Accurately predicting terrestrial ecosystem responses to climate change is crucial for addressing global challenges. This relies on mechanistic modelling of ecosystem processes through Land Surface Models (LSMs). Despite their importance, LSMs face significant uncertainties due to poorly constrained parameters, especially in carbon cycle predictions. This paper reviews the progress made in using data assimilation (DA) for LSM parameter optimisation, focusing on carbon-water-vegetation interactions, as well as discussing the technical challenges faced by the community. These challenges include identifying sensitive model parameters and their prior distributions, characterising errors due to observation biases and model-data inconsistencies, developing observation operators to interface between the model and the observations, tackling spatial and temporal heterogeneity as well as dealing with large and multiple datasets, and including the spin-up and historical period in the assimilation window. We then outline how machine learning (ML) can help address these issues, proposing different avenues for future work that integrate ML and DA to reduce uncertainties in LSMs. We conclude by highlighting future priorities, including the need for international collaborations, to fully leverage the wealth of available Earth observation datasets, harness machine learning advances, and enhance the predictive capabilities of LSMs.

Earth System Models (ESMs) project that the terrestrial carbon sink will continue to grow as atmospheric CO$_2$ increases, but this projection is uncertain due to biases in the simulated climate and how ESMs represent ecosystem processes. In particular, the strength of the CO$_2$ fertilization effect, which is modulated by nutrient cycles, varies substantially across models. This study evaluates land carbon balance uncertainties for the Canadian Earth System Model (CanESM) by conducting simulations where the latest version of CanESM’s land surface component is driven offline with raw and bias-adjusted CanESM5 climate forcing data. To quantify the impact of nutrient limitation, we complete simulations where the nitrogen cycle is enabled or disabled. Results show that bias adjustment improves model performance across most ecosystem variables, primarily due to reduced biases in precipitation. Turning the nitrogen cycle on increases the global land carbon sink during the historical period (1995-2014) due to enhanced nitrogen deposition, placing it within the Global Carbon Budget uncertainty range. During the future period (2080-2099), the simulated land carbon sink increases in response to bias adjustment and decreases in response to the dynamic carbon-nitrogen interaction, leading to a net decrease when both factors are acting together. The dominating impact of the nitrogen cycle demonstrates the importance of representing nutrient limitation in ESMs. Such efforts may produce more robust carbon balance projections in support of global climate change mitigation policies such as the 2015 Paris Agreement.