Quantifying universities’ direct and indirect carbon emissions – the case of Delft University of Technology

The purpose of this paper is to present a comprehensive analysis of the carbon footprint of the Delft University of Technology (TU Delft), including direct and indirect emissions from utilities, logistics and purchases, as well as a discussion about the commonly used method. Emissions are presented in three scopes (scope 1 reports direct process emissions, scope 2 reports emissions from purchased energy and scope 3 reports indirect emissions from the value chain) to identify carbon emission hotspots within the university’s operations.

The carbon footprint was calculated using physical and monetary activity data, applying a process and economic input-output analysis.

TU Delft’s total carbon footprint in 2018 is calculated at 106 ktCO₂eq. About 80% are indirect (scope 3) emissions, which is in line with other studies. Emissions from Real estate and construction, Natural gas, Equipment, ICT and Facility services accounted for about 64% of the total footprint, whereas Electricity, Water and waste-related carbon emissions were negligible. These findings highlight the need to reduce universities’ supply chain emissions.

A better understanding of carbon footprint hotspots can facilitate strategies to reduce emissions and finally achieve carbon neutrality. In contrast to other work, it is argued that using economic input-output models to calculate universities’ carbon footprints is a questionable practice, as they can provide only an initial estimation. Therefore, the development of better-suited methods is called for.

Reducing anthropogenic greenhouse gas (GHG) emissions to net-zero is the key strategy to limiting global warming to 1.5°C in the next century (IPCC, 2018; Kennelly et al., 2019; UNFCCC, 2015). To this end, the EU aims to be climate neutral by 2050, meaning emitting net zero GHGs (European Commission, 2019; Government of The Netherlands, 2019). While climate change was long considered an issue governments and international organizations had to tackle, all kinds of organizations are now taking up the responsibility to implement climate actions and policies themselves (UNEP, 2015). Universities, in particular, carry climate responsibility for educating future society, fostering innovation and demonstrating sustainable transitions themselves (Botero et al., 2017; Jain et al., 2017). For example, more than 1,000 universities and colleges worldwide officially committed to the UN’s “Race to Zero” with the goal of net-zero carbon emissions by 2050 (UNEP, 2021). This goal requires universities to be supported by all entities; faculties, corporate offices, administration, staff and students (Button, 2009).

Before engaging in carbon dioxide emission reduction strategies, organizations must assess their current carbon emissions to consider options, impacts and costs (Riddell et al., 2009). Carbon footprinting – assessing the carbon dioxide emissions of an organization and its supply chain – is gaining popularity as tools and standards are being developed to streamline the calculation process. The most popular standard that accounts for both direct and indirect GHG emissions is the GHG Protocol, which divides emissions into three scopes (1–3). Scope 1 accounts for direct emissions, such as combustion and process emissions; scope 2 accounts for those from the purchase of energy; and scope 3 accounts for all indirect upstream and downstream emissions embodied in the value chain (World Business Council for Sustainable Development and World Resources Institute, 2004). Gaining insight into an organization's complete carbon footprint is vital to identify emission sources and thus starting points for impactful reduction strategies.

Research into the carbon footprints of universities has revealed a diverse picture. Many higher education institutions (HEIs) voluntarily publish their carbon footprints (Udas et al., 2018). However, comparing them is difficult because of a lacking standard for HEIs and the variety of calculation methodologies, boundaries, functional units, inventories and published emission factors (Valls-Val and Bovea, 2021; Helmers et al., 2021). Especially scope 3 emissions are often only partially accounted for. Nevertheless, results show that scope 3 emissions, if comprehensively included, are higher than scopes 1 and 2. Therefore, investigating scope 3 emissions of universities is essential, as it unlocks an often unconsidered reduction potential. Hence, a standardized scope 3 approach considering all emission sources is important and called for (Robinson et al., 2015). Robinson et al. (2018) suggest a carbon footprinting standard for HEI, proposing two footprints. One comprehensive scope 1–3 footprint for internal carbon management use and one scope 1–2 carbon footprint for external reporting. However, this impedes the publication of full-scale carbon footprints, which are often stated to be lacking.

