Back to Top

INTRODUCTION

Introductions of non-native species are increasing in frequency and extent as human movements become more global and international trade increases. This globalization has caused a breakdown in the regional distinctiveness of the Earth's flora and fauna. Humans have become the most important mechanism for plant dispersal and carry plants to areas where they would not arrive without anthropogenic means. Additionally, species known to have dispersed great distances without human aid have not become detrimental (Mack and Lonsdale 2001). Many non-native species are not harmful, but some can become the cause of serious ecological and economic degradation. Non-native species invasions decrease human economical wealth by impacting agricultural land, rangeland, and forests; alter ecosystem function; and threaten native biodiversity that is important for economic, environmental, and ethical reasons (Vitousek et al. 1997, Mack et al. 2000).

For example, Pimentel et al. (2000) estimate that the economic losses and environmental damage from invasive non-native species total $137 billion per year in the United States. This number and the magnitude of the associated ecological effects is difficult to calculate, but even if it is off by a factor of five or ten, the amount is still very large. The non-native nitrogen fixing plant, Myrica faya Ait., in Hawaii Volcanoes National Park has been shown to alter ecosystem properties by changing the nutrient dynamics of the system, perhaps allowing plants that would have been excluded due to limited nitrogen to enter the area and more successfully compete with native species that are better adapted to lower nitrogen availability (Vitousek et al. 1987, Vitousek 1990). Non-native species can also dramatically alter the fire regime of a system, as cheatgrass (Bromus tectorum L.) has done in the western United States (D'Antonio and Vitousek 1992). Leafy Spurge (Euphorbia esula L.), a major problem in areas such as Theodore Roosevelt National Park, North Dakota, is a non-native plant species that has become established in the Northern Great Plains. This plant displaces native species, thereby decreasing herbivore habitat, and causes economic losses to farmers and ranchers whose land it occupies. E. esula can out compete native species because of its extensive root system (Ringwall et al. 2000), grazing tolerance mechanisms (Olson and Wallander 1999), grazing avoidance mechanisms (Trammell and Butler 1995, Lym 1998), and possible allelopathic effects (Steenhagen and Zimdahl 1979). Many non-native species such as the latter two form monocultures that decrease biodiversity. The impact of non-native species introductions on our ability to conserve native species is clear; almost half of the threatened and endangered species listed under the Endangered Species Act are listed due to competition with or predation by non-native species (Pimentel et al. 2000).

Back to Top

Non-native species invasions have always occurred without human aid. However, the rate at which they are currently occurring has drastically increased because humans have increased their movements thereby increasing the translocation of species. Additionally, anthropogenic spread of non-natives allows species to spread to areas where they would not have spread by non-human means. The change in the rate of introductions exacerbates these problems because new introductions continue to occur before a system has time to adapt to previous introductions of non-native species. If the time between disturbance events decreases to an interval shorter than the time to recovery, then major components of the entire system may change. It is this increase in the rate of disturbances and the distance of non-native species introductions that is currently alarming many scientists.

Back to Top

PROBLEM STATEMENT

As the spread of non-native species becomes increasingly more common, it is important to have early detection methods. Land managers currently control the spread of invasive non-native plant species after they have already become established instead of trying to prevent the establishment of the species. Lists of probable invasive species for many public land units (e.g. national wildlife refuges) do not exist, and distribution maps are not available for many non-native species. Models do not currently exist that predict areas vulnerable to invasion. Therefore, managers do not know where to focus their precious funds and efforts to control the spread of non-native plant species and the control methods are reactive, not proactive.

Back to Top

OVERVIEW OF CURRENT KNOWLEDGE

Vulnerability of habitats to invasion

Various studies have focused on identifying factors that influence the vulnerability of systems to invasion. These studies have been conducted at different scales, have concentrated on different factors, and have found varying results (Appendix A). A common factor first proposed by Charles Elton (1958) is species richness. Some researchers have found a positive correlation between species diversity and susceptibility to invasion (Robinson et al. 1995, Planty-Tabacchi et al. 1996, Wiser et al. 1998, Smith and Knapp 1999, Stohlgren et al. 1999a), while others have found a negative correlation (Elton 1958, Fox and Fox 1986, Tilman 1997, Levine and D'Antonio 1999, Dukes 2001). The conflicting results may arise because these studies were conducted at different scales and in different communities.

