Provincial nitrogen footprints highlight variability in drivers of reactive nitrogen emissions in Canada

We estimated VNFs at the national scale for Canada and separately for each of Canada's ten provinces (the three northern territories were excluded due to limited data). Our study encompasses 45 crops, which we aggregated into 5 crop groups: (a) grains, (b) fruits, (c) vegetables, (d) roots, and (e) legumes (see supplementary information S1, available online at stacks.iop.org/ERL/16/095007/mmedia, for food groups). Animal product VNFs were calculated for beef, pork, chicken, fish, milk, and eggs. We estimated VNFs as the sum of apparent Nr losses along the food supply chain from agriculture, processing, and food waste per kg of food consumed. Nr losses for crops include potential losses on agricultural fields and due to food processing and waste. Nr losses for animal products include the virtual N due to animal feed, manure Nr loss, and food processing and waste (figure 2).

At each step of the food supply chain, we estimated Nr release as the potential combined Nr losses to air and water (kg N). We estimated Nr losses as the difference between new Nr inputs and Nr outputs at a given step, and considered N recycling, leaching and denitrification (table 1). We subtracted inert N₂ from our emissions since our focus is on Nr.

2.1.1. Agricultural field losses

2.1.2. Animal feed and manure

2.1.3. Food processing and waste

We examined different scenarios to assess how assumptions and uncertainties about geographic food sourcing affect our provincial food N footprints. In our main analysis, we apply national weighted average VNFs based on the share of food production coming from each province. This assumes food is sourced nationally in proportion to provincial production. For example, the Canadian average VNF for beef is heavily weighted towards Alberta's beef VNF since this province produces the most beef in Canada. For context, Canada produces far more beef, pork and chicken domestically than it imports (imports account for 18%, 11%, and 13% respectively of the total supply of these foods) (Statistics Canada 2020a).

We applied two scenarios to examine uncertainty using alternative VNFs for each province. First, we used province-specific VNFs, assuming food is produced and consumed in the same province. If provincial VNFs were unavailable (e.g. provincial fruit VNFs could not be calculated for Newfoundland and Labrador due to data availability), we used the Canadian weighted average VNF. Second, because the United States is Canada's largest agricultural trading partner (FAOSTAT 2021b), we also estimated each provinces' food N footprint when using national weighted average VNFs for the U.S. from Leach et al (submitted), representing a scenario where food consumed in each province is sourced from the U.S.

Previous N footprint studies have typically used national average diets to estimate food intake, mainly from FAOSTAT (Leach et al 2012, Shibata et al 2014). For Canada, national food availability data from Statistics Canada (kg person⁻¹ yr⁻¹ adjusted for retail and household losses) (Statistics Canada 2021a) are available. Given the objectives of our study, we use a simple approach to incorporate variations in provincial diets into our model. First, to assess whether there are significant differences in food consumption patterns between provinces, we performed a Kruskal–Wallis statistical analysis using data from the 2015 Canadian Community Health Survey-Nutrition (CCHS-N). The CCHS is a national nutrition and health 24 h average food intake recall survey by Statistics Canada and Health Canada that included >35 000 respondents (Statistics Canada 2007). Several food groups, including beef, were consumed in significantly different quantities (p < 0.05) between provinces. However, when compared to national food availability data from Statistics Canada, the CCHS-N survey appeared to underrepresent absolute per capita meat consumption (in kg). Therefore, we scaled the national food availability values to the provincial level by using the relative provincial consumption values from the CCHS-N survey (see S2 for more detail). Consumed food weight was multiplied by the respective food N contents from the Canadian Nutrient File from Health Canada (2012) to estimate per capita N consumed.

We estimated potential Nr loss during wastewater treatment as the sum of Nr released to water and air during and after treatment as well as Nr emissions from waste incineration. To estimate influent Nr to wastewater, we assumed all food N consumed by the population of each province is excreted (Liang et al 2018b). To estimate aqueous Nr removal we used Nr influent and effluent measurements from wastewater treatment plants representing the three treatment levels in Canada (primary, secondary, and tertiary) as proxies (see S3). Aqueous Nr removal was estimated as the difference between total Nr (total Kjeldahl nitrogen, nitrates, and nitrites) in influent and effluent. We therefore assume that 75% of influent Nr is released to surface waters or groundwater under primary treatment, 66% after secondary treatment and 30% after tertiary. In the case of no treatment, which includes septic tanks, we assume that 95% of consumed Nr is released to water (Brandes 1978). Aqueous Nr removal was then weighted by the percentage of the population in a province that is served by each of three wastewater treatment levels using data from Statistics Canada (Statistics Canada 2021i).

