Ecological Archives E091-258-A3

Margaret R. Metz, Wayne P. Sousa, and Renato Valencia. 2010. Widespread density-dependent seedling mortality promotes species coexistence in a highly diverse Amazonian rain forest. Ecology 91:3675–3685.

Appendix C. Results tables for GLMM and GEE models of seedling survival.

TABLE C1. Effect of seedling and adult neighborhoods (of varying sizes) on individual seedling survival over the first year (GEE). Seedling neighbors are found within the same 1-m2 seedling plot as each focal seedling, and tree neighborhoods contain adults > 10 cm DBH in circles with the indicated radii surrounding each plot. See Methods for a description of the regression model. Coefficient estimates are expressed as odds ratios for the fixed-factor predictors in a generalized estimating equation model. Significant predictor variables are highlighted in bold. Odds ratios >1 have a positive relationship with increased survival, and ratios <1 have a negative relationship with survival. An asterisk (*) indicates the predictor had a P value <0.1, and is included for heuristic purposes to show changes in the strength of the signal with changes in the size of the adult neighborhood included in the model. Model ROC values > 0.7 indicate acceptable discrimination ability.

5-m radius

10-m radius

15-m radius

20-m radius

Conspecific adult basal area

0.001

0.030

* 0.096

* 0.252

Heterospecific adult basal area

0.715

0.844

1.142

0.950

Phylogenetic distance (adults) 1

0.909

0.945

0.950

0.958

Conspecific seedling density

0.979

0.978

0.973

0.975

Heterospecific seedling density

1.001

1.000

1.000

1.000

Phylogenetic distance (seedlings) 1

0.982

0.972

0.968

0.965

Seedling species diversity

1.005

1.011

0.973

0.999

Initial seedling height

1.032

1.032

1.048

1.007

Height × Conspecific adult basal area

1.300

* 1.122

1.015

1.020

Height × Heterospecific adult basal area

1.037

1.010

0.995

1.010

Height × Conspecific seedling density

1.002

1.002

1.002

1.002

Height × Heterospecific seedling density

1.000

1.000

1.000

1.000

Height × Seedling diversity

1.015

1.014

1.017

1.016

Recruit Year = 2004

0.385

0.379

0.382

0.379

Recruit Year = 2005

0.284

0.284

0.285

0.288

Recruit Year = 2006

0.251

0.248

0.243

0.244

Recruit Year = 2007

0.242

0.240

0.238

0.236

ROC

0.684

0.679

0.685

0.679

1 Phylogenetic distance measured as an effect size with values >0 indicating phylogenetic clustering (neighborhood is more closely related than expected by null model) and values <0 indicating phylogenetic dispersion (neighborhood is more distantly related than expected from null model).  Positive effects (odds ratios >1) indicate better survival with increased phylogenetic clustering; negative effects (odds ratios <1) indicate survival is decreased when neighbors are more closely related.

 

TABLE C2. Neighborhood effects on survival beyond the first year (GEE, 5-m adult radius). Regression coefficients expressed as odds ratios for, from left to right, survival of new recruits over the first two years, survival of new recruits over the first three years, and survival of one-year old seedlings through the second year. See Methods for a description of the generalized estimating equation model; the 5-m radius is used for the adult neighborhood in this model. Significant predictor variables are highlighted in bold. Odds ratios >1 have a positive relationship with increased survival, and ratios <1 have a negative relationship with survival. Model ROC values > 0.7 indicate acceptable discrimination ability.

Survival 0 to 2 1

Survival 0 to 3 2

Survival 1 to 2 3

Conspecific adult basal area

0.002

0.000

0.027

Heterospecific adult basal area

0.755

1.449

0.989

Phylogenetic distance (adults) 4

1.013

1.031

1.201

Conspecific seedling density

0.989

0.982

1.017

Heterospecific seedling density

0.997

1.001

0.991

Phylogenetic distance (seedlings) 4

0.921

0.870

0.889

Seedling species diversity

1.315

1.135

1.568

Initial seedling height

1.041

1.028

1.035

Height × Conspecific adult basal area

1.120

1.287

0.992

Height × Heterospecific adult basal area

1.012

0.978

0.956

Height × Conspecific seedling density

1.002

1.003

1.000

Height × Heterospecific seedling density

1.001

1.001

1.001

Height × Seedling diversity

1.004

1.017

0.999

Recruit Year = 2004 5

0.500

0.455

0.902

Recruit Year = 2005 5

0.343

0.326

0.646

Recruit Year = 2006 5

0.319

--

0.688

ROC

0.677

0.678

0.622

1 Analysis included 3690 seedlings from 160 species and four recruit cohorts.

2 Analysis included 2844 seedlings from 153 species and three recruit cohorts.

3 Analysis included 1950 seedlings from 151 species and four recruit cohorts.

4 Phylogenetic distance measured as an effect size with values >0 indicating phylogenetic clustering (neighborhood is more closely related than expected by null model) and values <0 indicating phylogenetic dispersion (neighborhood is more distantly related than expected from null model).  Positive effects (odds ratios >1) indicate better survival with increased phylogenetic clustering; negative effects (odds ratios <1) indicate survival is decreased when neighbors are more closely related.

