Breaking down barriers on PV trade will facilitate global carbon mitigation

The global trade of solar photovoltaic (PV) products substantially contributes to increases in solar power generation and carbon emissions reductions. This paper depicts global PV product trade patterns, explores emissions reduction potential, and evaluates the impeding effect of tariff barriers on global PV product trade and emissions reductions. Solar power generation will result in a reduction of emissions in a range of 50–180 gigatons of carbon dioxide equivalent (GtCO2e) between 2017 and 2060 in a business as usual (BAU) scenario. Compared with BAU, during 2017–2060, global total solar cell and module production and installation will increase by roughly 750 gigawatts (GW) if half of the status quo trade barrier are removed, while it will decrease by 160–370 GW under tensioned trade barrier scenarios. Trade barrier reduction by half from the 2017 status quo level will increase the net carbon emissions mitigation potential by 4–12 GtCO2e by 2060, while extra trade barrier imposition will result in global net carbon emissions mitigation potential decreasing by up to 3–4 GtCO2e by 2060. Well-coordinated policy and institutional reforms are recommended to facilitate PV product trade and to deliver the related global environmental benefits.

The paper abstracts from many other relevant factors-particularly going out to 2050. (1) How likely are today's trade patterns to hold over the next 30 years? Would China continue to have a dominant position on the markets irrespective of the trade wars? Would its aging population and rapidly rising middle-class price China out of the market to the benefit of competing producers? (2) To what extent is the trade war a temporary phenomenon and a function of the current political environment? Should the world's nations get serious about decarbonizing the global economy, wouldn't they seek policies that accelerate PV penetration? (3) What are the substitute technologies and how substitutable are they with PV? (4) What other factors will influence the future of PV? For example, the aforementioned R&D and product innovation.
There is some degree of sensitivity analysis with respect to the scenarios, but it would be useful to provide a more thorough and systematic sensitivity analysis as regards the key parameters: future trade shares, supply, demand and substitution elasticities.
The paper is a very difficult read. It is chalk full of numbers and figures (some of which are mislabeled), using more significant digits than necessary, and rarely contextualized-are the numbers large or small? and relative to what?
Reviewer #3 (Remarks to the Author): The manuscript entitled "Barriers on PV trade will impede global carbon mitigation and local pollutant emissions reduction" is focused on the emission potential reduction of global PV trade and the effect of extra trade barriers on PV product globally. The topic is very interesting and timely. However, crucial assumptions and data sources should be revised in the method section for the current and future scenarios under analysis. Below I list my specific comments/concerns.
In this study, PV life-cycle carbon emissions are obtained from various existing literature studies (some of them are not the latest available as detailed below in this report), reflecting the 2009-2013 silicon PV technology state of the art. In the developed future scenarios (from 2017-2050) the same PV carbon emissions are assumed. This assumption is a weak point because PV life-cycle carbon emissions are expected to decrease in the medium-long term mainly due to improvements in manufacturing processes. Specifically: (i) reduced silicon wafer thickness; (ii) reduced material usage (such as silicon, silver content in the metallization paste, copper), and kerf losses; (iii) electricity consumption reductions; (iv) PV module power conversion efficiency improvements.
As an example, please see the latest Fraunhofer report 2020, available at: https://www.ise.fraunhofer.de/content/dam/ise/de/documents/publications/studies/Photovoltaics-Report.pdf , which shows silicon usage reduction ranging from 150 g/Wp in 2013 to 80 g/Wp in 2019 as well as efficiency improvements in silicon-based PV module over the last years. Please check also the following IEA PVPS Task 12 Report, in which are listed potential future changes (such as reductions in kerf loss, wafer thickness, carbon emissions and power conversion efficiency improvements): Frischknecht, R., Itten, R., Wyss, F., Blanc, I., Heath, G.A., Raugei, M., Sinha, P. and Wade, A., 2015. Life cycle assessment of future photovoltaic electricity production from residential-scale systems operated in Europe (No. NREL/TP-6A20-73849). Available at: https://iea-pvps.org/keytopics/iea-pvps-task-12-life-cycle-assessment-of-future-photovoltaic-electricity-production-fromresidential-scale-systems-operated-in-europe-2015/ In this study, these potential changes are not discussed or included, such as providing a sensitivity analysis (or uncertainty analysis) for potential future emission changes due to production improvements. Also, it is not captured/discussed potential changes in the electricity grid mix compositions (such as increased penetration of renewable energies), which influence the amount of primary energy required for each PV production process as well as the associated carbon emissions.
