Biological Invasions and GIS

An Annotated Bibliography by Angela Brandt

GEO 565, Winter 2006, Oregon State University

Exotic invasive species represent one of the greatest threats to global biodiversity, as well as causing serious economic impacts. I am interested in applying community ecology theory to invasion and restoration ecology to understand the mechanisms driving invasion and continued dominance by exotic species. Better understanding of these mechanisms will lead to more efficient and successful restoration of invaded communities. Geographic information is an important component of studying communities and invasion, as well as for informing management decisions. Spatial and temporal variability are two important factors that can lead to coexistence of species within a community. Spatial data are crucial for documenting and predicting the spread of exotic species, control measures and their effectiveness, and correlating these variables with other environmental factors. A geographic information system (GIS) is an invaluable tool for compiling and analyzing such data. Below is a sample of studies that have used a GIS to investigate biological invasions.

Comparing distributions of invasive and non-invasive exotic species may contribute to understanding invasive success and doing so at a large scale may be useful for determining general trends. This study examined the regional distributions of invasive and non-invasive exotic plant species in California to determine relationships between these distributions and both environmental and anthropogenic factors. Invasive and non-invasive exotic species richness in each bioregion was compared to native species richness, road density, elevation, population density, precipitation, and area. A GIS was used to map these variables and obtain estimates of road density, population density, and area per bioregion. Two regression analyses were performed--a multiple linear regression and a spatial autoregressive model--to determine the impact of spatial dependence on a regression model. Coastal areas generally had the highest concentration of exotic species and invasive and non-invasive exotics had similar distributions, though annual grasses were underrepresented in the dataset. This result suggests that other plant attributes or factors on a smaller-scale determine whether a plant becomes invasive. In the ordinary regression model, invasive exotic richness was positively correlated with native richness and negatively correlated with elevation, while non-invasive exotic richness was positively correlated with native richness and road density and negatively correlated with elevation. Data in both models were determined to be spatially dependent. In the spatial autoregressive model, both invasive and non-invasive exotic richness were positively correlated with native richness and road density and negatively correlated with elevation. Accounting for spatial dependence was thus shown to be important in determining significant relationships between environmental and anthropogenic factors and exotic species distributions. This study's results support many previous findings of positive correlation between exotic species richness and both native species richness and anthropogenic disturbance.

Exotic dreissenid mussels can inhabit both hard and soft substrates. As the majority of Lake Erie's floor is composed of soft substrates and mussel colonization modifies these benthic habitats, the ability to analyze and predict the abundance of exotic mussels over space and time is important to understanding their ecosystem impacts. This study examined the spread of dreissenid mussels in Lake Erie over time in relation to water depth and substrate type. GIS analysis of bathymetry, substrate, and percent cover of dreissenid mussels obtained via side-scan sonar in two of Lake Erie's three basins from 1994 to 1998 was used to produce regression models that could predict the future percent cover of mussels. Models produced with data from 1994 to 1996 and 1994 to 1997 were able to predict observed cover of dreissenid mussels in 1997 and 1998, respectively, at least 80% of the time. The models produced were therefore useful for predicting the future spread of these exotic mussels. Further, these models can be applied to Lake Erie's third basin when dreissenid mussel cover data become available and the methods used to produce the models can be applied to other large lake systems.

Sampling methods, such as scale, are critical to accurately determine the factors influencing species richness using spatial statistics. Sampling at multiple scales may thus improve sampling efficiency and effectiveness. This study examined the effect of sampling at multiple scales on determining the relationships between environmental factors and plant species richness with spatial cross-correlation analysis. Random sampling locations were chosen within each of five vegetation cover types in the Beaver Meadows study area within Rocky Mountain National Park that had been delineated using aerial photos and GIS. Slope, aspect, and elevation were determined for each sampling location using GIS and soil samples were taken to determine soil composition and nutrient levels. Plots at four different scales (1-1000 square meters) were established in a nested design at each location and the species present in each plot and subplot recorded. Vegetative cover and height data were recorded in the smallest subplots. Most variables showed some type of spatial correlation at both the largest and smallest scale, though the type and significance of these correlations often differed at the various scales. Spatial autocorrelation was universal among variables at the smallest scale. Species distribution and richness were influenced more by topography at the large scale and more by species distributions at the small scale. The most useful information for exotic species management was obtained with small scale sampling, where native richness was observed to increase with soil clay content and exotic richness to increase with native richness and soil nitrogen, silt, and clay content. This suggests that managers need to focus control efforts on wetland areas and fertile soils with highest native species richness.