Research has also evaluated the relationship between disturbance regimes and invasion vulnerability. Exogenous disturbances, or disturbance regimes altered from those historically found in an area, have been found to increase the vulnerability to invasion of an area (Hobbs and Huenneke 1992). Similarly, Fox and Fox (1986) determined that a departure from the endogenous disturbance regime is necessary for invasion to occur. Looking at fire in the Kanza Prairie, Smith and Knapp (1999, 2001) found that deviation from the natural fire regime increased vulnerability to invasion. They found a similar trend for grazing, but Stohlgren et al. (1999b) did not find any significant effect from grazing in the Central Grasslands at large spatial scales. In contrast, Larson et al. (2001) found that the most common invasive species in Theodore Roosevelt National Park did not follow disturbance patterns. This study only incorporated anthropogenic physical disturbance such as roadways, however.

Back to Top

In Glacier National Park, Tyser and Worley (1992) reported a correlation between non-native species and disturbance from roads and trails. Greenburg et al. (1997) found that roadways facilitated non-native species spread, especially when roadside material differed from the native soil type. In a global study that accounted for size differences (Lonsdale 1999) and a more local study (MacDonald et al. 1989), high visitation to reserves was positively correlated with high invasion rates.

Soil type has also been proposed as a factor related to vulnerability to invasion. In the serpentine grasslands of California, Harrison (1999) and Huenneke et al. (1990) determined that invasion was greater on soils with greater levels of nutrients. Stohlgren et al. (1999a, 1999b) found a correlation between soil percent nitrogen and non-native species richness and cover at various spatial scales in plots throughout the central grasslands and Colorado Rockies.

Certain community types have also been found to be more invasible than others. In several studies in different areas of the world and at different spatial scales, riparian areas were more invaded than upland sites (Planty-Tabacchi et al. 1996, Kotanen 1997, Stohlgren et al. 1998, Levine 2000, Larson et al. 2001). Similar studies have shown that rare habitats are more heavily invaded (Stohlgren et al. 1999a, Stohlgren et al. 2001).

These studies indicate that factors influencing the abundance and distribution of non-native species vary at different spatial scales. For example, soil nitrogen trends are found at a much smaller scale than anthropogenic disturbance. These factors that influence vulnerability to invasion can also be used to predict what areas will be most heavily invaded.

Back to Top

Modeling invasions

Higgins and Richardson (1996) review traditional models predicting non-native plant spread including simple-demographic models, spatial-phenomenological models, and spatial-mechanistic models to aid in determining which model type to use given specific data at hand. Simple demographic models are suitable when population density relates to the area invaded (density dependence) or when determining if a species will successfully establish. Spatial phenomenological models formulate predictions based on the past. Mechanistic models do not require an empirical precedent. Mechanistic models discussed include reaction-diffusion, metapopulation, and individual-based models. Metapopulation models are less demanding computatively, but individual-based models should be employed when patterns of environmental heterogeneity are not captured with a metapopulation model, fine-scale ecological heterogeneity is important in invasion success, and interactions among individuals (e.g., competition) is important. The model chosen to represent a specific invasion should be chosen based on the characteristics of that particular invasion.

Higgins et al. (1996) assert that reaction-diffusion (R-D) models are probably the most widely applied invasion models, but have generally been used to model animal invasions (Higgins and Richardson 1996). The assumptions of this model prevent use of interactions between plant attributes and the environment (e.g., higher biomass occurs with higher nutrient levels), therefore relegating environmental heterogeneity and stochasticity as unimportant. The authors propose the use of a spatially explicit individual-based simulation (SEIBS) model as an alternative that integrates space, ecological processes, and stochasticity. They compared the two models and determined that there is a qualitative and quantitative difference in the predictions.

Back to Top

Higgins et al. (2000) scaled up a SEIBS model to create a spatially explicit landscape extent simulation (SELES) model to predict the rate of spread of non-native plant species. This model used the predicted probability of occurrence of a specific species from previous work (Higgins et al. 1999) to predict the spread of these occurrences. The native plant data were collected at a coarser scale (1 km2) than the grain chosen for the model (200 m), so they converted the data by assuming that if a species was present at 1 km2, then it would be present in each of the 25 squares (200 m x 200 m). The potential non-native plant distribution was based on a logistic regression model of environmental preferences (Higgins et al. 1999). This variable was used along with plant dispersal, mortality recruitment, and control efforts to create the SELES model. A comparison of the two models, SEIBS and SELES, did not reveal scaling artifacts.

Wadsworth et al. (2000) modeled the spread of two non-native plant species in England using MIGRATE, a spatially explicit model. They used 13 different reproductive and dispersal parameters in the model. Six different management strategies were simulated to determine the effectiveness of control strategies.

