The mating system of self‐compatible plants may fluctuate between years in response to ecological factors that cause variation in the deposition of self pollen vs. outcross pollen on stigmas. Such temporal variation may have significant ecological and evolutionary consequences, but it has rarely been studied, and the mechanisms that mediate temporal variation have almost never been investigated. We tested for variation in the proportion of seeds self‐fertilized (s) between two years within 19 populations of the short‐lived herb Aquilegia canadensis. Selfing varied widely among populations (range in s = 0.17–1.00, mean s = 0.82) but was inconsistent across years, indicating significant temporal variation. Three populations exhibited especially wide swings in the mating system between years. Mean s did not decrease with increasing population size (N), nor was the fluctuation in s associated with mean N or the change in N. As expected, s declined with increasing separation between anthers and stigmas within flowers (herkogamy), and s fluctuated to a greater extent in populations with more herkogamous flowers. Self‐compatible plants can experience wide temporal variation in self‐fertilization, and floral traits such as herkogamy may mediate temporal variation by forestalling self‐pollination and thus allowing outcrossing during periods when pollinators are frequent.
During May and June of 1995, 1996, 1998, and 1999, we sampled 19 natural populations of A. canadensis from the Thousand Islands of the Saint Lawrence River and the adjacent mainland in Leeds and Greenville and Frontenac Counties in eastern Ontario, Canada. Eight populations were located on islands in the Admiralty Archipelago, and the rest were on the mainland. Together, these populations spanned the range of typical population sizes and densities found in this species; from small scattered clusters of <100 plants to large dense patches of >1500 individuals (Appendix B). A population was defined as a discrete group of plants, usually confined to a rock outcrop, separated from other such groups by at least 100 m and usually much more. For each population in each year of study, we counted the number of flowering individuals at peak flowering as a measure of population size (N), and collected one mature undehisced fruit from at least 30 randomly chosen maternal plants per population. Fruits were air dried at room temperature and then stored at 5°C.
Ten of the study populations were surveyed in 1995 and then again in 1998 (three-year interval). Eight were surveyed in 1998 and again in 1999 (one-year interval), and the remaining population was surveyed in 1996 and 1998. To get an idea of how quickly individuals turn over in populations, we marked 147 reproductive individuals in a large population adjacent to QWM1 in 1998. The following year, 51.7% were still alive and 50.7% of the survivors were reproductive again. This suggests that after one year only ~25% of plants from the previous year would still be reproductive, and turnover would be nearly complete after three years. There was no difference between populations sampled one year apart and those sampled three years apart in terms of the absolute change in population size (t test, P = 0.28), the proportion of seeds produced via selfing (P = 0.84).
For each of 18 populations (all except QRR1), we measured herkogamy (i.e., the minimum distance between a receptive stigma and a dehiscing anther) on a single randomly chosen flower, using calipers (to 0.1 mm) for each of a sample of 17–144 randomly chosen plants (mean = 49) during one reproductive season. Previous analyses have shown that measures of herkogamy on A. canadensis are highly repeatable if herkogamy is measured at a standard stage of flower development, as it was here (Griffin et al. 2000, Herlihy and Eckert 2007). Owing to logistical constraints, floral measurements involved several observers and, in eight of the 18 populations, floral measurements and mating system parameters were estimated in different years (Appendix B). However, a comparison of herkogamy measured in two different years in 14 of these populations failed to detect a significant difference in herkogamy between years for nine populations, suggesting that herkogamy is usually fairly consistent between years. Year-to-year variation in herkogamy within some populations, perhaps caused by variation in environmental factors that affect plant growth, may weaken the correlation between population mean herkogamy and estimates of mating system parameters and between-year fluctuation in these parameters. Hence, our test of the predictions relating to herkogamy should be viewed as conservative. However, it is unlikely that the expression of herkogamy is greatly influenced by environmental factors in the populations studied here. First, Herlihy and Eckert (2007) showed that the herkogamy of maternal plants, measured in the field, correlated significantly with that of their offspring, which were measured in the greenhouse. Second, experimental manipulation of maternal resources by complete defoliation significantly affected some floral traits and seed production but did not affect herkogamy (Kliber and Eckert 2004). Finally, herkogamy did not covary with habitat variables like canopy cover or low vegetation cover that consistently affected plant size (Herlihy and Eckert 2004).