AI for Climate Peace: A Cooperative RDOC Monitoring System Based on Tokyo Bay Observations
As global warming intensifies, achieving carbon neutrality has become a shared strategic goal among nations. However, the ocean’s role as a long-term carbon sink—particularly through Refractory Dissolved Organic Carbon (RDOC)—remains poorly quantified due to its invisible nature and lack of direct observational data. This proposal designs an AI-driven monitoring and forecasting system based on 40 years of water quality data from Tokyo Bay. It aims to identify and predict RDOC as a stable carbon pool, support marine carbon accounting, and enhance transparency in climate data. The system is structured around six modular components and offers the following core approaches:
Automated RDOC Proxy Extraction:Uses multiscale decomposition (HP filtering, STL, EMD) and unsupervised clustering (GMM, DBSCAN) to isolate the long-term, low-variability fraction of DOC as a proxy for RDOC.
AI-Based Estimation and Forecasting:Employs supervised learning (Random Forest, XGBoost) to estimate the RDOC ratio and deep learning (LSTM, Transformer) to forecast future trends under different environmental scenarios.
Environmental Independence Verification:Applies Granger causality, mutual information, and Convergent Cross Mapping to ensure the RDOC proxy is minimally influenced by external drivers like temperature, salinity, and nutrients.
Transparency and Peacebuilding:Promotes cross-border trust through explainable AI models, shared analytical frameworks, and scenario-based insights, turning marine carbon management into a collaborative opportunity for climate peace.
