Methodology

Introduction

Following  the release of the Cambridge Blockchain Network Sustainability Index (CBNSI), we wanted to further investigate environmental externalities attributable to the Ethereum network and provide an emissions estimate for Ethereum before The Merge. Driven by a rise in popularity and mainstream adoption, the debate around the sustainability of cryptoassets has intensified. By adding this new index to the CBNSI, we hope to contribute to the debate and offer an approach to quantify Ethereum's historical emissions debt.

In our prior publication in April, we underscored Ethereum's pre-Merge electricity consumption as a pivotal milestone. Although this research signified a crucial preliminary step, it did not capture the geographical distribution of mining activity and thus missed a vital component for a more comprehensive environmental impact assessment. Addressing this, our current study builds on those initial findings, focusing on the historical geographical patterns of mining.

To this end, we are excited to unveil a major extension to our Ethereum research: a novel index tool that leverages information from a wide variety of sources to provide intricate  insights into Ethereum mining patterns, laying the groundwork for developing a pre-Merge greenhouse gas (GHG) emissions estimate.

Employing logic similar to that behind our Bitcoin GHG estimate, we performed a sensitivity analysis to assess how GHG emission estimates vary with changes in key parameters. Accurately calculating exact GHG emissions is challenging, prompting us to use a hypothetical range encompassing three scenarios. The two extremes constitute the upper and lower limits, encapsulating the most credible estimate.

It is vital to clarify that our index solely evaluates the environmental impact from Ethereum mining’s electricity consumption, not offering a full life-cycle assessment. Based on Köhler and Pizzol's (2019) research, the primary factors influencing Bitcoin's environmental footprint are geographical distribution and hardware efficiency, with other aspects having less than 1% impact. Under the assumption that these findings apply to Ethereum as well, we conclude that other factors would also have only a minor impact on Ethereum's environmental footprint.

Additionally, it is imperative to acknowledge key research that has informed our methodology and directed our perspective on this topic. Central to our references is the pioneering work of McDonald (2022), which underpins the block allocation algorithm we employed.

Key terms

Table 1: Key terms and abbreviations used

Model parameters and data sources

Table 1: GHG index model parameters

Life-cycle greenhouse gas emissions from electricity generation

Table 3 lists the median total of life-cycle emission factors (one-time upstream, ongoing combustion, ongoing non-combustion and one-time downstream) for selected electricity generation technologies. The values are based on a systematic review of published life-cycle assessment (LCA) studies conducted by the National Renewable Energy Laboratory (2021).

Table 1: Life-cycle emission factors for electricity generation technologies

^a Light-water reactor (including pressurised water and boiling water) only

^b Mean value of photovoltaic (thin film and crystalline silicon) and concentrating solar power (tower and trough)

^c Land-based and offshore

^d Mean value of geothermal, biomass and ocean

Methodology 

This section briefly outlines the core concept of our methodology before delving deeper into how the estimates are calculated.

Structurally, this study follows a similar approach to the CBECI by introducing hypothetical lower and upper bounds that form the estimated GHG emissions range. This is the basis of the sensitivity analysis, summarised in Figure 1. The two limits are extreme scenarios based on the hypothetical assumption that all the electricity used is generated by a single energy source (coal-only or hydro-only). This format is simply illustrative and provides context to make interpreting the results more intuitive.

It is essential to recognise that these two scenarios are based on stark simplifications, postulating that Ethereum miners exclusively rely on a single energy source. In reality, the electricity consumed is generated from various energy sources, referred to as the 'electricity mix'. Determining what this mix comprises is pivotal for calculating an appropriate emission intensity associated with the Ethereum network. This study defines emission intensity as the amount of GHGs emitted per kilowatt-hour (gCO₂e/kWh). Table 3 lists all the energy sources used in this study and their corresponding emission intensities

Figure 1: Sensitivity analysis

A crucial aspect of determining an Ethereum-specific electricity mix involves understanding the geographic distribution of mining activities. For instance, the emissions from mining activity in Sweden would significantly differ from those in Kazakhstan.

