Introduction

Nitrogen (N) is an essential nutrient and is one of the most important elements for plant development and growth, affecting yield and fruit quality (Greenwood et al., 1991; Lemaire et al., 2008). To guarantee high yields, N fertilizer is commonly applied in high amounts in vegetable production systems (Neeteson et al., 1999). Commonly, in vegetable production, the combined N supply, consisting of fertilizer N plus available soil N, appreciably exceed crop N requirements (Ju et al., 2006; Thompson et al., 2017b). The excess N is susceptible to nitrate (NO3−) leaching losses (Thompson et al., 2007; Zotarelli et al., 2009), and subsequent environmental contamination (Meisinger et al., 2008). Nitrate contamination of groundwater, from vegetable production, has been reported for diverse regions, such as south-eastern (SE) Spain (Pulido-Bosch et al., 2000), SE United States (Zotarelli et al., 2009), and China (Ju et al., 2006).

Matching N supply with crop N requirements quantitatively over time will optimize N use and decrease the associated environmental problems (Samborski et al., 2009; Thompson et al., 2017b; Zhang et al., 2012). Vegetable production systems increasingly have the technical potential to match N supply to crop requirements through the combined use of localized irrigation and fertigation systems (Thompson et al., 2017a, 2017b). Tools that can assess crop N status are required to ensure optimal N management in intensive production systems, by permitting correction of N fertilizer additions added frequently through combined fertigation and localized irrigation.

A convenient way to determine crop N nutrition status on farm is through the use of portable optical sensors. There are several types of proximal optical sensors with the capacity to evaluate crop N status; the most important are chlorophyll meters, canopy reflectance sensors, and flavonols meters (Padilla et al., 2018; Samborski et al., 2009; Tremblay et al., 2012). These sensors estimate the N content through the use of optical properties of the light which interact with compounds sensitive to N content (Fox & Walthall, 2008; Tremblay et al., 2012). The major advantages of these sensors are speed and simplicity of measurement, and that they can determine crop N status non-destructively (Samborski et al., 2009; Tremblay et al., 2012).

Two groups of compounds in plants are particularly sensitive to plant N content, chlorophyll and polyphenols (Agati et al., 2013; Fox & Walthall, 2008; Tremblay et al., 2012). Within polyphenols, flavonols are the compounds that are easier to estimate by using optical tools (Meyer et al., 2006) Fluorescence-based sensors can estimate chlorophyll and flavonols contents of leaf tissue in-situ (Padilla et al., 2018). These sensors estimate chlorophyll content from the chlorophyll fluorescence emission ratio of red (RF), and far-red radiation (FRF) emitted from chlorophyll after excitation with ultraviolet (UV), red, green or blue radiation (Tremblay et al., 2012). Flavonols content in leaves has the opposite behaviour to that of chlorophyll as function of N, and can also be directly affected by other environmental factors such as light. When N is deficient, flavonols content increases and chlorophyll content decreases (Bragazza & Freeman, 2007; Liu et al., 2010). The Nitrogen Balance Index (NBI) is calculated as the ratio between chlorophyll and flavonols contents; it has been shown to be a very sensitive indicator of crop N status (Cartelat et al., 2005; Padilla et al., 2016; Samborski et al., 2009).

Fluorimeters are a kind of optical sensors which provide indirect measurements of both chlorophyll and flavonols contents from the fluorescence properties of the leaves (Padilla et al., 2018; Thompson et al., 2017b). These sensors can also be used to estimate other crop properties not dependent on the chlorophyll content such as leaf health status and light use efficiency (Schreiber & Bilger, 1987). Two fluorimeters developed for agriculture are the Dualex, a leaf-clip sensor, and the Multiplex, both produced by Force-A (Orsay, France). The Multiplex is a proximal sensor that take measurements at a distance of 10 cm from the leaf (Padilla et al., 2018; Thompson et al., 2017b).

The evaluation of fluorescence measurements as predictors of crop variables that are important for crop N management has been made by establishing relationships between fluorescence measurements and crop variables (Padilla et al., 2016). Most of the literature has focused on the prediction of crop N status but it remains largely unaddressed if measurements of fluorescence-based sensors can predict other crop variables such as crop N uptake and yield (Huang et al., 2019).

This manuscript evaluates the capacity of five fluorescence indices, measured with the Multiplex sensor, to predict the following crop variables: dry matter production, crop N content, crop N uptake, Nitrogen Nutrition Index (NNI), and yield (both absolute and relative yield respect to the maximum yield), in sweet pepper crops. Subsequently regressions were validated for phenological stages. The work was carried out in an intensive vegetable production system in Almería, SE Spain. This area is one of the most important for production of vegetables crops in greenhouse in Spain, occupying 30,000 ha. Within the vegetable crops, sweet pepper is one of the most important crops, occupying an area of 8,000 ha (Valera et al., 2017).

