Reducing environmental risk by improving N management in intensive Chinese agricultural systems

Abstract

Excessive N fertilization in intensive agricultural areas of China has resulted in serious environmental problems because of atmospheric, soil, and water enrichment with reactive N of agricultural origin. This study examines grain yields and N loss pathways using a synthetic approach in 2 of the most intensive double-cropping systems in China: waterlogged rice/upland wheat in the Taihu region of east China versus irrigated wheat/rainfed maize on the North China Plain. When compared with knowledge-based optimum N fertilization with 30–60% N savings, we found that current agricultural N practices with 550–600 kg of N per hectare fertilizer annually do not significantly increase crop yields but do lead to about 2 times larger N losses to the environment. The higher N loss rates and lower N retention rates indicate little utilization of residual N by the succeeding crop in rice/wheat systems in comparison with wheat/maize systems. Periodic waterlogging of upland systems caused large N losses by denitrification in the Taihu region. Calcareous soils and concentrated summer rainfall resulted in ammonia volatilization (19% for wheat and 24% for maize) and nitrate leaching being the main N loss pathways in wheat/maize systems. More than 2-fold increases in atmospheric deposition and irrigation water N reflect heavy air and water pollution and these have become important N sources to agricultural ecosystems. A better N balance can be achieved without sacrificing crop yields but significantly reducing environmental risk by adopting optimum N fertilization techniques, controlling the primary N loss pathways, and improving the performance of the agricultural Extension Service.

The last 40 years have seen an extraordinary change in the global N cycle (1). As recently as the 1960s, N availability in most parts of the world was controlled by natural processes, but the expanded production of synthetic N fertilizer and release of N from fossil fuel combustion now matches the natural rate of formation of reactive N on a worldwide basis (2). The rate of change has been spectacular with half of the synthetic N fertilizer ever used having been applied during the last 15 to 20 years (3). Furthermore, because of the difficulty in accurately predicting N fertilizer requirements, rates exceeding plant requirements are often applied, thus inducing unintended environmental consequences such as leaching of nitrate and emission of nitrous oxide and ammonia. This has become a major concern for scientists, environmental groups, and agricultural policymakers worldwide.

From 1977 to 2005, total annual grain production in China increased from 283 to 484 million tons (a 71% increase) and the average grain production per unit area increased from 2,348 to 4,642 (a 98% increase). However, synthetic N fertilizer application increased from 7.07 to 26.21 million tons (a 271% increase) over the same period. This resulted in a partial factor productivity from applied N (PEP_N) that decreased from 55 to 20 kg/kg (4, 5). Large amounts of external N inputs with decreasing N use efficiency have contributed to severe environmental degradation since the 1990s (6, 7). For example, the annual application rate of synthetic N for conventional agricultural practices in east and southeast China as well as the North China Plain now ranges from 550 to 600 kg of N per hectare for typical double-cropping systems (5, 8).

The Taihu region and the North China Plain are 2 of the most intensive agricultural regions in China and the most economically developed areas (5, 9). Large inputs of synthetic N fertilizer, rapid development of intensive livestock production systems, and rapidly increasing consumption of fossil fuels have severely disturbed regional biogeochemical N cycling and led to a series of environmental problems including eutrophication of surface waters, nitrate pollution of groundwater, acid rain and soil acidification, greenhouse gas emissions, and other forms of air pollution. There have also been effects on human health and normal functioning of ecosystems (9, 10), details of which are given in SI Text.

Here, we provide the actual grain yields and total N losses associated with knowledge-based optimum fertilization strategies compared with experience-based N management practices from numerous on-farm field experiments. The ‘knowledge-based optimum N fertilization’ (11) hereafter refers to using large numbers of field-based academic research results or soil tests on quantity and timing of synthetic N fertilizer practically, using regional mean optimal N (RMON) application rate in rice/wheat systems and in-season N management based on the soil mineral N test in wheat/maize systems (5, 8, 12, 13). We use an integrated approach to measure the fate of ¹⁵N fertilizer and to document the major loss pathways in situ rather than by using individual measurements of loss processes reported in other studies to understand N behavior in these 2 rotations and thus attempt to control the environmental risk. We explain the reasons for the large external N inputs currently practiced in these intensive management systems and suggest strategies for balancing N management by applying knowledge-based optimum N fertilization techniques.

