*2.2. Experimental Design*

Four treatments were applied during the tomato and cucumber growing seasons following the local experts' recommendations: control (CK: 0 kg N ha−<sup>1</sup> ); 100% chemical N reduction fertilization, using a single application of organic fertilizer in the form of manure (M: 362 kg N ha−<sup>1</sup> ); reduced conventional fertilization (RCF: chemical N 992 kg ha−<sup>1</sup> + organic kg N 362 ha−<sup>1</sup> ), in which the chemical fertilizer N contents were reduced by, 38.7% and 28.8% on average, for tomatoes and cucumbers, respectively, relative to the CF treatment; conventional fertilization (CF: chemical N 1515 kg ha−<sup>1</sup> + organic N 362 kg ha−<sup>1</sup> ). Phosphate (triple superphosphate, 46% P2O5) and manure fertilizers (chicken manure, with average N, P2O5, and K2O contents of 1.00%, 1.32%, and 1.88%, respectively) were applied as base fertilizers before each tomato and cucumber season. Urea (46% N) and potassium sulfate (50% K2O) fertilizers were split into base and topdressing applications and were hand broadcasted. The vegetables were transplanted to seedbeds (600 cm long × 130 cm wide) with plant spacing of 40 cm and row spacing of 75 cm for tomatoes, and plant spacing of 30 cm and row spacing of 70 cm for cucumbers. There were three replicates for each treatment and were arranged according to randomized complete block design (RCBD). The greenhouse was covered with plastic sheet. The light of Ningxia greenhouse is sunlight without any artificial light. The indoor temperature was controlled at 14–18 ◦C at night and 25–30 ◦C during the day by sunlight, covering quilt and uncovering shed film. Weeding and pesticide application were according to the local conventional practices. Yellow River water and groundwater were used for irrigation.

#### *2.3. Evaluation of N Use Efficiency*

Tomato and cucumber samples were collected, and yields were measured for each harvest. The Kjeldahl method was used to analyze N content, as described by Yang et al. [15]. The following indicators were calculated, using data collected over the entire study, to evaluate NUE: apparent recovery efficiency of applied N (*REN*, %), agronomic effectiveness of applied N (*AEN*, kg kg−<sup>1</sup> ), physiological efficiency of applied N (*PEN*, kg kg−<sup>1</sup> ), and partial factor productivity of applied N (*PFPN*, kg kg−<sup>1</sup> ) [29]:

$$RE\_N = \frac{TUI\_N - T\_{CK}}{F\_N}$$

$$AE\_N = \frac{Y\_N - Y\_{CK}}{F\_N}$$

$$PE\_N = \frac{Y\_N - Y\_{CK}}{TUI\_N - T\_{CK}}$$

$$PFP\_N = \frac{Y\_N}{F\_N}$$

where *TU<sup>N</sup>* is TN uptake, *TCK* is TN uptake without N application, *F<sup>N</sup>* is applied fertilizer N, *Y<sup>N</sup>* is annual tomato/cucumber (tomato followed by cucumber) yield, and *YCK* is the tomato/cucumber yield without N application (all expressed in kg ha−<sup>1</sup> ). The relationships between these parameters and annual N application rate were examined using exponential, linear, logarithmic, and power functions.

#### *2.4. Leachate Collection and Measurement of Nitrate Losses Due to Leaching*

Using a leachate collection device as described by Zhao [23], leachate was collected 3 d after each irrigation. The leachate collection device and containers were cleaned before use. The samples were stored at −20 ◦C in a refrigerator, and the TN content of the leachate was analyzed using the alkaline potassium persulfate digestion UV spectrophotometric method [23]. Approximate N input and leaching were calculated as follows:

$$NL\_{\text{CC}}\left(\text{kg ha}^{-1}\text{ year}^{-1}\right) = NL\_{\text{CT}} - NL\_{\text{CO}}$$

