Journal Search Engine
Search Advanced Search Adode Reader(link)
Download PDF Export Citaion korean bibliography PMC previewer
ISSN : 1598-5504(Print)
ISSN : 2383-8272(Online)
Journal of Agriculture & Life Science Vol.54 No.3 pp.95-104

Yield Prediction of Chinese Cabbage (Brassica rapa var. glabra Regel.) using Narrowband Hyperspectral Imagery and Effective Accumulated Temperature

Ye-Seong Kang, Sae-Rom Jun, Si-Hyeong Jang, Jun-Woo Park, Hye-Young Song, Chan-Seok Ryu†*
Department of Bio-industrial Machinery Engineering, Gyeongsang National University (Institute of Agriculture and Life Science), Jinju, Gyeongsangnam-do, 52828, Republic of Korea

These authors contributed equally to this work.

*Corresponding author: Chan-Seok Ryu Tel: +82-55-772-1897 Fax: +82-55-772-1899 E-mail:
March 14, 2020 ; June 10, 2020 ; June 22, 2020


In this paper, the model for predicting yields of chinese cabbages of each cultivar (joined-up in 2015 and wrapped-up in 2016) was developed after the reflectance of hyperspectral imagery was merged as 10 nm, 25 nm and 50 nm of FWHM (full width at half maximum). Band rationing was employed to minimize the unstable reflectance of multi-temporal hyperspectral imagery. The stepwise analysis was employed to select key band ratios to predict yields in all cultivars. The key band ratios selected for each of FWHM were used to develop the yield prediction models of chinese cabbage for all cultivars (joined-up & wrapped-up) and each cultivar (joined-up, wrapped-up). Effective accumulated temperature (EAT) was added in the models to evaluate its improvement of performances. In all models, the performance of models was improved with adding of EAT. The models with EAT for each of FWHM showed the predictability of yields in all cultivars as R2≥0.80, RMSE≤694 g/plant and RE≤28.3%. Such as this result, if the yield can be predicted regardless of the cultivar, it is considered to be advantageous when predicting the yield over a wide area because it is not require a cultivar classification work as pre-processing in imagery.



    Chinese cabbage (Brassica rapa var. glabra Regel.) is a typically cultivated crop that serves widely in Asia and the increase in production have been required steadily by population and income increases. Nowadays, climate change such as an increase in temperature and precipitation has long been a threat to cultivation stability to maintain the production and quality of agricultural products with loss of agricultural land, water shortage, advancing the northern limit of cultivation, and new pests and weeds. Therefore, it is essential to establish a climate change management system to curb carbon emissions and increase carbon uptake, as well as research and development that can produce stable food for climate change. Autumn chinese cabbage has been primarily cultivated in central and southern regions, the Republic of Korea with optimal temperatures 18 to 20. The extreme temperature increase will adversely affect the yields of chinese cabbage (Song et al., 2017). Also, the blight rate will increase if the temperature and precipitation are above a certain level and the cultivation land will head north to the central region (Yoon et al., 2014). Therefore, it is important to monitor vegetation growth of chinese cabbage with climate factors (Kim et al., 2015).

    Regression models based on time-series data have been developed with climate factors and yield of major crops (Choi & beak, 2016). Applying remote sensing (RS) technology with the models can make it possible to non-destructively spatially quantify (Kang et al., 2017). RS is being studied in the agricultural field to monitor the vegetation growth by using an image- based sensor. There are hyperspectral, multispectral and thermal sensors in the image-based sensors for monitoring the structured attributes with spectral signatures of crop canopies (Zarco-Tejada et al., 2013). The attributes remotely monitored by sensors are influenced by differences in leaf shape and size, vegetative density, other climate factors, the availability of nutrients, moisture contents and soil properties (Yang et al., 2008). Particularly, there is a high potential for detecting or predicting vegetation growth in hyperspectral imagery named the high-dimensional data (Kang et al., 2018). It was tried to predict sweetness and amino acid in soybean crops from the hyperspectral sensor mounted on a crane (Monteiro et al., 2007), yields in crops using an aircraft (Luo et al., 2012) and rice protein content on the ground-base (Onoyama et al., 2011). As such, the sensors have been used with various platforms such as ground-based (tractor and crane) and air-based (unmanned aerial vehicles, aircraft, and satellite) to effectively monitor wide areas.

