Carbon footprint of global natural gas supplies to China

As natural gas demand surges in China, driven by the coal-to-gas switching policy, widespread attention is focused on its impacts on global gas supply-demand rebalance and greenhouse gas (GHG) emissions. Here, for the first time, we estimate well-to-city-gate GHG emissions of gas supplies for China, based on analyses of field-specific characteristics of 104 fields in 15 countries. Results show GHG intensities of supplies from 104 fields vary from 6.2 to 43.3 g CO2eq MJ−1. Due to the increase of GHG-intensive gas supplies from Russia, Central Asia, and domestic shale gas fields, the supply-energy-weighted average GHG intensity is projected to increase from 21.7 in 2016 to 23.3 CO2eq MJ−1 in 2030, and total well-to-city-gate emissions of gas supplies are estimated to grow by ~3 times. While securing gas supply is a top priority for the Chinese government, decreasing GHG intensity should be considered in meeting its commitment to emission reductions.


Supplementary Tables
Supplementary Table 1 Comparison of characteristics of natural gas pipeline system in different countries

Year of Pipeline installed
Operating pressure

Maintenance and mitigation practice
China's pipeline system 2000s 6.3~12 MPa No regulation and measurements on methane emissions U.S pipeline system Median at 1960s 3.5~10 MPa Well-maintained with deployment of emissions mitigation practices Russian gas export pipeline system Median at 1980s 7.5 Mpa Monitored leaks, promoted and deployed emissions mitigation practices Data sources: ANL 2007 37 , INGAA 38 , Lelieveld et al 14,15 , Balcombe et al 39 .

Natural gas extraction
Well drilling GHG emissions of well drilling arise from the emissions associated with diesel consumption of the drilling rig. The study uses the equation S1 to estimate the diesel consumptions of well drilling 40 : where, FC is the diesel consumption of drilling per well. RC is the drilling rig output power, which ranges from 1860 to 4920 kilowatt 40 . The drilling time T is calculated as the well measured depth D divided by the average penetration rate of the diesel drill PR, which varies from 14.6 to 16.8 meter per hour 40 . The measured depth D is the total length of the wellbore measured along the actual well path. For conventional vertical well, measured depth equals to the depth of the well (after subtracting water depth for offshore gas wells), while for the horizontal well of tight and shale gas extraction, measured depth D equals to well depth multiplied by the factor of 1.3 according to the practice experiences 40,84 .
Diesel consumption of drilling one well is then divided by the estimated ultimate recovery (EUR) per well to calculate the diesel consumption rate of well-drilling. The onsite and upstream GHG emissions associated with well-drilling diesel consumption is calculated as the diesel consumption rate multiplied by the emission factor of diesel combustion onsite and offsite production and transportation.
Field-specific values of well depth and EUR are used in the calculation. The onsite and upstream emission factors associated with diesel are obtained from the GREET ® 2018 60 .

Well completion and well workover
Well completion is the preparation process for gas production after well drilling 25,26 . Well workovers are the occasional operations for well cleaning and maintenance that happen during the production life of well 25,26 . The study adopted the method from the National Energy Technology Laboratory's (NETL) natural gas LCA model to calculate the GHG emissions of well completion and well workover. For well completion of conventional and coal bed methane (CBM), the potential gas venting (before flaring) per well is estimated as 1.05 and 1.40 thousand cubic meter, respectively 25,26 . For tight and shale gas well, because higher potential emissions arise from the additional hydraulic fracking and higher flow-back period, the estimation of well completion emissions would have significant impacts on the final LCA results. Here, we considered the effects of individual heterogeneity and calculate the potential emissions of well completion of tight and shale gas as the field-specific initial production rate multiplied by the duration of flow-back period 25,26 .
Not all wells require workover, while for some wells, workover can happen more than one time during the production life. The study adopts NETL's assumptions of average workover rate for different types of gas wells. The potential emissions of workover per episode for a conventional and CBM gas well are assumed as 0.069 and 1.40 thousand cubic meter, respectively 25,26 . The same estimation method is applied to estimate the potential emissions for tight and shale gas as the product of the initial production rate and flow-back period.
The potential emissions of well completion and well workover are then divided by the field-specific estimated ultimate recovery (EUR) to calculate the potential emission rate of well completion and workover. The potential emissions will be further collected for flaring or just venting depending on the flaring and venting practice of the well, which will be discussed in the Section of Fugitive emission below.

