Fertilizer nitrogen recovery efficiencies in crop production systems of China with and without consideration of the residual effect of nitrogen

As an essential constituent of the building blocks of life, i.e., DNA, RNA and proteins, nitrogen (N) is required by all organisms and ecosystems. However, human activities have more than doubled the annual amount of reactive N entering terrestrial ecosystems since the preindustrial era (Galloway 1998, Smil 1999, Green et al 2004). This added N has contributed to soil acidification, eutrophication, global warming and stratospheric ozone depletion, as well as particulate matter and photo-oxidant formation (Vitousek et al 1997).

Synthetic N fertilizer is by far the largest source of anthropogenic reactive N worldwide (Fowler et al 2013); its applications made the doubling of world food production possible in the past four decades (Tilman 1999). However, much of the fertilizer N intended for crops is instead lost to the environment via denitrification, NH₃ volatilization, surface runoff and leaching (Raun and Johnson 1999). The fraction of applied N actually utilized by crops is often low, and much of the remainder can be considered an expensive and environmentally damaging waste of N.

Nitrogen use efficiency (NUE) of crop production is often measured in term of the 'recovery efficiency' of applied N (RE_N), and is defined as (Cassman et al 2002):

$$\begin{eqnarray}&&\left( {{U}_{{\rm N}}}-{{U}_{0}} \right)/{{F}_{{\rm N}}}\end{eqnarray} \tag{ 1 }$$

where U_N is the plant N uptake (kg ha⁻¹) measured in aboveground biomass at physiological maturity in a plot that received N at the rate of F_N (kg ha⁻¹), and U₀ is the N uptake measured in aboveground biomass in a plot without the addition of fertilizer N. By definition, RE_N is relatively easy to measure and thus often used as the indicator of NUE.

China is the largest user of fertilizer N in the world, and many field experiments have shown low RE_N of 30–35% in the 1990s (Zhu and Chen 2002) and 26–28% in 2001–05 for major cereal crops (Zhang et al 2007), relative to 52% in America and 68% in Europe (Ladha et al 2005). This low RE_N has stirred up wide discussion on whether fertilizer N use can be significantly reduced in China (Ju et al 2009, Liu et al 2011, Qiao et al 2012). Long-term experiments on the other hand often yield higher RE_N results (Jensen and Schjoerring 2011), suggesting that part of the N previously applied could be recovered. Even the long-term experiments understate the contribution of fertilizer N because their zero-N plots continue to receive environmental contributions (e.g., dry deposition) of fertilizer N from neighboring plots. In addition, RE_N is dependent on fertilization rate; hence, the RE_N of such long-term experiments may not represent the present day situation if the fertilization rate has increased significantly since the beginning of the experiment.

Although residual N can be traced using the ¹⁵N-isotope technique, this method may incorrectly quantify the amount of residual fertilizer N because pool N substitution causes immobilization of ¹⁵N fertilizer in soil and thus overestimates residual fertilizer N (Krupnik et al 2004). Therefore, many of the data available in the literature are of RE_N and only a few studies have produced data to quantify residual fertilizer N recovery in subsequent crops (Ladha et al 2005), and this is especially true for large-area estimates.

Soil nitrogen budgets (Leip et al 2011, Eurostat 2013) on the other hand require a more thorough quantification of N inputs (as organic fertilizers, atmospheric deposition, and biological N fixation), but allow the estimation of NUE without the need of a 'no-fertilizer' reference.

We hypothesized that there is a significant residual effect of fertilizer N in areas with continuously high N input, such as China. Thus, the objectives of the present study were, by using long-term spatially extensive N budget data, to: (1) evaluate the relationship between crop N uptake and applied fertilizer N; (2) quantify the long-term effect of residual fertilizer N; and (3) illustrate the temporal trend in both RE_N and the residual fertilizer N effect.