This study examines double-crop rotations, which are 2 of the most intensive agricultural systems worldwide and are widely practiced in Asian countries. The frequent alternation between flooding and draining the rice/wheat rotations led to significant changes in soil N transformations as compared with upland crop rotation. In addition, the extremely high external synthetic N inputs used in these two rotations during the last decade provided a unique opportunity to study the effects of N on environmental degradation. A detailed knowledge of N cycling and environmental impacts of these two intensive agricultural regions may also provide options for more rational N management practices in other intensive management cropping systems around the world.

Results and Discussion

We developed knowledge-based optimum N management techniques in rice/wheat systems in south China and wheat/maize system in north China during the last decade. The regional mean optimal N (RMON) application rate was calculated from the average of economically optimum N rates (i.e., the point of the N rate where the marginal grain production value is equal to the marginal N fertilization cost) based on large numbers of on-farm field experiments in the rice-based rotations and in-season N management based on the soil mineral N (Nmin: NH₄⁺-N + NO₃⁻-N) test previously developed for upland crop rotations. In principle, RMON recommends a regional rational N rate in combination with fine tuning according to specific field conditions, taking into consideration the relative uniformity of climate, soil N fertility, and absence of a satisfactory soil N test index for rice-based rotations (5, 12). RMON gives the mean of economically optimum N rates and can be used as a reference for extension technicians. This recommendation method can be easily adapted to rural areas of China where soil testing facilities are usually not available and the cropping index is high. The short time interval between harvest of the first crop and sowing of the second makes soil testing impractical.

The improved Nmin method is based on synchronization of crop N demand and soil Nmin supply, and is considered to be a soil N supply index because of the high amounts of Nmin (mostly nitrate) frequently found in the root zone and its high availability for subsequent crop growth in the next phase of upland rotations. We found that high potential rates of mineralization and nitrification contribute to this high accumulation of nitrate in soils of the North China Plain (14). This method has been well established in wheat/maize rotations in recent years (8, 13, 15, 16) and can be easily adapted to individual fields because the nitrate test has become easy to perform in the laboratory and may even be feasible in the field (17). In the present study, we also evaluated the agronomic performance and potential environmental risk of these knowledge-based strategies in comparison with the farmer's experience-based practices.

We conducted numerous on-farm field experiments (Field Study 1) with different N levels in Taihu region and the North China Plain from 2003 to 2006. With the sole exception of wheat in Taihu region, the optimum N rate as determined using the regional mean optimal N application rate in Taihu region and the improved Nmin method on the North China Plain appeared to give slight increases in crop yield compared with farmers' current practices (Table 1). However, no significant differences (P > 0.05) were found, indicating that similar high crop yields can be achieved with lower N fertilizer rates because of increased soil indigenous N supply (18, 19). In contrast, total N losses (including NH₃ volatilization, denitrification and leaching from the top 1 m of the soil profile) increased significantly with increasing N inputs, indicating high environmental costs were caused by exceeding optimum N fertilizer rates.

Table 1.

Average grain yields and total N losses of the optimum N fertilization (ON) compared with farmers' N practices (FN) (Field Study 1 and 2).

Crop and site of field experiment	N fertilization	N rate		Grain yield		Total fertilizer N loss*
		Rate, kg of N per hectare	Ratio of FN to ON	Yield, kg·ha−1	Ratio of FN to ON	Total loss, kg of N per hectare	Ratio of FN to ON
Rice in Taihu (n = 26)	ON†	200		8,270		102
	FN	300	1.5	8,012	0.97	174	1.7
Wheat in Taihu (n = 9)	ON†	153		3,700		76
	FN	250	1.6	4,084	1.10	155	2.0
Wheat in NCP‡ (n = 121)	ON§	128		6,024		25
	FN	325	2.5	5,764	0.96	71	2.8
Maize in NCP‡ (n = 148)	ON§	158		8,900		52
	FN	263	1.7	8,500	0.95	108	2.1

*Total fertilizer N losses calculated with the models of Fig. 1B simulated from ¹⁵N field experiments.

