Introduction

Humans have for centuries been changing the composition of the Earth’s atmosphere, leading to significant climate change and air pollution, but the process has been particularly rapid since the 1950s1. To avoid the serious threat to the environment posed by exponential growth of greenhouse gas emissions, the international community has tried for 20 years to reduce global carbon emissions through sovereign state-level negotiations. One critical issue in these negotiations is to differentiate the historical responsibility for climate change and make a fair emission reduction program for different countries. Previous attribution studies of responsibility for climate change2,3,4,5,6 have been based on production-based emissions, i.e., accounted for using territorial boundaries7. These production-based emissions allow for convenient monitoring and regulation. However, international trade creates a geographic separation between the product’s final consumers and the carbon emitted in the production process, effectively shifting the CO2 associated with their consumption to distant lands8,9,10,11. This challenges the traditional principle of “the polluter pays”. One way of rectifying this problem is that responsibilities for climate change should be attributed in accordance with consumption-based accounting of carbon emissions that is defined as adding the emissions associated with imports and subtracting the emissions associated with exports, from production-based emissions12,13,14. Therefore, it has been argued that the current production-based carbon emission inventories should be replaced by consumption-based system in formulating emission reduction policies in post-Kyoto frameworks15,16,17.

Carbon emissions embodied in international trade rose by ~38% from 1990 to 200818 and the trend has continued in recent years11,19,20,21, motivating our quantification of its impact on the climate change attribution and responsibilities for mitigation. We present estimates of the role of carbon emissions in international trade using both a simple model (allowing calculations over a longer time interval) and three state-of-the-art Earth System Models (limited to a shorter study period by data availability) and explore the potential impact of transferred carbon emissions on the Kyoto Protocol (KP)22. We believe that the results will be useful for international negotiations in the future.

Results and Discussions

Carbon emissions via international trade potentially reduce the gap in historical responsibilities for CO2 loading between developed and developing countries. To investigate the influence of transferred carbon on historical responsibility for climate change, four experiments were designed and executed with a simple CO2 Impulse Response Function (IRF) model and three Earth System Models (Methods). The experiments are (i) PAX1: production-based CO2 emissions only allowed from developed countries (i.e., Annex I countries); (ii) PNX1: production-based CO2 emissions only allowed from developing countries (i.e., Non-annex I countries); (iii) CAX1: consumption-based CO2 emissions only allowed from the developed countries; and (iv) CNX1: consumption-based CO2 emissions only allowed from the developing countries. The simulations show that atmospheric CO2 concentrations would increase by 8.6–14.8 ppm (11.2–12.5 ppm) from 1990 to 2005 on conditions that production-based CO2 emissions are only allowed from the developed (developing) countries (Fig. 1). If consumption-based CO2 emissions are only allowed from the developed countries or the developing countries, atmospheric CO2 concentrations show an increase of 9.6–15.6 ppm and 8.9–11.5 ppm, respectively. Therefore, over the period 1990–2005, 0.8–2.3 ppm CO2 was transferred from the developed world to the developing world via international trade. This indicates that 3–9% of responsibility for the increased atmospheric CO2 concentration was shifted from the developed countries to the developing countries between 1990 and 2005 based on the normalized proportional method5(Table 1). These results suggest that transferred carbon emissions reduce the difference in historical responsibilities for CO2 loading between the developed and the developing countries, though these amounts are small.

Table 1 Contributions of the developed (AX1) and developing (NX1) countries to the rise in atmospheric CO2 concentration and the effectiveness of AKNP and AKNC scenarios (see text for definition).
Figure 1
figure 1

Temporal evolution of the simulated atmospheric CO2 concentration changes relative to 1990 using (a) IRF, (b) CESM, (c) BNU-ESM and (d) FGOALS-s2 under the PAX1, PNX1, CAX1 and CNX1 scenarios. (b) Shading shows the range of CO2 changes due to different initial conditions and lines are the ensemble mean.

Over the longer period (1990–2012), carbon emissions via international trade resulted in an increase of CO2 by ~1.4 ppm and hence a shift of historical responsibility by ~4% based on the IRF model. These numbers are quite similar to results for the 1990–2005 period from the Earth System Models. It should be noted that transferred carbon emissions account for a considerable proportion in production-based emissions for some regions and countries (e.g., China, USA and EU28). Based on the IRF model, carbon transfer (1990−2012) leads to a migration of CO2 by ~1.08 ppm (accounting for ~17.2% of CO2 rise that results from consumption-based emissions) from other countries to China, whereas a transfer of CO2 by ~0.33 and 1.17 ppm (4.6% and 19.3%) from USA and EU28 to other countries, respectively (Supplementary Fig. S1).