2.1. Crop N uptake and synthetic N input

2.2. Non-synthetic N input

2.3. Changes in soil N storage

Changes in soil N content are hard to detect over a short period of time because the change is usually small compared to the large soil N pool. China conducted a national soil survey during 1979–82, and the main results were published in the China Soil Series I–VI (National Soil Survey Office, 1993, 1994, 1995, 1996). These six books contain descriptions (including soil series, location and land use) and some analytical data for 2473 typical soil profiles, among which 1509 soil profiles were from croplands throughout the country. The dataset has been used to estimate the soil organic carbon (SOC) storage of Chinese croplands (Wu et al 2003, Xie et al 2007). Data of total N content for soil depths of 0–100 cm are available for 1213 cropland soil profiles. We conducted another sampling campaign in 2007–08, with soil profile samples (0–100 cm) of 1393 cropland sites across the country. Organic carbon and total N content along with other properties of the samples were analyzed. We have shown that, with these numbers of soil profiles, it is possible to detect statistically-robust changes in the average SOC content (0–100 cm) of Chinese croplands; the detailed methodology of the estimation process is described in Yan et al (2011). Using a similar method in the present study, we detected the changes in the average soil N content in Chinese croplands (0–100 cm) over the period 1980–2007. The following describes the statistical analysis method used to compare the changes in soil N content:

Soil N content does not usually follow a normal distribution. According to (Bernstein and Bernstein 1999), the confidence interval (CI) for the mean of such a population with large sample size (⩾30) can be estimated as

$$\begin{eqnarray}&&\bar{x}\pm {{z}_{\alpha /2}}\frac{s}{\sqrt{n}},\end{eqnarray} \tag{ 2 }$$

where $\bar{x}$, $s$ and $n$ are the mean, standard deviation and size of the sample, respectively; and ${{z}_{\alpha /2}}$ is the Z statistic of a critical level (α).

The CI of the mean difference of two independent large samples (size ⩾30) can be estimated as

$$\begin{eqnarray}&&\left( {{{\bar{x}}}_{1}}-{{{\bar{x}}}_{2}} \right)\pm {{z}_{\alpha /2}}\sqrt{\frac{s_{1}^{2}}{{{n}_{1}}}+\frac{s_{2}^{2}}{{{n}_{2}}}},\end{eqnarray} \tag{ 3 }$$

where ${{\bar{x}}_{1}}$ and ${{\bar{x}}_{1}}$ are the means of the two samples; ${{s}_{1}}$ and ${{s}_{2}}$ are the standard deviations of the two samples; and ${{n}_{1}}$ and ${{n}_{2}}$ are the sizes of the two samples.

The statistical significance of the mean difference of two large samples can be tested using the statistic

$$\begin{eqnarray}&&\frac{\left( {{{\bar{x}}}_{1}}-{{{\bar{x}}}_{2}} \right)}{\sqrt{(s_{1}^{2}/{{n}_{1}})+(s_{2}^{2}/{{n}_{2}})}},\end{eqnarray} \tag{ 4 }$$

and comparing with a standard normal distribution.

The sample taken between 1979–82 and that taken in 2007–08 can be viewed as two independent samples, so we were able to detect if the average soil N content has changed during this time period.

2.4. Statistical evaluation of the relationship between crop N uptake and N input

3.1. Soil N change, crop N uptake and cropland N input

3.2. Relationship between crop N uptake and N input at the provincial scale

We further calculated total crop N uptake and synthetic N input to croplands on a yearly basis from 1980–2010 for each province of China. At this scale, there was a generally linear relationship between crop N uptake and synthetic N input rate (figure 1). For a given year, the crop N uptake is derived from synthetic fertilizer N applied in that year and from a variety of other sources (soil organic N, fertilizer N applied in previous years, N deposition from the atmosphere, biological N fixation, applied crop residue and manure). The slope of the line can be viewed as the average fraction of applied synthetic N taken up by crops in the year, while the intercept of the line can be viewed as the average N uptake by crops from all other sources. Clearly, with the increase in N input, the fraction of applied synthetic N taken up by crop within a year decreased, from around 0.5 in the early-1980s to around 0.3 in the late-2000s (figure 2). By definition, the slope of the regression line should be similar to the RE_N calculated by the N difference method in one-year experimental plots, and indeed the slopes for recent years were (<0.30, figure 2), agreeing well with the widely reported RE_N of major crops in China for 2001–05 (Zhang et al 2007). Figure 3 shows that the determinant coefficient between fertilizer application rate and crop N uptake (R² in figure 2) generally decreased with fertilizer application rate, indicating the loosening of the link between fertilizer input and crop N uptake. The rapid decreasing in determinant coefficient when national average fertilizer N application rate was above 190 kg N ha⁻¹ probably indicates over fertilization at such rate (occurred in the 1990s). Similar phenomenon was revealed by Tian et al (2012) and as attributed to contributions of other factors such as genetic improvement to crop yield in addition to overuse of N fertilizer.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Crop N uptake from sources other than synthetic fertilizer applied to a site in a given year includes soil N, biologically-fixed N, atmospheric N deposition, and N in applied crop residues and manure. To distinguish soil N from other N sources, we regressed crop N uptake against total N input to croplands of different provinces in China for any given year (see figure 4 for selected years). Again, the slope of the regression line can be viewed as the fraction of total N input taken up by crops in the year, while the intercept of the line can be viewed as the average N uptake by crop from soils. The crop N uptake from soils increased continuously, from 8.0 kg N ha⁻¹ yr⁻¹ in 1980 to 46.4 kg N ha⁻¹ yr⁻¹ in 2010 (figures 2 and 4). This increase was obviously due to the residual effect of N input in previous years, especially synthetic fertilizer N since it dominated total N input.

