A Distributed Computational Model for Estimating the Carbon Footprints of Companies
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors1. Input-output is generally a macro sector model. How does this paper use the macro model to discuss the micro level of the firm? What is the specific basis in the model?
2. What is the basis of the EIN application protocol? Does it exist in reality?
3. The article assumes that companies produce a single product, but in reality many companies produce multiple products, which requires more sophisticated methods of cost accounting and data allocation.
4. Company-level emissions data needs to be audited by an independent body to ensure accuracy, which may require additional resources and costs. In reality, most companies will probably prefer not to participate in distributed systems. There is not much thought in this article about reducing the cost of data, and there is little thought in incentivizing companies to tell the truth.
Comments on the Quality of English LanguageI think it could have been revised better.
Author Response
- Input-output is generally a macro sector model. How does this paper use the macro model to discuss the micro level of the firm? What is the specific basis in the model?
This question is addressed in new section 3, previously numbered section 2 “Applying the input-output approach to the calculation of companies production footprints”
The section is divided in four subsections for better clarity.
Subsection 3.1 reviews the standard Leontief approach, which is a final demand driven approach working upstream to calculate the production of sectors, as endogenous variable.
Subsection 3.2 reviews two relevant alternative approaches, which are both scope 1 emissions driven approaches working downstream to calculate the “production footprints”. Production footprints can be termed more explicitly “total value chain production footprints” . They are the endogeneous variables. One option is to used absolute values, the other option is to use intensities.
We have added a reference to paper previous numbered [16] now numbered [32] by Hertwich and Wood, since they erived in 2018 the downstream equation on footprint monetary intensities (also termed embedded footprint per monetary production unit). They are intrinsically (using different notations) the same equations as those proposed by Kalckreuth in his 2021 and 2022 papers and Meunier in hits 2023 paper. The 2022 paper was already in our references as well as the Meunier paper, but we have added the 2021 paper and another relevant 2022 paper by Kalckreuth to the references.
The specific basis of the model we propose in the paper is described in the two following subsections.
Subsection 2.3 describes the iterative distributed mechanism which is key to the paper and explains its title. The iterative property of the algorithm makes it unnecessary to invert a very large dimension matrix (millions of vector components): this addresses the first issue of the macroeconomic approach: computational tractability. The distributed property of the algorithm makes unnecessary to centralize the data of the companies, on a central server, which addresses a second issue: confidentiality. Moreover the distributed processing load at the company level is light, since it can be handled instantaneously on a personal computer. The key references for the iterative solution to the input-output scope 1 driven equations are the papers of Kalckreuth just mentioned above.
Subsection 2.4 describes how the iterative distributed approach can be applied even though only a fraction of the world enterprises take part to the calculations. This is the key contribution of the paper. It addresses a key issue for real world feasibility. Another contribution is the fact that the footprint can be decomposed into three components with different uncertainties: two footprint components arising from the accumulation of respectively scope 1 and scope 2 data with typical 10% uncertainty, and a component arising from the accumulation of upstream scope 3 data of the companies that do not take part to the calculation, with higher uncertainty, the level of which is unknown because seldom calculated in LCA studies. Such scope 3 uncertainty is likely above 30%, probably higher for sophisticated products such as digital devices.
- What is the basis of the EIN application protocol? Does it exist in reality?
The need for such a protocol comes from the fact the system relies on software distributed across all enterprises. Such software must abide by rigourous specifications to implement the calculations, to synchronize and to exchange information between the software instances installed in the companies’ premises. The specifications and the protocol need to be standardized.
Such a protocol does not exist, nor does the EIN system, which has yet to undergo initial tests and field trials. Field trials may be conducted without a rigorous protocol, but on the basis of well-defined rules to be followed by the companies taking part to the trials. Such rules will probably evolve as the results of the field trials. For a large-scale deployment the rules shoiuld be standardized, including the standard protocol what information is exchanged, when and in what formats.
The protocol briefly outlined in the paper corresponds to a foreseen real-world implementation where the calculations are carried out at specific periods after the closing of accounting periods. The carbon intensities are managed outside the existing accounting systems.