Only very few universities present a carbon footprint also accounting for scope 3 emissions from university expenditures, for example, Yale University (Thurston and Eckelman, 2011), UC Berkeley (Doyle, 2012), De Montfort University (Ozawa-Meida et al., 2013), Norwegian University of Technology and Science (Larsen et al., 2013), Technical University of Madrid (School of Forestry Engineering) (Alvarez et al., 2014) and University of Castilla-La Mancha (Gómez et al., 2016). Emissions from expenditures account for a significant share in all studies, emphasizing the importance of including them in the carbon footprint of HEIs. However, here again, comparing those carbon footprints is difficult because of the variety of boundaries, methods used and unpublished emission factors.

This study investigates and quantifies the direct and indirect carbon emissions of the Delft University of Technology (TU Delft) in 2018, including emissions from procurement and related emission factors. The aim is to present the complete carbon footprint of the university to define starting points for reducing emissions, as the university aims to achieve CO₂ neutrality by 2030 (TU Delft, 2018b). Furthermore, the authors reflect critically on current calculation methods based on this study’s analysis.

To that end, a process and extended input-output life cycle analysis (EIOA–LCA) was applied for the consumption-based carbon footprint calculations. Whenever possible, physical activity data were used. This was the case for scopes 1 and 2 emissions and business flights and commuting, for example. When physical activity data were not available, monetary activity data were used. For procurement and catering emissions, data based on economic input-output (EIO) and hybrid multi-region (HMR) methods were applied (Defra, 2014; Vringer et al., 2010).

This study contributes to the literature on carbon footprinting by expanding the scope of analysis to include previously often neglected activities, such as procurement. This expansion has three implications. First, it could facilitate the comparison of the future carbon footprints of organizations. Second, it enables the identification of emission blind spots in organizational processes. Third, it calls again for developing HEI-specific carbon footprint guidelines.

In 2018, TU Delft had 24,703 students and 5,421 full-time equivalent (FTE) employees. The number of students is expected to grow significantly in the years to come (28,000 students expected in 2026 [¹]).

The university campus is connected to the Dutch city of Delft and covers an area of about 161 hectares. It has 73 buildings with a gross internal area of 612,000 m². The university has eight faculties: Aerospace Engineering; Applied Sciences; Architecture and the Built Environment; Civil Engineering and Geosciences; Electrical Engineering, Mathematics and Computer Science; Industrial Design Engineering; Mechanical, Maritime and Materials Engineering; and Technology, Policy and Management.

The technical state of a significant share of buildings is reasonable or moderate, with an aging process that has started locally or is already affecting constructions and installations. This can be linked to the construction years of the university's buildings, many dating to the 1960s and 1970s. The challenge for the coming years is, thus, the need to renovate the campus (Blom and van den Dobbelsteen, 2019).

TU Delft operates its own heating and electricity grids. The combined heat and power plant (CHP) supplies almost all the heat demand on campus, using natural gas-fired reciprocating engines (a small proportion comes from installed gas boilers). The university plans to drill a geothermal source to provide the campus with heat in 2022. Besides the share produced by the CHP, all electricity is bought from renewable sources (wind farms) in The Netherlands. Today, the installed capacity of solar photovoltaic (PV) panels on campus is about 1 MW (TU Delft, 2018a). The university’s main characteristics in numbers are shown in Table 1. The university’s consumption of electricity, natural gas, water, waste generation and travel data (business flights and commuting) is included in Table 2.

The university stated its aim to become a climate-neutral and circular campus by 2030 in its strategic framework for 2018–2024: “Develop and execute a sustainability plan for a CO₂ neutral and circular campus in 2030.” (TU Delft, 2018b, p. 45). The university has recently taken several strategic decisions concerning the sustainability of its operations following this framework. Moreover, in 2019, TU Delft defined its position on Climate Action, which is one of the UN’s Sustainable Development Goals: “TU Delft will harness its innovative powers to support the world-wide transition to non-fossil energy, and adaptation of the living environment to the consequences of global warming.” (TU Delft, 2019b) To do so, the university will use its “intellectual and innovative power for safeguarding the world population against the risks of climate change, by developing technologies and methods …” (TU Delft, 2019b).

