A Preliminary Open-field Study Investigating Commercial Smart Farming of a Potential Alternative Crop, Aster koraiensis

Abstract

Aster koraiensis, a plant indigenous to Korea, has recently attracted a great deal of attention as a potential functional food and pharmaceutical product due to its preventive and therapeutic efficacy against various diseases. Despite increasing demand, supply of A. koraiensis is limited to wild populations and small-scale wild plant growers. In an attempt to establish cultivation practices in open-field production platforms, this study investigated the effects of soil mulching, planting density, and fertilizer treatment on the above-ground dry matter yield of A. koraiensis for two years after transplantation. The maximum dry matter yield was obtained at 1,935 kg DW·ha⁻¹ and 3,803 kg DW·ha⁻¹ in the first and second years, respectively, under soil mulching and 15 × 15 cm planting density conditions. Soil mulching during the first year after transplanting significantly promoted plant growth, and increased yield, indicating its beneficial effects on the establishment and early growth of A. koraiensis plants. Fertilizer treatment increased yields up to 3.7-fold in the second year after transplantation. In addition, this study employed remote sensing technologies using a drone equipped with a multispectral sensor to evaluate the normalized difference vegetation index (NDVI) of experimental plots across different plant growth stages. The results revealed that NDVI values at 60 days after shoot emergence in the second year after transplantation produced the highest correlation with dry matter yield in a simple exponential regression model. Remote sensing technologies and the proposed regression model could be applied to optimize cultivation practices and enable precision agriculture for A. koraiensis crops.

Introduction

Aster koraiensis, also known as Gymnaster koraiensis, is a herbaceous perennial plant belonging to the Asteraceae family. A. koraiensis is a Korean native plant taxonomically distinct from other Aster species such as A. tataricus and A. yomena. A. koraiensis has attracted a lot of attention due to its high potential as a novel functional food, as well as for its pharmaceutical properties. In Korea, A. koraiensis has been occasionally consumed as a seasonal leafy vegetable, and has also traditionally been used as a folk medicine to control stroke, inflammation, and respiratory diseases such as pneumonia, pertussis, and chronic bronchitis (Ahn, 1998; Choi et al., 2017). It has been demonstrated that the extracts and compounds from A. koraiensis have various biological activities, including antioxidant, antidiabetic, antiangiogenic, antithrombotic, chemopreventive, and hepatoprotective activities (Choi et al., 2017; Jung et al., 2002; Kim et al., 2011; Kim et al., 2016; Kim et al., 2018; Lee et al., 2010; Lee et al., 2012; Sohn et al., 2010). Together with these pharmacological activities, its extracts exhibit preventive and therapeutic efficacy against glaucoma, hepatocarcinoma, diabetic retinopathy, and diabetic nephropathy both in vitro and in vivo (Kim et al., 2011; Kim et al., 2016; Kim et al., 2017; Kim et al., 2018; Lee et al., 2010; Sohn et al., 2010). In addition to its edibility and medicinal uses, A. koraiensis has recently been used as a cut flower and as a landscape and groundcover plant because of its long bloom period, vigorous rhizome growth and perennial nature.

Once a plant is identified as a novel resource for value-added functional foods and/or pharmaceuticals, commercialization of the product primarily depends on its stable and consistent supply. A. koraiensis is a wild plant, and much of its supply has relied on collection from wild populations and a few small-scale wild plant producers. Although A. koraiensis has been planted as a landscape and groundcover plant, it has not been produced on a large commercial scale. For efficient commercial production of a potential new crop, it is crucial to understand its behavior in an agricultural setting and establish effective cultivation practices. This process requires extensive optimization. For A. koraiensis, its photosynthetic characteristics and dry matter productivity were studied over different developmental stages and seasons (Nam et al., 2009, 2016). To date, however, studies related to establishing large-scale agricultural production have not been reported for A. koraiensis and its yield potential has not been addressed in an open-field production platform.

