On Conducting a Life Cycle Assessment of Network Traffic

(1)

IN

DEGREE PROJECT ENERGY AND ENVIRONMENT, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2021,

On Conducting a Life Cycle Assessment of Network

Traffic

A Qualitative Analysis of Current Challenges and Possible Solutions

TOVA BILLSTEIN

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ARCHITECTURE AND THE BUILT ENVIRONMENT

(2)

sara

On Conducting a Life Cycle

Assessment of Network Traffic

A Qualitative Analysis of Current Challenges and Possible Solutions

TOVA BILLSTEIN

Supervisor

ANNA BJÖRKLUND

Examiner

ANNA BJÖRKLUND

Supervisor at IVL Swedish Environmental Research Institute TOMAS RYDBERG

Degree Project in Sustainable Technology KTH Royal Institute of Technology

School of Architecture and Built Environment

Department of Sustainable Development, Environmental Science and Engineering SE-100 44 Stockholm, Sweden

(3)

i

Abstract

There is a growing demand for climate reporting of digital solutions and Internet services. However, the impacts of data transmission have historically been the least studied part of the ICT sector and in the few studies that exist, the magnitude of Internet energy intensity varies by a scale as large as 20,000.

This indicates that the assessment of network traffic is a complex task, and there is currently no consensus of how to correctly assess it.

In an attempt to guide process development within the area, this report sought to identify and address potential challenges with assessing the environmental impact of network traffic during its life cycle.

This was completed through a combination of a literature review and semi-structured interviews with experts in the field. Several areas in the form of knowledge gaps, unsolved methodological issues, and areas in need of further development were identified and addressed.

Eight key challenges were identified and relate to the areas of system boundaries, data collection methods, energy intensity metrics, transparency and data availability, age of data, allocation procedures, assumptions on inventory level, and impact categories. In an attempt to address said challenges, several suggestions on how to proceed were presented, as well as areas in need of further investigation. It was furthermore found that the sector should strive to agree upon a number of parameters of significance to enable future harmonized studies of the environmental impact of network traffic during its life cycle.

Keywords

Life Cycle Assessment, LCA, network traffic, data traffic, data transmission, network, data center, sustainability assessment, environmental impact, carbon emissions, ICT, information and communications technology, technology, electricity intensity, industrial ecology, climate change.

(4)

ii

Sammanfattning

Efterfrågan på klimatrapportering av digitala lösningar och Internettjänster ökar allt mer. Samtidigt är effekterna av datatrafik historiskt sett den minst studerade delen av IKT-sektorn, och i de få studier som finns varierar storleken på Internets energiintensitet med en skala på 20 000. Detta indikerar att bedömningen av nätverkstrafik är en komplex uppgift, och i nuläget saknas en konsensus kring hur det bäst kan mätas.

I ett försök att vägleda processutveckling inom området försökte rapporten identifiera och analysera potentiella utmaningar som kan uppstå när man bedömer miljöpåverkan av nätverkstrafik under dess livscykel. Med en kombination av en litteraturstudie och halvstrukturerade kvalitativa forskningsintervjuer med experter inom forskningsområdet identifierades och behandlades ett flertal områden i form av kunskapsluckor, olösta metodologiska frågor och områden i behov av vidareutveckling.

Resultatet visade att åtta utmaningar av hög relevans existerar inom områdena systemgränser, datainsamlingsmetoder, energiintensitetsmätvärden, transparens och datatillgänglighet, snabb teknikutveckling, allokering, antaganden och miljöpåverkningskategorier. I ett försök att ta itu med de nämnda utmaningarna presenterades ett flertal förslag till lösningar samt områden som behöver undersökas ytterligare i framtiden. Det konstaterades dessutom att sektorn behöver sträva efter att enas om ett antal parametrar av betydelse för att möjliggöra framtida harmoniserade studier av nätverkstrafikens miljöpåverkan under dess livscykel.

Nyckelord

Livscykelanalys, LCA, datatraffik, nätverk, datorhall, miljöanalys, miljöpåverkan, ICT, informationsteknik, teknologi, industriell ekologi, klimatförändring.

(5)

iii

Acknowledgements

This degree project of the second cycle (30 credits) completes a five-year study at the Royal Institute of Technology (KTH) at the program Energy and Environment, resulting in a Master of Science in Engineering in Sustainable Technology. The project was carried out during the spring of 2021 in collaboration with IVL Swedish Environmental Research Institute where Tomas Rydberg, researcher and LCA expert, shouldered the responsibility as the organization’s supervisor. Examiner and academic supervisor at KTH was Anna Björklund, teacher and director of studies at SEED.

First and foremost, I would like to express my gratitude towards Anna who contributed with a lot of valuable guidance and knowledge throughout the whole project. The feedback and support I received greatly aided the continued process and structure of the academic work. I furthermore wish to recognize the valuable guidance received from Tomas. Thank you for your assistance along the way, as well as your expertise and genuine interest in the subject. I also wish to thank the interviewees Dag Lundén, Jens Malmodin, Lorenz Hilty, Roland Hischier, Pernilla Bergmark, Vlad Coroama, Sara Gorton and the anonymous interviewee, for taking the time to share your expertise knowledge within the subject, as well as for the fruitful discussions and guidance on where to look next.

__________________________

Tova Billstein

Stockholm, May 2021

(6)

iv

List of Abbreviations

Abbreviation Meaning

AEC Annual Electricity Consumption

DC Data Center

CPE Customer Premise Equipment EPD Environmental Product Declaration

FU Functional Unit

GDPR General Data Protection Regulation

GB Gigabyte

ICT Information and Communication Technology ISO International Standards Organization

ISP Internet Service Provider

LCA Life Cycle Assessment

LCI Life Cycle Inventory

LCIA Life Cycle Impact Assessment

PC Personal Computer

PCR Product Category Rule

(7)

v

Table of Figures

Figure 1. Network traffic and included subsystems ... 4

Figure 2. The four phases of LCA. ... 6

Figure 3. The methodological steps included in the report. ... 10

Figure 4. Identified system boundaries of network traffic in previous literature. ... 18

Figure 5. Modified version of Figure 4. ... 19

Figure 6. Alternative version of Figure 4. ... 20

Figure 7. The preferable choice of metric per equipment. ... 24

Table of Tables

Table 1. Network traffic subsystems and descriptions. ... 5

Table 2. Included elements in the report that strengthen its validity and reliability. ... 14

Table 3. Identified challenges and summarized results. ... 15

Table 4. Interview guide questions. ... 39

Table 5. Chosen interviewees and other interview information. ... 41

Table 6. GDPR routine. ... 42

(10)

1

1 Introduction

During the last 50 years, a new industry emerged called the Information and Communication Technology (ICT) sector. These decades can be described as the period where humanity became dependent on information and communication technology, which now continues to grow and consume increasing amounts of energy and material (Appelman et al., 2013). The environmental footprint of the ICT sector itself is estimated to be significant (Coroama and Hilty, 2014), and further developments within network technologies and the continued growth of Internet services have led to increasing volumes of network traffic (Bhuyan et al., 2017, pp. 1-4), with an approximate annual growth of 30%

in Sweden (Malmodin et al., 2014). At the same time, the evolution of Industry 4.0, and related technologies such as the Internet of Things and big data analytics, are on the rise. However, their specific environmental impact is still unknown, with little research devoted to investigating the impact on sustainability aspects (Oláh et al., 2020, Bonilla et al., 2018), resulting in a greater need for climate reporting of all included parts.