$$NLIR\_{TN}\left(\%\right) = \frac{NI\_{TN}}{NL\_{TN}} \times 100\%$$

$$NLIR\_{ON}\left(\%\right) = \frac{NI\_{ON}}{NL\_{ON}} \times 100\%$$

$$NLIR\_{CN}\left(\%\right) = \frac{NI\_{CN}}{NL\_{CN}} \times 100\%$$

where *NLCC* is N leaching caused by input of chemical fertilizer N, *NLCT* is N leaching caused by TN input, *NLCO* is N leaching caused by organic N input, *NLIRTN* is the TN leaching-to-input ratio, *NLIRON* is the organic N leaching-to-input ratio, *NLIRCN* is the chemical N leaching-to-input ratio, *NITN* is TN input, *NION* is organic N input, *NICN* is chemical N input, *NLTN* is TN leached, *NLON* is organic N leached, and *NLCN* is the amount of chemical fertilizer N leached.

#### *2.5. Economic Analysis*

The N fertilizer economic benefit (*NEB*, in USD t−<sup>1</sup> N) and the input–output ratio, were used to calculate the economic benefits of N reduction.

$$NEB = \frac{B\_N - B\_{CK}}{TF\_I}$$

$$Input-to-output\ ratio = \frac{T\_I}{T\_E}$$

where *B<sup>N</sup>* benefits from N *input* (USD ha−<sup>1</sup> year−<sup>1</sup> ), *BCK* is benefited without N input, *TF<sup>I</sup>* is total fertilizer N input (t ha−<sup>1</sup> year−<sup>1</sup> ), *T<sup>I</sup>* is total income (USD ha−<sup>1</sup> year−<sup>1</sup> ), and *T<sup>E</sup>* is total expenditure (USD ha−<sup>1</sup> year−<sup>1</sup> ).

#### *2.6. Statistical Analysis*

Descriptive data analysis was conducted, and graphs were created, using Microsoft Excel 2013 (Microsoft Corp., Redmond, WA, USA). The results are expressed as means (with standard error, SE) of the three replicates. We estimated fruit yield, NUE, N leaching, and NEB. One-way ANOVA with Duncan multiple comparison test was used to assess differences among the treatments. All statistical analyses were performed using SPSS 19.0 (IBM Corporation, Armonk, NY, USA). *p*-values < 0.05 were considered statistically significant.

#### **3. Results**

#### *3.1. Tomato and Cucumber Yield*

The individual yield of tomato and cucumber, as well as the combined yield of both crops (Figure 1a–c), differed significantly (*p* < 0.01) between years, treatments, and with the interaction of year and treatment. The average annual yield for the treatments relative to the CF is shown in Figure 1d–f.

For the tomato rotations, there was no significant difference between the four treatments in the first year (2008). The yield of the control treatment was significantly lower in 2009 by 16.7% relative to the CF treatment, and the yield gap increased annually from 2009 to 2012. The yield of the M treatment did not decline until 2010, when it declined by 5.1%,

relative to the CF treatment. There was no significant difference in yield between the CF and RCF treatments.

**Figure 1.** Fruit yield of (**a**) tomato, (**b**) cucumber, and (**c**) tomato/cucumber rotation. Average annual yield (**d**–**f**) in the control (CK), manure (M), and reduced conventional fertilization (RCF), relative to conventional fertilization (CF) for 2008–2013. The error bars indicate the standard deviations. CK: 0 kg ha−<sup>1</sup> N; M: 362 kg ha−<sup>1</sup> organic N; RCF: 992 chemical N + 362 kg ha−<sup>1</sup> organic N; CF: 1515 kg ha−<sup>1</sup> chemical N + 362 kg ha−<sup>1</sup> organic N. The lowercase letters indicate the significant difference among the means each year, while each *p* value given upside right indicate the two-way ANOVA with year and treatment interaction.