    The prediction models using high-dimensional data are developed by various regression analyses. Multivariate regression analysis allows extracting information in the ability to predict vegetation growth from the full-spectral bands (Dinç et al., 2008). However, not all bands are essential for predicting vegetation growth due to a strong correlation between spectral bands (Medjahed et al., 2016). The irrelevant bands might interrupt to develop prediction models and to interpret attributes of crop canopies. Therefore, it is essential to select the key spectral bands predictable for vegetation growth (Kang et al., 2018b). Stepwise discriminant analysis of the subset selection method provides the best linear spectral combination for assessing vegetation growth (Huang & Townshend, 2003). It can be extremely useful for providing data interpretation and small broadband multispectral imaging sensor development (ElMasry et al., 2008). The sensors specialized in predicting the desired crop parameters (yield, moisture, and nutrients) will be widely available over-cultivation periods to facilitate decision making on cultivation management with various platforms (Jin et al., 2018).

    The objectives of this study were to develop models for predicting yields in different cultivars (joined-up in 2015 and wrapped-up in 2016) of chinese cabbage using hyperspectral imagery. Key band ratios were selected by stepwise analysis between bands merged with 10 nm, 25 nm, and 50 nm of full width at half maximum (FWHM) in remotely sensed canopies by hyperspectral imagery with 5 nm of FWHM. The yield prediction models were developed with key band ratios for each FWHM and Effective Accumulated Temperature (EAT) over-cultivation periods.

    Materials and Methods

    1. Study area

    Test fields located at Haeje-myeon (35°4′11.28″ N, 126°18′ 21.25″ E, about 64 m above sea level) and Hyeongyeong-myeon (35°4′11.28″ N, 126°18′21.25″ E, about 64 m above sea level), Muan-gun, Jeollanam-do, Republic of Korea. The test field has loamy alkaline soils containing gravel. Chinese cabbage (Wonjong in 2015 and Whistle in 2016, Brassica rapa var. glabra Regel) was sown in early August and planted in early September. The yield of chinese cabbage measured as fresh weight of all body above the ground regardless of the cultivation period in the same way as when supplied in the market just after the image acquiring. The measurement area was Bioenergy Crop Research Institute, National Institute of Crop Science, Rural Development Administration (34°58'08.9" N, 126°27' 18.4" E) and The number of samples was a total of 18 except for 2 that were wrong measured 5 by each experimental day in 2015 and was 30 in total, 10 for each experimental day.

    2. Hyperspectral image acquisition and processing

    The hyperspectral image sensor (PS, Specim, Finland) consists of 519 spectral bands with 5.5 nm of FWHM from 400 nm to 1000 nm. The spatial pixels were 656 pixels with the condition of a half resolution. The sensor has also a 23mm focal length objective lens that provides a 22.1° field of view. The images were acquired at a height of about 2 m above ground with an 18% white reference board (Ezybalance, Lastolite Ltd., England). The images were acquired on October 7th, October 21st, November 4th, and November 20th in 2015, and on October 27th, November 11th, and November 25th in 2016.

    Normalized difference vegetation index (NDVI) used to extract the vegetation area of canopies for chinese cabbages based on ostu’s threshold method using the image processing software (ENVI 4.7, Exelis Visual Information Solutions Inc., USA). The NDVI was calculated as Eq. 1.

    NDVI = ρ NIR ρ Red ρ NIR + ρ Red

    where ρNIR and ρred are 820 nm and 620 nm, respectively. Fig. 1 shows the selected areas of individual chinese cabbage extracted with NDVI. After image processing, the hyperspectral reflectance was merged as 10 nm, 25 nm and 50 nm of FWHM to find out the optimal combination of spectral bands.

