weak spatial structures (Heywood, 1991). Molecular markers such as microsatellites, which are highly polymorphic, provide abundant information to improve the power to detect spatial genetic trends by means of spatial data analysis. Efficient statistical approaches are needed to detect autocorrelations when the spatial structure is weak as it is expected for genetic data observed at short distances in field.