The Executive Board took another step by officially supporting the “Climate Letter” in 2019, as did all other Dutch universities (TU Delft, 2019a; VSNU, 2019). In the letter, scientists called on universities to reduce their carbon emissions by adopting and implementing ambitious climate agendas. Goals and measures should include reducing energy consumption, cutting back on flights, promoting sustainable modes of commuting, disinvesting in the fossil fuel industry, supporting environment-friendly food options and reviewing educational offers concerning energy efficiency (Klimaatbrief Universiteiten, 2019).

In 2021, the vision, ambition and action plan called “Sustainable TU Delft” was delivered to the Executive Board, comprising a comprehensive analysis of the current status, a lookout to the future and steps to be taken to reach the sustainability ambitions of the university (van den Dobbelsteen and van Gameren, 2021). The report includes education, research, valorization and funding, community and operations. For climate neutrality, key performance indicators for the campus buildings include reducing the university’s overall energy consumption by 50%, 50% on-campus generation of electricity and nearly 100% self-generation of heat on campus by 2030. Furthermore, ambitious targets for new buildings and renovations address circularity, heat and electricity consumption, electricity generation and carbon emissions in the building chain (Hänsch, 2020; Hänsch et al., 2020).

The emission scopes and sources were calculated according to the GHG Protocol of the World Business Council for Sustainable Development and World Resources Institute (2004). The choice of calculation method was influenced by data availability. When available, primary data in the form of physical activity and process data were used. This was the case for scopes 1 and 2 and for waste, business flights, water and commuting data (scope 3). To calculate procurement and catering emissions, we used a top-down spend-based method that considered the economic value of services and goods purchased by the university. These methods will be further explained in the remainder of this section.

Calculations for all emission sources followed the same pattern. First, activity or consumption data were collected. The data are presented in, for example, kWh used, km traveled, kg generated or euros spent. Second, specific, matching emission factors were derived from the literature to convert the data into GHG emissions. Emission factors indicate the amount of GHG emitted per data unit, for example, per liter of fuel or kWh consumed. They are presented in kilograms of CO₂ equivalents (kgCO₂eq) per unit. Then, the activity or consumption data were multiplied by the relevant emission factors to obtain the total CO₂eq emitted per emission source, which add up to the total carbon footprint (Figure 1).

Emissions can be calculated in two ways. Process analysis maps all physical flows of a particular product throughout its life cycle. This enables the precise calculation of environmental impacts. However, obtaining the necessary data can be challenging and time-consuming, making the method expensive. In contrast to process analysis, economic input-output (EIO) models describe an economy by mapping trades between economic sectors. All deliveries between producer, trader and consumer are shown in a matrix. These matrices facilitate quickly calculating a product’s or service’s environmental impacts along the whole supply chain in one specific sector. EIO tables, generally at the country level, allow for a fast overview; however, they are subject to a high level of aggregation (Kennelly et al., 2019; Thurston and Eckelman, 2011; Vringer et al., 2010). Hybrid models have been developed to combine the advantages of both models while avoiding their disadvantages. In those models, a process analysis is used for the primary process of a product’s life cycle; for secondary processes, an input-output analysis is used (Vringer et al., 2010).

Primary data from various university departments for the year 2018 were collected: Electricity, natural gas and water consumption data were provided by the Campus and Real Estate Department, flight data by the Human Resources Department, waste data by the Facility Management Department and commuting data by the Education and Student Affairs and the Human Resources Departments. All are specific activity or process data derived from bills, meter readings, registrations or purchase lists. For procurement and on-campus expenditures on food, financial data were obtained from the Finance Department and the university’s caterer. In this case, emissions are expressed per economic value spent, thus kgCO₂/€. Emission factors were derived from literature based on EIO models (Defra, 2014) and a HMR model (Vringer et al., 2010).

According to the GHG protocol, all university-relevant emission sources were included in the carbon footprint calculation process to obtain a comprehensive overview of the carbon emissions. In general, no scopes or emission sources were excluded. However, relevant emission sources for the university (e.g. canteens and restaurants on campus) were added to scope 3, whereas irrelevant ones (e.g. sold products, their use and end-of-life treatment) were disregarded. Figure 1 shows an overview of the calculation process and the emission sources considered in this study.

Table 2 explains the origin of used input data and assumptions around them per emission source. Emission factors are described, and physical activity data from TU Delft for 2018 are shown. A description of the monetary-based input data and the process of adapting and matching emission factors to procurement-based emission categories is explained in more detail later.

