**4. Discussion**

#### *4.1. The Soil-Specific Economic Optimal N Rate*

In this study, the black soil field was characterized by a higher water holding capacity and soil fertility than the aeolian sandy soil field (Table 1). This led to a more efficient nutrient supply to the maize crop during the growing season and resulted in larger AGB and NNI in the black soil field (Table 2). These findings were in agreemen<sup>t</sup> with previous studies conducted in this study region [47,48]. According to the relationship between grain yield and AGB or NNI (Figure 1), a higher yield was recorded in the black soil field than in aeolian sandy soil field (Table 2). That is despite the fact that the PNC was higher in the aeolian sandy soil field than in the black soil field (Table 2). The NNI is generally used during the growing season to diagnose crop N status (deficient, optimal or surplus) for guiding in-season N application [49], however, the concept can also be extended to the maturity stage to guide adjustment of N managemen<sup>t</sup> in the following season [50].

It is usually assumed that the ONRs are higher in coarse-textured soil fields than in fine-textured soil fields, due to the disability in coarse-textured soil fields to retain moisture leading to higher N leaching potential [51]. As a result, most farmers apply more N fertilizer in the aeolian sandy soil

field than in the black soil field [52]. Furthermore, in this study location, maize production is rain-fed and water deficit frequently occurs during the maize growing season, hence the drought has been the main limiting factor of crop growth in the aeolian sandy soil field [53,54]. In order to avoid the overuse of N fertilizers, many researchers tend to use the linear plus plateau model to determine the ONR in China [22,55]. In this study, according to the R<sup>2</sup> and RMSE, the quadratic-plus-plateau model had the best fit and was, therefore, used to calculate the EONR. The EONR across three years and three planting densities was considerably higher in fine-textured black soil field (265 kg ha−1) than in coarse-textured aeolian sandy soil field (186 kg ha−1) (Figure 3a). According to the relationships between yield and PNC or NNI (Figure 1), plants with a given level of PNC and NNI could produce much more yield in black soil than in aeolian sandy soil. The minimum NNI to obtain the maximum yield in the aeolian sandy soil field (0.81) was significantly lower than in the black soil field (0.95). In other words, adding more N fertilizer would not lead to substantial increase of maize yield in aeolian sandy soil field. Therefore, N fertilizer was not considered the main limiting factor there. This result was in agreemen<sup>t</sup> with the previous studies stating that the ONR was lower in coarse-textured soil fields than in fine-texture soil fields and showing grea<sup>t</sup> soil-specific variability [50,56].

#### *4.2. The Influence of Weather Conditions and Planting Density on Soil-Specific Economic Optimal N Rate*

The interaction between soil properties and weather conditions had the greatest influence on the response of crop yield to N fertilizer [23,24,57]. According to the previous research [58–60], the relationship between soil properties and yield was mainly a ffected by the spatial and temporal variability in soil water holding capacity and precipitation. Therefore, ONR should be adjusted based on the interaction between soil properties and weather conditions. Precipitation was significantly di fferent among three years covered by this study (Figure 1), and had a significant e ffect on yield, AGB, PNC, and NNI (Tables 1 and 2). Meanwhile, the minimum PNC and NNI to obtain the maximum maize yield also showed inter-annual variation in both fields (Figure 2). This resulted in the year-to-year variability of soil-specific EONR (Figure 3b–d). For the year of 2016, in black soil field with high soil bu ffering capacity and fertility (total N and SOM), the relatively high GDD with well-distributed precipitation would lead to a higher AGB and grain yield potential than in 2015 and 2017, a phenomenon noted also in several other studies [40,61,62]. Furthermore, the synchronization of high GDD and well-distributed precipitation in 2016 would lead to a higher soil nitrification rate [19,63] and would provide relatively more soil N for the maize growth than in 2015 and 2017. As a consequence, in 2016 the minimum NNI to obtained the maximum maize yield was the lowest among the three years. Therefore, the SS-EONR for the black soil was lower in 2016 than in 2015 and 2017. On the other hand, in the year of 2015, in aeolian sandy soil with low soil bu ffering capacity and fertility (total N and SOM), the severe drought restricted the crop growth and yield formation, a phenomenon well described in another study [64]. Due to the low AGB and yield potential, the minimum NNI to obtain the maximum maize yield in the aeolian sandy soil field was the lowest in 2015 among the three tested growing seasons. Therefore, the SS-EONR for the aeolian soil was lower in the dry year (2015) than in 2016 and 2017.