Similarly, for cucumber rotations, there was no significant difference between treatments in the first year (2008). The yields of the CK and M treatments were significantly lower in 2009 (by 24.6% and 11.5%, respectively) relative to the CF treatment. There were no significant differences in yields between the CF and RCF treatments throughout the experimental period.

For the first year of tomato/cucumber rotation, there was no significant difference in yield between the four groups. Significant differences occurred from the second year (2009) to the end of the experiment. Relative to the CF treatment, the yield was reduced in the CK and M treatments and the yield gap increased annually. On average, the fruit yield of the M treatment was significantly higher (by 8.2%) than CK but was significantly lower than CF (by 28.2%) and RCF (by 28.4%) treatments. Moreover, there was no significant difference between the CF and RCF treatments.

#### *3.2. Nitrogen Use Efficiency*

REN, AEN, PEN, and PFP<sup>N</sup> (Figure 2a–d) differed significantly (*p* < 0.01) between years and treatments. The REN, AEN, and PFP<sup>N</sup> of M showed a decreasing trend over the years. The REN, AEN, and PE<sup>N</sup> showed an upward trend to a relatively stable level year by year. By the end of this experiment, RE<sup>N</sup> was 11.0, 9.0, and 5.7%, AE<sup>N</sup> was 66.7, 43.2, and 18.5 kg kg−<sup>1</sup> , and PE<sup>N</sup> was 60.7, 43.2, and 31.6 kg kg−<sup>1</sup> in high to low order of RCF, CF and M, respectively. PFP was highest for M (17.9 kg kg−<sup>1</sup> ), followed by RCF (11.1 kg kg−<sup>1</sup> ) and CF (7.2 kg kg−<sup>1</sup> ).

**Figure 2.** N use efficiency for 2008–2013. (**a**) Recovery efficiency (REN); (**b**) agronomic effectiveness. (AEN); (**c**) physiological efficiency (PEN); (**d**) partial factor productivity (PFPN). Error bars reflect the standard deviations. Blue: M, 362 kg ha−<sup>1</sup> organic N. Red: RCF, 992 kg ha−<sup>1</sup> chemical N + 362 kg ha−<sup>1</sup> organic N; Green: CF, 1515 kg ha−<sup>1</sup> chemical N + 362 kg ha−<sup>1</sup> organic N. The lines indicate the annual general tendency.

#### *3.3. N Leaching*

The N leaching data are shown in Figure 3. For all rotations, the amount of leached N is the sum of the leaching results measured during all four irrigation cycles throughout the tomato production period. Because of a shortage of water from the Yellow River in 2008, no irrigation was applied during the fallow period in that year. From 2009 to 2012, TN leaching showed an annual cycle for the CF and RCF treatments: it was high during the tomato stage, lowest in the fallow stage, and highest in the cucumber stage, each year. The average annual TN leaching rates from high to low were CF (170.7 kg N ha−<sup>1</sup> ), RCF (130.2 kg N ha−<sup>1</sup> , 23.7% lower), M (92.0 kg N ha−<sup>1</sup> , 46.1% lower), and CK (69.0 kg N ha−<sup>1</sup> , 59.6% lower), respectively. For the CK, CF, RCF, and M treatments, the average annual leaching rates for the tomato stages were 28.5, 58.3, 45.4, and 35.7 kg N ha−<sup>1</sup> , accounting for 41.2%, 34.1%, 34.8%, and 38.8%, respectively, of the applied N; for the fallow period, they were 15.8, 26.9, 21.9, and 19.4 kg N ha−<sup>1</sup> , accounting for 22.9%, 15.7%, 16.8%, and 21.1%, respectively; for the cucumber stages, they were 24.8, 85.6, 63.0, and 39.6 kg N ha−<sup>1</sup> , accounting for 35.8%, 50.1%, 48.3%, and 40.1%, respectively.

