**3. Results**

#### *3.1. The Description of Maize Agronomic Parameters*

According to the results of ANOVA (Table 1), maize yield, AGB, PNC, and NNI, were all significantly a ffected by soil type, year with its weather pattern, N rate, and their interactions. However, the yield, PNC, and NNI were not directly a ffected by the planting density.

**Table 1.** Significance of mean squares in the analysis of variance (ANOVA) of yield, aboveground biomass (AGB), plant N concentration (PNC), and N nutrition index (NNI) across two soil types (S), three years (Y), three densities (D), and six N rates (N).


Note: \*, \*\*, and \*\*\* indicate significance at *p* < 0.05, *p* < 0.01, and *p* < 0.001 probability levels, respectively. ns was non-significant (*p* > 0.05).

The multiple comparisons of the analyzed agronomic parameters in data subsets aggregated by a given influencing factor are shown in Table 2. The maize yield, AGB, and NNI in the black soil field were significantly higher (by 3.43 t ha−1, 5.91 t ha−1, and 0.06) than in the aeolian sandy soil field. On the other hand, the PNC were significantly lower (by 1.31 kg kg−1) in the black soil field than in the aeolian sandy soil field. The relatively wet season of 2016 brought the highest yield and AGB

while NNI was the lowest among three years. In 2015, a relatively dry year, the yield and PNC were the lowest in the analyzed period. The yield, PNC, and NNI were not significantly affected by the three tested planting densities. The values of all of the parameters significantly improved with the increasing N rate, until the N4 treatment (240 kg ha−1).

**Table 2.** The multiple comparisons of maize yield, net return (NR), aboveground biomass (AGB), plant N concentration (PNC), N nutrition index (NNI), and N surplus (NS) at two soil types (S), three years (Y), three densities (D), and six N rates (N) respectively.


Note: the notation for treatments within soil (B: black soil, S: aeolian sandy soil), year, density (D1: 55,000 plant ha−1, D2: 70,000 plants ha−1, and D3: 85,000 plants ha−1), and nitrogen (N0: 0 kg ha−1, N1: 60 kg ha−1, N2: 120 kg ha−1, N3: 180 kg ha−1, N4: 240 kg ha−1, and N5: 300 kg ha−1). The number behind "±" is the standard error, and numbers for the same item followed by different letters indicate significant differences (*p* < 0.05).

An overview of the relationships between maize yield and agronomic parameters showed distinct crop response to growing conditions (Figures 1 and 2). Whether across the three years or in a specific year, for black soil and aeolian sandy soil fields the relationship between yield and AGB had a significant quadratic relationship. On the contrary, the relationships between yield and either PNC or NNI were modeled according to the linear-plus-plateau models. Across the three years, the yield was maximized when the PNC reached 9.6 kg kg−<sup>1</sup> and 10.1 kg kg−<sup>1</sup> in black soil and aeolian sandy soil, respectively. Correspondingly, in black soil field, the yield reached its maximum when the NNI was at 0.95, while in the aeolian sandy soil, the maximum yield was obtained at NNI of 0.81. Analyzed for a given year, in 2015, 2016, and 2017, the yield was maximized when the PNC reached 8.2, 9.2, and 9.9 kg kg−<sup>1</sup> in black soil and 9.7, 9.5, and 10.1 kg kg−<sup>1</sup> in aeolian sandy soil, respectively. Correspondingly, in the black soil field, the yield reached its maximum when the NNI was at 1.15, 0.84, and 0.90 in specific year of 2015, 2016, and 2017, while in the aeolian sandy soil, the yield was maximized at NNI of 0.74, 0.80, and 0.88, respectively.

**Figure 1.** The relationships between crop yield and aboveground biomass (**a**), plant N concentration (**b**), or N nutrition index (**c**) for two soils across three years and three planting densities. (Note: the "n" is the number of samples for each soil type, the first equation is for black soil, the second equation is for aeolian sandy soil, \* indicate significance at *p* < 0.05 probability level).

**Figure 2.** *Cont.*

**Figure 2.** The relationships between crop yield and aboveground biomass (**a**,**d**,**g**), plant N concentration (**b**,**e**,**h**), or N nutrition index (**c**,**f**,**i**) in year of 2015 (**a**–**c**), 2016 (**d**–**f**), and 2017 (**g**–**i**) for two soils across three planting densities. (Note: the "n" is the number of samples for each soil type, the first equation is for black soil, the second equation is for aeolian sandy soil, \* indicate significance at *p* < 0.05 probability level.

