Advancing Tree Growth Prediction with Interactive and eXplainable AI for Tackling Climate Change

Understanding the intricacies of tree growth is crucial for understanding vegetation dynamics, optimizing carbon sequestration, preserving biodiversity, and enhancing climate adaptation within forest ecosystems. Leveraging primarily time-series data from dendrometers and weather stations
provided by the International Cooperative Program for Forests (ICP-Forest), this study explores tree growth dynamics across diverse regions in Austria. Despite the value of this data, the nature of its collection introduces noise and errors, posing challenges for analysis. To address this, we employ advanced deep learning models within a machine and human interaction framework to
predict tree growth, complemented by state-of-the-art explainability AI techniques (e.g., SHAP and LIME). By analyzing dendrometer and weather data, the study specifically investigates the impact of environmental components’ fluctuations over time on tree growth, offering valuable insights
into forest ecosystem dynamics and their response to changing climatic conditions. We show that there is a strong correlation between soil moisture, temperature, and individual tree growth, emphasizing the importance of including these environmental factors in predictive models.
Furthermore, we underscore the necessity of calculating tree competition parameters (estimated using terrestrial laser scanning data collected for the project), which play a vital role in accurately modelling tree dynamics and growth patterns. Lastly, initial forecasting results demonstrated high
accuracy, providing a robust foundation and serving as a baseline for developing more sophisticated machine learning models. These insights collectively can advance the understanding of forest dynamics and offer a pathway toward enhancing global vegetation models and more effective data-driven decision-making in forestry.