# Testing a New Niche Hypervolume Technique for Predicting Susceptibility to Non-Indigenous Invasive Plants

A pre-proposal in response to SERDP SON Number CSSON-00-02:
"Natural Resources Management Control of Non-Indigenous Invasive Species"

SUBMITTED IN CONFIDENCE. PLEASE DO NOT CITE OR DISTRIBUTE.

Forrest M. Hoffman
Oak Ridge National Laboratory
Environmental Sciences Division
P.O. Box 2008
Oak Ridge TN 37831-6036
865-576-7680
forrest@esd.ornl.gov

William W. Hargrove
University of Tennessee
Energy, Environment, and Resources Center
Systems Development Institute
10521 Research Drive, Suite 100
Knoxville TN 37932
865-241-2748
hnw@ornl.gov

Diana Cooksey
Montana State University
Department of Land Resources and Environmental Sciences
Leon Johnson Hall, MSU-Bozeman
Bozeman MT 59717
406-994-5684
dcooksey@montana.edu

Anthony J. Krzysik
United States Army Corps of Engineers
Construction Engineering Research Laboratories
2902 Newmark Drive
Champaign IL 61822-1076
217-373-4561
krzysik@gis.uiuc.edu

Signature: __________________________________________

We propose to test a new Niche Hypervolume technique which will estimate the likelihood of non-indigenous species expansion into new habitats on military bases, based on the present locations of infestation. We believe that this Niche Hypervolume technique will identify the susceptibility of various ecosystems on federal lands to invasion, and will predict the likely extent of expansion. If practicable, this invasive species range prediction technique could fulfill DoD responsibilities to develop and maintain an integrated, programmatic approach for predicting the extent of biological pollution on federal lands, as specified under the National Invasive Species Act of 1996 and the Presidential Executive Order on Invasive Alien Species.

## Background

Using multivariate geographic clustering, we have produced several maps of ecoregions across the 48 conterminous United States at a resolution of one square kilometer per cell. Ecologists have long used the concept of the ecoregion, an area within which there are similar ecological conditions, as a tool for understanding large geographic areas. Multivariate geographic clustering employs a non-hierarchical grouping of individual cells in a digital map from a Geographic Information System (GIS) for the purpose of classifying the cells into types or categories based on characteristics identified as important to the growth of vegetation. Nine characteristics from three categories--elevation, edaphic (soil) factors, and climatic factors--were identified as important. These national data are represented as over 7.8 million map cells in a 9-dimensional data space (Hargrove and Luxmoore 1998).

Because multivariate geographic clustering with these large amounts of data is computationally intensive, we employ a parallel computer constructed from low-cost commodity personal computers. Working in unison, each of the standard PCs acts as a node in the composite parallel machine. By dividing the clustering problem into smaller tasks, each node can solve a small part of the overall problem, enabling the machine to quickly and successfully cluster the entire region of interest at very high resolution (Hoffman et.al., 1997).

## The Niche Hypervolume Susceptibility Analysis

The kernel of the Niche Hypervolume idea is a statistical embodiment of Hutchinson's (1957, 1965) definition of an organism's ecological niche as "an N-dimensional hypervolume." We propose to use detailed maps of the invasive organism's current distribution to define its niche, and then use the niche definition to "paint" the prediction of the full extent of the infestation back onto the map. We will statistically define the occupied part of the niche for several federally-recognized noxious weedy species in N-dimensional cluster or environmental space, then re-map all places that are suitable for colonization by the organism.

We will begin with the current distribution map for each of five weeds developed by one of us (Cooksey). These species, chosen by the Montana Noxious Weed Trust Fund Council, are leafy spurge, spotted knapweed, Russian knapweed, sulfur cinquefoil and Dalmatian toadflax, from the Montana section-based weed maps database (Cooksey et.al., 1998). We will submit all cells where an invasive species currently occurs to one of our multivariate spatial clustering analyses. Unlike our previous clustering studies, we will not perform a Principal Components Analysis, but will retain all of the raw environmental variables for clustering.