^†Regional mean optimal N application rate calculated from the mean of economically optimum N rates of field experiments in Taihu regain (5, 12).

^‡Data including Field Study 1 and also summarized from ref. (15, 16); NCP, North China Plain.

^§In-season nitrogen management based on soil Nmin test on the NCP (8, 13, 15, 16).

We used ¹⁵N in some of the field experiments (Field Study 2) to quantify differences in fertilizer N fate at different application rates among the four crops. We found that N recovery rates decreased with increasing N application rate for all of them (Fig. 1A). N rates showed polynomial, linear, polynomial, and logarithmic relationships with N recovery by rice, wheat (south), wheat (north) and maize, respectively. Each fitting produced a highly significant model (P < 0.01). At lower N input levels (<125 kg of N per hectare), the order was: wheat (north) > maize > wheat (south) > rice. At higher N input levels (>125 kg of N per hectare) the order was: wheat (north) > rice > maize > wheat (south), indicating that rice recovered more N at high N application rates. N loss rates increased with increasing N application rate in all four crops (Fig. 1B). A logarithmic model satisfactorily simulated the correlations between N rate and N loss rates for the four crops (P < 0.01). Rice and wheat (south) in the Taihu region showed much higher N loss rates than did wheat (north) and maize on the North China Plain. The trends of the N retention rates with increasing N rate differed among the four crops. The rice and wheat systems in the Taihu region showed lower N retention rates than wheat and maize systems on the North China Plain. Moreover, N retention rates in wheat (south) and wheat (north) decreased first and then increased with further increase in N rate, showing typical polynomial relationships (P < 0.01). However, N retention rates in rice and maize decreased by power function relationships when N rate increased (P < 0.01). These findings indicate that the upland wheat and maize rotation system has significantly higher N recovery rates, higher N retention rates, and lower N loss rates than the waterlogged rice and upland wheat rotation in the Taihu region under current soil N fertility and farmers' management practices. Moreover, on the North China Plain wheat has significantly higher N recovery rates, higher N retention rates, and lower N loss rates than maize, and in the Taihu region, rice has significantly higher N recovery rates and lower N loss rates than wheat. Rice/wheat rotations show high ‘leakage’ of external N inputs with resulting pollution threats to water and air. However, wheat/maize rotations show strong retention of external N inputs that can remain available for use by subsequent crops. Retention of N is favorable for long-term improvement of soil N fertility and further emphasizes the need to reduce N application rates. It should be noted that N rates applied to wheat in the Taihu region are similarly high to those on the North China Plain, but wheat grain yields are much lower than in the north and show much lower N recovery rates (Table 1). One explanation is that farmers in the Taihu region try to compensate for the unfavorable growing conditions for wheat (e.g., extremely wet conditions in the root zone due to high rainfall and groundwater) by increasing N fertilizer application rates (20).

Fig. 1.

Relationships of N recovery rate (A), loss rate (B), and retention rate (C) with N application rate in 4 crops (Field Study 2). Vertical bars denote standard deviation of the mean (average of 6 ¹⁵N field experiments). Each model fitting produced a highly significant model (P < 0.01).