**Figure 3.** N leaching for 2008–2013. (**a**) Relationship between the sampling period and total nitrogen (TN) leaching. (**b**) TN leaching by treatment and vegetable rotation. T: tomato; C: cucumber; F: fallow. CK: 0 kg ha−<sup>1</sup> N; CF: 1515 kg ha−<sup>1</sup> chemical fertilizer N + 362 kg ha−<sup>1</sup> organic N; RCF: 992 kg ha−<sup>1</sup> chemical fertilizer N + 362 kg ha−<sup>1</sup> organic N; M: 362 kg ha−<sup>1</sup> organic N.

#### *3.4. Economic Analysis*

Total expenditure, total income, and TN input are shown in Table 1. The costs for field management were USD 98, 210, 210, and 112, for the CK, CF, RCF, and M treatments, respectively, based on USD 14 d−<sup>1</sup> for labor times of 7, 15, 15, and 8 d, respectively. Each year, 45,000 tomato and 48,000 cucumber seedlings were planted per hectare, at a cost of USD 122.4 and 121.2 per thousand, respectively.

**Table 1.** Values used to calculate the average annual economic benefit (USD t−<sup>1</sup> ) based on six rotations (one rotation per year from 2008 to 2013) of greenhouse-cultivated tomato and cucumber plants.


<sup>a</sup> NEB: N fertilizer economic benefit.

The net benefit, NEB, and the input–output ratios (Table 1) were used as the main factors in the economic analysis of N fertilization. The highest net benefit was RCF, which was 1, 64.8, and 83.8% higher than CF, M, and CK, respectively. The RCF group produced the highest NEB, which is 41.8% higher than CF and 93.5% higher than M. The input−output ratio was highest for RCF, followed by CF, CK, and M.

#### **4. Discussion**

#### *4.1. Fruit Yield and N Use Efficiency*

Nitrogen fertilizer is typically used to improve crop yield [30,31]. However, excessive N input does not increase yields [31–33]. In our six-year greenhouse experiment, we found that reducing a certain amount of N content did not significantly reduce fruit yield, relative to the conventional fertilization used by local farmers. This indicates that the local conventional fertilization practice provides excessive fertilization. Reducing chemical fertilizer application from 1027 kg ha−<sup>1</sup> N to 692 kg ha−<sup>1</sup> N by 38.7% (335 kg ha−<sup>1</sup> N) for tomato and from 850 kg ha−<sup>1</sup> N to 662 kg ha−<sup>1</sup> N by 28.8% (188 kg ha−<sup>1</sup> N) for cucumber did not negatively affect fruit yield.

In the first year, the yield of the control treatment was not significantly different from that of the other treatments. This reflects previous excessive fertilization, leading to N accumulation in the soil, which supported growth during the first tomato season. However, the yield of the control treatment was lower in 2009, indicating that there was insufficient accumulated N in the soil to support tomato growth after N absorption by tomatoes and cucumbers in 2008. Based on this, crop yields could be severely reduced when no fertilizer is applied.

Organic fertilizer supplementation can improve soil biological quality and function, thereby further improving crop yield [34,35]. This view has been generally accepted by local farmers in the region. When we applied organic fertilizer only, with an average annual N input of 362 kg ha−<sup>1</sup> , tomato yield was the same as that of the reduced chemical fertilizer treatment for the first two years. The significant yield reduction occurred in the third year. We speculate that a single application of organic N can delay the reduction in vegetable yield, even when no chemical fertilizer N is used; however, the organic fertilizer level in this research was insufficient to support production in the third year.

The average fruit yield was 13.8% higher in the organic (M) treatment (at 119.2 t ha−<sup>1</sup> ) than in the control (CK) treatment (at 104.7 t ha−<sup>1</sup> ). However, only organic N application was substantially lower than those of the conventional and reduced fertilizer treatments (168.4 and 167.4 kg ha−<sup>1</sup> , respectively). These results indicate that reducing 100% of chemical fertilizer reduces yield. As expected, 27.8% reduced fertilization (from 1877 to 1354 kg ha−<sup>1</sup> ) produced similar yields to conventional fertilization. This is consistent with previous findings that yields can be maintained under appropriate chemical fertilizer reduction [36–38], this could be associated with the slow or insufficient N supply from the organic source.

