1. Introduction
Agricultural problems must be addressed internationally in order to solve the crisis of food shortage for humans on a global scale (FAO, 2017; Lobell et al., 2008; OECD and FAO, 2020). Major crop production management and yield prediction technology are important for world safety and national food security (Bryngelsson et al., 2016). The advancement of technologies related to crop production management are expected to increase in importance after COVID-19.
South Korea is currently in the preparation stage for launching a medium-sized satellite for agriculture and forestry, so expectations for its agricultural use are very high (Lee and Lee, 2017). The satellite will be used to understand the current status of crop cultivation and crop production, not only in South Korea but also in neighboring countries. This satellite is expected to play a role in real-time monitoring of information on the agricultural environment in Asia, where production and consumption of major crops are concentrated, along with satellites launched by other countries such as China and Japan. In addition, satellite data can observe a wide area of crop cultivation and have characteristics such as objectivity, repeatability, and ease of processing, particularly for wide-area data (Huete et al., 2002). It is necessary to obtain detailed terrestrial reference data on cultivated crops in order to determine the characteristics of the satellites scheduled to be launched in the future (Zhao et al., 2004).
Soybeans are used as a raw material for oil, food, and feed, and are one of the world’s major crops. Therefore, soybean production and demand have been steadily increasing worldwide since the 1960s (Hartman et al., 2011). Major grain-producing countries, such as the United States, China, and India, have increased production in response to the increase in demand. South Korea is a major importer of soybeans, whereby 87.9% of soybeans are imported from the United States, representing 85% total soybeans exported from the United States, and the remaining 12.1% from Brazil for oil harvesting. Prices of major crops, including soybeans, fluctuate significantly under the influence of weather conditions and trade disputes. Soybean has been affected by price increases due to the trade dispute between the United States and China (MAFRA, 2020). However, it appears that the main reason for price fluctuations thus far has been weather conditions. Therefore, South Korea has a weakness in that it can be immediately affected if the cultivated area is reduced or production is sluggish due to the influence of abnormal temperatures in the United States, a major source of Korean imports (KREI, 2020).
Although Korean soybean consumption is on the rise, there has been a decrease in the area used for domestic soybean cultivation area along with soybean production. The Rural Development Administration (RDA) has been continuously promoting the improvement of soybean quality to address the discordance between decreased production and increased consumption of soybeans. However, there are few studies on the systematic monitoring and analysis of the growth characteristics of soybeans following the development and improvement of new varieties (Holben et al., 1980; Hunt et al., 2010). In particular, the acquisition of accurate information tailored to the growth process of newly developed or improved varieties will help to minimize the impact of weather conditions and determine whether stable production is possible. Therefore, it is important not only for the rural economy but also for consumer prices to understand the trends in domestic soybean production, which is rising due to a shortage of supply compared to the increase in domestic demand, which is currently a problem.
The monitoring of major crops, including soybeans, has been carried out in various ways (Thenkabail et al., 2016; Zhou et al., 2019). Among them, the development of remote sensing (RS) technology performed in a noncontact method with crops has enabled efficient monitoring by extracting the vegetation index using the spectral characteristics of crops (Xue and Su, 2017). The combination of RS technology and unmanned aerial vehicle (UAV) technology is more precise than satellites, less affected by weather, and enables realtime confirmation from direct monitoring of the area of interest (Zhou et al., 2019). The convergence of various technologies, such as a combination of UAV, sensors, and deep learning, is being used to predict crop production and increase the accuracy of measurements (Lee et al., 2020; Maimaitijiang et al., 2020).
Solar energy acting on crops is a major factor in determining yield and quality because it affects crop growth, stem and leaf formation, and development. The noncontact crop spectral characteristic investigation method has the characteristic of being able to gather information on the physiological responses of cultivated crops by using the various wavelength ranges (Toscano et al., 2019; Turker-Kaya and Huck, 2017). The obtained spectral information is converted into a vegetation index, for example, and research has been conducted on this as a method for acquiring various types of information on crops (Na et al., 2016; Yamashita et al., 2021). This spectroscopic measurement method can mathematically process and interpret information such as values obtained by measuring the electromagnetic waves reflected or radiated from a measurement crop with each type of spectral sensor. In particular, the spectral information of crops can be used to determine variables such as the biomass of target crops, leaf area index (LAI), nitrogen content, crop moisture, crop diseases, and pest damage (Mariotti et al., 1996). The information obtained in this way is expected to help us improve the usability and accuracy of identifying vegetation information in a wide area with the advent and activation of devices that are good for acquiring spatial information, such as UAVs and satellites.
Zhou et al. (2019) conducted a study to predict the maturity period according to soybean varieties through multispectral images using UAV. As a result of that study, it was suggested that the method for predicting maturity through direct survey as determined by the observer’s estimation had low efficiency and economic feasibility. On the other hand, the UAV-based survey method was presented as a practical way to identify the maturity period in soybean cultivation because the area that can be surveyed is wider than that covered in a direct survey by a person. However, it is difficult to grasp the advantages of the varieties without detailed field survey data on the varieties that have been newly cultivated due to breed improvement.
This study aimed to investigate the hyperspectral characteristics of three newly developed soybean cultivars. The information obtained here is intended to be used as basic reference data for onsite growth surveys and for acquiring and analyzing wide-area information using UAV and satellite.
