This topic of discussion is most relevant for the East Asia region because the global FFC CO2 emission increase rate in the 2000s is mainly driven by the industrial CO2 emission from China (about 65% of 2.17 PgC year−1 increase in global emission during 2001–2010). Because our inverse model does not well constrain CO2 fluxes from China alone, due to lack of measurement sites within the source regions (Fig. 1), we aggregated six inverse model regions to discuss the change in fluxes over East Asia in this study over the past 2 decades.
Figure 3 shows the effect of different FFC CO2 emission a priori on the estimation of residual land CO2 fluxes estimated by the 84-region inverse model for the East Asia region. We find the FFC CO2 emissions as per the CDIAC inventory method are always higher compared to the CARBONES and IEA inventories, varying from 0.49 PgC in 2002 to a maximum of 0.76 in 2009 for CDIAC–IEA. The CDIAC–CARBONES differences decrease from a maximum of 0.47 in 2003 to 0.12 PgC in 2011 (Fig. 3a). A fairly compensatory land CO2 fluxes are estimated for the East Asia region with the interannual variations (IAVs) being opposite in phase. The mean (± 1−
σ standard deviations for the IAVs) differences of 0.59 ± 0.08 and 0.37 ± 0.12 PgC in FFC CO2 emissions lead to a mean uptake bias over the East Asia by −0.40 ± 0.10 and −0.21 ± 0.08 PgC, respectively, for the 2002–2011 period. This suggests about 60–67% of the FFC CO2 emission bias is transferred to land uptake increase for the East Asia region, and the rest of 33–40% of FFC CO2 bias affects inverse model results for other regions.
It would have been ideal to validate the a posteriori land CO2 fluxes using aircraft (independent) observations that are not used in the inversion system, as in the case of inverted CH4 emissions using vertical profiles over Sendai, Japan (Patra et al. 2016b). Such validation has also been attempted in Thompson et al. (2016), but large model–model differences in annual mean land fluxes for the East Asia region could not be separated using aircraft observations over Korea and Japan. Both the ACTM and MACC models with East Asian land fluxes of −0.94 and −0.44 PgC year−1, respectively, for 2008–2011 well-simulated the CO2 concentration within the planetary boundary layer and vertical gradients between 1 and 4 km (Thompson et al. 2016; their Fig. 2). Unlike CH4, which has sources predominantly on the Earth’s surface (a global surface sink of ~27 Tg-CH4 year−1, compared to ~550 Tg-CH4 year−1 of emission), CO2 has very strong source-sink variations over the land regions (Fig. 4). The annual mean biases for FFC CO2 emissions or land fluxes are much smaller than the seasonal variations in land CO2 exchange. In addition, our inversion system optimizes land-CO2 fluxes for a given FFC CO2 emission, where the inverted land fluxes can compensate for biases in an a priori FFC CO2 emission. Such a compensatory effect helps to simulate observed CO2 concentrations from aircrafts for very different source (FFC) and sink (land biosphere) combinations.
Figure 5 shows the quickest FFC CO2 emission increase in China is seen during 2003–2006 coincides well with the worsening emission intensity (CO2 emission per GDP produced), as used in preparation of EDGAR emission inventories (Olivier et al. 2014). The FFC CO2 emissions increase from China alone explains about 62% (1.72 out of 2.62 PgC year−1) global FFC emission increase in the period of 2002–2011 (Fig. 5a, b). In 2002–2003, the rate of FFC CO2 emission increase (~20% year−1) from China was twice its GDP growth rate (10% year−1). The CO2 emission increase is caused by a combined effect of fuel consumption and worsening emission intensity, from 216 kgC per thousand US$ in 2002 to 250 kgC per thousand US$ in 2005 (Fig. 5c). Note also that the emission intensity of China was about three and two times higher in 2002 and 2012, respectively, compared to those of India, Japan or USA. Given the available historical information, it is probably difficult to revise the inventory emissions retrospectively for the first half of the 2000s. Therefore we apply a CH4 emission scaling factor of 0.59 to the rate of anthropogenic CO2 emission increase for China (Ref. Fig. 2a and associated text). By talking into account the sequestered CO2 in carbonating cement materials (Xi et al. 2016), the FFC CO2 emission increase would be moderated by 0.09 PgC in the period 2002–2012, which is about 12% of the mismatch in FFC CO2 emissions between CDIAC/GCP and this study in 2012 (Fig. 5).