To further understand the different pathways of N loss from current N practices and to control them in the future, we conducted a lysimeter study (Lysimeter Study) in the Taihu region and a 4-year field experiment (Field Study 3) on the North China Plain, measuring in situ NH₃ volatilization, denitrification, N leaching, and N₂O emissions. We found that these N loss pathways differed sharply among the four crops because of differences in climate, soils and management practices (Table 2). Ammonia volatilization seems to be the main loss pathway for fertilizer N, with losses accounting for 19.4 and 24.7% of applied N in wheat (north) and maize seasons, respectively. Volatilization was much lower for rice and wheat (south), with losses accounting for 11.6 and 2.1% of applied N, respectively. Denitrification losses were quite small in the wheat and maize on the North China Plain (0.1 and 3.3% of applied N). In contrast, denitrification was the primary pathway for fertilizer N loss in rice and wheat within the Taihu region, accounting for 36.4 and 43.5% of applied N, respectively. Leaching accounted for only 0.3% of applied N from rice and 3.4% from wheat (south) season, but accounted for 2.7 and 12.1% of applied N for wheat and maize growing seasons on the North China Plain. We may be overestimating the denitrification loss in rice/wheat systems due to error accumulation with the different methods (5), but the order of magnitude agrees with previous studies (5).

Table 2.

Different N loss pathways expressed as a percentage (mean ± SD) of N application rate in farmers' N practices (Field Study 3, Lysimeter Study)

Component		Taihu region		North China Plain
		Rice	Wheat-south	Wheat-north	Maize
N rate (kg of N per hectare)		300	250	325	263
Recovery rate (%)*		29.6 ± 4.9	18.4 ± 6.3	31.0 ± 3.6	25.5 ± 5.2
Retention rate (%)*		21.7 ± 5.1	28.5 ± 4.6	45.7 ± 5.4	33.9 ± 2.3
Loss pathway	NH3 volatilization (%)	11.6 ± 4.7	2.1 ± 1.4	19.4 ± 5.2	24.7 ± 5.6
	Leaching out of 1 m soil depth (%)	0.3 ± 0.5	3.4 ± 2.1	2.7 ± 2.6	12.1 ± 8.5
	Denitrification (%)	36.4†	43.5†	0.1 ± 0.04	3.3 ± 1.6

*Measured from corresponding ¹⁵N field experiments.

^†Calculated by difference method.

Alternating upland wheat and waterlogged rice led to high denitrification of applied N and accumulated nitrate (after wheat harvest) in the rice season (20–22). The traditional practice of surface application of urea-N stimulated NH₃ volatilization from rice (23, 24). After plowing of paddy soil, a hard layer that formed below the plow layer due to clay deposition appeared to reduce leakage from the rice crop, but resulted in higher leaching losses during the wheat season when drainage prevails (25). The high denitrification loss for wheat (south) is attributed to wet soil conditions, higher temperatures and relatively high soil carbon (C) content in the Taihu region (26, 27). During the wheat season in the Taihu region, N fertilizer (urea) is readily nitrified and accumulates in the soil, consequently becoming available for loss through denitrification during the rice season (22). Thus the residual NO₃⁻-N in the soil profile is not fully available to the subsequent crop as is the case in upland crop rotations. We speculate that the very large N losses (corresponding with lower N retention rates) and large C inputs in straw under the anaerobic conditions in rice/wheat rotations may be another explanation for the structural changes in soil organic matter and increased C-rich humic acids found by Schmidt-Rohr et al. (28). On the North China Plain, with its calcareous soils, typical pH 7.5 to 8.5 and predominant use of urea or ammonium bicarbonate as fertilizer (accounting for 80%), NH₃ volatilization is generally considered to be the major pathway of N loss, especially in the hot maize growing season (29–31). Due to lower soil water content and soil organic matter (1.0–1.5%), most experimental results on the North China Plain have shown the magnitude of N loss by denitrification is no more than 10 kg of N per hectare and that denitrification accounts for less than 5% of the fertilizer N, most of which occurs during the hot and humid summer season (30, 32). We found that nitrate accumulation in the soil profile and gradual leaching below the root zone were the important N loss pathways in wheat and maize rotations under conventional agricultural N management practices due to concentrated rainfall in summer and flood irrigation (8, 33), contradicted conventional thinking that leaching losses are not an important pathway for N loss in semi-humid upland agricultural systems of the North China Plain.