The Finance Department provided monetary-based procurement data for 2018, comprising all goods and services procured by the university (ca. 1,400 entry points). The spend data were presented in three layers. Category level 1 was divided into eight aggregated categories (i.e. person-related matter, office and operational means, transportation and buildings and building-related installations and services). Category level 2 provided more specific accounts. Person-related matters, for example, contained ten sub-categories on the second level. Examples are: Study, coaching, training and education; Business trips, external accommodation, catering; and Recruitment, selection and outplacement. The most detailed level was Level 3, “Description.” The datasheet comprised 128 description titles at this level.

Emission factors were obtained from Vringer et al. (2010) and Defra (2014). Emission factors from Vringer et al. are based on a hybrid method model for households in The Netherlands, whereas Defra used an input-output model for the UK. As both sources use historic (and different) base years, the emission factors were adjusted with a correction factor based on the GHG/GDP ratio for the European Union (EU 28) (European Environment Agency, 2020; Eurostat, 2022). This ratio was chosen to account for the decrease in the carbon emissions of products and services over time and inflation. Trading balances of the European Union show that most products and services were traded within the Union (Eurostat, 2021). The calculated GHG/GDP ratio resulted in static correction factors for the year 2018: 0.57 for the emission factors from Vringer et al. (2010) and 0.81 for emission factors from Defra (2014).

The most detailed level (Level 3, Description) was considered to match specific emission factors to the spending (for the assigned emission factors see the Appendix). Matching was done in four ways (Figure 2):

1. If there was a direct match between the description item and an emission factor, then that emission factor was used.

2. If the description item matched different emission factors, then the average of those was used.

3. If no matching emission factors were available for an item, then the average of an emission factor group was used, for example, an average emission factor of all hardware emission factors or service-related emission factors.

4. If none of the above-mentioned ways was possible, then the average of all used emission factors was assigned to the remaining items.

The description items were recategorized from bookkeeping categories to the carbon footprint emission sources explained in Table 3 – reducing the number from 128 to 10. Several description items were disregarded. Cost accounting items purely for accounting and bookkeeping purposes were excluded, as no action and, thus, no additional carbon emissions result from them. This was the case for depreciation items, received advance payments and scholarships, for example. Items calculated separately based on physical activity data (electricity, natural gas, flights and water) were also deducted. Moreover, items considered the same for TU Delft and a third party (e.g. cooperation and collaboration with universities and guest lecturers) were disregarded to avoid the double-counting of emissions. Thus, it was assumed that TU Delft receives as many guest teachers and lecturers as it sends. Emissions are, therefore, already included in scopes 1 and 2 footprints.

Recategorization was done in three ways (Figure 2). In general, if one of the merged items within a description (originating from various category levels 1 and 2) contributed more than half of the financial sum of that description’s total, then the totality was assigned on that basis to one of the emission sources (Table 3):

1. The description items could be directly matched with a specific carbon footprint emission source.

2. Description items were traced back to their original category 2 level to assign them to the carbon footprint emission source.

3. When there was no single significant contributor and too many category 2 level relations, items were assigned individually to a carbon footprint emission source.

Catering spend data from on-campus canteens and restaurants were obtained from the catering company and internally, comprising a list of sold food and beverage items. Emission factors are based on Vringer et al. (2010), corrected as described above. Meal ingredients were approximated to match the emission factors, as received data were based on meals sold, not ingredients.

The calculated consumption-based carbon footprint of TU Delft in 2018 is 106,000 tCO₂eq. Divided into the scopes of the GHG protocol, scopes 1 and 2 together account for 17% of emissions, while scope 3 accounts for 83% (see also the Appendix for a comprehensive table with the detailed calculations of all emission sources). This distribution is similar to results from other organizations and universities that included procured goods and services in their carbon footprint calculations, which again emphasizes the importance of including scope 3 in an organization’s carbon emission reduction strategy and implementing practical reduction measures within that scope.