Page 2 Lines 31-35 It is stated that "studies have been conducted varying from calculating the lifecycle emissions and emission reduction potential of PV products (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21) to PV product trade patterns and structures (22) to the emissions embodied in the bilateral PV product trade23 and to the environmental impacts of the PV product trade (13,14,24). Studies have also been carried out in various countries (15-21,24), concluded that PV power generation could indeed help to reduce carbon and local air pollutant emissions obviously." The authors cite here previous PV life cycle studies, 7 of which are focused on China, 1 in the UK, 1 in Turkey, and 1 in the United States, and most of them are based on crystalline silicon technology. Since the scope of this paper is "global" and it includes future scenarios, it would be appropriate to extend the existing background references, integrating them with other studies that are focused also on: (i) PV produced and/or installed in other geographical areas, (ii) comparison between PV life cycle carbon emissions with other electricity generation technologies (fossil fuels and renewables) in specific electricity grid mixes, (iii) complete PV systems, including balance of system (BOS) (structures, inverters, cables, transformers). (iv) the additional life cycle contribution of potential storage and/or batteries needed for future scenarios with large renewable electricity penetration.
(v) PV end-of-life life-cycle contribution.
Page 2 Lines 45-47 "previous studies concentrated on only a specific PV product and a limited subset of countries/regions without full consideration for the whole production and supply chain from a global perspective" This statement is not correct. Please check the following literature studies: Page 20 Lines 481-482 It is stated that "PV power generation is relatively clean and near zero-emissions in its application, but certain amounts of GHGs and pollutants are emitted during the PV products production process" Life cycle analysis takes into the account not only the production processes, but the product's full life cycle from the extraction of resources and the production of raw materials, to manufacturing, distribution, use and re-use, maintenance, and finally recycling and disposal of the final productincluding transportation and use of energy carriers. Please check the following references, which are focused on LCA guidelines and standardization as well as specific PV LCA guidelines: In the supplementary information section 3 the considered emission factors taken from the literatures are listed, and it is stated that "because regional LCA studies of PV products were quite limited, we adopted the latest and up-to-date LCA emission coefficients available to carry out embodied emission calculation".
However, Ecoinvent database version 3 does not provide directly the PV "emission factors", but Ecoinvent is a database with specific processes. Emission factors should be calculated, considering specific processes (eventually updating them based on the geographical locations and their current grid mixes), and then applying an impact assessment method (such as CML). What processes from Ecoinvent were considered in this analysis? What are the life cycle inventories used in the considered Ecoinvent processes? The associated life cycle inventories are probably outdated, and they are not the latest. Please check this 2016 manuscript in which emission factors are calculated for Europe, USA and China, considering their respective electricity grid mix compositions: https://doi.org/10.3390/en9080622. Table 2 in the supplementary information shows the assumed carbon emissions for 4 products that are silicon, silicon wafer, solar cell, and PV module. However, there is a lack of transparency for reproducing the numbers. My main questions are: 1) What processes/contributions are included in each stage? Are the aluminum frame and encapsulations included in the PV module contribution? It would be useful to add a flow diagram to be transparent on what is included/excluded.
2) What is the considered silicon type (single-crystalline or multi-crystalline)? They have different efficiencies, energy demand and carbon emissions. Specifically, the Cz process in the single-silicon cell is more energy demanding and more impactful in terms of carbon emissions (https://doi.org/10.3390/en9080622).
