Abstract
Because the leaf area index (LAI) is an essential parameter for understanding the structure and growth status of plant canopies, nondestructive and continuous estimation methods have been required. Recently, an LAI estimation method using the ratio of near-infrared radiation (NIR; 700–1000 nm) to photosynthetically active radiation (PAR; 400–700 nm) (NIRin/PARin) transmitted through a canopy has been proposed. However, because previous studies on this NIRin/PARin-based LAI estimation method are limited to tall plants (e.g., forest and rice canopies), in this study, we applied this method to a short canopy (i.e., spinach) and investigated its validity. NIRin/PARin and three other traditional indices for indirect LAI estimation—relative PPF density (rPPFD), normalized difference vegetation index (NDVI), and simple ratio (SR)—were measured in 25 canopies with different LAI. NIRin/PARin showed better estimation sensitivity (R2 = 0.88) to the observed LAI than the other three indices, particularly when LAI was greater than 3 m2·m−2. In addition, the LAI estimated from NIRin/PARin measured at 10-min intervals in the entire growth period could capture an increasing trend in the measured LAI throughout the entire growth stage (mean absolute error = 0.87 m2·m−2). Errors in long-term LAI estimations may be caused by the sensor location and insufficient data due to unsuitable weather conditions for measuring NIRin/PARin. The current study demonstrates the merits and limitations of the NIRin/PARin-based LAI estimation method applied to low height canopies, thereby contributing to its practical use in horticultural crops.
Leaf area index (LAI) is a dimensionless variable that quantifies the amount of leaves in a canopy and is defined as half of the all-sided leaf area per unit ground area (Chen and Black 1992). LAI is an essential parameter that strongly affects the photosynthesis, respiration, and evapotranspiration of a canopy (Allen et al. 1998; Nomura et al. 2021; Sun et al. 2014). Therefore, accurate measurement of LAI is necessary for understanding physiological and ecological information about plants at a canopy scale.
Methods of measuring LAI can be categorized into direct and indirect methods (Yano et al. 2022). Direct methods, such as destructive sampling, are the most accurate; however, they are extremely labor-intensive, time-consuming (Daughtry 1990), and inapplicable to continuous observation. Thus, to overcome the drawbacks of direct methods, various indirect methods have been proposed (Gower et al. 1999; Jonckheere et al. 2004; Yan et al. 2019). One of the most typical indirect methods is using the canopy transmissivity of photosynthetic photon flux density (PPFD). This method is based on exponential radiation extinction in plant canopies (i.e., Lambert–Beer law) and can estimate LAI from the relative PPFD (rPPFD; the ratio of PPFD below a canopy to PPFD above the canopy) (Chason et al. 1991; Monsi and Saeki 2005). One of the main limitations of this method is that it requires PPFD measurement at two points (above and below the canopy) under near-identical light conditions. Generally, this limitation can be overcome via the calibration between two sensors; however, this calibration can be laborious at a location where a spatial difference of light environment fluctuates temporally owing to external factors (e.g., in greenhouse). Additionally, a previous study reported that the sensitivity of this estimation method is lower for a higher LAI canopy (Richardson et al. 2009). Vegetation indices such as the normalized difference of vegetation index (NDVI) and simple ratio (SR) are also used to estimate LAI (Hashimoto et al. 2019). These vegetation indices are usually calculated using the reflectance of red light and near-infrared radiation (NIR; 700–1000 nm) from the surface of canopies. The major problem with these methods is that they require measuring sample light from canopies and reference light from a reflective white standard plate for calibration. They are also affected by the light reflected from the background when plants are small (Nagai et al. 2010). Moreover, image-based LAI estimation methods have been successfully applied (Nomura et al. 2020; Sandmann et al. 2013); however, there are still some problems, such as difficulty in postprocessing after image acquisition (e.g., distinguishing leaf and background area) (Yan et al. 2019).
As a possible solution to the problems of the aforementioned conventional estimation methods, Kume et al. (2011) proposed an LAI estimation method using light with wavelengths between 400 and 1000 nm transmitted in a canopy. They found that the ratio of NIR to PAR of the transmitted light (NIRin/PARin) in a deciduous broad-leaved forest canopy had a logarithmic relationship with LAI (the coefficient of determination R2 = 0.96). This method has the advantages of not requiring two-point measurement and not considering the impact of the light reflected from the ground surface (e.g., soil). Recently, an LAI measuring sensor using this method has been developed for commercial use (MIJ-15LAI, Environmental Measurement Japan, Fukuoka, Japan). However, previous studies on the NIRin/PARin-based LAI estimation method are limited to tall plants (e.g., forest and rice canopy); to the best of our knowledge, this method has not been applied to short canopies. Moreover, Fukuda et al. (2021) noted that the relationship between NIRin/PARin and LAI could differ depending on the canopy structure; thus, further investigations are required for other types of plant canopies. In horticultural crop production, where precise management is essential for high-quality and labor-saving production, accurate leaf area information is required (Kitano et al. 2022). Notably, the nondestructive and continuous estimation of LAI is of practical value for facilitating improved crop management and productivity because LAI can increase rapidly within a few days in vegetable canopies.
