Online Resource 2: results of background data analyses ARTICLE

advertisement
Online Resource 2: results of background data analyses
ARTICLE TITLE: Effects of urbanization on herbaceous forest vegetation: the
relative impacts of soil, geography, biotic interactions, human access, and an
invasive shrub
JOURNAL: Urban Ecosystems
AUTHORS: Guy N. Cameron1, Theresa M. Culley1, Sarah E. Kolbe2, Arnold I. Miller2, and
Stephen F. Matter1
AFFILIATION: 1Department of Biological Sciences and 2Department of Geology
University of Cincinnati, Cincinnati Ohio 45221
Corresponding Author: G. Cameron (g.cameron@uc.edu)
Literature and figures cited below are in the published manuscript
Online Resource 2a: Assessing the relative importance of a site effect and a year effect
While preliminary analyses indicated that study site significantly affected herb diversity,
richness, and abundance, Christopher et al. (2014) demonstrated that herb diversity, richness, and
abundance in these deciduous forests in southwestern Ohio varied significantly among years.
Because MWW, MAF, and TRA were sampled 1-3 years later than BEN, EF, and EOA, we used
a General Linear Model (GLM) to determine the relative impact of study site and year of
sampling on these herb metrics. Based on when we sampled herbs in each study site, we coded
BEN, EF, and EOA as Year 1 (sampled in 2008), MWW and MAF as Year 2 (sampled in 2009),
and TRA as Year 3 (sampled in 2011) for the GLM analyses. For the count variables (e.g.,
richness and abundance) we specified a quasipoisson distribution because the data were
overdispersed. We analyzed the effect of study site and year of sampling separately for herb
diversity, richness, and abundance. We computed the amount of the deviance in these herb
metrics that was accounted for by study site and the amount that was due to year of sampling by
subtracting the residual deviance from the null deviance. We determined the relative importance
of study site and year of sampling by comparing the amount of deviance in herb diversity,
richness, and abundance that each factor explained. We also analyzed data on temperature and
rainfall for the months/ years of our study (May-September 2008-2011; data from
Cincinnati/Northern Kentucky International Airport) to determine whether differences among
years would affect our censuses.
Compared to year of sampling, study site explained 2-16 times the amount of deviance in
species diversity (deviance explained by study site= 2.06, deviance explained by year of
sampling = 1.18), richness (deviance explained by study site = 58.81, deviance explained by year
of sampling = 28.39), and abundance (deviance explained by study site = 2308.8, deviance
explained by year of sampling = 141.1).
There was little variation among years in temperature (20.8-23.0oC), but total MaySeptember rainfall was higher in 2009 (116.9 mm) and 2011 (147.1 mm) than in 2008 (80.3 mm)
or 2010 (85.0 mm). Higher rainfall during some years did not always result in higher herb
abundance (Fig. 2). TRA, the only site sampled in 2011, had very low abundance, but rainfall
was high. The sites sampled in 2008 (MWW and MAF) had low and high abundance,
respectively, but rainfall was low. From these analyses we concluded that a year effect would be
minor compared to a site effect.
Online Resource 2b: Variation in community metrics and environmental variables among
study sites
Results of Analysis of Variance to determine whether environmental and biotic variables
varied significantly among the 6 study sites and among Urban, Exurban, and Wildland
sites along the urbanization gradient. All variables were tested for significant variation
among study sites, whereas only variables identified in Generalized Linear Mixed Model
(GLMM) and General Linear Model (GLM) analyses as significantly affecting herb
species diversity, richness, or abundance were tested among Urban, Exurban, and Wildland
sites.