Field Study 1 and the Lysimeter Study also showed that NH₃ volatilization increased rapidly and represented the major pathway for loss in response to increasing N rate, but N₂O emission and leaching did not increase dramatically with increasing N rate during the rice growing season (Fig. S1, rice). Leaching did increase rapidly and accounted for most of the reactive N loss with increasing N fertilizer rate, while N₂O emissions and NH₃ volatilization increased very little in response to N rate during the wheat season in the Taihu region (Fig. S1, wheat-south). Ammonia volatilization and NO₃-N leaching increased significantly with increasing N rate and N₂O emissions also showed an upward trend in the wheat growing season on the North China Plain (Fig. S1, wheat-north). During the maize growing season, NH₃ volatilization, NO₃-N leaching, and N₂O emissions all increased significantly, with N leaching showing a particularly strong increase (Fig. S1, maize). These results confirm that emissions of reactive N to the environment will increase after N fertilization exceeds the optimum rate. However, we did not find large N₂O emissions among the four crops. The total N₂O emissions in rice, wheat (south), wheat (north), and maize with different N application rates were 0.13 to 0.50, 0.27 to 0.63, 0.13 to 0.24, and 0.16 to 0.65 kg of N per hectare, respectively. Expressed as average emission factors, this accounted for 0.20, 0.30, 0.12, and 0.23%, respectively, close the IPCC 2006 guideline of 0.3% for paddy rice soil and much lower than the guideline of 1% for upland soil (27, 34, 35). We explain these results by the low ratio of N₂O to N₂O+N₂ from denitrification products (36, 37) under waterlogged rice conditions, low available carbon sources and low water-filled pore space (WFPS) for denitrification in most semi-humid upland soils on the North China Plain (14).

To quantify atmospheric and irrigation water contributions (hereafter ‘environmental N inputs’) to agricultural systems in the Taihu region and the North China Plain, we measured deposition and N inputs from irrigation water (Monitoring Study). Total environmental N inputs reached about 89 kg of N per hectare in the Taihu region and about 104 kg of N per hectare on the North China Plain. Comparisons with observations made from both regions in the 1980s shows more than 2-fold greater environmental N inputs with time. N deposition increased from 15 to 33 kg of N per hectare and irrigation water N inputs from 15 to 56 kg of N per hectare in Taihu region (Fig. S2) (38). During the same period, N deposition increased significantly from 21 to 89 kg of N per hectare and irrigation water N inputs from 8 to 15 kg of N per hectare (Fig. S2) (38, 39) on the North China Plain. Inorganic N and soluble organic N are the primary environmental N sources that can be used directly by crops or after transformation in the soil (40). The absence of crop yield response to N application in most field experiments since the 1990s (5, 8, 15, 16, 20) is not only attributable to increased indigenous soil N supply, but also to high environmental N inputs (41, 42). Moreover, high N deposition may also contribute to surface water eutrophication, acid rain, and soil acidification (43) in these regions. High total N concentrations (7 to 8 mg·N·L⁻¹) in the irrigation surface water in the Taihu region and high NO₃⁻-N concentrations in the irrigation shallow groundwater on the North China Plain also reflect the heavy N pollution of water resources in both areas.

We calculated annual N balance for both rotations using two scenarios: conventional N practice versus optimum N fertilization (Tables S1 and S2). We computed an annual N surplus of 87 kg of N per hectare for current practices with large losses by denitrification for the rice/wheat system. Synthetic N fertilizer was the primary input followed by N in irrigation water, biological N₂ fixation, and N deposition. A better N balance can be achieved by adopting optimum N fertilization strategies designed to maintain relatively high yields but reduce environmental risk. However, dinitrification could be further reduced by improving carbon management and controlling the water regime (25–27). We also computed a 212 kg of N per hectare surplus for current practice with large losses by NH₃ volatilization within the wheat/maize system. A large proportion of surplus N accumulated as nitrate in the soil profile after harvest (33), and partly existed as N in organic form due to manure application. However, the unusually large quantity of annual surplus N might also be caused by underestimating the leaching loss due to drought conditions in our observation years with 24–46% reduction in rainfall (8). Previous studies showed that strong N leaching losses only occurred in some years with heavy summer rainfall (33), leading to high nitrate accumulation in the deep subsoil (8, 33) and the groundwater (43). In the future, a slightly negative N balance could be achieved using an optimum N rate of about 286 kg of N per hectare to maintain relatively high yields (15, 16). The slightly negative balance would be conducive to making sure the plants fully use accumulated nitrate and further reduce nitrate leaching (8, 33). A high N retention rate is conducive to maintenance of soil N fertility. Changing the N application techniques and using deep placement of urea or ammonium bicarbonate could substantially reduce NH₃ volatilization losses from calcareous soils (29).

