Geographic versus institutional drivers of nitrogen footprints: a comparison of two urban universities

Excess reactive nitrogen (N) is linked to a myriad of environmental problems that carry large social costs. Nitrogen footprint tools can help institutions understand how their direct and indirect activities are associated with N release to the environment through energy use, food, and transportation. However, little is known about how geographic context shapes the environmental footprints of institutions. Defining the system boundaries over which institutions are responsible and able to control individual drivers of N footprints is also a challenge. Here, we compare and contrast the circa 2017 N footprints for two research intensive universities located in Montréal, Canada, with a combined full-time equivalent campus population of ∼83 000. Our estimate of McGill University’s N footprint (121.2 t N yr−1) is 48% greater than Université de Montréal’s (74.1 t N yr−1), which is also reflected on a per capita basis (3.3 and 1.6 kg N capita−1 yr−1, respectively). Key institutional factors that explain the differences include McGill’s larger residential and international student populations, research farm, and characteristics of its on-campus fuel use. We use a series of counterfactual scenarios to test how shared urban geographic context factors lead to an effective reduction of the N footprints at both universities: the relatively small direct role of both institutions in food intake on campus (29%–68% reduction compared to a counterfactual scenario), energy from hydroelectricity (17%–21% reduction), and minimal car commuting by students (2%–3% reduction). In contrast, the near-zero N removal from the municipal wastewater system effectively increases the N footprints (11%–13% increase compared to a modest N removal and offset scenario). Our findings suggest that a shared geographic context of a dense city with plentiful off-campus housing, food options, and access to hydroelectricity shapes the absolute N footprints of Montréal’s two main universities more than the divergent institutional characteristics that influence their relative N footprints.


Introduction
Human activity has dramatically accelerated the nitrogen (N) cycle primarily via the conversion of inert N 2 gas to reactive forms of N through synthetic fertilizer production, growth of N-fixing crops, and fossil fuel combustion (Gruber and Galloway 2008). While increased use of reactive N in agriculture has been critical to increasing food production globally, it has contributed to a doubling of reactive N inputs to the environment toward levels that could have major consequences for global sustainability (Compton et al 2011, Fowler et al 2013, Steffen et al 2015. Reactive N can have many environmental and human health impacts depending on its form and location. For example, reactive N can contribute to eutrophication of coastal and other surface waters, groundwater contamination, the formation of smog and other atmospheric pollutants that negatively impact air quality (primarily as nitrogen oxides, NO x ), as well as acting as a potent greenhouse gas, nitrous oxide (N 2 O), that contributes to global climate change (Galloway et al 2003). Therefore, efforts to sustainably manage reactive N use and to mitigate N release to the environment have implications for meeting sustainability objectives ranging from the local scale (e.g. municipal or institutional net-zero emissions targets) to achieving the United Nations' sustainable development goals (SDGs) (Vanham et al 2019).
Nitrogen footprints estimate the release of reactive N to the environment as a result of individual or collective consumption activities . For example, the consumer-oriented N-Calculator model by Leach et al (2012) estimates the reactive N released to the environment on a per capita basis related to food production and consumption, household utilities, transportation, and the purchase of goods and services. National per capita N footprints have been calculated for more than 10 countries and vary between 15 and 47 kg N capita −1 yr −1 (Pierer et al 2014, Shibata et al 2014, Stevens et al 2014, Liang et al 2016, Shibata et al 2017, Guo et al 2017, Elrys et al 2019. Typically, these studies have emphasized the role of food production (e.g. excess fertilizer N use in agriculture and other N losses during food processing) and to a lesser extent, food consumption (N excretion by humans and wastewater treatment), as key aspects of national N footprints . Given the large variation in the N losses associated with producing different foods (e.g. animal> vegetal products), N footprints are strongly influenced by diet (Leach et al 2012. Nitrogen footprints can play an important role in raising consumer awareness about the impacts of reactive N and in understanding how our activities affect the N cycle , Shibata et al 2017, Einarsson and Cederberg 2019. However, defining system boundaries and the scope of activities to include in footprint assessments is a challenge. Past N footprint studies have had variable system boundaries that may lead to different conclusions about potential impacts of reactive N on the environment (Einarsson and Cederberg 2019). Such impacts include economic damage costs to society through the effects of reactive N on the environment and human health, which have been calculated at institutional (Compton et al 2017), state (Keeler et al 2016), and national scales (Sobota et al 2015).
As centers of knowledge creation and learning, higher-education institutions have an important role in catalyzing local sustainability efforts and by empowering sustainability solutions in communities (e.g., Colding andBarthel 2017, Withycombe Keeler et al 2018). As a complement to consumer-focused and national scale N footprint assessments, Leach et al (2013) adapted the N-Calculator model to the institutional scale (the 'Nitrogen Footprint Tool', NFT) and calculated the University of Virginia's N footprint. Many colleges and universities have subsequently applied the NFT approach to calculate their institutional N footprints. A comparison of the N footprints of seven US colleges and universities highlighted a wide range in institutional N footprints (from 7 to 27 kg N capita −1 yr −1 ), and showed that the magnitude of animal products in food services and the reliance on fossil fuels in purchased electricity are key factors affecting the absolute and relative magnitudes of the N footprints . Higher-education institutions therefore also make an impact directly through their operations and their efforts to address sustainability reporting around targets that might influence N footprints, such as the Association for the Advancement of Sustainability in Higher Education (AASHE) Sustainability Tracking, Assessment, & Rating System (STARS) (Fonseca et al 2011, Amaral et al 2015. While N footprint magnitudes vary considerably across higher-education institutions , less is known about the specific factors that shape this variation across different institutional and geographic contexts. For example, does a shared urban context outweigh the effects of university-specific differences on the overall N footprints for two schools of comparable size?We aim to compare and contrast the N footprints of two large, research intensive universities in Montréal, Québec, Canada, that have a shared urban setting but considerable variations in operations, governance, and student demographics. Greater Montréal has a high density of post-secondary schools, with 11 universities and a total of 155 000 students in 2015 (Board of Trade of Metropolitan Montréal, 2016). McGill University and Université de Montréal (UdeM) represent two of the four large universities in central Montréal and are located at the heart of the dense metropolitan region. Our specific objectives are to: (1) Calculate and compare the N footprint for these two universities; (2) Identify the institutional factors that affect the N footprints; and (3) Assess how shared geographic factors relating to the urban and provincial settings of these universities may either reduce or increase the N footprints based on counterfactual scenarios representing other geographic contexts. We further assess how operational boundaries affect estimates of N release to the environment associated with an institution, emphasizing food production linked to university-run food services and different scenarios of wastewater production and treatment. This research can help to inform the role of higher-education institutions in mitigating N release to the environment (i.e. academic researchers as well as university sustainability offices) but may also be relevant to other types of organizations seeking to incorporate new sustainability metrics to assess the environmental impacts of their operations.

