Tree of Heaven Habitat Suitability

Abstract

The invasion of ecosystems by non-native species is a major driver of biodiversity loss worldwide. A critical component of effective land management to control invasion is the identification and active protection of areas at high risk of future invasion. The Appalachian Trail Decision Support System (A.T.-DSS) was developed to inform regional natural resource management by integrating remote sensing data, ground-based measurements and predictive modelling products. By incorporating NASA’s remote sensing data and modelling capacities from the Terrestrial Observation and Prediction System (TOPS), this study examined the current habitat suitability and projected suitable habitat for the invasive species tree-of-heaven (Ailanthus altissima) as a prototype application of the A.T.-DSS. Species observations from forest surveys, geospatial data, climatic projections and maximum entropy modelling were used to identify regions potentially susceptible to tree-of-heaven invasion. The modelling result predicted a 48% increase in suitable area over the study area, with significant expansion along the northern extremes of the Appalachian Trail.

Objectives

• Relate field-based observations of the distribution of Ailanthus to a set of environmental variables.
• Map the current distribution of suitable habitats and identify high-risk regions along the A.T.
• Integrate projected precipitation and temperature data from TOPS based on IPCC climate change scenarios to simulate potential shifts in the distribution of Ailanthus habitats.

Tree-of-heaven

• Ailanthus altissima (Mill.) Swingle, a deciduous member of the Simaroubaceae family
• Native to the temperate regions of central China; populations established on every continent except Antarctica
• Exotic tree species pervasive throughout the U.S. due to its rapid growth, high fecundity, hardy tolerance, and strong competitive ability
• Quickly colonizes disturbed areas; suppresses growth of native species
• Vulnerability to frost damage restricts from higher latitudes and elevations
• Early detection crucial for minimizing the costs of control and risk of further dispersal and establishment

Geospatial Data Sources

• Forest Inventory & Analysis Database
  • Provides a nationwide systematic sample of forested ecosystems well-suited for broad scale ecological analysis.
  • Measurements include tree species, size, conditions, and physiographic site attributes.
  • Data for 3,926 FIA plots available within the A.T. shell between 2002 and 2010.
  • Ailanthus observations recorded at 136 of those FIA plot locations.
• Climate data and forecasts
  • Climate baseline (1950-2005) and projections (2090-2095) provided by NASA’s Terrestrial Observation and Prediction System (TOPS)
  • Downscaled from Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5) Coupled Model Intercomparison Project Phase 5 (CMIP5)
  • Representative Concentration Pathway (RCP) 6.0 selected for model projection (moderate-high emission scenario)
  • Ecologically relevant bioclimatic variables derived from monthly precipitation and temperature data using R package “dismo”
• Ancillary Geospatial Data
  • Elevation and derived topographic variables including slope, transformed aspect, and slope position from NED
  • Urban areas, agriculture, canopy cover, and wetlands from NLCD 2006
  • Soil hydrology from STATSGO

Maximum Entropy (MaxEnt) Modeling

• Machine-learning-based method
• Presence-only, absences do not indicate conditions unsuitable for invasive populations
• Maximum entropy distribution = least constrained
• Generates ‘features’ based on distribution of environmental variables across presence points
• Many iterations, balancing gain against regularization to prevent overfitting
• Variable ‘clamping’ restricts values of projected variables to range of current variables
• Suitability threshold applied to continuous probability distributions to derive binary classes

Variable Selection

• Eliminate highly correlated variables (Pearson correlation coefficients)
• Evaluate performance across model iterations based on variable response curves, contributions and permutation importance, and jackknifing
• Selections should make ecological ‘sense’ given known species characteristics

Model Evaluation

• Performance: 10-fold cross validation on test area under curve (AUC) of receiver operating characteristics (ROC)
• Complexity: sample size adjusted Akaike information criteria (AICc)
  • Simplicity particularly desirable when transferring (projecting) to new conditions
• Consistency: ecologically significant variables selected and resulting distribution in agreement with existing knowledge

Results

• Climatic variables consistently selected as highest performing
• Landcover variables performed poorly across broad geographic area; eliminated from final model
• Restricting MaxEnt to ‘hinge’ feature types increased model transferability
• Final model selected from array of >15
• Moderate complexity; incorporates 4 climatic and 4 topographic variables
• Strong indication potential extent of Ailanthus habitats likely to increase as climate changes
• Increasing invasive pressure on sensitive high elevation areas in northern ecosystems
• Independent test data needed to fully evaluate model performance