Material and methods

Experimental crops and N treatments

This study was carried with sweet pepper (Capsicum annuum cultivar ‘Melchor’) grown in three different years at the Experimental Station of the University of Almeria (36◦51’N, 2◦16’W and 92 m elevation), in Retamar, Almeria, Spain. The crops were grown in soil in a plastic greenhouse. The soil was an artificial layered “enarenado” typical of this area (Thompson et al., 2007). A complete description of the greenhouse and the enarenado soil are presented in Thompson et al. (2007). A complete list of acronyms used is presented in Table 1.

Table 1 List of acronyms used
Full size table

The sweet pepper crop was repeated over three crop** seasons (2014, 2016, and 2017; Table 2, Fig. 1), with a summer–winter cycle. Each crop was subjected to five different N treatments, applied throughout the crop in the nutrient solution applied by a combined drip irrigation and fertigation system. The N treatments were very N deficient (N1), N deficient (N2), conventional N (N3), excessive N (N4) and very excessive N (N5) (Table 2). The majority of N was applied as nitrate (90%), and the rest as ammonium (NH4+). In addition to N, the other macronutrients applied remained constant in all treatments in the following concentrations: H2PO4-, 1.75 mmol L−1; K+, 4 mmol L−1; Ca+2, 4 mmol L−1; Mg+2, 1.5 mmol L−1; SO4−2, 2.35 mmol L−1; on average for the three crop** seasons. The different N treatments were applied in every irrigation, which were made every 1–4 days. The irrigation was scheduled using tensiometers (Irrometer, Co., Riverside, Ca, USA) that were installed at 15 cm depth. Crop management followed local practices.

Table 2 Characteristics of the three crop** cycles and the corresponding N treatments applied in the five N treatments
Full size table
Fig. 1
figure 1

a Daily integral of solar radiation and b daily average air temperature, in the greenhouse during the three crop** seasons (2014–2015, 2016–2017, and 2017–2018), from transplant to end of the crop

Full size image

The experimental design was a randomized block design, with four replications per treatment. Each replicate plot measured 6 m by 6 m. There was a total of 20 plots. In each replicate plot there were 72 plants with a density of two plants m−2. One emitter with a flow of 3 L h−1 was immediately adjacent to each plant.

Dry matter production, crop N uptake, crop N content, NNI and yield

Periodic above-ground biomass samplings were made in each of the three crops. In the 2014–2015 crop, the biomass samplings were made every 21 days, in the 2016–2017 every 23 days, and in the 2017–2018 crop every 24 days.

At each sampling date, dry matter production (DMP) and N content were determined from two complete plants for each replicate. The dry weights (oven-dried at 65ºC) of the different plant organs (stems, leaves, and fruits) were determined. Additionally, harvested fruit and pruned shoot material were periodically removed from eight marked plants for replicate plot throughout each crop. The total amounts of harvest fruit and pruned material were calculated by summing the data per plot per treatment. Subsamples of oven-dry material were ground to a fine powder prior to analysis of N content (%) in a Dumas-type elemental analyzer (Rapid N, Elementar Analysensysteme GmbH, Hanau, Germany).

The total N in each organ was calculated multiplying the %N of the sub-sample by the corresponding dry matter mass. Total crop N uptake (kg N ha−1) was the sum of N in all relevant components. Total crop N content (%N) was calculated as total crop N uptake divided by total DM.

The NNI was calculated from the critical N curve derived for greenhouse-grown sweet pepper crop (Critical N = 4.71·DMP−0.22; Rodríguez et al., (2020)). The NNI was calculated as:

$$NNI = \frac{Nact}{{Nc}}$$
(1)

where Nact is the real N measured and Nc is the critical N content calculated from the critical N curve. The NNI value for each day of the crop was calculated by interpolating values between two consecutive biomass samplings (Padilla et al., 2017).

Multiplex measurements

A Multiplex® 3.6 sensor (Force-A, Orsay, France) was used to measure chlorophyll fluorescence throughout the three crop** seasons. This sensor is a portable device with four emission light sources (UV, green, red and blue) that induce fluorescence in plant tissues (Huang et al., 2019). Detailed descriptions of the sensor and its operation are available in Bürling et al. (2013) and Tremblay et al. (2012). The measurements were made weekly in the 2014–2015 crop and every two weeks for the other two crops. Measurements started at 15, 25 and 21 DAT in the first, second and third crops, respectively.

Measurements were made throughout the growing cycle in 16 marked plants per replicate plot. In each of these 16 plants, measurements were made on the most recently fully expanded leaf, following the protocols of Padilla et al., (2018). The measurement area was an 8 cm diameter circle of leaf surface.