    3. Stepwise Multiple Linear Regression

    Stepwise multiple linear regression(MLR) is the method to develop a model with key independent variables based on backward eliminating irrelevant ones and forward selecting significant ones. First, the forward selection is applied to select the significant ones from sequential independent variables entered in the model based on MLR using all independent variables. This step leads to the greatest improvement in the performance of the model. And then in the backward elimination, the insignificant independent variables can be dropped from the model without a significant decrease in the performance of the model. Finally, the model comprises the set of independent variables that best explain the response (Mundry & Nunn, 2008).

    4. Effective Accumulated Temperature

    Effective accumulated temperature (EAT) is one of the important factors for vegetation growth. The completion of growth stages or the entire generation requires the optimal EAT (Ma et al., 2008). Therefore, it is necessary to monitor the EAT for predicting the status of crop growth. The EAT was calculated as follows Eq. 2.

    EAT ( o C ) = i = 1 k 0.5 × ( H k +L k ) LT

    where Hk is the maximum temperature and Lk is the minimum temperature in k days. LT is the development of zero points of certain crops (Bunting, 1976). The climate data such as temperature and precipitation was collected at a weather station in Mokpo, Jeollanam-do, about 18 km from the test field (34°82'81.78" N, 126°38'33.43" E).

    5. Modeling performance analysis

    The reflectance can be unstable due to differences in the radiation of multi-temporal imagery even if the reflectance was normalized by the white reference board. Therefore, the MLR model for predicting the yields of chinese cabbage was developed based on the band ratios with the proportion between each reflectance of front and rear bands sequentially from single bands of 400 to 1,000 nm. The band ratios are useful to minimize differences in sunlight angle or intensity (Mather & Koch, 2011). The band ratios were calculated as follows Eq. 3.

    Band ratio =  R front band R back band eq : R 400 /R 402 , R 402 /R 404 R 998 /R 1000

    The key band ratios (independent variables) for each model of cultivar and all cultivars (joined-up & wrapped-up) were selected by stepwise MLR. The EAT was added to the models as an independent variable and the possibility of improvement in models was evaluated with R², root mean squares error (RMSE), and relative error (RE).

    Results and Discussion

    1. Effective accumulated temperature and yields of chinese cabbage

    Fig. 2 shows the EAT and accumulated precipitation depending on the growth duration after planting (GDP) for chinese cabbage in each year. Although there was a difference in EAT between 2015 and 2016 was 55℃ as the maximum in the middle of vegetation growth stages, the EAT during all vegetation growth stages was similar to each other between two years. However, the accumulated rainfall was almost twice in 2016 as 460 mm compared with it in 2015 as 233mm. Therefore, there was a difference in the average yield of autumn chinese cabbage between 2015 and 2016 as 11.3 ton/ha and 9.9 ton/ha, respectively (KOSIS, 2019)

    The yields of a joined-up cultivar in 2015 and a wrapped-up cultivar in 2016 are represented depending on GDP (Fig. 3). The yields of wrapped-up cultivar were lower than those of joinedup cultivar due to the attribute of cultivar with small inner leaves compared with outer leaves (Yun et al., 2012). The difference in yield between joined-up and wrapped-up cultivars was almost twice and it might be difficult to make a good model using the yield data of wrapper-up cultivar because the variation of yield in 2016 was lower compared with it of a joined-up cultivar in 2015.

    Fig. 4 shows the correlations between yields and EAT of chinese cabbage depending on the cultivars. Each graph has linearity but the slopes are completely different. Generally, the differences in the slopes or intercepts may be influenced by the environmental factors during the cultivation periods (Westcott, 1986). However, the difference in the slopes in Fig. 4 might be influenced by the cultivar because the EAT in 2016 was better than it in 2015 as shown in Fig. 3. It might also give an effect on the prediction model based on hyperspectral imagery.