The FFC CO2 emissions, before and after CH4 emission scaling, are compared with a scenario when the emissions are assumed to increase as per the GDP increase rate of China. This suggest that until about 2009, the FFC CO2 emission increase was faster than the GDP increase since 2003, while the scaled emission increase is still higher than that due to the GDP increase but until about 2006. The emission intensity probably improved at a faster rate leading to the Beijing Olympic in 2008 and the years that followed (ref. EDGAR42FT). Thus, the net FFC CO2 emission increase during 2003–2014 at the rate of GDP increase is in apparent violation of the common development mechanism of industrialization (Kaya Identity; Kaya and Yokoburi 1997). Emissions per unit of GDP should decrease with time as the industries strive to improve productivity, and more energy is produced through green technology investments (Raupach et al. 2007). About 20.3% of total energy generated in China (4.994 PWh) has been by renewable energy in 2012 compared to that of 17.6% (of 1.654 PWh) in 2002 (IEA, International Energy Agency 2016; www.iea.org/statistics/statisticssearch/). About 79 and 83% of China’s energy is generated from coal and crude oil in 2002 and 2012, respectively, which constitute about 89 and 87% of its total CO2 emissions (IEA 2016; Boden et al. 2016).
Choosing different time periods, we find that the rate of FFC CO2 emission increase (1.04 PgC as per CDIAC) is about 75% of Chinese GDP increase during 2007–2014, and our revised emissions are about 64% of the economic growth for another time period of 2002−2011. As a sensitivity case, we have checked the calculation for an arbitrary scaling factor of 0.69, and that produced an FFC CO2 emission of 2.19 PgC year−1 in 2012 compared to the case presented here with 2.02 PgC year−1. This maximum difference of 0.17 PgC year−1 in East Asian CO2 flux during the period of 2010–2012 (Fig. 8) does not affect most of the conclusions of this work, and the difference is much smaller than that derived using CDIAC FFC emissions (0.71 PgC year−1). Our revised FFC CO2 emissions show better agreement, compared to CDIAC, with a consumption-based emission inventory for China (Fig. 5b; IEA 2016). The difference between IEA and our revised FFC CO2 emissions has developed during the period of 2003–2007, when the emission intensity worsened for China, and only a systematic bias is seen for the later 5 years.
This correction to Chinese FFC CO2 emission increase after 2003 has large consequences for the global and regional CO2 budgets. Figure 6 shows the 84-region inverse model estimated global CO2 fluxes using three cases of FFC CO2 emissions as described in “Methods”. The FFC CO2 emissions are balanced well by the residual/inverted land sink estimation for simulating the global atmospheric burden increase, which is seen as the greatest and smallest global total sinks for CDIAC and IEA, respectively (further details in Saeki et al. 2017). More interesting is to note that the CO2 land sink is greater by 1.18 PgC year−1 for CDIAC case compared to IEA emission case, and that is at the expenses of a large reduction in oceanic CO2 sink by 0.51 PgC year−1. This is because the FFC CO2 biases are located over the land region, and thus the compensatory sink biases in inverse model occur over the land regions when constrained by observations. Some leakage of FFC CO2 is expected, e.g. 33–40% for the East Asia region, as most of the FFC CO2 emission signals are not strongly constrained by measurements within the region. The bias in global total CO2 fluxes for the CARBONES and IEA inversion cases arises from the land total alone, with the ocean total sinks being similar. Our results clearly suggest that the land–ocean partitioning of CO2 sinks is affected greatly by the uncertainties in FFC CO2 a priori emissions, which was more often linked to transport model uncertainties (Gurney et al. 2002; Peylin et al. 2013).
Further, the land CO2 sink bias due to uncertainties in FFC CO2 emission should influence our understanding of the global and regional carbon budgets. Figure 7 shows sectorial CO2 sources/sinks budget obtained from the Global Carbon Project (GCP; Le Quéré et al. 2015). The GCP estimated global land CO2 sink depends on the FFC CO2 inventory, model of land-use change emissions and atmospheric burden increase from measured CO2 concentration. We corrected the global FFC CO2 emission using scaled emissions for China only (broken red line; =Global − CDIAC China + CH4-Scaled China), and correspondingly calculated the corrected land CO2 sink (broken brown line; Fig. 7). Apparently, the recent release of CO2 emission inventory by the British Petroleum (BP 2016) also suggested a slower emission increase during 2007–2014, compared to the CDIAC inventory, and is comparable to our corrected FFC CO2 emission. However, the rate of BP emission increase is still slightly faster than CDIAC inventory and much faster than our corrected FFC emission for the problematic time period of 2002–2007. The higher emissions during the late 1990s in the latest BP emission inventory, compared to CDIAC, agree better with Francey et al. (2013), who suggested no stabilization of emissions during the period 1996–1999.
