Next Article in Journal
Statistical Analysis of Mesovortices during the First Rainy Season in Guangdong, South China
Next Article in Special Issue
Quantitative Assessment of Apple Mosaic Disease Severity Based on Hyperspectral Images and Chlorophyll Content
Previous Article in Journal
A Technology for Seismogenic Process Monitoring and Systematic Earthquake Forecasting
Previous Article in Special Issue
Forecasting Table Beet Root Yield Using Spectral and Textural Features from Hyperspectral UAS Imagery
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Evaluation of Absolute Measurements and Normalized Indices of Proximal Optical Sensors as Estimators of Yield in Muskmelon and Sweet Pepper

by
Cihan Karaca
1,2,
Rodney B. Thompson
2,3,
M. Teresa Peña-Fleitas
2,3,
Marisa Gallardo
2,3 and
Francisco M. Padilla
2,3,*
1
Department of Greenhouse Production, Kumluca Vocational School, Akdeniz University, Antalya 07059, Türkiye
2
Department of Agronomy, University of Almeria, 04120 Almeria, Spain
3
CIAIMBITAL Research Centre for Mediterranean Intensive Agrosystems and Agrifood Biotechnology, University of Almeria, 04120 Almeria, Spain
*
Author to whom correspondence should be addressed.
Remote Sens. 2023, 15(8), 2174; https://doi.org/10.3390/rs15082174
Submission received: 22 February 2023 / Revised: 11 April 2023 / Accepted: 17 April 2023 / Published: 20 April 2023
(This article belongs to the Special Issue Application of Hyperspectral Imagery in Precision Agriculture)

Abstract

:
The generally established protocol for leaf measurement with proximal optical sensors is to use the most recently fully expanded leaf. However, differences in the nitrogen (N) status of lower and upper leaves could possibly be used to enhance optical sensor measurement. Normalized indices that consider both upper and lower leaves have been proposed to improve the assessment of crop N status and yield estimation. This study evaluated whether normalized indices improved the estimation of crop yield from measurements with three different proximal optical sensors: (i) SPAD-502 leaf chlorophyll meter, (ii) Crop Circle ACS 470 canopy reflectance sensor, and (iii) Multiplex fluorescence meter. The study was conducted with sweet pepper (Capsicum annuum L.) and muskmelon (Cucumis melo L.) in plastic greenhouses in Almeria, Spain. Measurements were made on the latest (most recent) leaf (L1), and the second (L2), third (L3) and fourth (L4) fully expanded leaves. Yield estimation models, using linear regression analysis, were developed and validated from the absolute and normalized measurements of the three optical sensors. Overall, the calibration and validation results indicated that the absolute measurements generally had better yield estimation performance than the normalized indices for all the leaves and different leaf profiles. In both species, there was a better performance at the early phenological stages, such as the vegetative and flowering stages, for the absolute and normalized indices for the three optical sensors. Absolute proximal optical sensor measurements on the lower leaves (L2, L3 and L4) slightly improved yield estimation compared to the L1 leaf. Normalized indices that included the L4 leaf (L1–L4) had better yield estimation compared to those using L2 and L3 (e.g., L1–L2 and L1–L3). Of the normalized indices evaluated, the yield performance of the Relative Index (RI), Relative Difference Index (RDI), and Normalized Difference Index (NDI) were very similar, and generally superior to the Difference Index (DI). Overall, the results of this study demonstrated that for three different proximal optical sensors in both muskmelon and sweet pepper (i) normalized indices did not improve yield estimation, and (ii) that absolute measurements on lower leaves (L2, L3 and L4) slightly improved yield estimation performance.

Graphical Abstract

1. Introduction

Management of the crop nitrogen (N) supply affects crop yield, biomass production, fruit quality, plant development, and growth [1,2,3]. Insufficient N supply reduces profitability, while excessive N supply is associated with negative environmental impacts and can also have undesirable crop effects (e.g., lodging, excessive vegetative growth) [4].
Proximal optical sensors are an application of remote sensing in which sensors that measure the light properties of crops or sensitive compounds are positioned in contact with or close to the crop canopy [5]. Proximal optical sensors suitable for N management applications can be classified as (i) leaf chlorophyll meters, (ii) canopy reflectance sensors, and (iii) fluorescence-based chlorophyll and flavonols meters [5]. Proximal optical sensors provide rapid, non-destructive, and real-time assessment of crop N status [5]. They are being increasingly used for site-specific N management during cropping to optimize N management while ensuring maximum yield [6].
Several studies have examined the relationship between proximal optical sensor measurements and yield in a wide range of crops, such as maize [7], wheat [8,9], rice [10,11], cucumber [12], tomato [13,14], and sweet pepper [15]. These studies have shown that proximal sensor measurements are sensitive to N management and can be used to estimate yield. In many of these studies, equations were developed to estimate crop yield from proximal optical sensor measurements. However, these proximal optical sensor values can be affected by cultivar and cultural practices [2,9,16,17]. Additionally, there may be differences in optical sensor measurements between the lower and upper leaves of the crop [10,11,18,19,20]. These issues can limit the extrapolation of yield estimation models in different cultivars and growing conditions, and when consistent measuring protocols have not been used [5].
The standard measurement protocol for proximal optical sensors has traditionally been the measurement of the latest fully expanded and well-lit leaf [5,20]. However, various normalized indices have been developed that include measurements of different leaf positions [21,22,23]. The normalized indices are based on the N sufficiency index (NSI) developed by Blackmer and Schepers [24]. The N sufficiency index relates measurements of plants with a certain N treatment to measurements made on plants growing in a reference plot in which N is not limiting. Reference plots and the N sufficiency index have been developed for cereal crops but there are practical issues that limit their use with fertigated vegetable crops [25]. The primary limitation is the requirement for an additional fertigation sector, independent from that of the main crop. Additionally, the work and calculation involved in the periodic programming of fertigation would be doubled. For practical reasons, reference plots are not attractive for commercial fertigated vegetable crops [25]. Therefore, for these crops, there is a need to assess alternative normalization procedures, such as those based on different leaf positions.
Proximal optical sensor measurements can be affected by year, location, cultivar and phenological stage [26]. Recently, normalized indices have been used to minimize these effects [18,20,26]. Normalizing measured data may minimize the differences between cultivars, resulting in more accurate estimates of yield and other parameters.
The normalized indices based on different leaf positions relate measurements made on lower leaves (Lj) to those made on the latest fully expanded leaf (L1). By compensating for differences in optical sensor measurements between the upper and lower leaves, it is suggested that normalized indices reduce the effects of cultivars, environmental factors, or inconsistent measurement protocols [11,18,20].
The most commonly-used normalized indices for optical sensor measurements are the Difference Index (DI; [27]), Relative Index (RI; [21]), Relative Difference Index (RDI; [22]), and Normalized Difference Index (NDI; [23]). Studies on the normalized indices of proximal optical sensor measurements have been conducted on different cereal species such as rice [11,18,19,21,22,23,28], wheat [27,29], and barley [30]. However, there have been no studies on vegetable crops. Most of the studies on normalized indices have used the SPAD chlorophyll meter; the only study that addressed canopy reflectance measurements was that of Andleeb et al. [29]. There are no published studies evaluating the use of normalized indices with fluorescence-based leaf flavonols meters.
Various studies [11,18,26] have suggested that the normalized indices of optical proximal sensors can be used to improve the assessment of crop N status and estimation of crop yield. Andleeb et al. [29] noted that there is a strong positive correlation between RI and wheat grain yield, for both SPAD and NDVI measurements. These authors suggest that normalized indices can be used as efficient and precise criteria for identifying more efficient N-use wheat cultivar varieties under heat-stress conditions. Zhang et al. [11] examined absolute and various normalized indices and reported that the highest relationship between yield and SPAD measurements was with NDI for intensive rice crops.
This study examined the yield estimation models of absolute proximal optical sensor measurements and various normalized indices (DI, RI, RDI, NDI) for three different proximal optical sensors (leaf chlorophyll meter, canopy reflectance sensor, and fluorescence-based leaf chlorophyll and flavonols meter) in sweet pepper and muskmelon crops. The crops were grown in different years, with different cultivars, and with different N supplies. We hypothesized that normalized indices of proximal optical sensor measurements may improve yield estimation models when compared to absolute proximal optical sensor measurements. There is little published information, for all types of crops, examining whether normalized indices appreciably reduce differences between cultivars of optical sensor measurements. No studies have been conducted with vegetable crops.

2. Materials and Methods

2.1. Experimental Site and Crops

The study was conducted with sweet pepper (Capsicum annuum L.) and muskmelon (Cucumis melo L.) grown in two different years (2020 and 2021) in greenhouses at the University of Almeria Experimental Farm, in Almeria (SE Spain, 36°51′N latitude, 2°16′W longitude; 92 m above sea level).
In the study, two Mediterranean-type plastic greenhouses (U13 and U14) with the same characteristics, with a cropped area of approximately 1300 m2, were used [31]. The two muskmelon crops and the first sweet pepper crop were grown in greenhouse U13. The second sweet pepper crop was grown in greenhouse U14. Growing conditions were very similar in both greenhouses with the exception that the planting soil mineral N (0–0.4 m) in greenhouse U14 was appreciably higher than for the three crops in U13 (Table 1). This was because there was a delay in preparing the fertigation system of greenhouse U14, which prevented the application of irrigation prior to transplanting to leach nitrate (NO3) from the root zone [31], as was done before the three crops in greenhouse U13.
In both greenhouses, the growing medium consisted of a layer of 0.3 m thick imported silty loam soil that was placed over the original sandy loam soil. A mulch layer that consisted of either a layer of coarse sand (10 mm thickness) or a black polyethylene film was used in greenhouses U13 and U14, respectively [31]. In both greenhouses, surface drip irrigation was used. The distance between drippers (3 L h−1) was 0.5 m; there was 0.8 m spacing between the two paired drip lines, and 1.2 m between adjacent paired lines [31].
All cultural practices were consistent with local commercial crop management. Complete nutrient solution (macro and micronutrients) was applied by fertigation through the drip irrigation systems. Except for N (see Section 2.2), macronutrients were applied in non-limiting amounts at the following concentration: HPO4, 2 mmol L−1; K+, 4 mmol L−1; Ca2+, 4 mmol L−1; Mg2+, 1.5 mmol L−1; SO42+, 2.35 mmol L−1. Nutrient solution application began two and eight days after transplanting (DAT) for the muskmelon and sweet pepper crops, respectively. Nutrient solution was then applied in all irrigations. Irrigation was applied every 1–4 days, to keep the soil matric potential between −15 and −25 kPa, at 10 cm soil depth. Manual tensiometers (model SR, Irrometer Co., Riverside, CA, USA) were used to measure soil matric potential in each replicate plot.