Figure 3 shows the breakdown of TU Delft’s carbon footprint by scope and by emission source. Scope 3 emissions were divided into emissions influenced mainly by the university’s operation (Real estate and construction; Equipment; ICT; Facility services; Research expenses and consumables; Administration, Consultancy and auditing; Transportation and travel; Energy supply to third parties on campus; Other; Paper products; Finance and tax; Water; and Waste) and those mainly influenced by its staff and students (Business flights; Catering; and Commuting). The vast majority of the total carbon emissions are scope 3 operation related (69%), while only a small part is related to staff and students (14%).

Real estate and construction is the most significant emitter (18%), followed by Natural gas (17%), Equipment (13%), ICT (8%), Facility services (8%), Business flights (5%), Catering (5%) and Research expenses and consumables (5%). The “big five” emission sources are responsible for 64% of total carbon emissions. The eight emission sources contributing 5% or more account for almost 80% of the total footprint. The remaining ten account for only 21%. This highlights the need to address the most significant emission sources specifically. At the same time, the authors see the potential to significantly reduce the carbon footprint by focusing reduction strategies on the limited number of major emitters.

Some emission sources showing specificities concerning input data, their content, reduction plans or potentials are discussed in more detail in this section, as a framework for HEIs is missing.

Real estate and construction, the most significant emission source (18%), includes many service costs with relatively low emission factors, such as guarding buildings, rent and leasehold and daily maintenance. Although no major construction was carried out in 2018, the total emissions from the bookkeeping item “projects” account for almost 90% of the total Real estate and construction emissions. Attempts to investigate what kind of projects this entails were challenging. So far, the authors have been unable to discover the specific content as would be desirable.

Regarding Natural gas emissions (17%), TU Delft has decided to invest in a sustainable heat source, an on-campus geothermal well (TU Delft, 2022). Consequently, natural gas emissions will drop. However, with the geothermal energy, formation gas will be extracted from the earth, which will count toward the carbon footprint. To provide a CO₂ neutral campus, the university must develop plans to deal with this issue.

Equipment is the third most important carbon emitter (13%). It includes emissions from purchasing, maintaining and renting equipment and technical items. About 75% of the calculated emissions originate from the bookkeeping category “equipment.” As with “projects” in Real estate and construction, the exact content of the description item “equipment” is not always entirely transparent.

Business flights were responsible for 5% of the university’s emissions. In all, 70% of flights were long-distance (> 2,500 km). Short-distance flights (< 700 km) contributed only 10% of emissions. This means that a strict university regulation to justify the need for a flight will be a more effective reduction tool than the prohibition of business flights within a range of 700 km, for example. Schmidt (2022) discusses university’s air travel policies in detail.

Commuting by employees and students was another relatively small emission source (4%). The Dutch are known for being a biking nation, which benefits the commuting footprint. Thus, the most reduction potential is seen in the 32% of employees who currently come to the campus by car. TU Delft has set a 10% reduction target for car commuting by 2025, compared to the base year of 2018 (van de Klugt et al., 2018).

Electricity accounts for only 1% of TU Delft’s carbon emissions, including life cycle emissions from installed PV and purchased wind energy; thus, emissions from different scopes are combined to show the complete picture. If the university had bought its electricity from the grid, then it would have resulted in 34,139 tCO₂eq, almost double the biggest emission source. As the input for the CHP to generate electricity is natural gas, originating emissions are accounted for in Natural gas.

Surprisingly, although the authors expected the university to be a “paper organization” with a considerable amount of paper being bought and many books being produced and printed, emissions from Paper products play a negligible role in the overall footprint (1%).

Finance and tax (1%), Waste (0%) and Water (0%) are the emission sources with the least impact on the total footprint. However, this does not suggest that measures to reduce waste or increase waste sorting have no impact. Waste recycling can play a vital role in achieving carbon neutrality by closing material loops and avoiding embodied emissions. Additionally, waste should be investigated in relation to procured goods.

Knowing TU Delft’s carbon hotspots enables the university to develop reduction strategies that will have the biggest possible impact on the total footprint. However, the results are still at a high level of abstraction and subject to uncertainty.

The uncertainty of results is substantial for some emission sources – especially in the case of emissions calculated on a spend basis, which account for the most significant part of the footprint (70%). Consequently, variations in those calculations will have a significant impact.

The uncertainty of the input data and that of the used emission factors was considered to assess the results’ uncertainty level, according to the IPCC and GHG Protocol guidelines (IPCC, 2000). Uncertainties were estimated by emission source, and the IPCC error propagation equation was used to evaluate their impact on the results, as described in the following paragraphs.