Excessive fertilization not only contributes little in terms of increasing yield but also increases N accumulation in soil [39]. Due to the high background nitrogen accumulation in the soil, long-term observation is needed to obtain relatively accurate results in the comparison of different nitrogen application levels. As shown in Figure 2, the RE<sup>N</sup> and AE<sup>N</sup> levels in CF and RCF exceeded M in the fourth year. And the PE<sup>N</sup> level in CF and RCF exceeded M in the third year. Then, the situation was kept for the following years. Therefore, we believe that the evaluation of NUE with the results of the sixth year is reliable.

The relationship between CF and RCF for NUE in this research indicates that a moderate reduction in chemical fertilizer N application could improve NUE in an organic– inorganic fertilization situation. The main reason for this might be the balanced N supply from both sources, quicker at early stages from an inorganic source, and slow release throughout the cropping time by an organic source. Our observations are consistent with previous findings that moderate chemical fertilizer reductions can improve N use efficiency in areas that have been over-fertilized [40]. The NUE is typically negatively correlated

with the N fertilizer application rate [41]. The PFP in our research have the same situation. Contrary to expectations, in spite of the N application being lower in M than that of CF, the REN, AEN, and PE<sup>N</sup> were also lower than that of CF in the end. It can be inferred that compared with the application of organic fertilizer alone, the combined application of organic and inorganic fertilizer can effectively improve REN, AEN, and PEN.

## *4.2. N Leaching*

It has been reported that N leaching rates differ between years (2008–2014) for crops field [15]. Compared with short-term batch tests, long-term positioning tests are more reliable for comparing differences in N leaching among treatments [7,15]. Soil N leaching is related to excessive irrigation, heavy rainfall, over-fertilization, and poor tillage management [42]. In our six-year greenhouse positioning experiment, by controlling for irrigation and tillage mode, differences in leaching were explained mostly by differences in fertilizer input. Reducing chemical fertilizer N application by 34.5% from conventional levels reduced TN leaching significantly by 39.8%, and when using only organic manure fertilizer, reduced it by 77.4% (Figure 3b). This indicates that N leaching is directly related to the amount of chemical fertilizer used; therefore, reducing chemical fertilizer N input is an effective way to reduce N leaching, which is consistent with the previous studies [33,43].

During the fallow period, no fertilization was applied, and irrigation was reduced, causing TN leaching to be relatively low (15.7–22.9%). In the CK treatment, TN leaching was 5.4% higher for tomatoes than for cucumbers. In contrast, tomatoes had lower TN leaching in the CF, RCF, and M treatments by 16.0%, 13.5%, and 1.2%, respectively, compared with cucumbers. This indicates that relative to tomato cultivation, cucumbers have a higher risk of leaching caused by fertilization. Hence, reducing N application for cucumber cultivation has the potential to reduce N leaching.