Establishing cultivation practices for the production of new crops requires an optimizing process in which data related to various agricultural traits needs to be collected and analyzed under variable cultivation and environmental conditions. Remote sensing technologies coupled with an unmanned aerial vehicle (UAV) have become a powerful tool for acquiring in-field phenotypic traits in an efficient and high-throughput manner (Maes and Steppe, 2019; Yang et al., 2017). Among UAVs, commercial or customized drones are easily modified as sensor carriers, and have the extra advantages of low cost and simple operation. Multispectral imaging sensors capable of sensing visible and invisible spectra have been deployed for crop phenotyping. Multispectral data containing reflection characteristics of the crop canopy and adjacent environment against different spectra have been widely used to develop vegetation indices (VIs), which are designed for quantitative and qualitative evaluation of crop growth status (Bendig et al., 2015; Xue and Su, 2017; Yang et al., 2017). One of the most commonly used VIs has been a normalized difference vegetation index (NDVI) derived from the combination of spectra in the NIR and red channels (Friedl, 2018; Karnieli et al., 2010; Zaman-Allah et al., 2015). In a wide range of crops, including cereals and vegetables, yield estimation models have been developed using NDVI (Johnson and Trout, 2012; Labus et al., 2002).

The present study investigated the effects of soil mulching, planting density and fertilizer treatment on the above-ground biomass yield of A. koraiensis in the first and second years after transplantation under open-field conditions. Remote sensing technology using an unmanned aerial vehicle (UAV) and multispectral sensors was also applied to monitor growth performance. The relationship between the above ground biomass and NDVI was assessed, and a yield prediction regression model was proposed based on the NDVI.

Materials and Methods

Plant materials

Seeds of A. koraiensis were collected from four-year-old plants grown by the Pyeongchang Wild Plant Nursery and Farming Corporation (Pyeongchang, Korea) and stored at 4°C until use. In late March 2017, seeds were directly sown onto 105 cells of a vegetable nursery tray (10–15 seeds/cell) filled with nursery soil (Baroker-Horticultural Purpose, Seoul Bio, Korea). Seedlings were grown for five weeks in a polythene-covered growth house under natural photoperiod and ambient temperatures and irrigated twice a week with an irrigation depth of 10 mm. Ten seedlings per nursery cell were considered one plant, and they were transplanted in the experimental field in early May 2017.

Experimental sites and treatments

The experiments described here were conducted under rain-fed open-field conditions during the growing seasons in 2017 and 2018 at two Precision-Farming Experimental Stations (ES1: 37°47'N, 128°51'E and ES2: 37°30'N, 128°51'E) of the Institute of Natural Products, Gangneung, Korea. Soil was tilled to a depth of 30 cm, and planting beds were prepared in ES1 and ES2 using a tractor equipped with a bed shaper. Bed width and height were 1.2 m and 0.2 m, respectively, and alley spacing was 0.6 m.

Effects of soil mulching and plant density on growth performance and yield of A. koraiensis were investigated in the first and second year after transplanting at ES1. The experimental layout of ES1 was a split plot design in randomized complete blocks, with three replications. The primary treatment was soil mulching with two sub-treatments: 1) without mulching (WM) and 2) black polyethylene mulching (BM). Each primary treatment was assigned to two planting beds in each block, and each bed was 54 m long. The secondary treatment was planting density (PD) with three sub-treatments: 1) PD1, 9.3 plants·m⁻², 30 × 20 cm spacing; 2) PD2, 13.9 plants·m⁻², 20 × 20 cm spacing; and 3) PD3, 25.9 plants·m⁻², 15 × 15 cm spacing. Each secondary treatment had four pseudo-replications, and each pseudo-replication plot was 9 m long and randomly assigned into two planting beds within the primary treatment condition. Mulching was removed in the second year (2018) after transplantation to facilitate shooting from rhizomes. Weeds were removed manually three times a year from the beds without mulching.

The effects of fertilizer on the yield of A. koraiensis were investigated in the second year (2018) in ES2 for which the planting density was 13.9 plants·m⁻² (20 × 20 cm spacing). The experimental plot layout of ES2 was a randomized complete block design with three replications and five fertilizer level (FL) treatments: FL1, no fertilizer treatment; FL2, N:P:K = 21:17:17 kg·ha⁻¹; FL3, N:P:K = 42:34:34 kg·ha⁻¹; FL4, N:P:K = 84:68:68 kg·ha⁻¹; and FL5 N:P:K = 168:136:136 kg·ha⁻¹. There were two pseudo-replication plots for each fertilizer treatment (FL2–FL5), while FL1 had four pseudo-replications in each block. Planting beds for each plot were 5 m, 4 m, and 3 m long in blocks 1, 2, and 3, respectively. Weeds were removed manually three times a year.