Variation among
Variation among Urban,
6 study sites
Exurban, Wildland sites
Variable
F2,85
P
Variable
F2,85
P
Herb species diversity
10.4
<0.01
Herb species diversity
1.02
0.368
Herb richness
5.98
<0.01
Herb richness
10.52
<0.01
Herb abundance
5.85
<0.01
Herb abundance
5.47
<0.01
pH
4.48
<0.01
pH
7.06
<0.01
Percent C
8.93
<0.01
Percent C
5.14
<0.01
Percent N
5.12
<0.01
Slope
3.82
0.026
Soil bulk density
1.21
0.312
Aspect
7.62
<0.01
Slope
25.05
<0.01
Elevation
0.52
0.599
Aspect
3.63
<0.01
Tree diversity
3.94
0.023
Elevation
60.18
<0.01
Tree richness
5.98
<0.01
Tree diversity
2.71
0.026
Tree abundance
3.76
0.027
Tree richness
4.19
<0.01
Distance to nearest road
11.31
<0.01
Tree evenness
1.80
0.12
Distance to nearest major
109.9
<0.01
12.85
<0.01
road
Tree abundance
3.65
<0.01
Mean diameter of primary
stem
Distance to nearest road 16.66
<0.01
Distance to nearest
189.30
<0.01
4.85
<0.01
8.76
<0.01
major road
Honeysuckle
abundance
Mean diameter of
primary stem
Results of multiple comparison of means: After determining which community metric and
environmental variable varied significantly among our 6 study sites, we used multiple
comparisons of means (Tukey Honestly Significant Differences) for each variable to determine
which study sites were significantly different from each other. Herb diversity did not vary
significantly along the urbanization gradient, but diversity was highest at TRA and lowest at
EOA (Fig. 2). Herb richness was highest at TRA and EOA with MWW intermediate, and herb
abundance was highest at MWW, EOA, and TRA with MAF intermediate (Fig. 2). Among the
environmental variables (Fig. 3), pH was lowest at EF and MWW, percent C was highest at
MAF and EOA, and percent N was highest at MAF, MWW, and EOA with TRA intermediate.
Aspect at EOA and TRA was to the southwest, while aspect for the other sites was to the
southeast, elevation was lowest at MWW and EOA and highest at TRA, and slope was greatest
at MAF, MWW, and TRA. Tree diversity was highest at EF, MWW, EOA, and TRA, tree
richness was highest at TRA, EF, and MWW, and tree abundance was highest at TRA. Distance
to the nearest road was highest at MWW and EOA, and distance to the nearest major road was
highest at EF and TRA, with MWW at an intermediate level. Honeysuckle abundance was
highest at MAF and MWW and diameter of primary stem of honeysuckle was highest at MAF
and EF.
Online Resource 2c: Environmental variables selected by GLMM as important in
determining herb diversity
Effect of environmental and biotic variables on species diversity of forest-floor herbs. Edaphic,
Geographic, Forest Composition, Human Effects, and Honeysuckle thematic models containing
the variables listed below were constructed to test their effect on species diversity. Each model
was analyzed with a Generalized Linear Mixed Model (GLMM) that included the variables listed
below as fixed effects, and site as a random effect. To determine the significance of each factor
included in a model, comparisons were used to test the full model that included all fixed effects
against a model with an individual fixed effect removed. Tests between full and reduced models
were made using maximum likelihood tests (Crawley 2013). For each individual variable, a Chisquared value was computed as the deviance difference between the full model (all variables
included) and a model with one variable removed.
2
df
P
Soil density
1.432
1
0.23
pH
12.81
1
<0.01*
Percent N
0.06
1
0.81
Percent C
2.65
1
0.10
Aspect
3.04
2
0.18
Elevation
10.07
1
<0.01*
Slope
3.21
1
0.07
Tree Abundance
4.12
1
0.04*
Tree Evenness
1.01
1
0.32
Tree Richness
0.19
1
0.66
Thematic Model
Variable removed
Edaphic
Geographic
Forest Composition
Human Effects
Honeysuckle
Tree Diversity
0.18
1
0.67
Distance to Nearest Major Road
0.03
1
0.87
Distance to Nearest Road
9.41
1
<0.01*
Mean Diameter of Primary Stem
0.01
1
0.94
Honeysuckle Abundance
1.49
1
0.22
Online Resource 2d: Environmental variables selected by GLMM as important in
determining herb richness
Effect of environmental and biotic variables on species richness of forest-floor herbs. Edaphic,
Geographic, Forest Composition, Human Effects, and Honeysuckle models containing the
variables listed below were constructed to test their effect on species richness. Each model was
analyzed with a Generalized Linear Mixed Model (GLMM) that included the variables listed
below as fixed effects, and site as a random effect. To determine the significance of each
variable included in a model, comparisons were used to test the full model that included all
variables against a model with an individual variable removed. Tests between full and reduced
models were made using maximum likelihood tests (Crawley 2013). For each individual
variable, a Chi-squared value was computed as the deviance difference between the full model
(all variables included) and a model with one variable removed.