2. Materials and Methods
1) Study Area
As shown in Fig. 1, the study site was a test field located at the National Institute of Crop Science (NICS), Rural Development Administration (RDA), Miryang-si, Gyeongsangnam-do, Republic of Korea. Miryang-si, where NICS is located, has a high northeast and low southwest, mountains to the north and west, and Nakdong River to the south, so there are few problems with water access.
Fig. 1. Test-bed area in NICS (National Institute of Crop Science, Rural Development Administ - ration, South Korea).
In Miryang, the 30-year average temperature from June to October is 22.7°C, the highest average temperature is 33.6°C in July, and the average lowest temperature is 8.1°C in October (AW365, 2018). As of 2018, Miryang’s precipitation amounted to 915 mm from June to October. This is an increase of 580 mm over the previous year and 172.4 mm over a normal year. The average monthly temperature from June to October in Miryang in 2018 was 1.7°C higher than in a normal year, as shown in Fig. 2(a). As shown in Fig. 2(b), the amount of precipitation in September 2018 was 168.5 mm, 76.44% higher than in a normal year. This level of precipitation continued in early July and the temperature distribution was lower than normal; the soil moisture was insufficient due to 11 days of no rainfall in mid-July. The heavy rains were concentrated in August, which caused diseases to develop in soybean pods.
Fig. 2. Variations in precipitation and temperature distribution during the growing season of soybeans.
2) Materials
(1) Cultivated crops
The soybeans used in this study were of three varieties: Daewon (hereafter, DW), Pungsan (hereafter, PS), and Daepung 2ho (hereafter, DP). DW soybeans are widely cultivated for bean curd and soy sauce. PS soybeans are mainly used for growing bean sprouts and have high productivity and good palatability. The recently developed DP is a cultivar that has received multiple genes, directly or indirectly, and although the quality is somewhat inferior, it is a cultivar with excellent stability and yield. The test soybean is a cultivar that is an improved variety with a maturation period similar to that of the other tested varieties (within three days), and the unique characteristics of each variety are shown in Table 1. The length varies depending on the cultivation conditions, but is in the order DW (longest), PS, and DP.
Table 1. Characteristics of the three new soybean varieties (DW, PS, and DP)
(2) Composition of the test field
For analyzing the effect of the difference in sowing time, soybeans were sown twice, on June 5 and June 20, 2018, for each variety, and two seeds were sown per pod. The arrangement of soybeans for each variety was repeated three times for each variety, as shown in Fig. 3. The test field located in the south was sown on June 5, 2018, and the northern field was sown on June 20.
Fig. 3. Location of the three experimental soybean plots. DW, PS, and DP denotes the soybean varieties (The yellow number in the figure is the sowing date).
(3) Measuring equipment
The hyperspectral reflection characteristics of soybeans were investigated using a PSR-2500 (Spectral EvolutionTM, Haverhill, MA, USA). PSR-2500 is a spectroradiometer that can measure a wavelength range of 350 to 2500 nm. In this study, the hyperspectral reflection characteristics were measured at 1.5 nm intervals, as shown in Fig. 4, and a lens of 8° was used for FOV (Fig. 4). The measurement data obtained by the PSR-2500 can be checked directly through the PDA, so errors can be minimized. In addition, PSR- 2500 is easy to carry, making it easy to use to measure crops such as soybeans, and it has the advantage of being able to control the measurement area by using the distance from the crop.
Fig. 4. Equipment and method for measuring hyperspectral reflection characteristics of soybeans using a PSR- 2500 spectroradiometer.
3) Methods and Research Progression
(1) Spectral data acquisition and growth investigation
One of the problems in researching and interpreting crops using satellites and UAVs is that the reflective characteristics of crops vary according to the seasons and physiological changes in crops. In particular, the growing season for soybeans in South Korea is from July to October, which is a period where spectral measurements are greatly influenced by clouds and rainfall. In addition, research on the spectral reflection characteristics targeting soybeans has problems such as the fact that it is mainly conducted through outdoor measurements and the energy during the measurement time is not constant due to the influence of direct reflected light, so there are very few data. In this regard, quantifying variations in the hyperspectral reflectivity of soybeans will play an important role in obtaining more detailed information about soybeans using satellite and UAV image data in the future.
As shown in Fig. 5, field surveys on soybeans were carried out seven times, on July 19, August 6, August 20, September 5, September 19, October 10, and October 22, 2018. In Fig. 5, the growth stage of soybeans is divided into two parts according to the development (Fehr and Caviness, 1977). Table 2 shows the classification and characteristics of soybeans by growth stage. The first stage is the soybean vegetative (V) stage and the second stage is the reproductive (R) stage. Here, the number of growth stages is counted according to the number of three-leaf appearances of the soybean. The reproductive (R) stage includes the development of pods, seed development, and maturation from the flowering stage. The field survey on hyperspectral reflection characteristics conducted in this study was conducted three times in the vegetative (V) stage and four times in the reproductive (R) stage.
Fig. 5. Schematic of soybean growth phases and key stages in the life cycle of soybeans. The relationship between the growth process of soybeans and the date of field survey after sowing (DAS).