2.2. Experimental Design and N Treatments

For muskmelon, the first growing season was from 27 February 2020 to 11 July 2020 (105 days), and the second growing season was from 26 February 2021 to 8 July 2021 (102 days). Three cantaloupe-type muskmelon cultivars were grown in each season: Tezac (Seminis, Inc., Bayer AG, Leverkusen, Germany), Magiar (Nunhems, BASF SE, Ludwigshafen, Germany), and Jacobo (Semillas Fitó, Barcelona, Spain) in the first season, and Bosito (Seminis, Inc., Bayer AG, Leverkusen, Germany), Magiar, and Jacobo in the second season. Because of the discontinuation of the Tezac cultivar, it was replaced with the Bosito cultivar in the second season.
For sweet pepper, the first growing season was from 22 July 2020 to 28 January 2021 (190 days), and the second growing season was from 22 July 2021 to 9 January 2022 (171 days). Three different sweet pepper cultivars were grown: Melchor (Zeraim Iberica, Syngenta Crop Protection AG, Basel, Switzerland), Machado (Hazera Seeds Ltd., Limagrain Group, Saint Beauzire, France), and CLX PLRJ 731 (HM. CLAUSE SAS, La Motte, Portes-lès-Valence, France).
Both the U13 and U14 greenhouses had 12 experimental plots of 12 m long × 6 m wide. Three cultivars for each species were grown in each of the experimental plots. Each cultivar was planted in two paired lines of plants (i.e., four lines) in each plot. The positions of the paired lines for each cultivar in each plot were randomized.
In each crop, three treatments with different N concentrations were applied by fertigation. They were very deficient N (N1 = 2 mmol N L−1), deficient N (N2 = 8 mmol N L−1), and conventional N application (N3 = 14 mmol N L−1) in accordance with local commercial practice [32]. The N concentrations (mmol L−1) and amounts of mineral N (NO3–N + NH4+–N; kg N ha−1) applied to the different N treatments (N1, N2 and N3) of each crop are given in Table 1.

2.3. Proximal Optical Sensor Measurements

Measurements of relative leaf chlorophyll content were made with the hand-held leaf-clip SPAD-502 (Minolta Camera Co., Ltd., Tokyo, Japan). The Multiplex® 3.6 sensor (Force-A, Orsay, France) was used to measure relative chlorophyll and flavonols contents. This sensor is a portable device with four emission light sources (UV, green, red and blue) that induce fluorescence in plant tissues [33]. Of the various indices measured by the Multiplex sensor [33,34,35], the present work focused on the Nitrogen Balance Index under red excitation (NBI-R) [36]. The NBI is calculated as the ratio between relative chlorophyll and flavonols contents [35] and has been shown to be a very sensitive indicator of crop N status [15,36,37,38].
Measurements of canopy reflectance were made with a Crop Circle ACS-470 sensor (Holland Scientific Inc., Lincoln, NE, USA). This sensor has a light source that emits both visible radiation and NIR [39]. By modulating the light source, the sensor can distinguish reflected radiation from its own light source from that associated with ambient radiation [5]. With this sensor, filters were selected to measure reflectance at 550 nm (green), 670 nm (red), 730 nm (red-edge), and 760 nm (near-infrared, NIR).
Of the various vegetation indices that were calculated from measurements by the Crop Circle ACS-470, the present work focused on the Normalized Difference Vegetation Index (NDVI; [40]) because it is established as a sensitive indicator of crop N status [5,25].
NDVI was calculated as:
NDVI = ( R N I R R r e d ) / ( R N I R + R r e d )
where R is reflectance at each wavelength.
In addition to the NDVI, the following vegetation indices were calculated: GNDVI (Normalized Difference Vegetation Index on Green), RVI (Ratio Vegetation Index on Red), GVI (Ratio Vegetation Index on Green), RENDVI (Normalized Difference Vegetation Index on Red Edge), CI (Chlorophyll Index), CCCI (Canopy Chlorophyll Content Index), RENDVI/NDVI ratio, and MTCI (MERIS Terrestrial Chlorophyll Index). Exploratory analysis revealed no improvement of yield estimation with these indices, nor was there saturation of the NDVI at high N levels.
Canopy reflectance measurements were made by holding the sensor vertically parallel to the crop row. In this system, the cords used to vertically support these crops precluded downward measurement. Consequently, indices such as the Soil Adjusted Vegetation Index (SAVI) that consider the soil signal, were not assessed.
Measurements with SPAD and Multiplex were made on eight marked plants of each cultivar, in each replicate plot. They were made from 8:00 to 10:00 solar time, before fertigation was applied. On each plant on each measurement date, measurements were made on the latest (L1), second (L2), third (L3) and fourth (L4) fully expanded leaf, from the top of the crop [10,19]. Consistent with the protocol developed by Padilla et al. [41,42] for fully expanded leaf selection, individual leaf measurements were made on well-lit leaves, on the distal part of the adaxial side of the leaf, midway between the margin and the mid-rib of the leaves. Leaves with physical damage or with condensed water were not measured; in those cases, alternative plants were selected.
Measurements of canopy reflectance were made from 10:00 to 11:00 solar time. Measurements commenced once the crop had sufficient height to enable measurement considering (i) the 26 cm height of the field of view and (ii) that two passes, one above the other, were made. In each replicate plot, two passes of 4 m each were made for each cultivar at walking speed (approx. 1.5 km h−1). Each 4 m pass consisted of two 2 m passes on different lines of plants. There were 10 measurements per second and these were stored in a portable GeoScout GLS-400 data logger (Holland Scientific, Inc., Lincoln, NE, USA). The first pass was made with the top of the field of view at the height of the most recent fully expanded leaf (L1ref). The second pass was made immediately below the first pass, with the top of the field of view at the height of the fourth fully expanded leaf (L4ref).
Proximal sensor measurements were made every 28 days for sweet pepper, and every 14 days for muskmelon. Like most greenhouse-grown vegetable crops, muskmelon and sweet pepper can be considered to have four distinct phenological stages: vegetative, flowering, early fruit growth, and harvesting (or fruit maturity). Different plant physiological processes dominate each stage. For practical nutrient management, local vegetable growers generally manage each of these stages differently. Therefore, measurements were taken in each of the four phenological stages of vegetative, flowering, early fruit growth, and harvest, for each crop. The measurement dates of the proximal optical sensors based on the day after transplanting (DAT) are given in Table 2.
For SPAD and Multiplex, different normalized indices were calculated based on measurements at different leaf positions (L1, L2, L3 and L4). For the NDVI, normalized indices were calculated based on measurements at different profiles (L1ref and L4ref). The normalized indices used with the three sensors were Difference Index (DI; [27]), Relative Index (RI; [21]), Relative Difference Index (RDI; [22]), and Normalized Difference Index (NDI; [23]). The equations of these indices are given in Table 3.

2.4. Determination of Fruit Yield

Total fruit yield was determined by harvesting the fruit of the eight marked plants of each cultivar, in each plot [31]. Total yield was calculated by summing the fresh weight of the mature fruit of muskmelon and the red fruits of sweet pepper. There were two fruit harvests of muskmelon in 2020 (96 and 104 DAT) and one in 2021(101 DAT). In the 2020 sweet pepper crop, there were six harvests between 98 and 187 DAT. In the 2021 sweet pepper crop, there were four harvests between 90 and 160 DAT [31].

2.5. Data Analysis

Analysis of data of optical sensor measurements was done by pooling datasets of the two seasons of each crop, the three cultivars of each species, and the three different N treatments. The research diagram including the independent variables, dependent variables, calibration, and validation information used in the study is presented in Figure 1.
For relating the yield estimation models to a given absolute or normalized optical sensor measurement/index for each phenological period, there were a total of 18 data points. The 18 data points were pooled data for two seasons and three cultivars of each species, and three different N treatments. Each data point represented the mean of four values from each of the four replicate plots. Twelve (67%) of the 18 data points were used for each calibration analysis, and six (33%) were used for validation. The data used for the calibration and validation stages were randomly selected.
For calibration, the relationship between each variable and fruit yield was evaluated by linear regression analysis. The linear relationships between each variable and fruit yield were assessed using R2, RMSE, and p-values. The calibration model fitted to a linear regression is given in Equation (2):
y = a   x + b
where y is the total yield (kg m−2); x is the absolute measurement value or normalized index of the selected proximal sensor; a and b are the slope and intercept of the fitted line, respectively.
For validation of the derived calibration models, the yield was predicted with the pooled validation data of optical sensors and estimation models of each variable. Estimated yield was compared to observed (measured) yield using coefficient of determination (R2), root mean square error (RMSE), relative error (RE), mean bias error (MBE), and Willmott Index (d) values (Table 4).
R2 and d values equal to 1, and RMSE, RE, and MBE values equal to 0 indicate the best possible regression relationship. The classification of R2 values from Padilla et al. [43] was used, in which R2 values < 0.2, 0.5 > R2 ≥ 0.2, 0.7 > R2 ≥ 0.5, 0.85 > R2 ≥ 0.7, and R2 ≥ 0.85, indicate very weak, weak, moderate, strong, and very strong relationships, respectively. Linear regression models were developed using SPSS 25 (IBM Corporation, Armonk, New York, NY, USA). The p values were calculated for each linear regression to determine the significance level.

3. Results

3.1. Calibration Models to Estimate Fruit Yield from SPAD Measurements

3.1.1. Sweet Pepper

The linear relationships between the absolute SPAD measurements and the normalized SPAD indices, DI, RI, RDI, NDI, with fruit yield (kg m−2) at different phenological stages of sweet pepper, are presented in Figure 2.
The equations and calibration performances of these relationships are provided in Table S1. The strength of the relationships between absolute SPAD values and yield were very similar for different leaf positions (L1, L2, L3, L4) within each phenological stage, with R2 values of 0.50–0.75 (Figure 2). The best-performing calibration model was with the L3 (R2: 0.75; p < 0.001) and L4 (R2: 0.73; p < 0.001) leaves in the vegetative stage. The worst-performing calibration model was with the L1 (R2: 0.50; p < 0.05) and L2 (R2: 0.52; p < 0.05) leaves in the harvest stage.
The relationships between the normalized SPAD indices and yield varied with phenological stage and leaf positions. The calibration performance of the DI, RI, RDI and NDI indices were very similar to one another within each phenological stage. Of the normalized SPAD indices, the strongest relationship with yield occurred with DIL1–L3 (R2: 0.66; p < 0.01) in the vegetative stage. In the harvest stage, there were strong relationships between yield and all normalized SPAD indices (0.42 ≤ R2 ≤ 0.51; 0.01 < p < 0.05), except for DI (0.04 ≤ R2 ≤ 0.36). The R2 values for all the normalized SPAD indices were low during the flowering and early fruit growth stages.
In general, absolute SPAD measurement from different leaf positions at vegetative, flowering, and early fruit growth stages had stronger relationships with yield (0.62 ≤ R2 ≤ 0.75) than the equivalent normalized SPAD indices (Figure 2 and Table S1).

