Assessing future reactive nitrogen inputs into global croplands based on the shared socioeconomic pathways

Nitrogen (N) is a key nutrient for living organisms and has been exceedingly used in food production since the discovery of the Haber-Bosch process for fixing N₂ in the early part of the 20th century (Fowler et al 2013, Galloway et al 1995, Smil 2001). Since then, the application of synthetic fertilizers has boosted the global agricultural production and allowed for the exponential increase in the global population (Galloway et al 2004).

Since at the global scale the crop N yield is less than 50% of the total N inputs (Lassaletta et al 2014) there are major N losses to the environment. N is lost from the soil-plant system by denitrification in the form of dinitrogen (N₂), nitrous oxide (N₂O) and nitric oxide (NO), ammonia (NH₃) volatilization, nitrate (NO₃⁻) leaching, and via water runoff and soil erosion (Bouwman et al 2013). NO, N₂O, and NH₃ are gases that not only affect air quality, ozone chemistry and heat retention in the atmosphere, but also negatively affect human health (Sutton et al 2013). NO₃⁻ in drinking water has potential health impacts and has thus been subjected to regulations in several countries, (e.g. the USA and the European Union; Bryan and van Grinsven 2013, van Grinsven et al 2015a).

Increased N losses to the environment adversely affect freshwater systems, the coastal ocean, and terrestrial ecosystems (Bouwman et al 2005, Seitzinger et al 2010). Nutrients, such as N, stimulate primary production in the hydrosphere, which can ultimately lead to the proliferation of (harmful) algal blooms and the development of hypoxic dead zones, which are detrimental for aerobic organisms (Howarth and Marino 2006, Rabalais 2002, Rabalais et al 2010, Turner et al 2003).

Future global food demand will depend on many factors, including population change, technology advancement, trade and global cooperation, and dietary changes as a consequence of economic growth (e.g. Billen et al 2015, Lassaletta et al 2016, van Grisnven et al 2015b, Westhoek et al 2014). These elements have been integrated into the shared socioeconomic pathways (SSPs; Ebi et al 2014, Kriegler et al 2014, O'Neill et al 2014), which are global storylines that follow internally consistent assumptions on a range of socioeconomic factors, varying according to mitigation and adaptation challenges with a focus on climate change. These SSP scenarios have been implemented with the Integrated Model to Assess the Global Environment (IMAGE) version 3.0 from 2006 until 2050, resulting in energy, food production and trade, greenhouse gas emissions, climate change, and land use and cover projections based on prescribed population and economic growth for 26 global regions (van Vuuren et al 2017).

In this paper, we quantify and scrutinize the future N inputs into croplands required that satisfy the demanded agricultural N production on a global scale. N agricultural production was derived from IMAGE 3.0 assuming the contrasting social, agronomical, and technological evolutions of each SSP. We furthermore calculate the scenario-specific future need for synthetic fertilizers according to the regional agricultural N dynamics. Ultimately this analysis can pinpoint where specific regions may pose environmental issues related to N inputs into croplands and the feasibility for these regions to improve their N use efficiencies by improving N management in croplands.

2.1. General

We calculate future (2006–2050) global N inputs into croplands by assessing the crop requirements according to each SSP scenario and for each of the IMAGE world regions (table S2). The model uses cropland and grassland areas as simulated with the IMAGE model (Stehfest et al 2014) for the five SSPs (van Vuuren et al 2017). Synthetic N fertilizer (S, kg N ha⁻¹ yr⁻¹) inputs into croplands are estimated as the difference between total N inputs into croplands (I, kg N ha⁻¹ yr⁻¹), and the sum of manure (M, kg N ha⁻¹ yr⁻¹), atmospheric deposition (D, kg N ha⁻¹ yr⁻¹), biological N₂ fixation (F, kg N ha⁻¹ yr⁻¹) applied into or reaching croplands:

$$\begin{equation} S = I - (M + D + F) \end{equation} \tag{ 1 }$$

N inputs from manure, atmospheric deposition and biological N₂ fixation are taken from the IMAGE Global Nutrient Model (IMAGE-GNM, Beusen et al 2015, 2016, see SI4). N inputs in intensive grasslands are calculated as the N withdrawn from animal products (N in production of meat, milk, and carcass withdrawal, plus N excretion in grasslands) for intensive livestock production (SI6).