    2. Reflectance and ratio curves

    The reflectance and band ratio from 410 nm to 990 nm in the individual areas were represented in Fig. 5. The reflectance between 745 and 752 nm was removed due to the noise caused by equipment defect. However, the overall reflectance curves excluding the noising spectral ranges generally reveal the spectral attributes of the vegetation. Because of the degeneration at outer-leaves, the reflectance in 2015 and 2016 decreased gradually in the NIR region but increased in the visible region, except for the reflectance of 75 GDP in 2015. Even though 75 GDP days were just before harvest, the reflectance showed the characteristics of the healthy leaf with higher reflectance in the NIR region and lower reflectance in the visible region as shown in Fig. 5A and 5B (Haboudane et al., 2004). Because of the unusual reflectance in 75 GDP days in 2015, the reflectance of all GDP days was converted into the band ratios to minimize the unstable reflectance as shown in Fig. 5C and Fig 5D. The difference in band ratios depending on wavelength shown in both Fig.s C and D, but the difference in band ratios depending on GDP is shown only in Fig. D. Moreover, the movement of peaks for the band ratio between red and NIR regions was different depending on the varieties. In the red-edge region, the peaks of the band ratio increased sharply and then decreased in the joined-up variety but it decreased firstly and then severely increased in the wrapped-up variety. The pattern of peaks for the band ratio in the NIR region was also different depending on the varieties. Therefore, it might be possible to identify the varieties of chinese cabbage using the peak characteristics of band ratios.

    3. Prediction model without effective accumulated temperature

    Table 3 shows the key bands ratios selected by the stepwise MLR method depending on FWHM to predict the yields of chinese cabbage regardless of cultivars. The selected five bands ratios for 5 nm and 25 nm of FWHM were in the green, red, red edge and NIR regions, and those for 10 nm and 50 nm of FWHM were in the red and NIR regions. The band ratios selected by a combination of various band ratios existed in a wide range, except the blue region. When the combination of selected bands was considered based on fewer adjacent bands, the band ratios with 25nm of FWHM might be reasonable to predict the yields of chinese cabbage. The other combinations of band ratios had two or three adjacent combinations of band ratios, such as 820 nm, 850 nm, 860 nm for 10 nm of FWHM and 630 nm, 650 nm, 800 nm, 810 nm for 50 nm of FWHM.

    Table 4 shows the performance of stepwise MLR models without EAT using key bands ratios selected by 5 nm, 10 nm, 25 nm, and 50 nm of FWHM to predict the yields of chinese cabbage for each cultivar and all cultivars. The performances of stepwise models with 5 nm of FWHM were the highest for each cultivar and all cultivars. In all cultivars, the performances of models were reduced from 0.87 to 0.77 in R², 558 g/plant to 733 g/plant in RMSE and 22.8% to 29.9% in RE as the FWHM broadened. The performance of each model with 5 nm of FWHM for joined-up chinese cabbage in 2015 was the best as R² = 0.91, RMSE = 772 g/plant, RE = 26.9%, respectively. The performance of models with remaining FWHM decreased as R²= 0.85, RMSE = 1010 g/plant and RE = 35.1%. In the case of the performance of a model for wrapped-up chinese cabbage in 2016 shows the lower linearity as 0.61 in R² compared with it in 2015.

    Fig. 6 shows the yield prediction models for each cultivar and all cultivars depending on FWHM without EAT. The sensitivity of models was comparatively different in the mediate yields between 5 nm of FWHM and others not only for each cultivar but also for all cultivars. A, C, E, and G in Fig. 6 show that the slopes of yield prediction models between 2015 and 2016 were different depending on the cultivars as shown in Fig. 4. There was lower linearity of the yield prediction model in 2016 compared with it in 2015, because the variation of yield was comparatively lower as 1000 to 4000 g/plant than as 500 to 7000 g/plant in 2015. Therefore, it may be advantageous to develop yield prediction models with all data as shown in B, D, F, and H in Fig. 6. Also, the errors of models developed with all cultivars were lower than those in 2015 as below 30% in RE with the addition of the 2016 samples as shown in Table 4. It might be not necessary to classify chinese cabbage depending on cultivars because it is possible to predict the yields of chinese cabbage regardless of the cultivars.