In this study, we applied the NIRin/PARin-based LAI estimation method to the canopy of spinach, a short horticultural crop. In particular, we compare the relationship between NIRin/PARin and other conventional light indices—rPPFD, NDVI, and SR—and the LAI measured by direct method; moreover, and we estimate LAI continuously for an entire growth period using NIRin/PARin.
Materials and Methods
Plant materials and cultivation conditions.
Spinach plants (Spinacia oleracea L.) were used as plant materials and two cultivations were performed in winter and early summer (hereinafter, Cult. 1 and Cult. 2, respectively), in which ‘Wase Crone’ was used for Cult. 1 and ‘Summer Top’ for Cult. 2. In both cultivations, seeds were sown on urethane cubes (25 × 25 × 30 mm3) at a rate of three seeds per cube. Nursery plants were grown with a sufficient water supply for ∼10 d in a phytotron room (air temperature = 20 °C and relative humidity = 70%) at Environmental Control Center for Experimental Biology, Kyushu University (33°35′37″N, 130°12′49E). After the germination, the well-grown seedlings (plant height ∼20–30 mm) were transplanted into hydroponic panels (450 × 750 mm2) on a nutrient film technique (NFT) cultivation bed (15 × 1.27 m2) in a greenhouse located at Ito Plant Experimental Fields and Facilities, Faculty of Agriculture. In Cult. 1, plants were transplanted into 25 panels and grown in the NFT system for 15–51 d from 25 Nov 2020 to 1 Feb 2021; in Cult. 2, plants were transplanted into a total of 10 panels and grown for 10 to 32 d from 29 Apr to 31 May 2021. Each panel had 30 urethane cubes (planting density of 257 plants/m2). In this study, we defined one canopy as one panel. In the NFT system, plants were grown hydroponically with a nutrient solution (Otsuka AgriTechno Co. Ltd., Japan) at an EC of 2.0 dS·m−1 and pH of 6.0. The nutrient solution contained 3.9 mmol·L−1 Ca2+, 17. mmol·L−1 NO3, 8.4 mmol·L−1 K+, 1.6 mmol·L−1 SO4 2−, 1.1 mmol·L−1 PO43−, and 1.5 mmol·L−1 Mg2+. The greenhouse was equipped with environmental-control systems: roof and side-wall window ventilation occurred when the air temperature was higher than 22 °C, and heating with the heater (HK2027TEV, NEPON Inc., Tokyo, Japan) occurred when the air temperature was lower than 8 °C (Table 1).
Environment condition at each cultivation (Cultivations 1 and 2). Tmax, Tmin, Tavg represent the maximum, minimum, and average temperatures, respectively. PPFDout_avg represents the diurnal averaged photosynthetic photon flux density inside the greenhouse.
Model development for estimating LAI using different light indices.
To construct regression models for LAI estimation, measurements of NIRin/PARin, NDVI, SR, and rPPFD were performed at 10:00 AM–12:00 PM on the following dates during Cult. 1 (a total of 25 times): 10, 15, 18, 19, 22, 25, and 28 Dec 2020; 2, 7, 9, 11, 15, 17, 23, 26, 27, 29, and 31 Jan 2021; 1 Feb 2021. During measurement, to prevent the uneven spatial incidence of light into the canopy due to the greenhouse structure, diffuse illumination was created by closing a shading curtain in the greenhouse. Each measurement was performed three times at six points (a total of 18 times) per canopy, and we obtained the average as the representative value.
The spectral intensity (SI) of light for calculating NIRin/PARin, NDVI, and SR was measured using a multispectrometer (BIM-6002A-10-S03L00F06G02; Yixi Intelligent Technology, Hangzhou, China) with an optical fiber (SIM-6102-0615; Yixi Intelligent Technology, Hangzhou, China). This spectrometer can detect the light of 300 to 1100 nm with a spectral resolution of 0.35 to 1.0 nm. The field of view (FOV) of the spectrometer is 180° and 26° with and without a cosine corrector (SIM-6320-01, Yixi Intelligent Technology), respectively. Cosine correctors are optical diffusers that are coupled to fibers and spectrometers to collect signals from 180 °FOV. SI of light transmitted in a canopy (SIin) was measured at 6 cm above the cultivation panel inside the canopy using a spectrometer with a cosine corrector. SI of reflected light from a canopy (SIref) and SI of sample light from the reflective white standard plate (SIsam) were measured at 15 cm above the canopy by the spectrometer without a cosine corrector. This distance (15 cm) was determined based on the FOV and area of the reflective white standard plate. This plate was coated with a white reflectance coating (Pre-Mix White Reflectance Coating; Edmund Optics, Barrington, NJ, USA), which has high reflectance (>97%) in the range of 400 to 1000 nm. Because spectrometers will have a spectrally varying response of the detector, after the measurement of SI, we calibrated the raw SI data using tungsten halogen calibration lamp (SL1-CAL, StellarNet, Inc., Tampa, FL, USA).
