Using experimental seed banks, I examined the role of the size and timing ofdisturbance and seed density on seed bank formation.. I experimentally varied seed density, and soil disturban
THE MECHANISMS AND CONSEQUENCES OF SEED BANK FORMATION
Despite their importance to plant populations, it is unclear how seed banks form I experimentally varied seed density, and soil disturbance timing and size to examine the formation of seed banks and the resulting population dynamics in wild annual sunflowers (Helianthus annuus) Seed density treatments reflect the increase in wild sunflower reproduction expected due to hybridization with Bt-transgenic crop sunflowers Soil disturbances were applied either before or after seed dispersal I followed the experimental plots for 2 ẵ years, measuring seedling emergence, seeds in the soil, and reproduction, while preventing additional seed input I found that soil disturbances, regardless of timing, are required for successful germination and reproduction of sunflowers, based on the first year’s data Large disturbances were more effective, having greater seedling emergence and reproduction Reproduction on small disturbance plots was low however, due to low seedling survival Higher seed density resulted in more seedlings only in large-disturbance plots Seed bank formation was also enhanced by soil disturbances, specifically large disturbances occurring after seed dispersal Burial of seeds through these disturbances likely allowed seeds to escape predation Higher initial seed density increased seed bank size only in the presence of large, post-dispersal disturbances Although more seedlings emerged with the presence of large, post-dispersal disturbances, negative density dependence in seedling mortality and reproduction meant that there was no effect of higher initial seed density on reproduction Thus the positive effects of dependence To assess fully the impact of seed banks on population growth, studies should continue beyond the seedling stage My results suggest that increased seed production of wild populations resulting from crop transgenes would be unlikely to increase population growth on the scale of small patches However, increased seed production could result in dispersal of seeds across larger areas, and thus could lead to larger seed banks in disturbed environments The net result would be an increased in the prevalence of sunflowers across the landscape.
Persistent seed banks, seeds that remain viable on or in the soil for more than one year, are found in many plants (Fenner & Thompson 2005) In variable environments, seed banks enhance population growth rates and reduce extinction risks (Kalisz & McPeek 1993; Fisher & Mattheis 1998; Menges & Quintana-
Ascencio 2004; Adams et al 2005) Persistent seed banks are particularly important for annual species because they provide a mechanism for dispersal through time, allowing annuals to survive unfavorable periods Because soil disturbances can generate conditions favorable for seed germination and plant survival, the seed bank is expected to play a large role in the population dynamics of many disturbance specialists, weeds, and invasive species (Rees & Long 1992; Grigulis et al 2001; Claessen et al 2005a).
Soil disturbances can have apparently contradictory effects on persistence of seeds in the soil On one hand, they can facilitate entry of seeds into the seed bank through burial, which decreases both predation and germination (Weaver & Cavers 1979; Froud-Williams et al 1984; Hulme & Borelli 1999; Meyer & Schmid 1999).
On the other hand, soil disturbances can remove seeds from the seed bank by improving germination conditions (Eriksson & Eriksson 1997; Fisher & Mattheis 1998; Meyer & Schmid 1999; Fenner & Thompson 2005) The exact effect of a soil disturbance on seed bank formation will likely depend upon the timing of the disturbance relative to seed dispersal, and the size of the disturbance, as well as the landing on bare ground, increasing the risk of being eaten or washed away (García- Fayos & Cerdá 1997; Whittingham & Markland 2002) Only post-dispersal disturbances bury seeds, protecting them from germination cues and predation. Disturbances occurring at any time are likely to improve conditions for seed germination The importance of the timing of soil disturbance for seed bank formation will likely depend upon the relative strengths of predation, disease, and germination in removing seeds from the seed bank The size of the disturbance may also affect seed bank formation Large, continuous soil disturbances are likely to bury more seeds, yet also generate more uniformly positive conditions for germination compared to small, patchy soil disturbances.
