Nitrogen, phosphorus, and potassium not only affect crop growth rate and yield, but also can cause changes in related biochemical components in plants (Li, 2006, Science Press, pp.176-180; Tang et al., 2015). The overuse of fertilizers and pesticides has caused serious soil salinization. How to reduce fertilizer application without reducing crop yield, nutritional diagnosis has become an important means of real-time management of crop nutrients (Luo and Zhu, 2014). Spectral analysis technology has the advantages of good reproducibility, convenient testing, low cost, real-time, fast, non-destructive, and accurate (Racy et al., 1996; Bronson et al., 2003; Xue et al., 2003), making it an effective means to obtain crop nutritional status and growth information (Jiang et al., 2008).

Greenhouse vegetables need heavy phosphorus reduction, moderate nitrogen reduction, potassium regulation. Spectral diagnosis technology has the characteristics of high accuracy, simplicity and rapidity, which is of great significance for formulating scientific fertilization plan of nutrient profit and loss and establishing scientific management technology system of fertilizer and water for greenhouse vegetables. Predecessors have done a lot of research on monitoring the content of nitrogen, phosphorus and potassium in crops by using spectral analysis technology. Among them, there are many studies on estimating the content of potassium in crops using spectral reflectance and spectral analysis and diagnosis technology has been applied to the real-time detection and nutrition diagnosis of nitrogen nutrition in corn (Zhou et al., 2010; Sun et al., 2010; Liang and Liu, 2010; Wang et al., 2011), wheat (Liu et al., 2004; Hu et al., 2009; Zhang and Zhang, 2010), rice (Qin et al., 2011), tomato (Han et al., 2010; Li et al., 2010; Zhu et al., 2014; Yang et al., 2015; Zhao et al., 2018), cotton (Wang et al., 2011; Shi et al., 2014; Ma et al., 2016), ryegrass (Yang et al., 2011), Padus Virginiana‘Schubert’ (Li et al., 2011), peanut (Wang et al., 2014) and Chinese cabbage (Yang et al., 2013). These results showed that the spectral index could indirectly reflect the nitrogen nutrition status of crops by reflecting the chlorophyll content of crops; Using spectral characteristics, sensitive bands and sensitive periods can be screened out, which can quickly and accurately obtain the information of crop nutrition level.

At present, there are many problems in the application of chemical fertilizer for greenhouse vegetables, such as the imbalance of large proportion of fertilizer application, the overuse of nitrogen, phosphorus and potassium fertilizers, the unreasonable application mode of chemical fertilizer and organic fertilizer, the extremely lack of special new chemical fertilizer, the improper application method of chemical fertilizer, the lack of attention to the application of medium and trace element chemical fertilizer, and the low quality of organic fertilizer applied with chemical fertilizer and so on. Therefore, it is necessary to reduce the application of chemical fertilizers, regulate organic fertilizers, and coordinate the nutrient ratio of chemical fertilizers (Song et al., 2017). The sensitivity of leaf spectral reflectance to nitrogen, phosphorus, and potassium in different growth stages of peppers remains to be studied. This study investigated the sensitivity and correlation of leaf spectral reflectance to nitrogen, phosphorus, and potassium in different fertilizer treatments during pepper growth period. Using leaf spectral index to diagnose sensitive periods, the sensitivity of leaf spectral reflectance to nitrogen, phosphorus, and potassium in different growth stages of greenhouse Long pepper was analyzed, and the sensitive bands were explored, aiming to provide the best time and method for rapid, accurate and non-destructive nutritional diagnosis of greenhouse Long pepper.

1 Results and Analysis

1.1 Leaf spectral characteristics analysis of greenhouse Long pepper under different developmental stages and different fertilizer treatments

According to the results (Figure 1), during the initial flowering period, the spectral reflectance of greenhouse Long pepper under different fertilizer treatments showed a pattern of T8>T7>T6>T5>T4>T2>T1>T3, and the leaf spectral reflectance showed a basically consistent pattern. In the visible spectrum band (350~780 nm), the spectral reflectance of leaves under conventional fertilization was higher during the initial flowering period, followed by conventional nitrogen deficiency, reduced 30% to optimize nitrogen deficiency. On the basis of the same amount of phosphorus and potassium, reduced 30% to optimize treatment (0.016 kg/plot/time). On the basis of the same amount of phosphorus and potassium, reduced 30% to optimize treatment (0.044 kg/plot/time), reduced 15% to optimize fertilization, reduced 45% to optimize fertilization, reduced 30% to optimize fertilization, and the chlorophyll content of conventional fertilization was low, the chlorophyll content was higher under the treatment of reducing 30% to optimize fertilization.

During the initial fruiting stage, the spectral reflectance of greenhouse Long pepper leaves showed a consistent pattern of T7>T6>T1>T5>T8>T4>T3>T2. In the visible spectrum band (350~780 nm), the spectral reflectance showed T2>T3>T8>T4>T1>T5>T6>T7. It could be seen that the chlorophyll content of conventional nitrogen deficient fertilization was the lowest, and the chlorophyll content of the optimized fertilization treatment with a reduction of 15% was the highest, followed by a higher chlorophyll content of the optimized fertilization treatment with a reduction of 30%.

During the green ripening period, the spectral reflectance of greenhouse Long pepper showed T2>T5>T4>T3>T7>T6>T1>T8. In the visible spectrum band (350~780 nm), the spectral reflectance exhibited T7>T4>T5>T6>T2>T3>T8>T1. It could be seen that during the green ripening period, the optimized fertilization treatment with a 15% reduction had the lowest chlorophyll content in the leaves of Long pepper, while the conventional fertilization treatment had the highest chlorophyll content in the leaves.