There are few studies that comprehensively evaluate the impact of the whole ICT sector with all elements included (Malmodin et al., 2014), and the impacts of network transmission itself have historically been the least studied part of the whole sector (Malmodin et al., 2012). Even research on energy consumption that could be expected to provide insight into the impacts of network traffic does not do it justice, either by not separating the data transmission from the data center’s energy consumption, by using only a fixed percentage, or by not considering it at all (Coroama and Hilty, 2014).

In the few studies that exist, the magnitude of energy intensity, i.e. data transmission, varies by a scale as large as 20,000, indicating that the assessment of Internet transmission is a complex task. It is often a controversial subject of discussion (Coroama and Hilty, 2014), and there is currently no consensus on how to correctly measure it.

One well-established tool to measure contributions from different life cycle stages to facilitate relevant sustainable measurements is by performing a life cycle assessment (LCA) (Finnveden et al., 2009), a technique that addresses all potential environmental impacts of the product from raw material extraction to final waste disposal (ISO, 2006b). However, several challenges in the form of knowledge gaps, unsolved issues and methodological differences remain that hinder development within the area of network traffic (Malmodin et al., 2014, Malmodin et al., 2012, Achachlouei et al., 2013). More research is therefore needed that both explores the life cycle of network traffic and pinpoints the current challenges as well as addresses them in a way that enables future harmonized LCAs within the area.

(11)

2

1.1 Aim and Objectives

This master thesis aims to provide LCA practitioners, ICT experts and other relevant stakeholders with an overview of the current challenges, and possible solutions, in an attempt to guide process development and enable the future creation of life cycle assessment. Through a combination of a literature review and semi-structured interviews with experts within the field of research, the degree project seeks to identify knowledge gaps, unsolved methodological issues, and areas in need of further development when performing an LCA.

The aim was divided into two research objectives:

1. Identify, examine, and assess challenges that might occur when measuring the environmental impact of network traffic during a life cycle assessment

2. Investigate how the identified challenges could be addressed

(12)

3

2 Background

This chapter consists of three sections. Firstly, a brief introduction to network traffic is given where relevant terminology is explained. Thereafter, LCA and its four phases are explained in further detail, followed by a summary of previous LCAs on network traffic and other relevant studies within the ICT sector.

2.1 Network Traffic

Network traffic, also known as data traffic and data transmission, is defined as the data present in an active network. The network must contain a minimum of two devices and be connected through a link to fully function, the latter being a communication pathway that allows the data to transfer from one device to the other (Bhuyan et al., 2017, pp. 1-4). The constant transfer of information is moved in small units called packets, where the information initially is fragmented only to be reassembled at the destination (ibid, p. 30). The volume of the transferred data packets represents the load in the network, which if too large might lead to congestions. By adjusting the bandwidth, i.e. the maximum data transfer capacity, and congestions can be avoided (ibid, pp. 1-4). To ensure efficient communication, protocols are often required, which can be described as a set of rules that govern the format of the packet and messages (ibid, pp. 16 and 25). The data packets, i.e. network traffic, move through the Internet via a network of globally distributed routers connected via a fiber-optic communication technology (Schien et al., 2012).

2.1.1 The Network Subsystems

The Internet network contains many different subsystems that engage with each other to transfer the information from one destination to another (Aslan et al., 2018). There is no standard way of categorizing the different included components. However, three decompositions seem to be more frequently used than others in recently published research articles (Schien et al., 2015, Schien and Preist, 2014, Aslan et al., 2018, Malmodin and Lundén, 2018). A combination of the three, where all mentioned subsystems have been included, is further described in Figure 1 and in Table 1.

(13)

4

Figure 1. Network traffic and included subsystems

(Aslan et al., 2018, Malmodin and Lundén, 2018, Schien and Preist, 2014)

Network transmission, illustrated in Figure 1, originates in a consumer’s user equipment when data packets move to the customer premise equipment (CPE) (Malmodin and Lundén, 2018, Aslan et al., 2018). The information then moves via the access network, through the metro network towards the IP core network (also known simply as the core network or long-haul network) before reaching its final destination (Shien and Priest, 2014). The majority of the end-user traffic is directed to servers in a data center (DC), although it can also be directed to other user equipment (Schien et al., 2015). The two subsystems users and DCs are often categorized together as endpoints under the name “end devices”

(Coroama and Hilty, 2014, Schien and Preist, 2014).

While the above-mentioned subsystems are commonly accepted and defined the same way in many research articles, other subsystems are not as recurring. The components, illustrated in Figure 1 as

“Uncommon subsystems”, were not as recurring in literature. The IP core network, often defined as including all supporting infrastructure, is sometimes separated from the undersea cables which can be described as the intercontinental network transmission (Schien and Preist, 2014). Furthermore, the subsystem metro network is occasionally illustrated as a subsystem on its own, depending on the system boundaries and object of the study (Schien and Preist, 2014). Lastly, operator activities have also been identified as an important subsystem when assessing the whole life cycle of network traffic (Malmodin and Lundén, 2018), although it has also been known to be excluded from illustrations of the system in question.

(14)

5

Table 1. Network traffic subsystems and descriptions.

Subsystem Description Examples

User equipment Equipment used by a consumer to access the network¹

Game consoles, tablets, personal computers (PCs)^1,2 Customer premise

equipment (CPE)

Network connecting devices that link the user equipment to the network¹, operated by customers³

Wi-Fi routers and modems^1,3

Access network Equipment used by the consumers to connect to the Internet Service Provider (ISP)¹, connects to metro network³

Routers, DSLAMs and cables¹

IP core network ISP equipment that forms the network¹ (sometimes referred to as network core³)

Supporting infrastructure, routers, switches, etc.¹ Data center (DC) Large facilities that store data in servers and

house other equipment¹

Servers, power supply units, cooling systems¹ Undersea cables Cable infrastructure over long distances¹,

intercontinental³(sometimes grouped under IP core network¹)

Submarine cables, fiber optic cables¹

Metro network Sometimes referred to as part of IP core network, often defined as including the edge network³

Edge routers etc.³

Operator activities

- Offices, buildings, stores,

travel, vehicles²

1 (Aslan et al., 2018), ² (Malmodin and Lundén, 2018), ³ (Schien and Preist, 2014)

2.2 Life Cycle Assessment

A life cycle assessment (LCA) is a holistic, cradle-to-grave methodology used for identifying and assessing environmental aspects associated with a product or process throughout its entire life cycle.