3) The USA wafer carbon emissions listed are 302.83 kgCO2/kg wafer, while wafers produced in China cause 150.29 kg CO2/kg wafer. Why the USA emission factors is approximately twice as that for China? On the contrary, the carbon emissions of US silicon and US solar cells are lower compared to the ones in China (79.34 kg CO2/kg silicon compared to 129.86 kg CO2/kg silicon and 0.79 kgCO2eq/Wcell compared to 0.96 kgCO2eq/ Wcell)? What is the reason for such differences? 5) Solar cell and PV module carbon emissions are expressed per W cell. What are the assumed efficiencies? (Please see the Fraunhofer report in which the latest commercial c-Si efficiencies are listed on pag. 29: https://www.ise.fraunhofer.de/content/dam/ise/de/documents/publications/studies/Photovoltaics-Report.pdf) 6) The function units in Table 2 are different (kg and W). How did you combine those? What is the material usage (kg per silicon used per wafer and kg of wafer used per W cell)? 7) What is the thickness of the wafers?
8) The BOS contribution is not listed. BOS emissions factor ranges from 251-280 kgCO2eq/kWp (considering power conversion efficiencies of 20%-18%) (see figure 4 in the following reference: https://www.mdpi.com/1996-1073/9/8/622) Page 21 lines 499-507 Air pollutants (PM10, NOx and SO2) of PV trade were assumed from the literature. For China ref. n. 12 is assumed as data source, while for all the other countries Ecoinvent database is used. To be consistent it would be better to assume the same source of data.
Air pollutants are listed in Table 3 and 4 in the supplementary information. Please clarify which countries are included in "other". Also, other questions are? -Why do Europe and America have the same local pollutant emissions? The respective electricity grid mixes are different.
-What are the efficiencies assumed per W? -Wafer thickness? -In Table 4 (emissions of PV in China) two years are assumed as reference (2014 and 2017), while in Table 3 there is only one year. What is the reason? -Please explain the reason for huge differences between 2014 and 2017 values in Table 4 (such as PM10 from solar module: 0.103 g/W cell for 2014 compared to 0.024 g/W cell for 2017).
-It is also not clear the reason for differences between values in table 3 compared to those in  Page 11 lines 265-266 It is stated that "solar power replacement of fossil fuel-fired generation also leads to local air pollutant emissions abatement". As stated above in this report, the contribution of balance of systems should be added because it is a crucial part of PV systems for producing electricity. Also, electricity grids with a large penetration of PV (such as in your future scenarios) need storage systems, and their life-cycle emissions should be considered. Please see this 2020 manuscript: https://doi.org/10.1002/ente.201901146 and the other PV LCA literature studies on storage listed above.
Other comments: -In the manuscript, it is used "lifecycle" as a one word, but it is more common "life cycle" as two words in the LCA literature.
-Often, it is implicitly assumed in the manuscript that silicon is the only available technologies for PV panels, and this is incorrect. Some examples are: Lines 35-38 "Although PV power generation has nearly 'zero emissions' in its operation, pollutant emissions cannot be ignored when considering the whole lifecycle of PV products--from ore mining, silicon and silicon wafer processing, and solar cell and PV panel producing" This statement implicitly assume that PV systems can be only be fabricated with silicon and it is incorrect. It should be specified that your statement refers to PV crystalline-based technologies. Although the main commercial PV technology is currently based on silicon (~94% of the market)there are also other PV technologies commercially available, such as thin film cadmium telluride (CdTe) and copper indium gallium (di)selenide/(di) sulfide (CIGS) technologies [1]. Thin film technologies are more recent PV technologies compared to silicon-based architectures, and they offer potential benefits compared to wafer technology, including ~100x thinner absorber layers and opportunities for low cost. Thin film technologies also show reductions in life cycle carbon emissions and other environmental impacts [2]. There are also PV emerging technologies -such as perovskite PV (single junction and tandem architectures) -that have been attracting intensive attention due to notable laboratory efficiency improvements, and that may be implemented to industrial scale [3]. Their life-cycle impacts have been also evaluated in recent studies [4,5,6,7,8]. Lines 428-429 Please specify that it is referred only to crystalline-based technologies.
Lines 468-470 and lines 136-137 This is a repetition. The following sentence "this study converted all the currencies to US dollars first and then converted them to 2010 US dollars for comparability" is repeated two times.