Back to Top

Peterson and Vieglias (2001) used the Genetic Algorithm for Rule-set Prediction, or GARP, to model species distributions. This model used ecological niches to predict invasion by a specific non-native species. The program ran many iterations of multiple regressions to predict presence/ absence or intersection of ranges and compared these results to sets of resampled points from known occurrences. From these analyses, a set of five to 50 rules that together defined the species' ecological niche were obtained. These rules were then used to define the species distribution. Variables included in the model were vegetation type, slope, aspect, elevation, and soil type. Peterson and Vieglias based their model on the concept that 'ecological niches are stable and determine the set of possible conditions under which a species is able to invade a particular region'. Most data used for the model were derived from data sets obtained from museum collections. The model provided information on the distribution of species; no information on abundance or cover was provided.

All of the above studies model the distribution of a single species, predicting its presence or absence. None look at the abundance or cover of a single species or attempt to model the total number of non-native species distributed across an area (non-native species richness). These models do not attempt to identify hot spots of invasion or areas vulnerable to invasion.

Back to Top

Iverson and Prasad (1998) developed a new methodology, modeling native plant species richness in less sampled areas using data from better sampled areas. They took a county level dataset for Illinois and used stepwise regression to develop a model using variables from environmental and topographic variables from GIS. They then examined the response surface for possible spatial autocorrelation using Moran's I and Geary's C. In this way they would have included small-scale variability from neighbor interactions in addition to the large scale trend, but they did not find spatial autocorrelation in the response surface.

However, a spatial modeling process described by Reich and Bravo (1998) make use of a unique combination of spatial statistical methods to model landscape structure. The input data includes field data, remotely sensed imagery, and GIS data. Data is extracted from the GIS layers and images for each field point. A regression model can then be used to describe the large-scale variability in the data. If the residuals contain spatial autocorrelation a model can be developed to describe the small-scale variability in the data using kriging or co-kriging. Finally, the two surfaces (the large scale model and the co-kriged residuals) are combined to create a final trend surface model that best describes the spatial distribution of non-native species richness or the spatial distribution of an individual non-native species.
Kalkhan et al. (2000a, 2000b) and Chong et al (2001) used these methods to predict the distribution, presence and pattern of native and exotic plant species and soil characteristics. The spatial statistical models produced had higher R2 values than the simple regression models. The use of full-coverage, fine-scale data (e.g., Landsat TM data) ameliorates the problems associated with the other types of models arising from lack of empirical data and scale issues. These methods also allow for the development of maps of uncertainty based on subsampling. This surface provides land managers or others using the data with spatially represented confidence for the model, and also can direct further sampling efforts to areas with low confidence.

Back to Top

These methods integrating field data, remotely sensed data, GIS, and spatial statistical techniques provide a method that captures the patterns of species distributions. They also require less intensive data collection than individual based models reliant on demographic data. The predictions can cover large areas without losing resolution. For determining species distribution, abundance and patterns, these models appear the most appropriate for high accuracy without extensive (and cost prohibitive) field sampling.

Back to Top

RESEARCH HYPOTHESES

Primary objective:

Synthesize data from disparate sources into a geodatabase to: (1) illustrate the utility of data synergy; (2) test current modeling methodology; (3) model the probable abundance of non-native species and community vulnerability to invasion in Colorado; and (4) test the same modeling methods for modeling a single species distribution and make this information available to land managers. These four objectives can be divided into the following components:

METHODS

The study site area is comprised by the state of Colorado. It has a unique combination of ecosystems ranging from short grass steppe to tundra. For this project, non-native species will be defined as species that have been anthropogenically introduced to North America (either intentionally or accidentally) or Colorado. The USDA Natural Resource Conservation Service PLANTS database will be used as the authority for species origin. Datasets detailing non-native species distribution in the United States will be collected from different agencies and organizations (Table 1). These datasets have been compiled from fieldwork with different objectives using different methods and have different types of errors associated with them. However, these datasets will be more powerful when combined (provide more information than when separate) and will provide baseline information on non-native species, indicating overall gaps in knowledge. The datasets will be synthesized into a geodatabase created with ESRI's ArcGIS and Microsoft's SQL Server software (ArcGIS 2002, SQL Server 2003). Associated metadata will be developed. The datasets will be synthesized in general relational tables that will accommodate various data types. The TSN (Taxonomic Serial Number) codes from ITIS (the Integrated Taxonomic Information System) will by used to standardize species names as different projects include synonyms for the same species. The different datasets will be linked through nested locations of species. For example, a point may be located within a wildlife refuge which may be located within a county.

Back to Top

Table 1. Sources of datasets to be used.