The LCA will identify and quantify impacts, as well as potential transfers, of environmental impacts from one medium to another, and identify otherwise unnoticed trade-offs (ISO, 2006a). LCAs can often be divided into two different types: attributional and consequential LCA. The former can be described as considering the specific environmentally relevant physical flows that might occur in the life cycle (Ekvall et al., 2016), and has been deemed the more broadly applied method in society (Finnveden et al., 2009). Consequential LCA, in comparison, is often defined by its aim to describe how environmentally relevant flows might change in response to a decision (Ekvall et al., 2016). It is thus recommended to be used within decision-making, although not when the difference between the attributional and consequential LCA results is small or when the uncertainties outweigh the insights gained from the consequential LCA (Finnveden et al., 2009).

(15)

6

2.2.1 ISO Standards

In the 1990s, the International Standards Organization (ISO) created a series of standards named ISO 14040:1997, ISO 14041:1999, ISO 14042:2000 and ISO 14043:2000, to ensure that future LCAs were to be used for comparative assertions, restrict possible misuse of public applications, and protect third- party interests (Finkbeiner et al., 2006). A revised version of ISO 14040, as well as the new standard ISO 14044, was also released in 2006. The documents are recommended to be used as a reference for all users and practitioners of LCA to enable greater international and stakeholder acceptance (Finkbeiner et al., 2006). ISO 14040 details the principles and framework of an LCA, including all steps necessary within the LCA study, while ISO 14044 contains information relevant for performing the LCA (ISO, 2006b).

2.2.2 The Four Phases of LCA

An LCA study, as shown in Figure 2, consists of four phases: goal and scope definition, life cycle inventory (LCI), life cycle impact assessment (LCIA), and interpretation (ISO, 2006b). As seen in the Figure 2, the process is iterative to ensure continuous development (ISO, 2006a).

Figure 2. The four phases of LCA (ISO, 2006b).

Goal and Scope Definition

The first phase, the goal and scope definition, defines the depth and the breadth of an LCA (ISO, 2006a).

The goal of the LCA clarifies the intended application, reasons for carrying out the study and the intended audience, while the scope specifies the product and its functions, the functional unit (FU), the system boundaries, allocation procedures, the chosen impact categories, data requirements, assumptions made, potential limitations, etc. These can later be modified during the LCA as new data is revealed (ISO, 2006a).

(16)

7 Life Cycle Inventory

The life cycle inventory (LCI) is the second phase within an LCA study and contains an inventory of all input and output data relevant for the studied system, as well as the collection measures used (ISO, 2006b). The inputs and outputs from and within the system are quantitively modeled and often illustrated in some sort of flowchart. The gathered data must also be complete and unbiased, as it lays the foundation for the upcoming impact assessment and interpretation phases. While conducting the LCI, several methodological issues might occur that must be addressed. Some examples include cut-off rules, allocation schemes, recycling, and system expansion (ISO, 2006a).

Life Cycle Impact Assessment

The third phase, also called the life cycle impact assessment (LCIA), contains an assessment that aims to provide additional information to help understand the LCI (ISO, 2006b). It should include a selection of impact categories, classification, and characterization. The goal is to convert the input and output data from the previous step into relevant impact category indicators, to help the LCA practitioner understand and evaluate the significance and magnitude of the potential environmental impacts (ISO, 2006a).

Interpretation of the Result

The fourth and final step of an LCA study, also called the interpretation phase, includes a summary and discussion of the LCI or LCIA results, or both, which then act as the base for possible conclusions, recommendations and decision-making in accordance with the goal and scope of the study (ISO, 2006b).

The aim of the interpretation is to give credibility to the results, which are consistently linked to the goal and scope of the study (ISO, 2006a).

2.2.3 Allocation Procedures

It is not uncommon that a process within an LCA study fulfills more than one function. Such processes are classified as “multiproduct processes” and can exist in three different forms: multiple-output process, multiple-input process, or open-loop recycling. When multifunctions are identified, a methodological allocation problem occurs (Ekvall and Finnveden, 2001). Allocation within LCA has given rise to one of the biggest, long-standing controversial issues in LCA theory (Suh et al., 2010), and can be defined as a procedure of partitioning inputs and outputs of a multiproduct process or system over its multiple products (ISO, 2006b). The problems associated with allocation should be dealt with in the following order: by sub-division, system expansion, or based on physical or other relationships (ISO, 2006a). Sub- division is when the LCA practitioner divides the process into sub-processes, but it requires that the data exists for all produced sub-functions and that they thus are physically separated in space or time. System expansion means examining the boundaries to see if they can be expanded to include alternative production of exported functions, or if other activities can be included in the system (Ekvall and

(17)

8

Finnveden, 2001, ISO, 2006a). Allocation based on physical relationships, also called partitioning, is defined as assigning inputs and outputs of a process to their multiple products according to a ratio based on, for example, mass (Suh et al., 2010, Ekvall and Finnveden, 2001). Where physical relationships cannot be established, the inputs should be allocated by other means that fairly represent the relationship between them, for example, by an economic value (ISO, 2006a).

2.3 LCA of ICT and Network Traffic: An Overview

LCA is currently a widely used environmental assessment tool when studying electronic media and several parts of the ICT sector. The assessment of consumer products (including TVs and computers) are, for example, becoming increasingly popular while business products and TV peripherals are more rare (Arushanyan et al., 2014a). Several LCAs on semiconductors exist, a ubiquitous component within nearly all modern electronics (ICT network infrastructure included) that demand a great deal of resources as well as energy, and that generate significant waste (Krishnan et al., 2008, Boyd et al., 2009).

It is furthermore becoming increasingly popular to conduct LCAs of a comparative character where the differences between a specific development within ICT and the conventional way are studied (Moberg et al., 2010, Achachlouei et al., 2013). The number of LCAs of more sustainable data centers are also increasing (Masanet et al., 2020, Honée et al., 2012), and several research papers exist that study the life cycle of the ICT sector at a macro level (Malmodin et al., 2010, Malmodin et al., 2014).

Even though the LCA tool is widely used on a variety of objects and services, only a small number of research studies currently exist that exclusively study network traffic, while many other do not separate the data transmission from the data center’s energy consumption, use only a fixed percentage or do not consider it at all (Coroama and Hilty, 2014). In the sparse research that exists, several challenges are mentioned on more than one occasion. For example, several studies mention the usage of different metrics when measuring network traffic during transmission (Malmodin et al., 2014, Coroama et al., 2015, Aslan et al., 2018), and there is currently an ongoing discussion debating whether the most commonly used metric today is the most appropriate one (Malmodin, 2020, Arushanyan et al., 2014a).