During the color conversion period, the spectral reflectance of greenhouse Long pepper showed T3>T4>T6>T7>T5>T2>T1>T8. In the visible spectrum band (350~80 nm), the spectral reflectance exhibited T6>T5>T3>T2>T4>T7>T8>T1. It could be seen that during the color conversion period, the optimized fertilization treatment with a 30% reduction had the lowest chlorophyll content, while the conventional fertilization treatment had the highest chlorophyll content in the leaves of Long pepper.

During the red ripening period, the spectral reflectance of greenhouse Long pepper showed T4>T3>T6>T5>T7>T2>T1>T8. In the visible spectrum band (350~780 nm), the spectral reflectance exhibited T5>T6>T3>T4>T2>T7>T8>T1. It could be seen that during the red ripening period, the trend of reflectance change of Long pepper was relatively similar to that during the color conversion period. The reflectance of the optimized application treatment (0.044 kg/plot/time) reduced by 30% and the leaves chlorophyll content of the optimized fertilization treatment with a 30% reduction were lower. The optimized fertilization treatment by reducing 15%, the optimized fertilization amount by reducing 45%, and the conventional fertilization treatment resulted in higher chlorophyll content in the leaves.

The spectral reflectance of the leaves measured by the non fertilization treatment of greenhouse Long pepper showed consistent fluctuation trend of spectral curve of the leaves during the initial flowering, fruiting, green ripening, color conversion, and red ripening stages. In the short wavelength range of 310~350 nm, the initial spectral reflectance changed significantly, and the variation gradually decreased with the increase of wavelength. At 400~700 nm, the reflectivity was low, showing a trend of low blue violet band-high green band-low orange red band. The first strong reflection peak appears near 550 nm, and the reflection intensity was different at different development stages. The Red Valley appeared near 680 nm. From this, it could be seen that unique spectral characteristics such as "Blue Edge", "Green Peak", "Yellow Edge", and "Red Valley" appear within the visible range. In the 700~780 nm wavelength range, strong absorption in the red light band and strong reflection in the near-infrared band resulted in a steep climbing ridge on the curve, and the reflectivity rapidly increased. In the near-infrared band of 780~1 050 nm, a strong reflection platform was formed, with high and stable reflectivity. In the 1 050~1 110 nm wavelength range, the spectral reflectance of leaves varied significantly due to atmospheric absorption interference and instrument noise, and the variation gradually increased with the increase of wavelength (Zhuang et al., 2016). The research results were similar to those of almonds. In the entire near-infrared region, the spectral reflectance of the leaves of the five developmental stages of greenhouse Long pepper showed a pattern of initial flowering stage>red ripening stage>initial fruitingstage>green ripening stage>color conversion stage.

1.2 Effect of reduced fertilizer application on visible spectral characteristics of different fruit development stages

1.2.1 Study on the visible spectral characteristics of fruits at different development stages under the condition of habitual fertilization and fertilizer reduction by 30% (T4, T6, T7, T8)

According to the fertilization experiment design (Figure 2), in the visible spectrum band (350~780 nm), under conventional fertilization treatment and optimized fertilization with a 30% reduction in nitrogen deficiency conditions, the spectral reflectance showed a pattern of initial flowering stage>initial fruiting stage>red ripening stage>green ripening stage>color conversion stage. Under conventional fertilization with nitrogen deficiency and optimized fertilization with a 30% reduction, the spectral reflectance showed a pattern of initial flowering stage>initial fruiting stage>color conversion stage>red ripening stage>green ripening stage.

1.2.2 Visible spectral characteristics of fruits at different developmental stages under different fertilizer reduction treatments with conventional fertilization as the control

The results showed (Figure 3) that in the visible spectrum band (350~780 nm), under the conditions of conventional fertilization, 30% reduction of optimized fertilization, 15% reduction of optimized fertilization, and 45% reduction of optimized fertilization, the spectral reflectance all showed a trend of initial flowering stage>initial fruiting stage>red ripening stage>green ripening stage>color conversion stage.

1.2.3 The visible band spectral characteristics of fruits at different development stages under the treatment of comparison of nitrogen fertilizer treatments based on the same amount of phosphorus and potassium

The results showed (Figure 4) that in the visible spectrum band (350~780 nm), under the conditions of -30 optimized fertilization (-N), -30 optimized fertilization (N1), -30 optimized fertilization (N2), and -30 optimized fertilization (N3), the spectral reflectance showed a pattern of initial flowering stage>initial fruiting stage>red ripening stage>green ripening stage>color conversion stage.