Crop N yields (Y) represent the N in harvested products. We use historical and future aggregated crop N yields obtained from the production for all food, feed, fiber, biofuel and industrial crops and stimulants as well as the N content in the harvested product of each crop, as provided by IMAGE 3.0 for the five SSPs (van Vuuren et al 2017, see SI1). Y can be related to the N inputs in a variety of ways. For instance, the N use efficiency (NUE) is a measure that relates the total N inputs (I) to the crop N yields, that is NUE = Y/I. Similarly, Lassaletta et al (2016), (2014) related Y (expressed in kg N ha⁻¹ yr⁻¹) to I using a hyperbolic relationship as follows:

$$\begin{equation} Y = \frac{{{Y_{{\rm{max}}}}I}}{{{Y_{{\rm{max}}}} + I}} \end{equation} \tag{ 2 }$$

where Y_max (kg N ha⁻¹ yr⁻¹) represents the asymptote for the curve, or the hypothetical maximum yield (see SI2). Note that Y_max differs from the potential yield as defined by the agronomic yield gap approach (e.g. van Ittersum et al 2016). It rather represents a hypothetical value that fits the yield-input relation according to equation 2. This simple yet robust relation based on Y_max has been successfully used to describe input-yield curves at the country scale and for large global regions (Lassaletta et al 2014, Lassaletta et al 2016, Mueller et al 2017), whereby changes in agricultural management (e.g. fertilization technique, irrigation, rotations, crop varieties) are directly reflected in variation of the Y_max value (Lassaletta et al 2014, Lassaletta et al 2016, Zhang et al 2017, Mueller et al 2017). With an unchanged Y_max, N yield can only increase by adding N inputs. Improved N management or improved crop varieties lead to a larger Y_max value, such that with the same input level, N yield will be higher; or the same N yield can be obtained with less N inputs (see supplementary figure 1 available at stacks.iop.org/ERL/13/044008/mmedia). The merger of equations 1 and 2, provides the solution for the synthetic N fertilizer required for each region and SSP:

$$\begin{equation} S = \frac{{{Y_{{\rm{max}}}}Y}}{{{Y_{{\rm{max}}}} - Y}} - \left( {M + D + F} \right) \end{equation} \tag{ 3 }$$

Y-I curves with their respective Y_max have been established for the 26 IMAGE regions at the base year (2005) as a base to develop assumptions for scenario construction.

Production increase can be achieved not just by increasing the crop N yield, but also by expanding the cropland area. The methodology for differentiating the contribution toward future cropland N production (increasing area vs. increasing yield) is given in SI5.

Table 1. SSP descriptions (modified after Mogollón et al in preparation)

Property	2050 level
	SSP1	SSP2	SSP3	SSP4	SSP5
Keyword	Sustainability	Middle of the road	Fragmentation	Inequality	Conventional development
Technological development	Rapid	Medium	Slow	Slow	Rapid
Progress towards development goals	Good	Some	Failure to achieve goals	Highly unequal	Market-driven
Resource intensity	Low	Medium	Very low	Highly unequal	Conventional development
Population (106 inhabitants)	8531	9243	10038	9213	8629
GDP/capita (2005 US$, market exchange rate)	24563	17877	12024	17500	32449
Global greenhouse gas emissions (GtC-eq yr−1)	15.6	19.8	21.5	18.0	27.9
Global mean temperature increase relative to 1860 (°C)	2.1	2.3	2.3	2.2	2.5
Crop production (Mton d.m.a)	4749	5111	5069	4855	5215
Area arable and permanent crops (Mha)	1522	1735	1846	1732	1720
N Production (Tg N yr−1)	120	132	137	126	140
Ymax calculation (FYmax, equation S2)	$\frac{Y}{{\left( {1 - {{NUE}}} \right)}}$	equation S1	Same as previous year	wealth dependent	equation S1

^ad.m. = Dry matter.

2.2. The SSPs scenarios

2.3. Calculating N input and surplus in croplands

N in manure (M), N deposition (D), and biological N fixation (F) by crops were taken from the IMAGE-GNM model (Beusen et al 2016). For further details see SI4.

