Concealed nitrogen footprint in protein-free foods: an empirical example using oil palm products

In many cases, the N footprint is calculated on a national scale (e.g. Leach et al 2012, Shibata et al 2014, Hayashi et al 2018, Oita et al 2018). The N-calculator method expresses the food N footprint of item i as:

$$\begin{eqnarray}&&NF=N{F}_{{\rm{prod}}}+N{F}_{{\rm{cons}}}\end{eqnarray} \tag{ 1.1 }$$

$$\begin{eqnarray}&&N{F}_{{\rm{prod}}}=VNF\cdot ConsN\end{eqnarray} \tag{ 1.2 }$$

$$\begin{eqnarray}&&N{F}_{{\rm{cons}}}=\left(1-{R}_{denit}\right)\,ConsN,\end{eqnarray} \tag{ 1.3 }$$

where NF is the food N footprint of item i (kg N capita^–1 yr^–1), NF_prod is the food-production N footprint of item i (kg N capita^–1 yr^–1), and NF_cons is the food-consumption N footprint of item i (kg N capita^–1 yr^–1). VNF is VNF of item i (kg N kg^–1 N), which is defined as the amount of N released into the environment per unit N consumption of item i from its production until just before consumption (Leach et al 2012) (figure 1). VNF incorporates the effects of international trade, that is, N loads within a target country and those in food-exporting countries (e.g. Leach et al 2012, Shibata et al 2014). ConsN is the N consumed in item i (kg N capita^–1 yr^–1), and R_denit is the average conversion ratio of reactive N to N₂ by complete denitrification during sewage treatment (kg N kg^–1 N). N emission as N₂ is excluded from N loads because of the inert nature of N₂. For example, R_denit in Japan is estimated to be 32% (Oda and Matsumoto 2006).

Zoom In Zoom Out Reset image size

For protein-free foods such as oil and sugar, their VNF values become infinity by definition (figure 1), and the NF_prod is therefore expressed as ∞ × 0 and, for practical purposes, is treated as zero. Note that NF_cons of protein-free foods can be treated as zero because of its nonexistent N content.

The present study introduces a substitution factor for VNF, the VNF for protein-free foods (VNFree), to calculate NF_prod of protein-free foods. Oil palm products are used as an example. VNFree is defined as the environmental N load from production until just before consumption per unit weight of consumed food (kg N kg^–1 food). NF_prod using VNFree is expressed as:

$$\begin{eqnarray}&&N{F}_{{\rm{prod}}}=VNFree\cdot ConsA,\end{eqnarray} \tag{ 2 }$$

where VNFree is the VNFree of item i (kg N kg^–1 food) and ConsA is the consumed amount of item i (kg food capita^–1 yr^–1).

We aimed to obtain the global mean VNFree values of PO and PKO because most oil palm production is concentrated in three tropical countries (Indonesia, Malaysia, and Thailand) and oil palm products are available worldwide via international trade (Dittrich et al 2012). Ideally, estimation of environmental N loads accompanying oil palm production should be validated with field data. However, the collection of such field data is still limited (e.g. Pardon et al 2016a, 2016b, 2017). Furthermore, a discrepancy between respective field data and world mean values is a source of uncertainty. Thus, estimating global VNFree values of PO and PKO using national-scale data is consistent with the aim of the present study.

According to the nitrogen mass balance, the total environmental N loads associated with oil palm products are approximated by the new N input to their production while assuming a steady-state, that is, no stock change in N pools (e.g. soil and biomass) relevant to the production. The new N input is calculated by subtracting recycled N (e.g. N in manure and crop residues as organic fertilizer) from the total input. Enhancing N recycling can reduce chemical fertilizer demand and thus reduce the new N input. In the present study, chemical fertilizer, BNF, and atmospheric deposition were considered as components of the new N input. The VNFree of oil palm products as the sum of PO and PKO (VNFree_oilpalm, kg N kg^–1 oil) is expressed as:

$$\begin{eqnarray}&&VNFre{e}_{{\rm{oilpalm}}}={N}_{{\rm{new}}}\,/\,{Y}_{{\rm{oilpalm}}}\end{eqnarray} \tag{ 3.1 }$$

$$\begin{eqnarray}&&{N}_{{\rm{new}}}={N}_{{\rm{fert}}}+{N}_{{\rm{BNF}}}+{N}_{{\rm{dep}}}\end{eqnarray} \tag{ 3.2 }$$

$$\begin{eqnarray}&&{Y}_{{\rm{oilpalm}}}={Y}_{{\rm{PO}}}+{Y}_{{\rm{PKO}}},\end{eqnarray} \tag{ 3.3 }$$