1.3 Correlation between N, P, K content and spectral reflectance of different wavelengths in leaves of greenhouse Long pepper at different fruit developmental stages

1.3.1 The correlation between N, P, K content and spectral reflectance at different wavelengths under the conditions of habitual fertilization and fertilizer reduction by 30%

The results showed (Figure 5) that under the conditions of habitual fertilization and 30% reduction of fertilizer application, the N content in the leaves of greenhouse Long pepper showed a negative correlation with their spectral reflectance in the visible spectrum band of 310~731 nm during the initial flowering period. The correlation was strong in the 685~722 nm band, with a negative correlation coefficient of 0.95; There was a positive correlation between 732~800 nm, with a weak correlation and a positive correlation coefficient of 0.608. Therefore, the 685~722 nm (red light) band can be used as an indicator band for the N content in the leaves of greenhouse Long pepper during the initial flowering period. The P content was positively correlated with its spectral reflectance in the visible range of 310~510 nm, 659~675 nm, and 770~791 nm, with a positive correlation coefficient of 0.475. Among them, the negative correlation was strong at 722~728 nm, with a negative correlation coefficient of 0.997; Therefore, the 722~728 nm band can be used as an indicator band for the P content in the leaves of greenhouse Long pepper during the initial flowering period. The K content was negatively correlated with its spectral reflectance in the visible band of 310~753 nm, with a strong negative correlation in the 721~727 nm band, with a negative correlation coefficient of 0.887. The 754~800 nm band showed a positive correlation, but the correlation was weak. Therefore, the 721~727 nm band can be used as an indicator band for K content in the leaves of protected Long pepper during the initial flowering period.

Using the same analysis method, during the initial fruiting period, the negative correlation between the N content in the leaves of greenhouse Long pepper and its spectral reflectance was strong in the visible band of 310~505 nm, with a negative correlation coefficient of 1.0, 686~687 nm followed, and a negative correlation coefficient of 0.996; The positive correlation was strong at 792~800 nm, with a positive correlation coefficient of 0.702, indicating that the degree of positive correlation is weaker than that of negative correlation. Therefore, the 310~505 nm band can be used as an indicator band for the N content in the leaves of greenhouse Long pepper during the initial fruiting stage. The P content has a strong positive correlation with its spectral reflectance in the visible range of 511~515 nm, with a positive correlation coefficient of 0.927, 689~691 nm had a positive correlation coefficient of 0.984; The degree of negative correlation was weak, and the maximum negative correlation coefficient was less than 0.5. Therefore, the 689~691 nm band can be used as an indicator band for the P content in the leaves of greenhouse Long pepper during the initial fruiting stage. The positive correlation between K content and its spectral reflectance was strong in the visible band of 395~411 nm, with a positive correlation coefficient of 0.815. The negative correlation was strong in the 616~645 nm band, with a negative correlation coefficient of 0.751. It was evident that the degree of negative correlation was weaker than the degree of positive correlation. Therefore, the 395~411 nm band can be used as an indicator band for K content in the leaves of greenhouse Long pepper during the initial fruiting stage.

During the green ripening period, there was a strong negative correlation between the N content and spectral reflectance of greenhouse Long pepper leaves in the visible band of 347~357 nm, with a negative correlation coefficient of 0.90; The degree of positive correlation was weaker than that of negative correlation. Therefore, the 347~357 nm band can be used as an indicator band for the N content in the green ripening leaves of greenhouse Long pepper. There was a strong negative correlation between the P content in the leaves of greenhouse Long pepper and its spectral reflectance in the visible bands of 312~314 nm, 317~321 nm, 327~321 nm, 332~334 nm, and 358~361 nm, with negative correlation coefficients reaching 0.96, 0.99, 0.99, 0.941, and 0.945, respectively; The positive correlation was weak, with a maximum positive correlation coefficient of 0.622. Therefore, the 317~321 nm and 327~321 nm bands can be used as indicator bands for the P content in the green ripening leaves of greenhouse Long pepper. There was a strong positive correlation between K content and its spectral reflectance in the visible band of 339~340 nm, with a positive correlation coefficient of 0.945; The negative correlation was weak. Therefore, the 339~340 nm band can be used as an indicator band for K content in the green ripening leaves of greenhouse Long pepper.

During the color conversion period, the N content and spectral reflectance of greenhouse Long pepper leaves showed a negative correlation throughout the visible band. In the 510~612 nm and 691~742 nm bands, the negative correlation was strong, with negative correlation coefficients reaching 0.881 and 0.889, respectively. Therefore, the 691~742 nm band can be used as an indicator band for the N content in the leaves of greenhouse Long pepper during the color conversion period. There was a strong positive correlation between the P content and spectral reflectance of greenhouse Long pepper leaves in the visible range of 310~376 nm, with a positive correlation coefficient of 0.771; The negative correlation was weak, with a maximum negative correlation coefficient of 0.43. Therefore, the 310~376 nm band can be used as an indicator band for the P content in the leaves of greenhouse Long pepper during the color transition period. The positive correlation between K content and spectral reflectance of greenhouse Long pepper leaves in the visible band was weak, with the highest positive correlation coefficient less than 0.2, and the strong negative correlation coefficient between 438~505 nm, 658~682 nm band, with negative correlation coefficients reaching 0.771 and 0.86, respectively. Therefore, the 658~682 nm band can be used as an indicator band for K content in the leaves of greenhouse Long pepper during the color conversion period.

During the red ripening period, the N content in the leaves of greenhouse Long pepper showed a negative correlation with their spectral reflectance throughout the visible band, with strong negative correlation in the 422~435 nm and 718~800 nm bands, with negative correlation coefficients reaching 0.915 and 0.971. Therefore, the 718~800 nm band can be used as an indicator band for the N content in the leaves of greenhouse Long pepper during its red ripening period. The P content and its spectral reflectance showed a negative correlation throughout the visible band, with a strong negative correlation in the visible band of 424~539 nm, with a negative correlation coefficient of 0.994. Therefore, the 424~539 nm band can be used as an indicator band for the P content in the leaves of greenhouse Long pepper during its red ripening period. The K content exhibited a strong negative correlation with its spectral reflectance in the 310~370 nm wavelength range, with a negative correlation coefficient of 0.845; There was a strong positive correlation in the 679~704 nm band, with a positive correlation coefficient of 0.772. Therefore, the 310~370 nm band can be used as an indicator band for K content in the leaves of greenhouse Long pepper during its red ripening period.