PPFDs outside and inside the canopy (PPFDin and PPFDout; µmol·m−2·s−1) for calculating rPPFD were measured using a quantum sensor (PAR-02D; PREDE, Tokyo, Japan). PPFDin was measured at 3 cm above the cultivation panel inside the canopy. PPFDout was measured above the canopy. The data of the measurements were recorded by a datalogger (GL220; Graphtec Corporation, Yokohama, Japan). Each measurement required less than 20 s.
Validation test of estimation model of LAI by NIRin/PARin.
Continuous long-term measurement of NIRin/PARin.
At one panel in Cult. 2, the spectrometer was located at 6 cm above the cultivation panel, and SIin was measured continuously at 10-min intervals from 29 Apr to 30 May 2021(harvesting day). From this measurement, NIRin/PARin was obtained using Eq. [1] in a continuous manner, except for 10 d when we used the spectrometer for the validation measurements, as explained earlier.
Results
Spectral distribution of transmitted light at different growth stages.
Figure 1 shows the relationship between observed LAI (LAIobs) and SIin distribution in Cult. 1. SIin values were shown as the relative values to the maximum values, and the maximum values in all measurements existed in the NIR range. When the height of the sensing part of the spectrometer was higher than that of the canopy (LAIobs < 0.5 m2·m−2), the peak waveband was in the PAR region, and the spectral distribution showed a similar feature to the solar spectral distribution. However, because this experiment was performed in a greenhouse with a complicated light environment, the spectral distribution was not completely identical to the solar spectral distribution. When the height of the canopy became higher than that of the sensing part of the spectrometer (LAIobs > 0.5 m2·m−2), the light in the PAR region was largely absorbed, whereas the light in the NIR region was hardly absorbed by the canopy, resulting in a smaller relative intensity of light in the PAR region. When LAIobs > 5 m2·m−2, the light in the PAR region was mostly absorbed, and its relative intensity approached 0.
Relationship between observed leaf area index (LAIobs) and spectral distribution of transmitted light in the canopy at cultivation (Cult.) 1. The relative intensity (vertical axis) was calculated by setting values at the maximum values to 1.0.
Citation: HortScience 58, 1; 10.21273/HORTSCI16761-22
Relationship between LAI and light indices.
Figure 2 shows the relationship between LAIobs and NIRin/PARin, rPPFD, NDVI, and SR. In Cult. 1, NIRin/PARin was strongly correlated with LAIobs, either linearly (R2 = 0.91) or positive logarithmically (R2 = 0.88) (Fig. 2A). rPPFD was negative logarithmically correlated with LAIobs (R2 = 0.86), and when LAIobs >4 m2·m−2, rPPFD showed small change and approached zero (Fig. 2B). The relationship between LAIobs and NDVI was positive exponential (R2 = 0.78), and for LAIobs >2 m2·m−2, the values of NDVI were stagnant around 0.9 (Fig. 2C). SR was linearly correlated with LAIobs (R2 = 0.84); the plot variability increases when LAIobs >4 m2·m−2 (Fig. 2D).
Relationship between observed leaf area index (LAIobs) and light indices. (A) Near-infrared radiation (700–1000 nm) to photosynthetically active radiation (PAR; 400–700 nm) ratio (NIRin/PARin) (linear equation [dotted line]:
Citation: HortScience 58, 1; 10.21273/HORTSCI16761-22
Validation of estimation model of LAI by NIRin/PARin.
(A) Relationship between observed leaf area index (LAIobs) and near-infrared radiation (700–1000 nm) to photosynthetically active radiation (400–700 nm) ratio (NIRin/PARin) in Cult. 2. Two regression equations were constructed based on the data measured in Cult. 1. Open circle denotes outliers. The relationship between estimated LAI (LAIest) and LAIobs estimated from (B) linear and (C) logarithmic equations in Cult. 2. The 95% prediction interval was indicated by the gray hatch.
Citation: HortScience 58, 1; 10.21273/HORTSCI16761-22
Figure 3B and C show the relationship between LAI estimated using the two regression equations (LAIest) and LAIobs in Cult. 2. LAIest estimated using the logarithmic equation was strongly correlated with LAIobs (RMSE = 0.78 m2·m−2) (Fig. 3C) in a wide LAIobs range between 0.1 and 9.5 m2·m−2 compared with LAIest estimated using the linear equation (RMSE = 3.57 m2·m−2) (Fig. 3B). In addition, the 1:1 line in Fig. 3C was included in the 95% prediction interval.
Daily changes in NIRin/PARin and estimated LAI.
Figure 4 shows the daily changes in NIRin/PARin and LAIest, wherein those calculated from all data measured between sunrise and sunset are shown in Fig. 4A and C, and values extracted based on the thresholds are shown in Fig. 4B and D. CVavg was calculated by averaging the CV for each day over the entire growth period. In Fig. 4A, the NIRin/PARin values calculated from all data increased gradually in the range of 0 to 10 m2·m−2 at the early growth stage (DAT = 10–28 d) and increased rapidly at the late growth stage (DAT = 28–32 d). In Fig. 4C, the LAIest values calculated from all data gradually increased within 0 to 10 m2·m−2, but the CVavg was large. In Fig. 4B and D, the trend of the NIRin/PARin and LAIest corrected by the thresholds is similar to that in Fig. 4A and C, but the CVavg decrease considerably compared with Fig. 4A and C.