Zoom In Zoom Out Reset image size

Several studies have shown substantial increases in the average organic carbon content of Chinese cropland soils since the 1980s (Xie et al 2007, Piao et al 2009, Yu et al 2009). By comparing the data resulting from the national soil survey conducted in 1979–82 and soil samples collected in 2007–08 across the country, we found that the average C/N ratio of cropland soils did not change significantly, so we are confident that soil N has not depleted. For large-scale croplands, in the long-term, if soil N has not depleted, the N taken up by crops must all have come from N inputs, including synthetic fertilizer N, N in applied manure and crop residue, biologically-fixed N, and atmospheric N deposition. Then, the ratio between total crop N uptake and total N input is the average RE of the N input. The ratio between crop N uptake and total N input ranged between 0.35–0.50, averaging 0.42 (table 1 for selected years), higher than the RE_N of fertilizer N obtained in field experiments (Zhu and Chen 2002, Zhang et al 2007).

Given that synthetic N is more easily available to crops than other N sources (Zhu 1998), the ratio of crop N uptake to total N input (blue line in figure 5) can be viewed as the minimum proportion of synthetic N that is eventually used by crops. The ratio of crop N uptake to synthetic fertilizer N input (red line in figure 5) can be viewed as the maximum proportion of synthetic N that is eventually used by crops in China through uptake in the year, environmental effects in the year, and accumulated residual effects. The real fraction of synthetic N used by crops is between the maximum and minimum, including the fraction of synthetic N that is used in the year and the residual effect. Since there was no net decrease in soil N, the N taken up by crops from soil can be viewed as the residual effect of N input in previous years. Here, we therefore propose calculating the average accumulative RE of synthetic N of a country/region as

$$\begin{eqnarray}&&\left( {\rm R}{{{\rm E}}_{N-{\rm in}-{\rm year}}}\times {\rm F}{{{\rm N}}_{{\rm rate}}}+{{N}_{{\rm soil}}}*{\rm F}{{{\rm N}}_{{\rm rate}}}/{\rm T}{{{\rm N}}_{{\rm rate}}} \right)/{\rm F}{{{\rm N}}_{{\rm rate}}},\end{eqnarray} \tag{ 5 }$$

which is simplified as

$$\begin{eqnarray}&&{\rm R}{{{\rm E}}_{N-{\rm in}-{\rm year}}}+{{N}_{{\rm soil}}}/{\rm T}{{{\rm N}}_{{\rm rate}}},\end{eqnarray} \tag{ 6 }$$

where RE_N-in-year is the fraction of synthetic N taken up by crops in the year (the slope of the linear line in figure 1), FN_rate is the average rate of synthetic N input (kg N ha⁻¹), TN_rate is the average rate of total N input (kg N ha⁻¹), and N_soil is the N uptake by crops from soil (the intercept of the linear line in figure 4; in kg N ha^–1). Following this, the accumulative recovery efficiencies of synthetic N in China were estimated to have ranged between 0.40 and 0.68 during 1980–2010, averaging 0.51 and being around 0.43 in recent years (green line in figure 5). These estimated accumulative recovery efficiencies were 10%–46% higher than their corresponding recovery efficiencies in annual terms (blue line in figure 2), due to the residual effect of N.