1.3.2 The correlation between N, P, K content and spectral reflectance of different bands under different fertilizer reduction conditions with conventional fertilization as the control (T1, T5, T7, T8)

According to the results (Figure 6), under different fertilizer reduction research conditions with conventional fertilization as the control, the N content in the leaves of greenhouse Long pepper during the initial flowering period showed a stronger positive correlation than a negative correlation in the visible band, and a stronger positive correlation in the 672~675 nm band, with a positive correlation number of 0.932. Therefore, the 672~675 nm band can be used as an indicator band for the N content in the leaves of greenhouse Long pepper during the initial flowering period. The negative correlation between P content and its spectral reflectance was stronger than positive correlation in the visible band, with strong negative correlation in the 662~665 nm and 678~679 nm bands, and negative correlation coefficients reaching 0.988 and 0.944, respectively. Therefore, the 662~665 nm band can be used as an indicator band for the P content in the leaves of greenhouse Long pepper during the initial flowering period. The positive and negative correlation trend between K content and its spectral reflectance in the visible band was consistent with P content, with a stronger negative correlation than positive correlation. The negative correlation was stronger in the 661~664 nm and 678~680 nm bands, with negative correlation coefficients reaching 1 and 0.898, respectively. Therefore, the 661~664 nm band can be used as an indicator band for K content in the leaves of greenhouse Long pepper during the initial flowering period.

During the initial fruiting period, the N content in the leaves of greenhouse Long pepper had weak positive and negative correlation with its spectral reflectance in the visible band, and the correlation coefficient was also below 0.5. Therefore, it was not possible to screen out sensitive bands. The P content had a strong positive correlation with its spectral reflectance in the 428~510 nm and 658~687 nm bands, with positive correlation coefficients reaching 1.0 and 0.994, respectively; Negative correlation was weaker than positive correlation. Therefore, the 428~510 nm band could be used as an indicator band for the P content in the leaves of greenhouse Long pepper during the initial fruiting stage. The K content had a strong negative correlation with its spectral reflectance in the 736~800 nm band, with a negative correlation coefficient of 0.72; The maximum positive correlation coefficient in the 310~422 nm band was 0.45. Therefore, the 736~800 nm band could be used as an indicator band for K content in the leaves of greenhouse Long pepper during the initial fruiting stage.

During the green ripening period, the N and K content in the leaves of greenhouse Long pepper showed a negative correlation with their spectral reflectance in the visible band, while the N content showed a strong negative correlation with their spectral reflectance in the 526~569 nm and 698~736 nm bands, with negative correlation coefficients reaching 0.925 and 0.963, respectively. Therefore, the 698~736 nm wavelength band could be used as an indicator band for the N content in the green ripening leaves of greenhouse Long pepper. There was a strong negative correlation between K content and its spectral reflectance in the 495~661 nm and 688~716 nm bands, with negative correlation coefficients reaching 0.999 and 1, respectively. Therefore, the 688~716 nm wavelength band could be used as an indicator band for K content in the green ripening leaves of greenhouse Long pepper. The positive correlation between P content and its spectral reflectance was stronger than negative correlation in the visible band, and showed a strong positive correlation in the 666~686 nm band, with a correlation coefficient of 0.982. Therefore, the 666~686 nm wavelength band could be used as an indicator band for the P content in the green ripening leaves of greenhouse Long pepper.

During the color conversion period, the N content and spectral reflectance of greenhouse Long pepper leaves showed a stronger negative correlation than a positive correlation in the visible band, and a stronger negative correlation in the 729~800 nm band, with a negative correlation coefficient of 0.958. The positive correlation was in the 374~379 nm band, with the highest correlation coefficient of only 0.408. Therefore, the 729~800 nm band could be used as an indicator band for N content in leaves of greenhouse Long pepper during the color conversion period. The positive correlation between P content and its spectral reflectance was stronger than negative correlation in the visible band, showing a strong positive correlation in the 675~685 nm band, with a negative correlation coefficient of 0.825. Therefore, the 675~685nm band could be used as an indicator band for the P content in the leaves of greenhouse Long pepper during the color conversion period. The K content and its spectral reflectance showed a negative correlation throughout the visible band, with a strong negative correlation in the 747~800 nm band, with a correlation coefficient of 0.935. Therefore, the 747~800 nm band can be used as an indicator band for K content in the leaves of greenhouse Long pepper during the color conversion period.

During the red ripening period, the N content in the leaves of greenhouse Long pepper had a stronger positive correlation than a negative correlation with its spectral reflectance in the visible band, with a stronger positive correlation in the 642~695nm band and a positive correlation coefficient of 0.955. Therefore, the 642~695 nm band could be used as an indicator band for N content in the leaves of greenhouse Long pepper during its red ripening period. The negative correlation between P content and its spectral reflectance was stronger than positive correlation in the visible band, with strong negative correlation in the 535~559 nm and 727~800 nm bands, with correlation coefficients reaching 0.847 and 0.84, respectively. Therefore, the 535~559 nm wavelength band could be used as an indicator band for the P content in the leaves of greenhouse Long pepper during its red ripening period. The negative correlation between K content and its spectral reflectance was stronger than positive correlation in the visible band, and the negative correlation was stronger in the 772~800 nm band, with a negative correlation coefficient of 0.903. Therefore, the 772~800 nm band could be used as an indicator band for K content in the leaves of greenhouse Long pepper during its red ripening period.