Of the various indices measured by the Multiplex sensor (Huang et al., 2019), the present work focused on five indices that are most sensitive to plant N status (Tremblay et al., 2012). These indices were: i) the Simple Fluorescence Ratio (SFR), both under red (SFR-R) and green (SFR-G) excitation, indicative of leaf chlorophyll content (Tremblay et al., 2012); ii) the FLAV index, indicative of leaf flavonols content (Cerovic et al., 2002); and iii) the Nitrogen Balance Index (NBI) (Cartelat et al., 2005), either under red (NBI-R) or green (NBI-G) excitation.

Data analysis

Sweet pepper is an indeterminate crop with multiples harvests during the cycle. The major phenological stages considered were: vegetative, flowering, early fruit growth, and harvest (de Souza et al., 2019). Integrated values (Iv) of fluorescence indices, DMP, crop N content, crop N uptake and NNI for each phenological stage, were calculated as:

$$Iv \, = \, 1/D \cdot \Sigma \left( {V \cdot ds} \right)$$
(2)

where D was the duration of the phenological stage, V was the value of fluorescence indices or crop variable for each day of measurement, and ds was the duration between two successive measurements (de Souza et al., 2019).

Yield data were considered both in absolute terms (kg m−2) and in relative terms (%). In each crop, relative yield was calculated for each plot by dividing yield of each of the 20 plots by the maximum yield recorded in a plot of that particular crop, expressed as a percentage.

Data of integrated fluorescence indices and integrated crop variables were pooled across the three crops, for each phenological stage. Within each phenological stage, pooled data were randomly divided into two groups, one group was used for calibration and the other group was used for validation. The calibration group had approximately 2/3 of the data; the remaining data was in the validation group. For the calibration group, simple linear regression analyses were performed between each integrated fluorescence index, as the independent variable, and each integrated crop variable, as the dependent variable. These relationships assessed and compared the capacity of each fluorescence index to predict each crop variable. The CurveExpert Professional®2.2.0 software (Daniel G. Hyams, MS, USA) was used to retrieve statistical data of these relationships.

The relationships between each integrated fluorescence index and each integrated crop variable, at each phenological stage, were validated using the validation group. Validation consisted of calculating the predicted value of each integrated crop variable using the calibration equation, and then comparing with the observed values of each integrated crop variable of the validation dataset. Linear regression analysis was established between observed (independent variable) and predicted (dependent variable) values of each crop variable, and the root mean square error (RMSE) of the estimated crop variable was calculated as:

$$RMSE \, = \,\sqrt {\mathop \sum \limits_{1}^{n} \frac{{\left( {P_{i} - O_{i} } \right)^{2} }}{n}}$$
(3)

where n is the number of samples, Pi is the predicted value of the relationship, and Oi is the observed value (Zhao et al., 2018). Additionally, the relative error (RE) of the validation regression was calculated between values of each observed and predicted crop variables as:

$$RE:\,\frac{RMSE}{{\overline{Oi} }}$$
(4)

where Oi is the average of observed values.

The performance of fluorescence indices to predict crop variables was evaluated taking into account both calibration and validation results (** season advanced. The better performance found in the first half of the crop cycle for the chlorophyll index SFRi could be due to leaf chlorophyll content increasing steadily in the first half of the cycle (de Souza et al., 2019), and a decrease in chlorophyll content thereafter (de Souza et al., 2019), likely because of translocation of N from leaves to fruits (da Cunha et al., 2015). This could have caused the weaker performance of SFRi indices in the harvest stage. In contrast, there was better performance of FLAVi and NBIi in the harvest stage, compared to SFRi. A likely explanation for this contrasting behaviour may be in the dynamics of flavonols content throughout the crop. The FLAV index, measured with the Multiplex sensor as an indicator of flavonols content, was not different between N treatments in the vegetative stage (data not shown), but differences between treatments appeared at the middle of the crop and were maintained during the harvest stage and until the end of the crop (data not shown). This could explain the poor performance of FLAVi and NBIi in the vegetative stage and the better performance in later phenological stages.

The validation analysis showed that lower relative errors (mostly < 20%) were found in the prediction of crop N content, NNI and relative yield. According to the classification of Yang et al., (2016), these relative error values indicate good prediction capacity. However, relative errors were, in general, > 20% for the prediction of absolute yield, indicating a relatively poor capacity to predict absolute crop yield in sweet pepper. These results contradict Tremblay et al. (2010) who reported that fluorescence indices were able to predict yield in wheat. The likely explanation for this discrepancy may be the kind of crop; wheat is a determinate crop with a single harvest at the end of the crop, and greenhouse sweet pepper is an indeterminate crop with multiple fruit harvests that can span a period of 3–4 months.