Synthetic N fertilizer use is calculated with equation (3) based on the total N inputs (I). I for SSP1 (and the wealthy regions in SSP4) is calculated based on modifying the target NUEs of Zhang et al (2015) to our global regions to calculate Y_max. These NUEs assume, among other things, global improvements in water and nutrient management, in the timing of fertilizer application, the cultivation of crops best adapted to the local environment, global access to advanced agronomic technologies, and the implementation of policies to regulate N runoff (Zhang et al 2015). For SSP3 (and the SSP4 poor regions) we assume a constant Y_max (from 2005) provided a lowermost NUE of 20% is met. For SSP2, SSP5, and the middle-income regions in SSP4, equation (3) still contains two unknowns (S and Y_max). Equation (3) was thus solved for each region establishing a relation between Y_max and the gross domestic product—per capita (GDP-pc). The future Y_max for this correlation was calculated using the crop N yield, biological fixation, deposition, and manure of SSP2, together with the synthetic N fertilizer projections from FAO-2050 (Alexandratos and Bruinsma 2012). For further details see SI3. An additional assumption among scenarios is that world regions with Y < 30 kg N ha⁻¹ yr⁻¹ were inferred to adhere to a constant Y_max, provided it meets a minimum NUE constraint of 20% (see table 1). Furthermore a maximum NUE constraint of 85% is applicable to all scenarios (e.g. Giller et al 2004).

N surplus corresponds to the difference between N inputs and outputs. We compare the N surplus for each region and each SSP in 2050 with the maximum allowable threshold of 80 kg N ha⁻¹ yr⁻¹ according to EUNEP (EUNEP 2015). A summary of the various datasets used in this study is available in table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.1. GDP-pc—Y_max relation

3.2. Global N budget under the 5 SSPs scenarios

Figure 2(a) shows that all SSPs follow a modest increase in the crop N production until 2050. This increase, however, is highest for SSP5, which grows by 61 Tg N yr⁻¹, in comparison to most attenuated N production increase (SSP1 by 41 Tg N yr⁻¹). Figure 2(b) shows the N surpluses (positive budgets, inputs > outputs) up to the year 2050. The N surplus in 2005 (86 Tg N yr⁻¹) increases for SSP2 (163 Tg N yr⁻¹), SSP3 (250 Tg N yr⁻¹), SSP4 (173 Tg N yr⁻¹), and SSP5 (124 Tg N yr⁻¹) and only decreases under SSP1 (55 Tg N yr⁻¹).

Zoom In Zoom Out Reset image size

Figures 2(c)–(f) consists of the total N inputs into global croplands. In 2006, the sum of biological N₂ fixation (figure 2(d)), manure (figure 2(e)) and atmospheric N deposition (figure 2(e)) constituted about 46% of the total N inputs into global croplands. The share of these inputs, however, changes substantially toward 2050 for the five SSPs. In SSP1 they eventually represent at 51% of N inputs. SSP2, SSP3, SSP4 show a decrease in the proportion of these inputs because of increased use of synthetic N fertilizers (figure 2(c)), with values of 44%, 33%, 37%, respectively. The temporal pattern for the N surpluses in global croplands (figure 2(b)) strongly resembles the synthetic N fertilizer trends (figure 2(c)), indicating that the former is highly controlled by the latter. The proportion of these inputs for SSP5 decreases toward 2025, but eventually increases back to a value of 46%. SSPs 2–5 all follow a steady increase in manure, deposition, and fixation as N inputs into croplands. SSP1 shows other patterns that reflect changing practices under this sustainable pathway, notably full manure recycling and lower atmospheric N deposition.

Figure 2(c) shows that global synthetic N fertilizer use by 2050 between the SSP scenarios varies by a factor of three. The use in SSP1 remains for the most part stable, with values in the range of 85 Tg N yr⁻¹. SSP2 shows a rapid increase in global fertilizer use that reaches 180 Tg N yr⁻¹ by 2030, and around 185 Tg N yr⁻¹ by 2050. SSP3 is marked by the highest increase in synthetic N fertilizer inputs with values reaching 260 Tg N yr⁻¹. SSP4 shows an increase of global synthetic N fertilizer use that falls between those of SSP2 and SSP3 reaching values of 200 Tg N yr⁻¹ in 2040, before dropping to around 190 Tg N yr⁻¹. SSP5 follows a trend similar to SSP2, but after 2020 levels off to a range of 150–160 Tg N yr⁻¹.