The effect of disturbance on seed bank formation may also depend upon the number of seeds initially dispersed at the site One might expect that more seeds would result in larger seed banks However, many processes that remove seeds from the seed bank may be density-dependent In several species, post-dispersal seed predation rates increase with increasing seed density (Price & Heinz 1984; Hulme & Borelli 1999; Cabin et al 2000) and seeds at higher densities are more susceptible to disease (Van Mourik et al 2005) Seed density can also influence germination rates, although whether germination increases, decreases, or remains unchanged with increasing seed density appears to be species-specific (Linhart 1976; Rebollo et al. 2001).
Even if greater seed production and dispersal lead to larger seed banks, density-dependent processes at later life stages may temper the role of the seed bank in population dynamics For example, density-dependent reductions in survival and reproduction at high seedling densities could mean that larger seed banks are inconsequential for population growth, even if they lead to increased seedling emergence in favorable environments (Alexander & Schrag 2003; Maron &
Kauffman 2006) Most experimental seed bank studies do not follow plants past the seedling stage and thus effects of increased seed bank size on population dynamics remain unclear (but see Alexander and Schrag, 2003; Maron & Kauffman, 2006).
The wild annual sunflower (Helianthus annuus) is ideal for examining the role of soil disturbances in seed bank formation First, sunflowers form persistent seed banks, with individual seeds surviving from 1-2 to 10+ years (Burnside et al 1996; Teo-Sherrell 1996; Alexander & Schrag 2003) The contribution of the seed bank to the above-ground population can be relatively high (10-23%) compared to other species (Alexander & Schrag 2003) Second, because sunflowers are annuals, effects of experimental seed bank treatments can easily be followed through the entire life- cycle and multiple generations Third, wild sunflowers produce large seeds, allowing for relatively easy quantification of seed banks Finally, H annuus is found in disturbed areas; if disturbances are not maintained, the above-ground population eventually becomes extinct (Cummings & Alexander 2002).
I was specifically interested in the effects of increased seed density on seed bank formation in sunflowers, as they have become a model system for examining the risk of crop/wild hybridization (Snow et al 1998; Cummings & Alexander 2002). considered agricultural weeds (Humston et al 2005) Recent development of transgenic sunflower has caused concern about incorporation of crop transgenes into wild populations Crop and wild sunflowers commonly hybridize in nature and crop- specific genes can persist in wild populations for multiple generations (Arias & Rieseberg 1994; Whitton et al 1997) A recent experiment demonstrated that populations of wild sunflowers containing the Bt transgene, which reduces pre- dispersal seed predation, had a 55% increase is seed production under field conditions (Snow et al 2003) If the Bt gene were to become incorporated into wild populations of sunflowers, there is a risk that increased seed production may lead to the formation of “super-weed” populations, which can be characterized by faster population growth rate and/or an increase in the prevalence of populations across the landscape Whether increased seed production in wild populations due to transgenes results in increased weediness will depend, in part, on the influence of seed density on seed bank formation and the strength of density-dependent processes in regulating sunflower populations.
In this study, I investigate the roles of soil disturbances and seed density in seed bank formation of wild annual sunflowers By creating and following experimental plots for 2 ẵ years, I was also able to explore potential effects of my treatments on population dynamics Specifically, I addressed the following questions:(1) How does the timing of disturbance relative to seed-dispersal and the density of seeds initially dispersed influence seed bank formation? (2) Do different types of disturbance (large, continuous vs small, patchy) differentially impact seedling germination, seed bank formation, and reproduction? (3) Is it likely that density- dependent processes will temper the impact of increases in the size of the seed bank on population growth?
Wild annual sunflowers (Helianthus annuus, Asteraceae), native to the North American Great Plains, are found worldwide Though present on native prairies, they are more commonly found is disturbed areas such as open fields, roadsides, and railroad right-of-ways (Heiser 1954) The cypselas (hereafter “seeds”) are relatively large (7 mg, Alexander et al 2001) and nutrient rich, making them attractive to many vertebrate seed predators (Robel & Slade 1965; Teo-Sherrell 1996) In eastern
Kansas, USA, seedlings emerge from late March to early May (personal observation). Seed germination of H., annuus is strongly influenced by light availability and soil temperatures (Corbineau et al 1990; Baskin & Baskin 1998) The flowering period of sunflowers in eastern Kansas begins in July and continues into October Seed dispersal begins in September, though seeds are dispersed throughout the winter (Robel et al 1974) Sunflower seeds are dormant following dispersal and up to 40% of seeds may remain dormant during the spring germination period (Teo-Sherrell et al 1996; Snow et al 1998) Hybrid seeds have reduced dormancy and higher germination (Mercer et al 2006) This study was conducted in northeastern Kansas (USA) at the Nelson Environmental Studies Area (NESA), part of the University of Kansas’s Field Station and Ecological Reserves, in a field dominated by an introduced grass (Bromus inermis).