3.1.2. Muskmelon

The linear relationship between absolute SPAD measurements and the normalized SPAD indices, DI, RI, RDI, NDI, with fruit yield (kg m−2) for different phenological stages of muskmelon, are presented in Figure 3. The equations and indicators of calibration performances of these relations are given in Table S2.
There were very weak or weak relationships between the absolute SPAD values and yield for all leaf positions (L1, L2, L3, L4) and phenological stages, with R2 values of 0.16–0.46 (Table S2). There were slightly stronger relationships in the harvest stage (average R2 value for the L1, L2, L3 and L4 positions of 0.42; p < 0.01) than in the early fruit growth (average R2 of 0.39), flowering (average R2 of 0.34), and vegetative (average R2 of 0.29) stages (Figure 3).
In the vegetative stage, the normalized indices performed better than the absolute SPAD measurements. The relationships between the normalized indices and yield were negligible in the flowering stage. At the early fruit growth and harvest stages, the relationships between normalized indices and yield were generally lower than for the corresponding absolute SPAD measurements. At the vegetative stage, for all the normalized SPAD indices, the strongest relationships with yield were obtained with the L1–L3 leaf profile (R2 of 0.81 and p < 0.001 for DI, RI, RDI and NDI). The normalized index relationships with yield of the L1–L3 profile were also stronger than that with the other leaf profiles (i.e., L1–L2 and L1–L4) at the early fruit growth stage.

3.2. Calibration Models to Estimate Fruit Yield from NBI Measurements

3.2.1. Sweet Pepper

The linear relationships between the absolute NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI) with fruit yield (kg m−2) at different phenological stages of sweet pepper are presented in Figure 4. The equations and calibration performances of these relations are provided in Table S3.
For all the absolute and normalized NBI measurements made on different leaf positions, the strongest relationships were found at the vegetative stage for L2, L3 and L4 (R2 = 0.82; p < 0.001) (Figure 4; Table S3). The relationship at the L1 position was slightly lower, with R2 = 0.78 (Table S3). The performance of the calibration models decreased with crop growth. The average R2 value of each of the four leaf positions (i.e., L1 to L4) decreased from 0.81 in the vegetative stage, to 0.67 in the flowering stage, to 0.64 in the early fruit growth stage, to 0.36 in the harvest stage (Figure 4).
The performance of the calibration models for the normalized indices varied depending on the index and phenological stage. Overall, the R2 values for the normalized indices ranged from 0.01 to 0.74. The performance of the DI was the lowest of the normalized indices (R2 of 0.01–0.47). The RI, RDI and NDI performances were generally better and were very similar to each other for each phenological stage. In the vegetative stage for all the normalized indices, there were very weak relationships with fruit yield for all the leaf profiles (R2 of 0.01–0.29; p > 0.05). The highest relationship with yield occurred in the L1–L4 leaf profile in the RI, RDI and NDI indices during the flowering stage. The performance of the normalized indices RI, RDI and NDI decreased with crop growth, for all the leaf profiles.

3.2.2. Muskmelon

The linear relationships between the absolute NBI measurements and normalized NBI indices (DI, RI, RDI, NDI) with fruit yield (kg m−2), at different phenological stages of muskmelon, are presented in Figure 5. The equations and calibration performances of these relationships are provided in Table S4.
The relationships between the absolute NBI measurements and yield, at different individual leaf positions, had R2 values of 0.50–0.77 (Figure 5; Table S4). There were few differences in R2 values between phenological stages. Although there was no statistical difference between L1 and L4 (p < 0.001), the R2 and RMSE values showed that among the different leaf positions, the best performance was with L4 in the vegetative stage (R2 = 0.77; RMSE = 5.61). (Table S4). In the vegetative and flowering stages, the performance of the calibration model was strongest in L4 (R2 = 0.77 and 0.67, respectively) and L3 (R2 = 0.71 and 0.63, respectively). It was lowest in L1 in these stages (R2 = 0.50 and 0.59, respectively). In the harvest stage, the performance of the calibration models was significantly the same but according to the coefficients of determination, the L1 calibration model (R2 = 0.67) was stronger than that of L4 (R2 = 0.62) (Figure 5). In the early fruit growth stage, there was little difference in the performance of calibration models between leaf positions.
Overall, the normalized indices had R2 values of 0.01–0.44. Generally, DI had better performance than RI, RSI and NDI in the vegetative, flowering, and early fruit growth stages. In these phenological stages, for all indices, model performance was better in the L1–L4 combination than in L1–L3 and L1–L2. The normalized indices had similar performance in the harvest stage.
In all the phenological stages, the performance of calibration models with absolute NBI measurements was generally better, in terms of higher R2 values, than that of any of the normalized indices.

3.3. Calibration Models to Estimate Fruit Yield from NDVI Measurements

3.3.1. Sweet Pepper

The linear relationship between the absolute NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI) with fruit yield (kg m−2) at different phenological stages of sweet pepper are presented in Figure 6. The equations and calibration performances of these relations are provided in Table S5.
The relationships with the absolute NDVI measurements in different leaf profiles (L1ref and L4ref) had R2 values of 0.36−0.84. NDVI measurements taken at L4ref had higher R2 values than measurements at L1ref in the vegetative (0.64 vs. 0.36), flowering (0.84 vs. 0.77), and early fruit growth (0.71 vs. 0.51) stages (Table S5). However, at the harvest stage, the relationship at L4ref had a slightly lower R2 value than at L1ref (0.65 vs. 0.69). Regarding the phenological stages, the strength of the relationships decreased in the order of flowering, harvest, early fruit growth, and vegetative stages, with average R2 values of 0.81, 0.67, 0.61 and 0.50, respectively (Figure 6). None of the normalized NDVI indices were related to crop yield in the flowering, early fruit growth, and harvest stages, with average R2 values for the DI, RI, RDI and NDI indices of 0.02, 0.03 and 0.01, respectively. The exception was the vegetative stage when the average R2 value for the DI, RI, RDI and NDI indices was 0.68 (p < 0.01) in all cases. Generally, for three out of four phenological stages, the performance of the calibration models with absolute NDVI measurements was much better than for any of the normalized NDVI indices.

3.3.2. Muskmelon

The linear relationship between the absolute NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI) with fruit yield (kg m−2), at different phenological stages of muskmelon, are presented in Figure 7. The equations and calibration performances of these relationships are provided in Table S6.
The linear relationships between the absolute NDVI measurements in L1ref and L4ref leaf profiles and fruit yield had R2 values of 0.68−0.85 (p < 0.001) (Figure 7). There was a tendency for the absolute NDVI measurements at L4ref to have slightly higher R2 values than the measurements at L1ref in the vegetative (0.85 vs. 0.74), early fruit growth (0.79 vs. 0.72), and harvest (0.70 vs. 0.68) stages (Table S6). However, at the flowering stage, the relationship at L4ref had a slightly lower R2 value than at L1ref (0.69 vs. 0.74). The strength of the relationships of the absolute NDVI measurements was slightly higher in the vegetative (average R2 = 0.80) and early fruit growth (average R2 = 0.76) stages than in the flowering (average R2 = 0.72) and harvest (average R2 = 0.69) stages. None of the normalized NDVI indices were strongly related to crop yield in any of the phenological stages, with R2 values of 0.01–0.12.

3.4. Validation of Models to Estimate Fruit Yield from SPAD Measurements

3.4.1. Sweet Pepper

The results of the validation analysis of the models to estimate fruit yield from the absolute SPAD measurements and normalized indices in sweet pepper are presented in Figure S1 and Table 5.
For the absolute SPAD measurements, for all leaf positions, the validation analyses had R2 values of 0.62–0.96. R2 values increased with crop growth; the average R2 values of L4 being 0.67, 0.83, 0.88 and 0.90, for the vegetative, flowering, early fruit growth, and harvest stages, respectively (Figure S1). The absolute SPAD measurements at L3 at the harvest stage had the best validation results with respect to yield estimation (R2 = 0.96; p < 0.001) (Table 5). The measurements at L4 in the vegetative and flowering stages and at L2 in the early fruit stage had also high performance in the validation analysis.
The validation of yield estimation with the normalized SPAD indices had poorer performance than the absolute SPAD measurements. The MBE values of the normalized SPAD indices showed that yield estimation was generally higher than the observed yield (Table 5).

3.4.2. Muskmelon

The results of the validation analysis of the models, which estimated muskmelon yield from the absolute SPAD measurements and normalized indices, are provided in Figure S2 and Table 6.
The validation analysis for the absolute SPAD measurements on different leaf positions was weak and not statistically significant, with R2 values of 0.12–0.42 (Figure S2). The validation results were slightly better with the normalized SPAD indices than with the absolute SPAD measurements in terms of slightly higher R2 values, but the regressions were also not significant for the normalized indices. The best performance according to the R2 and d values (R2 values of 0.55–0.56 and d values of 0.35–0.83) occurred in the L1–L3 leaf profile for all the normalized SPAD indices in the vegetative stage, but the p-values for the relationship showed that the models were not significant and therefore not validated (Table 6). In the flowering stage, the R2 values of the L1–L2 and L1–L3 profiles of all the normalized SPAD indices were 0.49–0.51 and 0.48–0.49, respectively. The corresponding d values were 0.17–0.37 and 0.00–0.04, which were close to 0, with data distribution far from the 1:1 line (Figure S2). There was a generally weak performance of most normalized SPAD indices in the early fruit growth and harvest stages for the validation analysis, with R2 values < 0.21 and d values < 0.58.