Generally, the behaviour of fluorescence indices in the validation and calibration analysis were similar. The prediction of crop N content, crop N uptake, NNIi and relative yield, from SFRi values generally had slopes close to one and lower relative error values in the vegetative, flowering and early fruit growth stages. For FLAVi and NBIi, poorer prediction occurred in the vegetative and flowering stages, and the prediction notably improved (i.e., slopes took values closer to one and relative error values were lower) in the early fruit growth and harvest stages. Slopes close to one and lower RE values represent excellent validation of regression equations (Gallardo et al., 2014; Piñeiro et al., 2008). The analysis of the performance of fluorescence indices to predict crop variables considering both calibration and validation results was mostly consistent with the analysis conducted for calibration and validation separately (de Souza et al., 2020; **n-feng et al., 2013). The best predictions of crop N content and NNIi in the first half of the crop cycle (i.e., vegetative, flowering and early fruit growth stages) were with SFRi followed by NBIi and FLAVi. However, in the second half of the crop cycle (i.e., harvest stage), the best predictions of crop N content and NNIi were with NBIi and FLAVi, followed by SFRi.

These results show that it is possible to predict crop N status in sweet pepper, throughout the crop, using fluorescence indices. As NNI was accurately estimated by fluorescence indices, and NNI is a good indicator for diagnosis of crop N nutrition (Lemaire et al., 2008), values of fluorescence indices can be used to determine if the amounts of N being applied throughout the crop cycle are optimal or not. Since optimal crop N status occurs when NNI values are equal to 1, and deficient crop N status when NNI is < 1, on-going monitoring with fluorescence indices enables regular assessment of the adequacy of N fertilization. Knowing sweet pepper N status will contribute to appreciably improved crop N management, since vegetable growers commonly apply excessive amounts of N fertilizer as a risk avoidance strategy because they have no quantitative information of actual crop N status (Thompson et al., 2007, 2017b). This issue is more relevant in zones with a high density of sweet pepper crops, such as southeast Spain, where this crop is one of the most important with more than 8000 ha cultivated every year (Valera et al., 2017).

With all the fluorescence indices considered in this work, the SFR under both excitations was able to predict relative yield in the earlier phenological stages. This is important for growers because this early information on potential yield can guide N management at early phenological stages.

In general, in this work, the performance of SFR and NBI under red or green excitation were very similar in terms of R2 values for the relationships of these indices with crop variables, as was reported for maize (Quemada et al., 2014) and cucumber (Padilla et al., 2016). The SFR under both excitations are related to the leaf chlorophyll concentration (Tremblay et al., 2012). The difference between red and green light is that green light penetrates more in the leaf tissues (Terashima et al., 2009) giving slightly different values of SFR because of different reabsorption of fluorescence by chlorophyll. Since the SFR under red or green excitation was very similar in the current work, it shows that the structure and chlorophyll concentration of sweet pepper does not affect much the SFR indices. The similitude between NBI under red or green excitation is explained because they are derived from SFR-R and SFR-G divided by the same parameter (FLAV) so that their difference in the estimate performance is expected to be similar to SFR indices. Therefore, the use of either red or green excitation with these indices is equally effective for monitoring and predicting sweet pepper N status.

In terms of crop N management, the high cost of flavonols meters (3,000–14,000€ in Europe, depending on the model) may make them unattractive to farmers. It is necessary to determine if the use of these sensors could make it attractive given the economic benefit of the reduction in the use of nitrogen fertilizers when using these tools. An alternative for the use of flavonols meters in farmers’ best N management practices is through the service of technical advisors or consulting companies that provide flavonols meters measurements on farms.

Overall, the results of this work show the potential of fluorescence indices to predict crop N content, crop N uptake, NNIi, and relative yield. It has been shown that selecting an optimized index for each phenological stage could be a potentially useful approach to derive different crop variables related to crop N content of sweet pepper. The use of these indices would be useful for the estimation of N status in the field, for guiding farmers in an accurate application of N fertilisation throughout the crop cycle, and for increasing N use efficiency. The SFRi was most effective for prediction of crop N content and NNIi at earlier stages of the crop (i.e., vegetative and flowering phenological stages), with either red (SFR-Ri) or green (SFR-Gi) excitation. In the harvest stage, the best capacity for prediction was with NBIi, both under red (NBI-Ri) or green (NBI-Gi) excitation, and with FLAVi. However, the high economical cost of fluorescence sensors can be an issue that hampers adoption by farmers.

Conclusions

  • Selecting the best performing index in each phenological stage could improve the N use efficiency in sweet pepper crops.

  • The SFRi indices showed the best performance at earlier phenological stage to predict crop N status.

  • The NBIi indices and the FLAVi index showed the best performance at harvest stage to predict crop N status.

  • The SFRi indices could be useful to predict relative yield of sweet pepper at early growth stages.