Given the decentralised nature of the Ethereum network, obtaining direct data on the geographical spread of mining activity is challenging. We have tackled this data scarcity by harnessing publicly available information to create dynamic electricity mix estimates. This, in turn, allows us to compute a corresponding Ethereum-specific emission intensity.

The subsequent sections outline precisely how this is conducted for all scenarios.

Lower- and upper-bound estimates

To derive a fitting electricity mix profile for our best-guess estimate, we need detailed knowledge of the geographical distribution of mining activities. However, we do not need this information for the upper- and lower-bound estimates. Both scenarios act as hypothetical boundaries – representing a best case and worst case – encompassing our best-guess estimate.

In the upper-bound scenario, the premise is that Ethereum miners rely solely on coal-fired power, the energy source with the highest emissions per kilowatt-hour of generated electricity. Conversely, the lower-bound scenario assumes that all the electricity used stems from hydropower, the principal renewable energy source in mining.

This stark simplification, applied to both scenarios, makes translating electricity consumption to GHG emissions easier. Since we assume a single energy source is used, (I_s,d) remains constant, regardless of how the mining landscape evolved. Thus, it follows that changes in the geographical distribution of mining operations, which may otherwise have changed the network's electricity mix profile, do not influence the emission intensity. This assumption means that these two estimates are not affected by temporal adjustments in the methodology depending on data availability during specific periods. We further explain these adjustments in the best-guess section.

The values required to calculate GHG emissions for both scenarios – emission intensity and electricity consumption – are readily available. The GHG emissions estimate for any given day d is a function of the emission intensity of energy source s, where s ∈ S, multiplied by the corresponding electricity consumption of that day (Equation 1).

Best-guess estimate 

As highlighted earlier, an in-depth understanding of the geographical distribution of Ethereum mining activities is pivotal for determining our best-guess estimate, especially considering the stark variations in energy sources used to generate electricity across different regions. To address this need, we devised a strategy to gather available data from various sources, thereby reconstructing the geographical distribution of Ethereum mining activities.

This strategy primarily leverages information from Ethereum blocks to determine potential regions where a given block may have been mined. However, not every region identified in the block data directly aligned with individual countries, which could have then been easily mapped to their specific electricity mixes. Instead, regions such as Europe and Asia were identified, representing broader geographical territories comprising multiple countries. Consequently, we introduced a methodology to determine how mining activities were allocated within these territories.

This section explains all the steps necessary to derive a more reliable estimate of GHG emissions associated with Ethereum mining than what could be achieved using the stark simplifications in the lower- and upper-bound scenarios.

The section starts by introducing our block allocation algorithm. This methodology is central to our research, serving as the foundational data that informs our understanding of the geographical spread of Ethereum mining. In several instances, we needed supplementary data sources to estimate the mining distribution within certain identified regions. Accordingly, we adopted three distinct methods (referred to as ‘time intervals’) based on the data at our disposal. Each interval has a unique approach for gauging mining activity distribution within the regions discerned from our Ethereum ledger analysis. The following segments in this section describe the methodologies and data sources used.

Block allocation

Given the absence of a mining map for Ethereum, similar to that used for Bitcoin, which draws on data from mining pools to infer the geolocational spread of mining, an alternative method was required to gain insights into the spatial evolution of Ethereum mining.

Our analysis includes all Ethereum blocks from the genesis block (block 1) to block 15,535,776 (30 July 2015 to 14 September 2023). The data extracted forms the bedrock for mapping the geographical distribution of Ethereum mining. The analysis allocates Ethereum blocks into 18 distinct regions (Appendix 1) based on specific patterns that often indicate a country, territory or mining pool with which the block could be associated. Figure 2 illustrates the allocation algorithm used, which we shall now examine more closely.