Table 2. Classification and characteristics of soybeans by growth stage (Fehr and Caviness, 1977)
Field investigation items are hyperspectral reflection characteristics using PSR-2500 and growth investigation. The field survey using PSR-2500 was conducted in each plot of the three varieties shown in Fig. 3, and 2-4 objects were randomly selected for each plot. The hyperspectral reflectivity of soybeans was measured 60 cm above the leaves and stems of soybeans, the spectral reflectance of a 99.9% barium sulfate (BaSO4) white board close to the complete diffusion surface was set to 100%, and the spectral reflectance was measured as the ratio (Fig. 4).
The reflectance ρ(λ) is calculated as in Equation (1) via the wavelength (λ), energy of the incident intensity øI(λ) on the surface of the soybean sample, and the energy intensity reflected øR(λ) from the soybean surface:
\(\rho(\lambda)=\frac{\emptyset_{R}(\lambda)}{\emptyset_{l}(\lambda)} \times 100(\%)\) (1)
where ρ(λ) is the spectral reflectance, øI(λ) is the incident intensity of solar energy, and øR(λ) is the reflected intensity of solar energy.
The first derivative ρ′(λ) for the spectral reflection spectrum is the value obtained by differentiating Equation (1) by the wavelength (λ) and is the same as Equation (2):
\(\text { 1st derivative }=\rho^{\prime}(\lambda)=\frac{d \rho(\lambda)}{d(\lambda)}\) (2)
The growth survey was conducted three times, on August 20, September 5, and September 19, 2018, for five items: soybean height, stem length, thickness of stem, branching number, and number of stalks.
(2) Data analysis
The spectral reflection characteristics of plants have mainly been researched using these wavelength bands, with an interest in the difference in reflection characteristics between visible and NIR wavelengths. However, with the spread of hyperspectral measurement devices, research using various wavelength ranges is rapidly increasing.
Hyperspectral data acquired through PSR-2500 have the advantage of acquiring wideband wavelength information, as shown in Fig. 6. It can be seen that hyperspectral characteristics are directly related to the stress symptoms and physiological characteristics of crops, depending on the wavelength. In Fig. 6, chemical and physiological variations of leaves and canopy can be seen in the regions of 480-560 nm (green) and 600-680 nm (red), where chlorophyll absorption is strong (Carter and Knapp, 2001; Datt, 1998). However, variations in this wavelength region are relatively small compared to those between 750 and 1300 nm (Gitelson et al., 2005).
Fig. 6. Typical spectral reflectance curve for soybeans, with the main factors influencing each wavelength indicated.
The sharp rise in red-edge reflectivity between the red and NIR regions corresponds to changes in leaf chemistry and canopy, and the spectral slope and slope inflection point location change according to the growth process and density of soybeans (Cho and Skidmore, 2006). The mid-infrared (MIR) spectrum of 1650-2200 nm indicates vegetation stress signals, such as variations according to changes in the moisture content and chemical substances of soybeans.
In this study, in order to determine the difference in growth and activity between soybean varieties, the characteristics of the wavelength range of 350-900 nm, excluding the highly variable NIR and mid-infrared wavelengths, were analyzed.
3. Results and Discussion
1) Hyperspectral reflection characteristics by observation day and sowing date
The hyperspectral reflection characteristics of the DW, DP, and PS soybeans used in this study show the unique spectral characteristics of soybeans, and the degree of stress varies depending on the variety and growth conditions. The spectra of the three varieties have similar changes, but show different reflectance changes depending on the wavelength. Fig. 7 shows the average canopy reflection spectrum surveyed on July 19 (V1; DAS 29), and the initial stage of growth for the three soybeans. The initial hyperspectral reflection spectrum shows a relatively low contrast between the visible (500-700 nm) and NIR (750-900 nm) regions. In particular, in the case of sowing on June 5, all three varieties showed a mixture of soil and vegetation spectra when the canopy was opened. This trend (V1) is interpreted as a result of the flat conversion of the reflection curve in the wavelength region from visible to NIR in the case of soil. In addition, the red-edge points of the three varieties form different slopes and variation characteristics. This process continued during the V2 (DAS 44) stage.
Fig. 7. Hyperspectral reflection characteristic curves for each growth stage of soybeans. DAS is the number of days after sowing, and the symbol in parentheses indicates each growth stage of the soybean.
As shown in Fig. 7, from the point when soybeans have grown to some extent (V2), the reflectance of soybeans continues to decrease due to high chlorophyll absorption in the visible region as the canopy is closed, whereas the reflectance in the NIR region increased as the structural reflection of the canopy increased (V3; DAS 61). This trend is consistent with the results of previous studies on canopy reflectivity (Fehr and Caviness, 1977).
Meanwhile, in the early stage of reproductive growth(R1; DAS 76), the effect of bean pod, leaf sharp, flower, and trichome color was reflected, and the reflectance variations in the visible, red-edge, and NIR wavelength bands converge as a profile for the spectral reflection characteristics of soybeans (Sims and Gamon, 2002). In particular, the following differences were reflected in the spectral reflection of the three varieties in the visible, red-edge, and NIR. First, soybean flowers were the same in DW and DP in white, but differed in purple in PS. Second, the color of the trichome was gray in DW and PS, but different in DP (brown).