Methods
We define 'geographic context' as the urban and jurisdictional setting of a university that shapes shared factors such as off-campus housing availability, public transportation, and electricity sourcing. Universityspecific factors related to the campus operations and activities as well as student demographics are defined as 'institutional context'.
2.1. Study area and geographic context of Montréal Montréal (45°30′ N; −73°33′ W) is Canada's second largest city with ∼1.9 million and ∼4 million people in the urban and metropolitan areas, respectively, and with a high population density of almost 4,000 people km −2 on the Island of Montréal (Statistics Canada 2017). It has a comprehensive public transit system and more than 875 km of bike lanes (Vélo Québec 2016). Electricity is provided by the state-owned utility, Hydro-Québec, and renewable sources (typically >95% hydroelectricity) comprised 209 million MWh of the 212 million MWh produced in Québec in 2017 (Statistics Canada 2019). Montréal has a cool continental climate (average July/ January temperatures of 21.2°C and −9.7°C, respectively; Government of Canada 2019a) and about 54% of household energy consumption is used for heating or air conditioning (Hydro-Québec 2019).
McGill and the UdeM are the oldest of the four large universities in Montréal and are similar in size and budget when excluding affiliated institutes (table S1 is available online at stacks.iop.org/ERL/15/045008/mmedia). They are located centrally in the city, 2.5 km away from each other on either side of Mount Royal, a large hill located north of Montréal's central business district (figure 1). They are easily accessible by public transportation, including several bus lines, Montréal's subway (metro), and commuter rail; major bike paths also intersect both of the main campuses. Both universities are research-oriented institutions that comprise a medical school and both have a large international student population, although McGill's international student population is proportionally higher (27% of the full-time equivalent (FTE) student population at McGill versus 13% for UdeM).

