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.

The role of greenhouse gases (GHGs) in global climate change is now well recognised and there is a clear need to measure emissions and verify the efficacy of mitigation measures. To this end, reliable estimates are needed of the GHG balance at national scale and over long time periods, but these estimates are difficult to make accurately. Because measurement techniques are generally restricted to relatively small spatial and temporal scales, there is a fundamental problem in translating these into long-term estimates on a regional scale. The key challenge lies in spatial and temporal upscaling of short-term, point observations to estimate large-scale annual totals, and quantifying the uncertainty associated with this upscaling. Here, we review some approaches to this problem, and synthesise the work in the recent UK Greenhouse Gas Emissions and Feedbacks Programme, which was designed to identify and address these challenges. Approaches to the scaling problem included: instrumentation developments which mean that near-continuous data sets can be produced with larger spatial coverage; geostatistical methods which address the problem of extrapolating to larger domains, using spatial information in the data; more rigorous statistical methods which characterise the uncertainty in extrapolating to longer time scales; analytical approaches to estimating model aggregation error; enhanced estimates of C flux measurement error; and novel uses of remote sensing data to calibrate process models for generating probabilistic regional C flux estimates.

The Arctic is one of the regions in our planet with strongest warming observed and it is also almost certain to continue to change in the near future. The continuous change in key indicators of Arctic climate change (e.g. increase of temperature, intensification of the hydrological cycle, and shortening of the spring snow cover) will have marked consequences on ecosystem carbon (C) sink-source functioning. Such consequences are, however, broadly uncertain. Comprehensively integrated ecosystem models with long-term in-situ data are essential to understand the Arctic C cycle sensitivity to climate change and explore robust future scenarios. Our aim is to quantify the relative sensitivity of Greenland’s C balance to climate change based on regional variation in C and N cycling in a tundra gradient. The key roadblocks to this understanding have been limited time series of C fluxes, and limited regional data. Now with observations from multiple data streams measured by the Greenland Ecosystem Monitoring (GEM) program over the last two decades in conjunction with proven ecosystem and climate models we 1) analyse the underlying processes and links between present climate and terrestrial C and N cycling and 2) forecast the variation of plant phenology, productivity, and respiration forward in time. We use an established but novel C cycle model, the Soil-Plant-Atmosphere model, applied to two GEM wetlands relying on previous substantiated efforts on source-code model implementation, model calibration, and validation based on quality-controlled long-term data. Additionally, our modelling framework is now forced with future projections from the regional climate model HIRHAM5 specifically designed to characterize the Greenland domain (typically left behind in global modelling analyses) following the IPCC greenhouse gas emission scenarios. We ask the ecological question: How sensitive is the C balance expected to be under warmer and wetter conditions forecasted for the 21st century? Although still preliminary, we found strong evidence that the net C exchange will be significantly exposed to higher temperatures and intensified precipitation levels increasing 10-80% the C sink strength by the end of the century, but lengthening of the growing season and nutrient availability will also play a significant role.