Multi-country pine species allocation under different climate scenarios

• Machine Learning models applied to Pine species allocation in multiple countries. • High Precision maps of recommended Pine Species under different climatic scenarios. • Insights for Pine species testing and prioritization in Tree Breeding programs. Climate-change scenarios can expose forests to several environmental hazards and recommending the right tree species to be planted in the right place is a key factor. Tree breeding programs provide valuable information on species adaptability through field trials. Although data is available, there is a lack of studies that provide decision-support models capable of predicting the impact of climate change on site-species recommendations. This study aims to develop multi-country decision-support models for pine species that can assist in pine species (genetic material) allocation under past and future climate scenarios, utilizing machine learning techniques and environmental covariates. The variable selected to express growth potential was the dominant height at age 8 years (HT8). The source for environmental covariates used was WorldClim 2.1. Random Forest models were fitted for each genetic material and were used to build allocation maps to optimize HT8 growth under past and future climate scenarios. Model evaluation metrics were performed using R-squared (R²); Root Mean Square Error (RMSE); Mean Absolute Error (MAE). The RF models showed high accuracy, with a mean R² of 0.78, MAE of 6.4 %, and RMSE of 8.6 % across all species. The most widely allocated pure species across both scenarios were Pinus maximinoi, Pinus tecunumanii high elevation, and Pinus tecunumanii low elevation, covering 28 %, 16.9 %, and 4.2 % of the total area, respectively. Under the future scenario, the ranking of species remains consistent, while the proportions shift slightly. The proposed methodology provides a practical tool to help companies select the top potential pine species for development and planting.