Liquid uploading
Liquid uploading is the practice to remove water and other condensates from gas wells. Since there is no field-specific information about liquid uploading practices in China from both public and commercial database, we followed NETL's study of 2014 version to assume that one episode of liquid uploading released 102 cubic meter of natural gas, and a total of 930 unloadings happened over the well life 25,26 . The lack of field-specific data would increase the uncertainty of the final results considering the potential existing heterogeneity of liquid uploading among individual fields 85 , which can be a research focus in the future study.
Similar as the emissions from well completion and workover, the potential emissions of liquid uploading can be collected for flaring, and the emission rate associated with liquid uploading is calculated as the episode emissions after flaring divided by field-specific EUR. open-ended lines, pneumatic device, water tank, etc. Leakage associated with one type of device for a gas well production per day is calculated as the corresponding leakage rate of the device per hour multiple by the number of device connected and the total operational hours per day (24hr). Per day leakage of the device is then divided by the average production rate of the gas well to calculate the average leakage rate.
The average production rate of well is a field-specific parameter. The total leakage rate is the sum of leakage of all devices.

Gas gathering
At the wellhead, engine-driven compressors collect gas from multiple production wells and then transport gas to the processing facility. We follow NETL's assumption that the compressors' inlet and outlet pressures are 0.35 and 5.5 Mpa 25,26 , respectively, and then use the engineering model from the Oil Production Greenhouse gas Emissions Estimator (OPGEE) 28 to estimate the energy consumption for gas gathering at the wellhead.
Key parameters used to calculate emissions associated with gas extraction stage are listed in Supplementary Table 4.

Acid gas removal
Acid gas removal (AGR) is a process to remove the H2S and CO2 content in the raw gas to meet the pipeline gas quality or liquefied natural gas (LNG) specification. The most common technique for acid gas removal is amine treating, which removes the H2S and CO2 by amine absorption and chemical reaction. After absorption, the amine solution would be sent to a reboiler to separate the acid gas and regenerate amine. The separated acid gas (with a portion of CH4 absorbed from raw gas) is flared before venting to the atmosphere. GHG emissions of the acid gas removal process arise from the fuel consumption for reboiler and the residue gas after acid gas flaring. The fuel consumption of the amine reboiler is calculated by the parameter of heat duty, heat efficiency of the reboiler and the amount of amine consumed, which is determined by the acid gas content in the raw gas and the output gas quality requirement.
H2S and CO2 content of the raw gas are field-specific inputs for the calculation while the output gas quality requirement is determined by the industry standards. According to the Chinese standard 86 , for pipeline gas, CO2 content must be less than 3.0 vol% (equivalent to 0.47 wt%, which is the same as the U.S.' requirement 25,26 ), and H2S content must be less than 20 mg/m 3 . LNG has a more strict specification of less than 0.01 vol% CO2 and 0.0004% H2S 87 , for preventions of freezing and corrosion of the cryogenic liquefaction facilities.
We assume the reboiler use onsite-generated natural gas as fuel, and the GHG emission factor of natural gas combustion is then calculated based on its composition and carbon content, following the same calculation method in GREET ®60 . The consumption of onsite-generated natural gas also leads to feedstock loss which is tracked in the LCA model.
According to Chinese natural gas industry standards, the separated acid gas must be flared before venting to the atmosphere 54 . GHG emissions after flaring are calculated by the amount of CH4 and CO2 absorbed in the amine solution and the flaring efficiency.

Dehydration
Dehydration is the process to use glycol to absorb and remove the water content in the raw gas to meet the transportation and end-use requirement. Similar to the acid gas removal process, GHG emissions arise from the fuel consumption for the glycol-solution-reboiler and the residue gas after flaring.
The fuel consumption is a function of reboiler duty, glycol flow rate and the amount of water removed, which is determined by the mass balance of the raw gas water content and pipeline-required water content.
We adopted NETL's assumption for the water content of raw gas as 794 mg/m 3 . According to the Chinese natural gas industry standard, the pipeline-required water content is determined by the operational temperature and pressure of the connected pipeline of each gas field 86 . Supplementary Table 5 listed the pipeline water content requirement for major natural gas pipelines in China. The water content requirement for LNG is less than 0.1 ppm to avoid ice crystals from forming 87 .

