Digital imaging of coffee leaves under different nitrogen concentrations applied to soil

ABSTRACT

Brazil is the largest Coffea arabica L. producer and exporter in the world market. In view of the need for nitrogen fertilization, the objective of this work was to evaluate the use of digital images of coffee leaves as a function of nitrogen concentration and cultivars, as well as to evaluate the most adequate sampling of this technique to predict leaf nitrogen. The experiment was set in a greenhouse, at UNESP - Experimental Campus of Registro-SP, using 12L pots with medium sand. A completely randomized design was adopted, in a 4 x 2 factorial scheme, with four concentrations of N (0, 50, 100 and 200 mg dm^-3), and two coffee cultivars (Mundo Novo and Obatã), with ten replications. The variables evaluated were: hue, dark green color index (DGCI), leaf area and leaf nitrogen content. The hue and DGCI were influenced by N concentration, cultivar and are dependent on the day after treatment application and the number of leaves used. The use of five leaves per replication was more adequate to estimate leaf N content. Hue, DGCI and leaf area were higher in Obatã depending on the day after application of the treatment and N concentration.

Keywords
hue; dark green color index; Coffea arabica ; leaf nitrogen

INTRODUCTION

Brazil is the largest Coffea arabica L. producer and exporter in the world market, with an yield around 63.08 million bags of Arabica coffee in 2020 (Conab, 2020). In this context, nitrogen (N) is a macronutrient of fundamental importance, for being one of the elements most required by plants, due to its importance in the production of new cells and tissues (Santos et al., 2016). The content of this nutrient directly interferes with the growth, development and production of crops (Silva et al., 2018).

The low availability of N significantly alters the assimilation capacity of CO₂, affecting the production of photoassimilates and the growth rate as a whole, compromising the structural characteristics of the plants (Gimenes et al., 2017). Excess of nitrogen in the soil is also harmful for plants due to the imbalance and antagonism caused to other nutrients (Fagundes, 2011). The correct concentration of nutrients is essential to minimize losses and achieve good results for the crop. Fertilization with N has high mobility in the soil, requiring frequent monitoring of its availability to plants (Oliveira et al., 2020).

The monitoring of the crop and the diagnosis of N in the field mobile devices that estimate chlorophylls a and b and others devices that simultaneously capture the image and establish the RGB color indices, generating an estimate of leaf N, provide promising results (Wang et al., 2014). Show an association of chlorophyll with available nitrogen contents (Santos & Castilho, 2015), this correlation being associated with the hue and green color of the leaves.

In this context, digital image processing is a plausible technological option, capable of predicting the nutritional status of N in a simple and fast way, in comparison with laboratory analyses, based on the parameters of hue and dark green color index (DGCI) (Felisberto et al., 2016). Given the above, the objective of this work was to evaluate the use of digital images of coffee leaves as a function of nitrogen concentration and different cultivars, as well as to evaluate the most adequate sampling of this technique to predict leaf nitrogen.

MATERIAL AND METHODS

The experiment was set on January 5, 2015, in a greenhouse, at the Experimental Campus of Universidade Estadual Paulista, located in the city of Registro-SP, Southeastern Brazil (24º29'15" S, 47º50'37" W), with an average altitude of 25 m.

The experiment was carried out in a completely randomized design, in a 4 x 2 factorial scheme with four concentrations of nitrogen (0, 50, 100 and 200 mg dm^-3) and two coffee (Coffea arabica L.) cultivars (Mundo Novo and Obatã), with ten replicates for each treatment, totaling 80 plots. Each experimental plot contained a coffee seedling. Urea was used as a source of N, applied in coverage, divided into four plots, with an interval of 30 days between applications; the first was carried out on January 20, 2015. River sand was used as a substrate to conduct the experiment, and corrections were then made. Base saturation was increased to 50% using dolomitic limestone (65% PRNT) and the levels of P and K, for 200 and 100 mg dm^-3, using simple superphosphate and potassium chloride, respectively.

