Gabrielle Snider

GIS Applications in Invasive Species Management
GEO 565 Annotated Bibliography

By Gabrielle Snider
Contact Me

Invasive species have been identified as one of the four most critical threats to our Nation’s wildlands. A major difficulty land managers face is locating and monitoring weed invasions across landscapes. Understanding the spatial dynamics of source and satellite weed populations may help managers prioritize which weed populations are most important to control and which habitats are most at risk for invasion. Understanding invasive weeds in a spatial context will allow managers to combat weed spread more efficiently and economically.

Cole and Albrecht present a GIS-based cellular automata (CA) framework to study and create a model to predict spread of invasive plant species. Until recently, very few studies have incorporated both a GIS and some other form of dynamic modeling in one or two dimensional space to study species invasion. Cole and Albrecht suggest a dynamic model which itself can create a different model based on play and simulation. This paper explains what cellular automatas are (composed of cell, state, neighborhood and rules). It is a combination of a raster and vector data structure. It uses a grid system, but also is completely dependant on spatial relationships with each cells neighbor. In a CA, space is proximal, defined by its adjacency and influence of its neighbor. In GIS, space is absolute and determined by a coordinate. Cole and Albrecht use the two systems together to explore the complexity of ecological, spatial and temporal questions. As an example, CA’s could be used to obtain the optimal set of “rules” for any object or cell to go from one particular state to another. CA’s can also determine how many cycles it would take to get there. These authors emphasize scientific creativity through “play” and simulation as non-destructive way to manipulate objects, variables and relationships to explore an environment over time and space. This article references some key papers in the discussion about spatial analysis and invasive species .

Dark presents an analysis of invasive and noninvasive alien plant distribution in California using GIS. The study had three main objectives; demonstrate the potential for using existing data to better understand plant invasion at a bioregional scale 2. Examine the relationship between invasive and noninvasive alien plants distribution in relation to physical/environmental characteristics like elevation, disturbance, native plant species richness, climate and 3. Examine the effect of spatial autocorrelation in traditional statistical analysis vs. analysis using spatial autoregressive models. Dark found both invasive and alien non invasive had similar distribution, especially when looking at areas with either very high or very low numbers of species. The main difference between aliens was in the North Coast bioregion, where Dark found the greatest difference between the ratios of alien invasives (high) to alien non invasives (low). Other findings related to spatial autocorrelation and autoregressive modeling were; road density and population density were positively correlated with invasive species number, elevation was negatively correlated with invasive species number, road density was negatively correlated with elevation and positively correlated with population density, and elevation was negatively correlated with population density. OLS (ordinary least squares) regression was also used to look at the number of invasive alien plants relationship with native species richness, elevation and road density. Further, SpaceStat 1.9 indicated there was significant spatial dependence in the data and the spatial model was the most appropriate. This article showed the potentially misleading and inconclusive effects of not accounting for spatial dependence when needed. Check this out article out when doing statistical analysis on Liberty data. This article does a good job of referencing past important works in the field of invasive species, including referencing recent important issues in invasive species including; the relationship between species diversity and invasibility, invasive establishment in relation to disturbance, elevation and climate influence on invasibility, and affects of invasives on underground processes. Also, Dark’s explanation of statistical analysis is straightforward and easy to understand.

Understanding the susceptibility of a geographic area to weed invasion is a major concern for land managers and land owners. Accurate predictions of where a weed is likely to establish can provide managers with sufficient time to enact early, aggressive steps to prevent establishment and better equip them to allocate their resources. Gillham et al. present the Weed Invasion Susceptibility Prediction (WISP) model. It was developed as an extension of ArcView to predict potential risk of invasion by individual weed species in rangelands. First, a weed species database is created which defines the known environmental requirements (parameters) for establishment and growth of a particular weed. Parameters include soil pH, soil texture, distance from disturbance, distance from direct water, annual precipitation, associated land cover, elevation, slope and aspect. Next, nine geographic data layers are selected which represent the important factors for establishment. Layers include soil characteristics, texture and pH, land cover, precipitation, elevation, slope and aspect. In addition, road and water surface layers are included. For each species, a score-based “Existence Potential” layer can then be created within the user-defined area. For each raster grid cell, the geographic data are compared to the weed parameters. If the data of the geographic layer meet the criteria set by the parameter for that layer, one point is added to the total score for that raster cell. The final prediction for invasion is a new geographic data layer where each raster has a possible high score of 9 (one for each data layer). This final layer predicts those areas most likely to be invaded by a particular species. The real power in this model, is the ability of the practitioner to add, modify or delete layers and parameters as the information pool about an area or a particular weed grows.

Goslee et al. present their research on using a geographic information system and an individual based simulation model (ECOTONE) to investigate and predict risk of Acroptilon repens invasion across Colorado. As there are limited resources available to control invasive species, the authors assert that a predictive model and map are necessary to identify areas which have the greatest risk of invasion, as a way to prioritize the allocation of resources. They used an individual based model which incorporates traits of the native community, the invasive species, and a biotic factors to predict invasibility. Further, they created a coarse scale threat map which identified sites across Colorado which appeared to be at greatest risk of invasion. Goslee et al. compared their map to existing A. repens populations around the state to determine the accuracy of the model. The authors found that the performance of A. repens was significantly related to the three factors they tested; above ground biomass as a function of percentage clay, mean annual temperature and annual precipitation. Surprisingly, to me, soil clay content showed the strongest effect on A. repens, followed by precipitation and temperature. Beyond the actual study and results, the authors touted the benefits of using an individual based model (because it allows for inclusion of both spatial heterogeneity and ecological processes and can be used to make predictions for a specific area) and decried the shortcomings of models like the common diffusion model that cannot be used to make predictions at an individual location. The usual suspects arose when discussing the potential source of error in the model the authors used; inadequate resolution and lack of complete understanding of the ecology and physiology of A. repens.