Data acquisition

Relative growth rate (RGR) was investigated in the mulching and planting density treatment groups. Above-ground biomass from six plants was sampled in each treatment and block at the time of transplanting (0 DAT), and at 30 and 60 days after transplanting (30 DAT and 60 DAT, respectively). To measure dry weight, samples were dried under uniform conditions at 30°C with 20% relative humidity until a constant weight was achieved. RGR was calculated as $\mathrm{RGR}=(\stackrel{\_}{\mathit{ln}{W}_2}-\stackrel{\_}{\mathit{ln}{W}_1})∕(t_2-t_1)$, where W1 and W2 were dry weights of above ground biomass at times t1 and t2, and $\stackrel{\_}{\mathit{ln}{W}_1}$ and $\stackrel{\_}{\mathit{ln}{W}_2}$ were the means of the natural logarithm‐transformed plant dry weights (Hoffmann and Poorter, 2002).

To evaluate the yield of A. koraiensis crops, all of the above ground biomass that could be used as functional food and pharmaceuticals was collected from the whole experimental plot. In the first year, A. koraiensis did not produce flowering stems. Harvest of one-year-old plants was carried out at 150 DAT (late October, 2017) to maximize yields before leaf senescence occurred. In the second year, bolting occurred around 90 days after shoot emergence (90 DAS). Harvesting of two-year-old plants was conducted at 90 DAS (early June, 2018), before bolting.

For both years, total fresh weight (FW) for each treatment and block was measured at the time of harvest, and 500 g FW was sampled from each treatment and block, followed by drying in a drier at 30°C with 20% relative humidity until a constant weight was achieved. Total dry weight (DW) for each treatment and block was extrapolated from the dry weight of a 500 g FW sample.

Remote sensing

A drone (K-mapper X4 LTE quadcopter; Sistech Inc., Seoul, Korea) equipped with a multispectral camera, Parrot SEQUOIA (Parrot Drones SAS, Paris, France) was used to perform an aerial survey. The camera had a 16 mega pixel RGB sensor and four 1.2 mega pixel spectral sensors capturing Green (550/40 nm), Red (660/40 nm), Red edge (735/10 nm), and NIR (790/40 nm) spectral bands. The camera also had an on-board sunshine sensor allowing automatic normalization of irradiance levels for each survey. The aerial survey was conducted on ES2 (fertilizer-treated experimental field) at 53, 60, 67, and 80 days after shoot emergence (53, 60, 67, 80 DAS, respectively) in the second year after transplantation. Mission planner software (ver. 1.3.60) was used to set the waypoint and conduct automatic flights. Images were acquired around solar noon. The image sensor trigger interval was set to achieve a 77% forward and side image overlap. The speed and height of the UAV was 3 m·s⁻¹ and 30 m above ground level achieving a ground sampling distance of 2.83 cm.

The images for red and NIR spectral bands taken from each flight were georeferenced and processed into orthomosaic and reflectance maps using Pix4Dmapper Pro program (Pix4D SA; Lausanne, Switzerland). For radiometric calibration, images of a calibrated reference panel taken for each flight were provided when the reflectance map was generated. To designate regions of interest (ROIs) corresponding to each fertilizer-treated plot and to calculate the mean NDVI values for each ROI, orthomosaic and calibrate reflectance maps were added into a geographic information system, QGIS (ver. 2.18.23). ROIs defined by 0.2 m inside the outline of each plot to exclude border effects (Fig. 2A) were layered on the reflectance map using QGIS. Means of NDVI value (NDVI = NIR_790 nm − Red_660 nm/NIR_790 nm + Red_660 nm) for each ROI corresponding to each plot were calculated and extracted using Raster Calculator and Statistics functions in QGIS.