Natural gas liquid (NGL) separation
NGL separation is the process of removing the non-methane hydrocarbons of natural gas (also referred to as NGL). Not all natural gas processing plans have included the NGL separation and we include a parameter in the LCA model to determine whether the corresponding processing plant of a specific field has included NGL separation or not.
GHG emissions arise from the fuel (onsite-generated natural gas) and electricity input during the NGL separation. We adopted NETL's assumption on the electricity and fuel consumption rate for the NGL separation, and used emission factors from GREET ® to convert energy consumptions to GHG emissions. 90% of C2H6 and 100% of C3H8 and C4H10 are separated in the process 28,89 , which are treated as biproducts for the LCA analysis. We allocate GHG emissions between well extraction and NGL separation to the main product (natural gas) and NGL according to their corresponding energy contents.  Table 6). We also adopted the leakage rates of gas processing devices from NETL's model to estimate the leakage emissions.

Gas compression
As mentioned above, we assume the pressure of gas gathering pipeline is 5.5 Mpa. Additional gas compression is required for gas fields connected with higher-pressure pipelines, such as the Chinese West-East gas pipeline, of which the operating pressure is 10 Mpa due to the extremely long-distance transmission. We used the calculation method from OPGEE 28 to estimate the energy consumption and GHG emissions associated with the additional gas compression.
Supplementary Table 6 listed the key parameters used to calculate emissions associated with gas processing. GHG emissions associated with the natural gas combustion are calculated based on the NG use intensity, the carbon content of the feedstock NG and the combustion oxidation rate of different type of compressors, following the similar calculation method in GREET ®60 . The amount of natural gas used for combustion is deducted from the feedstock before entering to the next processing unit.

Liquefied natural gas storage and shipping
GHG emissions associated with LNG storage and shipping derive from the LNG boil-off and fuel consumption of LNG shipping. The boil-off effect is the phenomenon when the ambient heat is input to the cryogenic fluid to create vapors. In the study, we followed the calculation method in GREET ®60 to calculate the fugitive emissions due to the LNG boil-off, which is the function of the factor of boil-off rate per day, recovery rate of boil-off gas, and the duration of storage and shipping. The duration of LNG ocean shipping is calculated basing on the shipping distance from each LNG exporting terminal to receiving ports in China.
We adopted the method in GREET ®60 to calculate the combustion emissions associated with fuel consumption of LNG ocean shipping. LNG ocean tanker uses the transported LNG as fuel for shipping from origin to destination, and use heavy fuel oil for shipping back to the origin. The shipping-associated emissions are calculated based on fuel consumption and the corresponding combustion emission factors of LNG and heavy fuel oil. Similar to the feedstock natural gas combustion scenarios mentioned above, the CO2 emissions factor of the LNG fuel is calculated based on the carbon content of LNG. LNG combustion also leads to the feedstock loss, which is modeled specifically for each supply pathways of individual gas fields. The combustion and upstream emission factors of heavy fuel oil are obtained from GREET directly 60 .
Key parameters used to calculate emissions associated with LNG storage and ocean shipping are presented in Supplementary Table 9.

Regasification
Limited studies have reported GHG emissions for LNG regasification process, and data from those studies are generally outdated and aggregated without complete details [90][91][92]

Estimation of the missing value of fields-specific production profile
Field-specific production profile is important for the estimation of GHG emission intensity of different gas fields. In the study, EUR, average production rate and initial production rate (for tight and shale gas) are three field-specific production parameters input to the natural gas LCA model.
We gathered field-specific production data from industrial reports, journal publications, books, etc.
Detailed data sources for each gas fields are presented in Supplementary Data 1. Although a wide range of publications has been searched, there are still multiple missing data in the production profiles. For the missing production data for gas fields outside of China, commercial data from Wood Mackenzie 94 are used to fill the gap, which is colored in orange in Supplementary Data 1. For Chinese domestic gas fields, due to the limited data in both public and commercial database, usually only initial production rate or the stable production rate at the early stage can be found, while the data for EUR and average production rate over the well life is incomplete. We use the following assumptions and methods to estimate the missing data of EUR and the average production rate for Chinese domestic natural gas fields:

Chinese conventional gas
According to the previous research 95

and management documentation released by China National
Petroleum Corporation (CNPC) 96 , the common production mode for a high permeability conventional gas well in China is to maintain a stable production rate for approximate 10 years (by controlling well pressures), and then the production rate declines till the end of well life. By assuming a 30-years of production life, the EUR of a conventional natural gas well is thus estimated as the sum of cumulative gas production of the 10-years stable period and 20-years declining period. Arps's exponential production decline equations are used to estimate the cumulative production for the latter 20 years 97 : where is the production rate at the time of t for the decline period, which is the 10+t year from the beginning of well production. D is a constant parameter. is the initial production rate. Q is the total cumulative production of the 20 years of the decline period. By assuming that the production rate at the end of well life is 0.01 of the initial production rate, EUR can be estimated as a function of the well's stable production rate at the early stage. And the average production rate per is calculated as the EUR divided by 30 years and 365 days.

Chinese unconventional gas
Wei et al. 98  and Qin 100 , we assumed that the production curve of other shale gas wells in China have a similar shape as that for the typical shale gas well in Sichuan basin, and the missing EUR for other shale gas wells can be estimated by scaling the EUR of the typical shale gas well: Therefore, the missing EUR value of other shale gas wells can be calculated with the field-specific initial production rate. And the missing average production rate is calculated as the EUR divided by the total production days of the well life. A similar method is used to estimate the missing EUR and average production rate of tight gas wells in China by scaling the production data of the typical tight gas well in Ordos basin.
The estimated value of EUR and the average production rate for Chinese domestic gas wells are colored with green in Supplementary Data 1.

Supplementary Discussion 1: Sensitivity analysis
Due to the uncertainty of parameters input to the well-to-city-gate natural gas LCA model, we conduct a sensitivity analysis to exam how the uncertain input would affect the robustness of results. The well-tocity-gate natural gas LCA model contain 88 uncertain input parameters, including factors with heterogeneity of individual gas fields (e.g. raw gas composition, EUR, pipeline transmission distance, etc), as well as those are not varied with gas fields (e.g. compressor prime mover fuel use rate, boiler heat efficiency, LNG boil-off rate, etc). We vary one parameter each time while maintaining other parameters constant to evaluate the sensitivity of each parameter on the estimation of GHG emission intensity for different gas fields. For each simulation, we increase the value of a parameter by 20% unless there is a fixed upper limit for the parameter and then decrease 20% instead (e.g., 100% is the upper limit for flaring rate). To facilitate the comparison and identifications of the sensitive parameters, we calculate the elasticity of each parameter i to the estimation of the GHG emission intensity of gas filed j: where, is the GHG emission intensity of gas field j, is the value of parameter i. 0 and ̂ represent the estimation of GHG emission intensity for gas field j before and after the variation of parameter . 0 and ̂ represent the value input of parameter before and after variation.
In general, parameters' sensitivity varies for different gas fields while gas fields of the same category show a certain similarity. Technical parameters of compressors (e.g. compressor efficiency, prime mover fuel use rate of compressor engine) are sensitive parameters for gas fields of all categories because of the extensively used and energy-intensive of compressors throughout the natural gas supply chain. Parameters associated with pipeline transmissions (i.e. pipeline transport distance, energy intensity rate of pipeline transmission, pipeline leakage rate) are sensitive for carbon intensity estimation of pipeline sources gas fields, especially for international gas fields and domestic gas fields with extremely long transmission distance (e.g. Dina, Kela, Yakela, Yingmaili, Tazhong gas fields). The parameter of pipeline transport distance has the highest elasticity (~1.1) on the estimation of emission intensity for pipeline gas supply from Turkmenistan, Uzbekistan and Russia. The parameter of CO2 composition in raw gas is among the highest sensitive parameters for gas fields with high CO2 content (e.g. Dongfang, Ya, Wenchang, Liuhua, For most of the sensitive parameters mentioned above, we have conducted detailed analysis and identified the heterogeneity of parameters among individual gas fields, which would significantly reduce the uncertainty of the LCA results. For instance, raw gas composition, initial production rate, EUR, pipeline transport distance are field-specific input to the LCA model. These highly-sensitive and individualspecific parameters contribute most to the variability of emission intensity among different gas fields. Other parameters, for instance, compressor efficiency, energy intensity rate of pipeline transmission, pipeline leakage rate, LNG boil-off rate and recovery rate of boil-off gas, are set as constant for estimations across gas fields because no sufficient data suggests their variation among individual gas fields. Although we have referred to established model (e.g. GREET model, NETL's natural gas LCA model) and reliable data sources (e.g. EPA GHG inventory) to estimate the value of these parameters, the uncertainty inevitably exists in these parameters and the final LCA results. In the next section, we will further explore the uncertainty ranges of these parameters and thus estimate the overall uncertainty in the final LCA results. Tornado diagrams in Supplementary Fig. 4-7 show the variation of carbon intensities derived from uncertain inputs of top 5 parameters for four typical gas fields of the four gas categories