3.5. Validation of Models to Estimate Fruit Yield from NBI Measurements

3.5.1. Sweet Pepper

The results of the validation analyses of the models derived to estimate sweet pepper yield from the absolute NBI measurements and normalized indices are presented in Figure S3 and Table 7.
The performance of the validation analysis of the absolute NBI measurements did not vary notably with leaf position. However, it did change with the phenological stage, with better results in the flowering (average R2 of 0.95) and vegetative (average R2 of 0.87) stages than in the early fruit growth (average R2 of 0.79) and harvest (average R2 of 0.63) stages (Figure S3). The best validation result was found in the L4 position in the flowering stage (R2 = 0.98; p < 0.001) (Table 7).
The validation results for the normalized NBI indices were generally poorer than those of the absolute NBI measurements. Comparison between phenological stages showed very poor validation for all the normalized indices, at all leaf positions in the vegetative stage.

3.5.2. Muskmelon

The performance of the models to validate the estimated muskmelon yield from absolute NBI measurements and normalized indices are given in Figure S4 and Table 8.
The validation results for the absolute NBI measurements showed low R2 values of 0.20–0.46, which were not significant, regardless of leaf position and phenological stage (Figure S4). In general, the normalized NBI indices did not improve the validation results of the absolute NBI measurements, with the exception of DIL1–L4 at the vegetative stage, with R2= 0.64 and d = 0.87 (Table 8). At the harvest stage, the R2 values of the RI, RDI and NDI normalized indices in the L1–L3 leaf profiles were stronger than in any of the L1, L2, L3 and L4 absolute NBI measurements, but the d values were notably lower with the normalized NBI indices.

3.6. Validation of Models to Estimate Fruit Yield from NDVI Measurements

3.6.1. Sweet Pepper

The results of the validation analyses of the models derived to estimate sweet pepper yield from the absolute NDVI measurements and normalized indices are provided in Figure S5 and Table 9.
The absolute NDVI measurements had better validation performance at the L4ref leaf profile than at the L1ref profile in the vegetative (R2 = 0.93 vs. 0.40, respectively) and early fruit growth (R2 = 0.70 vs. 0.60, respectively) stages, whereas differences were negligible in the flowering and harvest stages (Figure S5).
The normalized NDVI indices had very similar validation performance to the absolute NDVI measurements at L4ref profile in the vegetative stage, with averaged R2 values of DIL1ref–L4ref, RIL1ref–L4ref, RDIL1ref–L4ref and NDIL1ref–L4ref of 0.94. However, all four normalized NDVI indices had worse validation performance (lower R2 values and d values closer to 0) than the absolute NDVI measurements for both the L1ref and L4ref profiles in the flowering, early fruit growth, and harvest stages (Table 9).

3.6.2. Muskmelon

The results of the validation analyses of the models derived to estimate muskmelon yield from the absolute NDVI measurements and normalized indices are given in Figure S6 and Table 10.
The absolute NDVI measurements at the L4ref profile had better validation results than at L1ref in all phenological stages. The exception was the vegetative stage, which had the best performance in the validation analyses at L1ref (R2 = 0.84; p = 0.01) (Table 10). In terms of the phenological stages, validation performance was higher in the vegetative and flowering stages than in the early fruit growth and harvest stages, with average R2 values for the L1ref and L4ref positions of 0.69, 0.62, 0.37 and 0.43, respectively (Figure S6). The normalized NDVI indices DIL1ref–L4ref, RIL1ref–L4ref, RDIL1ref–L4ref, and NDIL1ref–L4ref had worse validation performance than the absolute NDVI measurements at both the L1ref and L4ref profiles, regardless of the phenological stage.

4. Discussion

Previous studies [11,22,26] have suggested that the lower leaves in rice provide more reliable estimates of crop N status than the upper leaves when using proximal optical sensors, and that they could be used for the assessment of crop N nutrition status. Similarly, in the present study, the strongest calibration performances between absolute proximal optical sensor measurement and yield, in the muskmelon and sweet pepper crops, occurred with the lower leaves (i.e., third and fourth fully expanded leaves, L3 and L4, respectively). Therefore, although the present study focused on fresh fruit yield, the results were consistent with those that have focused on crop N nutrition. These results show that estimation accuracy can be improved by using lower leaves, rather than just using the latest fully expanded leaf, which is the standard protocol for proximal optical sensor measurement [5]. However, the improvement in the accuracy of yield estimation with lower leaves was generally very small, being on average 0.05 units of R2 values. Such small improvements are unlikely to impact the practical use of yield estimation models. Because of the small improvement, it is not justified to modify the standard measurement protocol for proximal optical sensors on the latest fully expanded and well-lit leaf [5,20].
Leaf chlorophyll content is significantly related to leaf age [44]. Once leaves are fully expanded, photosynthetic capacity declines linearly with leaf age in many plant species, despite favorable growing conditions [45]. At the same time, cultural practices such as irrigation, fertilization, and biotic and abiotic stresses affect leaf characteristics [46] and chlorophyll content [47,48,49]. Consequently, proximal sensor measurements of lower leaves may have higher performance in estimating crop yield because they represent a longer period of crop development. Yuan et al. [26] reported that the fourth fully expanded leaf is more reliable for estimating plant N status than the upper leaves. Similarly, Zhang et al. [11] reported that the sensor measurement values of lower leaves is better than the upper leaves in terms of assessing rice N status. Zhao et al. [30] observed that lower expanded leaves are more sensitive to the crop N status of winter barley. Some researchers [11,50] have suggested that the poorer relationships of the upper leaves with N indices may be due to the time difference in the maturity of the latest fully expanded leaf.
In general, in the present work, absolute optical sensor measurements had better performance (i.e., higher R2 values and significant relationships) than normalized indices for calibration and validation analysis in sweet pepper and muskmelon, for SPAD, NBI and NDVI measurement. Yuan et al. [26] reported similar results where absolute proximal optical sensor measurements in rice had higher correlations with leaf nitrogen concentration, plant nitrogen concentration, plant nitrogen accumulation, nitrogen nutrition index, and grain yield compared to normalized indices of DI, RI, RDI and NDI. Even though the present study focused only on fruit yield, the results are consistent with a study by Yuan et al. [26] which focused on crop N nutrition and grain yield. The findings of this study, therefore, did not confirm our hypothesis that normalized indices of proximal optical sensor measurements could improve yield estimation models when compared to absolute proximal optical sensor measurements. Assessing the differences in the optical sensor measurements of the upper and lower leaves did not show any improvement in the estimation of yield. Therefore, the calculation of normalized indices is not advised under the same conditions as the present study.
The performance of the fruit yield estimation models, both for absolute and normalized indices of proximal optical sensor measurements, varied depending on the phenological stage in both species. The higher performance in the vegetative and flowering stages, which are the earlier phenological stages compared to later stages, suggests there is potential to use proximal optical sensors in the early stages of crops to assess crop N status to ensure high yields [43].
In general, the calibration and validation performances of the fruit yield estimation models were higher for sweet pepper than for muskmelon. This may be due to differences in the length of the crop cycles of the two species [15] and the number of fruit harvests [31]. Both species are cultivated as indeterminate crops in greenhouses in SE Spain. However, sweet pepper has multiple fruit harvests throughout a notably longer crop cycle, whereas for muskmelon, there are only one or two fruit harvests [31]. The length of the cropping cycles in indeterminate crops is often determined by the farmer because of commercial reasons and planned crop sequences. It may be that the potential yield of muskmelon is somewhat restricted by the short crop cycles.
The normalized indices calculated with the lowest leaves, the L1–L3 and L1–L4 profiles, gave better yield estimation performance than the same indices using the upper leaves, such as L1–L2. These findings are consistent with research by Zhao et al. [30], who reported that the correlations between the normalized NDI indices of the L1–L4 leaf profiles using the crop nitrogen nutrition index were stronger than that of the upper leaf profiles. Light scattering and mobilization of N from lower leaves may cause greater gradients with the upper leaves of the crop canopy [30].
Numerous studies have been conducted with various greenhouse-grown vegetable species (e.g., [2,15,41,42]) to examine the relationships between optical sensor measurement and crop N status. The limited available data suggest that cultivar can affect optical sensor measurement [2]. The effectiveness of normalization procedures to minimize cultivar effects should be investigated in species other than muskmelon and sweet pepper, e.g., tomato.
The main focus of this study was to estimate crop yield by using the normalized indices of optical sensor data. Other important applications of normalized indices are for the estimation of leaf N content (LNC) and the nitrogen nutrient index (NNI) from optical sensor data. Previous studies have shown that these variables can be estimated using normalized data with rice [11,29], wheat [28], and barley [29]. It is anticipated that subsequently, the effectiveness of normalized indices for the estimation of LNC, from the same crops used in the present study, will be published. It is suggested that the estimation of NNI using normalized indices for muskmelon, sweet pepper, and other vegetable crops be assessed in future work.

5. Conclusions

Overall, this study demonstrated for three different proximal optical sensors in both muskmelon and sweet pepper that normalized indices did not improve yield estimation performance, but that absolute measurements on lower leaves (L2, L3 and L4) slightly improved yield estimation performance. Consideration of the differences in optical sensor measurements between the upper and lower leaves into normalized indices did not improve the estimation of yield, and therefore the calculation of normalized indices is not advised under the same conditions as in the present study. The estimation accuracy of crop yield could be improved by using lower leaves, but in this study, the improvement was too small to impact the practical use of yield estimation models or to suggest modifying the standard measurement protocol using the latest fully expanded leaf.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/rs15082174/s1, Figure S1: Distribution of the estimated fruit yield (kg m−2) from different SPAD measurements and several normalized SPAD indices (DI, RI, RDI, NDI) and the measured yield values, at different phenological stages of sweet pepper for validation data; Figure S2: Distribution of the estimated fruit yield (kg m−2) from different SPAD measurements and several normalized SPAD indices (DI, RI, RDI, NDI) and the measured yield values, at different phenological stages of muskmelon for validation data; Figure S3: Distribution of the estimated fruit yield (kg m−2) from different NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI) and the measured yield values, at different phenological stages of sweet pepper for validation data; Figure S4: Distribution of the estimated fruit yield (kg m−2) from different NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI) and the measured yield values, at different phenological stages of muskmelon for validation data; Figure S5: Distribution of the estimated fruit yield (kg m−2) from different NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI) and the measured yield values, at different phenological stages of sweet pepper for validation data; Figure S6: Distribution of the estimated fruit yield (kg m−2) from different NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI) and the measured yield values, at different phenological stages of muskmelon for validation data; Table S1: Equations, p-value, coefficient of determination (R2), and root mean square error (RMSE) of linear regressions between SPAD and several normalized SPAD indices (DI, RI, RDS, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for the calibration data; Table S2. Equations, p-value, coefficient of determination (R2), and root mean square error (RMSE) of linear regressions between SPAD and several normalized SPAD indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for the calibration data; Table S3. Equations, p-value, coefficient of determination (R2), and root mean square error (RMSE) of linear regressions between NBI and several normalized NBI indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for the calibration data; Table S4: Equations, p-value, coefficient of determination (R2), and root mean square error (RMSE) of linear regressions between NBI and several normalized NBI indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for the calibration data; Table S5: Equations, p-value, coefficient of determination (R2), and root mean square error (RMSE) of linear regressions between NDVI and several normalized NDVI indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for the calibration data; Table S6: Equations, p-value, coefficient of determination (R2), and root mean square error (RMSE) of linear regressions between NDVI and several normalized NDVI indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for the calibration data.