The usefulness of remote sensing and GIS in invasive species management is becoming increasingly recognized. With the large variety of remote sensing data available, land managers need to understand the differences in their capabilities to use them efficiently. This article provides a summary of the characteristics of different remote sensing technologies and how to determine which is the most appropriate for a given invasive species mapping project, as well as a list of projects in the southeastern U.S. that have used remote sensing. The four characteristics land managers should consider when selecting remote sensing data to use are (1) the ability to conduct ground-truthing to obtain the accuracy level required by the project, (2) spatial resolution, (3) spectral resolution, and (4) cost. Projects covering large areas can generally use satellite imagery, while projects requiring higher spatial resolution may need to use aerial photographs or high-resolution satellites. When scanning hardcopy photographs, the manager must consider both the resolution of the photograph itself and the scanning resolution. Remote sensing technology offers a wide range of spectral resolutions, some of which can distinguish certain types of vegetation better than others. Generally, multispectral and cover-infrared photographs are most suitable for distinguishing vegetation types. The cost of the project depends upon the technology used and the area needing to be covered. If large areas must be covered, a satellite image may be more cost-effective than many aerial photographs, for example. By understanding the characteristics of the remote sensing technologies available, land managers can make the best use of these for mapping invasive plant species.

The characteristics that make a species a good invader and a community more or less invasible are in debate, thus general predictions of the potential spread of invasive species are difficult to make. Further, the scale at which community invasibility is examined may affect a study's results, particularly regarding the relationship between invasive and native species richness. This study examined the relationship of native and invasive fish species' distributions in California to environmental variables and anthropogenic disturbance at the scale of watersheds. A GIS was used to determine the environmental characteristics and levels of both aquatic and terrestrial anthropogenic disturbance in each watershed and the factors significantly associated with fish species richness were determined. Several models based on hypotheses from the literature regarding the effects of these factors on invasive fish species richness were constructed and ranked based on how well they explained the data. Invasive species tended to be more strongly associated with anthropogenic disturbance than natives, but certain types of aquatic disturbance, such as the presence of dams, explained richness of stocked species better than other invaders. Stocked fish species were also more widespread than nonstocked invasive species. Invasive fish species richness was positively correlated with native fish species richness. Overall, this study concluded that anthropogenic disturbance contributes to fish invasion, stocked species are more likely to succeed at invasion, and current protected areas showed no correlation with invasive fish species richness. Different tactics may thus be necessary to more successfully combat fish invasions, such as restoring natural habitat characteristics.

Maps depicting the distribution, density, and characteristics of exotic plant stands within a managed area, such as a national park, are useful for land managers to focus control efforts and determine the effectiveness of those efforts. This article describes the steps in a mapping project of Melaleuca quinquenervia, an exotic invasive tree species, in a section of Everglades National Park. Large-scale color infrared aerial photographs were used to classify vegetation types in the study area. Stands of M. quinquenervia were additionally classified by density and height. Six sites infested with M. quinquenervia were characterized on the ground prior to photograph interpretation to increase accuracy. The vegetation classification overlays were scanned, converted from raster to vector format, and geocoded. Ground-truthing was performed after the final maps were produced to determine their accuracy, which was greater than 85% for all types of classification. The final map suggested that the eastern section of the study area contained an older infestation, possibly with a seed source outside the park, while the western section contained a potentially expanding infestation. The map also showed the locations of 162 seed trees, which provide a large seed source for expanding infested areas. These results provide park staff with important knowledge of where M. quinquenervia control is most needed and will likely produce the greatest results.

Quantitative risk assessment of the establishment potential of exotic plant pests is in growing demand, particularly in an increasingly global economy. Several methods for estimating the risk of establishment are currently used but may be limited by their inability to consider spatial and temporal variability of environmental factors. Here, a spatial simulation method was used to predict the area that would be affected by an introduction of Ralstonia solanacearum, a well-studied bacterial species that infects potatoes, in Norway, where it was not yet found. The pest organism's environmental requirements and the new region's environmental characteristics are the two general predictors for establishment potential, but the pest's method of spread is also important to predict the overall area it would affect. In this analysis, the host plant's distribution was used as a proxy for adequate environmental conditions and the location of streams and direction of water flow were used to determine where bacteria could spread from a point of introduction. A GIS was used to combine these spatial data, create a data layer of random points of introduction, and determine the location and area that would be affected by bacteria introduced at 1000 randomly selected points. The southeastern region of Norway was predicted to have the highest concentration of and largest affected areas. Shortcomings of the model include an incomplete set of predictor variables (e.g., the probability of infection of a host plant) and the inability to characterize the uncertainty present in the analysis, which is an inherent problem with GIS. However, this model still provides a useful tool to predict spread of a plant pest and the methods used to produce it can be easily applied to risk analysis for other pests.