Daily changes in near-infrared radiation (700–1000 nm) to photosynthetically active radiation (400–700 nm) ratio (NIRin/PARin) and estimated leaf area index (LAIest) (A and C: all values; B and D: valid values) for 10 to 32 d from transplantation. CVavg (%) is the average of the CV for each day. For the box plots, the horizontal lines inside the box denote the medians, the open circles denote the means of the day, and closed circles denote outliers. The outliers were defined as values for which the distance from the interquartile range exceeded 1.5 times the interquartile range.
Citation: HortScience 58, 1; 10.21273/HORTSCI16761-22
Figure 5A shows daily changes in LAIobs and the average values of LAIest. When DAT was 20 and 27 d, the LAIest was far away from LAIobs; however, the increasing trend of LAIest was similar to that of LAIobs in the range of 0.1–9.5 m2·m−2. Additionally, LAIest had a strong linear relationship with LAIobs (mean absolute error = 0.87 m2·m−2) in the LAIobs range from 0.1 to 9.5 m2·m−2, and the 1:1 line was included in the 95% prediction interval (Fig. 5B).
(A) Daily changes in observed leaf area index (LAIobs) and the average values of estimated leaf area index (LAIest) estimated from valid near-infrared radiation (700–1000 nm) to photosynthetically active radiation (400–700 nm) ratio (NIRin/PARin) for 10 to 32 d after transplanting. (B) The relationship between LAIest and LAIobs in the continuous measurements (Cultivation 2). The 95% prediction interval was indicated by the gray hatch.
Citation: HortScience 58, 1; 10.21273/HORTSCI16761-22
Discussion
The usefulness of LAI estimation using NIRin/PARin.
We compared LAI estimation models that were based on NIRin/PARin, rPPFD, NDVI, and SR, and the following was found in terms of estimation sensitivity and measurement method. The values of R2 in the entire growth stage in the four indices were 0.78 to 0.91 (Fig. 2) and did not differ significantly except for NDVI. In the late growth stage, the values of NDVI stagnated, and the estimation method using NDVI had low sensitivity when LAI >2 m2·m−2. Previous studies have also reported that the values of NDVI were saturated when LAI >2 m2·m−2 (e.g., Haboudane et al. 2004). Conversely, the values of SR still increased in the late growth stage, although SR-based LAI estimations fluctuated widely when LAI >4 m2·m−2. The values of NIRin/PARin showed high sensitivity in LAI and less fluctuation even in the late growth stage. In fact, for the late growth stage (LAI ≥3 m2·m−2), the RMSE of estimated LAI was lower using NIRin/PARin than using the other three indices (Supplemental Table 1), indicating that the estimation method of LAI using NIRin/PARin has good estimation accuracy not only in the early growth stage but also in the entire growth stage, including late growth stage, compared with the conventional methods. Meanwhile, because rPPFD and SR were superior in estimation accuracy in the early growth stage (LAI ≤3 m2·m−2) (Supplemental Table 1), it may be effective to use different estimation methods at each growth stage to achieve more accurate estimation in the entire growth stage.
Calculating rPPFD requires the calibration between two sensors to measure PPFD at two points under the same light conditions, which is difficult in a greenhouse because a spatial difference of the light environment fluctuates temporally due to the film and structural materials (Llorenc et al. 2016; Matsuda et al. 2020). In addition, if sensors are installed for LAI measurement over a long period, periodic sensor inspections are necessary because of the possibility of dirt on sensors. In the case of double point measurements, because a sensor that measures reference light (i.e., a light above a canopy) is generally installed at a high height, periodic inspections require labor. However, for measuring NIRin/PARin, we do not have to pay attention to the calibration, and the periodic inspection can also be less labor-intensive than two sensors.
In the measurement of canopy reflectance for calculating NDVI and SR, it is ideal to capture a broad area of a canopy using a sensor with a broad FOV (Kume et al. 2011). However, a sensor with a narrow FOV is used to prevent radiation from outside canopy (26° in this study). Further, the estimation of LAI using reflectance spectra is less versatile because the measurement of reflected light is strongly affected by the reflection from the background. Meanwhile, NIRin/PARin measurement can be obtained as a representative value because it is unaffected by radiation from outside a canopy, even with a sensor having a broad FOV (180° in this study). In addition, the sensing part of the sensor is located in the upward direction when measuring the transmitted light, so the NIRin/PARin-based LAI estimation method is unaffected by the reflection from the floor and can be used in various cultivation situations.
