Intelligent decision-making in smart port development in China through green finance instruments: a sustainable approach to the marine ecosystem

Introduction Despite the benefits of smart ports development for productivity, energy saving, and environmental improvement, an intelligent investment strategy should consider potential adverse effects on marine ecosystems during the construction and operation processes. To address this issue, this study aims to examine the integration of green finance instruments with artificial intelligence (AI)-driven intelligent decision-making (IDM), utilizing data on 15 major Chinese ports. Methods Employing machine learning (ML) models, alongside SHapley Additive exPlanations (SHAP) analysis, the research quantifies the impact of green finance on critical environmental metrics, including total organic carbon (TOC), carbon fluxes, carbon burial rate, pollution load index (PLI), flow velocity, and erosion/deposition rate (E/DR). First, ML models are employed to estimate these indicators based on historical data. Subsequently, SHAP is utilized to interpret the impact of financial instruments on ecological indicators. This enables the identification of financial instruments that positively influence ecological indicators in specific marine regions, thereby supporting IDM to prioritize those instruments in the corresponding areas. Results Findings highlight green bonds as the most influential, with SHAP values of 0.24-0.30 for carbon burial rate and 0.17-0.20 for PLI, particularly in advanced ports like Shanghai and Ningbo-Zhoushan, while eXtreme gradient boosting (XGBoost) achieves superior predictive accuracy. Discussion This study suggests that green bonds, green leasing, and green credit should be prioritized. Policymakers should establish a dedicated framework for green bonds and green leasing, specifically targeting ports with advanced smart infrastructure (L3-L4). Green credit schemes should be promoted to support infrastructure enhancement and renewable energy projects in L1-L2 ports.