Figure 2: Block allocation algorithm

Each block is first screened to determine if it contains any indication of the region where it may have been mined. This is achieved by examining the extraData field. The extraData field in Ethereum blocks is a component of the block header, capable of displaying arbitrary text limited to 32 bytes. This field enables mining pools or individual miners who have successfully mined a block to mark it for recognition and crediting, thereby promoting transparency within the Ethereum network.

The block is allocated to the relevant region if the extraData field contains any of the terms in Appendix 2.

If no match is found in this initial analysis, the next step is to try to identify the block's miner through the beneficiary field. The address or addresses used by mining pools are often public knowledge. If the address to which the block reward has been sent corresponds to any of those listed in Appendix 3, the block is allocated to the corresponding pool. The regional distribution associated with each mining pool is detailed in Appendix 4. Blocks that cannot be mapped to a region or mining pool are classified as ‘Others’.

Equation 2 shows the process of establishing the block count each day for all regions. Here, we aggregate the blocks allocatable via extraData (b_r,d) and miner addresses (m_r,d) to compute the total blocks attributable to region r (B_r,d).

Figure 3 shows the approximated geographical distribution of Ethereum mining activity over time based on the block allocation algorithm.

Figure 3: Geographical distribution of Ethereum mining activity

Note: The graph focuses on four key regions: ‘Asia’, ‘Europe’, ‘North America’ and ‘Others’. We have chosen to illustrate the data in this way to make it easier to interpret. Asia includes the identified regions (Asia, Eastern Asia, China, Japan, Singapore, Republic of Korea and Taiwan, Province of China. Europe includes (Europe, Eastern Europe, Northern Europe, Western Europe, Germany, Russia, Sweden and Ukraine). North America includes (Canada, North America, the US, Eastern US and Western US).

Computing the best-guess estimate

Two primary components are essential to compute an estimate for GHG emissions attributable to Ethereum for a given day d. The first is Ethereum’s annualised electricity consumption (E_d), which can readily be sourced from our previous research. The second is an Ethereum-specific emission intensity (I_d). Determining the latter requires a series of steps, which we outline in more detail later in this section. Once we have both components, the emission intensity associated with the network is multiplied by the annualised electricity consumption estimate (Equation 3).

As mentioned, in contrast to the simplified scenarios, I_d is not readily available, and a multi-step procedure is required to obtain a daily value. The foundation of that procedure is extracting information from Ethereum blocks to associate each one to a region (Appendix 1) where it has likely been mined. This helps us better understand the electricity mix that can be associated with Ethereum on any given day and how it has changed over time.

Yet, as mentioned, not every identified region corresponds to a single country, enabling it to be easily mapped to its respective electricity mix. Instead, some correspond to broader geographical territories such as Europe or Asia. For such cases, we needed a method to derive an emission intensity that could be associated with such regions.

However, before computing an emission intensity for all identified regions, they are separated into two categories (Table 4). This step aims to segregate blocks that contain useful information (Category 1) from those that do not (Category 2).

Table 4: Region categories

A simplified procedure was used for regions in Category 1 as we could not obtain any valuable data for those blocks. For regions in Category 1, we used a wide variety of data sources to compute the most reliable emission intensities possible at any given time.

Since only a few regions in Category 1 are countries, some had to be disaggregated into their constituent countries, provinces or states (shown in Appendix 6) to compute their emission intensities. Further, it is important to recognise that even after this breakdown an emission intensity cannot simply be derived as it does not give an indication on how mining activity is being distributed within a region.

Once an emission intensity for each region has been obtained, the Ethereum-specific emission intensity (I_d) can be computed by aggregating the proportional emission intensities for all regions (i_r,d) (Categories 1 and 2), as shown in Equation 4.

The next two sections, ‘Category 1: Regional data available’ and ‘Category 2: No regional data available’, will outline the intricacies of calculating i_r,d depending on whether data is available or not.