1.3.3 The correlation between N, P, K content and spectral reflectance of different bands under the same nitrogen fertilizer dosage based on the same phosphorus and potassium dosage (T3, T4, T5, T6)

According to the results (Figure 7), the study of nitrogen fertilizer application based on the same amount of phosphorus and potassium showed that during the initial flowering period, the correlation between the P and K contents of the leaves of greenhouse Long pepper and their spectral reflectance in the visible band showed a consistent trend, showing a positive correlation stronger than a negative correlation. The positive correlation between P content and spectral reflectance was strong in the 708~719 nm band, with a correlation coefficient of 0.996. The positive correlation between K content and spectral reflectance was strong in the 718~726 nm band, with a correlation coefficient of 0.99. Therefore, the 708~719 nm and 718~726 nm bands could be used as indicator bands for P and K content in the leaves of greenhouse Long pepper during the initial flowering period. The N content and spectral reflectance of greenhouse Long pepper leaves exhibited an opposite phenomenon to P and K in the visible band, with a stronger negative correlation than a positive correlation. The negative correlation was stronger in the 666~677 nm band, with a negative correlation coefficient of 0.915. Therefore, the 666~677 nm band could be used as an indicator band for the N content in the leaves of greenhouse Long pepper during the initial flowering period.

During the initial fruiting period, the correlation between the P and K contents in the leaves of greenhouse Long pepper and their spectral reflectance in the visible band showed opposite trends. The negative correlation between the P content and its spectral reflectance in the visible band was stronger than the positive correlation, and the negative correlation was stronger in the 670~684 nm band, with a negative correlation coefficient of 0.874. Therefore, the 670~684 nm band could be used as an indicator band for the P content in the leaves of greenhouse Long pepper during the initial fruiting stage. The positive correlation between K content and its spectral reflectance was stronger than negative correlation in the visible band, and the positive correlation was stronger in the 654~686 nm band, with a positive correlation coefficient of 0.996. Therefore, the 654~686 nm band could be used as an indicator band for K content in the leaves of greenhouse Long pepper during the initial fruiting stage. The negative correlation between N content and its spectral reflectance was strong in the visible band of 654~662 nm, with a negative correlation coefficient of 0.834; The 788~800 nm band had a strong positive correlation, with a positive correlation coefficient of 0.724; Therefore, the 654~662 nm band could be used as an indicator band for the N content in the leaves of greenhouse Long pepper during the initial fruiting stage.

During the green ripening period, the correlation trend between N and P content and their spectral reflectance in the visible band was relatively consistent, while it was opposite to K content. The negative correlation between N content and its spectral reflectance was strong in the visible band of 310~314 nm, with a negative correlation coefficient of 0.977; The 511~722 nm band had a strong positive correlation, with a positive correlation coefficient of 0.962; It could be seen that the negative correlation was stronger than the positive correlation, therefore, the 310~314 nm band could be used as an indicator band for the N content in the green ripening stage leaves of greenhouse Long pepper. The positive correlation between P content and its spectral reflectance was stronger than negative correlation in the visible band, and the positive correlation was stronger in the 409~497 nm band, with a positive correlation coefficient of 0.959. Therefore, the 409~497 nm band could be used as an indicator band for the P content in the green ripening leaves of greenhouse Long pepper. The K content had a strong positive correlation with its spectral reflectance in the visible band 311~313 nm, with a positive correlation coefficient of 0.969; The 507~736nm band had a strong negative correlation, with a negative correlation coefficient of 0.979; It could be seen that the negative correlation was stronger than the positive correlation, therefore, the 507~736 nm band could be used as an indicator band for K content in the green ripening leaves of greenhouse Long pepper.

During the color conversion period, there was a strong positive correlation between N content and its spectral reflectance in the visible bands of 311~314 nm, 324~327 nm, 332~335 nm, 352~354 nm, 366~367 nm, and 378~381 nm, with positive correlation coefficients reaching 0.935, 0.957, 0.912, 0.914, 0.921, and 0.929; The visible band of 520~576 nm had a strong negative correlation, with a negative correlation coefficient of 0.986; Therefore, it could be seen that the negative correlation was strong, and the 520~576 nm band could be used as an indicator band for the N content in the leaves of greenhouse Long pepper during the color conversion period. The negative correlation between P content and its spectral reflectance was strong in the visible band, with a negative correlation coefficient of 0.744 in the 751~800 nm band. Therefore, the 751~800 nm band could be used as an indicator band for the P content in the leaves of greenhouse Long pepper during the color conversion period. The negative correlation between K content and its spectral reflectance was strong in the visible band, with negative correlation coefficients reaching 0.776, 0.725, and 0.719 in the 501~514 nm, 632~662 nm, and 669~684 nm bands. Therefore, the 501~514 nm band can be used as an indicator band for K content in the leaves of greenhouse Long pepper during the color conversion period.

