Identifying the habitat suitability of Pteris vittata in China and associated key drivers using machine learning models.

Pteris vittata (P. vittata) possesses significant potential in remediating arsenic (As) soil pollution. Understanding the habitat suitability of P. vittata in China and pinpointing the key drivers that influence its distribution can facilitate the identification of optimal areas for using P. vittata as a remediation tool for As-polluted soils. In this study, a comparative analysis was conducted using ten machine learning models to assess the habitat suitability of P. vittata based on 744 specimen records and 20 environmental factors. The key drivers affecting the distribution of P. vittata were also investigated based on the optimal model. The results indicate that the XGBOOST model was the most reliable and stable, achieving a coefficient of determination of 0.95. Approximately 24.47 % of China's land area was identified as suitable for P. vittata. Particularly, it was predominantly found in Hainan (45.9 %), Guangxi (92.96 %), Guangdong (91.68 %), Hunan (91.26 %), Guizhou (90.83 %), Chongqing (88.17 %), Fujian (85.70 %), Yunnan (77.44 %), Jiangxi (73.99 %) and Zhejiang (57.05 %). Furthermore, this study pinpointed the lowest temperature, annual temperature range, and mining density as key drivers, contributing 45.9 %, 31.9 %, and 7.2 %, respectively. Spatial correlation analysis revealed a significant correlation between mining density and the habitat distribution of P. vittata (Moran' I = 0.519). This study confirmed that both natural factors and anthropogenic activities affect the distribution of P. vittata and provided valuable insights for optimizing the application of P. vittata in soil phytoremediation and reclamation.