Category 1: Regional data available

All blocks in Category 1 contain regional information indicating where the block may have been mined. This information stems from a geographic indicator in the extraData field or a miner for which geographic data was available (Appendix 5).

Knowing the geographical distribution of mining activity is vital to determining an appropriate electricity mix. However, even if we obtained regional data, it was still necessary to introduce certain assumptions about hashrate distribution for Category 1 regions that constituted a territory rather than an individual country, leading to an emission intensity not readily be available. To that end, we tried to leverage all information available to us for each period.

The approach we applied for Ethereum is similar to the methodology used for Bitcoin's GHG emissions. However, Ethereum's data has its own set of challenges. While we have regional information from 30 July 2015 to 14 September 2022, interpreting it is not always straightforward. Some identified regions, as opposed to specific countries, represent vast geographical areas. Therefore, breaking these regions into their constituents is imperative; a key challenge being determining the mining activity distribution within these territories. We devised three time intervals to navigate this, ensuring consistent data application.

Time window segmentation based on data availability

 This section delineates the three time intervals underpinning the emission intensity calculations for our best-guess estimate.

Throughout the observed period (30 July 2015 to 14 September 2022), data on region allocation of blocks was consistently available. However, these regions had to be broken down into their constituent parts to compute emission intensities. For example, when considering a broad region like Europe, developing a clear process to allocate weights to its constituent countries is vital to establish how mining activity is distributed within such regions.

To that end, we devised three approaches, now referred to as time intervals. Each interval uses a different method, depending on data availability, to allocate mining activity within a region, as shown in Figure 4.

Figure 4: Time intervals

Each interval has its inherent assumptions and limitations. We would like to reiterate here that these intervals refer to the distribution of mining activities within the identified regions (Appendix 6) and not how the blocks are distributed among the regions in Appendix 1. Below is an overview of the three intervals, followed by a more detailed explanation.

•  Interval 1: This method was employed from July 2015 to August 2019. With no proxy data available that provided insights into how mining activity was distributed within identified regions, we adopted a procedure similar to that used by prominent climate data providers when faced with comparable challenges.¹ Specifically, this method presumes that Ethereum mining operations are distributed proportionally based on electricity generation. For instance, if a country like France contributes 10% to a region's (in this case, Europe) electricity production, it is assumed to also host 10% of that region’s Ethereum mining operations. This approach enables us to determine an electricity mix for all regions, regardless of whether detailed data is directly available.

• Interval 2: This method was employed from September 2019 to January 2022. It leverages the geographical distribution of Bitcoin mining to supplement our Ethereum data. Our analysis of Ethereum blocks identifies 18 distinct regions (Appendix 1), and if used as a proxy for Ethereum mining, Bitcoin's data offers insights at the country or even intra-country level. By using both datasets, we obtain a richer, more comprehensive perspective. We use Bitcoin's data if an Ethereum region cannot be directly tied to a country-specific emission intensity. For instance, let us take the region Asia, which consists of multiple countries, as an example. While in Interval 1, a country's electricity generation determines its weighting, in this case, it is bitcoin mining activity. For example, if China is responsible for 70% of bitcoin mining activity in Asia, China's emission intensity is weighted with 70% when determining the emission intensity of the region Asia.

• Interval 3: This method was employed from February 2022 to September 2022. It follows the same approach as in Interval 2, except that we assume the last data points in Interval 2 remain the same due to the absence of more recent information on the geographical distribution of bitcoin mining.
  
  

Each interval's assumptions influence the emission intensity estimate for that duration. Compared to our Bitcoin analysis, the availability of regional data from 30 July 2015 to 14 September 2022 significantly reduces our dependence on broad generalisations like using the global emission intensity for an entire interval (Bitcoin: Interval 1). Figure 5 demonstrates that we could compute an Ethereum-specific emission intensity throughout the observed period.

Figure 5: Ethereum-specific emission intensity

The following sections delve into the rationale behind each interval, detailing the computational steps and data sources used.