Many authors also use different terminology to explain the included subsystems as well as mention the difficulty to define appropriate system boundaries as a result of too complex systems (Coroama et al., 2013, Koomey et al., 2004, Coroama and Hilty, 2014, Scharnhorst, 2008, Schien and Preist, 2014).

Several articles furthermore discuss the subject of data availability, where the lack of sufficient data especially along networks and data centers create great difficulties for the LCA practitioners (Itten et al., 2020, Malmodin et al., 2014). An extensive data collection methodology is required, which currently can be performed in several different ways, resulting in a variety of methods that greatly impact the results (Hischier et al., 2014, Koomey et al., 2004, Coroama et al., 2013). Several articles also mention the importance of the age of the collected data, as a result of the rapidly changing technology (Koomey

(18)

9

et al., 2004, Malmodin et al., 2012). The assumptions and decisions made by the LCA practitioners on an inventory level, such as the choice of energy efficiency in the routers and choice of the electricity mix, that both will greatly impact the result (Hischier et al., 2014, Schien and Preist, 2014).

(19)

10

3 Research Design and Method

This chapter consists of six sections. First, a brief introduction to the adopted choice of methodology is given, followed by three sections describing the methodological steps of the literature review, conducted interviews and procedure of selecting identified challenges. Lastly, an analysis of the empirical material and quality assurance is given in one section each.

3.1 The Adopted Exploratory Strategy

This master thesis aims to identify and evaluate potential challenges that occur when examining the environmental impact of network traffic during its life cycle, as well as investigate how the identified challenges could be addressed. As only sparse research literature within the subject exists, an exploratory, qualitative research method approach was adopted. The methodology and included steps are visualized in Figure 3 and consist of three main parts. Firstly, information about network traffic in the form of current challenges was collected from studies found in literature and peer-reviewed articles.

Semi-structured interviews with experts within the field were also conducted, where both challenges and solutions were discussed. The results from both steps were then analyzed and compared. The adopted exploratory strategy in the form of literature review combined with qualitative semi-structured interviews is often a preferable choice when there is little to no prior research within the studied subject, as it allows for a more unstructured approach and for new hypotheses to be formed (Bryman, 2012). The foundation of the research has been based on and originated from the LCA framework, where an attributional cradle-to-gate LCA of network traffic was assumed to be conducted.

Figure 3. The methodological steps included in the report.

(20)

11

3.2 Literature Review

A literature review was conducted with the purpose of summarizing current challenges and solutions identified in published research articles on the subject of LCA and the environmental impact of network traffic (Snyder, 2019). It consisted of an initially unstructured search query followed by a structured search and was conducted during the dates 2021-02-08 – 2021-03-04. The iterative process of snowballing was also performed to ensure that all relevant literature was identified (Wohlin, 2014), alongside the usage of Connected Papers (Connected Papers, n.d.). The databases Scopus and Web of Science combined with Primo (available via the KTH Library), as well as Google Scholar were used during the search, in which 44 scientific articles were retrieved, summarized, and assessed based on their relevance to the study. The following terms were used to find most of the reviewed articles: data traffic, network traffic, environmental impact, LCA, ICT, and energy intensity. The search was limited via the exclusion of articles published before 2000, with the oldest retrieved article dating back to 2004.

All considered articles were published in English.

3.3 Semi-Structured Interviews

As a complementary element to the literature review, several interviews were carried out with the purpose of discussing previously identified challenges, explore possible solutions to said issues as well as discuss potential new challenges. They also intended to give more insight into the world of network traffic and new literature recommendations. The interviews were conducted in the form of semi- structured interviews which is a qualitative collection strategy in which predetermined and open-ended questions are used, and a written interview guide is prepared beforehand (Given, 2008). The guide must not be followed in the fixed order it is written and allows for a deeper topic discussion as there is no fixed range of responses to each question.

3.3.1 The Interview Procedure

A preliminary written interview guide was prepared prior to the interviews (see Table 4 in Appendix A – Interview Guide). The guide was used as a template with the intention to provide a common framework of topics and questions that could be discussed during the interviews. As a result, the questions did not always follow the given order nor were they asked with the same specific phrasing during each interview. The questionnaire was structured into the following parts based on the findings from the literature review:

• Introductory questions

• System boundaries

• Data collection

• Electricity intensity metrics

(21)

12

• Transparency

• Age of data

• Allocation procedures

• Assumptions made on inventory level

• Overlooked impact categories

• Varied user practices

• Other challenges

Follow-up questions and probing questions were also asked to gain further insight into interesting topics, or when answers needed to be clarified. When discussing system boundaries, the interviewees were given the opportunity to both comment on the accuracy of a pre-created illustration as well as choose the most accurate alternative according to their opinion, knowledge, and previous experience. All interviews were conducted over the Internet in the form of video meetings over Microsoft Teams, and during the interviews, notes were taken. In agreement with the interviewees, all interviews were furthermore audio recorded which allowed for transcription and further analysis. The chosen participants were informed, both by email beforehand and orally during the meeting, about the purpose and structure of the interviews as well as of the chosen GDPR routine (see section 3.6.2). They were also given the choice of being anonymous, to which one person agreed. All interviews took place during 2021-03-18 – 2021-03-31 and were conducted in either English or Swedish. A summary of the interviewees, their roles and organizations, the date and duration of the interviews can be found in Table 5 in Appendix B – Overview of Interviews. After each interview, the recordings were listened to and compared to the notes that were taken during the interviews to pinpoint important results and to identify interesting aspects. The recordings were then transcribed which enabled a better familiarization with the subject at hand. If any uncertainties or questions remained after the interviews, the participants were contacted again via email to sort out the unclarities.

3.3.2 Selection of Interviewees

Interviews were held with eight individuals who were chosen for their expertise and previous work within the field of ICT, LCA, and network traffic. The interviewees were contacted via email after being identified as key individuals either via examined research articles or by recommendations by the supervisors or other interviewees. The following individuals were interviewed:

• Interviewee 1: Dag Lundén, Environmental Manager at Telia Company

• Interviewee 2: Sara Gorton, Head of Environmental Strategy at Telia Company

• Interviewee 3: Pernilla Bergmark, Principal Researcher, ICT Sustainability at Ericsson

• Interviewee 4: Jens Malmodin, Senior Specialist at Ericsson

(22)

13

• Interviewee 5: Lorenz Hilty, Professor of Informatics and Sustainability at UZH

• Interviewee 6: Roland Hischier, Head "Advancing Life Cycle Assessment" Group at Empa

• Interviewee 7: Vlad Coroama, Senior Research Associate at ETH

• Interviewee 8: Anonymous, Senior Life Cycle Expert at Telecom company

3.4 Selecting Significant Challenges

The identified challenges from the literature review and interviews were validated based on a chosen set of criteria. First, challenges were sought in research articles and documented based on their reoccurrence in literature. They were then presented during the interviews in which their recognition was discussed.