Supplementary Discussion 2: Uncertainty analysis
The above sensitivity analysis has identified parameters that have significant effects on the final results.
In this section, we will estimate the probability distributions for those sensitive parameters and then conduct Monte Carlo simulations to calculate the uncertainties of final results.
To have a clear understanding of the uncertainty of pipeline leakage rate and its impacts on our final results, here, we conduct a literature review on the pipeline leakage studies and incorporate the uncertainty of the pipeline leakage rate into the Monte Carlo simulation to generate robust results.
The majority of studies on pipeline transmission leakages were centered on the U.S. and Russia, which are the two largest natural gas producers that account for ~40% of gas production worldwide 39 . Pipeline leakage rates are also measured and reported in some European countries, such as United Kingdom, Germany and the Netherlands 30,104 , but since China does not import natural gas from these European countries, studies in those countries will not be discussed here. There is no publicly available report on the leakage rate of China's pipeline system. Russia's export corridors to Western European. The length of the export corridors is ~3000km within the Russian border (the distance of Northern Corridor is 3075km while the Central Corridor is 3376km) 13,15 .
The average transmission distance of the U.S natural gas system was estimated at ~1000km 25,26,60,111 . The much longer transmission distance of Russia's natural gas system leads to its higher leakage rate in terms of percentage throughput than that of the U.S. system in literature 14,15,39 . Yet on the per km basis, pipeline leakage rates of the two systems are comparable and their uncertainty ranges overlap mostly, as shown in Supplementary Fig. 3.
The difference in the estimations of pipeline leakage rate is caused by many reasons. Firstly, varied measurement approaches would result in different estimates 62  Secondly, characteristics of the pipeline systems, such as pipeline age, operating pressure, level of maintenances and mitigation practices, would affect the pipeline leakage rates 14,38,39,109,110 . Previous researches revealed that old compressor stations might lead to higher leaks, and pipeline system with higher operating pressures has the tendency to increase leaks, while well maintenance and mitigation measures can significantly reduce transmission leakages 14,38,39,109,110 .
Supplementary of the pipeline incidents are in irrespective of the pipeline age, with just 15% of the incidents related in some way to the pipeline age, and they concluded that periodically assessment, timely repairs with mitigation efforts can ensure aged pipeline's continued fitness for service 38 .
After carefully comparing the characteristics of the Chinese pipeline system to those of the U.S. and Russia, we are not expecting the Chinese pipeline leakage rate to be significantly higher or lower than that of the U.S or Russia. Actually, literature estimates of pipeline leakage rates for the U.S. and Russia have wide uncertainty ranges, which should be large enough to capture the possible value for the Chinese pipeline leakage rate. According to Lelieveld et al, the true value of leakage rate for Russia's gas export transmission system must be lower than their upper estimates of 1.6% (~7.1E-06 kg/kg-km, the highest estimate for Russia in Supplementary Fig. 3), because the upper estimation is calculated based on "worstcase assumptions" in numerous areas 14 . Balcombe et al believed that estimates of the leakage rate of pipeline transmission (exclude leakages from production and processing) above 1.6% are results of outdated data or flawed estimation methods 39 .
In the study, we apply the uncertainty estimates of Russia's pipeline leakage rate (which has a larger uncertainty range than that of U.S, as shown in Supplementary Fig. 3) to that of the Chinese pipeline system and the connected international pipeline system, such as the Central Asia-China pipeline, and conduct Monte Carlo simulation for uncertainty analysis.