Overview of SIMAP and system boundaries for institutional footprints
Most of our N footprint calculations were conducted with the Sustainability Indicator Management and Analysis Platform (SIMAP, https://unhsimap.org), an online tool developed by the University of New Hampshire's Sustainability Institute. The focal year of our calculation is 2017, but some data sources were obtained from earlier years. The institutional N footprint calculation is based on the Nitrogen Footprint Tool (NFT) methodology described by Leach et al (2013). SIMAP considers all N released to the environment in dissolved, solid, or gaseous forms resulting from university activities: fuel used by university-operated vehicles or boilers, the use of fertilizers on campus, and farm/research animal-related emissions. Utilities consumption by the university, the emissions generated by commuting as well as by academic and student mobility, and the N from the production and consumption of food sold on campus are also included (figure 2). Any N losses from off-campus housing and non-university administered food purchases, such as when students and staff bring lunches to the university or purchase food at independent or sub-contracted food vendors, are excluded here.
Different components of the institutional N footprint are divided into 'scopes' based on the operational boundaries of the institution (GHG Protocol 2004). This allows categorization of the potential N pollution contribution by level of institutional control over emission sources (figure 2). A critical aspect of N footprint calculation is therefore how institutional boundaries are set. We used a control approach for both universities (GHG Protocol 2004), where we included emissions from operations over which the university has some level of control, whether owned or not by the university (e.g. it may include leased facilities). Because the structures of the universities differ, including in terms of official data availability, campus boundaries for each university have some particularities. We excluded affiliated entities that are independently administered to ensure that the main operational teaching and research aspects of the universities were of comparable size. For example, UdeM has large affiliated schools (engineering and business schools) and both universities are affiliated with large teaching hospitals and health care centers, which are excluded from our analyses.
2.3. Details on N footprint calculations by component 2.3.1. On-campus fuel use Detailed inventories of stationary fuels (e.g. propane, natural gas, and diesel) as well as transport fuels for university vehicle fleets were provided by McGill's Greenhouse Gas Inventory report (Rivers and Conraud 2018). For UdeM, stationary fuels data were obtained from Québec's Ministry of Education and transport fuels data were provided by the University.

Fertilizers, research animals, and livestock
Neither university used nitrogenous fertilizers on grounds for landscaping in 2017. The quantity of fertilizers used on McGill's farm was directly available, with an average N content of 33%. Research and farm animals standing stocks were available for both universities. McGill has a research farm that contains poultry, swine, and dairy production, and also supplies campus food services with a considerable amount of food. We estimated the direct waste-related N footprint for animals by using the NFT methodology (Leach et al 2013). Our estimates of N losses for animals includes waste and incineration but omits the N use embodied in the production of animal feeds, which was unavailable for either university and to avoid double-counting of N with food production at McGill.
2.3.3. Purchased electricity and utilities N 2 O and NO x emissions associated with purchased electricity and chilled water were calculated in SIMAP using emission factors (EF) specific to Québec, 0.0001 g N 2 O kWh −1 and 0.025 g NO x kWh −1 (Government of Canada 2019b and Statistics Canada 2019) based on university reporting of purchased electricity. This is likely a conservative estimate that integrates the province's entire fuel mix, as the reported NO x EF for hydroelectricity is roughly onethird of the province-wide estimate (Hydro-Québec 2018). Default EFs in SIMAP were applied for purchased steam. Transmission and distribution (T&D) losses were ∼0.1 t N yr −1 at each university based on SIMAP defaults for this region.

Commuting emissions
Local travel-related emissions for all students, staff, and faculty were estimated based on student and staff mode shares derived from survey data at McGill (Shaw et al 2013) and UdeM (Université de Montréal, HEC Montréal and Polytechnique Montréal 2013). Commuting weeks per year and trips per week (10 one-way trips) were held constant between the two universities (faculty: 47 wks −1 yr −1 ; staff 46 wks −1 yr −1 ; full-time students 35 wks −1 yr −1 ). We applied an EF for public buses (0.0055 g N 2 O passenger km −1 ) used by Montréal's public transit authority (Environment and Climate Change Canada 2019). The metro (light rail subway) in Montréal is powered by hydroelectricity and assumed to have negligible NO x /N 2 O emissions (see supplementary information, SI, text for rationale).

Directly financed travel-related emissions
Statistics on directly financed air travel, and ground travel for sports teams, were available from McGill's Greenhouse Gas Inventory report (Rivers and Conraud 2018). However, no official data were available for other directly financed ground transportation for McGill, which is omitted here. Data for long-distance academic travel were unavailable for UdeM, so the annual passenger kilometers travelled by students, staff, and faculty was estimated based on an online

Student non-local travel-related emissions
Official statistics on international students by home country and out-of-province Canadian students by home province were used to estimate total passenger kilometers travelled by mode (air or ground transport) to and from each university following the approach of Arsenault et al (2019). Students are assumed to take one round-trip per year from capital cities of their home countries or provinces (or the national population centroid for US students). We applied a similar approach to estimate study-abroad travel and field trips, based on counts of students by destination country in 2017-2018. Information on home locations of Québec students was unavailable and is therefore omitted.