Source	Dataset
Biota of North America Program	Non-native plant species by county for the U.S.
Forest Health Monitoring Program	Point data (presence and cover) from USFS lands
FSVeg	Plot data
NPSpecies	Non-native species lists for National Park units
Graduate Students	Multi-scale plot data, point data
Research Scientists	Multi-scale plot data, other plots
County Weed Agencies	Point, line, and polygon and control data
San Luis Valley GIS Authority	Point, line, and polygon data
The Nature Conservancy	Point, line, and polygon data
Colorado Natural Heritage Program	Non-native plant point data
State of Colorado	Quarter quad data for at least 20 non-native species
U.S. Fish and Wildlife Service	Survey response from refuges
LTER Sites	Plot data and species lists
Bureau of Land Management	Point, line, polygon, and control data
USGS (Midcontinent Ecological Science Center) and the Natural Resource Ecology Laboratory (NREL) at CSU	Modified Whittaker plot data (presence and cover)

Back to Top

The first step will be to illustrate the utility of data synergy. I will use the BONAP dataset as a baseline. This dataset is a national county level dataset of species lists compiled from herbarium records. It was chosen as the baseline because it represents the largest spatial scale, is comprehensive, and is consistent because all of this data was collected in the same manner. Four other county level species lists will be generated from the other data types which are nested within counties. These data types will include: (1) plots, (2) GIS data including points, lines and polygons, (3) quarter quad survey data, and (4) area species lists (e.g., National Park species lists). Each dataset will differ in search intensity per county. The generated lists will be compared to the baseline county level species lists. I will examine how the various biases in each individual dataset can be ameliorated by using the combined data sets. Species accumulation curves will be used to determine which type of dataset adds more information to the baseline data.

The other components of analysis will include modeling techniques. Plot data will be used that can be reasonably extrapolated to a 30m x 30m grid of the state. For example, Modified Whittaker plots measure 20m x 50m, so their species richness could be extrapolated to a 30m x 30m grid. However, Daubenmire plots are only 0.25m x 0.25m, so extrapolating information from them to 30m x 30m cannot be assumed to be accurate for either the total number of non-native species or the absence of individual species. The small plot size would result in overlooking some species due to patchy species distributions.

Various environmental and anthropogenic independent variables will be derived from available GIS coverages of the state to use as covariates in the analysis (Table 2). The proposed variables were chosen based on an extensive review of non-native species invasion literature (Appendix A). Elevation will be interpolated with ARC/INFO (2002) from a digital elevation model. Slope and aspect (degrees from due south; 0 to 180) will be derived from the digital elevation model using ARC/INFO (2002) software functions. Other variables will be derived from available GIS coverages and Landsat Thematic Mapper imagery. GRIDS of the Euclidean distance from features (e.g., water, roads) will be created from GIS shapefiles using the map calculator function in ArcView GIS 3.2 (1999). Soil variables will be derived using the USDA State Soil Geographic Database (STATSGO) that has been synthesized by the Earth Systems Science Center into GIS coverages. TM Bands, NDVI, TNDVI, and Tassel Cap will be obtained using ERDAS IMAGINE software. Data will be extracted from these coverages for each grid cell with field data for the modeling process. Full coverages will be used in creating the final surface.

Back to Top

Table 2. Proposed independent variables.

* = Categorical variable; ** = Continuous variable.

Back to Top

MMoran's I will be calculated using Splus software for all variables to test for spatial autocorrelation, and the cross-correlation statistic will be calculated for all pairwise combinations of variables to examine the spatial dependency of the data across the landscape (Cliff and Ord 1981). This function will be preformed at multiple scales with the data by using different band widths to calculate the inverse spatial weights matrix, which uses distances as a measure of the spatial proximity between sample points. This matrix gives points farther away from the point of interest greater weight.

I propose to create a spatially explicit model of probable non-native species richness using trend surface analysis and kriging, co-kriging or regression tree methods to capture both large and small-scale variability (Figure 1). The models will be created using Splus and final predicted surfaces will be created in Arc/Info (2002). By using these models, the problems caused by spatial autocorrelation in the data by using classical statistics will be avoided and the natural spatial dependency will be taken into account. Inputs to the trend surface analysis will include the inverse weights matrix and covariates selected using either combinatorial screening, which compares all possible pairwise combinations of variables, for small numbers of possible covariates or stepwise selection for large numbers of covariates. An ordinary least squares regression will be used to create the trend surface.
If the residuals from the trend surface analysis are autocorrelated, then regression tree, kriging or co-kriging will be performed. If cross-correlation is also present, then co-kriging will be used; otherwise, one of the other methods will be used. Both kriging and co-kriging are interpolation methods. For co-kriging, primary (dependent) variables for individual models will include the residuals from the trend surface analysis. Secondary (independent) variables will include variables exhibiting minimally significant cross-correlation with the dependent variable of each model and will be selected by choosing the combination of variables that minimizes Akaike's Information Criteria. A variogram will be estimated for each of the dependent variables and the number of nearest neighbors will be chosen by minimizing the error and maximizing the R2 value using cross validation.