The result from the literature review indicated that eight key challenges exist, and among these eight, three were commented on in detail during all interviews. The rest were met by recognition in all but two interview each at most, deeming all eight challenges of high relevance. During the interviews, six other challenges were also brought up for discussion by the attendees. However, none of these reoccurred in more than two interviews in total, nor previously during the literature review, and they were thus not deemed relevant enough.

3.5 Analysis of Empirical Material

Before finalizing the report, the interviewees were contacted again by email and given drafts of the report where their statements were included in order to verify their oral statements and to avoid misunderstandings. Their comments were then amended into the report before publication. The empirical data were qualitatively analyzed, and the information gathered from the semi-structured interviews was used to complement the literature review. For the first objective – identifying challenges – information from the literature review was used to a larger extent and the information from the interviews was used as a complement. For the second objective – explore solutions – only sparse data could be found in the literature review and more information was thus used from the interviews.

3.6 Quality Assurance

Qualitative studies must maintain and demonstrate authenticity in their research process and outcomes in order to remain trustworthy (Reynolds et al., 2011). The rigor and quality can be secured and maintained through many different implemented strategies, of which enabling validity and reliability of the results and ethics in relation to GDPR were deemed of relevance in relation to this project. The two are discussed in further detail in the sections below.

3.6.1 Validity and Reliability

Validity and reliability are important aspects of research and it is, therefore, vital to implement research elements that help further enable the credibility of the study. Some common techniques include

(23)

14

triangulation, participant validation of findings, peer review, transparency, and a systematic approach to the design and conduct of the study (Reynolds et al., 2011). Other strategies also exist, of which the sampling of interviewees, standardization of procedures, and discussion of the potential influence of the participants' views and beliefs were of particular relevance (Sargeant, 2012). A selection of elements from both Reynolds et al. (2011) and Sargeant (2012) have been summarized in Table 2 together with informative text that in further detail describe how these strategies have been implemented in this study.

Table 2. Included elements in the report that strengthen its validity and reliability.

Element Implementation

Triangulation (i.e. using more than one method to collect data)

Both a literature review and semi-structured interviews were conducted.

Participant validation of findings Interviewees were able to validate their contributions by email after interviews.

Peer reviewing Implemented to some degree. Supervision and examination by KTH, as well as student opposition.

A clear and systematic approach to the design and conduct of the study

A detailed description of the different parts of the report and how they are interconnected was included. Figures and tables were used to further clarify.

Discussion of the potential influence of the participants' views and beliefs

A discussion of the possible profitability of the interviewees are mentioned briefly in section 5.2.

Mechanical data recording Interviews were recorded.

Standardization of procedures An interview guide and GDPR routine were established prior to the conducted interviews.

3.6.2 Ethics and GDPR

Any ethical aspects of the research that might arise concerning the methodology, process, and the project itself were addressed through attempts of transparency and thoroughly secured and verified empirical information. It is however also important to reflect upon any issues that might arise when including the methodological element of interviews, where the inclusion of informed consent is especially important.

The chosen interviewees were informed, both via email beforehand and orally at the start of the interviews, about the intention of the degree project and its scientific purpose. They were furthermore informed that their consent was a necessity to enable further participation, in light of the General Data Protection Regulation (GDPR). The involved actors were also informed that no sensitive information was to be saved longer than necessary and that they at any time during the interview could choose to not answer a question, withdraw their consent entirely or choose to end the interview. The full GDPR routine applied in the degree project can be found in Table 6 in Appendix C – GDPR Routine.

(24)

15

4 Results

This chapter is divided into ten parts. The first section provides an overview of the identified challenges, flowed by eight parts that represent the identified challenges deemed significant when attempting to conduct an LCA of network traffic. In each section, a review of previous findings in the literature review is presented together with findings from the interviews. The tenth section contains a summary of the identified challenges that have been deemed of less importance.

4.1 Overview of Identified Challenges

Challenges deemed of high significance were identified in the areas of system boundaries, data collection methods, energy intensity metrics, transparency and data availability, age of data, allocation procedures, assumptions, and impact categories. In an attempt to address said challenges, several research approaches were presented as well as areas in need of further development. An overview of the eight key challenges, as well as six other challenges deemed of less significance, can be found in Table 3.

Table 3. Identified challenges and summarized results.

Challenge Summary

System boundaries The complexity of the ICT sector, ever-changing components, the gap of knowledge, and the background of researchers were identified as causes of the shifting system boundaries. Two previously used system boundaries were found to be favorable in future studies. Future research could also model the system as three different components: user devices, network equipment, and DCs.

Data collection methods

Several different methods for collecting data exist, where the top-down method and bottom-up method were deemed to be most recurring even though being flawed and dependent on assumptions. A consensus was found that the best way forward would be to use a combination of the two methods to validate each other.

Energy intensity metrics

Three metrics currently exist in research studies: energy per data, energy per time, and energy per user. Per data was found to be preferable in the IP core and metro networks, while per time is more appropriate in the access network and CPE, due to the two latter subsystems not being dependent on data.

Transparency and data availability

A competitive market, limited research format, more pressing issues and lacking scientific validation processes have been identified as reasons for the currently low transparency within the ICT sector. Market observations show evident signs of development in the right direction; however, time is of the essence. The implementation of a separate, third-party entity was deemed unrealistic.

Age of data Limiting the validity of LCA results to one year of reference was deemed questionable due to included data of older age as well as previously low transparency. It was instead suggested to limit and compare studies via the efficiency gains in the equipment, however, how plausible the solution is remains to be seen due to the difficulty of making such estimations.

(25)

16 Allocation

procedures

Allocation was deemed particularly difficult at the endpoints of data transmission, as well as when assessing unused capacity in networks. A consensus was found that the former could be allocated by time, while the latter by even distribution among all users. More insight into the allocation procedures used for data centers is needed.

Assumptions on inventory level

Assumptions were discussed in relation to electricity mix and choices within bottom-up and top-down data collection methods. Further transparency and communication were indicated to help, however, continued transparent sensitivity analyses were also found to be of high importance.

Overlooked impact categories

The origin of the research field, lack of knowledge regarding LCA and measurement of other categories, the pressing matter of climate change, and a need to please the media have been identified as reasons as to why there is a strong bias towards the carbon dioxide impact. A clearer base for calculations is needed within the other categories, as well as further knowledge of LCA.

Other Several other challenges were identified but deemed of less importance, including the differing degree of knowledge between practitioners, the choice of FU, the lesser focus on mobile networks as well as the production stage, the usage of different models, and difficulty to present the results in a way that the public can understand.