Title：Barriers on PV trade will impede global carbon mitigation
Response to the reviewers' comments Reviewer #1: This paper tries to estimate the emissions reduction potential of PV of global PV product trade based on partial equilibrium analysis. In general, the topic is important. However, there are some major problems in your work.
Here are some specific comments: Reviewer comments: 1. The main tool of simulation analysis in this paper is GSIM model, which belongs to a short-term analysis. However, the author uses the simulation results of GSIM model to estimate the long-term environmental impacts of trade barriers of PV products. Is this reasonable? The author needs to explain this.

Response to reviewer comments:
Many thanks for the comments and suggestion. We have revised the manuscript and make up for this defect by deploying a dynamic model (integrated modelling system, IMS) in combination with GSIM model. GSIM model is used to simulate trade barriers impacts on global PV trade pattern, while IMS model is applied to project the long-term global and country/economy specific power production and supply composition and the relevant environmental impacts up to 2060. Although dynamic CGE model such as GTAP can also project long-term impacts of tariff rate imposition, it is very hard to separate and simulate the PV goods trade, which are only a small fraction of several sectors, such as Mineral products, Computer, electronic and optical products, and Electrical products. And it is not convenient for CGE to describe and simulate the detailed technology competition in the power generation sector. The detailed content is shown in Results "Projection to power market share and carbon emissions reduction potential of PV application" and "Barriers to the PV product trade would impede global emissions reduction potential", and in the Methods and Supplementary Information 5 and 6.

Reviewer comments:
2. Two basic facts that the author should pay attention to when using GSIM model to simulate. First, in January, 2018, when the U.S. government announced that it would impose 201 tariff on imported PV products, it will set a duty-free quota of 2.5 GW, and impose 30% tariff on imports exceeding this quota in the first year, and the tax rates will decrease to 25%, 20% and 15% in the following three years; Second, on June 12, 2019, the Federal Trade Department of the United States ruled that three types of PV products, including double-sided photovoltaic modules assembled by doublesided batteries, 250-900w flexible glass fiber solar panels and some optical thin-film cell panels, were exempted from the 201 tariff from June 13, 2019.
Many thanks for the comments and suggestion. In revising the GSIM model simulation and the policy scenario setting, we have tried our best to take the real world tariff barrier including the USA PV goods import tariff policy that the reviewer mentioned into consideration. However this study is to disclose the general rule that, trade barrier on PV goods in the long-run will inevitably impede global carbon reduction potential, rather than to evaluate the USA PV goods trade policy in very detail in a short term. Thus we set three policy scenarios based on the ongoing, foreseeable and possible tariff imposition situations, and also took uncertainties of tariff rates adjustment into account ( Supplementary Information 6). We would think that the very detailed PV goods tariff rate changes, for example, the exemption of thin-film cell panels, are beyond the research scope of this study, but we are willing to do further analysis in some other researches in the near future.

Reviewer comments:
3. This paper only evaluates the overall adverse impact of trade barriers on global carbon emissions, but ignores the differences in the impact of trade barriers on the carbon emission reduction of photovoltaic products exporting and importing countries. For example, although trade barriers are not conducive to the reduction of carbon emissions of countries such as the United States which impose high tariffs on photovoltaic products, they may be conducive to the reduction of carbon emissions of photovoltaic producing countries and exporting countries such as China. It is suggested that the author should evaluate the impact of trade barriers on carbon emissions of different countries, and then evaluate the overall impact from a global perspective.

Response to reviewer comments:
Many thanks for the great comments and suggestion. In the revision, we have evaluated the impacts of trade barriers on carbon emissions of different countries, and then evaluate the overall impacts from a global perspective in Results section "Barriers to the PV product trade would impede global emissions reduction potential", for example, " In escalated tariff scenarios TBS1 and TBS2, countries/economies that exert higher PV goods import tariff rates will experience larger carbon reduction potential losses; e.g., India and the USA are expected to lose net carbon reduction potential by 5.25-5.44 and 2.17-2.48 GtCO 2 e (SST), respectively……. Although trade barriers are not conducive to the reduction of carbon emissions of countries such as India and the United States which impose high tariffs on PV products import, they are conducive to the reduction of carbon emissions of PV producing and exporting countries such as the Republic of Korea and China, whose net carbon reduction potential will increase by 0.03-0.05 and 0.42-0.58 GtCO 2 e (SST), respectively. However, the overall impacts of trade barrier on PV goods cause the global carbon emission reduction potential to decrease. Escalated tariff scenarios TBS1 and TBS2 will decrease the global net carbon reduction potential by 3.71-6.34% or 3.26-6.77 GtCO 2 e (SSG-SST) and 4.37-6.26% or 3.22-7.98 GtCO 2 e (SSG-SST), respectively. (see Fig. 7b)"

Reviewer comments:
Many thanks for the comments and suggestion. In the revision, we have elaborated the methods in more details to describe how to predict/simulate tariff impacts on carbon emission reduction in Methods section "A technology-based integrated dynamic model to project power market share" and "Tariff barrier scenarios and computable partial equilibrium model simulation", and also in Supplementary Information 5 and 6.

Reviewer #2:
Reviewer comments: The authors provide a wealth of facts and detailed analysis, and the headline number is of plausible magnitude. At the same time, the paper is 33 pages long and one suspects that a reasonable range of the impacts of a tariff on total demand could be had with a simple back of the envelope calculation. How much does the tariff (or tariffs) raise the average end-user price of the relevant goods? What is the elasticity of demand for the product?

