Influence of sampling strategy on GWAS confounding effects. (a) Relationship between an adaptive trait and the position along a transect across the species distribution. The phenotype could, for example, be flowering time in A. thaliana, and accession lines could have been sampled along a transect from the south to the north of the species' distribution. The relationship is positive because the phenotype is adaptive to an environmental variable varying along this transect. (b) Some traits show a gradual change along the transect. In the example of flowering time in A. thaliana, environmental factors such as temperature and photoperiod would show continuous change along the latitudinal clines. But (c) the phenotypes also show extensive variation at a given position along the transect, suggesting that other ecological factors, acting at smaller scales, might also be acting as selective pressures on the phenotype. These local environmental variations could be related to soil quality, exposition, competition or predation. They can differ between sites that are close to one another without following a trend across the entire species' distribution range. (d) The genetic structure of a species can be represented as the proportion of individuals assigned to each of three structure groups along the species-wide transect. (e) A global sample covers the entire species repartition range; alternatively, local sub-samples can be taken at locations chosen with reference to the pattern of the population structure and to small-scale environmental variations that have the potential to act as selective pressures. (f-i) Effect of the sampling scale (from local to species-wide sampling) on LD and confounding factors. Sampling at a local scale should reduce the effect of most confounding factors but might decrease mapping precision.