Although not measured in this study, image-based LAI estimation methods (e.g., using digital directional photography and hemispherical photography) are also attracting attention as rapid and accurate LAI estimation methods in recent years (Jonckheere et al. 2004; Yan et al. 2019). Photography-based methods, such as the NIRin/PARin method, have a major advantage that they do not require any reference measurements (e.g., a measurement of the light above a canopy in the rPPFD method). Furthermore, according to the technological development of image sensors, improved camera devices are becoming affordable. Recently, even a smartphone can be used to estimate LAI (e.g., Confalonieri et al. 2013). However, there are some problems with image-based LAI estimation methods. The main disadvantage is the difficulty in distinguishing leaf and background area to determine gap fraction (the proportion of background in an image). For example, in the case of using a downward camera, this operation can be difficult because of the variety of possible backgrounds (e.g., fallen leaves) (Demarez et al. 2008). Additionally, if the amount of leaves is too large or the distance between canopy and camera is too large, the pixels of the gap will be too small to distinguish leaf and background area (Jonckheere et al. 2004). Moreover, in the case of using directional camera, the optimal setting angle of camera for photography is not yet known. Although it is generally believed that camera angle = 57.3° is better, it has reported that when LAI is too large, the gap will be too small, and the estimation of LAI is difficult (Liu et al. 2013). Therefore, photography-based methods are not yet ideal indirect estimation methods, and it is possible that the NIRin/PARin-based method is more useful for LAI estimation than photography-based methods in some situations.
These results suggest that the NIRin/PARin-based LAI estimation method may be more useful than the conventional indirect LAI estimation methods in terms of the estimation accuracy and measurement method.
Nevertheless, there are still some problems with the NIRin/PARin method. In this study, PARin was decreased because we measured NIRin/PARin in the light environment with closed shade curtain, resulting in larger values of NIRin/PARin than those reported in previous studies (Fukuda et al. 2021; Kume et al. 2011). Therefore, when measuring transmitted light in a greenhouse similar to this study, considering the influence of the greenhouse structure on illumination conditions is necessary. Additionally, the NIRin/PARin method requires a spectrometer that is more expensive or similar than a PAR sensor and difficult to operate. Therefore, the application of the NIRin/PARin method with a reasonable sensor and the manualization of the NIRin/PARin method are desired in the future.
The optimal regression equation between LAI and NIRin/PARin.
The relationship between LAI and NIRin/PARin in the spinach canopy was fitted to the logarithmic equation better than the linear equation (Fig. 3). Interestingly, despite the completely different canopy scale, the form of the regression equation between LAI and NIRin/PARin for this study (Eq. [12]) was similar to that of Kume et al. (2011), who focused on deciduous broad-leaved forests (LAI = 2.80ln (NIRin/PARin) + 0.69). Meanwhile, Fukuda et al. (2021) observed that the regression equation between LAI and NIRin/PARin was linear for a rice canopy.
The form and coefficient of the regression equation between LAI and NIRin/PARin should differ among plant types for the following reasons. First, the form of regression equation could be influenced by canopy structure. Based on the LAD, plants can be distinguished into various types such as erectophile, planophile, plagiophile, spherical, and extremophile (De Wit 1965). However, we distinguished plants into two types for simplification: the broad and grass-leaf types, following Monsi and Saeki (2005). The broad-leaf type (i.e., roughly corresponding to planophile), such as deciduous broad-leaved forests and spinach canopy, grows relatively horizontal. As LAI increases for such canopies, the light should be intercepted by those horizontal leaf layers and approach zero. Consequently, NIRin/PARin diverges infinitely; therefore, the regression between LAI and NIRin/PARin should be expressed as a logarithmic equation. Meanwhile, for grass-leaf type (i.e., roughly corresponding to erectophile), such as rice canopy, the narrow leaves grow relatively vertical, allowing incident light to penetrate the canopy, even when LAI is large. Consequently, a certain amount of PARin remains when LAI is large, and NIRin/PARin does not diverge infinitely. Therefore, the regression between LAI and NIRin/PARin becomes a linear equation.
Second, the coefficient of regression equation could be influenced by the spectral characteristics of leaves comprising the canopy. The amount of PAR and NIR absorbed by the canopy should change depending on the spectral characteristics of the leaves, resulting in a change in the NIRin/PARin, which may affect the regression equation coefficient. Fukuda et al. (2021) noted the change in the regression equation coefficient due to the spectral characteristics of leaves. In this study, two different species of spinach were grown, but there was no change in the regression equation coefficient depending on the species. This may be because there was little change in the spectral characteristics of the two species (data not shown). However, because the leaf spectral properties of different plant types differ considerably, the spectral characteristics of leaves should be considered when constructing regression equation. However, it may be difficult to consider leaf spectral characteristics in the model of NIRin/PARin method because leaf spectral characteristics are changed by various factors such as water status, nutritional status, and species.
These results suggest that the form and coefficient of the regression equation between LAI and NIRin/PARin is influenced by the canopy structure and spectral characteristics of the leaves comprising the canopy. In the future, it may be possible to construct a more robust model by considering the information of canopy structure. Moreover, statistically analyzing the factors that affect the regression equation would enable us to apply the NIRin/PARin-based LAI estimation method to various plant types.
Considerations and limitation in continuous NIRin/PARin measurements.
We extracted valid NIRin/PARin-based on the threshold values, following Fukuda et al. (2021), for estimating LAI accurately and stably. Consequently, we estimated continuous LAI more smoothly than LAI estimated by all NIRin/PARin (Fig. 4). However, even after extraction, some LAI values were found to deviate from the actual measured values (Fig. 5). This error may be due to some problems and limitations in the measurements of NIRin/PARin as follows.
