Tracy R. Holcombe

Caveats for correlative species distribution modeling

Abstract Correlative species distribution models are becoming commonplace in the scientific literature and public outreach products, displaying locations, abundance, or suitable environmental conditions for harmful invasive species, threatened and endangered species, or species of special concern. Accurate species distribution models are useful for efficient and adaptive management and conservation, research, and ecological forecasting. Yet, these models are often presented without fully examining or explaining the caveats for their proper use and interpretation and are often implemented without understanding the limitations and assumptions of the model being used. We describe common pitfalls, assumptions, and caveats of correlative species distribution models to help novice users and end users better interpret these models. Four primary caveats corresponding to different phases of the modeling process, each with supporting documentation and examples, include: (1) all sampling data are incomplete and potentially biased; (2) predictor variables must capture distribution constraints; (3) no single model works best for all species, in all areas, at all spatial scales, and over time; and (4) the results of species distribution models should be treated like a hypothesis to be tested and validated with additional sampling and modeling in an iterative process.

Remote Sensing and Mapping of Tamarisk along the Colorado River, USA: A Comparative Use of Summer-Acquired Hyperion, Thematic Mapper and QuickBird Data

Tamarisk (Tamarix spp., saltcedar) is a well-known invasive phreatophyte introduced from Asia to North America in the 1800s. This report compares the efficacy of Landsat 5 Thematic Mapper (TM5), QuickBird (QB) and EO-1 Hyperion data in discriminating tamarisk populations near De Beque, Colorado, USA. As a result of highly correlated reflectance among the spectral bands provided by each sensor, relatively standard image analysis methods were employed. Multispectral data at high spatial resolution (QB, 2.5 m Ground Spatial Distance or GSD) proved more effective in tamarisk delineation than either multispectral (TM5) or hyperspectral (Hyperion) data at moderate spatial resolution (30 m GSD).

Using State-and-Transition Modeling to Account for Imperfect Detection in Invasive Species Management

Abstract Buffelgrass, a highly competitive and flammable African bunchgrass, is spreading rapidly across both urban and natural areas in the Sonoran Desert of southern and central Arizona. Damages include increased fire risk, losses in biodiversity, and diminished revenues and quality of life. Feasibility of sustained and successful mitigation will depend heavily on rates of spread, treatment capacity, and cost–benefit analysis. We created a decision support model for the wildland–urban interface north of Tucson, AZ, using a spatial state-and-transition simulation modeling framework, the Tool for Exploratory Landscape Scenario Analyses. We addressed the issues of undetected invasions, identifying potentially suitable habitat and calibrating spread rates, while answering questions about how to allocate resources among inventory, treatment, and maintenance. Inputs to the model include a state-and-transition simulation model to describe the succession and control of buffelgrass, a habitat suitability model, management planning zones, spread vectors, estimated dispersal kernels for buffelgrass, and maps of current distribution. Our spatial simulations showed that without treatment, buffelgrass infestations that started with as little as 80 ha (198 ac) could grow to more than 6,000 ha by the year 2060. In contrast, applying unlimited management resources could limit 2060 infestation levels to approximately 50 ha. The application of sufficient resources toward inventory is important because undetected patches of buffelgrass will tend to grow exponentially. In our simulations, areas affected by buffelgrass may increase substantially over the next 50 yr, but a large, upfront investment in buffelgrass control could reduce the infested area and overall management costs. Nomenclature: Buffelgrass, Pennisetum ciliare (L.) Link Management Implications: Knowledge of where invasive species occur is often slim to nonexistent. In the face of this imperfect knowledge, land managers are still required to determine where to allocate their limited resources. Using a decision support model such as TELSA allows land managers to make a more informed decision on where to allocate funding. We addressed this imperfect knowledge in three ways. First, we acknowledged that there were many undetected buffelgrass plants on the landscape by stochastically adding and growing patches across the landscape throughout a 50-yr simulation. This is a way to see how populations that are not detected grow over time. We also developed a map of potentially suitable habitat to predict the future spread of buffelgrass patches. Finally, we calibrated spread rates by comparing past and current aerial photographs with simulation outputs. We found that areas invaded by buffelgrass may increase substantially over the next 50 yr, but that a large, upfront investment in buffelgrass control could reduce that area and overall management costs. The application of sufficient resources toward inventory is important because patches that remain undetected will tend to grow exponentially and, when eventually detected, will require substantially higher treatment efforts to control.