A question that needs to be addressed is why farmers in China continue to use such excessively high rates of N fertilizer. Before the 1990s, scientists, government and extension staff encouraged farmers to increase synthetic fertilizer inputs to increase yields and feed an increasing population. As a result, synthetic fertilizer inputs increased continually throughout the country, especially in the Taihu region and North China Plain. By 2000, the rates were far greater than crop demand and serious environmental degradation had begun. Although this trend has been recognized by the scientific community since the late 1990s, on-farm practices are difficult to reverse after 10 to 15 years of effort (43). Persuading farmers to limit fertilizer inputs is difficult because many of them still hold to now traditional opinions that higher crop yield will be obtained with more fertilizer. Application rates of N therefore often include an extra ‘insurance’ component to prevent yield loss rather than matching inputs to crop demand. The high off-farm incomes and relatively low retail prices for N fertilizers (with government subsidies for production and transportation) compared with U.S. and European prices are also important factors (43, 44). A third major reason is the poor infrastructure of the Extension Services and poor transfer of knowledge to farmers (43). A critical objective of the optimum N techniques developed using our studies are to substantially lower fertilizer N application rates by accounting for indigenous soil N and environmental N inputs for maintenance of the yields needed to feed an increasing population. This strategy also limits losses of total N to the water and atmosphere with the aim of establishing sustainable agricultural systems (8, 11, 12).

Crop production in intensive agricultural practice currently relies too heavily on synthetic N fertilizer inputs and cereal monocropping, especially in Asian countries under pressure to feed large and growing populations. It is now time to change this situation by balancing yield and environmental consequences. Integrated management packages need to be developed for major cropping systems that include efficient recycling of manures and crop residues, the use of legume crops in rotations to increase internal N cycling and further reduction in the reliance on synthetic N fertilizers (45).

Conclusions

It took nearly 50 years to achieve food sufficiency in China. An unanticipated cost has been that massive fertilizer inputs have led to significant environmental degradation (6). Over-fertilization is a serious problem in intensive agricultural production areas in China, resulting in enrichment of reactive N in the air, soil and water with consequent impairment of ecosystem services. Our studies show that more efficient use of N fertilizer can allow current N application rates to be reduced by 30 to 60%. This would still maintain crop yields and N balance in rotations, while substantially reducing N losses to the environment. The over-application of N also represents an unnecessary economic expenditure for farmers. The new recommendations should fully take into account the N supplying capacity of the soil and N deposited from air and irrigation water. The characteristics of N behavior among the 4 crops were sharply different depending on climatic, soil and management practices. This must be taken into consideration to further reduce N losses. Only by reducing fertilizer N inputs can degraded environments be gradually restored, enhanced and protected. Although several environmental standards have been set in the past, there are still no legislative controls in China equivalent to those in the European Union (33). China would benefit from adopting and enforcing relevant agricultural regulations (7). All these goals could be achieved by removing government subsidies, introducing an N fertilizer tax, improving local Extension Services, educating farmers for environmental awareness, and employing practices that avoid serious environmental degradation (43).

Materials and Methods

Study Areas.

Two different representative intensive agricultural regions were selected as study areas: (1) the Taihu region in east China (30–32 °N, 119–122 °E) in the alluvial delta of the Yangtze River and (2) the North China Plain in northeast China (32–41 °N, 113–120 °E) in the alluvial plain of the Yellow River (Fig. S3). Details of the climate, soils, and crops are given in SI Text.