During the red ripening period, the correlation trend between N and P content and their spectral reflectance in the visible band was relatively consistent. The positive correlation between N content and its spectral reflectance was strong in the visible band at 310~317 nm, 330~360 nm, 366~371 nm, with positive correlation coefficients reaching 0.94, 0.949, and 0.934. The negative correlation was strong in the 716~742 nm band, with the highest negative correlation coefficient reaching 0.73. Therefore, the 330~360 nm band could be used as an indicator band for the N content in the leaves of greenhouse Long pepper during its red ripening period. The P content had a strong positive correlation with its spectral reflectance in the visible band of 310~375 nm, with a positive correlation coefficient of 0.959; The visible band at 737~800 nm had a strong negative correlation, with a negative correlation coefficient of 0.873. Therefore, the 310~375 nm band could be used as an indicator band for the P content in the leaves of greenhouse Long pepper during its red ripening period. The K content and its spectral reflectance showed a positive correlation in the visible band, with a strong positive correlation in the 378~381 nm band, with a positive correlation coefficient of 0.764. Therefore, the 378~381 nm band could be used as an indicator band for K content in the leaves of greenhouse Long pepper during its red ripening period.

1.4 Comparison of nitrogen, phosphorus, and potassium content and yield in leaves of different fertilizer treatments and development stages of peppers

The results showed (Figure 8) that the nitrogen, phosphorus, and potassium contents in the leaves showed a pattern of initial fruiting stage>initial flowering stage>green ripening stage>color conversion stage>red ripening stage. During the initial fruiting stage, green ripening stage, color conversion stage, and red ripening stage, the leaves all showed potassium content>nitrogen content>phosphorus content.

2 Discussion

Different fruit development stages and fertilizer treatments resulted in different spectral characteristics of the leaves of greenhouse Long pepper, and the spectral reflectance of the leaves showed a basically consistent pattern. This was consistent with the spectral reflectance results of the leaves of Armeniaca Vulgaris ‘Luntaibaixing’ and Brassica rapa L. (Hu et al., 2013; Zhuang et al., 2016). Different fruit development stages, optimized fertilization treatment with a reduction of 30% and optimized fertilization treatment with a reduction of 45% showed lower spectral reflectance and higher chlorophyll content in the leaves of greenhouse Long pepper. The treatment with a 30% reduction in total chemical fertilizer dosage resulted in the highest yield of peppers, and the green ripening and color conversion stages were sensitive periods for N, P, and K nutrient diagnosis in this treatment. The bands of 347~357 nm, 317~321 nm, 327~321 nm, 339~340 nm can be used as indicator bands for N content, P content, and K content in green ripening leaves of greenhouse Long pepper; The 691~742 nm, 310~376 nm, and 658~682 nm bands can be used as indicator bands for N content, P content, and K content in the leaves of greenhouse Long pepper during the color conversion period.

Spectrometers are more sensitive to identifying nitrogen elements, followed by potassium and phosphorus. Therefore, there was limited research on using spectral reflectance to estimate P and K elements (Wang et al., 2007; Wang and Bai, 2007; Zhang et al., 2010). This study investigated the sensitive bands and sensitive periods of P and K elements in greenhouse Long pepper, which to some extent enriched the diagnostic examples of P and K elements.

The content of nitrogen, phosphorus, and potassium in leaves showed a pattern of initial fruiting stage>initial flowering stage>green ripening stage>color conversion stage>red ripening stage. During the initial fruiting stage, green ripening stage, color conversion stage, and red ripening stage, leaves all showed potassium content>nitrogen content>phosphorus content.

3 Materials and Methods

3.1 Experimental design

On February 28, 2018, a 'self pressure drip irrigation under film' experiment was conducted in a sunlight greenhouse at the Anning Canal Experimental Site in Urumqi to reduce the application of fertilizer in early spring cropping for greenhouse Long pepper. The experiment used a single factor randomized block design and set up 8 treatments: based on the conventional fertilization of greenhouse Long pepper in Xinjiang, the experiment was designed with four nitrogen fertilizer levels: conventional fertilization and N deficiency, optimized fertilization and N deficiency, and optimized fertilization conditions for N0, N1, N2, and N3. The optimized fertilization was reduced by 15%, 30%, and 45%. Each treatment was set up with 3 replicates, 24 plots in total, randomly arranged, with an area of 10 m² per plot (Table 1).

March 10th: raising ridges with soil moisture and installing the 'self pressure drip irrigation' water and fertilizer management system; March 22nd: final planting; April 13th: hardening of seedling and watering for the first time; July 13th: strike seedlings; The entire experimental period lasted for 136 d, with watering 16 times and an irrigation volume of 202 m³/667 m²; Drip irrigation and topdressing 10 times during the whole course; The first harvest was on May 2nd, and the last harvest was on July 12th, totaling 7 harvests.

3.2 Test materials

From April to August 2018, three healthy plants with the same size were selected for each fertilizer treatment. Healthy mature leaves were randomly selected from the middle, East, West, South and North, and 12 leaves were selected for each plant.

3.3 Spectral data acquisition

The spectral data was collected by UniSpec-SC portable spectral analyzer (produced by PP Systems in the United States). This model of spectral analyzer has its own light source, which can carry out continuous measurement in the wavelength range of 310~1 130 nm, and the spectral resolution is within λ/100, with a scanning wavelength of 3.3 nm (Zhuang et al., 2016). The experimental spectral data acquisition includes five developmental stages of greenhouse Long pepper, namely the initial flowering stage, initial fruiting stage, green ripening stage, color conversion stage, and red ripening stage. In selected sunny or windless weather, at the time 10:30-13:00 of Beijing, the healthy leaves in vivo were measured for three times. The middle and upper parts of the leaves (avoiding the probe directly facing the leaf vein) were selected for spectral measurement. Standard correction was carried out before each data acquisition. The spectral reflectance of leaves in visible band was used to analyze the spectral characteristics and sensitivity to N, P, K elements of greenhouse Long pepper leaves at different growth stages under the condition of reduced fertilizer application (Yang et al., 2013).

