2.7 Species Distribution Modeling
MAXENT 3.3.3e(Phillips, Anderson, & Schapire, 2006) was used to predict the distribution of Notholirion in China during 4 time periods: Last Interglacial (LIG), Last Glacial Maximum (LGM), and the present and future. The samples collected in this study covered the distribution of all 30 populations of the genus Notholirion throughout China for which specimens were recorded. Nineteen bioclimatic environment variables(Alemu & Wimberly, 2020) from four time periods above were downloaded from the WorldClim database(Fick & Hijmans, 2017) and employed in modeling analyses. In MAXENT, due to the small sample size of our data (15–79 localities), only the linear + quadratic + hinge functional forms were used. Here, models were built using linear + quadratic + hinge features, random training data was set to 25%, repeat run type was set to Subsample, maximum iterations were set to 1000, and other parameters were set by default. We also calculated the area under the Receiver Operating Characteristic (ROC)(Fawcett, 2005) and Operating Characteric Curve (AUC)(Ellis, Yahr, Belinchón, & Coppins, 2014) to observe the accuracy of each model prediction. Usually, good model performance was assessed by AUC values above 0.7(Elith & Leathwick, 2009; Fielding, 1997; Peterson, Pape, & Soberón., 2008). The spatial distribution range map of the final modeling results was plotted by the software DIVA-GIS v 7.5(Hijmans, Cameron, Parra, Jones, & Jarvis, 2010).