The goal of this step is to characterize and define the occupied (or occupiable) volume of environmental space by statistically cataloging all combinations of environmental conditions which are habitable by this species. Rather than specifying the number of clusters, we will use a technique in which we specify the radius of the clusters. We will specify a very small radius, resulting in a large number of clusters. These clusters will serve to precisely define the location and shape of the niche space for this particular invasive species.

Then, using all cells from a map of larger geographic extent, we will re-cluster using the same radius and the same cluster centroids as before. Map cells from the larger extent which are now inside the organism's niche space will be "captured" into those clusters. When the locations are re-assembled into the map, the maximum potential invasive spatial distribution for this species will be shown as the cells which were assigned to these clusters.

If the current distribution map is not simply present/absent, but has some surrogate measures of performance or "fitness" of the invasive organism at each location (i.e., abundance, density, biomass), it is possible to assign the same surrogate fitness measure to the new distribution map using the distance from this cell to the cluster centroids. The predicted distribution map following complete invasion will not be simply binary, but will project how well the invading plant will do in each location under the new conditions. New areas in the map which are similar to places known to be occupied by this species will be identified and mapped.

## Proposed Tests of the Niche Hypervolume Susceptibility Analysis

We propose two scales of approach which differ substantially in spatial extent and resolution: 1) a regional analysis, whose spatial extent will incorporate multiple military bases, national parks and forests, Indian reservations, and other federal land holdings, and 2) a within-base scale approach. The regional approach will be several states in extent, and we propose to perform this regional analysis at a resolution of 1 to 1.5 km. The objective of the regional approach is to provide early-warning of the impending arrival of weedy species, so that appropriate preparations can be made before the actual arrival of an invasive species at an installation.

Of course, an analysis resolution of 1 sq km could potentially miss many micro-sites which would be suitable for colonization by a particular invasive species. However, by considering a large extent, we will provide federal land managers with a "big-picture" strategic view of the progress of invasive plants into their region.

A finer, within-base scale approach will be employed and evaluated for areas of up to several thousand square kilometers. This approach will use environmental data at a resolution of ca. 30 m. It is envisioned that this finer-scale approach would be adopted by federal land managers once invasion of a base is imminent or has already occurred.

Field ecologists and botanists familiar with a particular base could probably make predictions about potential susceptibility to invaders based on a natural history perspective and personal expertise, assuming that they were familiar with both the base environmental conditions and the biology and requirements of the invader. However, it is unlikely that a single expert would have the requisite familiarity with a number of potential invaders, or that he/she could generate an accurate prediction of susceptibility at a scale of over several states. We are proposing an entirely objective and empirical technique which will result in the same prediction when applied by anyone, and which can potentially be applied at scales larger than those with which human experts are likely to be personally familiar.

Many invasive species are generalists which are able to thrive in many combinations of environmental conditions. For some invasive species, the majority of a region or base may be susceptible. Prediction of a severe potential impact is important, and such generalist invaders will demonstrate this capacity quickly via Niche Hypervolume Susceptibility analysis, since they will be seen to thrive in many combinations of conditions very soon after their initial appearance. Thus, the potential severity of an invader can be assessed quickly.

For less general invasive species, Niche Hypervolume analysis might rule out large parts of a base as unsuitable to a weedy species, allowing federal land managers to concentrate identification and eradication efforts and resources in more limited areas. Moreover, an analysis of niche breadth (the narrowest part of the niche hypervolume) will indicate the environmental characteristic providing the most constraint on this species, and may suggest eradication methods which are otherwise relatively benign.

## Procedures to Map Invasive Species on Federal Lands

Cooksey at Montana State has developed the Montana Noxious Weed Survey and Mapping System, and has published the {\it Weed Mapping Handbook (v1.9)}, a compendium of relevant and useful information on mapping weeds. She is also an author of an extension publication entitled {\it Mapping Noxious Weeds in Montana}, which contains statewide mapping procedures for weeds, including a weed data dictionary and metadata formats. We propose to modify and expand the procedures developed for Montana for specific use by personnel on federal lands.

## Incorporation of Biotic and Anthropogenic Environmental Characteristics

Species present on federal lands are subject to anthropogenic and biotic forces as well as abiotic environmental factors. These biotic factors can potentially be included in the description of the Niche Hypervolume as well as abiotic factors. For example, disturbance caused by training activities can be characterized as a component of the environment if the invasive potential of a species is to be predicted. We propose to develop and incorporate indices, classes, or severity of disturbance due to military training into the niche hypervolume definition.