Zoom In Zoom Out Reset image size

3.3. Regional N inputs under the 5 SSP scenarios

3.4. N use efficiencies and surpluses in 2050

Nitrogen fertilization has slowly increased since the early 20th century with a discernible acceleration post 1950. Different approaches have been used to estimate future total N inputs to croplands at the global scale. Bodirsky et al (2014), Liu et al (2016), and Lassaletta et al (2016) used food production, consumption and trade as a basis for forecasting the future N budget in croplands. Our methodology differs in that we do not model the crop N yields. Rather, we take the values generated by the integrated assessment model IMAGE 3.0, which has comprehensively incorporated the socioeconomic aspects of food production and trade to generate a regional N crop production demand. Our approach is similar to Lassaletta et al (2014), but whereas they used historical Y and I to calculate the evolution of Y_max from 1960–2010, we used GDP-pc to estimate Y_max, and calculated Y from IMAGE N production data and land use. These estimates were, in turn, used to calculate I (and eventually S) from 2005 until 2050.

Based on historical trends, Galloway et al (2004) predicted N inputs of 270 Tg N yr⁻¹ by 2050, a figure that compares well with our SSP5 prediction (264 Tg N yr⁻¹). Liu et al (2016) estimated that total N inputs to croplands will reach 270 Tg N yr⁻¹ by 2030, a value which falls between our SSP2 and SSP5 predictions (280 Tg N yr⁻¹ and 262 Tg N yr⁻¹). Excluding our SSP1 scenario, the range of N inputs in croplands globally (262–337 Tg N yr⁻¹) closely follow the estimates of Bouwman et al (2009) who projected future total N inputs into croplands in 2050 to be 292–378 Tg N yr⁻¹ for the four Millennium Ecosystem Assessment scenarios. Our SSP1 scenario, which follows the NUEs of Zhang et al (2015), has predicted total N inputs to croplands of (174 Tg N yr⁻¹) by 2050. These values are slightly larger than the values of this latter study (160 Tg N yr⁻¹). This discrepancy is due to the lower projected requirements in Zhang et al (2015) as compared to our SSP1 scenario (107 Tg N yr⁻¹ vs. 120 Tg N yr⁻¹).

Our estimates of fertilizer use in 2050 range from 85 Tg N yr⁻¹ (SSP1) to 260 Tg N yr⁻¹ (SSP3). The FAO projection for synthetic N fertilizer use of 138 Tg N yr⁻¹ in 2050 (Alexandratos and Bruinsma 2012) is lower than SSP3, SSP4 and SSP5. SSP1 (sustainability) is much lower than FAO, and SSP2 (using FAO trends in productivity) is very similar. Crucially, only SSP1 would be within the planetary boundaries established for N, which are defined as a safe operating space for human societies (de Vries et al 2013, Steffen et al 2015).

Advances in crop management can lead to both an upward shift in Y_max and also an improvement in NUE. Lassaletta et al (2014) determined four possible temporal trajectories for Y_max in various countries from 1960–2010. In their type 3 countries, there was a sufficient increase in Y_max leading to a noticeable improvement in NUE and thus increasing crop N yields were maintained even with decreasing N inputs. Type 3 particularly applies to Western Europe. Their type 2 suggested an increase in Y_max coupled to an increase in the N input, such that NUEs would improve despite an increase in synthetic N fertilizer usage, a situation observed in the USA and Brazil. For the future, a combination of these paths on a global scale is required in order to reach the NUEs suggested by Zhang et al (2015) by 2050. For instance, countries must increase their Y_max values by a factor ranging from 1.1 (Japan) up to 4.4 (Eastern and Western Africa) during this timeframe. On a country basis, Lassaletta et al (2014) found that most countries in their type 2 and type 3 increased their Y_max between 1960 and 2010 by a factor of 2.5, with increases of up to a factor of 4 for countries with low Y_max values (< 50 Tg N yr⁻¹) in 1960 (Lassaletta et al 2014). Thus, countries and regions with low Y_max values may be able to achieve much larger increases in Y_max toward the year 2050. Nevertheless, countries in Lassaletta's type 1 and type 4 had either a stable Y_max or were in the N mining realm. Thus, in order to reach the NUE of SSP1, Lassaletta's type 1 and type 4 countries must switch to type 2 or type 3.

High-income regions show little variations in N inputs between the five SSP scenarios because they are already characterized by high Y_max values and NUEs in 2005. Some low- and middle-income regions show little variation for the various SSPs due to low crop demand arising from slow economic and population growth (e.g. Russia, Ukraine, Southern Africa). The greatest potential to improve N yields due to an increase in Y_max thus lies in the remaining middle-income and the rapidly developing countries (figure 3). These regions include South Asia, India, and China, who are characterized by significant large populations and agricultural production, but whose Y_max values are still relatively low (< 110 Tg N yr⁻¹). These regions generally saw little variation in their Y_max development for the 1960 to 2000 timeframe (type 1; Lassaletta et al 2014). Nevertheless, they are predicted to see an acceleration in both population growth and the demand for animal proteins in the future. Therefore, improving agricultural management can mean the difference between only moderate increases in synthetic N fertilizer application (SSP1) and a strong increase in synthetic N fertilizer application coupled to low NUE (figure 4), which consequently would lead to large N surpluses and N losses to the environment.