2
df
P
Soil density
2.13
1
0.88
pH
14.98
1
<0.01*
Percent N
0.84
1
0.36
Percent C
3.36
1
0.07
Aspect
12.23
2
<0.01*
Elevation
0.14
1
0.70
Slope
0.04
1
0.84
Tree Abundance
2.12
1
0.15
Tree Evenness
<0.01
1
0.99
Thematic Model
Variable removed
Edaphic
Geographic
Forest Composition
Human Effects
Honeysuckle
Tree Richness
<0.01.
1
0.95
Tree Diversity
0.18
1
0.98
Distance to Nearest Major Road
6.82
1
<0.01*
Distance to Nearest Road
11.89
1
<0.01*
Mean Diameter of Primary Stem
0.21
1
0.65
Honeysuckle Abundance
0.01
1
0.93
Online Resource 2e: Environmental variables selected by GLMM as important in
determining herb abundance
Effect of environmental and biotic factors on abundance of forest-floor herbs. Edaphic,
Geographic, Forest Composition, Human Effects, and Honeysuckle models containing the
variables listed below were constructed to test their effect on abundance of forest-floor herbs.
Each model was analyzed with a Generalized Linear Mixed Model (GLMM) that included the
variables listed below as fixed effects, and site as a random effect. To determine the significance
of each variable included in a model, comparisons were made between the full model that
included all fixed effects and a model with an individual variable removed. Tests between full
and reduced models were made using maximum likelihood tests (Crawley 2013). For each
individual variable, a Chi-squared value was computed as the deviance difference between the
full model (all variables included) and a model with one variable removed.
2
df
P
Soil density
2.55
1
0.11
pH
89.09
1
<0.01*
Percent N
16.79
1
<0.01*
Percent C
24.75
1
<0.01*
Aspect
199.57
2
<0.01*
2.39
1
0.12
Slope
119.97
1
<0.01*
Tree Abundance
11.41
1
<0.01*
Tree Evenness
16.94
1
<0.01*
Tree Richness
149.81
1
<0.01*
Thematic Model
Variable removed
Edaphic
Geographic
Elevation
Forest Composition
Human Effects
Honeysuckle
Tree Diversity
73.47
1
<0.01*
Distance to Nearest Major Road
130.44
1
<0.01*
Distance to Nearest Road
59.84
1
<0.01*
Mean Diameter of Primary Stem
171.05
1
<0.01*
1.20
1
0.27
Honeysuckle Abundance
Online Resource 2f: Variation of community metrics and environmental factors among
Urban, Exurban, and Wildland sites
Results of multiple comparison of means. After determining which community metric and
environmental variable varied significantly among Urban, Exurban and Wildland sites (see
Online Resource 2b), we used multiple comparisons of means (Tukey Honestly Significant
Differences) for each variable to determine significant differences among Urban, Exurban, and
Wildland sites. Herb species richness and abundance were significantly higher in Wildland sites
than in Urban or Exurban sites which did not differ significantly from each other (Fig. 2). pH
and organic C were lowest in Exurban sites, slope was lowest in Wildland sites, and Wildland
sites faced southeast while Urban and Exurban sites faced southwest (Fig. 4). Tree diversity and
richness were highest in Exurban and Wildland sites and tree abundance was highest in Wildland
sites. Urban sites were closest to nearest roads and Exurban sites were farther from major roads,
with Wildland sites at intermediate distances. Diameter of primary stems of honeysuckle was
larger in Urban and Exurban sites than in Wildland sites.
Download