where N_new is the new N input to oil palm plantations (kg N ha^–1 yr^–1), Y_oilpalm is the oil yield as the sum of PO and PKO (kg oil ha^–1 yr^–1), N_fert is the chemical fertilizer application rate (kg N ha^–1 yr^–1), N_BNF is the BNF rate (kg N ha^–1 yr^–1), N_dep is the atmospheric deposition rate (kg N ha^–1 yr^–1), Y_PO and Y_PKO are the yield of PO and PKO (kg oil ha^–1 yr^–1), respectively. Factors such as unit price or production amount can be used as a weight to allocate VNFree_oilpalm to PO and PKO. In the present study, a production value (=oil yield × unit oil price) was chosen as the weight because we consider that the total N loss to the environment from each product should be proportional to its economic value. Using the production value as the weight, VNFree values of PO and PKO are obtained as follows:

$$\begin{eqnarray}&&VNFre{e}_{{\rm{PO}}}=VNFre{e}_{{\rm{oilpalm}}}\cdot P{V}_{{\rm{PO}}}/\,\left(P{V}_{{\rm{PO}}}+P{V}_{{\rm{PKO}}}\right)\end{eqnarray} \tag{ 4.1 }$$

$$\begin{eqnarray}&&VNFre{e}_{{\rm{PKO}}}=VNFre{e}_{{\rm{oilpalm}}}\cdot P{V}_{{\rm{PKO}}}/\,\left(P{V}_{{\rm{PO}}}+P{V}_{{\rm{PKO}}}\right)\end{eqnarray} \tag{ 4.2 }$$

$$\begin{eqnarray}&&P{V}_{{\rm{PO}}}={Y}_{{\rm{PO}}}\cdot {P}_{{\rm{PO}}}\end{eqnarray} \tag{ 4.3 }$$

$$\begin{eqnarray}&&P{V}_{{\rm{PKO}}}={Y}_{{\rm{PKO}}}\cdot {P}_{{\rm{PKO}}},\end{eqnarray} \tag{ 4.4 }$$

where VNFree_PO and VNFree_PKO are the VNFree of PO and PKO, respectively (kg N kg^–1 oil); PV_PO and PV_PKO are the production value of PO and PKO, respectively (e.g. USD ha^–1 yr^–1); and P_PO and P_PKO are the unit price of PO and PKO, respectively (e.g. USD kg^–1 oil).

Oil palm is an arboreal crop suitable for the tropics and subtropics. African oil palm (Elaeis guineensis), native to West Africa, is used for large plantations in Southeast Asia (FAO 2002). Oil palm fruits can be harvested year-round for several decades, although the yield is subject to factors such as plantation life cycle, tree growth, solar radiation, temperature, water regime, nutrient availability, and farm management (Pardon et al 2016a, Woittiez et al 2017). Malaysia and Indonesia accounted for 70% of the world's harvested area and 82% of the fruit production of oil palm in 2014; the fruit production accounted for 86% when including Thailand as the third-largest producer (FAO 2019). Thus, we expect that mean data obtained from these three countries approximate a representative global value.

Table 1 shows the primary data used to calculate each term in equations (3) and (4), and table 2 lists the calculated terms in equations (3) and (4) using the primary data. N_fert was obtained as the mean application rate in 2010 or 2014 by dividing the total chemical fertilizer consumption for oil palm (Heffer 2013, Heffer et al 2017) by the total harvested area of oil palm in the three countries in the same year from two data sources (FAO 2019, USDA 2019). Two conditions of N_BNF were set: either a plantation introduced leguminous cover on the forest floor to enhance BNF or it did not. N_BNF of an oil palm planation using cover legumes was estimated to be 25.6 kg N ha^–1 yr^–1 by dividing the total BNF of leguminous cover for the first 6 years from initiation of the plantation, 640 kg N ha^–1, by the mean plantation life span, 25 years (Pardon et al 2016a). N_BNF of a plantation without leguminous cover was set as the value of non-leguminous crops, 4 kg N ha^–1 yr^–1 (Shindo and Yanagawa 2017). The mean atmospheric deposition rate (wet and dry) in each country in 2009 was obtained from an international atmospheric transportation model intercomparison study (Lamarque et al 2013). Note that 2009 was the latest year in the historical dataset. N_dep in 2010 or 2014 was estimated as the weighted-mean value using the area of oil palm harvested in each country in 2010 or 2014 (FAO 2019, USDA 2019). National production of PO and PKO in each country in 2010 and 2014 were obtained from two data sources (FAO 2019, USDA 2019). The oil yields, Y_PO and Y_PKO, were obtained using economic data by dividing the total production by the total area harvested for the three countries (FAO 2019, USDA 2019). The same data source was used in the calculation (i.e. FAO production was divided by FAO area harvested, and vice versa for the USDA data). Annual mean prices of PO and PKO in 2010 and 2014 were obtained from the price variation charts of crude oil as the cost insurance and freight (CIF)–Rotterdam (Palm Oil Analytics 2019).