Food production
We estimated N losses embodied in production of the food that was served at university-run or universityaffiliated eateries by using detailed inventories of food purchases and default virtual N factors (VNFs) in SIMAP. Virtual N factors represent the N released to the environment per unit of food (Leach et al 2012), which includes excess N from fertilizer and manure management in agriculture as well as transportationrelated NO x emissions and N losses from food waste sent to landfills. Separate VNFs for Canada and Québec are not currently available so we use default US national average values in SIMAP; we expect that this may slightly overestimate N release for food production given the typically lower recommended N fertilizer application rates for key food crops in Québec compared to the US national average (as much as 62% lower on average in Québec for soybean, wheat, apples, tomatoes, lettuce, potatoes, and onion) despite typically lower crop yields (15% lower on average in Québec) (CRAAQ 2010, Leach et al 2012, ISQ 2019).
Detailed annual food purchases for McGill's five dining halls were available from the two main food suppliers for the 2015-2016 period, including the University's own farm. For UdeM, food purchase inventories from the university dining halls (annual basis for the main supplier and for a three-month period for others) and weekly purchases for 16 of the 20 student run cafés were available. These detailed inventories were subsequently scaled to represent food purchases during a full year across all food service locations (following Leach et al 2013). Further details are provided in the SI text on the food calculations, including estimates of the shares of local and organic food in overall sales.

Wastewater
Montréal has a combined sewage overflow system that drains to the Jean-R. Marcotte treatment plant, which is a chemically enhanced primary treatment (CEPT) Figure 2. Basic accounting framework in SIMAP, including each calculation conducted within scopes 1, 2, and 3 for the two universities in this study. The curved arrows represent the main form of N released to the environment associated with each footprint component in SIMAP. Scope 1 emissions are those resulting from upstream sources owned by the university, scope 2 includes emissions from sources that are not controlled or owned by the university but that depend on the university's energy consumption, and scope 3 includes downstream sources over which the university is considered to have less control (GHG Protocol 2004). system and the second largest wastewater treatment plant in the world (Canadian Water Network 2018). However, as with many wastewater treatment plants in eastern Canada, the plant is primarily designed to remove phosphorus (Oleszkiewicz and Barnard 2006). Primary treatment with coagulants to remove organic waste results in an incidental removal of about 20%-25% of total Kjeldahl nitrogen (TKN), which includes removal of ammonia and organic N, but excludes nitrates/nitrites (personal communication, Tony Di Fruscia, August 2019). However, because most sewage sludge is subsequently incinerated, we assumed that there was effectively zero net-removal of reactive N in our total N footprint calculation. We use three alternative approaches to estimate lower-, middle-, and upper-bound scenarios of wastewater in the total N footprint. These scenarios follow the 'food consumption' approach of Leach et al (2013), per capita N intake for the approximate total number of meals consumed annually on campus with the recommended daily allowance (RDA) for protein, and per capita wastewater discharge estimates for education institutions in Québec (detailed in the SI text). Since almost all N intake by adults is excreted, we assume in the 'food consumption' and RDA scenarios that N intake in food is excreted to wastewater (Liang et al 2018). We apply the upper-bound per capita wastewater discharge scenario except where specified.

Counterfactual scenarios
We use a series of counterfactual scenarios for food production, purchased electricity, commuting, and wastewater to examine how shared geographic context affects the total N footprint value of each university. When comparing these scenarios to the actual (baseline) circa 2017 N footprints, we estimate the magnitude by which each university's footprint is reduced or increased relative to an alternative situation, such as: low availability of off-campus housing and dining options; fossil fuel reliance in purchased electricity; lower walkability and public transit access; and municipal primary sewage treatment without incineration. We test scenarios that emulate the effect of alternative geographic contexts and use these to make inferences about the effects of shared city and provincial factors in our analysis: (1) A larger direct role of the university in food intake on campus: in this scenario, we assume that all food sales on campus are controlled directly by the university and that UdeM has a similar university food service configuration to McGill (i.e. more residential student dining halls).
(2) A lower share of renewables in purchased electricity: assumes average US eGrid emission factors instead of the current energy mix in Québec (i.e. emission factors per kWh for NO x 79 times higher and N 2 O 119 times higher than the baseline).
(3) A larger share of car commuting for students: assumes that student commuting mode share is equivalent to faculty mode share at each university, resulting in more car commuting and less active or public transportation (i.e. ∼30% personal car travel for student commuters instead of ∼8% in the baseline).
(4) Beneficial reuse of sewage sludge: the 20%-25% of the N influent to Montréal's wastewater treatment plant that is removed is mostly transferred to the atmosphere due to sludge incineration. In this scenario, we therefore assume that 25% of N influent is captured and beneficially reused thereby offsetting the N footprint as a credit (sensu Leach et al 2013).