Coffee seedlings produced in bags were transplanted to containers measuring 30 x 20 x 26 cm; upper, lower diameter and height, respectively, with 12 L of sand, where the treatments were applied. The pots were arranged on top of pallets, with spacing of 30 cm between plants and of approximately 80 cm between rows. Irrigation was by micro sprinklers and the ideal amount of water to conduct the experiment was based on the field capacity (FC) of the substrate, adopting 70% FC, reaching a value of 350 ml a day.

Digital images were obtained on the 60 and 90 DAA, from leaf newly expanded on valuation date. They were obtained with the use of a digital camera (Cybershot DSC-H200 Sony 20. Megapixels, Japan), with a resolution of 20 MP, without the flash, using the macro function. The image background was pink. The distance between the camera and the sheet was also standardized at 20 cm and the photos were taken from 9:00 am to 11:00 am. The images were read by RGB analysis, with the aid of Corel Photo Paint^® (Corel Draw Graphics Suite X4, COREL Corporation, Canada). In the case of an image that has information in different frequency intervals, an f (x, y) function is required for each band. This is the case of standard RGB color images, which are formed by the information of primary additive colors, such as red (R - Red), green (G - Green) and blue (B - Blue) (Filho & Neto 1999). Calculations defining by the RGB hue terms: (Paulidis & Liow, 1990) and dark green colors index (Wang et al., 2013).

In the last evaluation, 90 DAA of the treatment, the leaves from which the digital images were obtained, were collected, dried in a circulation oven air renewal for drying at 65 ^°C, for 72 hours, to determine the dry phytomass of the leaves, ground and sent to the laboratory for the determination of N content, from acid digestion and distillation (Kjeldahl) according to the methodology of Malavolta et al. (1997). Concerning the RGB, the green and circular cards used as color standards, known areas; therefore, with the aid of reading the pixels of the leaves and, it was possible to determine the leaf area. In the last leaf collection, a photo of one leaf was taken by replication (ten per treatment) and, subsequently, of a sample of five leaves per replication (fifty per treatment), in order to compare which is the most appropriate methodology. The images were all cropped in Windows Paint^® (Microsoft, EUA) and then, with the help of Corel Photo Paint^®, in the mask function, later on, color mask, it was possible to redefine the color of the leaves. The amount of RGB of each leaf was read in the image function and, later, a histogram of the same tool, in other words, the histogram of the distribution of RGB values ​​of the image of the leaves is a tool of Corel Photo Paint and, through this tool, it is obtained the average value of R, G and B of the image.

The green and yellow cards, according to the literature, are of known colors, thus enabling a correction factor and standardizing the photos, regardless of brightness or time of the sections (Rorie et al., 2011). The reference colors were selected in Corel Draw^® and the cards were printed and cut out.

The results were submitted to analysis of variance, regression adjustment by linear and quadratic models, and linear correlation using the Sisvar v program. 5.3. (Ferreira, 2011). The average values were compared by the test Tukey to 5% of significance. For the correlation, it was considered as variable x the content of leaf N and as variable y, the characteristics evaluated in the leaf image (hue, DGCI and leaf area) of one leaf or five leaves per repetition. As for the determination of the leaf N content, the samples of the repetitions were grouped two by two, for the correlation the other indexes were also grouped two by two, totaling 40 data for each variable.

RESULTS AND DISCUSSION

First evaluation date (60 days after application of the treatment – DAA)

For hue, there was an effect only for N concentrations. There was no significant interaction between the variation factors. Both cultivars showed equal responses to nitrogen concentrations (Table 1). The hue of the leaf on the cultivar Mundo Novo, plants showed a linear increase in relation to N concentrations, where the highest index was found at 200 mg dm^-3 (116 degrees), while the hue of Obatã shows a quadratic adjustment; the maximum point can be established, presenting the highest hue value (113 degrees) at 179 mg dm^-3 N. For the DGCI, there was an effect only for N concentrations. There was no significant interaction between factors. Both cultivars showed equal responses to nitrogen concentrations (Table 2). Both Mundo Novo and Obatã showed linear growth, where the highest DGCI observed was at 200 mg dm^-³ N. The behavior of linear regression is an indication that the increase in N concentrations improves the level of the DGCI in the cultivars. Hue and DGCI had similar behavior, with no distinction between cultivars and differences in N concentrations, in relation to the evaluation time of 60. That the application of concentrations of N influence the green color of leaves, due to the increase in the of chlorophyll, the higher concentration of N in tissues (Groff et al., 2020). Which implies predicting the correlation of hue and DGCI with leaf N availability.