In November 2001, experimental seed banks were created, with a total of 21 treatments based upon seed density, disturbance size, pre-dispersal disturbance, and post-dispersal disturbance (Fig 1-1) Experimental plots were 127 cm by 127 cm. Each treatment was replicated 10 times for a total of 210 seed bank plots These were laid out in a randomized block design with each of the 21 treatments represented in each of 10 blocks Plots were spaced 5 meters apart to reduce the potential effects of interplot seed movement (Teo-Sherrell 1996) In order to most closely represent natural conditions, I did not weed other species from my experimental plots, and even when completely tilled, plots were not sunflower monocultures.
Three seed densities were used: 0 seeds added (control), ~1700 seeds added (low density), or ~ 2600 seeds added (high density) The low seed-density treatment represents the expected number of seeds in the soil resulting from dispersal by a patch of about 20 plants (Pilson and Alexander unpublished) The high density treatment represents a 55% increase in seed production, the expected increase in seed production of crop-wild hybrid sunflowers containing the Bt transgene (Snow et al.2003) Seeds used in this experiment were collected from a natural population located about 13 km southwest of NESA in October and November of 2001 Seeds were combined into a single sample and the approximate number of seeds needed per plot was determined by weight Seeds were sown by hand, with care taken to ensure that seeds were evenly distributed within the plot.
SCALING-DOWN METAPOPULATION DYNAMICS OF TWO ROADSIDE
from coarse- to fine-spatial scales
As ecologists recognize that empirical data collection and practical applications occur at different spatial scales (i.e size of observational units and the total area of study), demand for methods to translate information between spatial scales has increased In particular, conservation decisions require data on fine-scale species abundances, but data are often available only as coarse-scale occupancy data derived from atlases or museum specimens Thus there has been great interest in methods that attempt to extract information on fine-scale species’ abundances from coarse-scale distributional maps Here, I describe the application of fractal-based scaling-down methods to long-term, large-scale metapopulation data sets of two roadside plant species, Helianthus annuus L., (Asteraceae) and Silene latifolia Poir. (Caryophyllaceae) My goal was to determine if fine-scale colonization and extinction rates can be predicted from coarser-scaled data using fractal scaling methodology. Fine-scale data were collected from roadside populations within a linear grid, and were aggregated to generate data at several increasingly coarse scales The relationships between occupancy, colonization, or extinction at the coarser-scales and the scale of measurement (scale-curves) were determined using the standard “fully- nested” method, in which all data are used at each scale, and a “stratified-sampling” method, in which data at each scale represent a randomly chosen subset of the entire data set These relationships were then used to predict the field-collected fine-scale data In general, scaling-down methods were successfully applied to dynamic metapopulation data However, scaling-down generated more accurate predictions in
Helianthus compared to Silene, and in both species, scaling-down generated more accurate predictions for occupancy and colonizations compared to extinctions Both ecological explanations and characteristics of the data set may explain these patterns.For example, the reduced success of scaling in Silene may result from Silene’s more clumped distribution Further studies, possibly incorporating computer simulations,will be useful in exploring processes that contribute to different scaling patterns.
While some ecological phenomena are scale invariant (Steele & Forrester 2005), many are dependent upon the spatial scale (i.e extent and resolution) of observations and analysis (Levin 1992; Zajac et al 1998; Murdoch & Aronson 1999; Englund et al 2001; Saunders et al 2002) For example, Norowi et al (2000), in a study examining wasp parasitism of plant-feeding weevils, found that the density- dependence of parasitism varied with the spatial scale: parasitism was inversely density-dependent at the seed-head scale, directly density-dependent at the scale of an individual plant, and density-independent at a scale of 729 m 2 Scale dependence of ecological phenomena is problematic because commonly, and out of necessity, ecological studies are generally conducted at single spatial scales, but predictions and management decisions occur at other spatial scales (Kareiva & Andersen 1986; May 1989; Weins 1989; Schulze 2000) Thus, there is great need for scaling methods that allow information to be translated between spatial scales (May 1989; Levin 1992; Root & Schneider 1995; Hobbs 2003).