Results and discussion
McGill's institutional N footprint is ∼48% larger than UdeM's and is consistently larger based on standardizations per capita, university budget, and built area bases. However, the N footprint estimates at both universities are highly sensitive to assumptions about food and wastewater. We therefore present scopes 1 and 2 separately and provide multiple estimates of the total institutional N footprint with scope 3 according to different system boundaries for wastewater (table 1). While several institutional factors affect the relative N footprint magnitudes (table 2), shared geographic context factors often have a relatively stronger role and effectively reduce or increase each of the university's footprint relative to institutions in other settings (table 3).

Breakdowns of McGill and UdeM N footprints
The total N footprint of McGill University with wastewater is ∼121.2 t N yr −1 , or 3.3 kg N per capita FTE, while UdeM's is ∼74.1 t N yr −1 , or 1.6 kg N per capita FTE ( figure 3). McGill's agricultural research farm contributes the largest share of scope 1 N release (11.4 t N yr −1 , mostly as N 2 O from N fertilizer use and other N from animal waste), followed by on-campus stationary fuels (6.4 t N yr −1 ), and research animals (3.8 t N yr −1 ). McGill's scope 2 emissions total ∼3.0 t N yr −1 , with roughly half coming from minor NO x and N 2 O emissions from hydroelectricity or associated with the small share of fossil fuels in Québec's electricity grid, and roughly another half coming from purchased steam and hot water from natural gas burners from a nearby hospital site. In contrast, the largest share of the UdeM's scope 1 and 2 emissions are from oncampus stationary fuels (4.4 t N yr −1 , including natural gas burners) and purchased electricity from the hydroelectricity-dominated grid (1.0 t N yr −1 ). Waste from and disposal of research animals contribute the next largest share at 2.4 t N yr −1 , while livestock animals at the University's veterinary college contribute a small share (<0.1 t N yr −1 ). Given the small contributions of both universities to scope 1 and 2 emissions, the largest share of their N footprints is linked to scope 3 emissions (79% and 89% of the N footprint for McGill and UdeM, respectively). Wastewater is by far the largest single source at ∼50.4 t N yr −1 for McGill and ∼35.2 t N yr −1 for UdeM. Our alternative lower-and middle-bound estimates for wastewater range from 5.2 to 22.3 t N yr −1 for McGill and from 2.0 to 24.0 t N yr −1 for UdeM (see section 3.3 for details).
Nitrogen release associated with producing food for McGill's university-run cafeterias comprises ∼27.1 t N yr −1 , of which about 7% (1.9 t N yr −1 ) is linked to food produced at the University's own farm. A substantial share of N linked to food is composted at McGill, which contributes a 'credit' of -3.4 t N yr −1 to the footprint. The UdeM's N release linked to food production for campus food services is about 8.4 t N yr −1 with a credit of -0.7 t N from composting.
McGill's non-local travel comprises 12.5 t N yr −1 , most of which is international student air travel to/from home, while local commuting comprises 9.0 t N yr −1 mostly from student commuting by bus and faculty/staff car travel. Non-local travel (11.3 t N yr −1 ) and local commuting (11.9 t N yr −1 ) contribute nearly equal contributions to UdeM's N footprint. Given uncertainty about the specific origin city of students and their travel habits, we examined the influence of our two major assumptions around international student air travel at each university. We varied the number of trips from 1 to 3 trips per student per year and tested different points of origin (capital city, largest city, population centroid and busiest airport) in home countries. In both cases, the assumption about the number of trips taken per year had a greater effect on the total N footprint (up to 11% and 16% increases for UdeM and McGill, respectively; see SI text for details) rather than the location of departure within origin countries.