Table 1
Summary of analysis of variance, regression analysis and comparison of hue averages, as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3), in Mundo Novo and Obatã coffee seedlings. Sixty days after application of the treatments (DAA). Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.5192 < 0.001 0.8032
CV, %		16.22
		Hue (degrees)
Mundo Novo Obatã		94.98 a 97.24 a
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	75.1 A 80.8 A 108.5 A 115.6 A	79.1 A 83.9 A 113.9 A 112.5 A	77.1 82.36 111.2 113.8
	Pr > t
Regression	0.001 L1	0.033 Q2	0.013 Q3
R²	86.1	83.1	87.9
1 ŷ = 0.2176x + 75.96 2 ŷ = -0.0012x2 + 0.429x + 75.36		3 ŷ = -0.001x2 + 0.407x + 73.67

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q – quadratic.

Table 2
Summary of analysis of variance, regression analysis and comparison of dark green colour index averages (DGCI), as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3), in Mundo Novo and Obatã coffee seedlings. Sixty days after application of the treatments (DAA). Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.3523 < 0.001 0.7638
CV, %		22.86
		Dark green color index
Mundo Novo Obatã		0.53 a 0.55 a
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	0.357 A 0.397 A 0.648 A 0.719 A	0.419 A 0.428 A 0.680 A 0.698 A	0.388 0.412 0.664 0.708
	Pr > t
Regression	< 0.001 L1	< 0.001 L2	0.0449 L3
R²	76.13	86.35	85.79
1ŷ = 0.0020x + 0.3588	3ŷ = 0.0018x + 0.3888
2ŷ = 0.0016x + 0.419

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q – quadratic.

As for the leaf area, there was an effect for coffee cultivars, and Obatã showed an average leaf area 10 cm² higher than Mundo Novo. There was no significant interaction between factors (Table 3). The coefficient of variation of this characteristic was high due to the evaluation of a single leaf per replication. The cultivar Obatã showed a linear increase in leaf area in relation to N concentrations, where the largest average was 62.26 cm² obtained at 200 mg dm^-³, while for Mundo Novo, there was no adjustment, neither linear nor quadratic, with an average of 42.67 cm².

According to Leghari et al. (2016), the increase in leaf area as a function of “soil N” was already expected, as it is a mineral nutrient that is more limiting to plant growth, as it is an integral part of molecules that are fundamental to metabolism. However, as seen in this research (Table 3), the response to nutrient uptake efficiency N can be variable for coffee cultivars.

Table 3
Summary of analysis of variance, regression analysis and comparison of leaf area averages, as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3), in Mundo Novo and Obatã coffee seedlings. Sixty days after application of the treatments (DAA). Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.0375 0.2186 0.1169
CV, %		43.05
		Leaf area (cm²)
Mundo Novo Obatã		42.67 a 52.36 b
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	45.83 A 40.82 A 46.10 A 37.87 A	41.27 A 43.61 A 62.34 A 62.26 A	43.550 42.244 54.218 50.069
	Pr > t
Regression	ns	0.0098 L1	ns
R²	ns	74.09	ns
1ŷ = 0.116x + 42.222

• Means followed by the same letter do not differ by the Tukey test at 5% significance. ns – not significant; L –linear; Q – quadratic.

Second evaluation date (90 days after application of the treatment – DAA)

For hue, there was an effect for cultivar and N concentrations. There was no significant interaction. The two cultivars showed different responses to nitrogen concentrations (Table 4). There was a difference in hue between the two cultivars in the treatments using 50 and 100 mg dm^-³, and Obatã showed higher hue values (Table 4). According to Felisberto et al. (2016), the hue of the leaf digital image was significantly influenced by N concentrations, ranging from 75.1 to 149.7 degrees in studies 51, 79, 93, 107, 121 and 149 days after transplantation, at 5; 10; 15; 20 and 25g of N for the pepper plant. The authors also mention that the lowest hue values corresponded to plants with N deficiency.

