Data-driven modeling of carbon dioxide and methane fluxes across the Arctic-boreal region: recent achievements and future opportunities

The Arctic-boreal region contains a significant portion of the global carbon reservoir, which is becoming increasingly vulnerable to atmospheric release due to climate change. Thus, it is vital to monitor and model Arctic-boreal carbon flux dynamics to understand the shifting carbon balance. Data-driven statistical and machine learning models are now commonly used to upscale carbon dioxide (CO₂) and methane (CH₄) fluxes in northern ecosystems from local to circumpolar scales. This review highlights recent progress, ongoing challenges, and new insights into data-driven Arctic-boreal carbon flux modeling. We identify five key areas for future model development: (1) developing comparable upscaling frameworks for terrestrial and freshwater ecosystems, especially for understudied systems like lakes and rivers; (2) reducing uncertainties and gaps in geospatial observations to better represent landscape heterogeneity; (3) fusing process-based with upscaling models to reduce computational demands, allowing to increase resolution and boost predictive power; (4) maximizing synergies between data-driven, process-based, and atmospheric inversion models using hybrid modeling; and (5) collaborating with field and remote sensing scientists to ingest observations in the upscaling process in near-real-time. These steps will reduce carbon budget uncertainty, advance our understanding of the carbon cycle, and support global climate policy.