3.4 Leaf acquisition

In 2018, at the greenhouse of the Horticultural Crop Research Institute of Xinjiang Academy of Agricultural Sciences (43°58′48.71″ N and 87°30′6.52″ E), the middle leaves of pepper shoots were collected at different developmental stages (initial flowering, initial fruiting, green ripening, color conversion, and red ripening stage); 3 plants for each treatment were selected, and healthy growing leaves were randomly selected from the East, West, South, and North; 3 leaves (with petioles) were selected form each plant; Sterilized at a constant temperature of 105 °C for 30 min, then baked at 70 °C to constant weight, crushed, and packed into a self sealing bag for analysis of nitrogen, phosphorus, and potassium content.

3.5 Data analysis

The total nitrogen content of leaves was determined by Kjeldahl method; The content of total phosphorus in leaves was measured by vanadium aluminum yellow method at 450 nm of Ultraviolet-visible Spectrophotometer(UV-1800); The total potassium content of leaves was determined by atomic absorption spectrophotometer (PE-Analysist100).

The data analysis was conducted by using DPSv9.5 statistical software and Origin9.0 software (Zhuang et al., 2016).

Authors’ contributions

ZHM and LPF are the executors of this experimental design and experimental study; ZHM and LHF completed the data analysis and and wrote the first draft of the paper; FGP and LP participated in the experimental design and analysis of the experimental results; ZHM, WQ and WH are the project conceptualizers and leaders, guiding the experimental design, data analysis, paper writing and revision. All authors read and approved the final manuscript.

Acknowledgments

This study was funded by the National Key R&D Program "Establishment and Demonstration of Chemical Fertilizer and Pesticide Reduction Technology Mode for Greenhouse Vegetables in Northwest Arid Areas" (2016YFD0201005).

References

Bronson K.F., Chua T.T., and Booker J.D., 2003, In-season nitrogen status in sensing irrigated cotton：Ⅱ．Leaf nitrogen and biomass, oil Science Society of American Journal, 67: 1439-1448.

https://doi.org/10.2136/sssaj2003.1439

Han X.P., Zuo Y.M., and Li L.Z., 2010, Study on the near infrared spectral prediction model of hydroponic tomatoes, Guangpuxue yu Guangpu Fenxi (Spectroscopy and Spectral Analysis), 30(9): 2479-2483.

Hu H., Bai Y. L., Yang L. P., Lu Y. L., Wang L., Wang H., and Kong Q.B., 2009, Red edge parameters of winter wheat canopy under different nitrogen levels, Zhiwu Yingyang yu Feiliao Xuebao (Journal of Plant Nutrition and Fertilizer), 15(6): 1317-1323.

Hu Z.Z., Pan C.D., Wang S.W., Guo Z.C., Wang Q.T., Ding F, and Li Y., 2013, Models for estimating foliar NPK content of Armeniaca Vulgaris ‘ Luntaibaixing’ using spectral reflectance, Xinjiang Nongye Kexue (Xinjiang Agricultural Sciences), 50(2): 238-248.

Jiang H.Y., Peng Y.S., Xie L.J., and Ying Y.B., 2008, Studies on impaction of scan times of tomato leaf diffuse reflection spectrum and model precision, Guangpuxue yu Guangpu Fenxi (Spectroscopy and Spectral Analysis), 28(8): 1763-1766.

Li L.Z., Guo R., Li H.P., and Cao Y., 2010, Effects of nitrogen concentration in hydroponics on growth and development of tomato and spectral characteristics of leaf in greenhouse, Zhiwu Yingyang yu Feiliao Xuebao (Journal of Plant Nutrition and Fertilizer), 16(4): 965-969.

Li X.F., Han T.T., Dong Y., Wu M., and Shen X., 2011, Relationships between spectral reflectance and pigment or nitrogen concentrations in leaves of Padus Virginiana’Schubert’, Linye Kexue (Scientia Silvae Sinicae), 47(8): 75-81.

Liang H.P., and Liu X.N., 2010, Hyperspectral calculation model of corn nitrogen nutrition index, Nongye Gongcheng Xuebao (Transactions of the CSAE), 26(1): 250-255.

Liu H.B., Zhang Y.G., Li Z.H., Zhang C.Y., and Hu D.Y., 2004, Application of spectroscopic techniques in nitrogen nutrition disputes in winter wheat, Zhongguo Nongye Kexue (Scientia Agricultura Sinica), 37(11): 1743-1748.

Luo X.N., and Zhu Y.J., 2014, Effect of nitrogen application rate on SPAD value of cotton leaf position and diagnosis of cotton nitrogen nutrition, Ganhang Diqu Nongye Yanjiu (Agricultural Research in Arid Regions), 32(1): 128-133.

Ma L.L., Lv X., Zhang Z., MA G.X., and Hai X.Y., 2016, Diagnosis model of nitrogen nutrition in drip irrigation cotton based on critical Nitrogen concentration, Nongye Jixie Xuebao (Chinese Journal of Agricultural Machinery), 49(2): 277-283.

Racy M.B., James S.S., and Garye V., 1996, Nitrogen deficiency detection using reflected short-wave radiation from irrigated corn canopies, Agronomy Journal, 88: 1-5.

https://doi.org/10.2134/agronj1996.00021962008800010001x

Shi H.G., Wang H.H., Lv X., LI X.W., Zheng Q., LI W.T., Yu L., and Shao Q., 2014, Diagnostic study of cotton nitrogen nutrition based on Green Seeker spectroscopy, Zhongguo Turang yu Feiliao (China Soil & Fertilizer), (5): 80-85.