Combined uncertainty levels were estimated to be high for emission sources calculated on a monetary basis, for Business flights and Commuting of staff and students (± 30% for most of them). For all emission sources calculated on a monetary basis, activity data uncertainty was considered 10% because of the recategorizations of bookkeeping categories and non-transparency of specific contents. Moreover, in 2018, the financial department's accounting system was renewed, resulting in some inconsistencies in bookkeeping categories. An activity data uncertainty of 30% was considered for Catering, Business flights and Commuting. Emission factor uncertainty was estimated to be 30% because of the correction of emission factors and their combination from different sources, often based on households. For Business flights, emission factor uncertainty was estimated at 20% because of detours, non-European departure locations, emissions in great heights and flight lengths. For Commuting, 10% were estimated.

Waste (± 14%), Electricity (± 10%) and Energy supply to third parties on campus (± 10%) are estimated to have moderate to low combined uncertainty levels from activity data and emission factor uncertainty. Natural gas and water are considered to have very low combined uncertainty levels (both ± 1%).

The authors estimate the combined uncertainty levels of this study to be moderate. Repetition of the calculation with precisely the same input data would lead to another calculated amount of carbon emissions because of different data allocation and (sub-)categorization; however, the deviation is estimated to be about 10%. Moreover, a significant shift in the order of contributing emission sources would not be expected. A previous study estimating TU Delft’s carbon emissions from procurement in 2015 came to about the same results (Mauro, 2017). Additionally, the result is in line with the calculations of other universities. Despite the uncertainty, the result is, thus, considered robust.

Nevertheless, the authors see a need for better-investigated input data and more specific emission factors, especially for procurement. TU Delft has started submetering buildings to investigate electricity consumption patterns inside buildings and a project to better register suppliers and their environmental emissions. In addition, a framework defining boundaries for HEIs’ scope 3 calculations (including the scope of the emission source itself) is needed to facilitate comparisons and benchmarking of carbon footprints in the sector.

The most common comparison ratios relate the carbon footprint to the number of students and staff, the gross internal area of campus buildings and the spending (Helmers et al., 2021; Valls-Val and Bovea, 2021). Compared to previous studies of universities, which included procurement emissions in their calculations, TU Delft’s emission ratios generally align. However, there are some exceptions, as described beneath. Table 4 compares the carbon emissions of the mentioned studies per gross internal area, per person (staff and students) and euro spent.

TU Delft’s footprint is 0.17 tCO₂eq/m², 19.54 tCO₂eq/FTE, 4.29 tCO₂eq/student and, thus, 3.52 tCO₂eq/capita and 0.44 kgCO₂eq/€ spent. Those numbers particularly align with the case of the Norwegian University of Technology and Science (Larsen et al., 2013). Previous studies have shown that social science faculties have a smaller footprint than their technical counterparts (Kulkarni, 2019; Larsen et al., 2013). Furthermore, Klein-Banai and Theis (2013) showed that laboratory spaces of research-intensive institutions affect the carbon footprint manifold more than offices, lecture halls and classrooms. This might explain the emission rates of both universities. However, Helmers et al.'s (2021) comparisons do not confirm this. Noteworthy is furthermore the high result per euro spent by Alvarez et al. (2014), for which they reason in their study.

In Helmers et al.’s (2021) rankings, which did not include procurement emissions, TU Delft would be situated in the top ten of the least emitting universities in all three ratios. Procurement emissions from the TU Delft’s carbon footprint were excluded for this comparison. Ranked by emission per capita, with 1.1 tCO₂eq/capita, TU Delft would come in the eighth or ninth best place [²] (meaning least emitting) from then 23 HEIs. However, it would come in the second-best place with 52 kgCO₂eq/m². Likewise, it would come in the second-best place relating emissions to university expenditure (without salaries and purchasing power corrected), namely, 90 kgCO₂eq/1,000$. The good rankings might be explained by the fact that TU Delft exclusively buys green electricity (to which life cycle emissions were assigned), which reduces the carbon footprint significantly compared to other universities.