Response to reviewer comments:
Many thanks for reminding us to provide the information of the impacts of escalated tariffs on the average price of the PV goods, which is the key to affect demand and supply and applications of PV solar cells and modules. Based on GSIM model simulation, PV products prices changes were obtained and are listed in Supplementary Information 6.2. The demand elasticity parameters involved in the GSIM model are also given in Supplementary Information 6.1 Supplementary Table  8.

Reviewer comments:
The paper abstracts from many other relevant factors-particularly going out to 2050. (1) How likely are today's trade patterns to hold over the next 30 years? Would China continue to have a dominant position on the markets irrespective of the trade wars? Would its aging population and rapidly rising middle-class price China out of the market to the benefit of competing producers? (2) To what extent is the trade war a temporary phenomenon and a function of the current political environment? Should the world's nations get serious about decarbonizing the global economy, wouldn't they seek policies that accelerate PV penetration? (3) What are the substitute technologies and how substitutable are they with PV? (4) What other factors will influence the future of PV? For example, the aforementioned R&D and product innovation.

Response to reviewer comments:
Many thanks for reminding us that this paper abstracts from many relevant factors.
(1) Trade pattern is indeed an important factor relevant to the projection results. We considered that trade pattern is largely decided by the differences in factor endowment of various countries/economies which are relatively stable, e.g., though East Asia countries including China have the problem of aging population and rapidly rising middle-class, their comparative advantages in PV goods supply over other regions should last long, and thus we would assume a stable trade pattern to facilitate the trade barrier scenario analysis. For the power demand predictions for various countries/economies in IMS model simulations, the population and income factors have been considered in the parameter setting. (2) In this study, we would like to disclose the fact/rule that, ongoing, foreseeable and possible trade barriers can impact on global PV trade and harm the global carbon mitigation capacity, and try to call on for freer trade of PV goods. We and many scholars are very much concern that trade war is not simply a temporary phenomenon, but a function of the current and long-lasting political environment in which the USA-Sino competitions are intensified, which is not conducive to global joint effort fighting against climate change. Recently, major carbon emission countries/economies, including China, the USA, Japan, EU, among others, made serious promise to achieve carbon neutrality by 2050 or 2060. We would like to take these promises as serious ones and have taken these factors into our dynamic model (IMS) simulation, referring to national energy plans, energy development outlooks, NDC target commitments, long-term energy strategic plans and decarbonisation pathway research reports of various countries/economies. (3) and (4) This study adopted IMS model to simulate the dynamic evolution of power generation and supply market in various countries/economies from 2017 to 2060, with an emphasis on the competition of solar PV with other power generation technologies, including coal, gas, oil, nuclear, solar thermal, geothermal, marine, hydropower, onshore wind, offshore wind, biofuel, biogas and waste incineration. The technology progress brought about by R&D were reflected in the production carbon emission coefficient decreases, PV panel conversion efficiency improvement, and performance rate increases which in turn will reduce the embodied carbon and increase the net carbon emission reduction potential of globally traded PV products (Supplement Information 3.3).

Reviewer comments:
There is some degree of sensitivity analysis with respect to the scenarios, but it would be useful to provide a more thorough and systematic sensitivity analysis as regards the key parameters: future trade shares, supply, demand and substitution elasticities.

Response to reviewer comments:
Many thanks for the comments and suggestion. In the revision, we have added more detailed analysis to test the sensitivities of simulation results, such as, future trade (imports and exports) volume and structure (shares) and consumptions (demand and supply), with respect to the tariff rates changes under different scenarios, varied substitution elasticities settings, among others ( Supplementary Information 6).

Reviewer comments:
The paper is a very difficult read. It is chalk full of numbers and figures (some of which are mislabeled), using more significant digits than necessary, and rarely contextualized-are the numbers large or small? and relative to what?

Response to reviewer comments:
Many thanks for the comments and suggestion. In the revision, we have improved the manuscript by deleting redundant numbers, unnecessary tables and figures, polishing the language, highlighting the most important results, and improving the contexture. We hope the revised manuscript is easier and better to read.

Reviewer #3:
The manuscript entitled "Barriers on PV trade will impede global carbon mitigation and local pollutant emissions reduction" is focused on the emission potential reduction of global PV trade and the effect of extra trade barriers on PV product globally. The topic is very interesting and timely. However, crucial assumptions and data sources should be revised in the method section for the current and future scenarios under analysis. Below I list my specific comments/concerns.