#### *3.2. The Response of Maize Agronomic Parameters to N Application Rate*

The maize yield was significantly higher in black soil than in aeolian sandy soil at each N application rate (Figure 3). According to the quadratic-plus-plateau model, the maize yield was maximized at the N rates of 285 and 201 kg ha−<sup>1</sup> in black soil and aeolian sandy soils across three years, respectively. Furthermore, the lowest N rate for obtaining the maximum yield, or the agronomic optimal N rate (AONR), was not stable in either the black soil field or the aeolian sandy soil field and was influenced by the weather pattern in a given season. For specific year of 2015, 2016, and 2017, the soil-specific AONR of black soil and aeolian sandy soil fields were 300, 243, and 277 kg ha−<sup>1</sup> and 112, 209, and 217 kg ha−<sup>1</sup> in 2015, 2016, and 2017, respectively. The average EONRs were 265 kg ha−<sup>1</sup> (276, 230, and 260 kg ha−1) and 186 kg ha−<sup>1</sup> (101, 193, and 203 kg ha−1) in black soil and aeolian sandy soil fields, respectively, across three years (in specific year of 2015, 2016, and 2017).

**Figure 3.** The responses of crop yield to N application rate across three years (**a**), and in specific year of 2015 (**b**), 2016 (**c**), and 2017 (**d**) in two soils across three planting densities. (Note: the "n" is the number of samples for each soil type, the first equation is for black soil, the second equation is for aeolian sandy soil, \* indicate significance at *p* < 0.05 probability level, the lowercase letters in the table indicate the significant difference at 0.05 level).

Moreover, the soil-specific EONR was also influenced by the year and planting density interaction (Figure 4). In the black soil field, the soil-specific EONR had a coe fficient of variation (CV) of 10% and reached 210, 225, and 240 kg ha−1, 234, 214, and 252 kg ha−1, and 266, 250, and 266 kg ha−<sup>1</sup> at the planting density of 55,000, 70,000, and 85,000 plants ha−<sup>1</sup> in year of 2015, 2016, and 2017, respectively. The NUE analysis using PFP and AE showed that the highest values were obtained at the planting density of 70,000 plants ha−<sup>1</sup> (64 and 42 kg kg−1) in all three years compared with 50,000 plants ha−<sup>1</sup> (54 and 20 kg kg−1) and 85,000 plants ha−<sup>1</sup> (50 and 37 kg kg−1). In the aeolian sandy soil field, the soil-specific EONR varied with the CV of 30% and reached 88, 150, and 96 kg ha−1, 158, 209, and 215 kg ha−1, and 168, 177, and 208 kg ha−<sup>1</sup> at the planting density of 55,000, 70,000, and 85,000 plants ha−<sup>1</sup> in year of 2015, 2016, and 2017, respectively. Interestingly, the NUE analysis (PFP and AE), showed that the highest values were obtained at the planting density of 55,000 plants ha−<sup>1</sup> (75 and 23 kg kg−1) in all three years compared with 70,000 plants ha−<sup>1</sup> (55 and 34 kg kg−1) and 85,000 (50 and 36 kg kg−1) (Figure 4). Meanwhile, according to the multiple linear regression (Figure 5), soil-specific EONR (R<sup>2</sup> = 0.77) and obtained yield (R<sup>2</sup> = 0.95) showed significant relationships with the soil total N, growing degree days, accumulated precipitation, and planting density.

**Figure 4.** The variation of economic optimal N rate (EONR) and N use e fficiency (NUE) (partial factor productivity (PFP) and agronomic e fficiency (AE)) in a specific soil (B: black soil, (**a**); S: aeolian sandy soil, (**b**), year (2015, 2016, and 2017), and planting density (D1: 55,000 plant ha−1, D2: 70,000 plant ha−1, and D3: 85,000 plant ha−1). (Note: the "n" is the number of samples for EONR, PFP, and AE respectively).

**Figure 5.** The relationships for measured and predicted soil-year-density specific economic optimal N rates (**a**) or yield (**b**) using soil, weather, and planting information (Note: the "n" is the number of samples, EONR = 101.24 × TN + 100.68 × GDD − 10.12 × APP + 0.57 × D − 164796.76, Y = 6.74 × TN + 0.91 × GDD − 0.09 × APP + 0.02 × D − 1486.83, where EONR is economic optimal N rate, Y is grain yield, TN is soil total N, GDD is growing degree days, APP is accumulated precipitation, D is planting density).