Another question faced by scientists and the farmers is how planting density should be adjusted for di fferent soil types and weather conditions. Although, in this study, the planting density did not have any significant e ffect on the yield, PNC, and NNI (Table 1), the soil-specific EONR still varied among three weather conditions and planting densities, along with PFP and AE (Figure 4). Also, the variability of the parameters was higher in the aeolian sandy soil field than in the black soil field. The bu ffering capacity mainly comes from the texture and organic carbon. Therefore, in the fertile black soil field with a higher bu ffering capacity, the production would in theory be less a ffected by the varying conditions than in the barren aeolian sandy soil field. The barren aeolian sandy soil field had a low yield potential and high variation in soil conditions, leading to high variation in AGB and yield, which translated to high variation in EONR. Due to the relatively higher N uptake and AGB accumulation at the relatively higher planting densities [37,65], the highest soil-specific AONRs were defined in this study under the high (85,000 plants ha−1) planting density in the fertile black soil field

and under the middle (70,000 plants ha−1) and high (85,000 plants ha−1) planting density in the aeolian sandy soil field. Therefore, according to the PFP and AE with the highest values among three planting densities, the middle (70,000 plants ha−1) and low (55,000 plants ha−1) planting densities with their corresponding SYS-EONR would be the optimal N managemen<sup>t</sup> strategy for maize production in the black soil and aeolian sandy soil fields, respectively. The SS-EONR could be adjusted based on the information about the soil properties, weather conditions, and planting density [66]. Through the multiple linear regression analysis (Figure 5) performed in this study, the SYDS-EONR and the obtained grain yield could be determined preliminarily using soil N, GDD, APP, and planting density.

#### *4.3. The Potential Benefits of Applying Soil-Specific Economic Optimal N Rate*

Currently, the RONR strategy recommended about 240 kg N ha−<sup>1</sup> with 70,000 plant ha−<sup>1</sup> for this study region [8,9]. In this strategy, the N fertilizer is applied at a fixed rate and timing without accounting for spatial and temporal variability in soil N supply and crop N demand. According to the results of this study, the EONR changed dramatically from the black soil field to the sandy soil field and from year to year, which confirmed the findings of the previous studies showing that an ONR varied significantly in space and time [33–35]. The previous research demonstrated that soil-specific N managemen<sup>t</sup> could adjust the N fertilizer application to match crop requirement by identifying the gap between soil N supply and crop N demand according to their spatial and temporal variation in a particular growing season for a specific soil type [67,68]. Therefore, it is of grea<sup>t</sup> interest to learn how much we can further improve N managemen<sup>t</sup> using alternative strategies that are more complex and accurate than the simple FNR and RONR strategies.

Across the two typical soils in this study region, with distinctly di fferent soil properties, compared with FNR, the soil-specific EONR strategies would decrease the N application rates with no negative e ffect on maize yield, while increasing NR and NUE (Table 3 and Figure 6). When compared with RONR, the soil-specific EONR strategies still showed the potential to decrease the N application rates and increase NUE but with no negative e ffect on maize yield and NR. Meanwhile, because the EONR showed higher variability in aeolian sandy soil than in black soil across di fferent weather conditions and planting densities (Figure 4), the soil-specific EONR strategies showed greater potential in decreasing N application rates and increasing NUE in aeolian sandy soil than in black soil. Therefore, the soil-specific EONR strategies have a grea<sup>t</sup> potential to be implemented to achieve high-yield and high-e fficiency maize production in China. Furthermore, because of the variation in weather conditions, especially precipitation, EONR varied among di fferent years (Figure 3). The SYS-EONR strategy would perform better in increasing NUE than the SS-EONR strategy. Although the planting density had no significant e ffects on grain yield and NR in this study (Table 1), the EONR was significantly a ffected by it and the interaction among soil type, weather conditions, and planting density (Figure 4). Therefore, the SYDS-EONR strategy would result in the highest potential benefits in reducing the N application rate, and increasing NUE than the SS-EONR and SYS-EONR strategies (Table 3 and Figure 6).

These results indicated that soil-specific N managemen<sup>t</sup> had the potential to increase N managemen<sup>t</sup> and improve NUE. The best improvement may be achieved in the coarse-textured aeolian sandy soils and implementing the soil-, year- and planting density-specific EONR strategy. However, it is a grea<sup>t</sup> challenge to determine soil- and year-specific planting densities and corresponding optimal N rates across di fferent farmers' fields. Future studies are needed to use crop-sensing technologies and crop growth modeling methods to guide in-season soil-specific N managemen<sup>t</sup> under on-farm conditions [5,43,69,70].