Statistical analysis

ANOVA and regression analysis were performed using the computing environment of R (R Development Core Team, 2005). ANOVA was conducted among the means of above-ground biomass yield from mulching, planting density treatment groups (six groups·year⁻¹) and fertilizer treatment groups (five groups·year⁻¹). Statistical significance among the means for the groups was determined using Fisher’s protected LSD test at P < 0.05 (n = 3).

In order to propose a yield estimation model, simple exponential and linear regression analyses were conducted with the mean of NDVI for each fertilizer-treated plot and the mean yield of each corresponding plot (n = 15).

Results and Discussion

Establishing generally applicable cultivation practices and investigating the potential yield are the first steps towards commercial production of A. koraiensis as a novel resource for functional foods and pharmaceuticals. In this study, five-week-old A. koraiensis seedlings were transplanted and cultivated under open field conditions in a raised bed cropping layout (1.2 m wide and 0.2 m high bed). Since transplanting and raised bed cropping are commonly used agricultural practices for leafy vegetable crops, they could be simply adapted for A. koraiensis production. Moreover, transplanting with precise spacing provides greater ability to produce uniform feedstocks, which is a prerequisite for the standardization of end products, such as functional foods and pharmaceuticals.

Determining the optimal planting density is important for improving land use efficiency and minimizing inter-plant competition. Soil mulching also provides various beneficial effects on crop growth by reducing weed infestation and retaining soil moisture and nutrients (Kader et al., 2017; Kasirajan and Ngouajio, 2012). Evaluation of the above effects on A. koraiensis productivity for two consecutive years showed that maximum dry matter yields were achieved under a planting density of 15 × 15 cm spacing (PD3) with black polyethylene mulching (BM). Under these conditions, yields were 1,935 kg DW·ha⁻¹ and 3,803 kg DW·ha⁻¹ in the first and second year after transplantation, respectively (Table 1). Regardless of soil mulching, planting density of 15 × 15 cm spacing (PD3) resulted in the highest yield compared with 30 × 20 cm (PD1) and 20 × 20 cm spacing (PD2) for both years (Table 1).

Table 1

Effects of soil mulching and plant density on the yield of Aster koraiensis.

Mulching significantly increased yields by 4.5-, 5.8-, and 3.1-fold in the first year under the planting densities PD1, PD2, and PD3, respectively, compared with those without mulching (Table 1). The positive effect of soil mulching was observed on the early growth of transplanted A. koraiensis plants. At 30 DAT and 60 DAT, means of the relative growth rate (RGR) with mulching were 2.2- and 1.4-fold higher than those without mulching (Fig. 1). Except for PD3 under mulching conditions at 30 DAT, planting density did not affect RGR in either mulching or no mulching treatments at both 30 DAT and 60 DAT (Fig. 1). Given the planting density conditions, there may have been negligible intra-plant competition during the early growth period. In the second year, although removal of mulch in the second season resulted in higher yields compared to no mulching, the increase in the annual yield rate was higher without mulching, ranging from 3.7 to 5.6 (Table 1). Mulching facilitates relatively vigorous growth during the establishment year, while possibly resulting in intra-plant competition in the following year.

Fig. 1

Effect of soil mulching and planting density on the early growth of transplanted A. koraiensis plants. The relative growth rate (RGR) was measured at 30 and 60 days after transplanting in 2017 (A). PD1, PD2, and PD3 indicate planting densities of 30 × 20 cm (9.3 plants·m⁻²), 20 × 20 cm (13.9 plants·m⁻²), and 15 × 15 cm (25.9 plants·m⁻²), respectively. PD1M, PD2M, and PD3M correspond to planting densities of PD1, PD2, and PD3 under mulching conditions. A. koraiensis plants at 60 DAT under PD3 (B) and PD3M (C).