    4. Prediction model with effective accumulated temperature

    Table 5 shows the performances of yield prediction model developed with EAT and key band ratios depending on FWHM in each cultivar and all cultivars for chinese cabbage. In all cultivars, the performances were improved slightly as 0.02 in R², 15 g/plant in RMSE and 0.6% in RE. The performances of the models with 10 nm, 25 nm, and 50 nm of FWHM decreased to 0.82 in R², 671 g/plant in RMSE and 27.4% in RE, compared to those with 5 nm of FWHM. However the performances of models with EAT are better than those without EAT because the added EAT, lowered errors of the model and leading to higher predictability of yields. In each cultivar, the models were also improved slightly in 2015 and 2016, but the performance of models in 2016 still is difficult to predict the yields because of the lower linearity as below 0.61 in R².

    In another study, physical data such as vegetation fraction and plant height calculated by unmanned aerial vehicle-based RGB imagery was used to predict yield and biomass of chinese cabbage without spectral data (Kim et al., 2018). Only using physical data may threaten the predicted performance if the cultivation type and cultivation change, but it can be insensitive to different illumination effects depending on time series that is problematic for spectral data normality. In the future, it is needed to develop a yield prediction model adding the physical factors of chinese cabbage along with the climatic factors and evaluate performance.

    Fig. 6 shows the yield prediction models with EAT for each cultivar and all cultivars depending on FWHM. Although it was difficult to visually identify the differences in the performance of models between A, C, E, and G in Fig. 6 and those in Fig. 7, the sensitivity was improved at the mediate yield, such as between 200 g/plant and 2100 g/plant, in the model with 25 nm of FWHM. It means that other environmental factors have to include independent variables to minimize the differences and improve the performance of the models.

    As a result, the models with five key band ratios depending each FWHMs were possible to predict yields of joined-up and wrapped-up chinese cabbage even if the linearity of models for wrapped-up chinese cabbage was low because of comparatively lower variation. Therefore, it is advantageous to develop the yield prediction model mainly with joined-up chinese cabbage as the high yield cultivar. Also, it is possible to improve the performance of models when the environmental factors, such as EAT, are added in the models as the independent variables. The models for all FWHMs with EAT were a possibility to predict yields of both joined-up and wrapped-up chinese cabbages.

    In this paper, yields of chinese cabbage in 2015 and 2016 of each different cultivars were predicted by developing an MLR model with key band ratios for each FWHM selected by stepwise analysis. The number of the key band ratios for each FWHM in all cultivars were selected in a combination with various band ratios in a wide wavelength without blue, respectively. When models were developed with the key band ratios for each FWHM in all cultivars, they were possible to predict yields of joined-up and wrapped-up chinese cabbage as R2≥0.77, RMSE≤733 g/plant and RE≤29.9%. However, in each cultivar, the models of wrapped-up chinese cabbage for each FWHMs in 2016 were difficult to predict it with low linearity below 0.61 in R2 due to low yield cultivar attribute. Therefore, it is advantageous to develop a yield prediction model with joined-up chinese cabbage of high yield cultivar (all cultivars).

    Adding EAT in the models in all cultivars and each cultivar can improve performance slightly and the sensitivity of predicted values on measured yields. In all cultivars, the performances for each FWHM were R2≥0.80, RMSE≤694 g/plant and RE≤ 28.3% in all cultivars. Like previous results, the models of wrapped-up chinese cabbage in 2016 were still low linearity below 0.61 in R2. Despite a slight improvement in performance, adding EAT in the models may be an important factor for predicting non-destructively yields of chinese cabbage in other years (Ryu et al., 2011). Therefore, it will be important to cross-validate this model with the spectral, environment and yield data of all cultivar collected in other years or situation.