Interval 1

This interval computes the proportional emission intensities for all Category 1 regions between the first block mined and 31 August 2019, the day before data on the geographical distribution of bitcoin mining became available.

With no detailed data on the distribution of mining activity within regions and no suitable proxy available, we reverted to an approach similar to what prominent climate data providers use in comparable circumstances,[3] namely, deriving the electricity mix of a region by aggregating the electricity generated by all energy sources from all its constituents.

Equations 5–9 show the derivations of region-specific emission intensities in Interval 1 for all r ∈ R | r ≠ Others. The proportional emission intensity of each region r (I_r,d) is determined by multiplying the region-specific emission intensity (I_r,d) by the region’s share of total block allocation (S_r,d) (Equation 5). This approach ensures that regional differences in the emission intensities of power generation are considered.


with,

Before computing the proportional emission intensity of all Category 1 regions, the emission intensity for each region I_r,d must be derived (Equation 7). (I_c,d) is the weighted sum derived from all the emission intensities of countries or intra-country regions (I_c,d) and their respective share of region r’s total electricity generation (Equation 8).

with,

Each country or intra-country region’s emission intensity is based on the local energy sources used in electricity generation. Here, the share of each energy source (S_c,s), derived from the electricity mix of every country, province or state, is multiplied by the corresponding emission intensity (C_s) of that energy source (Table 3). The weighted emission intensities of all sources are then aggregated to form the country or intra-country region’s distinct emission intensity (I_c,d) (Equation 9).

Interval 2

Interval 2 provides the proportional emission intensities of Category 1 regions between the first and the last data point available in our Bitcoin mining map (1 September 2019 to 31 January 2022). This interval uses data on the geographical distribution of bitcoin mining activity as a proxy.

We can now (unlike in Interval 1) base assumptions on how mining activity is distributed within regions on actual cryptocurrency mining data, assuming the geographical distribution of Ethereum mining mirrors that of bitcoin mining.

It should be noted that within this interval, the availability of state- and province-level mining activity data differs from that of country-level data. If the first available data point for a country, province or state does not coincide with the start of Interval 2 (1 September 2019), Interval 1 continues to apply. If the last data point does not coincide with the end of Interval 2 (31 January 2022), Interval 3 is adopted in advance.

Equations 10–14 show the derivations of region-specific emission intensities within Interval 2 for all r ∈ R | r ≠ Others. The proportional emissions intensity of each region r(Ir_,d) is determined by multiplying region-specific emission intensities (I_r,d) by the region’s share of total block allocation (S_r,d) (Equation 10). This approach ensures that regional differences in the emission intensities of power generation are considered.

with,

Before computing the proportional emission intensity of all regions in Category 1, the emission intensity for each region (I_r,d) must first be derived (Equation 12). (I_r,d) is the weighted sum derived from all the emission intensities of country or intra-country regions (I_c,d) and their share of the total mining activity in the region r (Equation 13).

with,

Each country or intra-country region’s emission intensity is based on the local energy sources used to generate electricity. Here, the share of each energy source (s_c,s), derived from the electricity mix of every country, is multiplied by the corresponding emission intensity (C_s) of that source. The weighted emission intensities of all sources are then aggregated to form the country's distinct emission intensity (I_c,d) (Equation 14).

Interval 3

Interval 3 defines the partial emission intensities for Category 1 regions between 1 February 2022 and 14 September 2022. While this time frame is currently integrated into the methodology, it is important to note that data up to September 2022 was unavailable at the time of its inclusion. As new data surfaces, we anticipate that this interval will become redundant so that Intervals 1 and 2 span the entire observation period from 30 July 2015 to 14 September 2022.

Occasionally, a whole region or constituent within a region will adopt Interval 3 earlier than 1 February 2022. This applies to the regions US, Eastern US, Western US, China, Asia and Eastern Asia, given that for the US and China, intra-country data on bitcoin mining up to the end of January 2022 is unavailable. Therefore, because the last data point does not coincide with the end of Interval 2, the assumptions made for Interval 3 apply before 1 February 2022.