Uncertainty of other sensitive parameters
Besides the transmission leakage rate, variations of some other parameters also have significant impacts on our LCA results. According to previous sensitivity analysis, parameters with elasticity greater than 0.1 (which means the final results would change more than 1% with 10% change in the parameter) include: CO2 content in raw gas, average production rate, initial production rate, EUR, pipeline transport distance, flaring rate at extraction stage, well completion flowback period, compressor efficiency, energy conversion factor of NG engine prime mover, energy conversion factor of NG turbine prime mover, flaring efficiency, energy intensity rate of pipeline transmission, LNG boil-off rate, recovery rate of boil-off gas, ocean tanker average speed. Among these parameters, CO2 content in raw gas, average production rate, initial production rate, EUR, pipeline transport distance are field-specific inputs, of which the individual variations have been analyzed thoroughly. Here we assumed ±10% variation for these field-specific inputs, except for the EUR, of which ±50% of variations have been assumed because of the relatively high uncertain feature of the parameter. For other sensitive parameters, we estimate their uncertain range through literature reviews, as shown in Supplementary Table 2.

Monte Carlo simulation
With the probability distributions of the sensitive parameters, we conducted 5000 times Monte Carlo simulations to calculate the uncertainty ranges of the final results. Well-to-city-gate GHG intensities with 90% confidence interval (CI) for each gas fields are presented as error bars in Fig.1, Fig.2, Fig. 4 in the main manuscript and in Supplementary Fig. 1, 10 and 11. Supplementary Fig. 8 and 9 show the probability density function (PDF) of the well-to-city-gate GHG intensity of gas supply from Galkynysh field.

Supplementary Discussion 3: Economic analysis
The present study focuses on providing climate-wise choices for China to minimize GHG emissions for its growing natural gas supply. One important question remains regarding the economic cost and its effects on China's choice of global natural gas consumption. In the section, we compare the cost of supply for different gas sources and thus provide further insight into the economic implications of China's natural gas supply policy.

Overseas liquefied natural gas
The cost of overseas LNG supply includes the import price and other additional costs, such as the cost of regasification 23,112 . The LNG import price accounts for more than 95% of the total supply cost of overseas LNG 23,112 . The LNG import price is positively related to the global oil price and also affected by the market supply and demand conditions 113 . The growth of Spot LNG price can break through the trend of global oil price when there are strong demand and tight supply in the LNG market. Impacted by the fluctuating oil price and uncertain market conditions, LNG import price varies in a wide range. As shown in Supplementary Fig. 12, from January to October 2019, China's LNG import price varied between 280~420 us dollar (USD) per thousand cubic meters (kcm, on the basis of gas volume after regasification) 22 .

International pipeline gas
The import price of international pipeline gas is determined by the pricing formula in the long-term contract, which has never been disclosed by China National Petroleum Corporation. According to the published statistics of the import prices of international pipeline gas in recent years 114,115 , the price of international pipeline gas has positive relationship with the global oil price and an extra price increase poses when there are tight supply and strong demand. Generally speaking, the average import price of pipeline gas is 20%~30% lower than that of overseas LNG during the same period [113][114][115] . However, since the international pipeline gas is mainly imported from the western border that is distant from the eastern coastal metropolitan areas, the extra cost of long-distance pipeline transmission significantly increases the total cost of supply. According to the study by Rioux et al 112 , after including the cost of delivering gas across ~4000 km, the total cost of pipeline gas imports from Central Asia to Shanghai is similar to the coastal LNG import price. Supplementary Table 3 shows the import price of international pipeline gas in 2018 and the estimations of gas delivering cost within China for different import sources.

Domestic gas
Supply cost for domestic gas is relatively low but its production capacity is limited and far below China's prospective gas demand in the future 23,116,117 . The average production cost of domestic conventional gas varies between 20~110 USD/kcm in different provinces 23,112 , as shown in Supplementary Fig. 13. Even with the cost of transmission, the total cost of domestic conventional gas supply is still much lower than the average import price of LNG in recent years 114,115 . China is now actively promoting the exploration of unconventional gas, such as shale gas in Sichuan Basin 100,118-120 . The production cost of shale gas in literature is estimated to be 130~400 USD/kcm 23, 24,112 . With the cost of transmission, the total cost of domestic shale gas can be close to the gas import price.
Supplementary Fig. 14 compares the supply costs of different sources of gas supply to Shanghai. Due to the lack of public data from natural gas production sector and the involving international gas market, high uncertainty exists in the estimations of natural gas supply costs.
The comparison of supply costs for different categories of gas sources varies for different regions of China owing to the different production cost of domestic gas and relatively transport distance of gas imports.
More in-depth economic analysis is required regarding the heterogeneity between different gas sources and different regions of gas consumption, which can be the focus for our future studies.