Forecasting Weed Distributions using Climate Data: A GIS Early Warning Tool

Abstract The number of invasive exotic plant species establishing in the United States is continuing to rise. When prevention of exotic species from entering into a country fails at the national level and the species establishes, reproduces, spreads, and becomes invasive, the most successful action at a local level is early detection followed by eradication. We have developed a simple geographic information system (GIS) analysis for developing watch lists for early detection of invasive exotic plants that relies upon currently available species distribution data coupled with environmental data to aid in describing coarse-scale potential distributions. This GIS analysis tool develops environmental envelopes for species based upon the known distribution of a species thought to be invasive and represents the first approximation of its potential habitat while the necessary data are collected to perform more in-depth analyses. To validate this method we looked at a time series of species distributions for 66 species in Pacific Northwest and northern Rocky Mountain counties. The time series analysis presented here did select counties that the invasive exotic weeds invaded in subsequent years, showing that this technique could be useful in developing watch lists for the spread of particular exotic species. We applied this same habitat-matching model based upon bioclimatic envelopes to 100 invasive exotics with various levels of known distributions within continental U.S. counties. For species with climatically limited distributions, county watch lists describe county-specific vulnerability to invasion. Species with matching habitats in a county would be added to that countys list. These watch lists can influence management decisions for early warning, control prioritization, and targeted research to determine specific locations within vulnerable counties. This tool provides useful information for rapid assessment of the potential distribution based upon climate envelopes of current distributions for new invasive exotic species.

Assessing range-wide habitat suitability for the Lesser Prairie- Chicken

Population declines of many wildlife species have been linked to habitat loss incurred through land-use change. Incorporation of conservation planning into development planning may mitigate these impacts. The threatened Lesser Prairie- Chicken (Tympanuchus pallidicinctus) is experiencing loss of native habitat and high levels of energy development across its multijurisdictional range. Our goal was to explore relationships of the species occurrence with landscape characteristics and anthropogenic effects influencing its distribution through evaluation of habitat suitability associated with one particular habitat usage, lekking. Lekking has been relatively well-surveyed, though not consistently, in all jurisdictions. All five states in which Lesser Prairie-Chickens occur cooperated in development of a Maxent habitat suitability model. We created two models, one with state as a factor and one without state. When state was included it was the most important predictor, followed by percent of land cover consisting of known or suspected used vegetation classes within a 5000 m area around a lek. Without state, land cover was the most important predictor of relative habitat suitability for leks. Among the anthropogenic predictors, landscape condition, a measure of human impact integrated across several factors, was most important, ranking third in importance without state. These results quantify the relative suitability of the landscape within the current occupied range of Lesser Prairie-Chickens. These models, combined with other landscape information, form the basis of a habitat assessment tool that can be used to guide siting of development projects and targeting of areas for conservation.