Zoom In Zoom Out Reset image size

At the end of the growing season, residual N from fertilizer can remain in the soil in the following forms: as mineral N, in roots, immobilized into the microbial biomass, or incorporated into other soil organic matter pools; and then become available for plant uptake during subsequent growing seasons (Dourado-Neto et al 2010). Although some ¹⁵N-isotope experiments have shown little residual effect of fertilizer N (Timmons and Cruse 1991, Xu et al 1993), this effect may depend on the amount of N input. Continuous high N input to croplands of China has led to large amounts of soil N accumulation. For example, by undertaking 269 on-farm demonstration trials in the North China Plain, Cui et al (2013) showed that NO₃-N in the top 90 cm of the soil profile averaged 191 kg N ha⁻¹ before the sowing of wheat, and 184 kg N ha⁻¹ before the sowing of maize. Therefore, testing of soil NO₃–N became an essential step in fertilizer recommendation in the North China Plain (Cui et al 2008, 2013). In the semiarid areas of northwestern China, Zhao et al (2013) conducted a ¹⁵N microplot experiment within long-term fertilization experiment plots and found that 7–15% of the ¹⁵N-labeled fertilizer was recovered in the two subsequent crop seasons.

The significant residual effect of N fertilizer in Chinese croplands is also reflected by the large differences in RE_N obtained from long-term and short-term experiments. Results of the China National Soil Fertility and Fertilizer Efficiency Long-term Monitoring Network showed that RE_N values of the maize–wheat system (uplands) had a range of 0.56–0.72,calculated from the difference between NPK and PK treatments (Liu et al 2010), while the RE_N of the wheat–rice system (paddy fields) was around 0.40 (Shen et al 2007, Zhao et al 2010). In a series (hundreds) of short-term field experiments during 2001–05, Zhang et al (2007) showed RE_N values of 0.283 for rice, 0.282 for wheat and 0.261 for maize. This increasing residual effect is also seen from the change in the yields of zero-N plots. For example, based on 7410 on-farm trials, Fan et al (2013) found that the yield of control plots (irrigated cereal-based cropping systems) was 0.73–1.76 Mg ha⁻¹ higher in the 2000s than in the 1980s, and this was primarily due to the enhancement of soil-related factors since the yield of control plots of long-term experiments did not change significantly.

In addition to the significant residual effect, the increased soil N content translates to 6.8% of the total N input to croplands during 1980 and 2007. This suggests that the use efficiency of synthetic N is much higher and the loss rate is less than generally perceived.

Nevertheless, there remain many uncertainties in our estimation. For example, biologically-fixed N is an important source of total N input, but it has a wide range of rates reported in the literature. Furthermore, N deposition, especially dry deposition, can contribute to uncertainties in N input due to the scarcity of measurements. Nevertheless, synthetic N fertilizer is by far the dominant source of N input of croplands and our analysis clearly shows the large residual effect of fertilizer N in the crop production systems of China.

Though the accumulative recovery efficiencies of synthetic N in China were estimated to have ranged between 0.40 and 0.68 during 1980–2010, decades of fertilizer N overuse have contributed to soil, water, and air pollution. For example, ammonia and nitrous oxide emissions dominated by agricultural soils in China are the major contributions of heavy wet deposition and global warming (Reis et al 2009, Liu et al 2013), and fertilizer runoff accounted for 22–57% of the total nonpoint N pollution in large river basins in China (Ti and Yan 2013). Experimental evidence indicated that cutting 30–60% N would reduce more than 50% N loss without significantly diminishing crop yields in some intensive cropping systems in China (Ju et al 2009). Therefore the residual effect of previously applied N, as identified in this study, should be given full consideration in determining N application rate.

In China, due to the continuously high N input to crop production systems, the fertilizer N RE in annual terms is indeed low (below 30% in recent years), but the residual effect of fertilizer N has been increasing continuously, to the point that 40–68% of applied fertilizer is taken up sooner or later. Therefore, while there is scope to reduce N fertilizer use in China, the residual effect of applied fertilizer must be accounted for in any such effort.

This research was funded by the Chinese Academy of Sciences (Grant No. KZCX2-YW-GJ01).