Calculating scope 3 emissions calls for the making of qualified boundary choices. It was chosen to integrate all emission sources to obtain a complete picture of the footprint, knowing that some uncertainty levels were elevated. Comprehensiveness versus accuracy is a debatable issue. Another point is boundary setting, that is, what to include in scope 3 emission sources without adding the emissions of whole supply chains and personal choices of employees and students to the university's account.

For example, many people working and studying at TU Delft come from abroad. Whereas business trips made on behalf of the university were included, trips to the home countries of staff and students were not. They were considered to be accounted for in personal carbon footprints. However, commuting was included in the university’s carbon footprint, so where people lived did impact the footprint. Another example is calculated catering emissions. Food and beverages sold on campus were considered. It is debatable whether food brought from home should also be included in the footprint, as people must eat to work. As these boundaries impact the results, the examples show that it is not enough only to define which emission sources to include or exclude. It is essential to provide guidelines in a HEI framework defining where to draw boundaries within those emission sources to assure comparability, also stated by Ozawa-Meida et al. (2013) and Valls-Val and Bovea (2021).

Regarding the calculation method, estimating the footprint based on spending might result in wrong conclusions for several reasons. Sustainable suppliers, for example, might charge more. Choosing such a supplier will result in higher calculated emissions when in reality, emissions might be reduced (Larsen et al., 2013). Also, economy-of-scale-effects, which might be substantial for a university, are not included (Larsen et al., 2013; Alvarez et al., 2014). In addition, emission reductions occurring over life cycles will not appear in future spend-based carbon footprints. This is especially the case for the construction and renovation of buildings. Next, large investments in a specific year affect and distort the carbon footprint of that year, as they are not spread over the lifetime. Thus, vast expenditures (like renovations or the purchase of large laboratory equipment, for example) will significantly increase calculated carbon emissions when in reality they might reduce scopes 1 and 2 emissions in the future (Ozawa-Meida et al., 2013). However, allocating historic emissions over the years may not solve this problem, as it distorts the momentary picture and prevents perspectives for immediate actions. Therefore, future research is called to investigate and develop a method to deal with extensive investments to level out underestimating and overestimating carbon emissions.

Likewise, spend-based emission factors could result in overestimating or underestimating carbon emissions. For example, Vringer et al. (2010) based their emission factors on Dutch households. Using them for an institution like a university might distort the results because of a scale-up that the authors neither intended nor included in their calculations. Nevertheless, they are the most detailed and specific to the Dutch system and culture at the moment.

Concise calculation of procurement-related emission sources involves specifying and investigating each one in depth. This makes the calculation process time-consuming. Moreover, various people's commitment in different departments is needed to thoroughly analyze and interpret the financial data, its layers and categories. The authors reached a point where they could not analyze the specific content of financial categories any further. Container terms like “projects,” “equipment” and “technical items” did not convey what was included and led, even after consultations, to investigative dead ends. As other case studies also stated the necessity to interpret spending categories and the need for a more detailed uniform category breakdown (Ozawa-Meida et al., 2013; Alvarez et al., 2014), the general suitability of the calculation method used is questioned by the authors.

Calculating on a spending base depends on the accounting system's consistency in the long term. A change of systems or categorization will also affect the footprint calculations. Therefore, accounting systems should not be the base for monitoring carbon footprints over time. Ideally, procured goods' physical activity data should be available, that is, material data stored in a material database. This aligns with the aim of a circular campus for which the university needs to know about its material stocks, inflows and outflows. Consequently, the carbon footprint could be calculated on a material base instead of a spending base, leading to more precision.

Another risk accompanying the chosen approach is the double-counting of avoided CO₂eq. First, avoided carbon emissions are included in the emission factor of waste streams. Second, avoided CO₂eq might be included in emission factors for products with a recycled material content. This would result in double-counting of the same avoided emissions. Therefore, organizations need to consider where avoided emissions are included to prevent whitewashing in upstream or downstream scope 3 calculations.