Although the three varieties have a different leaf shape and flower color, the unique growth characteristics of soybeans were reflected and continued until the R3 (DAS 91) stage, converging to the stable hyperspectral reflection spectrum of soybeans. In the R6 (DAS 106) stage, the characteristics of the three varieties began to become clear as the flowers fell and bean pods were formed. This trend was more pronounced in the NIR wavelength than in the other wavelength bands. The spectral reflectance in the R7 (DAS 124) stage was different from that of the healthy soybeans in the previous stage in all three varieties. In particular, as the soybeans matured, they began to show canopy reflection characteristics of withered leaves and dried stems. The biggest characteristic is that it shows a high reflectance variation in visible, the slope of the red-edge becomes dull, and the reflectance in the NIR wavelength also increases rapidly as the wavelength increases. From the R7 to R8 stage, the closer to the harvest, the better the characteristics of dry leaves, stems, and bean pods were reflected, and the R8 stage reflectance of visible and NIR became closer to that of the soil.
2) First derivative spectral characteristics
Fig. 8. shows the first derivative spectra for each growth stage of the hyperspectral reflection spectra of the three varieties presented in Fig. 7. Derivatives for crop spectra, including soybeans, play a role in reducing variability due to changes in the reflectance of factors other than vegetation, and have important advantages over spectral reflectance (Curran et al., 1991). Many researchers have used the derivative method to define the wavelength position of the red edge and analyzed the relationship between the red edge and the total chlorophyll (Chl tot) concentration for the leaves and canopies of plants (Mariotti et al., 1996).
Fig. 8. Characteristics of variation in the first derivative of hyperspectral reflectance for each growth stage of soybeans.
The existence of a peak in soybeans in the first derivative spectrum is interpreted as due to the characteristic absorption spectrum of chlorophyll, as a contribution of the fluorescence response of chlorophyll, unlike in other crops. In the case of soybeans, the variation in the first derivative in the green wavelength region was concentrated at 525 nm, from V1 to R8, which is relatively small compared to the change in the red-edge region for all three varieties.
The maximum value of the first derivative of the green wavelength showed a constant value of about 0.2-0.3 at the vegetative (V) stage, then increased to about 0.5 when entering the R7 stage. Finally, the R8 stage, it decreased to about 0.1. In particular, in the green wavelength region, the difference in the first derivative between the three varieties was most pronounced in the R1 stage, which changes from the vegetative (V) stage to the reproductive (R) stage and in the R7 stage, which varies from green canopy to chlorotic phenomenon. In addition, at the 550 nm point, + and –conversion of the first derivative was done, but from the R7 stage, the same pattern change as for the previous wavelength was shown, without + and – conversion. In the R7 stage, DW showed the largest variation in the first derivative at the green wavelength region, followed by DP and PS. In the R8 stage, the variation in DW was relatively large, and PS and DP showed similar fluctuation characteristics. This variation characteristic is interpreted as a result that reflects the leaf shape, structure, and bean pod color of the three varieties presented in Table 1.
South Korea is envisioning a red-edge wavelength range sensor on the platform of a medium-sized satellite for agriculture and forestry, which is scheduled to be launched. In order to provide basic reference information on this, the red-edge was specifically reviewed. The general method of defining the red-edge in previous studies is as an intermediate region in which the crop rapidly alternates from a red with low reflectivity to a NIR with high reflectivity, which refers to the wavelength range from 680 to 750 nm. The red-edge has been used to identify the wavelength position of the peak (maximum amplitude) of the first derivative in the red-edge region. It was reported that these peaks show one distinct maximum at about 710 nm or two distinct maxima in the 690-730 nm region (Zarco-Tejada et al., 2003). However, it was confirmed that the results obtained in this study appeared in a slightly different form from those of other researchers and crops.
The three soybean varieties applied in this study showed the following characteristics at the maximum value of the red-edge due to variations in chlorophyll concentration and pigment along with the growth of soybeans. As shown in Fig. 8, the maximum value of the red-edge occurs at 720-730 nm in the vegetative (V1-V3) stage, and in the reproductive (R1-R6) stage at 725 nm. In stage V1 (DAS 29), all three varieties showed different first derivative variation characteristics. The red-edge peak of the V1 stage was formed at 720 nm and slightly increased from 700 to 730 nm, then decreased. From this period, it can be seen that the variation of the first derivative begins to increase at the red-edge and 760 nm. The variation of the first derivative at the red-edge point was the largest in DP, followed by PS and DW. The red-edge peak of the V2 (DAS 44) stage was formed at 720 nm, but the variation in the 720 to 730 nm range was larger than the previous wavelength range. In stage V2, the variations in the first derivative of the red-edge and NIR transition points began to be noticeable, representing the growth characteristics of each variety.
From the red-edge peak point to the NIR, the change in the first derivative showed a large DW, followed by PS and DP. The peak of the first derivative at the red edge had a DP of about 1.3 in the V1 stage and a DW of about 1.0, the lowest. However, as V2 and V3 progressed, DW continued to increase to about 1.5, while PS and DP showed a change of about 1.4. It was found that this trend of variation proceeded similarly up to the V3 (DAS 61) stage. This variation is interpreted to reflect the crop length, etc., of the growth characteristics at this stage. The first derivative peak at the red-edge and NIR boundary was very high in the V2 stage, so the PS was about 1.6, and the other two varieties were relatively low. In the V3 stage, PS decreased and DP and DW increased to about 1.4. As described above, the variations in the red-edge and NIR regions were very large in the vegetative (V) stage, the period when the growth characteristics of each variety are expressed.