4.2 System Boundaries

The choice of system boundaries is known to be a source of complexity, and occurs when LCA practitioners seek to limit the system and choose the included subsystems (Aslan et al., 2018). The defined system boundaries depend on the purpose of the study, but even in cases where the intention is the same, the boundaries differ. One example of this is when researchers seek to study and measure electricity intensity, where many different combinations occur (Aslan et al., 2018). Since no standard way of categorizing the components exist, several decompositions have been used with significant differences (Schien et al., 2015, Schien and Preist, 2014, Aslan et al., 2018, Malmodin and Lundén, 2018), leading to incomplete studies and rendering the results incomparable (Aslan et al., 2018, Arushanyan et al., 2014a). The challenge has been deemed the most important methodological decision in literature (Aslan et al., 2018, Coroama and Hilty, 2014) as well as during interviews (Interviewee 8, 2021), and a more commonly accepted division thus needs to be developed to enable harmonized studies and better comparability within the research field.

4.2.1 Identified Causes and Previous Findings in Literature

Several causes for not reflecting the full system within the system boundaries were identified during the interviews, including the gap of knowledge among involved actors which has resulted in flawed images of the system (Bergmark, 2021). The shifting system boundaries were also deemed likely to occur due to the complexity of the ICT sector in the form of a large number of involved stakeholders on a global

(26)

17

level (Lundén, 2021), the different background of researchers (Hilty, 2021), and the inclusion of ever- changing components that has resulted in definitions changing ever so often (Lundén, 2021). It was furthermore found that articles often include illustrations of the chosen system boundaries within the particular studies (Aslan et al., 2018, Schien et al., 2015, Schien and Preist, 2014, Malmodin and Lundén, 2018), but that the phrasing and definitions of these subsystems differ and that their chosen system boundaries vary greatly.

Three attempts have previously been made to compare the system boundaries between several studies, of which several study objects were the same (Aslan et al., 2018, Coroama and Hilty, 2014, Coroama et al., 2013). In 2014, Coroama and Hilty summarized ten studies from 2004-2013 and compared the subsystems, which were divided into the categories networking equipment, optical fibers, and end devices, of which the latter included both user equipment and DCs. The results showed that nine out of the ten studies included the subsystems networking equipment and optical fibers, while end devices only were included in three (Coroama and Hilty, 2014). A similar comparison was conducted in 2015 by Coroama et al., where eleven studies from 2003-2013 were examined based on the subsystems CPEs, access network, core network, links (defined as optical fibers used for transmission) and DCs (Coroama et al., 2015, Schien et al., 2015). It was found that all studies included the subsystem core network, while nine articles also included the links, eight the access network, five the CPE, and three also included the DCs (Coroama et al., 2015). The last study, conducted in 2018, had summarized 14 research papers on network traffic from 2004-2016, where eight different combinations of subsystems were identified.

Twelve out of the 14 studies included the access network and IP core network, while only eight studies accounted for the DCs and even fewer included the user devices and CPEs (Aslan et al., 2018).

4.2.2 An Overview of Existing System Boundaries

In an attempt to identify an appropriate choice of system boundaries for future usage, the most recurring combinations of subsystems were identified and illustrated in a figure, based on the findings in the literature review Figure 1(Schien et al., 2015, Schien and Preist, 2014, Aslan et al., 2018, Malmodin and Lundén, 2018, Coroama and Hilty, 2014). The combinations were found to commonly occur in five different forms, illustrated in Figure 4 (alternatives A-E).

(27)

18

Figure 4. Identified system boundaries of network traffic in previous literature.

In an attempt to improve Figure 4 and its representativeness of reality, the interviewees were given the opportunity to comment on the illustration’s accuracy. Several participants found the layout to be representative of the system in question, while a few remarks were given by others. Two interviewees suggested that the subsystems metro network and undersea cables should be added under the IP core network (Malmodin, 2021, Coroama, 2021), with the arguments that the differences between the metro network and IP core network are becoming less and less pronounced as well as that undersea cables are a pure network component that often is categorized with the IP core network (Coroama, 2021). The suggestion was supported by a statement from another interviewee, who argued that no clear definition of the IP core network currently exists (Lundén, 2021). Due to the recurring nature of these suggestions, the illustration was modified to some degree by the input gained from the interviewees. The new version has been illustrated in Figure 5.

(28)

19

Figure 5. Modified version of Figure 4.

The interviewees were furthermore given the opportunity to choose the most accurate alternative when seeking to measure the complete environmental impact of network traffic. Several participants commented that the choice of appropriate system boundaries largely will depend on the purpose of the study as well as the research question (Hischier, 2021, Hilty, 2021). However, with this information in consideration, the preferable choice of system boundaries still differed. Two compilations in Figure 4 were favored: alternative C (with the addition of undersea cables) and alternative E (Coroama, 2021, Hilty, 2021, Hischier, 2021, Lundén, 2021, Malmodin, 2021, Interviewee 8, 2021). Alternative C in Figure 4, with the addition of the subsystem undersea cables (also known as alternative C2 in Figure 5), was deemed to be the preferable option by two interviewees (Coroama, 2021, Hilty, 2021), with one participant adding that the recommendation did not imply that user equipment and data centers should be excluded from the calculation altogether, but rather that they should be differentiated (Hilty, 2021).

Other interviewees preferred alternative E, with one arguing that it is necessary to get the complete picture of the system but that it might be hard to account for the operator activities (Hischier, 2021).

4.2.3 Possible Solutions and Further Suggestions

The interviewees who favored alternative C argued that the remaining subsystems should be differentiated from the rest (Hilty, 2021, Coroama, 2021), in order to make a distinction between the transmission subsystems and other subsystems (Hilty, 2021). It was furthermore suggested that the system should be modeled as three different components: user devices, network equipment and data centers as this division would enable a model that fairly reflects reality. One interviewee argued it would allow LCA practitioners to collect the essential puzzle pieces and put them together in the right combination based on the aim of the study (Coroama, 2021). The division has been illustrated in Figure 6 as the alternatives X1-X3.

(29)

20

Figure 6. Alternative version of Figure 4.

4.3 Data Collection Methods

When estimating the life cycle impact of network traffic, measuring, and collecting accurate data has been proven difficult due to the size and complexity of the ICT sector and its supply chain (Itten et al., 2020, Malmodin et al., 2014). As a result, there is a risk of increased usage of generic data which in turn may add to contradicting results between different LCAs (Arushanyan et al., 2014a). During the literature review, several data collection and calculation methods were identified (Coroama and Hilty, 2014, Coroama et al., 2015, Aslan et al., 2018, Chan et al., 2013), however, all methods were found to be flawed in some way (Aslan et al., 2018, Chan et al., 2013, Schien and Preist, 2014, Coroama and Höjer, 2016). The complete list of methods and linked flaws can be found in the section below.

