Results

Number of Flowers (1997)

Average flower number per plant did not respond significantly to snow removal manipulation for either of the two sites in 1997 (p > 0.05) (Table 5.4), nor was a trend evident. In the upper site, plants from the snow removal treatment tended to produce more flowers

Table 5.4. Two-way ANOVA. Effect of Snow Removal (1997, 2920 m and 3170 m) on the Average Number of Flowers per Plant.

Variance

Table 5.4. Two-way ANOVA. Effect of Snow Removal (1997, 2920 m and 3170 m) on the Average Number of Flowers per Plant.

Variance

Source

Effect

LSdf

Ddf

Component

F

P

Treatment

Fixed

1

12.5

0.46

0.5087

Site

Fixed

1

12.5

1.87

0.1953

Treatment x Site

Fixed

1

12.5

1.80

0.2040

Block (Site)

Random

4

0.0

Treatment x Block (Site)

Random

4

0.037

Error

Random

297

0.328

Treatment was treated as a fixed effect and blocks within site as a random effect. LSdf = least squares degrees freedom, Ddf = denominator degree of freedom. Significant effects are shown in boldface type. Variance component values are reported for random effect and F ratio and probabilities are reported for fixed effect.

Table 5.5. Two-way ANOVA. Effect of Snow Removal on Seed Number per Fruit, Upper Site 1996, 1997.

Variance

Table 5.5. Two-way ANOVA. Effect of Snow Removal on Seed Number per Fruit, Upper Site 1996, 1997.

Variance

Source

Effect

LSdf

Ddf

Component

F

P

Treatment

Fixed

1

4

0.36

0.5789

Year

Fixed

1

105

1.40

0.2390

Block

Random

4

0.0289

Treatment x Year

Fixed

1

105

4.00

0.0482

Treatment x Block

Random

4

0.0

Error

Random

105

1.7325

Treatment and year were treated as a fixed effect and block as a random effect. LSdf = least squares degrees freedom, Ddf = denominator degree of freedom. Significant effects are shown in boldface type. Variance component values are reported for random effect and F ratio and probabilities are reported for fixed effect.

Figure 5.4. Average number of flowers per plant 1997, 2920 and 3170 sites. US, upper site; MS, middle site. Error bars represent 95% comparison limits between control and treatments in the same year-site combination.

than the controls, whereas in the middle site this trend was reversed (Fig. 5.4).These results contrast with observations in the warming experiment (average flower number per plant is lower in the heated plots) indicating that flower number is not dependent on snow date alone (Saavedra 2000).

Seed Number (1996, 1997) An ANOVA of seed number for the upper site (1996 and 1997) shows a significant interaction effect of treatment x year (Table 5.5). Neither of these effects was statistically significant by itself. Plants from the controls produced more seeds per fruit than those from the snow removal in 1996, but the trend is reversed in 1997 (Fig. 5.5). Analyses of seed number for 1997 in the upper and middle site show a significant effect of treatment x site (Table 5.6). In the upper site, plants from the snow removal treatment produced more seeds than the controls, whereas in the middle site this trend was reversed (Fig. 5.5).

In both 1996 and 1997, in the upper site, the average seed number for the snow removal treatments did not differ significantly from the controls (i-tests, p > 0.05). In 1997, however, the middle site s

96-US

Control Snow Removal

97-US

97-MS

Year - Site

Figure 5.5. Average seed number per fruit 1996 and 1997, 3170 m. US, upper site; MS, middle site. Error bars represent 95% comparison limits between control and treatments in the same year-site conditions.

Table 5.6. Two-way ANOVA. Effect of Snow Removal on Seed Number per Fruit, Upper Site and Middle Site in 1997.

Variance

Table 5.6. Two-way ANOVA. Effect of Snow Removal on Seed Number per Fruit, Upper Site and Middle Site in 1997.

Variance

Source

Effect

LSdf

Ddf

Component

F

P

Treatment

Fixed

1

25.2

0.18

0.6789

Site

Fixed

1

25.2

0.61

0.4434

Block (Site)

Random

4

0.0

Treatment x Site

Fixed

1

25.2

7.19

0.0127

Treatment x Block (Site)

Random

4

0.0110

Error

Random

69

1.4429

Treatment and site were treated as a fixed effect and block within site as a random effect. LSdf = least squares degrees freedom, Ddf = denominator degree of freedom. Significant effects are shown in boldface type. Variance component values are reported for random effect and F ratio and probabilities are reported for fixed effect.

showed a response. The controls in the middle site produced fruits with significantly more seeds than the snow-removal treatments (p = 0.0204) (Fig. 5.5).

In summary, the effect of snow removal treatment on seed set per fruit varied between sites and years. Only in 1997 and in one site could one infer a negative effect of an early snowmelt date on seed number. It is possible that there is a strong relationship between snowmelt date and seed number, but the treatment did not shift snowmelt date sufficiently to detect this relationship. A regression analysis might be more sensitive to the responses of plants to snowmelt date across sites and could be used in the future to complement the ANOVA.

Seed Weight (1996, 1997) Seed weight was used as an estimator of seed quality (Stanton 1984). There is a significant year effect and a treatment x year interaction. Most of the weight variation can be explained by plant to plant variation (86% of the variance comes from the error term) (Table 5.7). Overall, seeds from the snow removal experiment tend to weigh more than the controls, but the difference is not statistically significant. Seeds in 1997 weighed significantly less than seeds from 1996.

Within-year and site comparisons indicate that in 1996, at both

0 0

Post a comment