Discussion of Empirical Studies

Results from the warming experiment (Price and Waser 1998) suggested that warming will trigger an early flowering of D. nuttallianum by shifting the date of snowmelt; however, the evidence was only indirect since warming altered many other microclimate parameters besides snowmelt (Harte et al. 1995). This snow removal experiment corroborated their findings and provided direct evidence that a change in snowmelt date will trigger changes in flower phenology. Moreover, the results suggest that shifting snowmelt date by, on average, 10 days, as could occur under a 2 x CO2 climate scenario, might affect some aspects of the reproductive biology of D. nuttal-

F.arly Mid Early Mid

Flowering Cohort Flowering Cohort

Figures 5.9. Average seed weight per fruit versus flowering cohort, two populations. For the early population: Early, flowers that opened between June 6 and 8; Mid, flowers that opened between June 14 and 18. For the late population: Early, flowers that opened between June 23 and 25; Mid, flowers that opened between June 30 and July 4. Error bars represent one standard error. (a) Early population 1997. (b) Late population 1997.

F.arly Mid Early Mid

Flowering Cohort Flowering Cohort

Figures 5.9. Average seed weight per fruit versus flowering cohort, two populations. For the early population: Early, flowers that opened between June 6 and 8; Mid, flowers that opened between June 14 and 18. For the late population: Early, flowers that opened between June 23 and 25; Mid, flowers that opened between June 30 and July 4. Error bars represent one standard error. (a) Early population 1997. (b) Late population 1997.

lianum, such as seed weight and seed set, and not others, such as flower production. Furthermore the results suggest that responses are year and site specific. Finally, this result indicates that other factors (e.g., moisture, pollinators, air temperature, etc.) modulate the responses to snowmelt date.

Contrary to what was expected from Inouye and McGuire's long-term studies (1991), in 1997 flower number did not respond significantly to a difference in snowpack (T versus C) (Table 5.4). A more detailed analysis of Inouye and McGuire's data revealed that the number of flowers per plant did not vary with snow accumulation, but the number of flowering plants per plot did. So apparently plant density seems to respond positively to snowpack across years. Observations from the warming experiment support this idea. The density of flowering D. nuttallianum was significantly lower in the heated plots (earlier snowmelt and drier conditions) than the control plots (Saavedra 2000). However in neither of the two sites from the snow removal experiment (1997 data) could I detect a significant difference in flowering plant density between snow removal treatments (lower snowpack) and controls, and it is possible that more years might be needed to see a density effect.

Decreasing snowpack had a significant effect on seed set and seed weight, but the direction of the change depended upon both site and year. In the upper site, seed number per fruit showed no significant response to treatment (1996, 1997), yet the trend was different for different years. In the upper site (1997) snow removal had a small, but significant, positive effect on seed weight. Seeds from the snow removal plots weighed on average 16% more than seeds coming from the control plots (the effect on weight was not compensated by seed number, as both were increased in the treatment plots). In 1996 the trend was the same but the difference is not statistically significant. The increase in seed weight from snow removal in 1997 is similar to that observed by Galen and Stanton (1991, 1993) when they extended the growing season for Ranunculus adoneus by 12 days. Advancing snowmelt led to an average increase in seed mass of 33% in their experiment. Although the trend is the same as that found in my study, the magnitude of the change is different. The sample size I used for the seed weight analysis was large and allowed detection of even small statistical differences between the treatments. Are these differences in seed weight biologically significant? Growing the seeds from the 1996 cohort in a common garden showed that after a year seeds that weighed more (from the snow removal treatment) had significantly larger seedlings. If seedling size is correlated to survival, then one could infer that having a larger snow-free season could induce an increase in D. nuttal-lianum seedling survival in a high altitude site. However, other effects may be important. Seedling establishment is dependent on the biotic and abiotic environment. One way to test this hypothesis would be to follow the fate of the seedlings through many years in controls and snow removal plots, instead of in common gardens. In doing so, both plants and their environment will change simultaneously. Even though such a plot is only a small proportion of the landscape, at least it would allow estimates of the seedling's response in a matrix that is closer to that expected under natural conditions. Experiments conducted by Stanton and Galen (1997) and Stanton et al. (1997) show that the problem is more complex. Sites that flowered earlier produced bigger seeds that have an advantage in all sites in the snowbed. Further, a snow removal experiment showed that seedling establishment is not a simple function of snowmelt regime; it varies significantly between early, more fertile sites and later, less fertile sites (Galen and Stanton 1999).

In the middle site, in 1997, seed number per fruit was negatively and significantly affected by the snow removal treatment. Fruits from the snow removal plots had on average 32% fewer seeds than the controls (this could be compensation, since seeds from the snow removal treatment weighed 12% more than the controls). In 1996 and 1997, snow removal did not have a statistically significant effect on seed weight. In 1996, seeds from the snow removal plots weighed on average 8% less than the seeds coming from the controls, whereas in 1997 the trend was reversed (seeds from the snow removal treatment weighed 12% more than the controls).

