A Machine Learning Framework to Evaluate Vegetation Modeling in Earth System Models

Vegetation gross primary productivity (GPP) is the single largest carbon flux of the terrestrial biosphere which, in turn, is responsible for sequestering 25%-30% of anthropogenic carbon dioxide emissions. The ability to model GPP is therefore critical for calculating carbon budgets as well as understanding climate feedbacks. Earth system models (ESMs) have the capability to simulate GPP but vary greatly in their individual estimates, resulting in large uncertainties. We describe a machine learning (ML) approach to investigate two key factors responsible for differences in simulated GPP quantities from ESMs: the relative importance of different atmospheric drivers and differences in the representation of land surface processes. We describe the different steps in the development of our interpretable ML framework including the choice of algorithms, parameter tuning, training and evaluation. Our results show that ESMs largely agree on the physical climate drivers responsible for GPP as seen in the literature, for instance drought variables in the Mediterranean region or radiation and temperature in the Arctic region. However differences do exist since models don't necessarily agree on which individual variable is most relevant for GPP. We also explore a distance measure to attribute GPP differences to climate influences versus process differences and provide examples for where our methods work (South Asia, Mediterranean) and where they are inconclusive (Eastern North America). Gross primary productivity (GPP) is the rate at which plants remove carbon dioxide from the atmosphere during photosynthesis. Carbon dioxide is a greenhouse gas and excess in the atmosphere causes global warming and climate change. Changes in the amounts of atmospheric carbon dioxide will impact the entire Earth System. We therefore need the ability to accurately calculate GPP, especially for different possible carbon usage pathways in the future. Earth system models or ESMs allow us to simulate various processes happening in the earth's atmosphere and biosphere including photosynthesis and can help us estimate GPP changes for such different pathways. However, ESMs can vary significantly in their simulated GPP estimates making it difficult to have confidence in using these estimates. We describe a ML framework to better understand where ESMs differ in calculating GPP so that we can address knowledge gaps in models. This approach allows us to understand the processes involved without having to run computationally expensive simulations. With improved models, we can also improve our ability to predict climate change outcomes for the future. A machine learning framework to advance our understanding of the terrestrial carbon cycle in earth system models or ESMs is proposed Differences in the relative importance of atmospheric drivers of gross primary productivity highlights differences across models A method to attribute differences in productivity estimates from ESMs due to process representation versus atmospheric forcing is demonstrated