In particular, in the R1 stage, which is the transition from the vegetative (V) stage to the reproductive (R) stage of soybeans, not only the spectral reflection curve but also the first derivative change was found to be very high at all wavelengths. These characteristics are interpreted as showing different variation characteristics depending on the wavelength because the structure, color, and shape of the leaves of the three varieties are different. The red-edge peaks from R1 to R3 were separated at both 720 and 730 nm, showing a characteristic that increased from about 1.3 in R1 to R3 and increased to 1.6. There was not much difference between the three varieties at these stages. The first derivative peak of the red-edge and NIR boundary showed more stable variation in the R1 stage than in the vegetative (V) stage. In the R3 stage, the variation in the first derivative from the red-edge peak point to the NIR was found to be high in all three varieties. This stage determines the characteristics of each variety, and results can be interpreted to reflect the major stresses for each variety.
In the R6 stage, the change in the first derivative of the red-edge was highest in DW, and DP and PS were almost the same. The variation from the red-edge peak point to converge in the NIR in almost the same manner for all three varieties. The variation in the red-edge of the R7 stage showed a large difference in the three varieties, and the peak shifted to 685 nm with the chlorotic phenomenon. The variation in DW was the largest, followed by DP, then by a clear difference in PS. On the other hand, the variation from the red-edge peak to the NIR of the R7 stage was similar to that of R6. In the R8 stage, the red-edge peak disappeared in all three cultivars. However, the variation from the red edge peak to the NIR showed a similar shape to that of R6, and the peak decreased toward R7 and R8.
Moreover, among the three varieties, the maximum value and wavelength position were slightly different in the reproductive (R) stage, and there was also a difference according to the sowing time.
3) Spectral response according to the difference in the sowing date
Multispectral remote sensing systems, such as Landsat Thematic Mapper (TM), have been operating for decades, and spectral-related indices have been developed based on broadband wavelengths to quantify plant biophysical properties such as NDVI and LAI. It is important to select the most appropriate wavelength, from the hyperspectral spectrum, to which crops respond for vegetation information and the monitoring of agricultural land and forests, even for medium-sized satellites for agriculture and forestry, scheduled to be launched by South Korea.
The conclusion of many studies is that broadband multispectral data are not suitable to remotely grasp the biochemical characteristics of vegetation, thus, a narrow bandwidth of 10 nm or less and high-resolution hyperspectral data that can distinguish vegetation information are required (Broge and Mortensen, 2002). In particular, in research using a satellite as well as UAVs, various wavelength bands are being used in the stage of developing a vegetation index with application to agriculture. In order to determine precise vegetation information, it is necessary to specify the energy absorption and reflection characteristics and reactions of various wavelengths.
The pigment corresponding to the molecule of green crops has evolved to effectively absorb energy in the visible region (350-700 nm). Chlorophyll a and b are among the most important pigments in crops. Chlorophyll a has peak absorption at 430 and 660 nm, and chlorophyll b absorbs the most energy at 450 and 650 nm (Croft and Chen, 2018). The peak point of the first derivative in the green discussed above, at 525-530 nm, is located between the absorption wavelength of chlorophyll a and b and shows relatively high energy reflection in the visible range.
In Fig. 9(a), the three varieties of soybeans show relatively high green reflectivity of about 10% during the growth period (V1 to R6), so they appear to be green in a healthy state. As for the characteristics of soybeans by growth stage, the reflectivity of soybeans decreased from about 10% to about 5% during the vegetative stages (V1 to V3). In the subsequent reproductive stage (R1-R6), the reflectivity showed a characteristic of recovering from around 5% in the initial stage to around 10% again. However, upon reaching R7, soybeans stopped growing and showed decreased chlorophyll and increased reflectance under the influence of other pigments such as β-carotene and anthocyanin (Neill and Gould, 1999). In stage R7-R8, soybean stems and leaves are dry and show signs corresponding to the chlorotic phenomenon, and water stress is at its peak during the time before harvesting.
Fig. 9. Characteristics of spectral response variation in four spectral wavelength bands for each number of days after soybean sowing.
The red wavelength region, 650 nm, corresponds to the absorption bands of chlorophyll a and b. As shown in Fig. 9(b), soybeans show a reflection characteristic of about 7%, similar to that for green, for each vegetative (V) stage. In the V1 stage, the variation in PS was large, but it was found that the variation in DW was large as it proceeded to stages V2 and V3. Moreover, from V1 to V3, the range of wavelength change tends to be very small. Therefore, the green and red wavelengths in the vegetative (V) stage showed a slight difference in the V1 and V2 stages, but there was no significant variation, and the difference between the varieties was not very large. In the red wavelength, the spectral reflectance was distributed in a very low and stable form of about 5% in the R1-R6 reproductive (R) stages. From R7 to R8, which is the late stage of reproduction, the amount of light absorption in the chlorophyll absorption band was reduced, and the reflectance was greatly increased. Soybeans also show a chlorotic phenomenon in which the reflectance increases in the green and red wavelength ranges of green plants and the leaves change to yellow in the late reproductive stage (R7-R8). The three soybean varieties showed similar distribution and variation in characteristics in the green and red wavelengths. At the red-edge (725 nm) (Fig. 9(c)), all three cultivars showed a tendency to decrease in reflectance from the vegetative V1 (about 40%) stage to V3 (about 25%). In the reproductive (R) stage, the reflectance from R1 (about 25%) to R6 increased with a narrow amplitude up to 40%, and from the R7 stage, the reflectance showed a difference between varieties, increasing and then decreasing to 25%.