The experimental approach in my study was complemented by observations of the phenology and the reproduction of two natural populations that naturally flower at different times (early and late snowmelt sites). The observations showed that in nature the phenology of different populations in a given area diverges in time more than the range produced by the experimental snow manipulation. A 10-day shift in snowmelt date is within the range of variation between close sites in an area, and certainly within the range of yearly variation (Barr, pers. comm. 2000. Moreover, if different flowering cohorts within each of the populations are considered analogous to the snow removal plot (early flowers) and control (mid/late flowers) from the snow removal experiment, one could reason that flowering earlier is advantageous. In both sites, early flowers produced more seeds and heavier seeds than middle flowers. This reasoning is contradictory to our findings in the middle site where flowering earlier (snow removal) had a negative and significant effect on seed number in 1997, but agrees with the results for the upper site, 1997, where snow removal had a positive effect on seed weight and number.

Although plant-pollinator interactions are in most cases not obligate (Waser et al. 1996), they could be negatively affected if global warming were to affect their synchrony. For instance, a lack of synchrony between hummingbirds and their host plant might not cause a total failure in sexual reproduction, yet it might decrease seed set (Waser 1979). The snow removal manipulation shifted the flowering time of D. nuttallianum but did not affect the time of emergence of bumblebees nor the arrival time of the hummingbirds (different spatial scales), and as such it offered the possibility of testing for hypothetical effect of changes in the synchrony between pollinator and their host plants.

Nesting patterns of broad-tailed hummingbirds (Selasphorus platycercus) at the RMBL are correlated with the temporal and spatial abundance of D. nuttallianum flowers (Waser 1976).The population size of broad-tailed hummingbirds seems to be correlated to floral abundance during the breeding season (Inouye et al. 1991) and nesting time parallels shifts in flower phenology between years (Waser 1976). Waser and Real (1979) reported that the years of lower density of D. nuttallianum were the years with lower activity of broad-tailed hummingbirds and lower fecundity of I. aggregata. If the distribution of D. nuttallianum were to be affected by climate change, we could expect changes in the abundance and fecundity of plants and pollinators that depend on it.

Indirectly, the fecundity of many plant species could be negatively affected if climate change affects the local and long-distance migratory pattern of hummingbirds. Rufous hummingbirds (Selaphorus rufus) breed in northwestern North America and arrive (on southward migration) after D. nuttallianum has flowered in the study area. Scarlet gilia (Ipomopsis aggregata) is the primary source of nectar for this hummingbird species (Calder 1987). There is evidence that the migratory pattern of rufous hummingbirds has changed in recent decades. Hill and collaborators (1998) reported that the number of wintering rufous hummingbirds has increased exponentially in the southeastern United States in the last 20 years. This evidence plus the capacity of hummingbirds to track changing resources (Bond 1995), suggests that if time or abundance of flower resources were to change with climate change, these hummingbirds might be able to shift their nesting sites and migration routes and move out of this habitat to forage elsewhere. Decreases in the local diversity of pollinators could probably have an impact on the resilience of the community (Aizen and Feinsinger 1994). It also could cause a decline of bird pollinators over the long term. For instance, Martin (1997) reported that a high elevation site in Arizona had in the past 5 years experienced both the wettest and the driest summers on record, and breeding birds at the site have moved up and down in elevation following the climate gradient. The changes in the preferred habitats have caused a significant decrease in their breeding success and will likely cause a population decline in the long term. Given that the local distribution of the birds along the altitudinal gradient is similar to their latitudinal distribution, Martin proposed that the reduction in nesting success might reflect larger shifts along latitudes related to climate change (Price 1997, Root 1997).

The reproductive responses of D. nuttallianum to snow manipulation are small and conflicting. One possibility is that snowmelt differences between controls and snow removal plots were only 4 to 8 days on average. Stronger treatment effects might be needed to simulate 2 x CO2 effects more accurately and to observe a response on seed parameters. Another possibility is that reproduction is not the most sensitive stage that may be affected by global warming. Observations in the warming experiment suggested that year to year flowering-plant emergence (density of flowering plants) could be more sensitive to warming than is reproduction (number of flowers per plant) (Saavedra 2000).

In light of the positive effect that a lower snowpack had on seed weight in the upper site (1997), or the negative effect on seed set in the middle site (1997), the long-term output is hard to predict. If timing of snowmelt were to change in a warmer climate, the effect might depend on the site as well as year, and on the correlated moisture patterns.These observations illustrate the importance of testing the impact of global warming at a larger temporal and spatial scale.

A warmer climate will likely induce a series of changes that will affect ecosystems through changes in the mean air and soil temperature, nutrient cycling, length of growing season, and phenology and relative fitness of species. In high latitude and altitude sites (like my study site) higher temperature should induce an early snowmelt and, as a consequence, induce a shift in plant phenology. The warming experiment (Harte et al. 1995) and the snow removal experiment (Dunne 2000) simulate only partially the effects of global warming. The warming experiment induces an early snowmelt and drier, warmer soil through a downward IR flux, and although it may be a reasonable mimic of some aspects global warming it has limitations (it doesn't warm the air over the unenclosed plots, it doesn't change CO2 concentrations, and it doesn't elicit potential covarying changes in precipitation; Price and Waser 1998).The snow removal experiment mimics the early snowmelt, which is a good simulation of the primary effect of warming expected in ecosystems with winter snowpack. But by changing the snow cover (snowmelt date), other variables were also changed (for instance decreases in water availability, drier soils in early summer), and other variables like total IR will be the same after the snowmelt for treatments and controls. These limitations and the small spatial time scale of the experiments limit our capacity to predict in the long term the direction of the changes, but they give us a starting point to explore possible consequences of climate change in alpine ecosystems.

0 0

Post a comment