Table 4
Summary of analysis of variance, regression analysis and comparison of hue averages, as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3), in Mundo Novo and Obatã coffee seedlings. Ninety days after application of the treatments (DAA)-Sample of one leaves per replication. Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.0006 < 0.001 0.1568
CV, %		7.17
		Hue (degrees)
Mundo Novo Obatã		87.77 a 92.96 b
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	78.38 A 85.52 A 90.26 A 96.94 A	80.1 A 93.44 B 99.37 B 98.95 A	79.24 89.48 94.82 97.95
	Pr > t
Regression	< 0.001 L1	< 0.001 Q2	< 0.001 Q3
R²	96.56	99.62	99.77
1ŷ = 0.09x + 79.9	3ŷ = -0.0007x2 + 0.2232x + 79.435
2ŷ = 0.001x2 + 0.2984x + 80.371

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q –quadratic.

The hue for leaves of Mundo Novo plants showed a linear increase in relation to N concentrations, where the highest hue index was found at 200 mg dm^-³ (97 degrees), while the hue value of Obatã demonstrated a quadratic adjustment, presenting the highest value (99.37 degrees) for hue at 100 mg dm^-³ N. The behavior of the cultivars for the hue parameter were similar for the leaf N content (Table 10). In this sense, the difference in concentrations of N provides different shades of green in the leaf. The difference in cultivar performance and the correlation with the of leaf N is evident. For DGCI, there was an effect for cultivar and N concentrations. There was no significant interaction. The two cultivars did not show equal responses for nitrogen concentrations (Table 5). There was a difference in the DGCI between the two cultivars, at all N concentrations, being higher in Obatã. The DGCI in the two cultivars showed a quadratic adjustment, as a function of N concentrations. The cultivar Obatã obtained a maximum DGCI point with a lower N concentration (150 mg dm^-³ N) in relation to Mundo Novo (200 mg dm^-³ N). Similarly, in the study by Bertani et al. (2019), the DGCI in relation to N concentrations also obtained a quadratic adjustment, with a maximum point of (680 mg dm^-3 N) in passion fruit seedlings. For leaf area, there was an effect for cultivar and N concentrations. There was no significant interaction. Both cultivars showed quadratic adjustment in terms of leaf area as a function of N concentrations, with a very similar maximum point, differing only at 50 mg dm^-³ N (Table 6). There was a difference in leaf area between the two cultivars, and the average leaf area of Obatã (46 cm²), was greater than that of Mundo Novo (39.02 cm²), as verified in the first evaluation. In this study, the cultivar Obatã was the most responsive (Table 6) to nitrogen concentrations and the leaf area followed the quadratic behavior of the leaf N content (Table 10), that is, greater leaf area resulting from higher leaf N content, which is favored by the availability of N in the soil. In the study of Schwers et al. (2016), they also verified the difference in leaf area in response to different plant cultivars as a function of nitrogen fertilizer doses.

Table 5
Summary of analysis of variance, regression analysis and comparison of dark green color index averages, as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3), in Mundo Novo and Obatã coffee seedlings. Ninety days after application of the treatments (DAA)-Sample of one leaves per replication Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.0012 < 0.001 0.5482
CV, %		12.49
		Dark green color index
Mundo Novo Obatã		0.509 a 0.560 b
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	0.406 A 0.489 A 0.554 A 0.590 A	0.433 A 0.556 B 0.631 B 0.621 B	0.420 0.523 0.593 0.606
	Pr > t
Regression	0.0260 Q1	< 0.001 Q2	< 0.001 Q3
R²	99.23	99.99	99.97
1ŷ = -0.000005x2 + 0.00201x + 0.4049	3ŷ = -0.000008x2 + 0.00251x + 0.4192
2ŷ = -0.000010x2 + 0.00301x + 0.4325

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q – quadratic.