We estimated gaseous Nr released as N₂O, NO_x and NH₃ emissions from wastewater treatment and waste incineration based on data from Canada's National Greenhouse Gas (GHG) Inventory, and from Canada's Air Pollutant Emissions Inventory (APEI). As with our agricultural estimates we assume a 1:1 N₂O to N₂ denitrification ratio. These atmospheric loss values include wastewater Nr from non-food sources (e.g. industry and other discharge), so they are likely proportionally larger than our estimates of Nr release to water, which are based on food N consumption.

The fate of remaining Nr after wastewater treatment (removed as biosolids or sludge and not incinerated) is uncertain. In our main analysis we assume 25% of this Nr is beneficially recycled. However, by some estimates, as much as 50% of wastewater biosolids may be recycled in Canada (Karimi et al 2020), with the rest being landfilled. Therefore, we performed a sensitivity analysis using a lower and upper bound of Nr recycling ranging from a low of 25% to a high of 50% 'beneficial' recycling to land (which lowers the N footprint).

To estimate fossil fuel-related Nr emissions we use a top-down approach, summing data on provincial N₂O and NO_x emissions (the forms of Nr released by burning fossil fuels). Data were obtained from greenhouse gas and air pollutant inventories, which account for N₂O and NO_x emissions, respectively. We then categorized these emissions into four subsectors based on their sources: (a) energy generation (emissions from public electricity and heat production, as well as commercial, institutional, and residential stationary combustion) (b) transportation (emissions from aviation, shipping, rail, trucks, and other vehicles), (c) the oil and gas industry, and (d) other industry sources (stationary energy generation and other emission sources from construction, mining, and other manufacturing industries). The oil and gas industry form a separate category because we expected them to be large sources of Nr emissions in Canada. However, for an N footprint, not all oil and gas industry emissions should be attributed to consumption in Canada since most of the oil is ultimately consumed abroad. Therefore, we only include 15% of the total sector emissions as approximately 85% of oil is exported (Statistics Canada 2016).

Diet is a major driver of N footprints in Canada. The average Canadian N footprint from food production is 12.9 kg N capita⁻¹ yr⁻¹, 50% of the total Canadian N footprint. However, in most provinces, other than Ontario and Quebec, less than 50% of the N footprint comes from food production. Ontario and Quebec are the two most populous provinces in Canada, and together account for nearly two-thirds of Canada's population (40% and 24% respectively). Therefore, the national weighted N footprint is disproportionately impacted by these provinces. Virtual nitrogen from meat production is a substantial contributor to Nr emissions across all provinces. Within the meat subsector, beef consumption is the largest driver. Beef consumption is the single largest overall driver of Ontario's, New Brunswick's, and Quebec's N footprints (5.4, 5.2, and 4.7 kg N capita⁻¹ yr⁻¹, respectively) and second largest in all other provinces except Alberta and Saskatchewan (figures 4(A) and (B)).

Our findings related to food production are broadly comparable to other N footprint studies. For example, Leach et al (submitted) estimated the food production N footprint for the U.S. to be ∼26 kg N capita⁻¹ yr⁻¹. Beyond differences in models, there are notable disparities in diets between the U.S. and Canada. For example, FAO estimates suggest Americans consume more meat than Canadians (e.g. the United States has a beef supply of 37.16 kg capita⁻¹ yr⁻¹ whereas Canada's is 26.78 kg capita⁻¹ yr⁻¹ (FAOSTAT 2021c)). The difference in beef consumption alone could explain ∼3 kg capita⁻¹ additional N in the U.S. N footprint as compared to Canada. We discuss other differences related to VNFs in the next section.

3.1.1. Virtual nitrogen factors

Beef, pork, chicken, have the highest average VNFs of all food groups, which is common among other studies and mainly due to relatively lower efficiency of converting feed crops into edible protein (Metson et al 2020). Crops tend to have lower VNFs than animal products, although there is a large range between provincial values (figures 5 and S4). While differences in types of crops grown and farm N management affect the variability in provincial crop VNFs, environmental factors, such as leaching potential, also play a role. Wetter, colder provinces have higher leaching potential (De Jong et al 2009), which partially accounts for the higher VNFs for most crop groups in Newfoundland and Labrador, Nova Scotia, and New Brunswick .