Organic fertilizer application at 100 and 200 kg N ha−<sup>1</sup> year−<sup>1</sup> N was shown to cause N leaching at 85.2 and 105.5 kg ha−<sup>1</sup> , which were 18.3 and 38.6 kg ha−<sup>1</sup> higher than control (66.9 kg ha−<sup>1</sup> ), respectively [44]. A greenhouse vegetable study found that applying poultry manure at less than 217.7 kg ha−<sup>2</sup> N did not negatively affect groundwater; however, double N application will lead to an increase in leaching N [45]. In our study, the organic fertilizer N input was 362 kg ha−<sup>1</sup> year−<sup>1</sup> , which carries the risk of enhancing N loss. Nitrogen leaching was 92 kg ha−<sup>1</sup> year−<sup>1</sup> in the M treatment and 69 kg ha−<sup>1</sup> year−<sup>1</sup> in the control counterpart (Table 2); note that this difference of 23 kg ha−<sup>1</sup> year−<sup>1</sup> was caused by organic N input. Then, the N leaching caused by the chemical N input part of CF and RCF were 78 and 39 kg ha−<sup>1</sup> year−<sup>1</sup> , respectively. Furthermore, the organic fertilizer N leaching-to-input ratio (NLIRON) was 6.4%, which was significantly higher (by 42.2%) than that of the conventional fertilizer treatment (4.5%), and 178.3% greater than that of the reduced fertilizer treatment (77.8%). The N leaching-to-input ratio (NLIRCN) in CF and RCF was 3.5% and 2.3%, which were 45.3% and 64.0% lower than NLIRON, respectively. Therefore, we found that organic fertilization carries a higher risk of N leaching than chemical fertilization, which is consistent with some earlier findings [46,47], despite more research having widely shown that the application of organic N fertilizer would cause a decrease in N leaching than inorganic [48–50]. Most of the nitrogen in organic fertilizer exists in the form of macromolecular, which can only be absorbed by plants after a certain period of dissolution by a series of microbial-mediated steps [51,52]. Therefore, compared with inorganic nitrogen from chemical N, it is difficult to be directly used by crops. These macromolecular N and dissolved N from organic fertilizer would also be lost by leaching when irrigation happened [53]. Therefore, we speculate that slow nitrogen dissolution, limited uptake by plants, and solubility of organic N are the main reasons for the high N leaching loss of organic fertilizer. However, further research is still needed to confirm the details. Previous studies have found that a combination of chemical and organic fertilizer is a sustainable fertilization approach [54], which can promote crop productivity and N uptake, and reduces N losses [55,56]. Similarly, our findings support combined chemical

and organic fertilization, while improving the activity of N-transforming microbes in the soil by organic manure, the deficiency of N leaching was balanced.

**Table 2.** Approximate fertilizer-N input and N leaching in the greenhouse vegetable cultivation system used in this study, from 2008 to 2013.


CK: control (0 kg ha−<sup>1</sup> N); CF: conventional fertilization (1515 chemical fertilizer + 362 kg ha−<sup>1</sup> organic N); RCF: reduced conventional fertilization (992 chemical fertilizer + 362 organic kg N ha−<sup>1</sup> ); M: manure fertilization (362 kg ha−<sup>1</sup> organic N). Results are expressed as means. Lowercase letters after means identify groups that differ significantly between treatments (*p* < 0.05), by row. *NLTN*: N leaching caused by TN input; *NLON*: N leaching caused by organic N input; *NLCN*: N leaching caused by chemical N input; *NLIRTN*: TN leaching-to-input ratio; *NLIRON*: organic N leaching-to-input ratio; *NLIRCN*: chemical N leaching-to-input ratio.

#### *4.3. Economic Benefit Analysis*

Fertilizer N input is the main factor affecting crop yield and economic benefits [57]. We found that reduced chemical fertilization produced a similar economic net benefit to conventional fertilization, whereas 100% chemical fertilization reduction produced 38.7% less net benefit than conventional fertilization (Table 1). Although organic fertilization alone reduces production costs (the chemical fertilizer part), it also significantly reduces profitability. Therefore, moderate chemical fertilizer reduction was an effective way to balance production costs and maintain profitability. Reduced chemical fertilization also increased NEB by 41.8%, whereas organic fertilization alone reduced it by 26.7%. It can be inferred that the NEB of a single application of organic fertilizer was much lower than that of a mixed application of organic and chemical fertilizer. Further, it can be seen that organic fertilizer alone is inferior to chemical fertilizer for increasing economic growth. Although the advantages of organic fertilizer application are generally recognized, its relatively low N content and the large amounts required make its cost far higher than that of chemical fertilizer [58–60].

