Ecological Archives E095-195-A2

Adam S. Hadley, Sarah J. K. Frey, W. Douglas Robinson, W. John Kress, Matthew G. Betts. 2014. Tropical forest fragmentation limits pollination of a keystone understory herb. Ecology 95:2202–2212. http://dx.doi.org/10.1890/13-0929.1

Appendix B. Complete predictor variable correlation matrix and nuisance variable justifications.

Table B1. Correlation matrix for all predictor variables. Pearson's r correlations are shown.

   

Config.

Local-scale

Landscape-scale

Nuisance variables

   

psize

PF100

PF250

pf500

pf1000

pf2000

ht500

ht1000

ht2000

mr500

mr1000

elevation

stream

edge

height

bracts

heto

richness

flowers

Composition and
configuration
metrics

psize

1.000

0.509

0.629

0.627

0.313

0.042

0.412

0.348

0.178

0.063

-0.042

0.134

0.253

0.439

0.082

0.124

0.055

0.134

0.059

pf100

0.509

1.000

0.753

0.563

0.236

-0.033

0.550

0.473

0.411

0.125

0.049

0.244

-0.153

0.369

0.252

0.261

-0.019

-0.086

0.055

pf250

0.629

0.753

1.000

0.879

0.470

0.098

0.550

0.393

0.251

-0.081

-0.038

-0.030

0.183

0.378

0.072

0.099

-0.217

-0.350

-0.174

pf500

0.627

0.563

0.879

1.000

0.771

0.402

0.656

0.523

0.372

-0.149

-0.166

0.054

0.345

0.408

0.134

-0.091

-0.098

-0.299

-0.178

pf1000

0.313

0.236

0.470

0.771

1.000

0.832

0.615

0.568

0.516

-0.299

-0.441

0.238

0.482

0.261

0.209

-0.265

-0.012

-0.128

-0.184

pf2000

0.042

-0.033

0.098

0.402

0.832

1.000

0.408

0.391

0.434

-0.443

-0.657

0.315

0.466

0.022

0.182

-0.211

0.057

0.065

-0.182

ht500

0.412

0.550

0.550

0.656

0.615

0.408

1.000

0.946

0.804

-0.202

-0.243

0.690

-0.036

0.309

0.364

-0.005

0.210

0.163

0.065

ht1000

0.348

0.473

0.393

0.523

0.568

0.391

0.946

1.000

0.915

-0.166

-0.221

0.769

-0.165

0.251

0.440

0.000

0.223

0.240

0.092

ht2000

0.178

0.411

0.251

0.372

0.516

0.434

0.804

0.915

1.000

-0.158

-0.285

0.805

-0.267

0.202

0.565

0.079

0.175

0.229

0.038

mr500

0.063

0.125

-0.081

-0.149

-0.299

-0.443

-0.202

-0.166

-0.158

1.000

0.693

-0.058

-0.118

0.288

-0.043

0.198

0.243

-0.139

0.129

mr1000

-0.042

0.049

-0.038

-0.166

-0.441

-0.657

-0.243

-0.221

-0.285

0.693

1.000

-0.274

-0.239

0.125

-0.211

0.077

0.053

-0.273

-0.030

Nuisance
variables

elevation

0.134

0.244

-0.030

0.054

0.238

0.315

0.690

0.769

0.805

-0.058

-0.274

1.000

-0.216

0.043

0.487

0.136

0.370

0.485

0.287

stream

0.253

-0.153

0.183

0.345

0.482

0.466

-0.036

-0.165

-0.267

-0.118

-0.239

-0.216

1.000

-0.050

-0.216

-0.245

-0.058

-0.137

-0.039

edge

0.439

0.369

0.378

0.408

0.261

0.022

0.309

0.251

0.202

0.288

0.125

0.043

-0.050

1.000

0.140

-0.009

0.131

-0.060

-0.143

height

0.082

0.252

0.072

0.134

0.209

0.182

0.364

0.440

0.565

-0.043

-0.211

0.487

-0.216

0.140

1.000

0.430

0.092

0.203

0.066

bracts

0.124

0.261

0.099

-0.091

-0.265

-0.211

-0.005

0.000

0.079

0.198

0.077

0.136

-0.245

-0.009

0.430

1.000

0.245

0.256

0.130

heto

0.055

-0.019

-0.217

-0.098

-0.012

0.057

0.210

0.223

0.175

0.243

0.053

0.370

-0.058

0.131

0.092

0.245

1.000

0.635

0.707

richness

0.134

-0.086

-0.350

-0.299

-0.128

0.065

0.163

0.240

0.229

-0.139

-0.273

0.485

-0.137

-0.060

0.203

0.256

0.635

1.000

0.609

flowers

0.059

0.055

-0.174

-0.178

-0.184

-0.182

0.065

0.092

0.038

0.129

-0.030

0.287

-0.039

-0.143

0.066

0.130

0.707

0.609

1.000

Landscape variables are – Configuration [psize (size of forest patch)]. Local-scale composition [pf100 (proportion of forest within 100 m), pf250 (proportion of forest within 250 m) and Landscape-scale composition [pf500 (proportion of forest within 500 m), pf1000 (proportion of forest within 1000 m), pf2000 (proportion of forest within 2000 m), ht500 (amount of heliconia habitat within 500 m), ht1000 (amount of heliconia habitat within 1000 m), ht2000 (amount of heliconia habitat within 2000 m), mr500 (ornithophilous flowers in matrix within 500 m), mr1000 (ornithophilous flowers in matrix within 1000 m)]. Nuisance variables are – elevation (elevation at the plant), stream (distance to the nearest stream), edge (distance to the nearest edge), height (plant height), bracts (number of bracts on the inflorescence), heto (number of flowering H. tortuosa plants within 20 m), richness (ornithophilous flower richness), flowers (total number of ornithophilous flowers).