Table 6
Summary of analysis of variance, regression analysis and comparison of leaf area averages, as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3), in Mundo Novo and Obatã coffee seedlings. Ninety days after application of the treatments (DAA)-Sample of one leaves per replication. Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.0207 < 0.001 0.5941
CV, %		31.83
		Leaf area (cm²)
Mundo Novo Obatã		39.02 a 46.19 b
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	21.53 A 32.25 A 50.31 A 51.89 A	27.15 A 46.39 B 53.33 A 57.90 A	24.34 39.34 51.86 54.90
	Pr > t
Regression	0.0249 Q1	0.0150 Q2	0.0012 Q3
R²	95.43	98.74	99.66
1ŷ = -0.0011x2 + 0.3827x + 19.9874		3ŷ = -0.0012x2 + 0.3870x + 23.9429
2ŷ = -0.0012x2 + 0.3899x + 27.9024

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q – quadratic.

Data obtained from a sample consisting of five leaves 90 days after application of the treatment – DAA)

For hue, there was an effect for cultivar and N concentration. There was no significant interaction. The cultivars showed different responses to nitrogen concentration (Table 7). There was a difference in hue between the two cultivars at concentrations of 50 and 100 mg dm^-³; Obatã stands out in relation to Mundo Novo one more time. The hue of the leaf from Mundo Novo plants showed a linear increase in relation to N concentrations, where the highest index was found at 200 mg dm^-3 N (108 degrees), while the Obatã demonstrated quadratic adjustment, presenting the highest value (113 degrees) at 153 mg dm^-³ N. The hue of the leaves mainly reflects the possible differences in the contents of leaf N, and, consequently, in the concentrations of chlorophylls, which may influence the photosynthetic capacity of the plant (Flexas et al., 2016). In this context, the hue and DGCI obtained from the digital image have shown to be promising, as they present high relationships with nitrogen concentrations and, consequently, an indicator of available leaf nitrogen.

Table 7
Summary of analysis of variance, regression analysis and comparison of hue averages, as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3) in Mundo Novo and Obatã coffee seedlings. Ninety days after application of the treatments (DAA)-Sample of five leaves per replication. Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.0003 < 0.001 0.1227
CV, %		7.81
		Hue (degrees)
Mundo Novo Obatã		93.01 a 99.36 b
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	78.18 A 87.18 A 98.90 A 107.77 A	80.73 A 96.78 B 110.14 B 109.80 A	79.46 91.99 104.52 108.72
	Pr > t
Regression	0.001 L1	< 0.001 Q2	< 0.001 Q3
R²	95.36	99.43	99.25
1ŷ = 0.14859x + 80.006	3ŷ = -0.000948x2 + 0.3397x + 78.8911
2ŷ = -0.00140x2 + 0.4290x + 80.2120

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q – quadratic.

For DGCI, there was an effect for cultivar and N concentrations. There was no significant interaction. The two cultivars did not show equal responses for nitrogen concentrations (Table 8). There was a difference in the DGCI between the two cultivars at 50 and 100 mg dm^-³ N. The intensity of the leaf green presented by the cultivars, is an important value to estimate the indices of photosynthetic pigments (Conceição et al., 2019), so that changes in the photosynthetic process may have consequences on coffee yield. The two cultivars demonstrate quadratic adjustment, where the maximum point of Mundo Novo was 200 mg dm^-³ N, with DGCI of 0.677, and that of Obatã was 100 mg dm^-³ N, to obtain the DGCI 0.690.

Table 8
Summary of analysis of variance, regression analysis and comparison of dark green color index averages (DGCI), as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3), in Mundo Novo and Obatã coffee seedlings. Ninety days after application of the treatments (DAA)-Sample of five leaves per replication. Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		< 0.001 < 0.001 0.1259
CV, %		10.68
		Dark green color index
Mundo Novo Obatã		0.534 a 0.593 b
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	0.396 A 0.483 A 0.592 A 0.667 A	0.430 A 0.567 B 0.690 B 0.687 A	0.413 0.525 0.641 0.709
	Pr > t
Regression	0.0203 Q1	< 0.001 Q2	< 0.001 Q3
R²	99.07	99.14	99.09
1ŷ = -0.000005x2 + 0.00241x + 0.3903	3ŷ = -0.000007x2 + 0.00295x + 0.4082
2ŷ = -0.000012x2 + 0.00376x + 0.4243

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q – quadratic.

