Climate change is altering forest ecosystems and challenging the stability of plant-microbe interactions. Microbial communities residing inside plant tissues, especially in woody organs, play a crucial role in plant health, yet their responses to climatic shifts and management interventions remain poorly understood. This project aims to identify microbial bioindicators sensitive to climate variability and treatment conditions such as soil chemistry and water deficit. It also seeks to develop interpretable machine-learning models to predict ecosystem responses based on microbial community profiles. Metagenomic datasets were combined with high-resolution meteorological records and treatment metadata were combined with high-resolution meteorological records and treatment metadata. Alpha and beta diversity metrics, ordination methods, and statistical models (OLS regression, Random Forest) were applied to explore the dynamics of microbial assemblages. Significant interannual differences in microbial diversity were observed, particularly in Aesculus and Fagus. Temperature and precipitation explained up to 89% of the variability in diversity indices, with a negative impact of excess rainfall on Fagus. Random Forest classification based on microbial genera achieved 92% accuracy in year prediction and identified genera such as Bulbomeridium and Ophiocordyceps as potential temporal indicators. Microbial communities inside plant tissues are highly responsive to environmental fluctuations and treatment-related stressors. Their structure reflects both climatic variability and host-specific conditions. Integrating metagenomics, environmental metadata, and machine learning offers a robust approach for monitoring ecosystem health and supporting climate-adaptive strategies.