Abstract

Accurately quantifying carbon fluxes is critical for understanding carbon cycle dynamics and predicting carbon-climate feedbacks in rapidly-warming Arctic and boreal ecosystems. Here we evaluate net ecosystem exchange (NEE) estimates derived from three distinct approaches: atmospheric inversions, upscaled flux measurements, and terrestrial biosphere models. We leverage atmospheric CO2 observations collected during airborne campaigns organized as part of the Carbon in Arctic Reservoirs Vulnerability Experiment (CARVE) and the Arctic Carbon Atmospheric Profiles (Arctic-CAP) campaigns during 2012–2014 and 2017, respectively. We find that the ability to reproduce observed atmospheric CO2 variability varies substantially across the three modeling approaches. Model-data consistency is strongly associated with estimated NEE seasonality; models showing less consistency with atmospheric observations often exhibit earlier spring onset, earlier carbon uptake peak, or delayed autumn senescence. Furthermore, we find that these NEE seasonality biases result largely from the seasonality of gross primary production and, to a lesser extent, that of heterotrophic respiration. These findings underscore the need for integrated frameworks that incorporate diverse approaches to generate carbon flux estimates that are constrained at large scales while providing detailed process-level insights. Our results provide insights to guide model refinement and highlight the need to expand in situ measurements in Arctic and boreal regions and derive region-specific constraints on the relationships among vegetation phenology, carbon fluxes and environmental drivers.