<p>Greenhouse gas cutting and air pollutant reduction are crucial for effectively addressing climate change and regulating air quality. This study aims to contribute to sustainable green development by constructing an explainable spatiotemporal model averaging method for point and interval prediction of the pollution reduction and carbon reduction collaborative control degree at the provincial level in China. Specifically, this study proposes a novel graph convolutional long transformer as the core for spatiotemporal prediction, which integrates graph convolutional network to acquire spatial relationship, simplified transformer block to capture long-term temporal pattern, and skipping connection to decrease the gradient vanishing. Then, graph convolutional long transformer enhanced by the STIRPAT economic theory model generates multi-view predictions, which fully consider the neglected socioeconomic dynamic relations between nodes. This study integrates multi-view predictions through the spatiotemporal model averaging method, which combines linear operator, spatial prediction error correction, and weight coefficient optimizer. This study evaluates the proposed system using pollution reduction and carbon reduction data from China provinces. The proposed forecasting system demonstrates superior point and interval prediction performance compared to other methods. The application and policy implication of the proposed system are discussed in this study to provide actionable support for promoting sustainable green development and controlling energy consumption.</p>