Genotype-by-environment interactions for fitness of Erigeron annuus show fine-scale selective heterogeneity

The results of natural selection depend critically on whether variation in fitness is fine-grained or coarse-grained with respect to dispersal, but little is known of the spatial scale of fitness variation in natural populations. For most evolutionary questions, environmental heterogeneity must be defined by reversals in the relative fitness of genotypes; absolute fitness may vary, but if genotypes respond in parallel then selection is uniform. Thus, measurements of genotype-by-environment (G x E) interactions for fitness are necessary to understand patterns of variation in natural selection. Survivorship and fecundity of three genotypes of Erigeron annuus were measured at 630 locations within a 0.5-ha oldfield. The nested design allowed measurements of variation in lifetime fitness over scales spanning three orders of magnitude (10 cm to 100 m). G x E interactions for fitness were large relative to variation in mean fitness among genotypes. Almost all of the G x E interactions occurred at the smallest spatial scale, showing reversals in the relative fitness of genotypes on a scale of only 10 cm. A weak G x E interaction occurred among sections of the field (25+ m), but the statistical interaction was caused primarily by scaling effects, not changes in rank. Spatialautocorrelation analysis showed that relative fitness was spatially unpredictable at all scales greater than 10 cm. Seed-dispersal distances provide a natural scale for interpreting the effects of selective heterogeneity. Spatial autocorrelations of relative fitness, using seed-dispersal probabilities as the weighting function, were not significantly different from zero. Thus, progeny from a given plant are likely to disperse to microsites where their relative fitness is unpredictable. In contrast to the small-scale heterogeneity of relative fitness, absolute fitness was positively autocorrelated at distances up to 25 m, showing larger-scale variation in general habitat quality. Soil nutrients and the percent cover of surrounding species were measured to identify abiotic and biotic causes of selective heterogeneity. All showed patchiness on a scale of 5 m or more. Genotypes were sensitive to different sets of environmental factors, but no set of external measurements of the environment explained more than 12% of the G x E variance. Thus, the "ecological environment" varies on a scale of meters and affects the absolute fitness of Erigeron, but most variation in relative fitness occurs on a scale