#### *3.3. The Potential Benefits of Site-Specific N Management Strategies*

Based on the EONR specific to different soil types, years and planting densities, as described above (Figure 4), three site-specific N managemen<sup>t</sup> strategies were proposed. The results of the SS-EONR, SYS-EONR, and SYDS-EONR strategies with their explicit N-rates and optimal planting densities at 70,000 and 55,000 plants ha−<sup>1</sup> for the black and aeolian sandy soil fields, respectively, were averaged across the soils and years (Table 3). This facilitated the comparison with FNR at 300 kg N ha−<sup>1</sup> and 55,000 plants ha−<sup>1</sup> and RONR at 240 kg N ha−<sup>1</sup> and 70,000 plants ha−1. The variation between the different strategies at the two soil types are given in Figure 6.

**Table 3.** The comparison of the N rate, yield, net return, partial factor productivity (PFP), agronomic efficiency (AE), and recovery efficiency (RE) from different N managemen<sup>t</sup> strategies across soils and years.


Note: CK: check, zero N rate; FNR: farmer N rate; RONR: regional optimal N rate, SS-EONR: soil-specific economic optimal N rate; SYS-EONR: soil-, and year-specific economic optimal N rate; SYDS-EONR: soil-, year- and density-specific economic optimal N rate. The number behind "±" is the standard error. Different lowercase letters in the same column indicate significant difference at 0.05 level (*p* < 0.05).

In comparison with FNR across the two soil types and three years (Table 3), the SS-EONR, SYS-EONR, and SYDS-EONR strategies significantly reduced N rate by 25%, 30%, and 38%, increased NR by 155, 176, and 163 \$ ha−1, and improved NUE parameters (PFP and AE) by 37–42%, 52%, and 67–71%, respectively. Meanwhile, these three strategies showed no significant effects on maize yield. When compared with RONR, the SS-EONR, SYS-EONR, and SYDS-EONR strategies significantly reduced N rate by 6%, 12%, and 22%, and improved NUE parameters (PFP and AE) by 7–8%, 16–19%, and 28–34%, respectively, without significantly affecting maize yield and NR.

Analyzed for each soil type separately, the SS-EONR, SYS-EONR, and SYDS-EONR strategies performed differently when compared with FNR and RONR across the three years (Figure 6). In the black soil field, in comparison with FNR, the SS-EONR, SYS-EONR, and SYDS-EONR strategies significantly reduced N rate by 12%, 15%, and 22%, increased NR by 201, 212, and 193 \$ ha−1, and improved PFP by 20%, 26%, and 35%, respectively, without significantly affecting maize yield, AE, and RE. However, when compared with RONR, the SS-EONR, SYS-EONR, and SYDS-EONR strategies did not significantly affected N rate, maize yield, NR, and NUE parameters (PFP, AE, and AE).

In the aeolian sandy soil field, in comparison with FNR, the SS-EONR, SYS-EONR, and SYDS-EONR strategies reduced N rate by 38%, 45%, and 54%, increased NR by 109, 140, and 133 \$ ha−1, and improved NUE parameters (PFP and AE) by 62%, 76–93%, and 105–126%, respectively. When compared with RONR, the SS-EONR, SYS-EONR, and SYDS-EONR strategies reduced N rate by 23%, 31%, and 42%, increased NR by 95, 126, and 119 \$ ha−1, and improved NUE parameters (PFP and AE) by 31–33%, 44–56%, and 68–83%, respectively. It is worth noting that the SS-EONR, SYS-EONR, and SYDS-EONR strategies showed no significant difference in maize yield whether comparing with FNR or RONR in the aeolian sandy soil field.

**Figure 6.** The comparison of the N rates (**a**), yield (**b**), net return (**c**), partial factor productivity (PFP, **d**), agronomic efficiency (AE, **e**), and recovery efficiency (RE, **f**) from different N managemen<sup>t</sup> strategies at specific soil type across three years. (Note: CK: zero N rate, FNR: farmer application N rate, RONR: regional optimal N rate, SS-EONR: soil-specific economic optimal N rate, SYS-EONR: soilyear-specific economic optimal N rate, SYDS-EONR: soil-year- density-specific economic optimal N rate. The "n" is the number of samples for each N managemen<sup>t</sup> strategies at specific soil type respectively. The lowercase letters in the table indicate the significant difference at 0.05 level (*p* < 0.05)).