Author Contributions

Conceptualization, F.M.P., R.B.T. and M.G.; Methodology, M.T.P.-F., F.M.P. and C.K.; Software, C.K.; Validation, C.K.; Formal Analysis, C.K. and F.M.P.; Investigation, M.T.P.-F.; Resources, R.B.T. and M.G.; Data Curation, F.M.P. and M.T.P.-F.; Writing—Original Draft Preparation, C.K.; Writing—Review and Editing, F.M.P. and R.B.T.; Visualization, C.K.; Supervision, R.B.T., M.G. and F.M.P.; Project Administration, R.B.T. and M.T.P.-F.; Funding Acquisition, R.B.T. and F.M.P. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by grant RTI2018-099429-B-100 of the Spanish Ministry of Science, Innovation and University. FMP acknowledges the support of a Ramón y Cajal grant (RYC-2014-15815). CK acknowledges funding from the Scientific and Technological Research Council of Turkey (TUBITAK-BIDEB) under International Post-Doctoral Research Fellowship Program 2219.

Data Availability Statement

The data are not publicly available. The data presented in this study may be available on request from the corresponding author.

Conflicts of Interest

The authors declare that they have no known competing financial interest or personal relationships that could appear to have influenced the work reported in this paper.

References

  1. Yang, H.; Yang, J.; Lv, Y.; He, J. SPAD Values and Nitrogen Nutrition Index for the Evaluation of Rice Nitrogen Status. Plant Prod. Sci. 2014, 17, 81–92. [Google Scholar] [CrossRef]
  2. de Souza, R.; Grasso, R.; Peña-Fleitas, M.T.; Gallardo, M.; Thompson, R.B.; Padilla, F.M. Effect of Cultivar on Chlorophyll Meter and Canopy Reflectance Measurements in Cucumber. Sensors 2020, 20, 509. [Google Scholar] [CrossRef] [PubMed]
  3. Lemaire, G.; Jeuffroy, M.H.; Gastal, F. Diagnosis Tool for Plant and Crop N Status in Vegetative Stage. Theory and Practices for Crop N Management. Eur. J. Agron. 2008, 28, 614–624. [Google Scholar] [CrossRef]
  4. Thompson, R.B.; Tremblay, N.; Fink, M.; Gallardo, M.; Padilla, F.M. Tools and Strategies for Sustainable Nitrogen Fertilisation of Vegetable Crops. In Advances in Research on Fertilization Management in Vegetable Crops; Tei, F., Nicola, S., Benincasa, P., Eds.; Springer: Heidelberg, Germany, 2017; pp. 11–63. [Google Scholar]
  5. Padilla, F.M.; Gallardo, M.; Peña-Fleitas, M.T.; de Souza, R.; Thompson, R.B. Proximal Optical Sensors for Nitrogen Management of Vegetable Crops: A Review. Sensors 2018, 18, 2083. [Google Scholar] [CrossRef] [PubMed]
  6. Lu, J.; Miao, Y.; Shi, W.; Li, J.; Hu, X.; Chen, Z.; Wang, X.; Kusnierek, K. Developing a Proximal Active Canopy Sensor-Based Precision Nitrogen Management Strategy for High-Yielding Rice. Remote Sens. 2020, 12, 1440. [Google Scholar] [CrossRef]
  7. Corti, M.; Cavalli, D.; Cabassi, G.; Marino Gallina, P.; Bechini, L. Does Remote and Proximal Optical Sensing Successfully Estimate Maize Variables? A Review. Eur. J. Agron. 2018, 99, 37–50. [Google Scholar] [CrossRef]
  8. Sultana, S.R.; Ali, A.; Ahmad, A.; Mubeen, M.; Zia-Ul-Haq, M.; Ahmad, S.; Ercisli, S.; Jaafar, H.Z.E. Normalized Difference Vegetation Index as a Tool for Wheat Yield Estimation: A Case Study from Faisalabad, Pakistan. Sci. World J. 2014, 2014, 725326. [Google Scholar] [CrossRef]
  9. Monostori, I.; Árendás, T.; Hoffman, B.; Galiba, G.; Gierczik, K.; Szira, F.; Vágújfalvi, A. Relationship between SPAD Value and Grain Yield Can Be Affected by Cultivar, Environment and Soil Nitrogen Content in Wheat. Euphytica 2016, 211, 103–112. [Google Scholar] [CrossRef]
  10. Zhang, K.; Liu, X.; Ma, Y.; Zhang, R.; Cao, Q.; Zhu, Y.; Cao, W.; Tian, Y. A Comparative Assessment of Measures of Leaf Nitrogen in Rice Using Two Leaf-Clip Meters. Sensors 2020, 20, 175. [Google Scholar] [CrossRef]
  11. Zhang, K.; Liu, X.; Ata-Ul-Karim, S.T.; Lu, J.; Krienke, B.; Li, S.; Cao, Q.; Zhu, Y.; Cao, W.; Tian, Y. Development of Chlorophyll-Meter-Index-Based Dynamic Models for Evaluation of High-Yield Japonica Rice Production in Yangtze River Reaches. Agronomy 2019, 9, 106. [Google Scholar] [CrossRef]
  12. Padilla, F.M.; Peña-Fleitas, M.T.; Gallardo, M.; Thompson, R.B. Determination of Sufficiency Values of Canopy Reflectance Vegetation Indices for Maximum Growth and Yield of Cucumber. Eur. J. Agron. 2017, 84, 1–15. [Google Scholar] [CrossRef]
  13. Güler, S.; Büyük, G. Relationships among Chlorophyll-Meter Reading Value, Leaf N and Yield of Cucumber and Tomatoes. Acta Hortic. 2007, 729, 307–311. [Google Scholar] [CrossRef]
  14. Mukherjee, J.; Sastri, C.V.S. Fruit Yield Predicting Model of Tomato Using Spectral and Hyperspectral Indices. J. Indian Soc. Remote Sens. 2004, 32, 301–306. [Google Scholar] [CrossRef]
  15. De Souza, R.; Peña-Fleitas, M.T.; Thompson, R.B.; Gallardo, M.; Grasso, R.; Padilla, F.M. Use of Fluorescence Indices as Predictors of Crop N Status and Yield for Greenhouse Sweet Pepper Crops. Precis. Agric. 2022, 23, 278–299. [Google Scholar] [CrossRef]
  16. Kurtar, E.S.; Odabas, M.S. Modelling of the Yield of Cucumber (Cucumis sativus L.) Using Light Intensity, Temperature and SPAD Value. Adv. Food Sci. 2010, 32, 170–173. [Google Scholar]
  17. Le Bail, M.; Jeuffroy, M.H.; Bouchard, C.; Barbottin, A. Is It Possible to Forecast the Grain Quality and Yield of Different Varieties of Winter Wheat from Minolta SPAD Meter Measurements? Eur. J. Agron. 2005, 23, 379–391. [Google Scholar] [CrossRef]
  18. Wang, S.; Zhu, Y.; Jiang, H.; Cao, W. Positional Differences in Nitrogen and Sugar Concentrations of Upper Leaves Relate to Plant N Status in Rice under Different N Rates. Field Crop. Res. 2006, 96, 224–234. [Google Scholar] [CrossRef]
  19. Zhao, B.; Ata-Ul-Karim, S.T.; Liu, Z.; Zhang, J.; Xiao, J.; Liu, Z.; Qin, A.; Ning, D.; Yang, Q.; Zhang, Y.; et al. Simple Assessment of Nitrogen Nutrition Index in Summer Maize by Using Chlorophyll Meter Readings. Front. Plant Sci. 2018, 9, 11. [Google Scholar] [CrossRef]
  20. Yuan, Z.; Ata-Ul-Karim, S.T.; Cao, Q.; Lu, Z.; Cao, W.; Zhu, Y.; Liu, X. Indicators for Diagnosing Nitrogen Status of Rice Based on Chlorophyll Meter Readings. Field Crop. Res. 2016, 185, 12–20. [Google Scholar] [CrossRef]
  21. Shen, Z.Q.; Wang, K.; Zhu, J.Y. Preliminary Study on Diagnosis of the Nitrogen Status of Two Rice Varieties Using the Chlorophyll Meter. Bull. Sci. Technol. 2002, 18, 174–176. [Google Scholar]
  22. Wang, S.H.; Cao, W.X.; Wang, Q.S.; Ding, Y.F.; Huang, P.S.; Ling, Q.H. Positional Distribution of Leaf Color and Nitrogen Nutrition Diagnosis in Rice Plant. Agric. Sin. 2002, 35, 1461–1466. [Google Scholar]
  23. Lin, F.F.; Qiu, L.F.; Deng, J.S.; Shi, Y.Y.; Chen, L.S.; Wang, K. Investigation of SPAD Meter-Based Indices for Estimating Rice Nitrogen Status. Comput. Electron. Agric. 2010, 715, 60–65. [Google Scholar] [CrossRef]
  24. Blackmer, T.M.; Schepers, J.S. Use of a Chlorophyll Meter to Monitor Nitrogen Status and Schedule Fertigation for Corn. J. Prod. Agric. 1995, 8, 55–60. [Google Scholar] [CrossRef]
  25. Padilla, F.M.; Farneselli, M.; Gianquinto, G.; Tei, F.; Thompson, R.B. Monitoring Nitrogen Status of Vegetable Crops and Soils for Optimal Nitrogen Management. Agric. Water Manag. 2020, 241, 106356. [Google Scholar] [CrossRef]
  26. Yuan, Z.; Cao, Q.; Zhang, K.; Ata-Ul-Karim, S.T.; Tan, Y.; Zhu, Y.; Cao, W.; Liu, X. Optimal Leaf Positions for SPAD Meter Measurement in Rice. Front. Plant Sci. 2016, 7, 719. [Google Scholar] [CrossRef] [PubMed]
  27. Wang, Z.; Wang, J.; Zhao, C.; Zhao, M.; Huang, W.; Wang, C. Vertical Distribution of Nitrogen in Different Layers of Leaf and Stem and Their Relationship with Grain Quality of Winter Wheat. J. Plant Nutr. 2005, 28, 73–91. [Google Scholar] [CrossRef]
  28. Ata-Ul-Karim, S.T.; Zhu, Y.; Cao, Q.; Rehmani, M.I.A.; Cao, W.; Tang, L. In-Season Assessment of Grain Protein and Amylose Content in Rice Using Critical Nitrogen Dilution Curve. Eur. J. Agron. 2017, 90, 139–151. [Google Scholar] [CrossRef]
  29. Andleeb, T.; Ali, Z.; Mahmood, Z.; Latif, S.; Quraishi, U.M. Wheat Varietal Response to Relative SPAD Index (RSI) and Relative Normalized Difference Vegetation Index (RNDVI) under Variable Nitrogen Application and Terminal Heat Stress along with Yield Repercussion. Agronomy 2022, 12, 1538. [Google Scholar] [CrossRef]
  30. Zhao, B.; Liu, Z.; Ata-Ul-Karim, S.T.; Xiao, J.; Liu, Z.; Qi, A.; Ning, D.; Nan, J.; Duan, A. Rapid and Nondestructive Estimation of the Nitrogen Nutrition Index in Winter Barley Using Chlorophyll Measurements. Field Crop. Res. 2016, 185, 59–68. [Google Scholar] [CrossRef]