    As a result, the models for all FWHMs with each five band ratios in the combination of various band ratios in a wide wavelength without blue were the possibility to predict yields of chinese cabbage, which it should always be collected with samples of joined- up chinese cabbage of high yield cultivar and environment factors over the cultivation periods. If using the collected data to continually verify key band ratios along with model performance, it will enable the development of cheaper and small multispectral image sensors specialized in yield prediction of chinese cabbage.


    This work was performed with the support of ‘Advanced Production Technology Development Project’ (Project title: Development of a UAV-based remote sensing technology for monitoring the growth of major upland crops, Project No.: 315011- 03-3-HD020), IPET (Korea Institute of Planning and Evaluation for Technology of Food, Agriculture, Forestry and Fisheries), MAFRA (Ministry of Agriculture, Food and Rural Affairs).



    Individual canopy areas extracted by image processing.


    Effective accumulated temperature over growth duration after planting in 2015 and 2016.


    Yields of joined-up (A) and wrapped-up (B) chinese cabbage.


    Relationship between effective accumulated temperature and yields of chinese cabbage.


    Reflectance and ratio curves in joined-up (A), (C) and wrapped-up chinese cabbage (B), (D) depending on growth duration after planting.


    Yield prediction models without effective accumulated temperature depending on FWHM for each cultivar (A, C, E, and G) and all cultivars (B, D, F, and H).


    Yield prediction models with effective accumulated temperature depending on FWHM for each cultivar (A, C, E, and G) and all cultivars (B, D, F, and H).


    Cultivation information of chinese cabbage

    Specification of hyperspectral image sensor and acquisition dates

    Key band ratios selected to predict yields of chinese cabbage with each and all cultivars

