Identification of suitable areas for the growth of Teucrium polium species using machine learning models (a case study of Khalil Abad and Kashmar Counties, Iran)

With the advancement in computer technologies, the prediction of ecological niches for various plant species has become possible. The impact of climate change on the distribution of plants could be investigated using species distribution models. This study aimed to identify the climatic and environmental factors influencing the distribution of Teucrium polium species and determine its geographic range in Kashmar and Khalilabad counties, Khorasan Razavi province, Iran. To achieve this, 75 bioclimatic variables encompassing soil, topography, climate, and geology factors were initially analyzed for correlation, and variables with correlation coefficients higher than 0.80 were eliminated. The data of 57 GPS-recorded presence points were collected from two areas during 2021 −2022. The environmental data and presence points were processed and predicted using the BIOMOD2 package within the R software, which encompasses 11 models. The models were evaluated using Cohen’s kappa coefficient (KAPPA), True Skill Statistic (TSS), and Receiver Operating Characteristic (ROC) indices. Model accuracy assessment revealed that the Random Forest (RF) model achieved 99.7% accuracy while the Ensemble model achieved 99.4% accuracy, indicating excellent modeling precision. The relative importance of different variables in the selected models were as follows: in the RF model, silt at a depth of 30 − 60 and 15 − 30 cm, topographic humidity index, annual mean temperature, and daily temperature range; and for Ensemble model, nitrogen at a depth of 15 − 30 cm, topographic humidity index, soil bulk density at a depth of 5 − 15 cm, silt at a depth of 30 − 60 cm, and nitrogen at a depth of 15 − 5 cm, highlighting the influence of soil factors on the species distribution. The obtained results could be utilized for the conservation, management, and expansion of Teucrium polium habitats in similar areas.