Transcript Document

Characterizing Bentgrass Distribution with Spatial and Biological
Data To Support Ecological Risk Assessment in Connecticut
1
C,
Ahrens,
J.
2
Chung ,
T.
2
Meyer ,
and C.
1
Auer
1
Department of Plant Science, University of Connecticut
2 Department of Natural Resources and the Environment, University of Connecticut
INTRODUCTION
Photographs: Habitats that could be
All new technologies carry both benefits and risks. For example, medical research
tests the effects of new drugs before they are approved for sale. Likewise, geneticallymodified (GM) crops must be assessed before they are released into the environment.
Our research examines the potential for gene flow and negative ecological impacts
from the release of GM plants. Gene flow can produce hybrid offspring with
transgenes and novel traits that could change the ability of the plant to survive and
spread. If hybrid offspring have an advantage in the environment, they could become
invasive and/or affect other components of our ecosystems.
Creeping bentgrass (Agrostis stolonifera) is a common, non-native turfgrass that is a
weed and could hybridize with other Agrostis species. At present, United States
Department of Agriculture (USDA) is considering an application to allow
commercialization of genetically-modified herbicide-resistant (HR) creeping
bentgrass. If approved, there is a probability that the transgenic HR trait would move
into feral bentgrass populations and could create environmental hazards over various
temporal and spatial scales. Connecticut has three native bentgrasses and four nonnative bentgrasses that form a hybridizing complex. The non-native bentgrasses (A.
stolonifera, A. gigantea, A. canina and A. capillaris) are highly adaptable, weedy species
that are widely distributed in wet meadows, agricultural fields, lawns, and roadside
areas.
The escape of HR creeping bentgrass could lead to risks that are difficult to accurately
predict such as more harmful weed management practices, genetic pollution, increased
‘weediness’, and loss of community and ecological diversity (see photographs). This
project aims to predict potential ecological risk by modeling bentgrass distribution in
New England, and the probability that gene flow and seed dispersal will move the HR
trait into feral and cultivated bentgrass populations. This poster reports on the
development of a Habitat Suitability Model using site visits, ecological variables and
GIS data for a golf course study site in central Connecticut.
RESULTS and discussion
The survey of 290 randomly-chosen plots in the HSM study site showed that 77% of
bentgrass populations were associated with little to no tree canopy (Table 1). This is
consistent with the requirement that most grass species have for high light
conditions. Over 92% of the non-native species (A. stolonifera, A. gigantea and A.
capillaris) were found in areas covered predominantly by herbaceous plants (Table
1). The survey suggests that there is positive correlation between non-native
bentgrasses and invasive plant species (Table 1). More importantly, some bentgrass
populations lie within areas of critical habitat for state-listed animal and plant
species, including species that are endangered in Connecticut (Figure 1, B-C.).
These observations are important because species at both ends of the spectrum
(invasive species and endangered species) may need management regimes that
could include herbicide. If HR creeping bentgrass escapes from the golf course,
these management practices could be compromised.
impacted by release of herbicide-resistant
bentgrasses. The freshwater marsh (left) has
invasive purple loosestrife and non-native
bentgrasses. Spraying glyphosate to control
purple loosestrife could allow spread of HR
bentgrass populations.
Roadside plant
communities (right) could be altered by
weed control with glyphosate, a herbicide
commonly used by homeowners.
In
Connecticut, many sunny roadside areas
have one or more bentgrass species in the
plant community.
Figure 1. The Habitat Suitability Model involved surveys of 290 plots. A) Occurrence of bentgrass species. Yellow dots = no bentgrasses, red dots = bentgrasses
present. B) Study site with overlay of Connecticut Natural Diversity Database. The pale circles represent areas where state-listed species are found. C) CoC
A
B
occurrence
of critical habitat for state-listed species and bentgrass
populations found during the surveys.
A
C
B
If a HSM for bentgrass is validated, it suggests that simple environmental variables
can be useful for land managers and government agencies tasked with predicting
bentgrass gene flow or monitoring transgene movement. These models could help
predict the degree of risk to critical habitats that need herbicide management to
help preserve state-listed species.