In the NIR (800 nm) wavelength region (Fig. 9(d)), the spectral reflectance variation in the vegetative stage (V1 to V3) showed an almost constant distribution. In the reproductive stages from R1 to R6, the reflectance showed a tendency to slightly increase. Afterwards, in the late reproductive stage (R7-R8), the reflectance for all three varieties was significantly lowered. Among the three soybean varieties, PS had the highest reflective variation, followed by DW and DP.
Although there is no consensus among researchers about the optimal wavelength that can be applied to satellites and UAVs, it was agreed that they react very sensitively to the absorption state of pigments such as chlorophyll according to the growth conditions and time (Croft and Chen, 2018; He et al., 2020).
4) Variation characteristics of the first derivative by growth stage of soybean
It was found that the soybeans show optical variation due to the action of various crop pigments and physiological factors during the growth process. These characteristics were analyzed by changing the first derivative of the spectral spectrum, as shown in Fig. 10.
Fig. 10. Characteristics of first derivative variations for each of the four (green, red, red-edge, NIR) wavelength bands according to the growth of soybeans.
The variations in the vegetative (V) stage of the first derivative in green and red, which are visible wavelength bands, were similar for all three cultivars. The variation patterns of the three varieties from the sowing stage to the vegetative stage (V1 to V3) were also fairly similar. At the red-edge and NIR wavelengths, the three varieties showed similar variation patterns in the first derivative. The reproductive stages R1 to R6 showed similar variations in green, red-edge, and NIR. In the R7 to R8 stages, they were found to vary greatly depending on the growing environment and soybean conditions. In particular, in the R1 stage, the three varieties showed different responses and forms. PS had a positive change in the first derivative, DW had little change, and DP had a negative change. These characteristics are interpreted as reflecting the color of leaves, trichome, and pod color for each variety presented in Table 1.
The obtained results confirmed that the soybean vegetative (V) and reproductive (R) stages were clearly distinguished in terms of their spectral reflection characteristics. Moreover, it was found that the boundary between the vegetative (V) and reproductive (R) stages was closely related to the absorption of pigments such as chlorophyll along with the growth of soybeans. Especially, it was found that this occurred at the boundary between the period when pigments such as chlorophyll were an active component of the growth activity and the period when the absorption of pigments such as chlorophyll decreased (Croft and Chen, 2018).
In the red wavelength, the first derivative showed significant variation from R7 to R8, in the reproductive (R) stage. This phenomenon is interpreted as showing a large reaction in the red wavelength because of the chlorotic phenomenon due to xanthophyll, etc., which shows the stress state of soybeans, proceeding rapidly from the R7 stage (Croft and Chen, 2018). The red edge and NIR first derivative variations also showed very similar trends. From sowing to the vegetative (V) and reproductive (R) growth stage, R6, there was a large change in the first derivative with soybean growth, but from R7 onward, there was little variation due to continued stress from the drying of leaves and stems.
The spectral variation for the three soybean varieties showed the following characteristics. In the case of green and red, which are visible, large fluctuations were observed in the order of PS, DW, and DP in the vegetative stage (V1 to V3) according to decreasing magnitude. On the other hand, in the reproductive stage (R1-R6), it was confirmed that the DW with the smallest flower and leaf color difference among the three cultivars showed little change, but the two cultivars with distinct color variations showed significant changes. However, in the red wavelength, there was no specific trend, but there was a variation trend similar to the variation pattern of the vegetative (V) stage. On the other hand, in R7-R8 of the reproductive stage, large fluctuations were observed in the order of DW, PS, and DP. This phenomenon is interpreted as being greatly influenced by pigments such as chlorophyll in the early stage of growth and converging to a similar stress response in the latter half of the reproductive stage (R7- R8). In addition, the growth characteristics of the late reproductive stage (R7-R8) were consistent with the enteral conditions of the three cultivars presented in Table 1.
This study analyzed the spectral reflection characteristics of three new soybean varieties, which can be used to quickly and nondestructively examine the leaf and canopy conditions during the growth process. The obtained results are expected to provide useful basic reference information for the use of sensors such as medium-sized satellites for agriculture and forestry and the UAVs scheduled to be launched in the future.
4. Conclusions
Currently, in South Korea’s agriculture, resolving the issue of workforce shortage in rural areas and improving the working environment is a priority. To this end, spatial information on agricultural and rural areas is one important way to understand the work process and replace agricultural patterns and manual labor through mechanization, automation, and monitoring systems.
In particular, in South Korea’s open-field agriculture, there are many jobs that depend on workers and involve heavy labor and monotonous work, so it is necessary to understand when these work processes are required and to automate the corresponding technologies. However, the agricultural environment responds to various conditions and is complex, so it is necessary to consider not only the engineering approach but also the biological approach. Therefore, in the future, it is necessary to solve the problems with field agriculture by applying convergence technologies in various research fields, such as the grafting of ICT and spatial information utilization technology. In particular, it is important to secure technology and accumulate reference data that can provide agricultural and forestry information on the payload, including the planned red edge band, to the medium-sized satellite for agriculture and forestry scheduled to be launched.