    Yield prediction model for each FWHM without effective accumulated temperature

    Yield prediction model for on each FWHM with effective accumulated temperature


    1. Bunting ES. 1976. Accumulated temperature and maize development in England. The J. of Agricultural Science 87: 577-583.
    2. Choi S and Baek J. 2016. Onion yield estimation using spatial panel regression model. Korean J. of Applied Statistics 29: 873-885.
    3. Dinç E , Baleanu D , Ioele G , De Luca M and Ragno G. 2008. Multivariate analysis of paracetamol, propiphenazone, caffeine and thiamine in quaternary mixtures by PCR, PLS and ANN calibrations applied on wavelet transform data. J. of Pharmaceutical and Biomedical Analysis 48: 1471-1475.
    4. ElMasry G , Wang N , Vigneault C , Qiao J and ElSayed A. 2008. Early detection of apple bruises on different background colors using hyperspectral imaging. LWT-Food Science and Technology 41: 337-345.
    5. Haboudane D , Miller JR , Pattey E , Zarco-Tejada PJ and Strachan IB. 2004. Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: Modeling and validation in the context of precision agriculture. Remote Sensing of Environment 90: 337-352.
    6. Huang C and Townshend JRG. 2003. A stepwise regression tree for nonlinear approximation: applications to estimating subpixel land cover. International J. of Remote Sensing 24: 75-90.
    7. Jin X , Kumar L , Li Z , Feng H , Xu X , Yang G and Wang J. 2018. A review of data assimilation of remote sensing and crop models. European J. of Agronomy 92: 141-152.
    8. Kang YS , Kim SH , Kang JG , Sarkar TK , Kwon YS , Jun SR , Kim WJ and Ryu CS. 2017. Model assessment multi-temporal monitoring of Chinese cabbage growth using low altitude remote sensing system. J. of Agriculture & Life Science 51: 149-160.
    9. Kang YS , Ryu CS , Jun SR , Jang SH , Park JW , Song HY , Sarkar TK , Kim SH and Lee WS. 2018a. Distinguishing between closely related species of Allium and of Brassicaceae by narrowband hyperspectral imagery. Biosystems Engineering 176: 103-113.
    10. Kang YS , Ryu CS , Kim SH , Jun SR , Jang SH , Park JW , Sarkar TK and Song HY. 2018b. Yield prediction of Chinese cabbage (Brassicaceae) using broadband multispectral imagery mounted unmanned aerial system in the air and narrowband hyperspectral imagery on the ground. J. of Biosystems Engineering 43: 138-147.
    11. Kim DW , Yun HS , Jeong SJ , Kwon YS , Kim SG , Lee WS and Kim HJ. 2018. Modeling and testing of growth status for Chinese cabbage and white radish with UAV-based RGB imagery. Remote Sensing 10: 563.
    12. Kim YS , Shim KM , Jung MP and Choi IT. 2015. Study on meteorological factors affecting estimation of Chinese cabbage yield. J. of Climate Change Research 6: 277-281.
    13. KOSIS.2019. Vegetable yield(Leafy vegetables in Korea) (2019.12.23.).
    14. Luo B , Yang C , Chanussot J and Zhang L. 2012. Crop yield estimation based on unsupervised linear unmixing of multidate hyperspectral imagery. IEEE Transactions on Geoscience and Remote Sensing 51: 162-173.
    15. Ma LQ , Gao SJ , Wen JB , Zong SX and Xu ZC. 2008. Effective accumulated temperature and developmental threshold temperature for Semanotus bifasciatus (Motschulsky) in Beijing. Forestry Studies in China 10: 125-129.
    16. Mather PM and Koch M. 2011. John Wiley & Sons. Computer processing of remotely-sensed images: an introduction.
    17. Medjahed SA , Saadi TA , Benyettou A and Ouali M. 2016. A new post-classification and band selection frameworks for hyperspectral image classification. The Egyptian J. of Remote Sensing and Space Science 19: 163-173.
    18. Monteiro ST , Minekawa Y , Kosugi Y , Akazawa T and Oda K. 2007. Prediction of sweetness and amino acid content in soybean crops from hyperspectral imagery. ISPRS J. of Photogrammetry and Remote Sensing 62: 2-12.
    19. Mundry R and Nunn CL. 2008. Stepwise model fitting and statistical inference: turning noise into signal pollution. The American Naturalist 173: 119-123.
    20. Onoyama H , Ryu C , Suguri M and Iida M. 2011. Estimation of rice protein content using ground-based hyperspectral remote sensing. Engineering in Agriculture, Environment and Food 4: 71-76.
    21. Ryu C , Suguri M and Umeda M. 2011. Multivariate analysis of nitrogen content for rice at the heading stage using reflectance of airborne hyperspectral remote sensing. Field Crops Research 122: 214-224.
    22. Song EY , Moon KH , Wi SH , Kim CH , Lim CK , Oh S and Son IC. 2017. Impact of elevated temperature in growing season on growth and bulb development of extremely early-maturing Onion (Allium cepa L. cv. Singsingball). Korean J. of Agricultural and Forest Meteorology 19: 223-231.
    23. Westcott B. 1986. Some methods of analysing genotype—environment interaction. Heredity 56:243-253.
    24. Yang Z , Willis P and Mueller R. 2008. Impact of band-ratio enhanced AWIFS image to crop classification accuracy. In Proc. Pecora (Vol. 17, pp. 18-20).
    25. Yoon DH , Oh SY , Nam KW , Eom KC and Jung PK. 2014. Changes of cultivation areas and major disease for spicy vegetables by the change of meteorological factors. Climate Change Res. 4: 47-59.
    26. Yun HB , Lee JS , Lee YJ , Kim MS and Lee YB. 2012. Effect of different colored polyethylene mulch on the change of soil temperature and yield of chinese cabbage in autumn season. Korean J. of Soil Science and Fertilizer 45: 511-514.
    27. Zarco-Tejada PJ , Guillén-Climent ML , Hernández-Clemente R , Catalina A , González MR and Martín P. 2013. Estimating leaf carotenoid content in vineyards using high resolution hyperspectral imagery acquired from an unmanned aerial vehicle (UAV). Agricultural and Forest Meteorology 171: 281-294.
    오늘하루 팝업창 안보기 닫기