Back to Top

Figure 1. Steps for model development.

Back to Top

If there is no autocorrelation present in the trend surface residuals, regression tree analysis will be used. This method splits data into groups by using a binary partitioning algorithm that maximizes the differences between groups. The groups continue to be split until they are similar or there are less than 5 observations per group. The final surface will be created by adding together the trend surface and the grid capturing the small-scale variability.

The first model analysis will include a comparison of different models created for Rocky Mountain National Park's non-native species richness. Generally, these spatial statistical models include field data, GIS data, and remotely sensed data. I will create three models for the Park. The first will include all three types of data (field data, GIS data, and remotely sensed imagery) from one year. The second model will include the same data as the first, but will use average pixel values from remotely sensed data from each year that field surveys were conducted. The third model will include data from only one year of remotely sensed imagery. The three models will be compared for significant differences to determine the benefit of including remotely sensed parameters into the model.

I will also attempt to identify hot spots of invasion and will predict habitats vulnerable to invasion using my geodatabase and high-resolution GIS vegetation coverages for the state of Colorado. I will not include remote sensing information in the statewide model due to the high cost and computing power required. Certain areas of the state may exhibit a high error because of the scarcity of field data available. Hot spots of invasion will be determined from the combined trend surface and co-kriged surface described above. These areas will be those that have the highest predicted number of non-native species present (e,g, the highest non-native species richness). Habitats vulnerable to invasion will also be determined using the predicted surface created above. Analysis of Variance (ANOVA) will be used to statistically compare the predicted number of non-native species (determined above) between community types (as defined by available GIS coverages) associated with each predicted point.

I will also choose a non-native species to test a spatial statistical model for an individual species. The species will be chosen based on data availability for individual species. Presence data will be derived from all data types for each grid cell, whereas absence data will be taken only from the plots that are used in other models. So, absence will only be assumed from plots large enough to represent an entire grid cell.

The model validity for all above models will be tested using presence/ absence data from other datasets not included in the model development (e.g., area species lists where the areas are much larger than the specified grid cell size). Also, a 10-fold cross validation technique will be used to validate models. Here, the model will be developed using a random 90% of the data and tested against the remaining 10%. This process will be repeated 10 times.

As a final product, I will make the database and models available to land managers and the public on the web. An individual will be able to choose a species and get the county level distribution in Colorado or click on a county or public land unit (national park, national wildlife refuge, etc.) and obtain a species list with probabilities for each species and the vulnerability of the area to invasion.