VNFree was calculated using the results of each term (table 2), for which the following 64 combinations were used: 2 years × 8 results of N_new × 2 results of Y_oilpalm × 2 results of the weight. As the representative value from 2010 to 2014, global mean VNFree_PO and VNFree_PKO values were derived by averaging the 64 combinations of VNFree values.

Five-year means of domestic production of PO and PKO and domestic consumption of PO and PKO for food, industry, and feed in each country in the world from 2010 to 2014 were calculated using statistics of the USDA (2019) which treat trade (import and export) separately. Five-year mean per capita consumption of PO and PKO for food, industry, and feed was also obtained by dividing the domestic consumption by the national population (UN 2017). These data are presented in the supplementary information.

As a global mean, VNFree_PO was estimated to be 0.0241 ± 0.0032 kg N kg^–1 oil (±SD, n = 64; range, 0.0188–0.0300 kg N kg^–1 oil), and VNFree_PKO was estimated to be 0.0037 ± 0.0005 kg N kg^–1 oil (range, 0.0028–0.0047 kg N kg^–1 oil). VNFree_PO was 6.5 times larger than VNFree_PKO because of the large difference in their yields: 3600–3800 kg oil ha^–1 for PO versus 400–430 kg oil ha^–1 for PKO (table 2). The price of PKO was 1.3–1.4 times higher than that of PO (table 1). The term that mostly contributed to N_new was N_fert, accounting for 76% on average (65%–86%; table 2). VNFree decreases as oil yield increases (equation (3.1)), suggesting that the VNFree values of other oil crops are larger than those of oil palm for several reasons. First, oil palm has the highest oil yield per unit area among oil crops, with yields 5.5–6.2, 7.7–8.8, and 9.7–10.2 times those of rapeseed, sunflower, and soybean, respectively (Basiron 2007, Thomas et al 2015). Second, the oil yield per unit N input, which is calculated by dividing the world oil production (FAO 2019) by the world chemical N fertilizer consumption (Heffer 2013, Heffer et al 2017), for oil palm, soybean, and other oil crops was 49.1, 39.9, and 11.0 kg oil kg^–1 N, respectively; here, oil palm also had a yield 1.2 and 4.5 times those of soybean and other oil crops.

N_fert, N_BNF, N_dep, Y_PO, and Y_PKO calculated using the national-scale data were compared with other available information. N_fert ranged from 71 to 104 kg N ha^–1 yr^–1 (table 2) and was within the range of values reported previously (0–212 ha^–1 yr^–1; Pardon et al 2017) and comparable with actual or recommended fertilizer application rates (e.g. Goh and Po 2005, Comte et al 2012). For N_BNF, the mean value between the N_BNF with cover legume to enhance BNF (25.6 kg N ha^–1 yr^–1) and that without such enhancement (4 kg N ha^–1 yr^–1; table 2) was reflected in the mean VNFree. The collection of BNF data in oil palm plantations is needed because of the lack of field measurements. The annual mean wet deposition monitored at five sites in Indonesia, four sites in Malaysia, and six sites in Thailand in 2010 and 2014 was 8.0 kg N ha^–1 yr^–1 (EANET 2012, 2015). That value seems reasonable as compared with the N_dep as the sum of wet and dry deposition based on the dataset of an atmospheric model intercomparison study, which ranged from 12.0 to 13.2 kg N ha^–1 yr^–1 (table 2). Although the potential oil yield of oil palm is expected to be 9000–10 000 kg oil ha⁻¹ yr⁻¹, average productivity worldwide has stagnated at around 3000 kg oil ha⁻¹ yr⁻¹ (Woittiez et al 2017). An average oil yield of oil palm of 3680 kg oil ha⁻¹ yr⁻¹ was also reported (Basiron 2007). The mean oil yield (PO + PKO) in the three Southeast Asian countries, 4000–4200 kg oil ha⁻¹ yr⁻¹ (table 2), was higher than these world averages but still lower than the potential yield. In addition, the oil yields in the three countries were in the order Malaysia > Indonesia > Thailand, with values of 4500, 3900, and 3100 kg ha⁻¹ yr⁻¹, respectively.

