Large divergence of satellite and Earth system model estimates of global terrestrial CO2 fertilization

Satellite-derived estimates of increases in terrestrial net primary productivity are less than half of those derived from Earth system models. This discrepancy is explained by over-sensitivity of Earth system models to atmospheric CO2 concentrations. Atmospheric mass balance analyses suggest that terrestrial carbon (C) storage is increasing, partially abating the atmospheric [CO2] growth rate1, although the continued strength of this important ecosystem service remains uncertain2,3,4,5,6. Some evidence suggests that these increases will persist owing to positive responses of vegetation growth (net primary productivity; NPP) to rising atmospheric [CO2] (that is, ‘CO2 fertilization’)5,6,7,8. Here, we present a new satellite-derived global terrestrial NPP data set9,10,11, which shows a significant increase in NPP from 1982 to 2011. However, comparison against Earth system model (ESM) NPP estimates reveals a significant divergence, with satellite-derived increases (2.8 ± 1.50%) less than half of ESM-derived increases (7.6 ± 1.67%) over the 30-year period. By isolating the CO2 fertilization effect in each NPP time series and comparing it against a synthesis of available free-air CO2 enrichment data12,13,14,15, we provide evidence that much of the discrepancy may be due to an over-sensitivity of ESMs to atmospheric [CO2], potentially reflecting an under-representation of climatic feedbacks16,17,18,19,20 and/or a lack of representation of nutrient constraints21,22,23,24,25. Our understanding of CO2 fertilization effects on NPP needs rapid improvement to enable more accurate projections of future C cycle–climate feedbacks; we contend that better integration of modelling, satellite and experimental approaches offers a promising way forward.