On a global scale, little to no temporal improvement in Y_max represents the main assumption for our SSP3 scenario and our SSP4 scenario for low-income regions. These scenarios portray a world that will increasingly rely on synthetic N fertilization with NUEs significantly diminishing (figure 4). The SSP5 scenario represents a world with a population of similar proportion to SSP1 (2050) but with less environmental concern and more wealth, which directly translates into more animal protein consumption. While not as substantial as in SSP1, economic improvements in NUEs and Y_max also occur in SSP5. However, the crop N production demanded from agricultural systems is highest of all SSPs, and the N budget difference between SSP5 and SSP1 typically exceeds 60 Tg N yr⁻¹. Likewise, the N budgets increase with time for the SSP2 scenario, despite an increasing Y_max. These scenarios demonstrate that Y_max improvements that are solely tied to economic development are also not sufficient to keep up with the food demand and thus will require an increase in the future global synthetic N fertilizer use in agriculture.

Zoom In Zoom Out Reset image size

The significant increase in crop production from higher N fertilization does not prevent the need for considerable expansion of agricultural land, which takes place under all scenarios except SSP1 (figure 5(b)). While at the IMAGE region scale, no N mining takes place, Lassaletta et al (2014) found that several individual countries such as Morocco and Nigeria have been characterized by negative soil N budgets since the 1960s. This type 4 Y_max path is associated with soil N mining due to low fertilizer application. Depletion of N and other nutrients such as P could trigger cropland expansion. N mining, coupled to increasing N crop demand may pose environmental challenges, as it entices land encroachment and deforestation to make way for new agricultural land. Land expansion may significantly contribute to additional crop production, with up to 32% contribution for SSP3. In contrast, improvements in Y_max for SSP1 lead to no requirements for areal cropland expansion (figure 5(a)).

Ultimately, the dramatic increase in the areal N surplus for most of the studied scenarios threatens the sustainability of food production systems. Only SSP1 reconciles crop production with water and air quality preservation, while in the other storylines the environmental burden would exceed acceptable limits of N pollution and additional environmental impacts due to new agricultural land acquisitions at the expense of natural areas.

Additional N fertilizer application is also required in intensive grasslands due to human consumption of ruminant products. Our estimated N extraction from grasslands with intensive ruminant production was roughly 6.5 Tg N yr⁻¹ in 2005, which is somewhat higher than the 3–4 Tg N yr⁻¹ previously estimated (Bouwman et al 2013). This N imbalance increases for most scenarios, with estimates in 2050 of 4.9 Tg N yr⁻¹, 23 Tg N yr⁻¹, 45 Tg N yr⁻¹, 42 Tg N yr⁻¹, and 28 Tg N yr⁻¹ for SSP1, SSP2, SSP3, SSP4, and SSP5, respectively. N extraction from intensive systems could eventually lead to insufficient N availability for grass production. Therefore, additional fertilization and/or better legume management of intensive grasslands to meet these removed amounts must take place to maintain a balanced system. These values represent 6%–22% of the predicted synthetic N fertilization use in global croplands and thus potential future locations where N application may be necessary.

In order to achieve a global sustainable balance between cropland production and ecosystem preservation, it is important to limit the amount of N emitted to the environment. Following the NUE targets of Zhang et al (2015) for SSP1, the global anthropogenic reactive N produced can be stabilized at current levels (and even slightly decrease) on the global scale. This scenario also leads to a marked decrease in the reactive N demand per capita in agriculture. In contrast, all other scenarios forecast an increase in synthetic N fertilizers to boost agricultural production. The increase in demand due to poor socioeconomic choices could produce a severe environmental impairment in 2050.

Funding for this research was provided by Marie Skłodowska-Curie Individual Fellowship Grant number 661163 awarded to JMM. AFB and AHWB received support from PBL Netherlands Environmental Assessment Agency through in-kind contributions to The New Delta 2014 ALW projects no. 869.15.015 and 869.15.014.