Reviewer comments:
In this study, PV life-cycle carbon emissions are obtained from various existing literature studies (some of them are not the latest available as detailed below in this report), reflecting the 2009-2013 silicon PV technology state of the art. In the developed future scenarios (from 2017-2050) the same PV carbon emissions are assumed. This assumption is a weak point because PV life-cycle carbon emissions are expected to decrease in the medium-long term mainly due to improvements in manufacturing processes. Specifically: (i) reduced silicon wafer thickness; (ii) reduced material usage (such as silicon, silver content in the metallization paste, copper), and kerf losses; (iii) electricity consumption reductions; (iv) PV module power conversion efficiency improvements.
As an example, please see the latest Fraunhofer report 2020, available at:https://www.ise.fraunhofer.de/content/dam/ise/de/documents/publications/studies/Photovoltaics -Report.pdf, which shows silicon usage reduction ranging from 150 g/Wp in 2013 to 80 g/Wp in 2019 as well as efficiency improvements in silicon-based PV module over the last years. Please check also the following IEA PVPS Task 12 Report, in which are listed potential future changes (such as reductions in kerf loss, wafer thickness, carbon emissions and power conversion efficiency improvements): Frischknecht, R., Itten, R., Wyss, F., Blanc, I., Heath, G.A., Raugei, M., Sinha, P. and Wade, A., 2015. Life cycle assessment of future photovoltaic electricity production from residential-scale systems operated in Europe (No. NREL/TP-6A20-73849). Available at: https://iea-pvps.org/keytopics/iea-pvps-task-12-life-cycle-assessment-of-future-photovoltaic-electricity-production-fromresidential-scale-systems-operated-in-europe-2015/ In this study, these potential changes are not discussed or included, such as providing a sensitivity analysis (or uncertainty analysis) for potential future emission changes due to production improvements. Also, it is not captured/discussed potential changes in the electricity grid mix compositions (such as increased penetration of renewable energies), which influence the amount of primary energy required for each PV production process as well as the associated carbon emissions.
Many thanks for the comments and suggestion. In the revision, we have fully considered the suggested aspects, i.e., reduced silicon wafer thickness; reduced material usage (such as silicon, silver content in the metallization paste, copper) and kerf losses; electricity consumption reductions; potential changes in the electricity grid mix, etc., taken the technology progress of PV products manufacture into consideration to update the emission coefficients in the Methods part, and added an uncertainty analysis in Supplementary Information 3.3 to address the sensitivity of embodied carbon emission and the net carbon emission reduction potential of global traded PV products with respect to the carbon emission coefficients (CE) decreases, PV module power conversion efficiency (CE) improvements, and PV performance rate (PR) increases. In Supplementary Information 5 we also addressed the projected electricity power supply composition in various countries/economies and the world, and its impacts on the PV production process carbon emission were reflected in the carbon emission coefficients (CE) decreases.

Reviewer comments:
It is stated that "studies have been conducted varying from calculating the lifecycle emissions and emission reduction potential of PV products (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21) to PV product trade patterns and structures (22) to the emissions embodied in the bilateral PV product trade (23) and to the environmental impacts of the PV product trade (13,14,24). Studies have also been carried out in various countries (15-21, 24), concluded that PV power generation could indeed help to reduce carbon and local air pollutant emissions obviously." The authors cite here previous PV life cycle studies, 7 of which are focused on China, 1 in the UK, 1 in Turkey, and 1 in the United States, and most of them are based on crystalline silicon technology.
Since the scope of this paper is "global" and it includes future scenarios, it would be appropriate to extend the existing background references, integrating them with other studies that are focused also on: (i) P V produced and/or installed in other geographical areas, (ii) comparison between PV life cycle carbon emissions with other electricity generation technologies (fossil fuels and renewables) in specific electricity grid mixes, (iii) complete PV systems, including balance of system (BOS) (structures, inverters, cables, transformers).
(iv) the additional life cycle contribution of potential storage and/or batteries needed for future scenarios with large renewable electricity penetration.
(v) PV end-of-life life-cycle contribution.