Scaling can occur in two directions Scaling-up methods use data collected at fine scales to infer coarse-scale patterns and processes, while scaling-down methods use data collected at course scales to infer patterns and processes at finer scales. Although scaling-up can be challenging (Schulze 2000; Englund & Hambọck 2004), such methods are commonly applied to a wide range of ecological questions
(Rastetter et al 1992; Inouye 2005; Urban 2005; Barnes et al 2006; Bergstrửm et al.2006) For example, photosynthetic rates of individual leaves can be incorporated into models that predict the CO2uptake of an entire forest (Rastetter et al 1992) Scaling- down had generally been thought impossible (Hartley et al 2004) However, recent studies have demonstrated that fine-scale species’ abundances can be predicted from coarse-scale occupancy data gathered from distributional maps (Kunin 1998; He & Gaston 2000; Kunin et al.; Kallimanis et al 2002; Cousens et al 2004; Halley et al. 2004; Hartley et al 2004; Tosh et al 2004) Scaling-down methodologies could be particularly useful to conservation efforts For example, in many cases fine-scale data on rare or invasive species’ abundance is needed, but such data are difficult to collect and knowledge of most species is limited to coarse-scale distributional maps (Kunin 1998; Tosh et al 2004) The need for scaling-down methods is only expected to increase as technological advances in GPS, GIS, and remote sensing increase the availability of large-scale ecological data, most often collected at coarse grains (Lonsdale 1993; Castle 1998; Withers & Meentemeyer 1999).
The use of fractal methodologies for scaling-down species’ distributions have been particularly successful (Kunin 1998; Cousens et al 2004; Hartley et al 2004;Tosh et al 2004) Fractal geometry was developed to describe the complexity of nature, which does not neatly fit into strictly Eucledian geometries (i.e line, plane,solid) (Mandelbrot 1983; Sugihara & May 1990; Halley et al 2004) Fractal objects have a non-integer, and thus non-Euclidean, dimension, and the fractal dimension measures the ability of an object of fill the Euclidean space in which it is embedded(Mandelbrot 1983) Another key feature of fractal phenomena is that they are self- same object For truly fractal objects, the fractal dimension remains constant across all scales (scale invariance) (Sugihara & May 1990; Hamburger et al 1996; Avnir et al 1998; Gonzato et al 1998; Halley et al 2004) However, most natural phenomena, including species’ distributions, are not expected to be self-similar across all spatial scales (Sugihara & May 1990; Avnir et al 1998; Gonzato et al 2000; Halley et al. 2004), and the fractal model should be viewed as a simplification of a complex natural world.
The scaling down of species’ distributions typically utilizes the box-counting method of estimating the fractal dimension, in which a species’ distribution is examined using different-sized quadrants (“boxes”) and scale-area curves are generated through linear regression of the log area-occupied as a function of the log scale of analysis (Fig 2-1 a-b) The slope of the resulting line measures the degree to which a species fills its geographic range; sparser distributions result in steeper slopes (Kunin 1998) Since few data sets exist where the species distribution is measured at multiple scales, this method is applied by generating increasingly coarse-resolution distribution maps from finer-resolution field-collected data covering the same spatial extent (Kunin, 1998; Tosh et al 2004; Pocock et al 2006, but see Cousens et al. 2004; Hartley et al 2004) Ideally, the linear relationship among the coarse-scale data and scale of measurement can then be used to generate accurate predictions of the species’ abundance at resolutions finer than the available data.
In the study of species distributions and abundances, fractal methodology has been applied to static occurrence data However, the distribution and abundance of a species depends in part upon metapopulation dynamics (Prince et al 1985; Hanski 1999; Menéndez & Thomas 2000; Silvertown & Antonovics 2001; Hanski &
Gaggiotti 2004), which are determined by population colonizations and extinctions.
To test the scaling-down of metapopulation dynamics, data must be collected yearly at both a fine spatial resolution and a large spatial extent Roadside surveys of plant populations offer an unusually rich source of this kind of data, as roadsides allow rapid assessment of populations, at least for highly visible plant species.
Consequently, these types of survey have been the source of plant metapopulation data for several species (Prince et al 1985; Antonovics et al 1994; Crawley & Brown 1995; Antonovics et al 1998; Antonovics et al 2001) Specifically, Janis Antonovics (University of Virginia) and colleagues have used roadside surveys to document the annual metapopulation dynamics of Silene latifolia (White campion,
Caryophyllaceae) in Virginia since 1988 (Antonovics et al 1994; Thrall &
Antonovics 1995; Antonovics et al 1998; Antonovics et al 2001; Antonovics 2004). Similar surveys have been conducted by Helen Alexander (University of Kansas) and Diana Pilson (University of Nebraska) for Helianthus annuus (wild sunflower,
Asteraceae) in the Great Plains from 1999-2003 These data sets are unique in both their spatial and temporal scales In this study, I use these two extensive databases on roadside plant metapopulation dynamics to determine if fine-scale colonization and extinction rates can be predicted from coarser-scaled data using the fractal-based scaling methodology.
Silene latifolia (hereafter Silene) is a ruderal, short-lived perennial native to Europe (McNeill 1977) In Virginia, where the metapopulation has been surveyed annually since 1988, Silene populations are frequently restricted to along roadsides. The survey route covers approximately 150 km of roadside in the Allegheny Mts of western Virginia The roads included in the survey routes were chosen without specific prior knowledge of Silene distributions along the road As roadsides represent a linear transect of habitat across environmental gradients, plants were censused in a one-dimensional grid overlying the roadside; the length of the grid segments is 40 m and the width is defined by the road verge (3-5 m) Endpoints of individual segments were defined using specific local landmarks, and were relocated each year using GPS coordinates and odometer readings The surveys used in this study were conducted from 1989-2002, during the flowering season in June when plants are most conspicuous Teams of two to three people, through a combination of driving and walking, surveyed the route by counting the number of individual Silene plants within each 40 m segment Only flowering individuals were counted Data were collected along both sides of the road A second survey has been conducted later in the summer each year to verify colonizations and extinctions within segments These surveys represent “snap-shots” in time of annual changes in occupancy Although the entire metapopulation data set contains several thousand points over 150 km, I applied scaling-down methods to a smaller sub-section of the routes, approximately 74 km, representing the longest continuous segment of the route (Table 2-1) For more detail about the Silene metapopulation see Antonvoics et al 1994, Thrall & Antonovics
1995, Antonovics et al 1998, Antonovics et al 2001, Antonovics 2004.
Helianthus annuus (hereafter Helianthus) is an annual species with a persistent seed bank (Burnside et al 1996; Teo-Sherrell 1996; Alexander & Schrag 2003) Native to the North American Great Plains, Helianthus is currently found throughout the continent (Heiser 1954) Helianthus plants were surveyed annually in eastern Kansas (KS) between 1999 and 2004 and in western Nebraska (NE) between
2001 and 2004, using methods modelled after the Silene metapopulation surveys The survey routes were located along rural roads, typical of the region In Kansas, the route followed roads near University of Kansas’ Kansas Field Station and Ecological Reserves, and the surrounding landscape was a mixture of residential areas, forest fragments, rangeland, and cropland In Nebraska, the survey route followed roads near University of Nebraska’s Cedar Point Biological Station, and the surrounding landscape was dominated by rangeland and crops The Helianthus routes were shorter than in Silene: 23.6 km in Kansas and 18.6 km in Nebraska The length of the segments in the Helianthus routes was 80 m, with width equal to that of the road verge (3-5 m) Surveys were conducted in late August and early September, when
THE REGIONAL ECOLOGY AND GEOGRAPHY OF WILD SUNFLOWERS (HELIANTHUS ANNUUS): LINKING MODELS OF PRESENCE WITH MODELS OF ABSENCE
presence with models of absence
Modeling ecological niches and predicting distributions of widespread, generalist species has often been problematic, as distributions of such species may be determined more by local-scale heterogeneity than by broad-scale climatic parameters Furthermore, many modeling environments assume that a species is absent at geographic locations where occurrences are not known; for widespread species, this assumption is unlikely to be robust, especially with small samples I challenged the Genetic Algorithm for Rule Production (GARP) to model the ecological niche of the generalist plant Helianthus annuus L (wild sunflower,