Song L.J., Ye W.J., and Zheng Y.Y., 2017, Research progress of crop nitrogen nondestructive rapid nutrition diagnosis, Zhongguo Daomi (China Rice), 23(6): 19-22.

Sun H., Li M.Z., Zhang Y.E., Zhao Y., and Wang H.H., 2010, Spectral characteristics of corn under different nitrogen treatments, Guangpuxue yu Guangpu Fenxi (Spectroscopy and Spectral Analysis), 30(3): 715-719.

Tan X., Wang S.H., and Xue L.H., 2011, Nitrogen nutrition diagnosis of early rice with NDVI and its application for nitrogen top dressing recommendation at Yingtan, Jiangxi Province, Zhongguo Nongye Kexue (Scientia Agricultura Sinica), 44(4): 691-698.

Tang M.Y., Zhang Y., Hu W., Hu G.Z., and Li Q.J., 2015, Study on nitrogen fertilizer use and nutritional diagnosis of processed tomato, Zhongguo Turang yu Feiliao (China Soil & Fertilizer), (4): 82-87.

Wang K.R., Pan W.C., Li S.K., Chen B., Xiao H., Wang F.Y., and Chen J.L., 2011, Study on nitrogen concentration in height spectral characteristics of cotton crow, Guangpuxue yu Guangpu Fenxi (Spectroscopy and Spectral Analysis), 31(7): 1868-1872.

Wang L., and Bai Y.L., 2007, Nutrients change and spectral response of spring corn leaf for varying amounts of potassium fertilization, Yaogan Xuebao (Journal of Remote Sensing), 11(5): 641-647.

Wang L., Bai Y.L., Lu Y.L., Wang H., and Yang L.P., 2011, Nitrogen nutrition diagnosis for corn based on spectral analysis, Zhiwu Yingyang yu Feilao Xuebao (Journal of Plant Nutrition and Fertilizer), 17(2):333-340.

Wang L., Bai Y.L., and Yang L.P., 2007, Spectral response and diagnosis of phosphorus nutrition in corn, Zhiwu Yingyang yu Feilao Xuebao (Journal of Plant Nutrition and Fertilizer), 13(5): 802-808.

Wang Y.N., Zhao L.Y., Zhang A.Q., and Liu N., 2014, Nutritional diagnosis of peanut hyperspectral leaf, Jiangsu Nongye Kexue (Jiangsu Agricultural Sciences), 42(12): 129-131.

Xue L.H., Cao W.X., Luo W.H., Jiang D., Meng Y.L., and Zhu Y., 2003, Diagnosis of nitrogen status in rice leaves with the canopy spectral reflectance, Zhongguo Nongye Kexue (Scientia Agricultura Sinica), 36(7): 807-812.

Yang H.L., Chen G., and Wu J.F., 2011, Plant nitrogen content of annual ryegrass and spectral reflectance response to nitrogen application level, Caoye Xuebao (Acta Prata Culturae Sinica), 20(3): 239-244.

Yang H., Cao H.X., Liu M.Y., and Liu S.H., 2015, Establishment of critical nitrogen concentration model and nitrogen nutrition diagnosis of tomato under water-nitrogen coupling condition, Zhiwu Yingyang yu Feilao Xuebao (Journal of Plant Nutrition and Fertilizer), 21(5): 1234-1242.

Yang J.B., Du C.W., Shen Y.Z., and Zhou J.M., 2013, Characteristics of middle infrared spectrum of Chinese cabbage and its application in nitrogen nutrition diagnosis, Zhiwu Yingyang yu Feilao Xuebao (Journal of Plant Nutrition and Fertilizer), 19(4): 1012-1017.

Zhang B.H., and Zhang J.B., 2010, Response of the spectral reflectance to total N and NO₃-N of winter wheat, Ganhan Diqu Nongye Yanjiu (Agricultural Research in the Arid Areas), 28(1): 104-110.

Zhang J.H., Zhang J.B., and Qin S.W., 2010, Spectral reflectance characteristics of summer maize under long-term fertilization, Zhiwu Yingyang yu Feiliao Xuebao (Journal of Plant Nutrition and Fertilizer), 16(4): 874-879.

Zhao W., Zhang Z.P., and Miao M.M., 2018, Establishment of critical nitrogen concentration model in tomato and nitrogen nutrition diagnosis, Zhongguo Nongxue Tongbao (Chinese Agricultural Science Bulletin), 34(22): 37-44.

Zhou L.L., Feng H.Y., Yan Z.M., Liu K., and Zhou S.L., 2010, High spectral estimation and varieties of maize leaf nitrogen content, Nongye Gongcheng Xuebao (Transactions of the CSAE), 26(8): 195-199.

Zhu W.J., Mao H.M., Zhou Y., and Zhang X.D., 2014, Diagnosis of nitrogen nutrition in tomato leaves based on hyperspectral imaging technology, Jiangsu Daxue Xuebao (Journal of Jiangsu University (Natural Science Edition)), 35(3): 290-294.

Zhuang H.M., Lu C.S., Gong P., Xu Y.T., Xie H., and Fan D.Y., 2016, Nitrogen, Phosphorus and potassium spectral characteristics of leaves of 'Ye Erqiang' almond fruit at different growth stages, Zhiwu Yingyang yu Feiliao Xuebao (Journal of Plant Nutrition and Fertilizer), 22(4): 1079-1090.