Creating this interval allows us to estimate emission intensities for all Category 1 regions despite the lack of concurrent insights into the geographical distribution of bitcoin mining activity post-January 2022. To bridge this data gap, we use the most recent data points available as a proxy. This method permits the ongoing calculation of emission intensities for all regions, presuming that the latest data remains a valid representation of the geographical distribution of Ethereum mining activity.

This implies that if, for instance, Bitcoin mining map data is available until date , but not until 14 September 2022 (t₁), the geographical hashrate distribution for date is used as a proxy for the period between the two dates. Figure 6 shows a graphical representation. Here, date t represents the latest data point available on our bitcoin mining map for any country or intra-country region (in most cases, 31 January 2022), and date t₁ represents the 14 September 2022. Interval 3 applies for all time points between t and t₁. However, Interval 3 should be viewed with caution, as previous estimates will be retroactively revised when the actual data becomes available.

Figure 6: Shift from Interval 2 to Interval 3 after new mining map data is released

Equations 15–17 show the derivations of region-specific emission intensities within Interval 3 for all r ∈ R | r ≠ Others. The approach used to compute i_r,d is similar to that used for Interval 2 in that Equations 10 and 11 still apply. The region-specific emission intensities i_r,d is then calculated by multiplying the latest available shares of all country or intra-country regions (s_c,t) by the corresponding emission intensities.

with,

and,

Category 2: No regional data available

Unlike the methodologies employed for Category 1 regions, Category 2 adopts a more streamlined approach. The ‘Others’ region includes all blocks that cannot be identified through the extraData field information or the miner address (Figure 1). In the absence of specific regional data, we resort to the global electricity generation mix as a proxy. This implies that Ethereum mining is globally distributed in proportion to each country's share of total electricity generation.

Equations 18 and 19 show how the proportional emission intensity i_r,d (where r = Others) is derived.

with,

Limitations of the model

Since our estimate of Ethereum's GHG emissions is partly based on previous research on Ethereum electricity consumption and Bitcoin mining map data, the associated assumptions and limitations of both data sources also apply. In addition, a set of new assumptions and limitations have been introduced, specific to determining GHG emissions.

As mentioned in the introduction, this study addresses the environmental impact resulting from the use of dedicated mining hardware, in other words, the emissions produced by the electricity consumed to operate the equipment. (A more detailed explanation of what this comprises can be found on our website.) Therefore, this study cannot be considered a life-cycle assessment. Furthermore, for simplification, we assume that marginal emissions equal average emissions, meaning that a given location's emission intensity (in terms of gCO₂e/kWh) is not influenced by any additional demand from Ethereum mining operations. Below is a brief overview of key assumptions carried over from previous studies.

Electricity rates greatly influence electricity demand. This study uses the default average electricity cost per kilowatt-hour on our website. Other costs are not considered. Another crucial parameter that affects electricity consumption is the average efficiency of the mining hardware.

Besides electricity demand, the location of Ethereum mining operations is imperative. To that end, we partly depend on geolocational data of Bitcoin mining operations, which has been used as a proxy. As outlined in the corresponding methodology section, creating the Bitcoin mining map depends on a sufficiently large sample size and the non-existent or insignificant use of VPNs or proxy services by bitcoin miners.

In addition to the assumptions and limitations inherited from previous studies, estimating Ethereum’s GHG emissions required introducing a new set of assumptions, which include the following:

• Central to our model is block metadata related to information in the extraData field. This field allows miners to add up to 32 bytes of arbitrary text, which they frequently use to identify themselves. Some miners also added geographical information that we use as an indication of the mining pool server’s location and, by extension, a proxy for the location where the block has been mined. While there are trade-offs, such as latency, that grow with distance, the indication of the mining pool server location may not necessarily coincide with the location of the mining node.

