Ecological Archives A018-009-A1

Jean H. Burns. 2008. Demographic performance predicts invasiveness of species in the Commelinaceae under high-nutrient conditions. Ecological Applications 18:335–346.

Appendix A. Matrix models for invasive and noninvasive congeners were parameterized with germination experiments in a growth chamber.

Appendix A. Matrix models for invasive and noninvasive congeners were parameterized with germination experiments in a growth chamber. Here, I describe the source for the seeds used in these experiments, including (1) the types of seeds used in these experiments, (2) the effect of maternal growing environment and seed position on seed size and germination, and (3) model parameterization.

(1) Types of seeds.

Commelina benghalensis produced different seed types in a single capsule. Thus matrix models were parameterized with separate germination and viability data for the different seed types where possible. First, seed types and their influence on seed properties, such as germination, are described. The information about seed types was used to make decisions about how to parameterize the models (i.e., how many seed classes to have). In addition, seed size variation and its influence on germination were examined, to determine whether different parameterizations were necessary across different environments (e.g., high and low water). Seed sizes differed by maternal environment and position in the seed capsule for some species, and a series of analyses were used to determine when seeds from different maternal environments or positions could be pooled for model parameterization. The results of these analyses are presented here. All analyses were conducted on SAS (SAS version 8.01, SAS Institute, Cary, North Carolina, USA) unless otherwise noted.

Commelina benghalensis produces two types of flowers (chasmogamous (CH) and cleistogamous (CL)). Each flower type produces a capsule with both dehiscent (D) and indehiscent (I) locules. Seeds in the dehiscent locule(s) are typically smaller than seeds in the indehiscent locule. Previous studies have shown that the four seed types (e.g., CH-D seeds, which are produced by chasmogamous flowers in the dehiscent portion of the capsule) have different dormancy properties (Walker and Evenson 1985).

Commelina bracteosa, Murdannia bracteata, and Murdannia nudiflora produce only CH-D seeds, which was confirmed by visual inspection of seeds at harvest and seed size distributions.

(2) Effects of maternal environment and position in the capsule on seed size and germination.

Effects of maternal environment on seed size were examined to help determine whether to pool seeds across maternal treatments. Commelina benghalensis produced a total of 945 seeds across all maternal environments, with a greater number of CH-D (629) and CH-I (208) seeds than CL-D (62) or CL-I (46) seeds (cleistogamy: F1,942 = 110, P < 0.001; dehiscence: F1,942 = 42, P < 0.001; Proc GLM, SAS). Dehiscent seeds were smaller on average than indehiscent seeds (F1,138 = 55, P < 0.001, Proc GLM). There was also a significant water × CH × D effect (F1, 138 = 11, P < 0.01; Proc GLM, SAS), a marginal water × nutrients × CH × D effect in the full model on seed size (F1, 138 = 2.35, P = 0.099; Proc GLM, SAS), and an effect of locule position on seed germination (Table A2), suggesting that seeds from different seed types and treatments could not be pooled for germination experiments.

TABLE A1. Two-way fixed effect ANOVA for effects of maternal environment on average seed mass (g). The maternal environment consisted of two levels of water and nutrient availability in a factorial design. Degrees of freedom are 1 for all sources of variation.

 

Commelina benghalensis

Murdannia nudiflora

Source

df

MS

F

df

MS

F

Water

1

0.78

7.20**

1

0.99

17.00***

Nutrients

1

0.02

0.15

1

0.07

1.20

W × N

1

0.06

0.59

1

0.01

0.10

Residual

46

0.11

 

56

0.06

 

m P < 0.10, * P < 0.05, ** P < 0.01, *** P <0.001

Maternal environment influenced seed size (Table A1), but did not significantly influence seed germination (Table A3). CH-D and CL-I seeds did not differ in size among treatments (F1,44/1,32 < 2.0, P > 0.20, PROC GLM, SAS) and were therefore pooled across treatments. CH-I seeds differed in size between water treatments (F1,45 = 4.5, P < 0.05, PROC GLM, SAS) and nutrient treatments (F1,45 = 8.3, P < 0.01, PROC GLM, SAS) and could not be pooled. CL-D seeds also differed in size between watering treatments (F1,22 = 32, P < 0.001; PROC GLM, SAS) and were not pooled across maternal growing environments.

Commelina benghalensis CH-D seeds were scarified, because scarification enhances germination (CH-D seeds: proportion germinating scarified = 0.90 (± 0.03 SE), unscarified = 0.40 (± 0.03 SE); one-way ANOVA: F1,8 = 110, P < 0.001). For CL-I seeds, two replicates of 20 scarified seeds were germinated. For the other two seed types of Commelina benghalensis, a single replicate of up to 20 scarified seeds (minimum = 7) was germinated for each treatment. A limited number of seeds for some growing treatments limited replication.

TABLE A2. Categorical model results for seed germination for Commelina benghalensis with maximum likelihood estimate of chi-squared.


Source

df

Chi-square

P value

Flower

1

0.39

0.531

Locule

1

3.95

0.047

Flower × locule

1

1.20

0.274

Intercept

1

41.78

<0.001

Commelina bracteosa produced only nine seeds total across all experimental treatments. In order to get an estimate of seed germination proportions, these seeds were supplemented with 39 seeds collected in the greenhouse under uniform (high nutrients and water) conditions from February 2003 through February 2005. Four replicates of 10 scarified seeds each were germinated under standard conditions. Since seeds of Commelina bracteosa are large with a thick seed coat, and since seeds for its congener, Commelina benghalensis, were scarified, all seeds of Commelina bracteosa were scarified.

Murdannia nudiflora produced 7,906 seeds total with from 338 (Low/Low) to 4,448 (High/High) per treatment. Preliminary germination experiments found high proportions of germination without scarification and no effect of chemical scarification (Burns, unpublished data), so germination experiments were conducted on unscarified material. Five replicates of 20 seeds each were germinated from seed from each of the four treatments.

Murdannia simplex has a very low rate of seed set in the greenhouse, and no seeds were available for this experiment.

Maternal environment did not affect seed germination (Table A3), but power was low for detecting such differences, and separate parameterizations were used for each matrix model for each of the four maternal environments wherever replication allowed.

TABLE A3. Categorical models were used to determine the affect of maternal growing environment on seed germination.

 

Commelina benghalensis

Murdannia nudiflora

 

Invasive

Invasive

Source

df

Chi-square

df

Chi-square

Water

2

1.84

1

2.32

Nutrients

1

2.49

1

1.75

W × N

1

0.32

1

0.02

Intercept

1

5.93*

1

9.63**

m P < 0.10, * P < 0.05, ** P < 0.01, *** P <0.001

(3) Model parameterization.

Placeholders (0 value entries) were used in matrices smaller than the largest matrix, to make matrices comparable in size. This has no effect on estimates of λ. Seeds of other species were placed in the same position as the chasmogamous, dehiscent seeds for Commelina benghalensis. Because all seed transitions were summed (e.g., Fig. 3d), results presented should not be dependent on the choice of placement.


[Back to A018-009]