The first is that the transmitted light spectra in the canopy vary significantly spatially and temporally. Particularly, in the case of the medium growth stage at which leaves are sparsely distributed, the light environment in the canopy changed significantly by leaf movement during the day, resulting in NIRin/PARin dramatically varying in space. In this study, it was difficult to obtain representative values of NIRin/PARin of the canopy because we measured it at only one point for the canopy. To solve this problem, it is desirable to set multiple sensors in the canopy, which would be cost-intensive. Another solution to the problem is increasing the amount of data in a day and using its daily mean value. Fukuda et al. (2021) successfully estimated continuous LAI, sampling the NIRin/PARin at 1-min intervals throughout the day. Meanwhile, we sampled the NIRin/PARin at 10-min intervals and succeeded in estimating the increasing trend of LAI. Therefore, the 10-min sampling interval is adequate; however, it is possible that we may obtain more stable values of estimated LAI by increasing the number of samples.
Second, the height of the spinach canopy was low (∼35 cm), and the leaves might have been in close contact with the spectra sensor. This may have caused a rapid decrease in the amount of light received by the sensor, inducing measurement errors. In fact, we confirmed that the leaves were in full contact with the sensor during the day when DAT was 27 d, resulting in unnaturally large values of NIRin/PARin and LAIest (Fig. 4 and Supplemental Fig. 3). This problem would not be severe in canopies with large internal space, such as forests (Tanioka et al. 2020), or canopies with a certain height of grass, such as paddy rice (Fukuda et al. 2021), but it can be a problem for a low height canopy, such as spinach in this study. To apply this method to the canopy having a low height, the sensor should be placed lower and kept at a larger distance from the leaves. In this study, the sensor was placed at a height of 6 cm from the bottom level of the canopy, but it is ideal to place it as close to the ground surface as possible.
Third, the amount of valid NIRin/PARin data that can be extracted depends on the weather condition on the measurement day. In this study, valid NIRin/PARin were extracted following Fukuda et al. (2021); however, many NIRin/PARin were deleted in the period when the light environment was unsuitable for NIRin/PARin measurement (see Results). Consequently, the amount of extracted data was small; therefore, a reasonably valid NIRin/PARin could not be obtained. The solution to this problem, as well as the solution to the first problem of spatial and temporal difference of transmitted light spectra in the canopy, is increasing the amount of data by shortening the data measurement interval. Another solution is to improve the method of extracting valid NIRin/PARin. In this study, we set a threshold value for PPFDin as the minimum light intensity at which reasonable NIRin/PARin can be detected, but the threshold value of PPFDin was too low, and a large quantity of NIRin/PARin data were deleted. Because the spectrometer used to measure NIRin/PARin can adjust the integration time that light is received, there is a possibility that a large amount of valid NIRin/PARin data can be extracted by increasing the integration time, even under extremely low PPFDin conditions.
These results suggest that there are several problems in the continuous measurements of NIRin/PARin, and further studies are necessary to consider the installation method of the sensor and the suitable extraction method of valid NIRin/PARin.
Conclusions
In conclusion, we succeeded in applying an estimation method of LAI using transmitted light spectra (NIRin/PARin), as suggested by Kume et al. (2011) for a spinach canopy with a low height. Compared with conventional estimation methods, the new NIRin/PARin-based method could estimate LAI, even in the late growth stage (LAI ≥3 m2·m−2), suggesting that it is useful as a new estimation method (Fig. 2). In addition, we attempted continuous estimation of LAI for the entire growth period. The results show that an increasing trend of LAI could be monitored using NIRin/PARin, whereas some plots deviated significantly from the measured values (Fig. 5). These measurement errors could be attributed to the installation strategy of the sensor and the insufficient amount of data, which could be solved by installing the sensor considering the canopy structure and shortening the measurement interval. It is essential to apply the new method to various plant species and cultivation conditions to determine its validity for practical use in horticultural crops in the future.
In recent years, several studies have reported various ecophysiological information about plants that can be evaluated using spectral characteristics (reflectance and transmittance) of leaves and canopies (Singh et al. 2022). However, in this field of study, most work has been conducted using not transmitted but reflected light spectra because reflected light can be measured from the sky using satellites or drones for large-scale canopies. Meanwhile, our results suggest that for small-scale canopies, such as horticultural crops, transmitted light spectra may be useful compared with reflected light in terms of measurement technique and estimation accuracy. Further, transmitted light, which penetrates an entire canopy layer, may provide more information than reflected light, which is mainly reflected on the upper layer. In this study, the spectral distribution of transmitted light in a plant canopy was revealed as basic information (Fig. 1). However, because few studies have clarified the spectral characteristics of transmitted light in the canopy (e.g., Lao et al. 2014), we should collect more basic spectral characteristics data of transmitted light in the future. Further, the transmitted light spectra of a plant canopy may be applicable to analyzing various plant ecophysiological information in addition to LAI in future investigations.