The corrected FFC CO2 emissions may also help resolve one the mismatches in the GCP’s residual land CO2 sink budget and ensemble mean of the TRENDY DGVM simulations. A comparison suggests the GCP land sink increased by 2.04 PgC during the period 2001–2014, while that increased by only 1.34 PgC in the case of the TRENDY S2 simulation. The revised GCP land sink, using corrected FFC CO2 emissions, is calculated to be 1.27 PgC, which is in closer agreement with the TRENDY simulation (Fig. 7). The assessment of the CO2 sink simulation, by the DGVMs, has large implications for future prediction of carbon–climate feedback.
The policymaking of emission mitigation or verifying the success of INDCs, towards reduction of global levels of GHGs concentration, rely on our ability to estimate the time evolution of regional sources and sinks. However, estimations of FFC CO2 emissions and the land sink in China has been one of most discussed issues in recent times (Liu et al. 2015; Thompson et al. 2016; Jiang et al. 2016). Jiang et al. (2016) estimated a large sink in China’s terrestrial biosphere, in the range of 0.39–0.51 PgC year−1 for the period 2006–2009, which agrees well with the ensemble mean CO2 sink of 0.46 PgC year−1 estimated by Thompson et al. (2016). Strong regional CO2 sink over a short period of time is in no apparent contradiction with our state-of-the-art knowledge, when all the component fluxes have uncertainties. However, we are unable to propose a simple biospheric mechanism for increasing CO2 uptake systematically by 0.26 PgC within a decade in East Asia, from a mean flux of −0.20 PgC year−1 in 2000–2002 to −0.46 PgC year−1 in 2009–2011. The CO2 uptake increase of similar amount is also calculated over a longer time period, from 0.08 ± 0.14 (mean ± 1−
σ for IAVs) PgC year−1 during 1993–2002 to 0.36 ± 0.14 PgC year−1 during 2003–2012. The corrected CO2 uptakes for the same periods are calculated as 0.08 ± 0.14 and 0.03 ± 0.16 PgC year−1, respectively, using the CH4-scaled FFC CO2 emissions.
The TRENDY S2 simulation, including CO2 fertilization and climate effects, produced CO2 uptakes of 0.09 ± 0.11 and 0.13 ± 0.10, respectively, for the periods 1993–2002 and 2003–2012 (Fig. 8). The TRENDY simulations showed excellent agreement with inversion results for a slower rate of FFC CO2 emission increase for both the interannual variations and sink magnitudes, except for magnitudes during 2010–2012. These CO2 uptakes are also in good agreement with the average biomass carbon sink of 0.17 PgC year−1 during 1999–2008, estimated based on forest stands inventory in China (Zhang et al. 2013), or the net carbon sink in the range of 0.19–0.26 PgC year−1 using inventory, biogeochemical models and inverse models (Piao et al. 2009).
The anomalously high East Asian emission of ~0.2 PgC year−1 during 2010–2012 is likely to arise from an underestimation of the rate of FFC CO2 emission increase. Although the Chinese economic growth showed no sign of slowing down, the CDIAC estimated FFC CO2 emission showed sharp decrease in the rate of emission increase since 2011 (Fig. 5). As per the Chinese economic growth, a proportional increase in FFC CO2 emission would raise the emission by 0.3 PgC in the period 2009–2012. The proposed emission increase is consistent with one of the scenarios in Korsbakken et al. (2016), i.e. the National apparent consumption (pre-third National Economic Census, NEC) and that of the IEA. The a posteriori CH4 emissions also show continued increase during the years 2010–2012 (Fig. 2a, blue line). The use of FFC CO2 emissions from IEA show excellent agreement between corrected land sinks and the TRENDY simulation for the period 2010–2012, but produced a too strong sink increase during 2005–2009 (Fig. 8). Resolving the rate of increase in FFC CO2 emission from China for different time segments is beyond the scope of this work. We believe large investment is required to develop independent MRV system involving atmospheric measurement and model for tracking CO2 emissions from industrial and biospheric activities.