Our predictive understanding of land biogeochemistry has been improved greatly owing to substantial increases in data availability and advances in biogeochemical models. However, quantitative forecasting of biogeochemistry is still a big challenge, as reflected by large uncertainties across Earth system models. To meet the challenge, it will likely be necessary to develop new theories on carbon cycle, invent new modelling techniques, devise innovative tools for data-model integration, and/or use new artificial intelligence approaches (machine learning and deep learning). This session aims to include presentations on new advances in carbon cycle science and techniques to improve carbon cycle models. We invite contributions from experimentalists, modelers, and theoreticians to this session on improving land carbon cycle modeling.