Zoom In Zoom Out Reset image size

We compare our Canadian VNFs to findings for the U.S. from Leach et al (submitted) (see vertical lines in figure 5) and find that they are of comparable magnitudes for most food items, especially beef (our average weighted Canadian beef VNF is 401 g N kg food⁻¹, the U.S. beef VNF is 456 g N kg food⁻¹ . Our VNFs for pork and chicken are about 45% lower than those for the U.S., while our Canadian fruit and vegetable VNFs tend be higher (figure 5). These differences in the U.S. and Canadian VNFs are partly attributable to differences in crops considered in each food group (e.g. cereals in Canada include corn, rye, oats, triticale and wheat, whereas cereals in the U.S. include corn, wheat, and rice), and agricultural practices between the two countries, such as differences in crop fertilizer recommendation rates (e.g. Ontario, Canada recommends 135 lbs N acre⁻¹ of corn (Ontario Ministry of Agriculture, Food and Rural Affairs 2018) whereas Minnesota, U.S. recommends 145–195 lbs N acre⁻¹ (University of Minnesota Extension 2021)). Differences are also a reflection of methodologies, including data sources, Nr loss accounting, and livestock feed conversion efficiencies. For example, in our approach we use fertilizer sales data whereas Leach et al (submitted), used recommend fertilizer rates by crop and state mainly from university agricultural extension agencies.

Our VNF food origin scenarios resulted in a Canadian average food N footprint ranging from 12.4 kg N cap⁻¹ yr⁻¹ (using provincial VNFs) to 16.7 kg N cap⁻¹ yr⁻¹ (using U.S. VNFs) (figure 6). The Canadian average is heavily weighted towards Ontario, which has roughly half of Canada's total population. This province has relatively low provincial VNFs for most food groups (figure 5), resulting in a 37% increase in virtual nitrogen for the scenario assuming sourcing food from the United States (U.S. VNFs) over the scenario with local food sourcing (provincial VNFs) (figure 6). However, seven out of the ten provinces have their lowest relative food N footprints when assuming food is consumed proportionally to overall Canadian production (Canadian VNFs), which is the approach we use in our main results (figures 3–5). Since beef has a substantially higher VNF than other foods (four times that of pork, the next largest VNF), this was also strongly reflected in our food origin sensitivity analysis. In all scenarios and provinces 35%–40% of virtual nitrogen was from beef consumption.

Zoom In Zoom Out Reset image size

3.1.2. Wastewater treatment

Nr release from wastewater treatment accounts for 3.1 kg N cap⁻¹ yr⁻¹ of the Canadian average N footprint, which makes it the third largest driver of the national and most provincial N footprints. Nitrates and nitrites are not regulated at a federal level across Canada, and ammonia regulation varies by province (Oleszkiewicz 2015). Geographical proximity to the ocean may be reflected in wastewater regulation for relatively low-population density coastal provinces (Nova Scotia, New Brunswick, Prince Edward Island and Newfoundland) as well as British Columbia, which discharge primarily to coastal waters and have generally lower N effluent standards than interior provinces (Ontario, Manitoba, Saskatchewan and Alberta) (Oleszkiewicz 2015). Only ∼30% of Canada's population is connected to municipal wastewater systems that receive tertiary treatment (the treatment level that may include biological nitrogen removal) (Statistics Canada 2021i).

Our wastewater treatment scenarios reflect the relatively low level of Nr removal in wastewater treatment across Canada: when using high (50%) and low (25%) N recycling scenarios, the average Canadian wastewater footprint decreases from 3.1 kg N cap⁻¹ yr⁻¹ to 2.7 kg N cap⁻¹yr⁻¹ (±13%). The largest variation was for Alberta (2.5–3.0 kg N cap⁻¹yr⁻¹, ±17%), which is the province with the highest tertiary treatment coverage (84%). Conversely, in British Columbia only 10% of the population is covered by tertiary treatment (McCourt Sibeal 2021), and it is also the province with the largest wastewater N footprint, varying between 3.1 and 3.2 kg N cap⁻¹yr⁻¹ (±3%) for the high/low scenarios (figure 7).

Zoom In Zoom Out Reset image size

Burning fossil fuels accounts for 11.1 kg N capita⁻¹ yr⁻¹, or 40% of the average Canadian N footprint. Larger provincial fossil fuel footprints mainly coincide with the makeup of energy grids and the presence or absence of fossil fuel or other natural resource extraction. The average energy N footprint in Canada is 1.8 kg N capita⁻¹ yr⁻¹ but varies considerably by province. In Nova Scotia, Saskatchewan and Alberta, 63%, 40% and 43% (Canada Energy Regulator 2021), respectively, of the energy grids are powered by burning coal (figure 1). These three provinces have a combined average energy N footprint of 6.9 kg N capita⁻¹ yr⁻¹, whereas other provinces are powered primarily by renewables and nuclear energy and have an average energy N footprint of 1.4 kg N capita⁻¹ yr⁻¹.