In the case of soybeans, international demand is increasing while domestic production is stagnating or declining. Various soybean varieties are being improved in an attempt to meet food needs, but little research has been done on the characteristics of the cultivation process. Considering that existing methods have limitations in identifying the growth characteristics of soybeans, further efforts are needed to secure methods and technologies that utilize noncontact spatial information in line with the launch of medium-sized satellites for agriculture and forestry and the activation of UAV use. In order to increase the usability of satellites and UAVs, it is necessary to identify and analyze accurate vegetation information on the ground in parallel.
This study reviewed the spectral reflection characteristics of the new varieties of soybeans, information which can be used to identify vegetation information using satellite and UAV imagery in the future. In particular, the obtained spectral reflection characteristics for each wavelength can be converted into a vegetation index, and other useful tools (He et al., 2020). The periodic monitoring data for each growth stage gathered in this study provides useful information for analyzing the characteristics of soybeans according to the growth process. The growing season for soybeans is rainy and cloudy, which poses constraints on the use of satellites and UAVs. It is thought that the data collected in this study, which avoided this period, can provide useful ground reference information for image utilization.
The research on the spectral characteristics of soybeans by growth stage to date is insufficient. The results of first derivative analysis obtained in this study are presented by specifying the characteristics of each wavelength region. In the future, it will be possible to use it more effectively through conversion to a vegetation index. In particular, the medium-sized satellite for agriculture and forestry can be used for accurate information extraction and monitoring on crops and forest vegetation, which is being promoted as a next-generation medium-sized satellite project, but additional research is needed on various major crops to guide us in how to best utilize satellite data.
Acknowledgments
This research was funded by the Cooperative Research Program for Agriculture Science & Technology Development (Project No. PJ01404902) from the Rural Development Administration, Republic of Korea.
References
- Agricultural Weather (AW) 365, 2018. Agricultural weather analysis, http://weather.rda.go.kr/analysis/inquiry.jsp,Accessed on Jan. 26, 2021 (in Korean).
- Broge, N.H. and J.V. Mortensen, 2002. Deriving green crop area index and canopy chlorophyll density of winter wheat from spectral reflectance data, Remote Sensing of Environment, 81(1): 45-57. https://doi.org/10.1016/S0034-4257(01)00332-7
- Bryngelsson, D., S. Wirsenius, F. Hedenus, and U. Sonesson, 2016. How can the EU climate targets be met? A combined analysis of technological and demand-side changesin food and agriculture, Food Policy, 59(C): 152-164. https://doi.org/10.1016/j.foodpol.2015.12.012
- Carter, G.A. and A.K. Knapp, 2001. Leaf optical properties in higher plants: linking spectral characteristics to stress and chlorophyll concentration, American Journal of Botany, 88(4): 677-684. https://doi.org/10.2307/2657068
- Cho, M.A. and A.K. Skidmore, 2006. A new technique for extracting the red edge position from hyperspectral data: The linear extrapolation method, Remote Sensing of Environment, 101(2): 181-193. https://doi.org/10.1016/j.rse.2005.12.011
- Croft, H. and J.M. Chen, 2018. Leaf Pigment Content, Reference module in Earth System and Environment Sciences, 3: 117-142.
- Curran, P.J., J.L. Dungan, B.A. Macler, and S.E. Plummer, 1991. The Effect of a Red Leaf Pigment on the Relationship between Red Edge and Chlorophyll Concentration, Remote Sensing of Environment, 35(1): 69-76. https://doi.org/10.1016/0034-4257(91)90066-F
- Datt, B., 1998. Remote sensing of chlorophyll a, chlorophyll b, chlorophyll a+b, and total carotenoid content in eucalyptus leaves, Remote Sensing of Environment, 66(2): 111-121. https://doi.org/10.1016/S0034-4257(98)00046-7
- FAO(Food and Agriculture Organization of the United Nations), 2017. The future of food and agriculture: trends and challenges, Food and Agriculture Organization of the United Nations, Rome, IT.
- Fehr, W.R. and C.E. Caviness, 1977. Stages of Soybean Development, Iowa State University Digital Repository, Ames, IA.
- Gitelson, A.A., A. Vina, V. Ciganda, D.C. Rundquist, and T.J. Arkebauer, 2005. Remote estimation of canopy chlorophyll content in crops, Geophysical Research Letters, 32(L08403): 1-4.
- Hartman, G.L., E.D. West, and T.K. Herman, 2011. Crops that feed the World 2. Soybean-worldwide production, use, and constraints caused by pathogens and pests, Food Security, 3(1): 5-17. https://doi.org/10.1007/s12571-010-0108-x
- He, L., X. Ren, Y. Wang, B. Liu, H. Zhang, W. Liu, W. Feng, and T. Guo, 2020. Comparing methods for estimating leaf area index by multi-angular remote sensing in winter wheat, Scientific Reports, 10(1): 1-11. https://doi.org/10.1038/s41598-019-56847-4
- Holben, B.N., C.J. Tucker, and C.J. Fan, 1980. Spectral Assessment of Soybean Leaf-Area and Leaf Biomass, Photogrammetric Engineering and Remote Sensing, 46(5): 651-656.
- Huete, A., K. Didan, T. Miura, E.P. Rodriguez, X. Gao, and L.G. Ferreira, 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices, Remote Sensing of Environment, 83(1-2): 195-213. https://doi.org/10.1016/S0034-4257(02)00096-2
- Hunt, E.R., W.D. Hively, S.J. Fujikawa, D.S. Linden, C.S.T. Daughtry, and G.W. McCarty, 2010. Acquisition of NIR-Green-Blue Digital Photographs from Unmanned Aircraft for Crop Monitoring, Remote Sensing, 2(1): 290-305. https://doi.org/10.3390/rs2010290
- KREI(Korea Rural Economic Institute), 2020. Bean, Korea Rural Economic Institute, Naju, KR (in Korean).