As may be expected given the relatively small levels of CO2 involved, the climate system shows little response to the carbon transferred via international trade. The modeled warming of global atmosphere and oceans and the melting of sea-ice in Northern Hemisphere are similar under all scenarios between 1990–2005 (Fig. 2); also borne out by differences in initial conditions being comparable with differences between the experiments (Figs 1 and 2). If a longer history of trade was available then climate effects due to trade may be more discernable, although trade has only grown rapidly in recent decades. The amount of transferred carbon emissions and their proportion in global total carbon emissions are gradually increasing18,19,21 and so traded carbon is likely to become more significant in future.

Figure 2
figure 2

Temporal evolution of annual mean surface air temperature, upper ocean heat content (0–700 m) and Northern Hemisphere sea ice fraction simulated by CESM (left panel), BNU-ESM (middle panel) and FGOALS-s2 (right panel) under the PAX1, PNX1, CAX1 and CNX1 scenarios.

Left panel: shading shows the range of values due to different initial conditions and lines are the ensemble mean.

Transferred carbon emissions will, to an extent, affect the effectiveness of the Kyoto Protocol. To investigate the impact of transferred carbon on the KP, we construct three CO2 emission pathways for 1990–2005 depending on whether carbon transfers are allowed between the developed and the developing countries while following CO2 mitigation protocols in KP (Methods). Under the scenario that the developed and the developing countries ignore their pledges and follow their production-based emissions (APNP; equivalent to the CMIP5 historical experiment), the simulated CO2 concentration in 2005 is 23.5–30.6 ppm higher than in 1990 (Fig. 3). If the developed countries follow the KP and the developing countries pursue their production-based emission (AKNP; as is specified by the KP), the increase of CO2 is simulated as 23.2–30.8 ppm. Therefore, actual global carbon emissions are seemingly in keeping with the KP. When the developed countries follow the KP and the developing countries pursue their consumption-based emission (AKNC; equivalent to the KP without counting carbon transfers from the developed to the developing countries), simulated CO2 concentration increases by 22.4–29.2 ppm, 0–1.6 ppm less than that simulated by AKNP. We now define the relative change of CO2 concentration under a mitigation scenario (i.e. AKNP or AKNC) to that under the observed emission scenario (i.e. APNP) as the effectiveness of the mitigation scenario. These 1990–2005 simulations indicate that the effectiveness of AKNP and AKNC is −0.7–8.9% and 4.6–8.9%, respectively (Table 1). This result indicates that the trade between the developed countries and the developing world contributed up to 5.3% of CO2 concentration increases from 1990–2005 (Table 1). This is the contribution to CO2 rise from items actually used in the developed world but which were produced in the developing world and hence escaped the limitations of the KP. Over the whole first commitment period of KP (1990–2008) simulated by the IRF model, 3.7% of CO2 increase can be similarly attributed. The accumulated sum of transferred emissions (0–1.6 ppm) from 1990 to 2005 is, however, small: less than the annual increase of CO2 (~1.7 ppm/yr) over the same period. The climate system hence shows little response to the transferred emissions (Fig. 4). Overall, the effectiveness of the Kyoto Protocol may have been potentially increased during 1990–2005 if the transferred carbon emissions are taken into account, though the resulting CO2 concentrations reduction and climate responses are tiny.

Figure 3
figure 3

Same as in Fig. 1 but under the APNP, AKNP and AKNC scenarios.

Figure 4
figure 4

Same as in Fig. 2 but under the APNP, AKNP and AKNC scenarios.

Numerical simulations with the IRF and three Earth System Models reveal that including carbon in international trade reduces the gap of historical responsibilities between the developed and the developing countries and the effectiveness of the KP. Although the climate change caused by the transferred carbon emissions (1990−2005) is almost negligible, the climatic effects of embodied emissions is expected to be more profound in future as global trade appears set to continue to grow. International trade also results in transfer of polluting gases which has additional environment and health hazards to the regions where goods are produced. For example, we estimate that the developed countries transferred 2.26 teragrams of SO2 to the developing world in 1990, which grew to 3.28 teragrams by 2005 (Supplementary Fig. S2). In addition, international trade potentially increases global carbon emissions as carbon-intensive manufacturing in emerging countries (e.g., China) entails more carbon emissions than would making the same product in the developed (importing) countries23. Given continuous growth of transferred carbon emissions and likely more significant impact on climate change, future climate negotiations should take into consideration embodied emissions in international trade. This entails accurate national carbon emissions accounting24 and implementation of incentives to make a feasible, fair emissions reduction policy.