Institutional drivers of differences in the institutional N footprints
Although UdeM has a larger campus population in terms of both student and staff numbers and McGill has a larger built area as well as budget (table S1), we find that McGill's N footprint is consistently greater than UdeM's (from 1.4 to 2.0 times larger) according to different standardizations per unit population, dollar, and area (figure 4). For both McGill and UdeM, the standardized footprints are highly sensitive to assumptions about N release linked to wastewater, which results in intra-university variability that is of a comparable magnitude to the inter-university differences (1.7and 1.9-times greater N release when including wastewater for McGill and UdeM, respectively). Since urban geographic context is similar between the two universities, we systematically compared each of the N footprint components by scope to examine how institutional factors determine the disparities in the total N footprints. The share of McGill's N footprint from scopes 1 and 2 is roughly double that of UdeM's, mainly due to McGill's use of purchased steam, more oil consumption on campus, and the presence of a research farm. McGill's research farm considerably elevates its footprint relative to UdeM (by 11.3 t N yr −1 ), which does not have a farm but has a small livestock animal population at its veterinary college (<0.1 t N t N yr −1 ). If N release linked to livestock feed production were considered, McGill's total N footprint would be at least 33% higher (an additional ∼48.5 t N yr −1 release linked to animal feed production, based on the approximate weight of animal production in 2015 and the virtual N factor approach of Leach et al 2013).
The largest factors that affect the relative N footprints relate to scope 3, including differences in the magnitude of on-campus food services and residential student populations among the two universities (table 2). The types of foods served at the universities differ considerably, with more full meals represented in McGill's data for residential dining halls; McGill also has a much larger share of animal products (dairy, eggs, and meat) in its food purchases (∼35%) compared to UdeM (∼16%) on a mass-basis, which contributes proportionally to its larger food production N footprint. McGill's larger overall role in campus food services and different types of foods served therefore results in a per capita N release for food production ∼4 times that of UdeM (0.86 kg N versus 0.22 kg N student −1 yr −1 , respectively). Since our 'per capita discharge' method for estimating wastewater N is determined by whether students live on-or off-campus, McGill's larger residential student population also considerably elevates the total N footprint (by 12.6 t N). However, when using a 'per capita daily protein intake' method, UdeM's larger population results in a slightly higher wastewater estimate (by 1.7 t N yr −1 ).
Other institutional characteristics that increase McGill's N footprint include its larger share of international students than UdeM (27% versus 13%, respectively), which mainly elevates NO x emissions related to international air travel. McGill is primarily an English language institution that draws a considerable student population from other Canadian provinces and countries; in contrast, UdeM is primarily a French language institution where the student base is primarily from Québec. These factors in turn influence student housing, air travel, and commuting. UdeM has slightly more car commuting, elevating its Table 2. Effects of university-specific factors (institutional context) as drivers of the differences in the N footprint results. The effect of each driver is estimated in the third column, sorted from highest to lowest magnitude.

Difference in N footprints
Institutional context driver Likely magnitude of the effect on each university's N footprint

Large difference in food production N (McGill>UdeM)
McGill has more food services on-campus due partly to mandatory meal-plans for residential students (Scope 3). McGill also has a different food composition (e.g. more full meals and meat served), which increases the food production N footprint.
Food production is ∼18.7 t N yr −1 higher for McGill than UdeM, ∼4 times greater per student FTE.

Large difference in wastewater N (McGill>UdeM)
McGill has a larger residential student population, therefore more water discharge per capita (Scope 3).
Wastewater discharge is ∼12.6 t N yr −1 greater at McGill than UdeM; UdeM has slightly higher wastewater N based on a per capita  Limited direct role of campus food services in total food intake by the university population reduces Scope 3.
Inferred reduction effect of between 29%-68% a assuming a more widespread direct role in food consumed on campus.
Small role of purchased electricity (−)

Reliance on renewables in
Quebec's electrical grid.
Inferred reduction effect of 17%-21% assuming average US eGrid emission factors instead of the current energy mix in Quebec. Relatively small role of commuting (−) Dense urban setting, walkability and plentiful public transit options, often powered by hydroelectricity or biodiesel.
Large % mode share of walking/biking/transit and low % of car travel reduces Scope 3 emissions for students.

Large role of wastewater (+)
Municipal combined overflow sewage system with primary treatment and incineration of sludge.
Effective lack of N removal by central treatment system increases Scope 3 emissions.
Inferred increasing effect of 11%-13% assuming 25% N removal with beneficial reuse of sewage sludge instead of incineration. a A multiplication factor of ∼2.5 times the current food production N footprint is used for McGill, which results in a per capita food production N footprint of 1.8 kg N capita FTE −1 yr −1 . This assumes that the current food mix is held constant (i.e. no change in the types of foods being sold) but that there is an increase in the volume of food services controlled directly by the university (see SI text for details). For UdeM, we apply the same per capita food production footprint as in McGill's scenario (above) to UdeM's population to mimic a situation with more residential student dining halls, more full meals (and animal products), as well as a larger share of food services controlled directly by the university. . We estimate that ∼7% of the virtual N release in the food production component for McGill's campus food services occurs on the campus research farm, which sells produce, eggs, and milk to campus cafeterias; as a result, there may be some minor double-accounting of N with the scope 1 'fertilizer & livestock' component but it is currently not possible to adjust for this with the available data. * N release in wastewater is the most uncertain term in each of the university N footprints, so a range of values is given (see figure 5 for details). McGill's estimate of directly financed travel (business & sports travel) does not include ground transportation as no data were available, except in the case of sports teams' travel by charter bus.

