Species Distribution Models for Five Rare Plant Species within the Blackfoot Swan Landscape Restoration Project
Presentation Type
Oral Presentation
Abstract/Artist Statement
Purpose – This study aimed to design a framework for identifying rare plant habitats at the landscape scale using fine-resolution remotely sensed products to try to capture complex ecological relationships between plant species with limited distribution and their environment. Using the Maximum Entropy algorithm (MaxEnt), relationships to and between each environmental variable were quantified for each species in the Blackfoot Swan Landscape Restoration Project (BSLRP) in western Montana. This project is part of a greater vegetation assessment for BSLRP that utilizes remotely sensed raster products for planning and management purposes. It may also serve as a framework for other resources and projects to assess risk, prioritize work, and provide additional information on species distribution and ecology. Methods – The five rare plant species studied in this analysis were common camas (Camassia quamash), clustered lady’s-slipper (Cypripedium fasciculatum), western pearlflower (Heterocodon rariflorum), Howell’s gumweed (Grindelia howellii), and crested shieldfern (Dryopteris cristata). Known occurrence data for these five species were taken from multiple state and federal databases and assessed for taxonomic and locational accuracy. Environmental covariates hypothesized to affect plant distribution included elevation, topographic wetness index, solar insolation, precipitation, landcover, and geology. This analysis tested the efficacy of LiDAR-derived canopy cover to improve prediction accuracy for the crested shieldfern. Species occurrence data and environmental covariates were used as input into the MaxEnt algorithm, which produced a potential distribution map for each species. The accuracy of the predicted distribution maps were assessed by field sampling ten 210 m transects for each species.Originality – Typically rare plant populations are spatially represented as polygons in state and federal databases. However, MaxEnt requires occurrence data to be in point form. Species Distribution Models (SDM) traditionally address this problem by taking the center of a polygon feature and representing it as a single point. This methodology can be ecologically misleading since it fails to acknowledge the environmental gradient found across a plant population. This analysis addressed this by placing a proportional number of points randomly within the boundary of each polygon. Results – The field surveying identified 11 new populations of common camas and 2 new populations of crested shieldfern. Comparison between predictions based upon centroids versus distributed points yielded an improved AUC and reduced standard deviation. The LiDAR data defined a narrower niche and had improved AUC and lower standard deviation but did not lower assessed accuracy for crested shieldfern. Overall accuracy for the predictions ranged from 26% to 69%. Significance – Rare plant management is contingent on understanding species distribution across the area of interest. For many rare plants, there is little knowledge on distribution or ecology. Additionally, field studies designed to collect more information on rare plants are both time and labor intensive. SDM studies are one way for resource managers to identify potential habitat for rare species before going into the field. They can use this information to prioritize field surveys and inform management decisions. Concomitantly, SDM studies provide a wealth of information on species’ environmental associations and can be used to further understand species ecology.
Species Distribution Models for Five Rare Plant Species within the Blackfoot Swan Landscape Restoration Project
UC Ballroom, Pod #3
Purpose – This study aimed to design a framework for identifying rare plant habitats at the landscape scale using fine-resolution remotely sensed products to try to capture complex ecological relationships between plant species with limited distribution and their environment. Using the Maximum Entropy algorithm (MaxEnt), relationships to and between each environmental variable were quantified for each species in the Blackfoot Swan Landscape Restoration Project (BSLRP) in western Montana. This project is part of a greater vegetation assessment for BSLRP that utilizes remotely sensed raster products for planning and management purposes. It may also serve as a framework for other resources and projects to assess risk, prioritize work, and provide additional information on species distribution and ecology. Methods – The five rare plant species studied in this analysis were common camas (Camassia quamash), clustered lady’s-slipper (Cypripedium fasciculatum), western pearlflower (Heterocodon rariflorum), Howell’s gumweed (Grindelia howellii), and crested shieldfern (Dryopteris cristata). Known occurrence data for these five species were taken from multiple state and federal databases and assessed for taxonomic and locational accuracy. Environmental covariates hypothesized to affect plant distribution included elevation, topographic wetness index, solar insolation, precipitation, landcover, and geology. This analysis tested the efficacy of LiDAR-derived canopy cover to improve prediction accuracy for the crested shieldfern. Species occurrence data and environmental covariates were used as input into the MaxEnt algorithm, which produced a potential distribution map for each species. The accuracy of the predicted distribution maps were assessed by field sampling ten 210 m transects for each species.Originality – Typically rare plant populations are spatially represented as polygons in state and federal databases. However, MaxEnt requires occurrence data to be in point form. Species Distribution Models (SDM) traditionally address this problem by taking the center of a polygon feature and representing it as a single point. This methodology can be ecologically misleading since it fails to acknowledge the environmental gradient found across a plant population. This analysis addressed this by placing a proportional number of points randomly within the boundary of each polygon. Results – The field surveying identified 11 new populations of common camas and 2 new populations of crested shieldfern. Comparison between predictions based upon centroids versus distributed points yielded an improved AUC and reduced standard deviation. The LiDAR data defined a narrower niche and had improved AUC and lower standard deviation but did not lower assessed accuracy for crested shieldfern. Overall accuracy for the predictions ranged from 26% to 69%. Significance – Rare plant management is contingent on understanding species distribution across the area of interest. For many rare plants, there is little knowledge on distribution or ecology. Additionally, field studies designed to collect more information on rare plants are both time and labor intensive. SDM studies are one way for resource managers to identify potential habitat for rare species before going into the field. They can use this information to prioritize field surveys and inform management decisions. Concomitantly, SDM studies provide a wealth of information on species’ environmental associations and can be used to further understand species ecology.