Parameterizations in Earth System Models (ESMs) are subject to biases and uncertainties arising from subjective empirical assumptions and incomplete understanding of the underlying physical processes. Recently, the growing representational capability of machine learning (ML) in solving complex problems has spawned immense interests in climate science applications. Specifically, ML-based parameterizations have been developed to represent convection, radiation and microphysics processes in ESMs by learning from observations or high-resolution simulations, which have the potential to improve the accuracies and alleviate the uncertainties. Previous works have developed some surrogate models for these processes using ML. These surrogate models need to be coupled with the dynamical core of ESMs to investigate the effectiveness and their performance in a coupled system. In this study, we present a novel Fortran-Python interface designed to seamlessly integrate ML parameterizations into ESMs. This interface showcases high versatility by supporting popular ML frameworks like PyTorch, TensorFlow, and Scikit-learn. We demonstrate the interface’s modularity and reusability through two cases: a ML trigger function for convection parameterization and a ML wildfire model. We conduct a comprehensive evaluation of memory usage and computational overhead resulting from the integration of Python codes into the Fortran ESMs. By leveraging this flexible interface, ML parameterizations can be effectively developed, tested, and integrated into ESMs.

Plant and microbial nitrogen (N) dynamics and nitrogen availability regulate the photosynthetic capacity and capture, allocation, turnover of carbon (C) in terrestrial ecosystem. It is important to adequately represent plant N processes in land surface models. In this study, a plant C-N framework was developed by coupling a biophysical and dynamic land surface processes model, SSiB4/TRIFFID, with a soil organic matter cycling model, DayCent-SOM, to fully incorporate N regulations to investigate the impact of N on plant growth and C cycling. To incorporate the N limitation in the coupled system, the parameterization for dynamic C/N ratios for each plant functional type (PFT) was developed first. Then, after accounting for plant/soil N-cycling, when available N is less than demand, N would restrict the plant growth, reducing the net primary productivity (NPP), but also impact plant respiration rates and phenology. The improvements of the newly-developed model, the SSiB5/TRIFFID/DayCent-SOM, was preliminary verified at three flux tower sites with different PFTs. Furthermore, several offline global simulations were conducted from 1948 to 2007 to predict the long-term mean vegetation distribution and terrestrial C cycling, and the results are evaluated with satellite-derived observational data. The sensitivity of the terrestrial C cycle to N processes is also assessed. In general, new model can better reproduce observed emergent properties, including gross primary productivity (GPP), leaf area index (LAI), and respiration. The main improvements occur in tropical Africa and boreal regions, accompanied by a decrease of the bias in global GPP and LAI by 16.3% and 27.1%, respectively.

It is important to adequately represent plant nitrogen (N) biogeochemistry and its respective processes in land surface models. Thus far, various N representations in models lead to uncertainty in estimating model responses to global warming. Through plant and microbial N dynamics, nitrogen availability regulates the capture, allocation, turnover of carbon (C), and photosynthetic capacity. In this study, to fully incorporate these N regulations, we have developed a plant C-N framework by coupling a biophysical and dynamic land model, SSiB4/TRIFFID, with a soil organic matter cycling model, DayCent-SOM, to simulate the impact of nitrogen on the plant growth and C cycling. To incorporate the N limitation in the coupled system, we first developed the parameterization for the C/N ratios. Then, after accounting for daily plant/soil N-cycling, N will not only limit the plant growth when not sufficient, causing the net primary productivity (NPP) to be down-regulated, but will also impact plant respiration rates and phenology. Using this newly-developed model named SSiB5/TRIFFID/DayCent-SOM, we conduct several simulations from 1948 to 2007 to predict the global vegetation distribution and terrestrial C cycling, and the results are evaluated with satellite-derived observational data. The sensitivity of the terrestrial C cycle to N processes is also assessed. In general, the coupled model can better reproduce observed emergent properties, including gross primary productivity (GPP), NPP, leaf area index (LAI), and respiration. The main improvement occurs in tropical Africa and boreal regions, accompanied by a decrease of the bias in global GPP and LAI by 16.3% and 27.1%, respectively.