It is undeniable that international trade affects global carbon emissions, air pollution and countries historical responsibility by redistributing emissions related to production of goods and services. But countries with net exports profit while bearing the extra climatic and environmental burden. Whether the profits compensate for the damage, especially over the long run, is still an open question which has many other dimensions and cannot be properly addressed by simple measurement or models.

Methods

Model Description

We use a CO2 Impulse Response Function (IRF) and three Earth System Models that have participated in the Coupled Model Intercomparison Project Phase 5 (CMIP5). The IRF is used to calculate CO2 concentration by a sum of exponentially decaying functions, one for each fraction of the additional concentrations, which should reflect the time scales of different sinks25.

where is CO2 concentrations, is a constant and set to approximately 0.47 ppmv/GtC, is the emission of CO2, is the atmospheric exponential decay time of the sth fraction of the additional concentration CO2 (171.0, 18.0 and 2.57 years), is the first fraction (0.152) and is the respective fraction (0.253, 0.279 and 0.316). The coefficients are based on the impulse response of the Bern model26 as used in the IPCC–SAR and IPCC–TAR.

The three Earth System Models are the Community Earth System Model (CESM)27, the Beijing Normal University-Earth System Model (BNU-ESM)28 and the Flexible Global Ocean-Atmosphere-Land System model (FGOALS-s2)29. Each of the three Earth System Models contains an interactive carbon cycle module in the land component and an ecosystem-biogeochemical module in the ocean component. The simulated atmospheric CO2 concentrations are fully coupled to the land and ocean surface CO2 fluxes and are used directly to compute the radiative forcing, hence forming a complete carbon cycle process. In this study, the atmospheric horizontal resolution of the CESM, BNU-ESM and FGOALS-s2 is ~0.9° × 1.25°, ~2.8° × 2.8° and ~2.81° × 1.66°, respectively. The ocean component has a nominal 1° resolution for the CESM and 1° × 1° for the BNU-ESM and FGOALS-s2.

Experimental design

Two groups of numerical experiments were designed to investigate the influence of transferred carbon emissions. In group I, the influence of transferred carbon emissions on historical climate change is examined. We design four scenarios in which production-based/consumption-based emissions are allowed only from either the developed countries or the developing countries (Supplementary Table S1). The production-based carbon emissions fluxes are available at 1° × 1° spatial resolution from 1751 to 1949 at annual resolution and from 1950 to 2007 at monthly resolution30. The national inventories of consumption-based carbon emissions18 cover 113 regions and extend from 1990 to 2008. We use the regional distribution of production-based carbon fluxes to construct gridded consumption-based carbon fluxes at monthly and 1° × 1° spatial resolution. The cumulative transferred carbon emissions in the developed and the developing countries are shown in Supplementary Fig. S3. The CESM was first integrated over the period of 1850–1990 under the PAX1 and PNX1 scenarios (Supplementary Table S1), respectively. Initialized from the year at 1990 in the PAX1 (PNX1) experiment, CESM was further run from 1990 to 2005 under the PAX1 and CAX1 (PNX1 and CNX1) scenarios. Other forcings varying over the historical period (1850–2005) include CH4, N2O, halocarbons, aerosols, solar irradiance and volcanoes. The same method is used for the BNU-ESM and FGOALS-s2. Note that we run all the sensitivity experiments (groups I and II) with three different initial conditions using CESM. For the IRF, we construct four time series of carbon emissions (1850–2012) based on the designed emissions scenarios (Supplementary Table S1).

In group II, the effect of mitigation of production-based and consumption-based counting on the KP is investigated. We assume that each developed country decreases (or increases) its annual carbon emissions linearly and achieves its reduction commitment in 2008 according to the KP—whose purpose is reducing the overall emissions of anthropogenic greenhouse gases of the developed world by at least 5% below the 1990 levels in the commitment period from 2008 to 2012. We construct three emission inventories (Supplementary Table S1) for each developed country from 1990 to 2005 at monthly and 1° × 1° spatial resolution (Supplementary Fig. S4). For the CESM, BNU-ESM and FGOALS-s2, each model is first integrated over the period of 1850–1990 under the scenario that all countries follow their production-based carbon emissions (equivalent to the CMIP5 historical experiment). Starting from the end of this experiment, each model was then run from 1990 to 2005 under the APNP, AKNP and AKNC scenarios (Supplementary Table S1). For the IRF, we create three time series of carbon emissions (1850–2008) based on the designed emissions pathways (Supplementary Table S1).

Additional Information

How to cite this article: Wei, T. et al. Quantitative Estimation of the Climatic Effects of Carbon Transferred by International Trade. Sci. Rep. 6, 28046; doi: 10.1038/srep28046 (2016).