Species distribution modeling of Hyrcanian boxwood (Buxus Hyrcana Pojark) using the random forest model in the forests of northern Iran

Background and Objectives: The stability and sustainability of ecosystems require understanding the relationships between the distribution of plant species and environmental factors. Species distribution modeling techniques are recognized as powerful tools to support forest management strategies in the context of climate change. Boxwood is a shade-tolerant species that is distributed in the understory of lowland and middle-altitude forests of northern Iran, where it creates a dark environment in the forest by forming a dense canopy. Due to the widespread destruction of boxwood habitats in recent decades, this species has been considered part of conservation communities to prevent its extinction. The main objective of this research is to model the distribution of this species in the Hyrcanian forests.
Methodology: In this study, the distribution of boxwood was modeled using 570 actual occurrence points in Hyrcanian forests. Bioclimatic variables were extracted from the WorldClim database and topographic variables from the digital elevation model. VIF test was used to reduce collinearity. 70% of the samples were assigned as training data for model development and the remaining 30% as test data for model validation. Modeling was performed using the random forest algorithm in R environment and using presence and pseudo-absence data, adjusting the model parameters (500 trees, mtry=2, min_n=5) and validating it by using the 10-fold Cross-Validation method. Model performance was evaluated with AUC, TSS, overall accuracy, and kappa coefficient.
Results: The results of the Random Forest model showed that the model has high accuracy in predicting the distribution of boxwood (AUC=0.98, Accuracy=0.95, Kappa=0.72, TSS=0.63). The random forest model, using bioclimatic and topographic variables, showed very favorable performance in predicting the distribution of Hyrcanian boxwood. Evaluation of variable importance indicated that the bioclimatic variables bio3 (isothermality), bio12 (annual precipitation), bio8 (mean temperature of the wettest quarter) and bio1 (annual mean temperature) had the greatest impact on the distribution of the boxwood species. The response curve of boxwood to the four important variables was also plotted. The boxwood habitat suitability map identified areas with suitable ecological conditions, including large parts of Mazandaran province and the border areas of the Hyrcanian forests.
Conclusion: Analysis of variable importance showed that isothermality (Bio3) was the most effective variable, indicating the role of thermal stability in the viability of boxwood; an evergreen and shade-loving species that is sensitive to temperature fluctuations and is more present in areas with a balanced temperature range (15–20 degrees). Annual precipitation (Bio12) was also of high importance, and the response curve showed that the probability of boxwood presence is much higher in areas with precipitation above 300 mm; because it provides sufficient moisture for growth, photosynthesis, and reduction of water stress. Variable Bio8 (mean temperature of the wettest quarter) by affecting vegetative growth and resistance to fungal diseases, provides optimal conditions for this species in the temperature range of about 5 to 10 degrees. Also, Bio1 (annual mean temperature) showed that boxwood is more distributed in areas with temperate climates and an average annual temperature between 8 and 13 degrees. In contrast, topographic variables such as slope length and solar radiation played a complementary role and moderated microclimatic conditions at the local level. The habitat suitability map also showed that the central and western regions of the Hyrcanian forests, especially in the provinces of Mazandaran and Gilan, have the highest probability of species presence. Also, in some areas of Golestan province, medium to high suitability was predicted, which shows the potential for boxwood restoration in these areas. These results confirm the key role of climate, especially the combination of temperature and humidity, in explaining the distribution of boxwood and emphasize the need to protect habitats with stable climatic conditions.