Response to reviewer comments:
Many thanks for the comments and suggestion. In revision, we have extended the existing background literature to studies on PV produced and/or installed in other geographical areas, PV life cycle carbon emissions, complete PV systems (including balance of system (BOS), storage and/or batteries, the additional life cycle contribution of potential storage and/or batteries, PV endof-life life-cycle contribution, etc., in the Introduction section, for example, "Studies have been conducted ranging from calculating the life cycle emissions and emissions reduction potential of PV products (8,10-16,19-24) to PV systems including balance of system (BOS) and storage and/or batteries (25-29) and their end-of-life emission contribution (30,31), life cycle impact comparisons between solar power and other power generation technologies (32,33), and the impacts of PV application on future electricity grids (34,35) in various geographical areas." We have also used the updated literature to calibrate the emission coefficients and to calculate the carbon emission and carbon reduction potential.

Reviewer comments:
Page 2 Lines 45-47 "previous studies concentrated on only a specific PV product and a limited subset of countries/regions without full consideration for the whole production and supply chain from a global perspective" This statement is not correct.

Response to reviewer comments:
Many thanks for the comments and suggestions. In the revision, we have deleted the inappropriate statement and cited most of the literatures recommended by the reviewer.

Reviewer comments:
Lines 78-88 Please list the references for your trade flow matrix (TFM).

Response to reviewer comments:
Many thanks for the suggestion. We have listed the detailed literatures and data sources for building the trade flow matrix in Method section titled "Multilateral PV product trade and trade flow matrix (TFM) construction".

Reviewer comments:
Page 20 Lines 481-482 It is stated that "PV power generation is relatively clean and near zero-emissions in its application, but certain amounts of GHGs and pollutants are emitted during the PV products production process" Life cycle analysis takes into the account not only the production processes, but the product's full life cycle from the extraction of resources and the production of raw materials, to manufacturing, distribution, use and re-use, maintenance, and finally recycling and disposal of the final productincluding transportation and use of energy carriers. Please check the following references, which are focused on LCA guidelines and standardization as well as specific PV

Response to reviewer comments:
Many thanks for the comments and suggestions. We have carefully checked all the literatures recommended by the reviewer, and cited the most up-to-date PV LCA guidelines and revised the statement as "Although PV power generation is nearly 'zero emissions' during operation and could indeed help to substantially reduce carbon emissions (8-13), its emissions should not be ignored when the whole life cycle of PV products is considered (14-18)." We have also added consideration of BOS and storage system related emissions in the revision, referring to the guidelines recommended by the reviewer and up-to-date researches to obtain emission coefficients and calculated the carbon emissions. The detailed information is shown in Supplementary Information 3.2 "carbon embodied in BOS and storage system". In the supplementary information section 3 the considered emission factors taken from the literatures are listed, and it is stated that "because regional LCA studies of PV products were quite limited, we adopted the latest and up-to-date LCA emission coefficients available to carry out embodied emission calculation". However, Ecoinvent database version 3 does not provide directly the PV "emission factors", but Ecoinvent is a database with specific processes. Emission factors should be calculated, considering specific processes (eventually updating them based on the geographical locations and their current grid mixes), and then applying an impact assessment method (such as CML). What processes from Ecoinvent were considered in this analysis? What are the life cycle inventories used in the considered Ecoinvent processes? The associated life cycle inventories are probably outdated, and they are not the latest.

Response to reviewer comments:
Many thanks for the kind comments and suggestions. We used Ecoinvent database to inform emission coefficient of a PV product. When typing in a specific PV product name, e.g., silicon wafer, in the database interface, a life cycle inventory (LCI) considering emissions of all the upstream activities/processes of the product were presented, which are used as PV product emission coefficients to calculate the embodied carbon emissions. However, Ecoinvent only distinguishes the data applicability for Europe and the world and claims a long applicable period (for example 1.1.2005-12.31.2020). Since as the reviewer indicated, the life cycle inventories in the Ecoinvent are outdated, in the revision, we adjusted the original coefficients from Ecoinvent according to the PV production technology improvement (e.g., reduced silicon wafer thickness; reduced material usage and kerf losses; electricity consumption reductions; potential changes in the electricity grid mix, etc.) described by the IEA PVPS trend report. The detailed information can be seen in Methods section "Accounting embodied carbon emissions in PV products and their global trade" and Supplement Information 3 "Embodied carbon emission accounting". Table 2 in the supplementary information shows the assumed carbon emissions for 4 products that are silicon, silicon wafer, solar cell, and PV module. However, there is a lack of transparency for reproducing the numbers. My main questions are:

Response to reviewer comments:
Many thanks for the great comments. Please see in below the point-by-point response to the question: 1) What processes/contributions are included in each stage? Are the aluminum frame and encapsulations included in the PV module contribution? It would be useful to add a flow diagram to be transparent on what is included/excluded.