Urban geographic context drivers of institutional footprints
Both Montréal universities have much lower per capita N footprints than US universities with published N footprint values (from 7 to >25 kg N per full time student ;Castner et al 2017). This is largely due to the very small scope 2 emissions with hydroelectricity and the limited role of food services, which we attribute to geographic context with university locations in a dense city with plentiful off-campus food options and modest campus residential student populations.
In table 4, we compare the two Montréal university N footprints in this study to three universities with comparable overall populations. Of these, two are in much smaller cities (Charlottesville and Fort Collins, USA) that have a large share of coal in their fuel mix, which considerably elevates their total and per capita N footprints compared to the Montréal universities, despite having much higher rates of N removal in wastewater treatment. The University of Melbourne, Australia, is located in a city of similar size to Montréal, with a total and per capita N footprint similar to that of the Montréal universities (i.e. 139 t N yr −1 and 2.6 kg N per capita). Liang et al's (2018) assessment of the University of Melbourne N footprint found that electricity generation represented 29% of the total given use of fossil fuels (e.g. brown coal) compared to just 3% for the food consumption component due to secondary wastewater treatment; however, their study omitted non-local student travel despite a student population that is 36% international (University of Melbourne 2018). Our N footprint comparison illustrates how geographic and institutional context interact with system boundaries of the analysis in complex ways to affect institutional N footprints. Both of the Montréal universities considerably lower N footprints than would be likely if located in a different geographic context, such as in a rural or suburban area with higher residential student populations, or in a region with a lower share of renewables in the electrical grid. We examine these effects based on the counterfactual scenarios described next (table 3).

Campus food services
Based on a simple counterfactual scenario where each university is assumed to have a more widespread direct role in the total food intake on campus, the N footprint of McGill would be 40.7 t N yr −1 greater and UdeM's would be 70.6 t N yr −1 greater. We therefore interpret that the current campus food service arrangements reduce the theoretical potential institutional N footprints for McGill and UdeM by 29% and 68%, respectively. Of the seven initial US colleges and universities with N footprint calculations using NFT methodology, all have much higher shares of food production in their totals  likely given a greater role of the institutions in food services and higher proportions of students living in campus residence.

Utilities
Assuming a counterfactual scenario where the electricity purchased by McGill and UdeM is sourced with the US average energy mix, McGill's N footprint would be 22.4 t N yr −1 greater and UdeM's 17.6 t N yr −1 greater-meaning that the reliance on renewables in purchased electricity reduces McGill's and UdeM's footprint by 17% and 21%, respectively. Therefore, access to hydroelectricity effectively reduces the Montréal university N footprints relative to contexts with lower shares of renewables in the electricity grid.

Commuting
A large share of students at both Montréal universities commute via active transportation (walking or cycling) or via subway, powered by hydroelectricity. We expected that this would reduce the footprints of both universities relative to a setting where there is more car commuting. For example, in their comparison of the N footprint for seven US universities and colleges, Castner et al (2017) found per capita commuting values around 1 kg N per FTE student for institutions in less dense urban or suburban settings compared to our estimates of 0.22-0.23 kg N per FTE student at McGill and UdeM. In our counterfactual scenario for commuting, we therefore assume that students commute with the same mode share as faculty (e.g. ∼30% car mode share instead of the actual ∼8% car mode share for students, with lower active transport mode shares). Contrary to our expectations, in this scenario reflecting a context where students have less access to public and active transit options, the Montréal university N footprints would be only ∼2%-3% greater (∼3.1 t N yr −1 for McGill and ∼1.4 t N yr −1 for UdeM). The effect of urban context on N footprints due to commuting is likely relatively small as long as car mode share is held at a moderate level.