5 Change in intercept relative to baseline cohort (recruit year = 2003).

 

TABLE C3. Neighborhood effects on survival beyond the first year (10-m adult radius). Regression coefficients expressed as odds ratios for, from left to right, survival of new recruits over the first two years, survival of new recruits over the first three years, and survival of one-year old seedlings through the second year. See Methods for a description of the GLMM and GEE models. Significant predictor variables are highlighted in bold. Odds ratios >1 have a positive relationship with increased survival, and ratios <1 have a negative relationship with survival. Model ROC values > 0.7 indicate acceptable discrimination ability.

Survival 0 to 2 1

Survival 0 to 3 2

Survival 1 to 2 3

Analysis

GLMM

GEE

GLMM

GEE

GLMM

GEE

Conspecific adult basal area

0.069

0.130

0.091

0.211

0.962

0.768

Heterospecific adult basal area

0.828

0.668

0.994

0.792

0.782

0.733

Phylogenetic distance (adults) 4

1.084

1.081

1.146

1.128

1.221

1.204

Conspecific seedling density

0.978

0.990

0.967

0.980

1.011

1.017

Heterospecific seedling density

0.995

0.997

1.000

1.001

0.992

0.993

Phylogenetic distance (seedlings) 4

0.872

0.885

0.789

0.825

0.860

0.872

Seedling species diversity

1.180

1.351

1.081

1.158

1.597

1.674

Initial seedling height

1.039

1.041

1.040

1.035

1.042

1.040

Height × Conspecific adult basal area

0.967

0.968

0.951

0.931

0.871

0.892

Height × Heterospecific adult basal area

1.001

1.009

0.984

0.995

0.990

0.996

Height × Conspecific seedling density

1.002

1.002

1.004

1.003

1.000

1.000

Height × Heterospecific seedling density

1.001

1.001

1.001

1.001

1.001

1.001

Height × Seedling diversity

1.012

1.003

1.023

1.016

0.997

0.994

Recruit Year = 2004 5

0.444

0.502

0.403

0.461

0.915

0.929

Recruit Year = 2005 5

0.285

0.342

0.261

0.325

0.616

0.642

Recruit Year = 2006 5

0.261

0.307

- -

- -

0.679

0.679

ROC

0.808

0.672

0.833

0.674

0.750

0.619

1 Analysis included 3690 seedlings from 160 species and four recruit cohorts.

2 Analysis included 2844 seedlings from 153 species and three recruit cohorts.

3 Analysis included 1950 seedlings from 151 species and four recruit cohorts.

4 Phylogenetic distance measured as an effect size with values >0 indicating phylogenetic clustering (neighborhood is more closely related than expected by null model) and values <0 indicating phylogenetic dispersion (neighborhood is more distantly related than expected from null model).  Positive effects (odds ratios >1) indicate better survival with increased phylogenetic clustering; negative effects (odds ratios <1) indicate survival is decreased when neighbors are more closely related.

5 Change in intercept relative to baseline cohort (recruit year = 2003).

 

TABLE C4. Effect of seedling and adult neighborhoods (5–10 m radius) on individual seedling survival, excluding the most common species. Analysis includes 154 morphospecies and 2997 individuals; the nine most common seedling species (each with > 100 recruits) were excluded. Seedling neighbors are found within the same 1-m2 seedling plot as each focal seedling, and tree neighborhoods contain adults > 10 cm DBH in circles with the indicated radii surrounding each plot. See Methods for a description of the regression model. Coefficient estimates are expressed as odds ratios for the fixed-factor predictors in a GEE or GLMM model. Significant predictor variables are highlighted in bold. Odds ratios >1 have a positive relationship with increased survival, and ratios <1 have a negative relationship with survival.

 

GLMM

GEE

Adults: 5-m

Adults:10-m

Adults: 5-m

Adults:10-m

Conspecific adult basal area

0.001

0.022

0.003

0.041

Heterospecific adult basal area

0.612

0.949

0.613

0.903

Phylogenetic distance (adults) 1

0.810

0.861

0.815

0.871

Conspecific seedling density

0.958

0.952

0.962

0.960

Heterospecific seedling density

0.992

0.991

0.992

0.991

Phylogenetic distance (seedlings) 1

0.942

0.936

0.953

0.942

Seedling species diversity

0.689

0.676

0.702

0.695

Initial seedling height

1.002

1.003

1.000

0.999

Height × Conspecific adult basal area

1.386

1.176

1.298

1.145

Height × Heterospecific adult basal area

1.043

1.006

1.043

1.009

Height × Conspecific seedling density

1.000

1.000

1.001

1.001

Height × Heterospecific seedling density

1.001

1.001

1.001

1.001

Height × Seedling diversity

1.038

1.040

1.036

1.037

Recruit Year = 2004 2

0.250

0.237

0.269

0.256

Recruit Year = 2005 2

0.190

0.183

0.208

0.202

Recruit Year = 2006 2

0.143

0.137

0.165

0.162

Recruit Year = 2007 2

0.141

0.136

0.159

0.156

ROC

0.791

0.797

0.724

0.719

1 Phylogenetic distance measured as an effect size with values >0 indicating phylogenetic clustering (neighborhood is more closely related than expected by null model) and values <0 indicating phylogenetic dispersion (neighborhood is more distantly related than expected from null model).  Positive effects (odds ratios >1) indicate better survival with increased phylogenetic clustering; negative effects (odds ratios <1) indicate survival is decreased when neighbors are more closely related.

2 Change in intercept relative to baseline cohort (recruit year = 2003).


[Back to E091-258]