  31. Grasso, R.; Peña-Fleitas, M.T.; de Souza, R.; Rodríguez, A.; Thompson, R.B.; Gallardo, M.; Padilla, F.M. Nitrogen Effect on Fruit Quality and Yield of Muskmelon and Sweet Wepper Cultivars. Agronomy 2022, 12, 2230. [Google Scholar] [CrossRef]
  32. Fernández, E.J.; Camacho-Ferre, F. Manual Práctico de Fertirrigación en Riego por Goteo, 2nd ed.; Ediciones Agrotécnicas: Madrid, España, 2007. [Google Scholar]
  33. Huang, S.; Miao, Y.; Yuan, F.; Cao, Q.; Ye, H.; Lenz-Wiedemann, V.I.S.; Bareth, G. In-Season Diagnosis of Rice Nitrogen Status Using Proximal Fluorescence Canopy Sensor at Different Growth Stages. Remote Sens. 2019, 11, 1847. [Google Scholar] [CrossRef]
  34. Bürling, K.; Cerovic, Z.G.; Cornic, G.; Ducruet, J.-M.; Noga, G.; Hunsche, M. Fluorescence-Based Sensing of Drought-Induced Stress in the Vegetative Phase of Four Contrasting Wheat Genotypes. Environ. Exp. Bot. 2013, 89, 51–59. [Google Scholar] [CrossRef]
  35. Tremblay, N.; Wang, Z.; Cerovic, Z.G. Sensing Crop Nitrogen Status with Fluorescence Indicators. A Review. Agron. Sustain. Dev. 2012, 32, 451–464. [Google Scholar] [CrossRef]
  36. Cartelat, A.; Cerovic, Z.G.; Goulas, Y.; Meyer, S.; Lelarge, C.; Prioul, J.L.; Barbottin, A.; Jeuffroy, M.H.; Gate, P.; Agati, G.; et al. Optically Assessed Contents of Leaf Polyphenolics and Chlorophyll as Indicators of Nitrogen Deficiency in Wheat (Triticum aestivum L.). Field Crop. Res. 2005, 91, 35–49. [Google Scholar] [CrossRef]
  37. Padilla, F.M.; Peña-Fleitas, M.T.; Gallardo, M.; Thompson, R.B. Proximal Optical Sensing of Cucumber Crop N Status Using Chlorophyll Fluorescence Indices. Eur. J. Agron. 2016, 73, 83–97. [Google Scholar] [CrossRef]
  38. Samborski, S.M.; Tremblay, N.; Fallon, E. Strategies to Make Use of Plant Sensors-Based Diagnostic Information for Nitrogen Recommendations. Agron. J. 2009, 101, 800–816. [Google Scholar] [CrossRef]
  39. Solari, F.; Shanahan, J.; Ferguson, R.; Schepers, J.; Gitelson, A. Active Sensor Reflectance Measurements of Corn Nitrogen Status and Yield Potential. Agron. J. 2008, 100, 571–579. [Google Scholar] [CrossRef]
  40. Sellers, P.J. Canopy Reflectance, Photosynthesis and Transpiration. Int. J. Remote Sens. 1985, 6, 1335–1372. [Google Scholar] [CrossRef]
  41. Padilla, F.M.; Peña-Fleitas, T.M.; Gallardo, M.; Thompson, R.B. Evaluation of Optical Sensor Measurements of Canopy Reflectance and of Leaf Flavonols and Chlorophyll Contents to Assess Crop Nitrogen Status of Muskmelon. Eur. J. Agron. 2014, 58, 39–52. [Google Scholar] [CrossRef]
  42. Padilla, F.M.; Peña-Fleitas, M.T.; Gallardo, M.; Thompson, R.B. Threshold Values of Canopy Reflectance Indices and Chlorophyll Meter Readings for Optimal Nitrogen Nutrition of Tomato. Ann. Appl. Biol. 2015, 166, 271–285. [Google Scholar] [CrossRef]
  43. Padilla, F.M.; Peña-Fleitas, M.T.; Gallardo, M.; Giménez, C.; Thompson, R.B. Derivation of Sufficiency Values of a Chlorophyll Meter to Estimate Cucumber Nitrogen Status and Yield. Comput. Electron. Agric. 2017, 141, 54–64. [Google Scholar] [CrossRef]
  44. Datta, U.; Das, R.; Chakraborty, K. Comparative Assessment of Leaf Chlorophyll Content of Seven Selected Vegetable Crop by Two Alternative Methods at Murshidabad, West Bengal. Int. J. Pharm. Biol. Sci. 2018, 8, 570–578. [Google Scholar]
  45. Bauerle, W.L.; McCullough, C.; Iversen, M.; Hazlett, M. Leaf Age and Position Effects on Quantum Yield and Photosynthetic Capacity in Hemp Crowns. Plants 2020, 9, 271. [Google Scholar] [CrossRef] [PubMed]
  46. McMurtrie, R.E.; Dewar, R.C. Leaf-Trait Variation Explained by the Hypothesis That Plants Maximize Their Canopy Carbon Export over the Lifespan of Leaves. Tree Physiol. 2011, 31, 1007–1023. [Google Scholar] [CrossRef]
  47. Guo, Y.; Ren, G.; Zhang, K.; Li, Z.; Miao, Y.; Guo, H. Leaf Senescence: Progression, Regulation, and Application. Mol. Hortic. 2021, 1, 5. [Google Scholar] [CrossRef]
  48. Wen, B.; Li, C.; Fu, X.; Li, D.; Li, L.; Chen, X.; Wu, H.; Cui, X.; Zhang, X.; Shen, H.; et al. Effects of Nitrate Deficiency on Nitrate Assimilation and Chlorophyll Synthesis of Detached Apple Leaves. Plant Physiol. Biochem. 2019, 142, 363–371. [Google Scholar] [CrossRef]
  49. Zakari, S.A.; Zaidi, S.H.R.; Sunusi, M.; Dauda, K.D. Nitrogen Deficiency Regulates Premature Senescence by Modulating Flag Leaf Function, ROS Homeostasis, and Intercellular Sugar Concentration in Rice during Grain Filling. J. Genet. Eng. Biotechnol. 2021, 19, 177. [Google Scholar] [CrossRef]
  50. Prost, L.; Jeuffroy, M.H. Replacing the Nitrogen Nutrition Index by the Chlorophyll Meter to Assess Wheat N Status. Agron. Sustain. Dev. 2007, 27, 321–330. [Google Scholar] [CrossRef]
Figure 1. Research diagram including the independent variables, dependent variables, calibration, and validation information. N1 is the very deficient N treatment, N2 is the deficient N treatment, and N3 is the conventional N treatment. L1 (L1ref), L2, L3 and L4 (L4ref), refer to the latest (most recent) second, third and fourth fully expanded leafs, respectively.
Figure 1. Research diagram including the independent variables, dependent variables, calibration, and validation information. N1 is the very deficient N treatment, N2 is the deficient N treatment, and N3 is the conventional N treatment. L1 (L1ref), L2, L3 and L4 (L4ref), refer to the latest (most recent) second, third and fourth fully expanded leafs, respectively.
Remotesensing 15 02174 g001
Figure 2. The relationships between absolute SPAD measurements and several normalized SPAD indices (DI, RI, RDS, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Figure 2. The relationships between absolute SPAD measurements and several normalized SPAD indices (DI, RI, RDS, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Remotesensing 15 02174 g002
Figure 3. The relationships between absolute SPAD measurements and several normalized SPAD indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent) second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Figure 3. The relationships between absolute SPAD measurements and several normalized SPAD indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent) second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Remotesensing 15 02174 g003
Figure 4. The relationships between absolute NBI measurements and several normalized NBI indices (DI, RI, RDS, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Figure 4. The relationships between absolute NBI measurements and several normalized NBI indices (DI, RI, RDS, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Remotesensing 15 02174 g004
Figure 5. The relationships between absolute NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent) second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Figure 5. The relationships between absolute NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent) second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Remotesensing 15 02174 g005
Figure 6. The relationships between absolute NDVI measurements and several normalized NDVI indices (DI, RI, RDS, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Figure 6. The relationships between absolute NDVI measurements and several normalized NDVI indices (DI, RI, RDS, NDI) with total fruit yield (kg m−2) at different phenological stages of sweet pepper, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, p > 0.05.
Remotesensing 15 02174 g006
Figure 7. The relationships between absolute NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent) second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; ns, p > 0.05.
Figure 7. The relationships between absolute NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI) with total fruit yield (kg m−2) at different phenological stages of muskmelon, for calibration data. L1, L2, L3 and L4, refer to the latest (most recent) second, third and fourth fully expanded leaf, respectively. Asterisks show significance of regressions: ***, p < 0.001; ns, p > 0.05.
Remotesensing 15 02174 g007
Table 1. Mineral N (NO3–N + NH4+–N) in the soil (0–0.4 m depth) at the beginning of each crop and year, N concentration in the applied nutrient solution, and amount of mineral N applied in fertilization.
Table 1. Mineral N (NO3–N + NH4+–N) in the soil (0–0.4 m depth) at the beginning of each crop and year, N concentration in the applied nutrient solution, and amount of mineral N applied in fertilization.
CropGreenhouse NumberYear of TransplantN TreatmentMineral N at Planting (kg N ha−1)N Concentration in Nutrient Solution (mmol L−1)N Amount Applied