4.3.1 The Different Methods and Identified Flaws

The different methods of calculation and data collection identified have commonly been divided into different categories based on their character; wherein the three categories top-down, bottom-up, and model-based often reoccurred (Coroama and Hilty, 2014, Coroama et al., 2015). Four other terms were also found, including the methods modeling, annual electricity consumption, direct measurement, and extrapolation (Aslan et al., 2018).

The Top-Down Method

The top-down approach is defined as measuring the total electricity consumption of the Internet or traffic for a region within a defined time, and then dividing the former with the latter, thus yielding an average number of the energy consumption per data transferred (Chan et al., 2013). Later research suggested specifying the approach by clarifying that it is based on two different estimations: (1) measuring the

(30)

21

total network traffic of the region, and (2) measuring the overall energy demand of the Internet or its regional part (Coroama et al., 2015). Flaws were identified to occur in the form of relatively large overestimation errors (Chan et al., 2013, Aslan et al., 2018), too wide system boundaries, i.e. too many included elements (Coroama, 2021), and large allocation problems when the large number of activities must be sorted (Hilty, 2021). The method has furthermore been deemed limited due to the sparse descriptions of the methodology used in the studies (Aslan et al., 2018). It has been suggested that the method should be used when data cannot be collected at the equipment class level (Chan et al., 2013).

The Bottom-Up Method

The bottom-up method is defined as using direct observations from one or several individual case studies followed by a discussion on the result’s generalizability (Coroama and Hilty, 2014). It has also been described as an approach that directly measures the power consumption and data trafﬁc of equipment within a specific network (Aslan et al., 2018). Flaws were identified to occur in the form of relatively large underestimation errors (Aslan et al., 2018), since researchers often try and model the system ideally, forgetting equipment that should be included but may not seem as important (Coroama, 2021), and there is a constant risk of missing some spare equipment that might affect the results (Schien and Preist, 2014). There has also been a recurring lack of methodological descriptions (Aslan et al., 2018).

In theory, however, the bottom-up method is known to be the most accurate estimation as it collects and utilizes direct measurements from the individual equipment units, and the method is thus recommended as reference method (Chan et al., 2013).

The Model-Based Method

The model-based analysis will model part of the system based on different network design principles.

By combining the model with specific information from manufacturers’ consumption data, the LCA practitioner can estimate an overall energy consumption which in turn can lead to estimates of the network traffic (Coroama and Hilty, 2014, Coroama et al., 2015). The method will result in detailed mathematical equations that can be used for future predictions of electricity intensity as well as to estimate the impact of changes in specific variables. However, the method also requires many assumptions of the characteristics of the network and the data traffic, and it is therefore deemed highly sensitive to the accuracy of said assumptions, as well as to the chosen boundaries (Aslan et al., 2018).

During the interviews, it was argued that due to the complexity of energy consumption in particular networks and to some extent data centers, theoretical models based on one or a few parameters (such as data traffic) would not give an accurate representation of the actual consumption which represent the majority of emissions for such products (Bergmark, 2021). It would also be too complex to build (Malmodin, 2021). It must be noted that Coroama et al. (2015) include the model-based approach into

(31)

22

their definition of the bottom-up methodology, but since no such occurrence could be found in any other literature, it has been described in its own paragraph.

Other Identified Methods

Two other methods were identified, called “annual electricity consumption (AEC) analysis” and

“extrapolation”. The former is defined as using data from usage, power consumption and stock of equipment within a specific network to estimate the total energy used over a specific period of time.

Most often, the annual electricity is measured and then divided with the estimated annual traffic. The AEC-method is preferable over model-based methodology as it requires fewer assumptions and can provide more accurate estimations (Aslan et al., 2018). By using the extrapolation method, LCA practitioners extrapolate already existing values to derive an estimate for a different base year based on changing factors in energy use of equipment or network traffic. It has been deemed a complex method due to the high dependency on the accuracy of the original estimations as well as the assumed factors of change (Aslan et al., 2018), and it has also been described as a method that involves uncertainties of fundamental nature (Coroama and Höjer, 2016).

4.3.2 The Preferable Choice of Method

During the literature review, two data collection methods were found to be more recurring: the top-down method and the bottom-up method (Aslan et al., 2018, Chan et al., 2013, Schien and Preist, 2014, Coroama and Höjer, 2016). Even though flaws were identified in both methods, a consensus was found among the interviewees when asked about their preferred choice of method. The combination of the methods top-down and bottom-up, to validate each other, was deemed most appropriate if aiming for future harmonization (Lundén, 2021, Hilty, 2021, Coroama, 2021). One interviewee furthermore argued that theoretical models should be avoided due to the complexity of building them, and that data should be collected from operators to the largest extent possible (Malmodin, 2021).

4.3.3 Possible Solutions and Further Suggestions

In a research article from 2014, Schien and Priest presented a bottom-up “meta-model” and a reworked top-down model. Their results showed that the discrepancy was substantially reduced, however, the two models were still not found to be comparative (Schien and Preist, 2014). During the interviews, suggestions of further tweaking the two methods, in an attempt to minimize their flaws, were discussed.

One interviewee argued that it would depend on the specific additions made (Coroama, 2021), while another participant noted that it may implicitly happen in the future by itself anyway and that LCA practitioners instead should aim to be consistent in the way the methods are used (Hilty, 2021).

(32)

23

4.4 Electricity Intensity Metrics

There is currently no standardized way of measuring the average electricity intensity, i.e. the measurement of data transmission (Coroama and Hilty, 2014, Aslan et al., 2018). As a result, three different measurement units commonly occur in studies: energy per data, energy per time, and energy per subscriber (Aslan et al., 2018, Coroama and Hilty, 2014, Coroama et al., 2013, Malmodin et al., 2014, Coroama et al., 2015, Malmodin and Lundén, 2018). These findings were further strengthened by two interviewees (Coroama, 2021, Hilty, 2021). The occurrence of the different metrics has resulted in several research papers highlighting the importance of choosing the correct one (Aslan et al., 2018, Malmodin, 2020), as results of LCAs continue to differ widely, weakening the robustness of any conclusions drawn from the results (Schien and Preist, 2014).

4.4.1 Metrics Identified in Previous Literature

During the literature review, it was found that electricity intensity commonly is measured via the unit one kilowatt-hour per gigabyte [kWh/GB] (Aslan et al., 2018, Coroama and Hilty, 2014, Coroama et al., 2013), but the unit J/bit also occur (Koomey et al., 2004, Schien and Preist, 2014). The second method of measuring electricity intensity is by estimating the energy consumed during a period of time rather than per data transmitted (Malmodin et al., 2014, Coroama et al., 2015). Since different network components scale differently with different dimensions, some argue that the subsystem core network is best modeled as energy per data while CPEs and access networks should be modeled as energy per time, since the latter two are largely traffic-independent (Coroama et al., 2015). Lastly, the electricity intensity metric has been discussed in relation to the user, e.g., the environmental impact per human, user, and subscription (Malmodin and Lundén, 2018). However, user behavior is often linked to many uncertainties (Arushanyan et al., 2014a).