Field Study 1.

In the Taihu region, 26 rice (8 or 9 locations each year) and 9 wheat (3 locations each year) on-farm field experiments with 7 (rice) or 5 (wheat) N rates were conducted from 2003 to 2006. The 7 rice N treatments were 0, 100, 150, 200, 250, 300, and 350 kg of N per hectare and the 5 wheat N treatments were 0, 100, 150, 200, and 250 kg of N per hectare. The plots (42 m² in area, 7 × 6 m) were arranged in a randomized complete block experimental design with 4 replicates.

On the North China Plain 6 winter wheat and 6 summer maize on-farm field experiments with 5 (winter wheat) or 6 (summer maize) N rates were conducted from October 2003 to October 2005 at Dongbeiwang near Beijing, Huimin County in Shandong province, and Baoding County in Hebei province. The 5 winter wheat N treatments were 0, 75, 150, 225, and 300 kg of N per hectare and the 6 summer maize N treatments were 0, 40, 80, 120, 160, and 240 kg of N per hectare. The plot size was 63 m² (9 × 7 m). The plots were arranged in a randomized complete block experimental design with 3 replicates. The fertilizer types and application times are given in the SI Text. Except for fertilizer application and grain harvest, each experimental field was managed using the individual farmer's current management practices in both rotations. Aboveground biomass was measured by hand in all plots. The central area of 16 m² (4 × 4 m) in each plot was harvested to determine grain dry matter yield.

Field Study 2.

In the Taihu region, 6 rice (2 locations per year) and 6 wheat (2 locations per year) field experiments were selected to conduct ¹⁵N studies in microplots to determine the fate of N fertilizer at different N levels. On the North China Plain, the ¹⁵N study was carried out in microplots in the on-farm field experiments (SI Text). In both rotations plant and soil sampling and analysis for total N and ¹⁵N abundance are described as in (44). The N recovery rate is expressed as the percentage of applied ¹⁵N fertilizer taken up by the aboveground plant parts and the N retention rate as the percentage of applied ¹⁵N fertilizer recovered in the top 100 cm of the soil profile. The loss rate was calculated by subtracting the recovery rate and retention rate from 100. The mean and standard variation was calculated across all of the experimental years and sites in same crop species.

Lysimeter Study.

Twenty-four lysimeters containing undisturbed soil profiles were used at the research station at Changshu, Jiangsu province (31° 31.93′ N, 120° 41.88′ E) in the Taihu region. They contained a Typic Epiaquept (46) formed on alluvial loess with a silty clay loam (46) texture. A steel cylinder 1-m deep with 1.14-m inner diameter was pushed into the soil, cut at the base, and then removed. After collection, the lysimeters were prepared for free drainage by replacing approximately 0.08 m of soil at the base of each column with a nylon mesh, a gravel layer, and a porous plastic sheet (47). The lysimeters were then placed in plots permanently installed belowground at the station. Each lysimeter was used to grow a waterlogged summer rice/upland winter wheat rotation from October 2003 to October 2005 with 2 rice and 2 wheat growing seasons. Details of treatments are given in SI Text. Water leaching through the soil columns was collected in plastic containers placed in an underground measuring station. The leachate volume was measured at 3-day intervals and subsamples were taken from the accumulated leachate and stored at −18 °C before chemical analysis. During the winter wheat growing seasons, leachate could be collected only during rainfall periods. Leachate NO₃⁻, NH₄⁺, and total N concentrations were determined using a continuous flow analyzer (TRAACS 2000, Bran and Luebbe Inc.). The amount of N leached at each collection time was calculated by multiplying the volume and the N concentration of leachate. The total amount of N leached over the whole growing period was obtained by summing the individual amounts of N leached at the different collection times. The N leached from fertilizer was calculated as the difference in leached N between fertilized treatments and unfertilized controls. The mean and standard variation of each N treatment was calculated across 3 replicates over 2 years.