(kg N ha−1)
Sweet pepperU132020Very deficient (N1)312.266
Deficient (N2)188.4428
Conventional (N3)4214.2704
U142021Very deficient (N1)3401.970
Deficient (N2)3468.2337
Conventional (N3)29014.2615
MuskmelonU132020Very deficient (N1)632.761
Deficient (N2)508.3302
Conventional (N3)3814.0582
U132021Very deficient (N1)132.757
Deficient (N2)778.0228
Conventional (N3)7914.6515
Table 2. Measurement dates of proximal optical sensors based on the day after transplanting (DAT).
Table 2. Measurement dates of proximal optical sensors based on the day after transplanting (DAT).
CropYear of TransplantPhenological Stage (DAT)
VegetativeFloweringEarly Fruit GrowthHarvest
Sweet pepper2020427099127
2021416998125
Muskmelon202035496378
202132466074
Table 3. Equations of normalized indices calculated for different proximal optical sensors. L1 is the latest fully expanded leaf. Lj is the leaf position and varies from 2 to 4.
Table 3. Equations of normalized indices calculated for different proximal optical sensors. L1 is the latest fully expanded leaf. Lj is the leaf position and varies from 2 to 4.
Normalized IndexEquation
Difference Index D I = L 1 L j
Relative Index R I = L 1 / L j
Relative Difference Index R D I = L 1 / ( L 1 + L j )
Normalized Difference Index N D I = ( L 1 L j ) / ( L 1 + L j )
Table 4. Statistical parameters used to evaluate yield models from optical sensor measurements. R2 and RMSE were used for calibration analyses. R2, RMSE, RE, MBE, and d, were used for validation analyses. Xi, Yi, X ¯ i , and Y ¯ i represent sensor measurements, yield values, mean sensor measurements, and mean yield values, respectively, in the calibration stage. In the validation stage, Xi, Yi, X ¯ i , and Y ¯ i represent observed yield, estimated yield, mean observed yield, and mean estimated yield. n is the number of observations.
Table 4. Statistical parameters used to evaluate yield models from optical sensor measurements. R2 and RMSE were used for calibration analyses. R2, RMSE, RE, MBE, and d, were used for validation analyses. Xi, Yi, X ¯ i , and Y ¯ i represent sensor measurements, yield values, mean sensor measurements, and mean yield values, respectively, in the calibration stage. In the validation stage, Xi, Yi, X ¯ i , and Y ¯ i represent observed yield, estimated yield, mean observed yield, and mean estimated yield. n is the number of observations.
ParameterCalculation
Coefficient of Determination R 2 = [ i = 1 n ( X i X ¯ i ) ( Y i Y ¯ i ) ] i = 1 n ( X i X ¯ i ) 2 i = 1 n ( Y i Y ¯ i ) 2
Root Mean Square Error R M S E = i = 1 n ( X i Y i ) 2 n
Relative Error R E = R M S E Y ¯
Mean Bias Error M B E = i = 1 n ( X i Y i ) n
Willmott Index d = 1 i = 1 n ( X i Y i ) 2 i = n n ( ( X i Y ¯ i ) + ( Y i Y ¯ i ) ) 2
Table 5. Results of the validation analysis of the estimated fruit yield (kg m−2) from SPAD measurements and several normalized SPAD indices (DI, RI, RDI, NDI), at different phenological stages of sweet pepper. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively.
Table 5. Results of the validation analysis of the estimated fruit yield (kg m−2) from SPAD measurements and several normalized SPAD indices (DI, RI, RDI, NDI), at different phenological stages of sweet pepper. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively.
IndexVegetativeFloweringEarly Fruit GrowthHarvest
R2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREd
SPADL10.690.0420.83−0.140.150.850.780.0200.72−0.080.130.890.850.0090.570.000.100.940.860.0080.630.040.120.92
SPADL20.620.0440.91−0.310.160.840.830.0110.66−0.090.120.910.920.0030.470.020.090.960.890.0040.580.000.100.94
SPADL30.670.0450.85−0.350.150.870.820.0130.610.040.110.930.910.0030.48−0.080.090.960.96<0.0010.430.030.080.97
SPADL40.710.0350.77−0.230.130.900.900.0040.47−0.140.080.960.850.0090.58−0.190.100.940.890.0050.57−0.010.100.94
DIL1–L20.310.2481.38−0.790.220.600.110.5291.36−0.150.240.040.030.7501.51−0.020.270.320.330.2371.22−0.400.210.71
DIL1–L30.420.1621.69−0.990.260.690.030.7601.46−0.320.250.030.050.6772.06−1.220.310.190.090.5641.38−0.240.240.44
DIL1–L40.390.1841.19−0.340.200.750.220.3531.55−0.060.280.040.010.8601.400.190.260.240.010.8851.53−0.270.260.06
RIL1–L20.300.2631.36−0.740.220.550.120.4961.35−0.160.240.080.220.3511.450.050.270.650.610.0660.91−0.310.160.87
RIL1–L30.430.1601.44−0.870.230.710.120.5031.48−0.330.250.000.210.3671.82−0.620.300.640.370.1981.21−0.180.210.76
RIL1–L40.370.1991.14−0.270.200.730.320.2421.57−0.090.280.030.030.7571.37−0.160.240.020.050.6571.85−0.450.310.45
RDIL1–L20.290.2661.37−0.740.220.550.120.4941.35−0.150.240.080.210.3581.460.040.270.650.600.0700.93−0.320.160.86
RDIL1–L30.420.1671.48−0.890.230.700.120.5011.48−0.340.250.000.190.3821.87−0.650.300.630.350.2121.24−0.190.220.75
RDIL1–L40.370.2031.15−0.290.200.730.320.2411.57−0.090.280.030.030.7621.37−0.160.240.030.060.6521.86−0.480.310.46
NDIL1–L20.290.2661.37−0.740.220.550.120.4941.35−0.150.240.080.210.3581.460.040.270.650.600.0700.93−0.320.160.86
NDIL1–L30.420.1671.48−0.890.230.700.120.5011.48−0.340.250.000.190.3821.87−0.650.300.630.350.2121.24−0.200.220.75
NDIL1–L40.370.2031.15−0.290.200.730.320.2411.57−0.090.280.030.030.7621.37−0.160.240.030.060.6521.86−0.480.310.46
Table 6. Results of the validation analysis of the estimated fruit yield (kg m−2) from SPAD measurements and several normalized SPAD indices (DI, RI, RDI, NDI), at different phenological stages of muskmelon. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively.
Table 6. Results of the validation analysis of the estimated fruit yield (kg m−2) from SPAD measurements and several normalized SPAD indices (DI, RI, RDI, NDI), at different phenological stages of muskmelon. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively.
IndexVegetativeFloweringEarly Fruit GrowthHarvest
R2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREd
SPADL10.300.2561.580.020.280.610.160.4371.780.180.320.570.160.4271.91−0.120.320.610.150.4421.860.130.330.59
SPADL20.330.2371.54−0.110.260.690.270.2911.640.190.300.680.130.4761.93−0.010.340.570.180.3951.830.060.320.63
SPADL30.420.1641.430.120.250.750.230.3361.750.040.310.670.120.4961.88−0.080.320.540.230.3361.690.240.310.65
SPADL40.410.1681.430.080.250.770.200.3791.860.050.330.640.270.2941.680.150.300.690.220.3471.690.100.300.64
DIL1–L20.450.1451.67−0.760.260.560.490.1211.680.070.300.370.140.4721.82−0.060.310.150.010.8672.010.210.360.30
DIL1–L30.550.0901.510.760.300.830.430.1581.890.060.330.000.060.6481.980.030.350.510.020.7882.07−0.150.350.29
DIL1–L40.490.1201.400.300.260.810.260.2961.850.040.320.030.150.4413.15−1.060.460.020.010.8802.150.150.380.29
RIL1–L20.390.1841.79−0.670.280.350.510.1091.790.060.310.170.210.3661.79−0.100.310.190.010.8532.000.220.360.31
RIL1–L30.560.0891.550.860.320.820.490.1232.140.180.380.040.090.5612.01−0.050.350.580.020.8132.07−0.120.350.25
RIL1–L40.440.1521.460.340.270.770.350.2192.110.130.380.030.150.4543.23−1.200.460.030.010.8702.210.210.400.28
RDIL1–L20.390.1841.79−0.680.280.360.510.1131.790.060.310.170.210.3671.79−0.120.310.200.010.8572.000.220.360.31
RDIL1–L30.550.0911.560.860.320.820.480.1252.140.180.380.040.090.5552.01−0.050.350.580.020.8132.05−0.120.350.24
RDIL1–L40.440.1491.460.360.270.780.340.2262.130.130.380.030.150.4473.20−1.190.460.030.010.8732.200.210.400.28
NDIL1–L20.390.1841.79−0.670.280.360.510.1131.790.060.310.170.210.3671.79−0.120.310.200.010.8572.000.220.360.31
NDIL1–L30.550.0911.560.870.320.820.480.1252.140.180.380.040.090.5552.01−0.050.350.580.020.8132.05−0.120.350.24
NDIL1–L40.440.1491.460.360.270.780.340.2262.130.140.380.030.150.4473.20−1.190.460.030.010.8732.200.210.400.28
Table 7. Results of the validation analysis of the estimated fruit yield (kg m−2) from NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI), at different phenological stages of sweet pepper. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively.
Table 7. Results of the validation analysis of the estimated fruit yield (kg m−2) from NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI), at different phenological stages of sweet pepper. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively.
IndexVegetativeFloweringEarly Fruit GrowthHarvest
R2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREd
NBIL10.870.0070.57−0.180.100.940.97<0.0010.40−0.060.070.970.710.0350.730.010.130.910.620.0610.88−0.050.160.82
NBIL20.870.0070.54−0.110.100.950.940.0010.47−0.120.080.960.760.0240.670.020.120.920.600.0700.88−0.130.160.84
NBIL30.880.0060.59−0.140.100.930.920.0020.44−0.110.080.970.820.0120.600.000.110.940.660.0480.84−0.030.150.84