Back to Top

APPENDIX A

Study	Type	Scale	Results
(Bergelson et al. 1993)	Experiment	Multiple	Rates of spread of a weed depends changes with different spatial patterns of disturbance
(Burke and Grime 1996)	Experiment	Local	Limestone grassland in United Kingdom. Most invasible with high levels of disturbance. Even more invaded when disturbance associated with eutrification.
(Dukes 2001)	Experiment	Local	Grassland community microcosms invisibility by Centaurea solstitialis. High functional diversity decreased success. Species diversity not correlated with invasibility. Species diversity positively correlated with persistence of species after invasion
(Fox and Fox 1986)	Review	None	Invasion occurs with at least minimal alteration of endogenous disturbance regimes
(Greenburg et al. 1997)	Observational	Local	Compared invasion along roadsides with native and non-native soils and clearcuts to hold disturbance constant in xeric Florida sand pine scrub. Modified soil type was more invaded than native, even with disturbance. Roadways facilitate non-native propagule transport.
(Harrison 1999)	Observational	Local	California serpentine and non-serpentine grasslands compared indicate that invasion is less in poor soils (but whether because poor soils are less invasible or because current non-natives are unsuited is unknown).
(Hobbs and Huenneke 1992)	Review	None	Exogenous disturbance such as fire, grazing, and soil disturbance outside of the natural cycle (either greater or less) leads to increased invisibility
(Huenneke et al. 1990)	Experiment	Local	Applied plant nutrients (N and P) to plots in a Californian annual serpentine grassland. Grazing exclusion resulted in decreased native forb presence. Adding nutrients resulted in increased invasion (reduced total richness, increased non-native dominance).
(Larson et al. 2001)	Observational	Regional	Mixed grass prairie (THRO, both units, 11 plant communities). Vegetation type explained more variance in invasibility than disturbance or park unit. Three way model with vegetation type, park unit, and disturbance best explains frequency and number of non-natives in aggregate- perhaps a reflection of nitrogen and water availability differences. Riparian zones are more invaded. Different species respond differently to disturbance and to vegetation type.
(Levine 2000)	Experiment	Local	California riparian area diversity is positively correlated with invisibility.
(Levine and D'Antonio 1999)	Review		Generally, species rich communities are more invasible than species-poor communities.
(Lodge 1993)	None	None	None
(Lonsdale 1999)	Observational	Global	Variances in non-native species number explained by number of native species, island or mainland status, and reserve or not reserve status. On the mainland degree of invasion increased with latitude. In reserves number of visitors also explained number of non-native species.
(MacDonald et al. 1989)	None	None	North American nature reserves
(Planty-Tabacchi et al. 1996)	Observational	Watershed and patch	Studied Adour River in southwestern France, McKenzie River in Oregon, and the Dungeness and Hoh Rivers in Washington. The McKenzie and Adour similarities indicated that invasion along riparian areas can be predicted by species richness. At the patch scale, riparian zones are more invaded than upland zones.
(Rejmanek and Richardson 1996)	Review	Global	Island isolation from the mainland is correlated with their invasibility. Species rich islands are more invaded, but could result because they are more suited to agriculture.
(Robinson et al. 1995)	Experiment	Community	California winter annual grassland with plots 2m2 to 32m2. Invaded plots typically had greater disturbance (trampled vegetation, bare soil), lowers levels of dominance, and higher species richness. Plot size was not significant. However, 70% of the variation was not explained.
(Smith and Knapp 1999)	Experiment	~60ha	Kanza Prairie, C4 dominated grassland. Frequent fire reduced the number of non-natives found (e.g., invasibility). Non-native richness negatively correlated with ANPP and aboveground grass production. Grazing treatments both showed a positive correlation between native and non-native species richness.
(Smith and Knapp 2001)	Experiment	Local	Northeastern Kansas Kanza Prairie grassland. Fire reduced invasibility. Size of the local pool of non-native species was positively correlated with non-native species, and appears to counter the negative effect of fire on number of non-natives.
(Stohlgren et al. 1998)	Observational	Multiple	Central grassland riparian versus upland areas with similar grazing intensities. 1m2 scale no patterns. 1000m2 scale showed riparian zones had higher native and non-native richness than upland sites. Landscape scale:
(Stohlgren et al. 1999a)	Observational	Multiple	Rocky Mountains and central grasslands. At the 1m2 scale, non-native species richness was negatively correlated with plant species richness and cover. At the 1000m2 scale, non-native species richness correlated with soil % N and native cover. At the landscape scale, non-native species cover correlated to total foliar cover, mean soil % N, and total number of non-native species.
(Stohlgren et al. 1999b)	Observational	Multiple	Central grassland grazed and grazing exclosure areas. At 1m2 and 1000m2 scales across all areas, grazed and ungrazed plots did not differ significantly in non-native species richness. Elevation was strongly correlated with non-native diversity. Non-native species richness was positively correlated with soil % N.
(Tyser and Worley 1992)	Observational	Community	Glacier National Park invasion of roadsides and trails in the backcountry. Non-natives decrease with distance from roadsides. Trails exhibited a similar trend, though the effect extended less distance.

Back to Top

---

REFERENCES

ArcGIS. 2002. ArcGIS. Environmental Systems Research Institute, Redlands, California.
ArcView GIS 3.2. 1999. ArcView GIS 3.2. Environmental Systems Research Institute, Redland, California.
Bergelson, J., J. A. Newman, and E. M. Floresrous. 1993. Rates of weed spread in spatially heterogeneous environments. Ecology 74:999-1011.
Burke, M. J. W., and J. P. Grime. 1996. An experimental study of plant community invasibility. Ecology 77:776-790.
Chong, G. W., R. M. Reich, M. A. Kalkhan, and T. J. Stohlgren. 2001. New approaches for sampling and modeling native and exotic plant species richness. Western North American Naturalist 61:328-335.
D'Antonio, C. M., and P. M. Vitousek. 1992. Biological invasions by exotic grasses, the grass/fire cycle, and global change. Annual Review of Ecology and Systematics 23:63-87.
Dukes, J. S. 2001. Biodiversity and invasibility in grassland microcosms. Oecologia 126:563-568.
Elton, C. 1958. The ecology of invasions by plants and animals. Meuthuen and Company, LTD, London.
Fox, M. D., and B. J. Fox. 1986. The susceptibility of natural communities to invasion. Pages 57-66 in R. H. Groves and J. J. Burdon, editors. Ecology of biological invasions: and Australian perspective. Cambridge University Press, Cambridge.
Greenburg, C. H., S. H. Crownover, and D. R. Gordon. 1997. Roadside soils: a corridor for invasion of xeric scrub by nonindigenous plants. Natural Areas Journal 17:99-109.
Harrison, S. 1999. Local and regional diversity in a patchy landscape: native, alien, and endemic herbs on serpentine. Ecology 80:70-80.
Higgins, S. I., and D. M. Richardson. 1996. A review of models of alien plant spread. Ecological Modelling 87:249-265.
Higgins, S. I., D. M. Richardson, and R. M. Cowling. 1996. Modeling invasive plant spread: the role of plant-environment interactions and model structure. Ecology 77:2043-2054.
Higgins, S. I., D. M. Richardson, and R. M. Cowling. 2000. Using a dynamic landscape model for planning the management of alien plant invasions. Ecological Applications 10:1833-1848.
Higgins, S. I., D. M. Richardson, R. M. Cowling, and T. H. Trinder-Smith. 1999. Predicting the landscape-scale distribution of alien plants and their threat to plant diversity. Conservation Biology 13:303-313.
Hobbs, R. J., and L. F. Huenneke. 1992. Disturbance, diversity, and invasion: implications for conservation. Conservation Biology 6:324-337.
Huenneke, L. F., S. P. Hamburg, R. Koide, H. A. Mooney, and P. M. Vitousek. 1990. Effects of soil resources on plant invasion and community structure in Californian serpentine grassland. Ecology 71:478-491.
Iverson, L. R., and A. Prasad. 1998. Estimating regional plant biodiversity with GIS modelling. Diversity and Distributions 4:49-61.
Kalkhan, M. A., G. W. Chong, R. M. Reich, and T. J. Stohlgren. 2000a. Landscape-scale assessment of mountain plant diversity: integration of remotely sensed data, GIS, and spatial statistics. in ASPRS 2000 Proceedings: Start the 21st century: launching the geospatial information age, Washington, D.C.
Kalkhan, M. A., T. J. Stohlgren, G. W. Chong, L. D. Schell, and R. M. Reich. 2000b. A predictive spatial model of plant diversity: integration of remotely sensed data, GIS, and spatial statistics. in J. D. Greer, editor. Eighth Biennial Remote Sensing Applications Conference: Remote Sensing and Geospatial Technologies for the New Millennium. American Society for Photogrammetry and Remote Sensing, Albuquerque, NM.
Kotanen, P. M. 1997. Effects of experimental soil disturbance on revegetation by natives and exotics in coastal Californian meadows. Journal of Applied Ecology 34:631-644.
Larson, D. L., P. J. Anderson, and W. Newton. 2001. Alien plant invasion in mixed-grass prairie: Effects of vegetation type and anthropogenic disturbance. Ecological Applications 11:128-141.
Levine, J. M. 2000. Species diversity and biological invasions: relating local process to community pattern. Science 288:852-854.
Levine, J. M., and C. D'Antonio. 1999. Elton revisited: a review of evidence linking diversity and invasibility. Oikos 87:15-26.
Lodge, D. M. 1993. Biological invasions: lessons for ecology. Tree 8:133-137.
Lonsdale, W. M. 1999. Global patterns of plant invasions and the concept of invasibility. Ecology 80:1522-1536.
Lym, R. G. 1998. The biology and integrated management of leafy spurge (Euphorbia esula) on North Dakota rangeland. Weed Technology 12:367-373.
MacDonald, I. A., L. L. Loope, M. B. Usher, and O. Hamann. 1989. Wildlife conservation and the invasion of nature reserves by introduced species: a global perspective. Pages 215-255 in J. A. Drake, H. A. Mooney, F. di Castri, R. H. Groves, F. J. Kruger, M. Rejmanek, and M. Williamson, editors. Biological invasions: a global perspective. John Wiley and Sons, New York.
Mack, R. N., and W. M. Lonsdale. 2001. Humans as global plant dispersers: getting more than we bargained for. BioScience 51:95-102.
Mack, R. N., D. Simberloff, W. M. Lonsdale, H. Evans, M. Clout, and F. A. Bazzaz. 2000. Biotic invasions: Causes, epidemiology, global consequences, and control. Ecological Applications 10:689-710.
Olson, B. E., and R. T. Wallander. 1999. Carbon allocation in Euphorbia esula and neighbors after defoliation. Canadian Journal of Botany 77:1641-1647.
Peterson, A. T., and D. A. Vieglais. 2001. Predicting species invasion using ecological niche modeling: new approaches from bioinformatics attack a pressing problem. BioScience 51:363-371.
Pimentel, D., L. Lach, R. Zuniga, and D. Morrison. 2000. Environmental and economic costs of nonindigenous species in the United States. BioScience 50:53-65.
Planty-Tabacchi, A.-M., E. Tabacchi, R. J. Naiman, C. DeFerrari, and H. Decamps. 1996. Invasibility of species-rich communities in riparian zones. Conservation Biology 10:598-607.
Reich, R. M., and V. A. Bravo. 1998. Integrating spatial statistics with GIS and remote sensing in designing multiresource inventories. Pages 202-207 in North America Science Symposium: Toward a unified framework for inventorying and monitorin forest ecosystem resources. USDA Rocky Mountain Research Station Proceedings, RMRS-P-12, Guadalajara, Jalisco, Mexico.
Rejmanek, M., and D. M. Richardson. 1996. What attributes make some plant species more invasive? Ecology 77:1655-1661.
Ringwall, K., M. Biondini, and C. Grugiel. 2000. Effects of nitrogen fertilization in leafy spurge root architecture. Journal of Range Management 53:228-232.
Robinson, G. R., J. F. Quinn, and M. L. Stanton. 1995. Invasibility of experimental habitat islands in a California winter annual grassland. Ecology 76:786-794.
Smith, M. D., and A. K. Knapp. 1999. Exotic plant species in a C-4-dominated grassland: invasibility, disturbance, and community structure. Oecologia 120:605-612.
Smith, M. D., and A. K. Knapp. 2001. Size of the local species pool determines invasibility of a C-4-dominated grassland. Oikos 92:55-61.
SQL Server. 2003. SQL Server. Microsoft Corporation, Redmond, WA.
Steenhagen, D. A., and R. L. Zimdahl. 1979. Allelopathy of leafy spurge (Euphorbia esula L.): breeding system analysis. Canadian Journal of Botany 77:1684-1688.
Stohlgren, T. J., M. Lee, K. A. Bull, Y. Otsuki, and C. A. Villa. 1998. Riparian zones as havens for exotic plant species in the central grasslands. Plant Ecology 138:113-125.
Stohlgren, T. J., Y. Otsuki, C. A. Villa, M. Lee, and J. Belnap. 2001. Patterns of plant invasions: a case example in native plant species hotspots and rare habitats. Biological Invasions 3:37-50.
Stohlgren, T. J., L. D. Schell, K. A. Bull, Y. Otsuki, G. Newman, M. Bashkin, Y. Son, D. Binkley, G. W. Chong, and M. A. Kalkhan. 1999a. Exotic plant species invade hot spots of native plant diversity. Ecological Monographs 69:25-46.
Stohlgren, T. J., L. D. Schell, and B. Vanden Heuvel. 1999b. How grazing and soil quality affect native and exotic plant diversity in rocky mountain grasslands. Ecological Applications 9:45-64.
Tilman, D. 1997. Community invasibility, recruitment limitation and grassland biodiversity. Ecology 78:81-92.
Trammell, M. A., and J. L. Butler. 1995. Effects of exotic plants on native ungulate use of habitat. Journal of Wildlife Management 59:808-816.
Tyser, R. W., and C. A. Worley. 1992. Alien flora in grasslands adjacent to road and trail corridors in Glacier National Park, Montana (U.S.A.). Conservation Biology 6:253-262.
Vitousek, P. M. 1990. Biological invasions and ecosystem processes: towards an integration of population biology and ecosystem studies. Oikos 57:7-13.
Vitousek, P. M., C. M. D'Antonio, L. L. Loope, M. Rejmanek, and R. Westbrooks. 1997. Introduced species: a significant component of human caused global change. New Zealand Journal of Ecology 21:1-16.
Vitousek, P. M., L. R. Walker, L. D. Whiteaker, D. Mueller-Dombois, and P. A. Matson. 1987. Biological invasion by Myrica faya alters ecosystem development in Hawaii. Science 238:802-804.
Wadsworth, R. A., Y. C. Collingham, S. G. Willis, B. Huntley, and P. E. Hulme. 2000. Simulating the spread and management of alien riparian weeds: are they out of control? Journal of Applied Ecology 37:28-38.
Wiser, S. K., R. B. Allen, P. W. Clinton, and K. H. Platt. 1998. Community structure and forest invasion by an exotic herb over 23 years. Ecology 79:2071-208

Back to Top