In an attempt to deploy remote sensing technologies to establish cultivation practices for A. koraiensis crops, we first developed regression models to estimate the above ground dry matter yield based on NDVI values extracted from aerial multispectral images. To create variations in vegetation and yield, five different fertilizer treatments were applied to ES2 in the second year. Depending on fertilizer levels, the mean of the above-ground dry matter yield was variable, ranging from 916 to 3,559 kg DW·ha⁻¹ (Table 2). During the cultivation period, aerial multispectral images were collected at different growth stages, and the mean NDVI values were extracted for each growth stage and fertilizer treatment plot (Fig. 2A, B). After acquiring the dry matter yield for each plot, simple regression analysis (both exponential and linear regression) was conducted between NDVI and yield (Table 3). In all growth stages, exponential regression models had higher R² and lower RMSE compared to linear regression models, indicating its superior accuracy for yield estimation. The exponential regression model developed using NDVI at 60 DAS showed the best fit with the highest R² = 0.752 and lowest RMSE of 640 kg DW·ha⁻¹ (Table 3; Fig. 2C). The pattern of our yield estimation models is consistent with previous reports that canopy biomass tends to be more accurately estimated by the exponential relationship with NDVI (Baret and Guyot, 1991; Kawamura et al., 2005; Richter et al., 2016). This is because NDVI reaches a plateau as leaf canopies become denser, and it is inappropriate to represent harvestable biomass under leaf canopies since NDVI is an estimate of the surface spectral reflectance (Haboudane et al., 2004; Pontailler et al., 2003). Although the NDVI values for A. koraiensis showed an asymptotic saturation problem after a certain density of vegetation was reached, its coefficient of determination with yield was maintained above 0.666 in the exponential regression models (Table 3; Fig. 2C). The NDVI values of A. koraiensis at or before 60 DAS provided a better estimation performance than those at 67 and 80 DAS (Table 3), suggesting that dates of harvesting and/or image acquisition could be revised in the yield estimation modeling. In addition, a combination of other VIs to improve predictive strength awaits further investigation.

Table 2

Effects of fertilizer treatment on the yield of Aster koraiensis.

Fig. 2

RGB aerial images over a fertilizer-treated experimental plot (ES2) at 60 days after shoot emergence (60 DAS) in the second year after transplanting (2018) (A). Each box with a black line indicates each fertilizer treatment plot with different levels; 0, 21, 42, 84, and 168 indicate the amount of fertilizer used (0, no treatment, 21, N:P:K = 21:17:17 kg·ha⁻¹; 42, N:P:K = 42:34:34 kg·ha⁻¹; 84, N:P:K = 84:68:68 kg·ha⁻¹; and 168, N:P:K = 168:136:136 kg·ha⁻¹). Normalized difference vegetation index (NDVI) images created using multispectral data over ES2 at 60 DAS (B). The proposed yield estimation regression model based on remotely sensed NDVI at 60 DAS (C).

Table 3

Coefficient of determination (R²) and root mean square error (RMSE) for regression between NDVI (x) and dry matter yield (y).

Remote sensing technologies and the regression model proposed here can be applied to optimizing cultivation practices or precision agriculture for A. koraiensis. Remotely sensed NDVI and yield estimation could facilitate the selection of the best possible cultivation and environmental conditions in a high-throughput manner. Furthermore, during commercial production, these technologies and growth models could enable variable rate applications of various agricultural inputs such as water and agrochemicals.

Conclusion

To provide optimal practices for A. koraiensis farming, this study examined the effects of various agricultural inputs on the crop yield based on open-field experiments over a two-year period. The maximum production of A. koraiensis was achieved in the second year of cultivation, and soil mulching, higher planting density and fertilizer application contributed to an increased yield. As a first step towards precision agriculture for A. koraiensis, a yield prediction model was also generated using NDVI values obtained from UAV-based remote sensing. Further research on monitoring additional vegetation indices at different growth stages and/or conditions is necessary to create advanced prediction models that could help growers to make optimal management decisions.