Asteraceae), using both presence and absence data from coarse-scale roadside surveys covering 5330 km across the Great Plains of the United States, and relatively high spatial resolution environmental data generated from MODIS satellite imagery
(Normalized Difference Vegetation Index and Enhanced Vegetation Index) A jackknifing technique was used to determine which environmental data were most informative for generating accurate models This study is unique in that I incorporated explicit models of known sunflower absences as a key element of the final predictions, thereby decreasing commission error (false positives) The joint presence-absence model generated better predictions of sunflower distribution in the study region than did the presence-only model, but most accurate models made predictions of fewer locations within the study region, indicating a trade-off between accuracy and applicability Surprisingly, environmental data corresponding to the and during seedling emergence are most important in determining sunflower distributions These methods offer promise not only in modeling widespread species,but also in generating testable hypotheses about environmental factors determining their distributions.
Ecological niche modeling (ENM) is a useful tool for ecologists and evolutionary biologists Recently, these models have been used to locate potential populations of rare species (Raxworthy et al 2003), forecast changes in species’ distributions due to global climate change (Roura-Pascual et al 2004; Skov &
Svenning 2004; Thuiller et al 2005), predict the potential spread of invasive species (Peterson 2003; Peterson et al 2003), and anticipate areas of high disease risk
(Levine et al 2004) Various methodologies have been applied to ENM development, ranging from simple climate envelope models to much more complex evolutionary- computing approaches (Peterson 2001; Pearson & Dawson 2003) In general, these methods model species’ ecological requirements by combining known geographic occurrences of the species with environmental data for those locations, typically broad-scale climatic and topographic data The ENM can then be reprojected to predict the species’ geographic distribution One successful ENM tool is GARP (Genetic Algorithm for Rule-set Prediction), an evolutionary-computing program developed specifically for the ENM challenge (Stockwell & Peters 1999).
Modeling widespread species accurately has been a difficult challenge for ENM (Anderson et al 2003; Guisan & Hofer 2003; Segurado & Araújo 2004; Luoto et al 2005) The difficulty arises, in part, because the broad climatic tolerances of widespread species result in their occurrences being determined by environmental heterogeneity at spatial resolutions finer than the climatic data used in creating the finer-resolution environmental data from satellite imagery into their models (Roura- Pascual et al 2004) The use of such data sets should, by better capturing landscape heterogeneity, improve accuracy of distributional models of widespread species.
In general, two types of error are encountered in ENMs or any type of geographic predictions (Fielding & Bell 1997) First, omission errors occur when the model fails to predict locations where the species is known to occur (false negatives). Second, commission errors occur when the model predicts presence where the species does not occur (false positives) Modeling widespread species has been problematic, in particular because of the difficulty of interpreting high commission error
(Anderson et al 2003), which in fact may be simply broad distributional potential.
Three components contribute to commission error First, the model may predict the species as present in locations where it cannot survive, which reflects a poor quality model and represents real commission error (Peterson 2001) Second,commission “errors” may occur because organisms have restricted dispersal abilities(Peterson 2001; Anderson et al 2003), and “erroneous” presence predictions may occur in locations that represent suitable habitat, but are uninhabited because the species cannot physically get there Metapopulation dynamics may also result in suitable habitats being unoccupied at any single point in time (Hanski & Gilpin1991) Short of transplantation experiments, such commission “error” resulting from dispersal limitation cannot be easily factored out, although dispersal limitation is unlikely to be an issue for widespread species The third component of commission error results from under-sampling or non-detection, which again is a source of apparent commission error (Anderson et al 2003).