Nr emissions related to transport are the largest portion of the fossil fuel N footprint sector in Canada (7.4 kg N capita⁻¹ yr⁻¹ on average) and in six provinces (figure 4). Canada is one of the least densely populated countries on the planet, has large distances between cities, cold winters, and a strong natural resource sector. This means Canadians drive more than in many other nations (on average Canadians drive ∼15 200 km yr⁻¹ (Natural Resources Canada 2008), Americans, by comparison, drive ∼13 500 km yr⁻¹ (United States Department of Transportation 2018)). In addition, heavy-duty (>3.9 tonne) diesel vehicles are a major contributor to transport emissions in several provinces, typically associated with trucking, mining and other resource extraction activities (Environment and Climate Change Canada 2020a). While resource extraction contributes to a province's economy and thus indirectly benefits Canadians, these activities may be associated with foreign exports and thus consumption activities abroad.

Our top-down approach to N footprints (national and provincial territorial emissions divided by population) differs from bottom-up N footprint studies, which estimate N emissions using personal consumption data (e.g. household electricity use, and personal distances driven or flown), multiplied by emissions factors (Leach et al 2012, Pierer et al 2014, Shibata et al 2014, Liang et al 2016). Our top-down methodology, drawing from territorial N emission datasets, therefore captures additional N emissions when compared to a bottom-up approach, particularly in the case of fossil fuels. Since some territorial emissions may ultimately be attributed to consumption activities in other countries, they should be discounted from our N footprint. However, the specific amount to discount is uncertain. For example, the APEI and GHG territorial inventories report Nr emissions for 'aviation', which includes emissions associated with direct personal consumption (e.g. vacation flights), indirect consumption (cargo flights bringing in imports), and with exports (which should be excluded from an N footprint). A bottom-up approach only captures the emissions from direct personal consumption, whereas a top-down approach potentially captures emissions from all three areas.

To examine this uncertainty, we compared our fossil fuel N footprint results across these different approaches in our national N footprint, and for two resource-dependent provinces (table 2). These two provinces produce most (∼90%) of Canada's oil and gas (Statistics Canada 2016). Alberta alone produced 80% of Canada's total oil in 2019 (∼200 000 m³). We approximated a bottom-up fossil fuel N footprint by including only categories from the APEI and GHG inventories that were similar to those in the N-Calculator (a bottom-up N footprint model, Leach et al 2012), and a strictly territorial per capita N emissions estimate (including all categories with no adjustment for exports). The largest differences occur when discounting exports from the oil and gas industry in Alberta (the main approach we used in this study), and whether emissions from all forms of transport (like heavy machinery) are included in Saskatchewan.

In their top-down global input-output study of 188 countries, Oita et al estimated Canada's per capita N footprint to be ∼63 kg capita⁻¹ yr⁻¹, over twice our estimate of 27.1 kg capita⁻¹ yr⁻¹ (Oita et al 2016). The difference in scopes and methodologies of various N footprint studies means that comparisons need to be done carefully, however, valuable insights can still be gained. Oita et al accounted for a larger range of consumables than in our study, including non-food agricultural products like cotton. We plan to incorporate additional non-food agricultural goods in future assessments of Canada's N footprint, such as textile crops (e.g. hemp and flax), timber products and pet food (e.g. canary seed, of which Canada is the largest exporter in the world (FAOSTAT 2021a)).

Given the relative contribution of meat consumption and transport emissions to all provincial N footprints, moving towards more plant-based diets, and promoting the use of electric vehicles are among policy options that would be relevant across most of the country. Canada has already made strides in this direction by updating Canada's Food Guide in 2019 to encourage Canadians to eat a more plant-based diet (Health Canada 2021c). Furthermore, British Columbia and Quebec have passed legislation to ban the sale of new gas vehicles by 2040 (Canadian Press 2019) and 2035 (Lampert 2020), respectively, and both provinces already provide financial incentives for purchasing electric vehicles. Improving VNFs through better nitrogen use efficiency is also important, particularly for beef, as ammonia emissions from livestock are increasing in Canada (Environment and Climate Change Canada 2021a). Future research studying the change in N footprints over time, in relation to improved N use efficiency in agriculture, changing dietary patterns, and a general movement away from fossil fuels as an energy source, would also be insightful for policy recommendations to lower N emissions. Given the difference in our results between bottom-up and top-down N footprint approaches, additional case studies would be useful to examine how to equitably attribute emissions from transportation, manufacturing, and industrial sectors to a nations' N footprint.