• In instances where we could not derive meaningful information from the extraData field about the location of the mining pool server or mining node, we resorted to identifying the miner via the miner address. Appendix 4 details the regional allocation of blocks identified using this method. The key assumption in this case is that the utilised regional allocation for a given pool mirrors that of the actual distribution.

• While undertaking best efforts to reduce misallocation, there is a remaining chance that blocks whose regions were identified via the extraData field contain a pattern that randomly occurred and may match any of those in Appendix 2, but has not been captured by any of our misallocation mitigation measures.

• The GHG emission intensities presented in Table 3 are standardised values and apply uniformly to all regions belonging to one of the two categories listed in Table 4.

• Before data on the geographical distribution of bitcoin mining was available, for all regions defined as Category 1 in Table 4, the power generation mix of a region, determined by the electricity generation profile of its constituents, is assumed to be a fair approximation of the mix used by Ethereum miners.

• Throughout the observed period, the world electricity mix is assumed to be an appropriate reflection of Ethereum mining for all Category 2 blocks and those countries for which no electricity mix could be obtained (Appendix 7).

• When data on the geographical distribution of Bitcoin mining is available, it serves as a proxy to gauge the distribution of Ethereum mining within regions. In its absence, we default to the most recent available data, assuming it continues to be a fair approximation.

• It is assumed that the Ethereum mining load remains constant throughout the day and the electricity mix does not vary between daytime and night-time.

• Our estimates do not account for any activities that could reasonably be expected to reduce emissions, such as using flare-gas, off-grid (behind the meter) mining operations, waste heat recovery or carbon offsetting.

   

  While the majority of limitations may not substantially affect the model's efficacy, we are cognisant of its imperfections. The CBNSI is an ongoing project, being constantly updated in response to shifting circumstances. All modifications are transparently documented in the Change Log.

  If you have suggestions on how we could improve the Index, please feel free to send us a message.

  How does our estimate compare to other estimates?

  Historically, numerous studies have sought to assess Ethereum's electricity consumption. Yet, there is a paucity of published studies that focus on Ethereum’s emissions. Table 5 showcases the estimates of several of those available. Notably, other studies have solely considered carbon dioxide (CO₂) in their evaluations.
  
  

Table 1: Available studies and articles on Bitcoin's environmental footprint

^a The Merge took place on 15 September 2022. Owing to the lack of data for the entire day, 14 September 2022 has been selected as the final day for our estimate.

Acknowledgements

We want to express our sincerest gratitude to the following individuals who helped us by reviewing our work and providing invaluable feedback:

Christian Stoll, Lena Klaaßen, and Ulrich Gallersdörfer from Crypto Carbon Ratings Institute, Lambis Dionysopoulos (Henley Business School, University of Reading), York Rhodes (Microsoft), and Zining Wang (University of Bristol)

Appendices

 Appendix 1

Regions identified

Appendix 2

Regional identifiers

 Source: McDonald (2022) and own observations

Appendix 3

Mining pools and corresponding addresses

Source: McDonald (2022) and own observations.

Appendix 4

Mining pool region distribution

[4] In their research, Silva et al. (2020) designate the region as Central Europe. However, the United Nations Standard country or area codes for statistical use (M49), employed in this study, do not recognise a region termed Central Europe. Considering the node identified as Central Europe operated in Czechia, the region has been reclassified as Eastern Europe.

Appendix 5

Categorisation of regions

Appendix 6

Countries or intra-country regions associated with each identified region

 Note: We adhere to United Nations classifications for countries within broader geographical regions or national classifications for states or provinces.

 Appendix 7

Countries not considered in the analysis

 Note: No electricity mix profile could be obtained for the countries above.

[2] https://ember-climate.org/app/uploads/2022/03/GER22-Methodology.pdf

[3] https://ember-climate.org/app/uploads/2022/03/GER22-Methodology.pdf