Ensemble of machine learning and global circulation models coupled with geospatial databases for niche mapping of Bell Rhododendron under climate change

Himalayan species conservation faces major challenges due to unprecedented climate change. Alpine Rhododendrons are crucial components of Himalaya, yet their vulnerability to climate change remains poorly understood. This study examines niche shifting of Rhododendron campanulatum, a keystone species of alpine treeline, under different climate change scenarios using ensemble models. The study presents extensive use of four machine learning models and three global circulation models for niche modelling. Models achieved True Skill Statistic >= 0.8, Area Under Curve >= 0.9, Cohen's Kappa >= 0.7, and overall accuracy of >= 0.9. Results showed distribution of R. campanulatum is governed by annual temperature range, minimum temperature of coldest month and precipitation of warmest quarter. Analyses revealed niche contraction and expansion of a 3-5%. Contractions are particularly evident at lower treeline boundaries. Both upward and downward shifts are anticipated under future climatic scenarios.