Figure 2 shows the top 20 countries of the food N footprint for oil palm products. Relevant data including oil palm products for industrial use are provided in the supplementary information. PO was the dominant component of the food N footprint; VNFree_PO was 6.5 times that of VNFree_PKO and PO showed higher food consumption than PKO (supplementary information). Indonesia, Malaysia, and Thailand, the three countries with the greatest oil palm production, were ranked within the top ten. African and Latin American countries were also included in the top 20.

Zoom In Zoom Out Reset image size

The food N footprint of oil palm products calculated in the present study was compared with previously reported N footprints of several countries and the world average (Gu et al 2013, Galloway et al 2014, Pierer et al 2014, Shibata et al 2014, Stevens et al 2014, Liang et al 2016, Oita et al 2016, 2018, Hutton et al 2017, Shindo and Yanagawa 2017, Hayashi et al 2018) (table 3). The food N footprint of oil palm products in these countries ranged from 0.07 to 0.18 kg N capita^–1 yr^–1, whereas their food N footprint ranged from 8.9 to 32 kg N capita^–1 yr^–1. Therefore, the food N footprint of oil palm products represented less than 2% of the food N footprint for these countries. Note that the reported N footprints using a bottom-up approach do not include the N footprint of protein-free products like oil palm products. Shindo and Yanagawa (2017), using a top-down approach based on the national and international statistics, estimated that the Japanese food N footprint attributed to oil crops was 0.5 kg N capita^–1 yr^–1. The Japanese food N footprint of oil palm products estimated in the present study corresponded to 20% of that, 0.1 kg N capita^–1 yr^–1 (table 3). On average, oil palm products accounted for 18% of vegetable oil consumption in Japan from 2010 to 2014 (USDA 2019). Thus, the percentages agree well, although it should be noted that the VNFree values of other oil crops might be different from those of oil palm products. The comparison of N footprint including other uses of oil palm is shown in the supplementary information.

The nitrogen investment factor (NIF) (Leip et al 2014) is a similar factor as VNFree used in the present study. The NIF of a product is calculated by dividing the new N input to production by the product N content. However, the NIF of protein-free products is calculated by substituting the raw crop N content for the product N content, which corresponds to the NUE of the new input N to the harvested crop. When applying the NIF concept to a protein-free product in the N-calculator method, conversion of the final product consumption to raw crop consumption is necessary. The VNFree concept suggested in the present study is advantageous for calculating the N footprint of protein-free products directly if these consumption data are available.

The VNFree concept is applicable to several main products used for food and other uses and can be extended to byproducts and recycled materials by consistently using their values (e.g. monetary value in the present study) as the weight to allocate the total N footprint to each product or material. When recycling is costly, that recycled material can have a negative value. However, caution is needed when the main product(s) do not contain N but byproduct(s) and/or recycled material(s) do, as is typical of oil and sugar crops. The current N-calculator method allocates the production N footprint only to products containing N; this implies that protein-free main products do not contribute to N loads to the environment, whereas byproducts and recycled materials containing N are a source of the N loads. Thus, an objective and reasonable rule about how to allocate the N footprint to each main product, by-product, and recycled material is needed. Using available data, the present study employed a monetary value to allocate the total environmental N load for each product. Relative values among products and/or recycled materials vary with domestic and international supply and demand. As a next step, connecting a model of dynamic changes in value to the N footprint calculation would be an effective means of improving the VNFree concept.

Oil palm has the highest oil yield among oil crops both in per unit area and per N input. However, suitable production areas of oil palm are limited to pantropical regions (Pirker et al 2016). Changing pristine and secondary forests into oil palm plantations deteriorates biodiversity and ecosystem functions in these regions (e.g. Fitzherbert et al 2008, Dislich et al 2017). The expansion of oil palm plantations has progressed rapidly since the 1980s in Malaysia and the 1990s in Indonesia (Koh et al 2011, Sayer et al 2012). Ecological impacts are also a concern in Latin America and Africa, where oil palm production continues to expand (e.g. Furumo and Aide 2017, Strona et al 2018). Although the high yield of oil palm appears to be positive from the viewpoint of the N footprint, methodological development is necessary to better assess the benefits and costs of oil palm production and consumption by linking the N footprint concept with other indicators, such as biodiversity, ecological, carbon, and water footprints (Wackernagel and Rees 1998, Hoekstra and Chapagain 2007, Hertwich and Peters 2009, Hillier et al 2009, Hoekstra and Mekonnen 2012, Lenzen et al 2012, van Noordwijk and Brussaard 2014, Nishijima et al 2016). The present study provides a methodological basis to incorporate N loads to the environment towards sustainable oil crop production and consumption.