Wastewater treatment
Our final counterfactual scenario addresses the effect of beneficial recycling of N from wastewater. If ∼25% of N were captured in the wastewater treatment plant and recycled to, for example, agricultural lands instead of lost to the atmosphere via incinerated sewage sludge, it would offset McGill's N footprint by ∼12.6 t N yr −1 and UdeM's by ∼8.8 t N yr −1 . We therefore infer that wastewater increases our institutional N footprints by 11%-13% relative to a municipal context with basic aerobic wastewater treatment focused on reducing reactive N losses to the environment. However, this is not an ambitious scenario, as municipal wastewater treatment plants with tertiary treatment remove as much as 90% of reactive N and involve complete denitrification to inert N 2 gas (Oleszkiewicz et al 2015).
3.4. Complexities of institutional system boundaries for N footprints Food and wastewater are the single largest fluxes in McGill's N footprint, but our counterfactual scenario for food production indicates that the food component is likely heavily reduced since the university controls only a fraction of food consumed on campus. While wastewater is also the largest flux for UdeM, the university oversees an even smaller fraction of all food theoretically consumed on-campus. Therefore, N discharged in wastewater at both universities likely results in large part from food that was sourced from off-campus or from food services not controlled by the institutions. Figure 5 illustrates the mismatch between the food production and wastewater system boundaries. Comparison of our three wastewater N scenarios shows the strong effect of McGill's residential student population in the per capita discharge scenario (∼50.4 t N yr −1 ) compared to limiting wastewater N to excretion from food consumed at university-run food services (∼5.2 t N yr −1 ); at UdeM, its larger overall campus population drives a much larger value (∼24.0 t N yr −1 ) in the protein intake (RDA) scenario compared to food consumed at university-affiliated food services (∼2.0 t N yr −1 ).
Institutional N footprints were developed for US universities and colleges that often have a large proportion of students living in residence. While both Montréal universities have students in residence, UdeM's >1000 residential students mostly prepare their own food while McGill's >3000 residential students have mandated meal plans (table S1). The use of scope 3 scenarios that estimate all food consumed on campus by the university population regardless of the direct institutional involvement may therefore ensure fairer comparisons with universities in diverse contexts. While the Montréal universities source purchased electricity from renewable sources historically subsidized by the provincial government that reduce NO x and N 2 O emissions, the absence of effective N removal in wastewater treatment by the municipal government results in tradeoffs in terms of N release to the environment that influence potential damage costs (i.e. riverine N inputs and atmospheric N 2 O emissions).
To assess the potential to mitigate N losses linked to activities at McGill and UdeM, we categorized each of the components according to the degree of institutional control versus the potential for additional change in N footprints (figure 6). Scope 3 components have the greatest potential to reduce the N footprints of these Montréal universities, since emissions related to purchased utilities are already small in Québec. UdeM has considerable opportunity to increase composting, with a moderate mitigation potential, while both universities could further address academic travel (e.g. through tele-conferencing or shifting shorthaul flights to rail). Both universities are currently undertaking new initiatives that could affect their Future research on the effects of institutional context should explore how university structure and governance affect N footprints and mitigation potential, including via sustainability commitments under frameworks such as the AASHE STARS system, which also differ among UdeM and McGill. A detailed consideration of how these universities can learn from each other and partner to make an impact on each other's N footprints warrants more systematic consideration, particularly as each school continues to adopt new sustainability strategies.

Conclusions
We identified institutional and geographic factors that affect the relative and absolute N footprints of the two major Montréal universities. McGill's larger N footprint is mainly due to its greater share of residential and international students compared to UdeM Figure 5. Depiction of the system boundary issues faced for food production and food consumption footprint components in the McGill (A) and UdeM (B) institutional N footprints. Each square represents ∼1 t N in the footprint for each university. The left panels show the virtual N release linked to food sold at campus food services, which are assumed to be a relatively small share of total food intake on campus (no data are available for all food consumed on either campus). The right panels show the potential N released from wastewater following the three alternate approaches in this study: the N release just from food consumed at campus food services (light green), the N release linked to all meals consumed on campus (light-yellow, where each meal served is ∼1/3 of the recommended daily allowance for protein intake, 17 g of protein per person per day with protein assumed to be 16% N), and the total N release associated with wastewater discharged to the central treatment system (brown).
(affecting the number of meals served at campus cafeterias, wastewater discharge, and air travel, respectively) as well as the presence of a research farm. While these university-specific factors shape many N losses under the university's control, geographic factors typically beyond the university's control have an overarching influence by either reducing or increasing the N footprints. Our findings therefore highlight that while N footprints are useful campus sustainability benchmarks, inter-university comparison is challenging due to the strong effect of geographical setting and system boundaries on scope 2 and 3 emissions. Higher-education institutions can play an important role in sustainability efforts, for which the N footprint represents one example. Estimating how much of an individual consumer's per capita N footprint (Leach et al 2012) is accounted for within an institution's N footprint could allow for investigation of the mitigating potential of alternative scenarios (e.g. the effects of students living on-campus versus off-campus). When comparing institutional N footprints, consideration of the role of universities in food consumption on campus, residential student population, wastewater treatment, and the degree of renewables in utilities should be considered. Our analysis also stresses how a university's international ties that may represent a strength academically can also increase its N footprint-in particular, the share of international students and degree of international research travel that drive NO x emissions for air travel. With a community equal to roughly 5% of the City of Montréal's population, our case study of these urban universities can also inform efforts that institutions can make toward reducing city-level N footprints. Figure 6. Qualitative ranking of key N footprint components from figure 2 according to the potential to change the institutional N footprint magnitudes for each university versus the relative degree of institutional control.