- Lee, D.H., H.S. Shin, and J.H. Park, 2020. Developing a p-NDVI Map for Highland Kimchi Cabbage Using Spectral Information from UAVs and a Field Spectral Radiometer, Agronomy, 10(11): 1798. https://doi.org/10.3390/agronomy10111798
- Lee, H.S. and K.S. Lee, 2017. Effect of Red-edge Band to Estimate Leaf Area Index in Close Canopy Forest, Korean Journal of Remote Sensing, 33(5-1): 571-585 (in Korean with English abstract). https://doi.org/10.7780/KJRS.2017.33.5.1.10
- Lobell, D.B., M.B. Burke, C. Tebaldi, M.D. Mastrandrea, W.P. Falcon, and R.L. Naylor, 2008. Prioritizing climate change adaptation needs for food security in 2030, Science, 319(5863): 607-610. https://doi.org/10.1126/science.1152339
- MAFRA(Ministry of Agriculture, Food and Rural Affairs), 2020. Soybean production, purchase and consumption by use, Ministry of Agriculture, Food and Rural Affairs, Sejong, KR(in Korean).
- Maimaitijiang, M., V. Sagan, P. Sidike, A.M. Daloye, H. Erkbol, and F.B. Fritschi, 2020. Crop Monitoring Using Satellite/UAV Data Fusion and Machine Learning, Remote Sensing, 12(9): 1357. https://doi.org/10.3390/rs12091357
- Mariotti, M., L. Ercoli, and A. Masoni, 1996. Spectral properties of iron-deficient corn and sunflower leaves, Remote Sensing of Environment, 58(3): 282-288. https://doi.org/10.1016/S0034-4257(96)00070-3
- Na, S.I., S.Y. Hong, C.W. Park, K.D. Kim, and K.D. Lee, 2016. Estimation of highland kimchi cabbage growth using UAV NDVI and agro-meteorological factors, Korean Journal of Soil Science and Fertilizer, 49(5): 420-428 (in Korean with English abstract). https://doi.org/10.7745/KJSSF.2016.49.5.420
- Neill, S. and K.S. Gould, 1999. Optical properties of leaves in relation to anthocyanin concentration and distribution, Canadian Journal of Botany-Revue Canadienne De Botanique, 77(12): 1777-1782. https://doi.org/10.1139/cjb-77-12-1777
- OECD(Organisation for Economic Co-operation and Development) and FAO(Agriculture Organization of the United Nations), 2020. OECD-FAO Agricultural Outlook 2020-2029, OECD and Food and Agriculture Organization of the United Nations, Paris, FR.
- Sims, D.A. and J.A. Gamon, 2002. Relationships between leaf pigment content and spectral reflectance across a wide range of species, leaf structures and developmental stages, Remote Sensing of Environment, 81(2-3): 337-354. https://doi.org/10.1016/S0034-4257(02)00010-X
- Thenkabail, P.S., J.G. Lyon, and A. Huete, 2016. Hyperspectral remote sensing of vegetation, CRC press, Boca Raton, FL, US.
- Toscano, P., A. Castrignano, S.F. Di Gennaro, A.V. Vonella, D. Ventrella, and A. Matese, 2019. A Precision Agriculture Approach for Durum Wheat Yield Assessment Using Remote Sensing Data and Yield Mapping, Agronomy, 9(8): 437. https://doi.org/10.3390/agronomy9080437
- Turker-Kaya, S. and C.W. Huck, 2017. A Review of Mid-Infrared and Near-Infrared Imaging: Principles, Concepts and Applications in Plant Tissue Analysis, Molecules, 22(1): 168. https://doi.org/10.3390/molecules22010168
- Xue, J.R. and B.F. Su, 2017. Significant Remote Sensing Vegetation Indices: A Review of Developments and Applications, Journal of Sensors, 2017: 1-17.
- Yamashita, H.,R. Sonobe,Y. Hirono,A. Morita, andT. Ikka, 2021. Potential ofspectroscopic analyses for non-destructive estimation of tea quality-related metabolites in fresh new leaves, Scientific Reports, 11(1): 4169. https://doi.org/10.1038/s41598-021-83847-0
- Zarco-Tejada, P.J., J.C. Pushnik, S. Dobrowski, and S.L. Ustin, 2003. Steady-state chlorophyll a fluorescence detection from canopy derivative reflectance and double-peak red-edge effects, Remote Sensing of Environment, 84(2): 283-294. https://doi.org/10.1016/S0034-4257(02)00113-X
- Zhao, D.H., J.L. Li, and J.G. Qi, 2004. Hyperspectral characteristic analysis of a developing cotton canopy under different nitrogen treatments, Agronomy, 24(8): 463-471. https://doi.org/10.1051/agro:2004042
- Zhou,J., D.Yungbluth, C.N. Vong, A. Scaboo, and J.F. Zhou, 2019. Estimation ofthe Maturity Date of Soybean Breeding Lines Using UAV-Based Multispectral Imagery, Remote Sensing, 11(18): 2075. https://doi.org/10.3390/rs11182075