Sensitivity and uncertainty of modelled terrestrial net primary productivity to doubled CO2 and associated climate change for a relatively large perturbed physics ensemble

Net primary productivity (NPP) is often modelled explicitly in general circulation models (GCMs) utilising process models that may include plant photosynthesis, respiration, allocation of photosynthates, phenology, mortality and competition between plant functional types. It is an important measure for understanding the role of terrestrial vegetation in the global carbon cycle, and useful for gaining insights into the large-scale, integrated effects of climate and atmospheric changes on potential plant productivity and associated impacts, i.e. food security and carbon cycle feedbacks. However, there are simplifications and uncertainties in GCM projections of future climate change, as well as further uncertainties involved in modelling the associated terrestrial vegetation responses. In particular, it is important to highlight that many GCM simulations, including the ones used in this study, do not model nutrient limitation, even though primary plant nutrients, e.g. nitrogen and phosphorus, are key limiting factors on plant productivity. Here, we examine sensitivities and uncertainties in large(global)-scale modelled NPP to climate and atmospheric carbon dioxide concentration [CO2], utilising a relatively large perturbed physics ensemble (PPE) of simulations generated from the HadSM3 GCM under equilibrium doubling of pre-industrial atmospheric [CO2]. We also exploit the ensemble design to highlight the relative importance of two, often opposing, forcings on NPP: (i) plant physiological responses to CO2, termed ‘Phys’; and (ii) plant responses to physical drivers of climate, termed ‘Rad’. It is important to note that this is a sensitivity study that provides useful guidance on the relative importance of the Rad and Phys drivers and their uncertainties. The results cannot be considered quantitatively realistic, particularly because the equilibrium experimental design and lack of nutrient limitation in the model are important limitations that prevent such interpretation. We find that doubled [CO2] and associated climate changes ultimately increase potential global average NPP by 57%, from 0.293 kg cm−2 yr−1 (∼36 PgCyr−1) to 0.460 kg cm−2 yr−1 (∼57 PgCyr−1). Spatially, the largest decreases (∼−0.45 kg cm−2 yr−1) occur across the north-east of South America in association with the largest decreases in precipitation. The largest increases (up to ∼0.75 kg cm−2 yr−1) occur across tropical Africa and Indonesia, where NPP is already high, and both temperature and precipitation increase under doubled [CO2]. In most regions where NPP shows an increase the changes are significantly larger than the ensemble standard deviation, indicating that increases in global NPP under doubled [CO2] are reasonably robust. However, in some regions, particularly north-eastern South America and Central America, where NPP decreases are projected, the standard deviation across the ensemble is larger than the average NPP change, indicating that even the sign of the NPP sensitivity to doubled [CO2] and climate is uncertain. These uncertainties are shown to be highly dependent on the relative sensitivities of NPP to the Phys and Rad forcings. Highlights , Global ensemble average potential NPP increases by 57% under doubled atmospheric [CO2] conditions (in equilibrium), and these increases occur in just over 91% of the current vegetated land area. , Radiative (Rad) and physiological (Phys) forcings account for a 21% decrease, and 75% increase, respectively in the ensemble average global potential NPP. , Global average potential NPP changes for the RadPhys ensemble are generally greater than the ensemble SD, indicating that the direction of simulated NPP responses to doubled [CO2] are reasonably robust. , In some regions, particularly in north-east South America, the SD of the RadPhys ensemble is larger than the average NPP change, highlighting that even the direction of potential NPP responses in these regions are highly uncertain. , In some regions, i.e. the northern boreal zone, temperate North America, and parts of eastern Europe, uncertainties in radiative forcing (Rad) are the dominant influence on the uncertainties in potential NPP responses, whereas, in other locations, i.e. Central America and north-eastern South America, the large uncertainties in the potential NPP response are dominated by physiological forcing (Phys).