NBIL40.870.0070.62−0.280.110.930.98<0.0010.33−0.200.060.980.860.0080.550.060.100.950.650.0510.84−0.130.150.85
DIL1–L20.030.7251.38−0.130.240.020.330.2311.30−0.180.230.210.840.0100.99−0.090.180.690.320.2451.41−0.620.230.68
DIL1–L30.000.9421.42−0.100.250.230.350.2193.923.531.980.110.750.0250.88−0.100.160.800.470.1321.04−0.030.190.69
DIL1–L40.020.7651.40−0.250.240.260.910.0031.24−0.210.220.330.760.0240.750.080.140.880.500.1161.04−0.400.180.79
RIL1–L20.450.1441.08−0.290.190.690.830.0120.740.050.130.880.550.0940.96−0.060.170.760.470.1301.040.330.200.79
RIL1–L30.220.3531.29−0.150.230.290.780.0200.67−0.010.120.910.630.0600.850.110.160.880.400.1751.09−0.070.200.77
RIL1–L40.130.4911.32−0.110.240.190.96<0.0010.530.120.100.940.590.0750.880.000.160.860.580.0770.900.200.170.85
RDIL1–L20.480.1271.06−0.290.180.700.850.0090.730.020.130.880.550.0900.97−0.090.170.750.500.1171.050.400.210.81
RDIL1–L30.250.3151.28−0.140.230.300.810.0140.65−0.030.120.920.650.0520.840.140.160.890.390.1871.12−0.030.200.77
RDIL1–L40.140.4571.32−0.110.240.180.97<0.0010.500.080.090.950.610.0670.860.030.160.870.590.1080.910.230.170.86
NDIL1–L20.480.1271.07−0.290.180.700.850.0090.730.020.130.880.550.0900.97−0.090.170.750.500.1171.050.400.210.81
NDIL1–L30.250.3151.28−0.140.230.300.810.0140.65−0.030.120.920.650.0520.840.140.160.890.390.1871.12−0.030.200.77
NDIL1–L40.140.4571.32−0.110.240.180.97<0.0010.500.080.090.950.610.0670.860.030.160.870.590.1080.910.230.170.86
Table 8. Results of the validation analysis of the estimated fruit yield (kg m−2) from NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI), at different phenological stages of muskmelon. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively.
Table 8. Results of the validation analysis of the estimated fruit yield (kg m−2) from NBI measurements and several normalized NBI indices (DI, RI, RDI, NDI), at different phenological stages of muskmelon. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1, L2, L3 and L4, refer to the latest (most recent), second, third and fourth fully expanded leaf, respectively.
IndexVegetativeFloweringEarly Fruit GrowthHarvest
R2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREd
NBIL10.220.3521.85−0.070.320.670.210.3552.09−0.160.350.670.270.2871.88−0.120.320.710.240.3191.820.070.320.70
NBIL20.270.2871.79−0.100.310.710.240.3211.98−0.100.340.700.240.3211.96−0.070.340.690.240.3191.820.040.320.70
NBIL30.390.1821.66−0.140.280.780.260.3001.96−0.110.330.710.220.3472.03−0.020.350.670.220.3431.88−0.030.320.68
NBIL40.460.1361.52−0.160.260.820.300.2621.84−0.030.320.730.240.3251.99−0.110.340.680.200.3681.99−0.130.340.65
DIL1–L20.170.4211.80−0.040.310.620.060.6542.250.340.420.490.030.7582.340.260.430.470.020.7791.940.160.350.41
DIL1–L30.570.0831.24−0.090.210.830.220.3461.690.120.300.630.020.8032.310.400.430.410.220.3451.820.220.330.16
DIL1–L40.640.0541.13−0.130.190.870.190.8441.810.320.330.610.100.5452.03−0.030.350.540.010.8241.950.240.350.07
RIL1–L20.010.8261.91−0.020.330.280.010.8351.87−0.010.320.260.030.7502.320.580.450.510.060.6371.880.170.340.49
RIL1–L30.170.4101.72−0.010.300.430.010.8281.860.050.330.040.010.9992.370.720.470.340.450.1441.770.240.320.25
RIL1–L40.230.3401.65−0.130.280.600.010.9011.870.090.330.160.150.4521.840.140.330.600.010.8941.940.320.360.26
RDIL1–L20.020.7981.89−0.020.330.290.010.8421.87−0.010.320.260.030.7442.290.560.440.510.060.6421.870.170.340.49
RDIL1–L30.180.3961.710.000.300.430.020.8121.860.050.330.030.010.9832.330.690.460.340.460.1381.760.230.320.25
RDIL1–L40.230.3321.64−0.100.280.600.010.8811.870.080.330.150.140.4631.850.140.330.590.010.8891.930.320.360.26
NDIL1–L20.020.7981.89−0.020.330.290.010.8421.87−0.010.320.260.030.7442.290.550.440.510.060.6421.870.170.340.49
NDIL1–L30.180.3961.710.000.300.430.020.8121.860.050.330.030.010.9832.330.690.460.340.460.1381.760.230.320.25
NDIL1–L40.230.3321.64−0.100.280.600.010.8811.870.080.330.150.140.4631.850.140.330.590.010.8891.930.320.360.26
Table 9. Results of the validation analysis of the estimated fruit yield (kg m−2) from NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI), at different phenological stages of sweet pepper. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1ref and L4ref, refer to the latest (most recent) and fourth fully expanded leaf, respectively.
Table 9. Results of the validation analysis of the estimated fruit yield (kg m−2) from NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI), at different phenological stages of sweet pepper. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1ref and L4ref, refer to the latest (most recent) and fourth fully expanded leaf, respectively.
IndexVegetativeFloweringEarly Fruit GrowthHarvest
R2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREd
NDVI L1ref0.400.1751.10−0.230.190.650.870.0070.57−0.100.100.940.600.0720.87−0.090.160.840.840.0100.58−0.130.100.94
NDVI L4ref0.930.0020.74−0.300.130.880.870.0060.57−0.080.100.940.700.0380.77−0.070.140.880.830.0120.57−0.030.100.95
DIL1ref–L4ref0.920.0030.73−0.310.130.880.130.4831.34−0.150.240.130.280.2781.52−0.130.270.000.140.4641.39−0.170.250.00
RIL1ref–L4ref0.940.0020.74−0.340.130.880.190.3951.28−0.160.230.320.300.2571.50−0.140.260.000.180.4041.40−0.180.250.00
RDIL1ref–L4ref0.940.0020.73−0.330.130.880.190.3931.28−0.150.230.320.300.2601.50−0.140.270.000.180.4011.40−0.180.250.00
NDIL1ref–L4ref0.940.0020.73−0.330.130.880.190.3931.28−0.150.230.320.300.2601.50−0.140.270.000.180.4011.40−0.180.250.00
Table 10. Results of the validation analysis of the estimated fruit yield (kg m−2) from NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI), at different phenological stages of muskmelon. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1ref and L4ref, refer to the latest (most recent) and fourth fully expanded leaf, respectively.
Table 10. Results of the validation analysis of the estimated fruit yield (kg m−2) from NDVI measurements and several normalized NDVI indices (DI, RI, RDI, NDI), at different phenological stages of muskmelon. DI, Difference Index; RI, Relative Index; RDI, Relative Difference Index; NDI, Normalized Difference Index, R2, coefficient of determination; RMSE, root mean square error; RE, relative error; MBE, mean bias error; d, Willmott Index. L1ref and L4ref, refer to the latest (most recent) and fourth fully expanded leaf, respectively.
IndexVegetativeFloweringEarly Fruit GrowthHarvest
R2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREdR2p ValueRMSEMBEREd
NDVI L1ref0.840.0100.750.140.130.950.530.1001.30−0.220.220.830.320.2411.60−0.200.270.720.370.2041.54−0.140.260.75
NDVI L4ref0.540.0951.460.450.280.850.710.0351.06−0.200.180.880.410.1741.500.070.260.790.490.1191.44−0.270.240.83
DIL1red–L4ref0.210.3641.720.140.310.420.010.9212.010.210.360.200.010.9732.62−0.670.410.320.010.8901.890.250.340.20
RIL1ref–L4ref0.240.3291.730.110.310.360.010.8882.030.190.370.190.010.9762.30−0.490.370.300.010.9301.900.240.340.17
RDIL1ref–L4ref0.220.3491.730.100.310.380.010.8922.030.220.370.190.010.9762.31−0.510.370.300.010.9244.20−3.750.440.07
NDIL1ref–L4ref0.220.3491.730.100.310.380.010.8922.030.200.370.190.010.9762.31−0.510.370.300.010.9241.900.240.340.18
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Karaca, C.; Thompson, R.B.; Peña-Fleitas, M.T.; Gallardo, M.; Padilla, F.M. Evaluation of Absolute Measurements and Normalized Indices of Proximal Optical Sensors as Estimators of Yield in Muskmelon and Sweet Pepper. Remote Sens. 2023, 15, 2174. https://doi.org/10.3390/rs15082174

AMA Style

Karaca C, Thompson RB, Peña-Fleitas MT, Gallardo M, Padilla FM. Evaluation of Absolute Measurements and Normalized Indices of Proximal Optical Sensors as Estimators of Yield in Muskmelon and Sweet Pepper. Remote Sensing. 2023; 15(8):2174. https://doi.org/10.3390/rs15082174

Chicago/Turabian Style

Karaca, Cihan, Rodney B. Thompson, M. Teresa Peña-Fleitas, Marisa Gallardo, and Francisco M. Padilla. 2023. "Evaluation of Absolute Measurements and Normalized Indices of Proximal Optical Sensors as Estimators of Yield in Muskmelon and Sweet Pepper" Remote Sensing 15, no. 8: 2174. https://doi.org/10.3390/rs15082174

APA Style

Karaca, C., Thompson, R. B., Peña-Fleitas, M. T., Gallardo, M., & Padilla, F. M. (2023). Evaluation of Absolute Measurements and Normalized Indices of Proximal Optical Sensors as Estimators of Yield in Muskmelon and Sweet Pepper. Remote Sensing, 15(8), 2174. https://doi.org/10.3390/rs15082174

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop