Results and discussion Validation Since the exact consumption based

The 37 regions that are covered both by LCBA and the other 3 datasets are further analyzed. The relative errors for LCBA optimal values in production-based emissions and emissions transfers are calculated separately according to the 3 existed studies (positive indicates greater values than the LCBA). The absolute values of the 3 relative errors are then averaged to reflect the mean level for each of the 37 regions, and this 4E1RCat mean value is named as the “average relative error”. Although average relative errors can be greatly influenced by large values in the original 3 errors, as in the case of the US, the tendency can be roughly shown. It is found that the average relative errors for emissions embodied in trade are normally larger than those for production-based emissions. For most regions, the average relative errors in production-based emissions remain low, but those regions that are highly dependent on trade show abnormally large values, such as Singapore, Hong Kong and Netherlands. This can be explained by their huge values of trade in contrast to GDP in those economies. As for average relative errors in emissions transfers, most of the top emitters show moderate results (China, 15%; Russia, 18%; Japan, 30%; Germany, 37%). The abnormal value in Canada (1300%), for example, might be ascribed to the different calculation frameworks of the 4 models and different original trade data since they share similar production-based emissions and all 3 original errors are much smaller than 0. So does it to US (46%), of which the original errors vary from −9.3% to 63%. To further clarify these average relative errors, specific regions, namely US, China, India, Singapore, Hong Kong and 2 other groups (AX1 and NX1), are selected for illustration.