The present study endeavors to delve into the integration of advanced digital artificial intelligence methodologies within the realm of landscape design. By employing a hybrid approach utilizing convolutional neural networks (CNNs) in conjunction with long short-term memory (LSTM) networks, we aim to scrutinize the intricate relationship between landscape configurations and their influence on vegetation-based carbon sequestration. The ultimate objective is to furnish evidence-based insights that can inform and optimize the strategic planning of urban landscapes. This article uses the CNN-LSTM model to conduct in-depth research on the landscape pattern of regional carbon sequestration land from two scales: landscape and type, and from four levels: morphology, composition, distribution, and structure. Using remote sensing image data, through preprocessing steps such as geometric correction and image dehazing, the landscape pattern index is extracted to explore the relationship between landscape pattern and vegetation carbon sequestration. Research has found that the ED, LSI, IJI, SHEI, AI indices at the landscape scale, as well as the correlation indices of cultivated land, forest land, and grassland at the type scale, are significantly correlated with vegetation carbon sequestration. The impact weights of landscape form, composition, distribution, and structure on vegetation carbon sequestration are 58.69%, 15.79%, 19.36%, and 2.83%, respectively. The application of digital artificial intelligence technology in landscape design helps to enhance the creative, planning, and integration capabilities of landscape architecture, creating practical, aesthetically pleasing, and entertaining landscapes. By optimizing the landscape pattern, especially in terms of form, composition, and distribution, the carbon sequestration effect of vegetation can be effectively improved, promoting the protection of urban ecological environment.