References Cited
Allen, R.G., Pereira, L.S., Raes, D. & Smith, M. 1998 Crop evapotranspiration—Guidelines for computing crop water requirements FAO Irrigation and Drainage Paper 56 Food and Agriculture Organization of the United Nations Rome, Italy
Chason, J.W., Baldocchi, D.D. & Huston, M.A. 1991 A comparison of direct and indirect methods for estimating forest canopy leaf area Agric. Meteorol. 57 107 128 https://doi.org/10.1016/0168-1923(91)90081-Z
Chen, J.M. & Black, T.A. 1992 Defining leaf area index for non-flat leaves Plant Cell Environ. 15 421 429 https://doi.org/10.1111/j.1365-3040.1992.tb00992.x
Confalonieri, R., Foi, M., Casa, R., Aquaro, S., Tona, E., Peterle, M., Boldini, A., Carli, G.D., Ferrari, A., Finotto, G., Guarneri, T., Manzoni, V., Movedi, E., Nisoli, A., Paleari, L., Radici, I., Suardi, M., Veronesi, D., Bregaglio, S., Cappelli, G., Chiodini, M.E., Dominoni, P., Francone, C., Frasso, N., Stella, T. & Acutis, M. 2013 Development of an app for estimating leaf area index using a smartphone. Trueness and precision determination and comparison with other indirect methods Comput. Electron. Agric. 96 67 74 https://doi.org/10.1016/j.compag.2013.04.019
Daughtry, C.S.T. 1990 Direct measurements canopy structure Remote Sens. Rev. 5 45 60 https://doi.org/10.1080/02757259009532121
De Wit, C.T. 1965 Photosynthesis of Leaf Canopies Agricultural Research Reports 663 Centre for Agricultural Publications and Documentation Wageningen, The Netherlands
Demarez, V., Duthoit, S., Baret, F., Weiss, M. & Dedieu, G. 2008 Estimation of leaf area and clumping indexes of crops with hemispherical photographs Agric. Meteorol. 148 4 644 655 https://doi.org/10.1016/j.agrformet.2007.11.015
Fukuda, S., Koba, K., Okamura, M., Watanabe, Y., Hosoi, J., Nakagomi, K., Maeda, H., Kondo, M. & Sugiura, D. 2021 Novel technique for non-destructive LAI estimation by continuous measurement of NIR and PAR in rice canopy Field Crops Res. 263 108070 https://doi.org/10.1016/j.fcr.2021.108070
Gower, S.T., Kucharik, C.J. & Norman, J.M. 1999 Direct and indirect estimation of leaf area index, fAPAR, and net primary production of terrestrial ecosystems Remote Sens. Environ. 70 29 51 https://doi.org/10.1016/S0034-4257(99)00056-5
Haboudane, D., Miller, J.R., Pattey, E., Zarco-Tejada, P.J. & Strachan, I.B. 2004 Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: Modeling and validation in the context of precision agriculture Remote Sens. Environ. 90 337 352 https://doi.org/10.1016/j.rse.2003.12.013
Hashimoto, N., Saito, Y., Maki, M. & Homma, K. 2019 Simulation of reflectance and vegetation indices for unmanned aerial vehicle (UAV) monitoring of paddy fields Remote Sens. 11 1 13 https://doi.org/10.3390/rs11182119
Jonckheere, I., Fleck, S., Nackaerts, K., Muys, B., Coppin, P., Weiss, M. & Baret, F. 2004 Review of methods for in situ leaf area index determination Part I. Theories, sensors and hemispherical photography Agric. Meteorol. 121 19 35 https://doi.org/10.1016/j.agrformet.2003.08.027
Jones, H.G. 1992 Plant and microclimate 2nd ed. Cambridge University Press Cambridge, UK
Kitano, M., Nomura, K., Yamazaki, T., Iwao, T., Saitou, M., Mori, M., Yasutake, D., Kaneko, T., Ukeda, H., Ishizuka, S., Fujiwara, T. & Okabayashi, T. 2022 Internet of Plants (IoP) empowers bottom-up innovations in greenhouse horticulture Environ. Control Biol. 60 3 12 https://doi.org/10.2525/ecb.60.3
Kume, A., Nasahara, K., Nagai, S. & Muraoka, H. 2011 The ratio of transmitted near-infrared radiation to photosynthetically active radiation (PAR) increases in proportion to the adsorbed PAR in the canopy J. Plant Res. 124 99 106 https://link.springer.com/article/10.1007/s10265-010-0346-1
Lao, C.L., Xu, Z.L., Guo, Y., Jin, Y. & Yang, Y.H. 2014 Characterization of spatial variability of spectral irradiance in tobacco canopy Appl. Ecol. Environ. Res. 12 727 740
Liu, J., Pattey, E. & Admiral, S. 2013 Assessment of in situ crop LAI measurement using unidirectional view digital photography Agric. Meteorol. 169 0 25 34 https://doi.org/10.1016/j.agrformet.2012.10.009
Llorenc, C.B., Christian, F., Nicolas, B., Claude, W., Benoit, S. & Francois, T. 2016 High-throughput estimation of incident light, light interception and radiation-use efficiency of thousands of plants in a phenotyping platform New Phytol. 212 269 281 https://doi.org/10.1111/nph.14027
