Current and future projections of forest aboveground carbon storage over Northeast China using an advanced deep convolutional neural network

Forests are the largest carbon reservoirs in terrestrial ecosystems, with climate and anthropogenic disturbance have considerably reshaped the size of this system. However, remote sensing estimates to quantify current and future carbon stock are characterized by large uncertainty. Here, we developed a new spatially explicit map of forest aboveground carbon storage at a 30 m spatial resolution across Northeast China, by integrating an image-to-image U-Net deep convolutional neural network algorithm with in situ measurements, Global Ecosystem Dynamics Investigation observations, Landsat, and PALSAR2 images. The result yields a high coefficient of determination (R2) of 0.83 and a relatively low relative root mean squared error of 24.02%, and is superior to traditional pixel-based and in situ-based methods. The forest aboveground biomass carbon stock was estimated as 2.74 ± 0.38 PgC, with precipitation and forest age being important determinants. By developing a climate and age-driven predictive model, local forest biomass C sink is estimated to be 7.42 and 2.51 TgC year-1 for 2060s and 2100s, a significant decline due to aging forests. The study highlights the importance of forest protection and management to optimize forest age structure and sustain natural forest carbon sinks.