Literature Cited

• Ahn, D. K. 1998. Illustrated Book of Korean Medicinal Herbs. Kyo-Hak Publishing. Seoul (In Korean).
•  Baret,  F. and  G.  Guyot. 1991. Potentials and limits of vegetation indices for LAI and APAR assessment. Remote Sens. Environ. 35: 161–173.
•  Bendig,  J.,  K.  Yu,  H.  Aasen,  A.  Bolten,  S.  Bennertz,  J.  Broscheit,  M. L.  Gnyp and  G.  Bareth. 2015. Combining UAV-based plant height from crop surface models, visible, and near infrared vegetation indices for biomass monitoring in barley. Int. J. Appl. Earth Obs. Geoinf. 39: 79–87.
•  Choi,  J.-H.,  K.-J.  Kim,  S.-J.  Kim and  S.  Kim. 2017. Novel protease from the leaves of edible medicinal plant Aster koraiensis Nakai with antithrombotic activity: Purification and partial characterization. J. Food Biochem. 41: e12334. DOI: 10.1111/jfbc.12334.
• Friedl, M. A. 2018. Remote Sensing of Croplands. p. 78–95. In: S. Liang (ed.). Comprehensive Remote Sensing. Elsevier. Oxford.
•  Haboudane,  D.,  J. R.  Miller,  E.  Pattey,  P. J.  Zarco-Tejada and  I. B.  Strachan. 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.
•  Hoffmann,  W. A. and  H.  Poorter. 2002. Avoiding bias in calculations of relative growth rate. Ann. Bot. 90: 37–42.
•  Johnson,  L. F. and  T. J.  Trout. 2012. Satellite NDVI assisted monitoring of vegetable crop evapotranspiration in California’s San Joaquin valley. Remote Sens. 4: 439–455.
•  Jung,  H.-J.,  B.-S.  Min,  J.-Y.  Park,  Y.-H.  Kim,  H.-K.  Lee and  K.-H.  Bae. 2002. Gymnasterkoreaynes A–F, cytotoxic polyacetylenes from Gymnaster koraiensis. J. Nat. Prod. 65: 897–901.
•  Kader,  M. A.,  M.  Senge,  M. A.  Mojid and  K.  Ito. 2017. Recent advances in mulching materials and methods for modifying soil environment. Soil Till. Res. 168: 155–166.
•  Karnieli,  A.,  N.  Agam,  R. T.  Pinker,  M.  Anderson,  M. L.  Imhoff,  G. G.  Gutman,  N.  Panov and  A.  Goldberg. 2010. Use of NDVI and land surface temperature for drought assessment: Merits and limitations. J. Climate 23: 618–633.
•  Kasirajan,  S. and  M.  Ngouajio. 2012. Polyethylene and biodegradable mulches for agricultural applications: a review. Agron. Sustain. Dev. 32: 501–529.
•  Kawamura,  K.,  T.  Akiyama,  H.-O.  Yokota,  M.  Tsutsumi,  T.  Yasuda,  O.  Watanabe and  S.  Wang. 2005. Comparing MODIS vegetation indices with AVHRR NDVI for monitoring the forage quantity and quality in Inner Mongolia grassland, China. Grassland Sci. 51: 33–40.
•  Kim,  J.,  K.  Jo,  C.-S.  Kim and  J. S.  Kim. 2017. Aster koraiensis extract prevents diabetes-induced retinal vascular dysfunction in spontaneously diabetic Torii rats. BMC Complem. Altern. Med. 17: 497. DOI: 10.1186/s12906-017-1998-3.
•  Kim,  J.,  K.  Jo,  I.-S.  Lee,  C.-S.  Kim and  J. S.  Kim. 2016. The Extract of Aster Koraiensis prevents retinal pericyte apoptosis in diabetic rats and its active compound, chlorogenic acid inhibits AGE formation and AGE/RAGE interaction. Nutrients 8: 585.
•  Kim,  J.,  Y. M.  Lee,  W.  Jung,  S.-B.  Park,  C.-S.  Kim and  J. S.  Kim. 2018. Aster koraiensis extract and chlorogenic acid inhibit retinal angiogenesis in a mouse model of oxygen-induced retinopathy. Evid. Based Complement. Alternat. Med. 2018: 8. DOI: 10.1155/2018/6402650.