Matsuda, S., Yoshikoshi, H., Suzuki, T., Ohta, Y., Chiba, A., Arima, H., Kumagai, H., Yasutake, D. & Kitano, M. 2020 Calculation of the irradiance of solar radiation in a greenhouse with a complex structure using a diagram for sky view factor Nogyo Kisho 76 44 52 https://doi.org/10.2480/agrmet.D-19-00043
Miao, G., Guan, K., Yang, X., Bernacchi, C.J., Berry, J.A., DeLucia, E.H., Wu, J., Moore, C.E., Meacham, K., Cai, Y., Peng, B., Kimm, H. & Masters, M.D. 2018 Sun-Induced Chlorophyll Fluorescence, Photosynthesis, and Light Use Efficiency of a Soybean Field from Seasonally Continuous Measurements J. Geophys. Res. Biogeosci. 123 610 623 https://doi.org/10.1002/2017JG004180
Monsi, M. & Saeki, T. 2005 On the factor light in plant communities and its importance for matter production Ann. Bot. 95 549 567 https://doi.org/10.1093/aob/mci052
Nagai, S., Nasahara, K.N., Muraoka, H., Akiyama, T. & Tsuchida, S. 2010 Field experiments to test the use of the normalized-difference vegetation index for phenology detection Agric. Meteorol. 150 152 160 https://doi.org/10.1016/j.agrformet.2009.09.010
Nomura, K., Takada, A., Kunishige, H., Ozaki, Y., Okayasu, T., Yasutake, D. & Kitano, M. 2020 Long-term and continuous measurement of canopy photosynthesis and growth of spinach Environ. Control Biol. 58 21 29 https://doi.org/10.2525/ecb.58.21
Nomura, K., Yasutake, D., Kaneko, T., Iwao, T., Okayasu, T., Ozaki, Y., Mori, M. & Kitano, M. 2021 Long-term estimation of the canopy photosynthesis of a leafy vegetable based on greenhouse climate conditions and nadir photographs Scientia Hort. 289 110433 https://doi.org/10.1016/j.scienta.2021.110433
Richardson, J.J., Moskal, L.M. & Kim, S.H. 2009 Modeling approaches to estimate effective leaf area index from aerial discrete-return LIDAR Agric. Meteorol. 149 1152 1160 https://doi.org/10.1016/j.agrformet.2009.02.007
Sandmann, M., Graefe, J. & Feller, C. 2013 Optical methods for the non-destructive estimation of leaf area index in kohlrabi and lettuce Scientia Hort. 156 113 120 https://doi.org/10.1016/j.scienta.2013.04.003
Sinclair, T.R. 2006 A reminder of the limitations in using beer’s law to estimate daily radiation interception by vegetation Crop Sci. 46 6 https://doi.org/10.2135/cropsci2006.01.0044
Singh, L., Mutanga, O., Mafongoya, P., Peerbhay, K. & Crous, J. 2022 Hyperspectral remote sensing for foliar nutrient detection in forestry: A near-infrared perspective Remote Sens Appl: Soc Environ. 25 100676 https://doi.org/10.1016/j.rsase.2021.100676
Sun, J., Wu, J., Guan, D., Yao, F., Yuan, F., Wang, A. & Jin, C. 2014 Estimating daytime ecosystem respiration to improve estimates of gross primary production of a temperate forest PLoS One 9 11 https://doi.org/10.1371/journal.pone.0113512
Tanioka, Y., Cai, Y., Ida, H. & Hirota, M. 2020 A spatial relationship between canopy and understory leaf area index in an old-growth cool-temperate deciduous forest Forests 11 1 11 https://doi.org/10.3390/f11101037
Yan, G., Hu, R., Luo, J., Weiss, M., Jiang, H., Mu, X., Xie, D. & Zhang, W. 2019 Review of indirect optical measurements of leaf area index: Recent advances, challenges, and perspectives Agric. Meteorol. 265 390 411 https://doi.org/10.1016/j.agrformet.2018.11.033
Yano, T., Yasutake, D. & Kiyosue, Y. 2022 Characterization of canopy structure for high-yield performance of greenhouse-grown satsuma mandarins using direct measurements and indirect estimations Nogyo Kisho 78 19 30 https://doi.org/10.2480/agrmet.D-21-00018
Root mean square error (RMSE) at each growth stage in four indices. The early growth stage [leaf area index (LAI) ≤3 m2·m−2] has 11 plots, and the late growth stage (LAI ≥3 m2·m−2) has 14 plots.
Leaf area (LAleaf) as a function of leaf dry weight (DWleaf) in (A) Cultivation 1 and (B) Cultivation 2.
Citation: HortScience 58, 1; 10.21273/HORTSCI16761-22
(A) Average PPFDin (Avg_PPFDin) during the measurement of near-infrared radiation (700–1000 nm) to photosynthetically active radiation (400–700 nm) ratio (NIRin/PARin) in Cultivation 2. (B) Diurnal change in PPFDin when days after transplanting = 25. The measurement was performed for 2 h (10:00 AM–12:00 PM). PPFD = photosynthetic photon flux density.
Citation: HortScience 58, 1; 10.21273/HORTSCI16761-22
Diurnal change in near-infrared radiation (700–1000 nm) to photosynthetically active radiation (400–700 nm) ratio (NIRin/PARin) when days after transplanting = 27 (Cultivation 2).
Citation: HortScience 58, 1; 10.21273/HORTSCI16761-22