Similarly, the biotic effects caused by predators, parasitoids, and competitor species can be considered as important environmental characteristics. At the extreme, these biotic agents represent potential biological control of invasive species. We propose to incorporate such biotic effects into the niche hypervolume definition in order to address the third objective of this SON. Krzysik at CERL has expertise in environmental disturbance due to military training, and is also interested in individual biotic interactions, and will lead the effort to incorporate biotic and anthropogenic factors into the niche hypervolume definitions.

## Technology Transfer to Federal Land Managers

Initially, the large number of cells in the regional or multiple-base scale and the relatively high number of environmental niche dimensions will present a computational challenge, and we have included in the budget additional node capacity for our parallel supercomputer to meet these requirements. However, as the dimensionality of the niche hypervolume definition and the geographic extent of the analysis decreases, we expect the computational requirements to substantially relax.

Federal land managers will not require a parallel supercomputer to successfully perform Niche Hypervolume Susceptibility analysis, particularly at within-base scales. Once the niche hypervolume has been defined for a species, a rough susceptibility analysis can be performed by base planners using a standard GIS system by specifying the range of conditions within which the plant has been found, parameter by parameter. While not as accurate as the full niche volume location and shape specification that we propose to initially employ, the Boolean combination of these ranges essentially re-forms the niche hypervolume. Application of the set of ranges can be accomplished by any GIS to identify all of the local susceptible areas, simplifying the technical transfer of the Niche Hypervolume analysis to federal land planners. Because the initial phase of niche hypervolume definition is conceptually similar to an unsupervised image classification, which can be achieved with standard, commercially-available image processing software, it may be ultimately possible to also transfer the statistical niche-defining ability to federal land managers.

## Specific Work Proposed

An important part our testing of Niche Hypervolume analysis will be to determine whether the niche hypervolume for a particular invasive species defined at a regional scale can be adequately and accurately applied at within-base scales. We propose to develop niche hypervolume definitions for each of the five Montana weeds from the state-scale township-resolution maps of these species, and then to test these coarse-resolution, wide-extent niche hypervolumes for making high-resolution predictions of invasion susceptibility at within-base scales. Our hope is that, by considering a large extent, we will still be likely to statistically encounter locations representing a full set of combinations of environmental growing conditions for the invaders. If so, it may not be necessary to re-develop niche definitions to be used at particular bases at a finer spatial resolution. This top-down approach to invasive species, if successful, would represent an important cornerstone in a centrally-integrated programmatic approach to management of invasive species on federal lands. Centrally-developed niche hypervolume definitions for particular species of federal importance could be distributed to individual bases, which could then apply them to test for local susceptibility.

First, we will define the niche hypervolume for each species using weed map data from the state of Montana at 1.5 km resolution. Next, we will use the statistical niche definition developed using these state-scale data to predict ultimate expansion for each of the five weed species over a portion of the western United States. Then we propose to use the niches defined above to make specific predictions at high-resolution for Malmstrom Air Force Base, near Great Falls, Montana. Finer resolution maps of the five weed species have already been developed for Malmstrom AFB. We will hold these maps in reserve, make blind predictions for Malmstrom, and then compare with actual results. We will also use the wide-extent niche hypervolume definitions to make specific predictions for the susceptibility to invasion by these weeds within the Marine Corps Air Ground Combat Center (ca. 2400 sq km) in the southern Mojave desert. If we are invited to prepare a full proposal for this SON, we will add collaborators from one or both of these specific military bases to facilitate the testing of the Niche Hypervolume analysis at the within-base scale.