For generalist, widespread species, under-sampling is most likely to inflate commission errors in ENM, as ENMs are usually built using relatively few presence points In many ENM applications, it is standard to assume that the species is absent at locations where collections have not been made, and models are built based on these “pseudoabsences” (Stockwell & Noble 1992; Stockwell 1999) This assumption is likely to be robust for rare species or restricted-range species For generalist or widespread species, however, this assumption can be dangerous, as such species are likely to be common where data have not been collected Increasing sampling and incorporating data on known absences may reduce commission error due to under- sampling, providing a more accurate representation of the species’ ecological and geographic distribution.
Coarse-scale sampling of species’ distributions is time-consuming and expensive Plant species found in roadside habitats offer a unique opportunity for collecting both presence and absence data for ENM Roadsides represent transects across broad environmental gradients, and presences, absences, and coarse-level abundance can be assessed rapidly, at least for highly visible species Over 6.2 million km of roads exist in the United States alone (Forman & Alexander 1998) By providing habitat and facilitating dispersal, roads can act as a conduit for invasive species (Parendes & Jones; Gelbard & Belnap 2003; Pauchard & Alaback 2004), while in areas of intensive agriculture, roadside vegetation may also be valuable reservoirs of biological diversity (Forman & Alexander 1998).
In this study, I use GARP to model the generalist species, Helianthus annuus
L (wild sunflower, Asteraceae), based on roadside survey data covering a large portion of the species’ geographic range Helianthus annuus is common and highly visible along roadsides throughout the Great Plains of North America, the central portion of its native range Sunflowers appear to have expanded their range thanks to human activity (agriculture, building of railways and roads) and currently have a near-global distribution (Renửfọlt et al 2005) Sunflowers are not typically considered invasive, although they are exotic components of many floras and can be agricultural weeds (Bauer et al 1991; Humston et al 2005) Crop sunflowers (also Helianthus annuus) are cultivated worldwide for seed and oil Recent development of several transgenic lines has raised concerns about risks of crop/wild hybridization increasing the invasiveness of wild populations This risk is not trivial: crop and wild sunflowers commonly hybridize in nature, and crop-specific genes can persist in wild populations for multiple generations (Arias & Rieseberg 1994; Whitton et al 1997). Incorporation of the Bt transgene into wild sunflower populations increases seed production (Snow et al 2003), which could increase the size of the seed bank
(Chapter 1), resulting in increased prevalence across the landscape Understanding the regional ecology of sunflowers is thus an important component of assessing the risks of crop/wild hybridization and the spread of transgenes through wild populations.
My goal in this chapter was to generate an accurate ENM, predicting the potential geographic range, of the widespread plant species, Helianthus annuus, by incorporating both fine-resolution MODIS data and known locations of sunflower absences in their native range I also wanted to identify variables that may influence the regional distribution of H annuus, thereby generating new testable hypotheses regarding the species ecology and distribution.
Helianthus annuus is an annual plant that forms a persistent seed bank
(Alexander & Schrag 2003; Chapter 1) Seedlings emerge in the spring, between late March and early May Flowering begins in late July and continues through the autumn until there is a hard freeze (Chapter 1) Plants are quite showy as individual plants have multiple bright yellow inflorescences and are typically at least 1 m tall.
Roadside surveys were conducted between July 25 th and September 28 th ,
2003, over 5333 km across Kansas, Missouri, Nebraska, Iowa, and Oklahoma (Fig 3- 1a) These surveys cover a large portion of the species’ historical range and include areas of high and low occupancy at the county level Sunflower abundance was estimated for each 1.6 km segment on a logarithmic scale (0, 1-9, 10-99, 100-999, 1000+) on the passenger side of the road All routes were located along interstate and state highways A GPS unit (Garmin GPSmap 76) was used to track mileage along pre-programmed routes For this study, I used a subset of the total data set to examine the extreme abundance values, focusing on the strictest possible definitions of presence and absence: only observations of zero abundance were considered as absences, and only observations of 1000+ sunflowers were considered as presences under the assumption that areas with highest abundances best represent the ecological niche A total of 1131 segments fell into either the zero or 1000+ abundance categories, representing 1809.6 km.
I began model building with 28 environmental coverages; four topographic related coverages (compound topographic index, slope, aspect from the US