All other studies, which included procurement emissions, call for adjustments in the calculation methods. These include: Hybridization also for scope 3 emission sources (thus using a process approach); the development of a set of indicators for the most significant contributors calculated on an EIO basis (Larsen et al., 2013); a common reporting framework for HEI with defined organizational boundaries and a uniform breakdown of procurement categories, considering product carbon footprints of goods and services and LCAs of waste streams and recycled materials; monitoring embodied emissions and refurbishments (Ozawa-Meida et al., 2013); and the consideration of the geographic location, more recent IO data and economies of scale (Alvarez et al., 2014). Nevertheless, all consider their approach practical and applicable for other HEIs, which the authors of this study question for the reasons mentioned above.

The calculated direct and indirect carbon emissions of TU Delft were 106 ktCO₂eq in 2018. Of the total footprint, 83% were scope 3 emissions, highlighting the need to consider organizations’ upstream and downstream activities to achieve carbon neutrality. This 20/80 distribution across the three scopes was also seen in other cases that included emissions from procurement. The five most significant emission sources (Real estate and construction, Natural gas, Equipment, ICT and Facility services) were responsible for 64% of the total carbon footprint. Efficient carbon emission-reducing strategies can, therefore, focus on these hotspots.

The authors see several limitations in this study. First, as in other studies, activity data lacked accuracy or had a high aggregation level. Second, the latter was also true for the emission factors from EIO and hybrid method models. Therefore, they cannot account for product differences, production processes and recycled material content. The elevated uncertainty levels of some emission sources and the limitations of the calculation process imply several avenues for future research. The authors call to discuss and develop calculation methods that improve results’ accuracy and precision. Those methods should clarify emission source boundaries and consider life cycle carbon emissions and reductions.

This study adds value by reviving the discussion about better-suited calculation approaches, including issues related to spend-based calculation methods; for example, the difference between calculated and actual emissions for (eventually) pricier sustainable products or the increase of the footprint because of substantial investments, which however might lead to emission reductions in the long term.

Real progress regarding these issues only seems possible when suppliers make their product’s carbon footprint or material data available. Hence, calculating scope 3 emissions on a material or physical activity data basis would be possible, enabling more precise indirect carbon footprint calculations. Universities can then take up their role model function by including scope 3 emissions in their climate neutrality goals and lead the way in their realization to mitigate climate change.

The authors thank TU Delft offices and faculty staff and the university service providers for waste and catering, for their help in providing data and assisting with its analysis, calculation and interpretation. This work would have been impossible without them. The authors also thank the two anonymous reviewers and the deputy editor for their constructive feedback.

Funding: This research did not receive any specific grant from public, commercial or not-for-profit funding agencies.

EF Reference

a Milieu Centraal; Stichting Stimular; Connekt; SKAO; Ministerie van Economische Zaken en Klimaat. (2018). “CO2emissiefactoren 2018”, available at: Www.co2emissiefactoren.nl/wp-content/uploads/2019/01/co2emissiefactoren-2018.pdf

b Adjusted from Vringer, K., Benders, R., Wilting, H., Brink, C., Drissen, E., Nijdam, D., and Hoogervorst, N. (2010), “A hybrid multi-region method (HMR) for assessing the environmental impact of private consumption”, Ecological Economics, Vo. 69 No. 12, pp. 2510-2516

c Adjusted from Defra. (2014), “Table 13 - Indirect emissions from the supply chain. UK Department for Environment, Food and Rural Affairs”, available at: Www.gov.uk/government/statistics/uks-carbon-footprint

d Adjusted from Milieu Centraal; Stichting Stimular; Connekt; SKAO; Ministerie van Economische Zaken en Klimaat (2018). “CO2emissiefactoren 2018”, available at: Www.co2emissiefactoren.nl/wp-content/uploads/2019/01/co2emissiefactoren-2018.pdf

e Stichting Stimular. (2018). “Actuele CO2-parameters - 2018: Milieubarometer”, available at: Www.milieubarometer.nl/CO2-footprints/co2-footprint/actuele-co2-parameters-2018/

f Renewi, TU Delft's waste processor

g Turner, D. A., Williams, I. D., and Kemp, S. (2015). “Greenhouse gas emission factors for recycling of source-segregated waste materials”, Resources, Conservation and Recycling, Vol. 105, pp. 186-197.

h Pulselli, R.M., Marchi, M., Neri, E., Marchettini, N. and Bastianoni, S. (2019). “Carbon accounting framework for decarbonisation of European city neighbourhoods”, Journal of Cleaner Production, Vol. 208, pp. 850-868.