Ranking the treatments in terms of their input–output ratios, in ascending order, yielded the following order: M, CK, CF, and RCF. This result showed that only organic fertilizer will not bring economic benefits (M < CK), and an appropriate reduction in chemical fertilizer in areas where there has been excessive fertilization can effectively improve the economic benefit of N fertilization (RCF < CF). Similarly, for a rice–wheat rotation system, Wang et al. [61] found that a 50% reduction in chemical fertilizer N raised NEB by 320.8% and increased the input–output ratio from 1:3.0 to 1:4.0, relative to conventional tillage [62].

#### *4.4. Selecting the Optimum N Application Rate*

In conventional agricultural production in China, the yield and economic benefits related to the N application rate are the most important driving factors [63–65]. To reduce the high production costs and environmental pollution risks caused by excessive N input, sustainable agriculture aims to balance the N application rate, with both ecological and agronomic benefits [66–68]. Therefore, we aimed to determine the optimum range of N

application rates to balance N leaching, vegetable yield, net benefit, and NEB in greenhouse vegetable farming. Organic fertilizer was applied at the same rate in the organic fertilization, conventional fertilization, and reduced conventional fertilization treatments. Figure 4 illustrates the effects of each treatment on yield, N leaching, net benefit, and the N fertilizer economic benefit.

**Figure 4.** The optimum range of N application. (**a**) Regression models of tomato/cucumber yield, N leaching, N fertilizer economic benefit (NEB), and net benefit against the annual N application rate (2008–2013). Regression models of net benefit, N fertilizer economic benefit (NEB), tomato yield (**b**), and cucumber yield (**c**) against the annual N application rate (2008–2013).

Yang et al. [15] found that TN leaching at less than 170 kg ha−<sup>1</sup> would cause no environmental harm. In our study, the highest average annual N leaching rate was 170.7 kg ha−<sup>1</sup> in conventional fertilization treatment, at the N application rate of 1877 kg ha−<sup>1</sup> , and N leaching was lower when the N application rate was reduced. Therefore, we consider a total average annual N input of less than 1877 kg ha−<sup>1</sup> to be safe for the environment. In the tomato/cucumber rotation, the maximum yield was obtained at an annual N application rate of 1691 kg ha−<sup>1</sup> , using 12.3% less chemical fertilizer N than conventional fertilization (Figure 4a). Generally, the N application rate that maximizes the crop's economic benefit is lower than that which maximizes yield [62,69]. Our findings were similar: the maximum net benefit occurred at an annual N application rate of 1583 kg ha−<sup>1</sup> (108 kg ha−<sup>1</sup> lower than the rate that maximized fruit yield), and using 19.4% less chemical fertilizer

N than conventional fertilization. The maximum NEB was obtained at the annual N rate of 1143 kg N ha−<sup>1</sup> , using 48.4% less chemical fertilizer N than conventional fertilization. Maximum NEB reflects the most economical use of chemical N fertilizer. Further, using less N reduces N leaching and the risk of environmental pollution; hence, NEB may become a key indicator for determining chemical fertilizer N use in the future. We, therefore, recommend N application rates of 1143 and 1583 kg ha−<sup>1</sup> for tomato and cucumber greenhouse farming, respectively, and reducing chemical fertilizer N application rates by 19.4% and 48.4%, respectively, relative to conventional fertilization. The maximum yield, net benefit, and NEB for tomatoes occurred at N application rates of 918, 821, and 634 kg ha−<sup>1</sup> , respectively (Figure 4b); for cucumbers, they occurred at 871, 778, and 556 kg ha−<sup>1</sup> , respectively (Figure 4c). In summary, our findings indicate that the optimal ranges of N application rates for greenhouse production of tomatoes and cucumbers are 634–821 kg ha−<sup>1</sup> and 556–778 kg ha−<sup>1</sup> , respectively, under current experimental conditions.