4.4.2 Possible Solutions and Further Suggestions

The unit energy per data has recently been deemed an outdated metric which can lead to large overestimations when discussing high-use data cases. It is often based on network averages or future assessments, thus not fairly representing the longer periods of no data or high data usage services which in turn creates a large error in the results (Malmodin, 2020). The different usages of the metrics were also discussed in depth during the interviews. One participant argued that LCA practitioners would be able to get a more adequate picture of the average situation with the metric per time (Hischier, 2021).

Another interviewee noted that if an LCA study was to be compared to other studies, the metric energy per data would be useless as it does not work if compared to a non-IT-related service (Lundén, 2021).

The metric energy per data was furthermore found to be more uncertain, meaning that a small calculation error may result in large differences (Malmodin, 2021). Lastly, it was argued that the metric energy per

(33)

24

data is not suitable when creating a model with the purpose of estimating future development, as both the system composition and independently collected data show that there is no direct correlation between carbon emissions and data growth (Bergmark, 2021).

Results from the interviews also indicated that the choice of metric preferably should depend on the type of subsystem in question (Coroama, 2021, Hilty, 2021, Malmodin, 2021), i.e. if a device were to use the same amount of electricity regardless of its load, it should be measured based on time. If the energy consumption instead were to be proportional to the load, the device should be calculated per data volume (Hilty, 2021). As a result, several interviewees argued that the IP core network should be modeled using the unit energy per data (Hilty, 2021, Coroama, 2021), whilst the access network, customer premise equipment and end-user devices fall under the category of energy per time (Coroama, 2021, Malmodin, 2021). This division has been illustrated in Figure 7.

Figure 7. The preferable choice of metric per equipment.

4.5 Transparency and Data Availability

There is currently low transparency within the research field with a lack of published data and methodological descriptions (Itten et al., 2020, Malmodin et al., 2014). As a result, LCA practitioners risk having to make further assumptions of their own or adopt from past studies without detailed analysis of the underlying assumptions, which in turn might jeopardize the representativeness of future LCAs (Arushanyan et al., 2014a). The information available is furthermore often not very comprehensive (Malmodin et al., 2014), and more transparent and sufficient data linked to the networks and data centers are therefore needed (Itten et al., 2020).

(34)

25

4.5.1 Identified Causes

Several causes of the low transparency within the research field were identified. Two interviewees mentioned the possibility that companies may be hesitant to share data they do not want competitors to get hold of (Bergmark, 2021, Hilty, 2021), and it was also argued that the word limit of research papers does not support a detailed disclosure of studies of complex products and systems (Bergmark, 2021).

More pressing issues within the sector, including IT security issues, cyber-attacks, personal integrity, and espionage issues, were also identified as a possible cause (Hilty, 2021), as well as a lacking scientific validation processes (Coroama, 2021).

4.5.2 Possible Solutions and Further Suggestions

During the literature review, it was found that better stakeholder involvement has been deemed a necessity for future development within the area (Coroama and Hilty, 2014). A more transparent and up-to-date inventory data of the use phase of ICT-related devices is also required, however how this is supposed to be implemented in reality was not disclosed (Itten et al., 2020). During the interviews, one interviewee found the issue of low stakeholder involvement to be a matter of maturity of the market (Lundén, 2021). The participant had notedly observed market developments which indicated that stakeholders are becoming increasingly aware of the benefits of sharing data (Lundén, 2021, Bergmark, 2021) while another had observed an increased request for transparency related to sustainability which could benefit the willingness to share data positively (Bergmark, 2021). However, one interviewee argued that it may not be the best solution to leave the market to sort itself out due to the increasingly pressing matter of climate change, which has resulted in increasing demands for quicker results regarding the ICT sector’s environmental impact (Coroama, 2021).

On the topic of a more transparent and up-to-date inventory data of the usage phase, a discussion was held during interviews regarding the possible implementation of a separate entity within the ICT sector that would act as an independent and trustworthy third party, collecting the required information from relevant stakeholders. The suggestion was however deemed unrealistic by several interviewees (Lundén, 2021, Hilty, 2021), with the argument that its purpose would be too narrow, and that there currently is no market interest in such facility. It was furthermore argued by one interviewee that it should not be a requirement to share that knowledge, but that it instead should be given free of will (Lundén, 2021).

4.6 Age of Data

Increased developments within technology and improved equipment, e.g. in the form of energy efficiency gains in transport equipment, have a significant impact on the results of recent studies of network traffic (Coroama and Hilty, 2014) and researchers have since made a point of clarifying to what exact year the data is referring to in an attempt to be as transparent as possible (Coroama et al., 2015,

On Conducting a Life Cycle Assessment of Network Traffic

On Conducting a Life Cycle Assessment of Network

Traffic

A Qualitative Analysis of Current Challenges and Possible Solutions

TOVA BILLSTEIN

On Conducting a Life Cycle

Assessment of Network Traffic

A Qualitative Analysis of Current Challenges and Possible Solutions

TOVA BILLSTEIN

Supervisor

ANNA BJÖRKLUND

Examiner

ANNA BJÖRKLUND

Supervisor at IVL Swedish Environmental Research Institute TOMAS RYDBERG

Abstract

Sammanfattning

Acknowledgements

List of Abbreviations

Table of Contents

Table of Figures

Table of Tables

1 Introduction

1.1 Aim and Objectives

2 Background

2.1 Network Traffic

2.1.1 The Network Subsystems

2.2 Life Cycle Assessment

2.2.1 ISO Standards

2.2.2 The Four Phases of LCA

2.2.3 Allocation Procedures

2.3 LCA of ICT and Network Traffic: An Overview

3 Research Design and Method

3.1 The Adopted Exploratory Strategy

3.2 Literature Review

3.3 Semi-Structured Interviews

3.3.1 The Interview Procedure

3.3.2 Selection of Interviewees

3.4 Selecting Significant Challenges

3.5 Analysis of Empirical Material

3.6 Quality Assurance

3.6.1 Validity and Reliability

3.6.2 Ethics and GDPR

4 Results

4.1 Overview of Identified Challenges

4.2 System Boundaries

4.2.1 Identified Causes and Previous Findings in Literature

4.2.2 An Overview of Existing System Boundaries

4.2.3 Possible Solutions and Further Suggestions

4.3 Data Collection Methods

4.3.1 The Different Methods and Identified Flaws

4.3.2 The Preferable Choice of Method

4.3.3 Possible Solutions and Further Suggestions

4.4 Electricity Intensity Metrics

4.4.1 Metrics Identified in Previous Literature

4.4.2 Possible Solutions and Further Suggestions

4.5 Transparency and Data Availability

4.5.1 Identified Causes

4.5.2 Possible Solutions and Further Suggestions

4.6 Age of Data