This study predicted potential habitat shift under climate change of the subalpine rare plant Rhododendron sohayakiense var. koreanum using only 13 limited occurrences. We constructed an ensemble of small models (ESM) based on bivariate combinations to compensate for the degradation in predictive power due to limited sample size. Seven high-resolution environmental variables were used: four climate metrics from CHELSA v2.1 (BIO2, gst, gsp, and swe) and three topographic indices derived from the 30 m NASA SRTM (Shuttle Radar Topography Mission) DEM (Digital Elevation Model) (slope, TPI, and TWI). With the ecospat package in R, the ESM was projected to 2100 for three shared socioeconomic pathways (SSP1-2.6, SSP3-7.0, and SSP5-8.5). All scenarios showed progressive habitat contraction, and the high emission SSP5-8.5 scenario produced an almost complete loss of suitable area by the end of the century. The results demonstrate that ESMs can provide reliable vulnerability assessments for narrowly distributed species even when occurrence data are scarce, offering a quantitative basis for conservation planning for Rhododendron sohayakiense var. koreanum and similar alpine rarities.