Predictive models describing where exotic species are most likely to occur are useful to land managers with limited resources for intensive monitoring for and control of invasives. This study used environmental factors associated with cover of groups of exotic species in sampled plots in Yosemite National Park to predict the likely distributions of these species groups throughout the park. A GIS was used to combine elevation, slope, aspect, and vegetative cover and structure data for these plots. Exotic species richness and cover increased with decreasing elevation and number of herbaceous species in the sample plots. Different groups of exotic species tended to occur at different elevations and slope angles. The distribution of exotic species was correlated with cover of trees, shrubs, and herbaceous species. Thus, elevation, slope, and percent cover of the three vegetation types were used in the predictive model. The predicted distributions of the four groups of exotic species were supported by previous documentation of their habitat preferences and occurrences. The predictions were determined to be 76% correct based on accurate classification of a subset of sample plots excluded from the model. Maps were produced depicting the probability of exotic species presence throughout the park. Generally, the lower elevation and western portions of the park were predicted to have the highest probability of exotic species presence, which may be a concern as these areas are the most impacted by visitors. The authors warn that predictive models have shortcomings, i.e. they are only as good as the data informing them, but further tests of model accuracy via ground-truthing and more extensive mapping of locations of exotic species in the park are two important steps to improve the model. This model will still be useful, however, for allowing park staff to focus invasive species management on areas where these species are most likely to occur. Similar models would thus be feasible and beneficial in other parks and managed lands as well.

Siam weed (Chromolaena odorata) is a climbing perennial native to the Americas that has heavily invaded Africa and Asia. In South Africa, it is considered both an agricultural and environmental problem and control of it is required by law. This study examined the percent cover of Siam weed within protected areas where control methods were used to determine the effectiveness of the methods and provide management recommendations. The control methods were a combination of mechanical and chemical control applied to each site up to six times at one to three-month intervals. Records for management blocks and the dates of applying Siam weed control had been maintained by the control agency in a GIS. These data were used to randomly sample points within each management block to estimate percent cover of Siam weed and bare ground. Control methods generally appeared to reduce percent cover of Siam weed, with the most effective treatments being plots where control methods were applied only once or twice and plots where the interval between the repeat application was only one month. Thus, recommendations from this study were to limit control of Siam weed to one application and to revegetate cleared areas following weed control.

Identification of habitats that are most vulnerable to both invasion and negative impacts of invasive species would allow managers to most efficiently use the resources available to them for invasion prevention and control. This study focused on predicting lakes in Ontario that would be vulnerable to smallmouth bass invasion because (1) the species' range is expanding due to climate change and accidental or purposeful introduction and (2) the species can affect other fish populations to varying degrees based on the prey species available in the lake. First, a model predicting lakes where smallmouth bass could establish was produced. This model was based on several environmental variables, including community composition, mean annual temperature, and mean annual precipitation. Temperature and precipitation values for each lake were determined using an inverse distance weighted interpolation in a GIS with values obtained at weather stations over 30 years. Second, a model predicting which lakes would be most impacted by smallmouth bass invasion was produced. This model used fish species distribution data and food web analyses from a subset of lakes to determine the circumstances in which smallmouth bass would most affect lake trout populations. The two models were combined to identify lakes that were most vulnerable to smallmouth bass invasion and should thus be the initial focus of prevention efforts. Of the 3046 lakes used to produce the models, 854 already contained smallmouth bass and 69 had potential for smallmouth bass invasion. Of the lakes susceptible to invasion, 48 were predicted to be most impacted. The locations of all lakes classified by the two models were mapped using a GIS. Though 48 lakes were predicted to be most vulnerable to smallbouth bass invasion, only 31 are located outside of a protected area and thus may be more susceptible to bass introductions.

Freshwater marsh vegetation is affected by various disturbances, which may be natural in origin or anthropogenically-related. Different types of disturbance in combination may have synergistic effects on native plant communities, or a greater effect than predicted by the sum of their individual effects. This study examined the effects of fluctuating water levels, human population growth as a proxy for eutrophication, and exotic species abundance on the native plant community of a freshwater marsh. Percent cover of plant species and open water were estimated in a GIS from digital versions of field vegetation maps created between 1946 and 1979. A digital elevation model was used to estimate mean water depths for the most abundant species in these years. The change in species' distributions between 1946 and 1979 was determined by a union of the vegetative cover layers for these two years in the GIS. Native cattails (Typha latifolia) and exotic manna grass (Glyceria maxima) were the most abundant species in the marsh. Cattail cover was negatively correlated with water level, human population growth, and manna grass cover, with a synergistic effect of water levels and manna grass cover detected. Manna grass cover was negatively correlated with water level and positively correlated with human population growth. Though both cattails and manna grass declined over the study period due to high water levels in these years, manna grass replaced 24% of the 1946 cattail cover by 1979, suggesting it is a better colonizer and benefits from current conditions in the marsh, e.g., greater nutrient availability. These results suggest that marsh restoration will be impeded by the presence of invasive species and conditions associated with urbanization.