NH₃ volatilization and N₂O measurements were conducted in both rice and both wheat growing seasons from October 2003 to October 2005. Ammonia volatilization rate was measured with a continuous airflow enclosure method (48). The total amount of NH₃ volatilized over the whole growing period was obtained by summing NH₃ volatilization measured each day. N₂O fluxes were usually measured once a week throughout each crop growing period with static closed chambers, and measurements were made more frequently after basal fertilization and topdressing. Detail measurements and calculations are described in (49). Similar NH₃-volatilization and N₂O measurements at different N levels were conducted in the on-farm field experiment at Baoding County in Hebei province (Field Study 1). The amount of fertilizer derived NH₃ volatilization or N₂O emission was calculated as the difference between volatized NH₃ or N₂O in fertilized and unfertilized treatments. The mean and standard variation of each N treatment was calculated across 3 replicates over 2 years.

Field Study 3.

A field experiment was conducted from September 2000 to September 2004 at Dongbeiwang research station near Beijing (40.08 °N, 116.28 °E) on a calcareous alluvial soil [Fluvaquents (46)] considered typical of the North China Plain. Eight successive crops (4 winter wheat and 4 summer maize) were grown on the same plots over the 4-year period. Detailed descriptions of the field experiment were reported previously (8).

¹⁵N studies in microplots were also conducted in the conventional N and optimized N fertilization treatments from September 2002 to September 2004. NH₃ volatilization was measured from September 2002 to September 2004 with an automatic wind tunnel system that covered an experimental soil area of 0.7 × 1.5 m (1.05 m²) (50). Field denitrification was monitored by a soil core incubation system using the C₂H₂-inhibition method (51) during the 2-year study from September 2002 to September 2004 (32). Nitrate leaching was quantified using TerrAquat passive samplers filled with ion-exchange resin. These were used to quantify the total annual amount of leached nitrate below the rooting zone with ten replicates per plot (52). The mean and standard variation of NH₃, N₂O, and nitrate leaching were calculated across 4 replicates over 2 years.

Monitoring Study.

Measurements of wet and dry deposition were conducted from October 2003 to October 2005 in the 3 representative locations of the Taihu region, that is, Changshu, Nanjing, and Hangzhou. Three automatic wet-only samplers (APS series, Wuhan Tianhong Inc.), which collected precipitation samples only during precipitation events (controlled by sensors) were separately installed at the above 3 sites for the collection of wet-only deposition (53). Particulate dry N samples were collected using a 0.5-m² polyethylene sheet-based sampler (41). Before sunrise (about 6:00 AM in winter and around 5:00 AM in summer), 250 mL of 2 M KCl solution was used to collect deposited dew and the particles in the collector. In the present experiment, the N in dry deposition is defined as only the part of the inorganic N adsorbed to particles and exchangeable with 2 M KCl.

Precipitation samples were collected from 11 monitoring sites located on the North China Plain during the period 2003–2006. Bulk deposition (wet deposition plus sedimenting dry deposition such as dust during dry periods) was collected (41, 53). The mean of the bulk deposition was calculated across all of the sites over the measuring years. Total deposition was determined by the ITNI system at Dongbeiwang, Beijing and Wuqiao and Quzhou in Hebei province. The ITNI system is based on ¹⁵N isotope dilution and allows the inclusion of direct N uptake by the plant as well as the absorption of gaseous N compounds by the soil-plant system (40, 54, 55).

The N inputs from irrigation were calculated from the sum of the amounts of water in the periods of irrigation and the corresponding concentrations of ammonium and nitrate in the irrigation water. The measurements were conducted from October 2003 to October 2005 at 12 locations in the Taihu region and 10 locations on the North China Plain. The mean value was calculated across all of the years and sites.

Statistical Analysis.

Most of the data are presented as mean ± SD (standard deviation) in the Tables and Figures. All data were analyzed using the SAS software package (SAS Institute). Means were compared using LSD (Least Significant Difference) by one-way ANOVA analysis. Regression analysis of N fertilizer rate vs. recovery rate, retention rate, and loss rate was performed using the PROC REG procedure.