METHODS
The Habitat Suitability Model (HSM) project was conducted in central
Connecticut in 2008. The study site (8.5 km2) surrounded a golf course that uses
creeping bentgrass in putting greens and fairways (Figure 1A). Other criteria used
in selecting the study site were the presence of natural areas, roads, utility right-ofways, old agricultural fields, and recreation areas. These diverse land uses were of
interest because they are managed ecosystems that could have herbicide
applications to remove weeds or invasive plants. The research team visited 290
random plots (10 m2) in the study site to collect botanical and ecological
information (Figure 1A). Environmental variables were recorded at each plot
including percent tree canopy cover (sun/shade), percent area covered by leaf litter,
wetlands status, invasive plants present, bentgrass species present and other
features.
The HSM is built upon survey data and remotely-sensed GIS layers. This task can
be couched as a statistical problem using multivariate logistic regression (MLR):
given an observed dichotomous outcome (the presence or absence of bentgrass
species) and a set of observations of explanatory variable factor, the model will
estimate the probability of finding bentgrass in an unobserved location subject to
the same explanatory factors.
This poster presents the first efforts to generate a statistical model using the survey
data collected in 2008. Future efforts will improve and validate the model.
A Habitat Suitability Model (Multivariate Logistic Regression using three
environmental variables) was generated to examine the ability of specific
environmental features to predict the presence/absence of bentgrass populations.
The most complex model had the lowest AIC number and explained the most
variability (Table 2). Based on that model, locations with little or no forest litter, the
absence of wetlands, and little or no tree canopy (high light conditions) have the
highest probability of supporting native and/or non-native bentgrass species.
Observations by the research team (Table 1) support this conclusion. In future,
efforts will be made to test other environmental variables and validate the model.
This work is supported by grants from the USDA Biotechnology Risk Assessment
Grant Program and the University of Connecticut.
Table 2. Models produced using bentgrass survey data and STATA software. Three parameters were run in the model (Column 1): litter
Table 1. The first column lists all bentgrass species found in the study. “Unclassified” means that
cover, tree canopy cover, and wetlands status. The interaction effects used in the model are represented by letters (l, t and w). Log
likelihood (ll), number of parameters (k), and Akaike Information Criteria (AIC) are shown. Models are ranked by the AIC values, a
measure of the goodness of fit of an estimated model. The lowest AIC value represents the model with the best explanatory value, but
models within 2 units provide the same explanatory value.
plants could be identified to the genus Agrostis but not a specific species. The “unclassified” plants
were mostly found in sites where mowing prevented the development of flower panicles which are
helpful for species identification. Column two shows the number of plots containing each species.
The four other columns show the percentage of the bentgrass species (n) found in association with:
open sunny habitat, invasive plant species, area with active management practices, or area with
predominantly herbaceous plants.
Model
litter, tree,
litter, water
litter, tree,
litter, tree,
litter, tree,
litter, tree,
litter, tree,
litter, tree,
litter, tree,
litter, tree,
lree, water
litter, tree
litter
tree
water
ll(model)
water, l*t, l*w, t*w, l*t*w
water
water,
water,
water,
water,
water,
water,
water,
l*t
l*w
t*w
l*t,
l*t,
l*w,
l*t,
l*w
t*w
t*w
l*w, t*w
-103.0749
-108.5988
-108.2004
-107.7019
-107.7744
-108.1588
-107.2971
-107.6635
-107.6785
-107.1792
-116.0785
-118.283
-120.0326
-123.6708
-142.5174
K
7
2
3
4
4
4
5
5
5
6
2
2
1
1
1
AIC
220.1497
225.1976
226.4008
227.4038
227.5488
228.3176
228.5942
231.327
231.357
232.3583
240.157
242.5659
244.0653
251.3415
291.0347
Species
A. stolonifera
A. gigantea
A .perennans
A. scabra
A. capillaris
A. canina
Unclassified
Total
n
13
19
9
4
5
2
17
69
Tree canopy
25% or less
77%
90%
33%
75%
80%
.
82%
77%
Found with Found near Herbaceous cover
invasives managed areas ove r 50%
77%
62%
92%
80%
68%
100%
55%
22%
33%
25%
50%
50%
60%
60%
100%
.
.
.
18%
94%
88%
53%
61%
90%