A new record of Stylophorum diphyllum (Michx.) Nutt. in Canada: A case study of the value and limitations of building species distribution models for very rare plants

Abstract

Stylophorum diphyllum (Michx.) Nutt. is an endangered plant of rich floodplain forests in southern Ontario, Canada. Prior to 2015 there were only four known populations in Ontario. I built a species distribution model (SDM) based on the known occurrences, and tested it by surveying 156 forest sites that varied in their predicted suitability. An indicator species analysis showed that sites predicted to be suitable had significantly higher frequency and abundance of common species usually associated with S. diphyllum, demonstrating the ability of the SDM to pinpoint similar habitat, although none of these sites contained S. diphyllum. The most important predictors used by the SDM to determine habitat suitability were growing season precipitation, surficial geology, and soil texture. I discovered a new population of S. diphyllum more than 50 km north of the known populations, at one of the sites not predicted to be suitable. This demonstrates a clear example of SDM overfitting, which may occur when models are built based on few, spatially limited occurrence records. Nonetheless, the key environmental predictors remained the same in an updated SDM including the new record. Stylophorum diphyllum provides a case study of both the value and the limitations of using SDMs to predict suitable habitat for very rare and geographically restricted plants, and the need for more rare plant surveys even in human-dominated landscapes.