## References

Hargrove, W. W. and R. J. Luxmoore. 1998. A New High-Resolution National Map of Vegetation Ecoregions Produced Empirically Using Multivariate Spatial Clustering. http://www.esd.ornl.gov/~hnw/esri98

Hoffman, F. M., W. W. Hargrove, and A. J. Schultz. 1997. The Stone SouperComputer: ORNL's First Beowulf-Style Parallel Computer. http://www.esd.ornl.gov/facilities/beowulf

Cooksey, D., E. Roberts, and R. Sheley. 1998. Montana Noxious Weed Survey and Mapping System. http://www.montana.edu/places/mtweeds/slides/slide43.html

Hutchinson, G.E. 1957. Concluding Remarks. Cold Spring Harbor Symp. Quant. Biol. 22:415-427.

Hutchinson, G.E. 1965. The niche: An abstractly inhabited hypervolume. In: The Ecological Theatre and the Evolutionary Play. New Haven, Yale University Press, pp. 26-78.

## Attachments

1. Clustering Technique Diagram
2. Sample Data Layers Used in Multivariate Geographic Clustering
3. Multivariate Geographic Clustering for Determination of Ecoregions

## Personnel

Forrest M. Hoffman is a member of the staff at the Environmental Sciences Division of Oak Ridge National Laboratory. Mr. Hoffman is responsible for the development of the parallel supercomputer and much of the parallel multivariate geographic clustering code which will be modified and used for the Niche Hypervolume Susceptibility analysis.

William W. Hargrove, Ph.D., is a landscape ecologist with expertise in GIS, remote sensing, and simulation modeling with the University of Tennessee in Knoxville, TN. Dr. Hargrove is also a subcontractor within the Geographic Information Science and Technology group of the Computational Physics and Engineering Division of Oak Ridge National Laboratory.

Diana Cooksey is a Land Resources Project Coordinator at Montana State University-Bozeman. Among the projects she coordinates are the Montana Noxious Weed Survey and Mapping System, the Montana Cooperative Agricultural Pest Survey (CAPS) program, and the Montana Agricultural Potentials System (MAPS). Ms. Cooksey has a plant science degree from Colorado State University.

Anthony J. Krzysik, Ph.D., is Senior Research Ecologist at the U.S. Army Construction Engineering Research Lab in Champaign, IL, and is an adjunct associate professor in the Department of Natural Resources and Environmental Sciences at the University of Illinois. Dr. Krzysik has expertise in field, quantitative, and statistical ecology, with an emphasis in arid and semiarid landscapes, and has worked extensively on military lands in the Mojave desert.

## Proposed Budget and Milestones

The proposed project has a duration of 3 years.

### Year 1:

• Delivery and incorporation of Cooksey's Montana weed maps.
• Purchase and installation of supercomputer upgrades.
• Hoffman and Hargrove modify current parallel clustering code to define niche hypervolumes.
• Krzysik locates fine-scale environmental data for southern Mojave.
• Krzysik begins development of disturbance index for incorporation into niche definition.
• Cooksey begins modifications of Weed Mapping System for use on federal lands.
• First attempts at niche definition for at least one of the five weed species by the end of first year.
 Hoffman and Hargrove/ORNL $175,000 (1/3 time each) Parallel computer upgrades 75,000 Cooksey/MSU 75,000 Krzysik/CERL 75,000 Total Cost:$400,000

### Year 2:

• Krzysik incorporates disturbance index due to military training into environmental conditions.
• Develop niche definitions for all five invasive species.
• First regional prediction maps of susceptibility potential.
• Initial tests of coarse-scale niche definitions for within-base predictions.
• Initial susceptibility prediction maps for Malmstrom AFB.
• Initial susceptibility prediction maps for Marines Air Ground Combat Center.
 Hoffman and Hargrove/ORNL $175,000 (1/3 time each) Cooksey/MSU 75,000 Krzysik/CERL 75,000 Total Cost:$325,000

### Year 3:

• Final predictions for regional scale for all five species.
• Final prediction maps for Malmstrom AFB and Marines Air Ground Combat Center.
• Cooksey obtains fine-scale maps of five weed species for Malmstrom AFB.
• Comparison of results of "blind test" predictions with actual weed maps.
• Tech transfer of Niche Hypervolume analysis to federal land managers.
• Transfer of Federal Noxious Weed Survey and Mapping system to military.
 Hoffman and Hargrove/ORNL $175,000 (1/3 time each) Cooksey/MSU 75,000 Krzysik/CERL 75,000 Total Cost:$325,000