For leaf area, there was an effect for cultivar and N concentrations. There was no significant interaction. Both cultivars showed equal responses to N concentrations (Table 9). There was a difference in leaf area averages between cultivars Mundo Novo and Obatã; the latter showed the largest leaf area, with a gain of 10.65%. The leaf area of Mundo Novo and Obatã showed a quadratic adjustment in relation to N concentrations. In the study of Colodetti et al. (2015), the leaf area of the coffee tree also increased with the concentrations of N, which compromises the efficiency of the photosynthetic rate of the plant, as well as its yield.

Table 9
Summary of analysis of variance, regression analysis and comparison of leaf area averages, as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3) in Mundo Novo and Obatã coffee seedlings. Ninety days after application of the treatments (DAA)-Sample of five leaves per replication. Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.0496 < 0.001 0.1498
CV, %		25.25
		Leaf area (cm²)
Mundo Novo Obatã		34.95 a 39.12 b
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	15.49 A 30.47 A 43.09 A 50.74 A	20.86 A 38.62 A 50.39 A 46.61A	18.17 34.55 46.74 48.67
	Pr > t
Regression	0.0057 Q1	< 0.001 Q2	< 0.001 Q3
R²	99.86	99.85	99.85
1ŷ = -0.00095x2 + 0.3681x + 15.203	3ŷ = -0.00129x2 + 0.4117x + 17.9005
2ŷ = -0.00162x2 + 0.4551x + 20.605

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q – quadratic.

For N content in the coffee leaf, there was an effect only for N concentrations (Table 10). Bertani et al. (2019), stated that the N content was influenced in the leaves of passion fruit seedlings, that is, leaf N content increased with the increment in nitrogen fertilization represented by the quadratic model. The dose 0.34 g L^-1 had the lowest content, with an average of 21 g kg^-1, and can be indicated, considering that it reached the greatest plant development, the highest DGCI and leaf area, very close at the maximum dose of 0.68 g L^-1. This result corroborates those found in this experiment for hue and leaf N content, in which the maximum value for Mundo Novo, was found at the highest N concentration, while for Obatã, the maximum values of hue and DGCI were obtained lower N concentration.

Table 10
Summary of analysis of variance, regression analysis and comparison of leaf N averages, as a function of nitrogen concentrations (0, 50, 100 and 200 mg dm^-3), in Mundo Novo and Obatã coffee seedlings. Ninety days after application of the treatments (DAA)-Sample of five leaves per replication. Registro, SP, Brazil.

FV		Pr > Fc
Cultivars (C) Doses (D) C X D		0.2459 < 0.001 0.0987
CV, %		9.03
		Leaf N content (g kg-1)
Mundo Novo Obatã		21.85 a 22.60 a
Doses (mg dm-3)	Mundo Novo	Obatã	Mean
0 50 100 200	18.4 A 20.2 A 22.4 A 25.2 A	18.6 A 21.0 A 25.8 B 26.0 A	18.5 20.7 24.1 25.6
	Pr > t
Regression	< 0.001 L1	0.0157 Q2	0.0282 Q3
R²	97.39	89.91	97.69
1ŷ = 0.034171x + 18.56	3ŷ = -0.000165x2 + 0.0700x + 18.2572
2ŷ = -0.000275x2 + 0.0949x + 18.1472

• Means followed by the same letter do not differ by the Tukey test at 5% significance. L – linear; Q – quadratic.

Regarding regression, leaf N content of cultivar Mundo Novo showed linear development in response to N concentrations, where the point of maximum response was in relation to 200 mg dm^-³ of 25.2 g kg^-1. In Obatã, there was a quadratic adjustment; the maximum point was the concentration 160 mg dm^-³, with a content of 26 g kg^-1. These data corroborate those of Teixeira et al. (2020), a quadratic adjustment was observed with increasing N concentration, inducing a greater accumulation of leaf N in the lemon tree.