Note on correlated variables in models:

Some predictor variables are highly correlated, but we never include these in the same model. We preselected the top nuisance and composition variables before final models (See Methods and Appendix E)

Additional background on variable choice:

Biological justification of local and nuisance variables:

To control for plant vigor:

Plant height – Plant height is a good indication of plant age and vigor (Bruna and Kress 2002). Older and more vigorous plants might be able to attract more pollinators and also produce more seeds. Height was measured from the ground to the top of the tallest petiole.

Number of bracts on the inflorescence - This is an indication of plant vigor. Inflorescence size also has been shown to affect attractiveness to pollinators and number of flowers visited in other plant species (Ghazoul 2005). We counted all of the bracts on the inflorescences of marked plants.

To control for differences in local flowering community:

Conspecific flower densities – Conspecific flower densities have often been shown to be an important aspect of pollination. We counted all flowering H. tortuosa within 20 m of each focal plant.

Flowering community richness – Other co-flowering species can compete for pollinator services and impede pollination (Murcia and Feinsinger 1996). This provides an index of floral species competing for hummingbird visits or conversely available for attracting hummingbirds. We counted all species of ornithophilous flowers within 20 m of each focal plant. Ornithophilous: Since we were primarily interested in flowering resources that were available to hummingbirds as alternative food sources, we considered ornithophilous flower species to be "any flowering species visited by hummingbirds." This included species know to be visited by hummingbirds from existing literature and any species local naturalists or authors had seen hummingbirds using.

Heterospecific flower densities – Local flowering resources can be important for both attracting pollinators or as competition for their visits. We counted all ornithophilous flowers within 20 m of each focal plant.

To control for differences in plant environment:

Elevation - H. tortuosa spans a fairly broad elevation range within our study area. Elevation affects temperature, moisture, plant species richness, phenology and potentially hummingbird species richness. Changes in any of these factors could result in differences in plant fitness. We determined the elevation of each plant from a 10 m elevation DEM layer. Note: We also stratify sites across elevation in study design.

Distance to nearest stream - H. tortuosa requires a fairly moist environment and often is at a higher density along streams. Hummingbirds also tend to follow streams during their movements (Hadley, unpublished data). We measured the distance to the nearest stream using GIS.

Distance to nearest edge - Edges frequently have effects on plants, animals and their interactions (Murcia 1995). We measured the distance to the nearest forest edge using GIS.

Literature cited

Bruna, E. M., and W. J. Kress. 2002. Habitat fragmentation and the demographic structure of an Amazonian understory herb (Heliconia acuminata). Conservation Biology 16:1256–1266.

Ghazoul, J. 2005. Pollen and seed dispersal among dispersed plants. Biological Reviews 80:413–443.

Murcia, C. 1995. Edge effects in fragmented forests: implications for conservation. Trends in Ecology & Evolution 10:58–62.

Murcia, C. and P. Feinsinger. 1996. Interspecific pollen loss by hummingbirds visiting flower mixtures: Effects of floral architecture. Ecology 77:550–560.


[Back to E095-195]