Response to reviewer comments:
In the Supplementary Information Table 2, we have added a column to explain the processes/contributions included in each stage.
2) What is the considered silicon type (single-crystalline or multi-crystalline)? They have different efficiencies, energy demand and carbon emissions. Specifically, the Cz process in the single-silicon cell is more energy demanding and more impactful in terms of carbon emissions (https://doi.org/10.3390/en9080622).

Response to reviewer comments:
We consider the mix of the two dominant silicon types (single-crystalline or multi-crystalline) to calculate the emission coefficients for the PV products. 3

Response to reviewer comments:
In the last version, we employed carbon emission coefficients from different data sources/literature which caused such differences. In the revision, we applied the uniform data source, i.e., Ecoinvent to avoid such differences. 5) Solar cell and PV module carbon emissions are expressed per W cell. What are the assumed efficiencies? (Please see the Fraunhofer report in which the latest commercial c-Si efficiencies are listed on pag. 29: https://www.ise.fraunhofer.de/content/dam/ise/de/documents/publications/studies/Photovoltaics-Report.pdf)

Response to reviewer comments:
For the 2017 status quo, the corresponding conversion efficiencies were assumed to be 16-25% for mono-crystalline silicon technology and 14-18% for multi-crystalline silicon technology, respectively (IEA-PVPS. Trends in Photovoltaic applications 2018 (2018)). Table 2 are different (kg and W). How did you combine those? What is the material usage (kg per silicon used per wafer and kg of wafer used per W cell)?

Response to reviewer comments:
The emission coefficients based on different function units in Supplement Information Table 3 (revised coefficients)were all converted to emission coefficients based on status quo price and trade value. In Methods section, it is mentioned "since the PV product trade volume in the constructed TFMs are in US$ but not in physical quantities, the emissions coefficients measured by PV product quantity were converted into emissions per unit value in US$ based on the prices of PV products." The Life cycle inventory of the present study were drawn from Ecoinvent, that, 1 m 2 of mono-crystalline wafer and multi-crystalline wafer use 1.07 kg and 1.14 kg silicon respectively. For multi-crystalline wafer production, the 1 m 2 of wafer surface is sawn into square wafers with a size 156x156 mm 2 (0.0243 m 2 ) and a thickness 240 μm, and the weight is 559 g/m 2 . For monocrystalline wafer production, the 1 m 2 of wafer surface is sawn into square wafers with a size 156x156 mm 2 (0.0243 m 2 ) and a thickness of 270 μm, and the weight is 629 g/m 2 . 7) What is the thickness of the wafers?

Response to reviewer comments:
The thickness of the mono-crystalline wafer is 270 μm and the thickness of the multi-crystalline wafer is 240 μm.

Response to reviewer comments:
Emission coefficients of BOS and battery storage were drawn from the recommended literature and listed in Supplementary Table 4.

Reviewer comments:
Page 21 lines 499-507 Air pollutants (PM10, NOx and SO2) of PV trade were assumed from the literature. For China ref. n. 12 is assumed as data source, while for all the other countries Ecoinvent database is used. To be consistent it would be better to assume the same source of data.

Response to reviewer comments:
Many thanks for the kind comments and suggestion. In the revision, we focused only on carbon emission mitigation, and deleted the discussion on the air pollutant issue.

Reviewer comments:
Air pollutants are listed in Table 3 and 4 in the supplementary information. Please clarify which countries are included in "other". Also, other questions are? -Why do Europe and America have the same local pollutant emissions? The respective electricity grid mixes are different.
-What are the efficiencies assumed per W? -Wafer thickness? -In Table 4 (emissions of PV in China) two years are assumed as reference (2014 and 2017), while in Table 3 there is only one year. What is the reason? -Please explain the reason for huge differences between 2014 and 2017 values in Table 4 (such as PM10 from solar module: 0.103 g/W cell for 2014 compared to 0.024 g/W cell for 2017).
-It is also not clear the reason for differences between values in table 3 compared to those in table 4. As an example, NOx associated to solar module in Europe are 0.721 kg/W module (which is similar to the value for "other" countries), while in China in 2014 are 1.389 and in 2017 are 0.215.