Pearson correlation between the parameters evaluated

There was a significant and positive correlation of the nitrogen content in the leaf with hue (H), DGCI and leaf area (AF), with the exception only for AF, 60 DAA (Table 11). The highest correlation coefficients between leaf N content and hue, DGCI and leaf area were obtained when five leaves were used per replication. The highest correlation coefficient with leaf N content was obtained with DGCI, 90 DAA, when five leaves were used per replication in the analysis. In general, the correlations of hue, DGCI and leaf area were associated with N concentrations and influenced by leaf N content. The results corroborate Gazola et al. (2016), this is, emphasize the efficiency of the use of digital imaging in recommending nitrogen fertilization.

Table 11
Summary of data correlation between nitrogen content and hue, dark green color index and leaf area in Mundo Novo and Obatã coffee seedlings at different collection times, 60 and 90 days after the application of treatments. Registro, SP, Brazil

Leaf N content	One leaf / replication, 60 DAA	One leaf / replication, 90 DAA	Five leaves / replication, 90 DAA
Hue	0.53***	0.71***	0.75***
Dark green color index	0.57***	0.71***	0.76***
Leaf area	0.30 ns	0.52***	0.59***

• ^{*, ***}
  significant by t-test at 5 and 0.1% probability;
• ^ns
  not significant.

CONCLUSIONS

The hue and DGCI obtained in the digital image analysis of the coffee leaf were influenced by N concentrations, cultivar and are dependent on the day after treatment application and the number of leaves used.

The use of five leaves per replication in digital image analysis was more adequate to estimate leaf N content.

Hue, DGCI and leaf area were higher in cultivar Obatã compared to Mundo Novo, and are dependent on the day after application of the treatment and N concentrations.

ACKNOWLEDGEMENTS, FINANCIAL SUPPORT AND FULL DISCLOSURE

There is no conflict of interests in carrying out the research and publishing the manuscript. The authors are grateful to the São Paulo Research Foundation (FAPESP)[Grant 2020/14029-4].