Remote sensing has been used to map the distribution of invasive species which dominate the understory. However, 67 of the worlds 100 most invasive species are found in the understory. The authors suggest an indirect method of mapping understory invasive species which uses the ecological relationships between a species and its environment to predict the distribution of the species. The premise is that spatial information on environmental conditions affecting invaders’ reproductive and competitive traits could be used to predict its distribution. The paper discussed a study in which the invasive shrub species, Chromolaena odorata, was mapped using this indirect method.
Previous studies suggested light intensity to be the most important environmental trait that drove competitiveness and reproduction of C. odorata. A three layer feed-forward back-propagation artificial neural network consisting of an input, a hidden and an output layer was used to map light intensity using the first seven bands of Landsat ETM+image. Light intensity reaching the understory was derived from hemispherical photographs and used to train the artificial neural network. Of 302 observations, half were randomly selected to be used in the training of the artificial neural network and half were used for the actual testing. Results from the study showed a strong relationship between light and cover abundance and seed production. The authors exposed how GIS and remote sensing technologies can map and localize source populations of invasive species through indirect mapping techniques as well as provide land managers and ecologists with an innovative tool with which to better understand invasive species.

Kalkhan and Stohlgren used the cross correlation statistic, Iyz¬ to test for spatial cross correlations between combinations of several variables including topographic variables, plant species richness, soil characteristics, and plant cover in a 754 ha study site in Rocky Mountain National Park, Colorado. Kalkhan and Stohlgren demonstrate one use of spatial cross correlation analysis to determine (quantify) invasive species patterns on a landscape scale based on aerial photo cover and GIS. The objective of this study was to use the cross correlation statistic in combination with linear correlation to analyze spatial relationships between species richness of native and non native plant species with topographic and edaphic factors with large and small plot sampling techniques. Eight of the 12 variables showed significant cross correlation with other variables in the 1000m2 plots, and 6 of 12 variables showed significant spatial auto-correlation. Much more analysis was performed looking at environmental factors, soil, topography etc. The same analysis was used in the small plot sampling (250 1m2 subplots). In this smaller plot design, all the variables showed significant positive spatial auto-correlations. The authors did a good job of integrating GIS capabilities with spatial correlation statistical analysis to present a case for understanding spatial relationships and plant distribution on a landscape scale. However, they emphasize the importance of study design (plot size and sample size) when using these statistical techniques, as their findings differed between their two approaches.

Radosevich et al. discuss several approaches to illuminate the factors responsible for the spread of invasive species. A GIS reconstruction, DNA analysis and species demography were explored as possible ways to better predict where invasion is most likely to take place and where land managers should focus their efforts. Radosevich et al. discuss the three phases of invasion; introduction, colonization and naturalization as well as how and why previous predictive models have fallen short. The authors suggest more research should be done to determine the role of satellite populations of invasive species as areas with the most potential for rapid spread. Looking at these satellite populations, spatially, in relation to their source populations will provide a more descriptive picture of the process and patterns of species invasions. A GIS which incorporates the interplay of extrinsic (environmental) and intrinsic (biological) factors to assess the potential risk to land area may prove to be a vital tool for land managers in combating invasive species. This article talks about; safe sites, sources vs. sinks, the diffusion estimator, successful life history characteristics and general processes and patterns of biological invasion

Determining where non indigenous species are, and which species are present, is one of the most important management goals in invasive species management. Little research has been done to determine which survey method is the most consistent and accurate. Rew et al. use a GIS model to evaluate seven different survey methods. They graded each method on consistency and reliability of intersecting non indigenous species (NIS) patches, producing samples which reflect the spatial distribution of the population, and can be implemented in the most time and cost efficient way. The GIS model was developed to create NIS populations which were then sampled using the different survey methods and the results were recorded. Grid and random points, and targeted (stratified continuous) transects (starting on a road and following a 2km transect) methods provided the most consistent samples of the population. The targeted transect method was the most resource efficient.

This paper evaluates the circumstances under which control programs reduce the range of two widespread invasive weeds of riparian habitats (Himalayan balsam and giant hogweed). The spread of the species was modeled using MIGRATE, a model that uses realistic demographic parameters and multiple dispersal mechanisms. In addition, simulations of range of control scenarios were run with a GIS using real landscapes based on topographic, hydrological and land cover maps of the area. Wadsworth et al. looked at six strategies for weed control including; at random, in relation to human population density, by size, by age (young or old), and by spatial distribution of the weed. The strategies were evaluated on area treated per year and proportion of plants destroyed. Timeliness of implementation was also evaluated. The most successful strategies were those based on weed population and spatial characteristics. Plant population size and spatial distribution were also key factors. A positive finding was a reduction in weed range after any control strategy was implemented. Wadsworth et al. assert that successful control of both species is only possible when strategies based on species distribution data are used, and when they are undertaken at relatively high intensities and efficiencies.