•  Kim,  K.-A.,  K. D.  Kang,  E. H.  Lee,  C. W.  Nho and  S. H.  Jung. 2011. Edible wild vegetable, Gymnaster koraiensis protects retinal ganglion cells against oxidative stress. Food Chem. Toxicol. 49: 2131–2143.
•  Labus,  M. P.,  G. A.  Nielsen,  R. L.  Lawrence,  R.  Engel and  D. S.  Long. 2002. Wheat yield estimates using multi-temporal NDVI satellite imagery. Int. J. Remote Sens. 23: 4169–4180.
•  Lee,  J.,  Y. M.  Lee,  B. W.  Lee,  J.-H.  Kim and  J. S.  Kim. 2012. Chemical constituents from the aerial parts of Aster koraiensis with protein glycation and aldose reductase inhibitory activities. J. Nat. Prod. 75: 267–270.
•  Lee,  S. B.,  K.  Kang,  S.  Oidovsambuu,  E. H.  Jho,  J. H.  Yun,  J.-H.  Yoo,  E.-H.  Lee,  C.-H.  Pan,  J. K.  Lee,  S. H.  Jung and  C. W.  Nho. 2010. A polyacetylene from Gymnaster koraiensis exerts hepatoprotective effects in vivo and in vitro. Food Chem. Toxicol. 48: 3035–3041.
•  Maes,  W. H. and  K.  Steppe. 2019. Perspectives for remote sensing with unmanned aerial vehicles in precision agriculture. Trends Plant Sci. 24: 152–164.
•  Nam,  H.-H.,  H.-S.  Lee and  J.-H.  Woo. 2009. Seasonal changes in dry matter productivity of Korean native plants, Aster koraiensis, Iris ensata, and Hemerocallis fulva. Korean J. Hort. Sci. Technol. 27: 371–379 (In Korean with English abstract).
•  Nam,  H.-H.,  C.-K.  Son,  J.-H.  Lee and  J.-B.  Kwon. 2016. Photosynthetic characteristics of korean endemic plant, Aster koraiensis Nakai according to growth and development conditions. Korean J. Plant Resour. 29: 393–399 (In Korean with English abstract).
•  Pontailler,  J.-Y.,  G. J.  Hymus and  B. G.  Drake. 2003. Estimation of leaf area index using ground-based remote sensed NDVI measurements: validation and comparison with two indirect techniques. Canadian J. Remote Sens. 29: 381–387.
• R Development Core Team. 2005. R: A language and environment for statistical computing. R foundation for statistical computing, Vienna, Austria. <http://www.R-project.org>.
•  Richter,  G. M.,  F.  Agostini,  A.  Barker,  D.  Costomiris and  A.  Qi. 2016. Assessing on-farm productivity of Miscanthus crops by combining soil mapping, yield modelling and remote sensing. Biomass and Bioenergy 85: 252–261.
•  Sohn,  E.,  J.  Kim,  C.-S.  Kim,  Y. S.  Kim,  D. S.  Jang and  J. S.  Kim. 2010. Extract of the aerial parts of Aster koraiensis reduced development of diabetic nephropathy via anti-apoptosis of podocytes in streptozotocin-induced diabetic rats. Biochem. Biophys. Res. Commun. 391: 733–738.
•  Xue,  J. and  B.  Su. 2017. Significant remote sensing vegetation indices: A review of developments and applications. J. Sensors 2017: 17. DOI: 10.1155/2017/1353691.
•  Yang,  G.,  J.  Liu,  C.  Zhao,  Z.  Li,  Y.  Huang,  H.  Yu,  B.  Xu,  X.  Yang,  D.  Zhu,  X.  Zhang,  R.  Zhang,  H.  Feng,  X.  Zhao,  Z.  Li,  H.  Li and  H.  Yang. 2017. Unmanned aerial vehicle remote sensing for field-based crop phenotyping: Current status and perspectives. Front. Plant Sci. 8: 1111. DOI: 10.3389/fpls.2017.01111.
•  Zaman-Allah,  M.,  O.  Vergara,  J. L.  Araus,  A.  Tarekegne,  C.  Magorokosho,  P. J.  Zarco-Tejada,  A.  Hornero,  A. H.  Albà,  B.  Das,  P.  Craufurd,  M.  Olsen,  B. M.  Prasanna and  J.  Cairns. 2015. Unmanned aerial platform-based multi-spectral imaging for field phenotyping of maize. Plant Methods 11: 35. DOI: 10.1186/s13007-015-0078-2.