REFERENCES

• Conab - Companhia Nacional de Abastecimento (2020) Acompanhamento da safra brasileira. Available at: <file:///C:/Users/Usuario/Downloads/CAFEZDEZEMBRO.pdf>. Accessed on: January 01^rst, 2021.
  » file:///C:/Users/Usuario/Downloads/CAFEZDEZEMBRO.pdf
• Conceição WB, Machado CMM, Macedo LA, Ferreira BO & França AC (2019) Crescimento inicial do cafeeiro em um latossolo vermelho amarelo sob diferentes umidades do solo. Revista Craibeiras de Agroecologia, 4:715-720.
• Bertani RMA, Silva SP, Deus ACF, Antunes AM & Fischer IH (2019) Doses de nitrogênio no desenvolvimento de mudas altas de maracujá-amarelo. Revista de Agricultura Neotropical, 6:29-35.
• Colodetti TV, Rodrigues WN, Martins LD, Brinate SVB, Tomaz MA, Amaral JFT & Filho ACV (2015) Nitrogen availability modulating the growth of improved genotypes of Coffea canephora African Journal of Agricultural Research, 32:3150-3156.
• Fagundes AV (2011) Cuidados com o equilíbrio nutricional do cafeeiro. Available at: <https://www.cafepoint.com.br/radares-tecnicos/solos-e-nutricao/cuidados-com-o-equilibrio-nutricional-do-cafeeiro-70285n.aspx>. Accessed on: January 20^th, 2018.
  » https://www.cafepoint.com.br/radares-tecnicos/solos-e-nutricao/cuidados-com-o-equilibrio-nutricional-do-cafeeiro-70285n.aspx
• Felisberto PAC, Godoy LJG & Felisberto G (2016) Índice de cor da folha para monitoramento nutricional de nitrogênio em plantas de pimentão. Científica, 44:207-216.
• Ferreira DF (2011) Sisvar: a computer statistical analysis system. Ciência e Agrotecnologia, 6:1039-1042.
• Flexas J, Díaz-Espejo A, Conesa MA, Coopman RA, Douthe C, Gago J, Gallé A, Galmés J, Medrano H, Ribas-Carbo M, Tomàs M & Niinemets Ü (2016) Mesophyll conductance to CO2 and Rubisco as targets for improving intrinsic water use efficiency in C3 plants. Plant Cell Environ, 39:965-982.
• Filho OM & Neto HV (1999) Processamento digital de imagens. Rio de Janeiro, Braspot. 331p.
• Groff VLF, Garcia C, Groff DGF, Avila FW & Santos LA (2020) Influência do teor de nitrogênio no efeito verde causado por piraclostrobina em plantas de trigo. Applied Research & Agrotechnology, 13:e5964.
• Gazola RPD, Buzetti S, Gazola RN, Castilho RMM, Teixeira FMCM, Celestrino TS & Dupas E (2016) Nitrogen and type of herbicide used for growth regulation on the green coloration intensity of Emerald grass. Ciência Rural, 6:984-990.
• Gimenes FMA, Barbosa HZ, Gerdes L, Giacomini AA, Mattos WT, Batista K, Premazzi LM & Miguel ANV (2017) The utilization of tropical legumes to provide nitrogen to pastures: a review. African Journal of Agricultural Research, 2:85-92.
• Leghari SJ, Wahocho AW, Laghari GM, Hafeezlaghari A, Mustafabhabhan G, Talpur KH, Bhutto TA, Wahocho SA & Lashari AA (2016) Role of nitrogen for plant growth and development: a review. Advances in Environmental Biology, 9:209-218.
• Malavolta E, Vitti GC & Oliveira AS (1997) Avaliação do estado nutricional das plantas: Princípios e aplicações. 2ª ed. Piracicaba, Potafos. 319p.
• Oliveira EJ, Melo HC, Alves FRR, Melo APC, Trindade KL, Guedes TM & Souza CM (2020) Morphophysiology and yield of green corn cultivated under diferente water depths and nitrogen doses in the cerrado conditions of Goiás, Brazil. Research, Society and Development, 10:e6179108857.
• Paulidis T & Liow YT (1990) Integrating region growing and edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12:225-233.
• Rorie LR, Larry CP, Morteza M, Douglas EK, Andy K, Matthew CM & David EL (2011) Association of “Greenness” in Corn with Yield and Leaf Nitrogen Concentration. Agronomy Journal, 103:529-535.
• Santos PLF & Castilho RMM (2015) Relação entre teor de clorofila e nitrogênio foliar em grama esmeralda cultivada em substratos. Revista Tecnologia & Ciência Agropecuária, 9:51-54.
• Santos R, Barros AAG, Graça DCS & Cardoso G (2016) Obtenção e avaliação de nutriente de nitrogênio em plantio de quiabeiro (Abelmoschus esculentus L Moench). Revista Técnica Científica, Edição especial: 01-09.
• Schwers F, Caron BO, Eli EF, Oliveira DM, Monteiro V & Souza FQ (2016) Evaluation of the effect of doses and sources of nitrogen on morphological variables, radiation interception and yield of sunflower. Revista Ceres, 63:380-386.
• Silva TA, Júnior JOS, Mielke MS & Hernández COA (2018) Uso de imagem digital para estimar o teor foliar de nitrogênio em cacaueiros. Agrotrópica, 30:15-24.
• Teixeira PTL, Schafer G, Back MM, Petry HB & Souza PVD (2020) Interactions bet ween nitrogen fertilization with the growth and leaf macronutrients of citrus root stocks. Revista Pesquisa Agropecuária Gaúcha, 1:273-287.
• Wang Y, Wang D, Zhang G & Wang Jun (2013) Estimating nitrogen status of rice using the image segmentation of GR thresholding method. Field Crops Research, 149:33-39.
• Wang Y, Wang D, Shi P & Omasa K (2014) Estimating rice chlorophyll content and leaf nitrogen concentration with a digital still color camara under natural light. Plant methods, 1:01-11.

Publication Dates

• Publication in this collection
  09 Jan 2023
• Date of issue
  Nov-Dec 2022

History

• Received
  23 Apr 2021
• Accepted
  01 Feb 2022