Other protocols are possible including the integration of carbon intensities in the billing systems. In such a case the protocol may seem simpler, but it alters the relatively simple mechanism we have described, so we yet have to simulate the mechanism and compare the convergence properties to the implementation we have simulated.
- The article assumes that companies produce a single product, but in reality many companies produce multiple products, which requires more sophisticated methods of cost accounting and data allocation.
The mono-product company assumption is indeed a simplification. It helps in simplifying the equations. But there is no loss in generality as for the feasibility of the system, since the general case of multiproduct companies can fit into the same input-output model. It is possible to model a multi-product enterprise with m products as a set of m mono-product enterprises, by sharing the purchased goods of the company between the m products. What is needed is a sharing key for each couple (purchased product, produced product). The set of sharing keys between purchased products and produced products comes easily from the cost accounting tables, when they exist. If the company does not carry out cost accounting, an alternative solution is as follows: (a) to calculate only the company footprint aggregating all the products; (b) to calculate an alternative value of the footprint by using emissions factors from existing LCAs; (c) to adjust such LCA emission factors by the ratio of the footprint calculated at step (a) to the footprint calculated at step (b). If LCA data is missing for some products at step (b), the monetary emissions products calculated by macroeconomic input-output can be used. We addressed this issue in previous section 6 (discussion) as follows:
A second issue is the need to calculate product footprints rather than company level footprints. In our paper we have assumed that all companies produce a single product. In reality we must handle the multi-product case. It is theoretically possible to desaggregate an enterprise into single product sub-companies. The accounting solution to this is cost accounting, whereby the procurement and expenses are attributed to specific products of the company, while general expenses are divided between all products using allocation rules. But only a few companies produce such cost accounting data. Another approach, although less precise, would be to share the total footprint of the company between its products based on the a priori knowledge contained in emissions factors from available databases.
We now have added to previous section 2, now numbered section 3 as a subsection 3.5 to explain more precisely how to handle multiproduct companies, and we have modified the paragraph in the discussion (section 7) accordingly.
- Company-level emissions data needs to be audited by an independent body to ensure accuracy, which may require additional resources and costs. In reality, most companies will probably prefer not to participate in distributed systems. There is not much thought in this article about reducing the cost of data, and there is little thought in incentivizing companies to tell the truth.
The article focuses on the soundness of the underlying input-output model, as a preliminary step to the specification of a real system to be deployed in real life. It is necessary to conduct field trials which will help in identifying bottlenecks and hurdles to a full deployment. To reduce costs, standards are mandatory (general requirements for the companies, use of standard software and standard protocol). The audit is necessary as for financial accounting but need not be done exhaustively, but rather on a sampling basis. As for incentives, we contemplate working with a kernel of large companies that are engaged in adapting their processes to environmental issues, and that would pass the requirement to their suppliers. Some of our partners contemplate applying penalizing scope 3 emissions factors for companies that don’t take part to the distributed system. We believe such recommendation is too strong, but in any case we feel this discussion is too early and beyond of the scope of this article. We believe the most relevant next steps are (a) to design and conduct a field trial (b) to start and take part into an international standardization effort.
Reviewer 2 Report
Comments and Suggestions for AuthorsComments for author File: 
Comments.pdf
Author Response
We thank the reviewer for the encouraging review. We remain available to answer questions.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe article discusses current issues.
The tools used are correct. The results are well described.
The conclusions were well formulated. The most important limitations and possible applications of the results were also indicated.
The theoretical part seems to be an issue that needs improvement. In the article, I propose to add a section "Review of literature (research)" that has been conducted so far and is consistent (similar) to the article. This will improve the quality of the article and provide a stronger justification for the need for this type of research.
Additionally, a review of research will allow for a broader approach to the problem.
The number of cited literature items seems too small. It is worth expanding it.
The language of the article is correct. No major linguistic errors were found.
Comments on the Quality of English LanguageThe language of the article is correct. No major linguistic errors were found.
Author Response
We thank the reviewer for the review and the suggestions. We acknowledge that the review of literature was too light. We have added a dedicated section numbered section 2 for that purpose.
Round 2
Reviewer 3 Report
Comments and Suggestions for AuthorsThe corrections made take into account previous indications.