Cross-Scale Assessment of Potential Habitat Shifts in a Rapidly Changing Climate

Abstract We assessed the ability of climatic, environmental, and anthropogenic variables to predict areas of high-risk for plant invasion and consider the relative importance and contribution of these predictor variables by considering two spatial scales in a region of rapidly changing climate. We created predictive distribution models, using Maxent, for three highly invasive plant species (Canada thistle, white sweetclover, and reed canarygrass) in Alaska at both a regional scale and a local scale. Regional scale models encompassed southern coastal Alaska and were developed from topographic and climatic data at a 2 km (1.2 mi) spatial resolution. Models were applied to future climate (2030). Local scale models were spatially nested within the regional area; these models incorporated physiographic and anthropogenic variables at a 30 m (98.4 ft) resolution. Regional and local models performed well (AUC values > 0.7), with the exception of one species at each spatial scale. Regional models predict an increase in area of suitable habitat for all species by 2030 with a general shift to higher elevation areas; however, the distribution of each species was driven by different climate and topographical variables. In contrast local models indicate that distance to right-of-ways and elevation are associated with habitat suitability for all three species at this spatial level. Combining results from regional models, capturing long-term distribution, and local models, capturing near-term establishment and distribution, offers a new and effective tool for highlighting at-risk areas and provides insight on how variables acting at different scales contribute to suitability predictions. The combinations also provides easy comparison, highlighting agreement between the two scales, where long-term distribution factors predict suitability while near-term do not and vice versa. Nomenclature: Canada thistle, Cirsium arvense (L.) Scop., reed canarygrass, Phalaris arundinacea L, white sweetclover, Melilotus albus Medik. Management Implications: Effective and proactive management of invasive species requires information on both current and potential future distributions. Alaska, similar to other high latitude areas, is relatively invasion free (Lassuy and Lewis, 2013). The rapidly changing climate in this region, however, is expected to increase the area suitable for establishment for a larger number of invasive species. Here, we present results for habitat suitability models of highly invasive plants in the southern coastal region of Alaska, creating climate driven models at a regional scale and physiographic and anthropogenic models for two local regions. Using these types of models for targeted sampling of invasive plants detected more locations with less effort than nontargeted sampling (Crall et al., 2013). Our local scale models can be thought of as predicting near term establishment and distribution (potential early detection locations for management), while longer term trends in distribution may be driven by climate, especially related to the future climate scenarios at the coastal scale (potential distributions). Locations where models at both scales indicate high habitat suitability values are more appropriate targets for current control and monitoring efforts than locations identified by a model that considers factors operating at a single scale. Additionally, evaluating the areas of future suitable habitat among early invaders can help prioritize which species should be targeted for control first. If two species have similar initial distributions and similar ecological impacts, management efforts should be directed to the species with the largest possible future distribution. These models, when incorporated into an iterative sampling approach, can guide future sampling efforts. The new sampling is then used to generate future model iterations, which then provide evolving distribution models that help prioritize locations for control and restoration efforts

Ecology and Space: A Case Study in Mapping Harmful Invasive Species

The establishment and invasion of non-native plant species have the ability to alter the composition of native species and functioning of ecological systems with financial costs resulting from mitigation and loss of ecological services. Spatially documenting invasions has applications for management and theory, but the utility of maps is challenged by availability and uncertainty of data, and the reliability of extrapolating mapped data in time and space. The extent and resolution of projections also impact the ability to inform invasive species science and management. Early invasive species maps were coarse-grained representations that underscored the phenomena, but had limited capacity to direct management aside from development of watch lists for priorities for prevention and containment. Integrating mapped data sets with fine-resolution environmental variables in the context of species-distribution models allows a description of species-environment relationships and an understanding of how, why, and where invasions may occur. As with maps, the extent and resolution of models impact the resulting insight. Models of cheatgrass (Bromus tectorum) across a variety of spatial scales and grain result in divergent species-environment relationships. New data can improve models and efficiently direct further inventories. Mapping can target areas of greater model uncertainty or the bounds of modeled distribution to efficiently refine models and maps. This iterative process results in dynamic, living maps capable of describing the ongoing process of species invasions.

Impacts of Land-Use Change to Ecosystem Services

Increasing human populations on the landscape and globe coincide with increasing demands for food, energy, and other natural resources, with generally negative impacts to wildlife habitat, air and water quality, and natural scenery. Here we define and describe the impacts of land-use change on ecosystem services – the services that ecosystems provide humans such as filtering air and water, providing food, resources, recreation, and esthetics. We show how the human footprint is rapidly expanding due to population growth, demand for resources, and globalization. Increased trade and transportation has brought all the continents back together, creating new challenges for conserving native species and ecosystems.