search for




 

Image Based-Phenotyping and Selection Index Based on Multivariate Analysis for Rice Hydroponic Screening under Drought Stress
Plant Breed. Biotech. 2021;9:272-286
Published online December 1, 2021
© 2021 Korean Society of Breeding Science.

Adinda Asri Laraswati1, Rusnadi Padjung2, Muh Farid2*, Nasaruddin Nasaruddin2, Muhammad Fuad Anshori2, Amin Nur3, Andi Isti Sakinah1

1Agricultural Systems Study Program, Graduate School Hasanuddin University, Makassar 90245, Indonesia
2Department of Agronomy, Faculty of Agriculture, Hasanuddin University, Makassar 90245, Indonesia
3Assessment Institute for Agriculture Technology of Gorontalo, Gorontalo 96583, Indonesia
Corresponding author: Muh Farid, farid_deni@yahoo.co.id, Tel: +62-813-5504-1712, Fax: +62-813-5504-1712
Received May 4, 2021; Revised August 21, 2021; Accepted October 12, 2021.
This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
The development of rice varieties that are tolerant of drought stress needs to be detected with Image-based phenotyping. This Image-based phenotyping method in combination with selection index and multivariate analysis can characterize the morphological response easily within a short time, which makes it is suitable for rice screening under drought stress. Therefore, this study aims to determine the selection index based on multivariate analysis and assess the effectiveness of using image-based phenotyping in drought rice screening. This study was conducted in two stages, the first was in static hydroponic and the second was in dynamic hydroponic. In static hydroponic, a split-plot design was used, where the levels of drought were the main plots and varieties were the subplots. However, in dynamic hydroponic, a nested design was used, where the replicates were nested in the drought level treatments. Also, The drought level factors used were PEG 0%, PEG 10%, and PEG 20%, and the variety factor consisted of 5 varieties which were repeated three times. The results showed that the selection index for static hydroponic consisted of shoot area (0.421), green shoot area (0.4177), and the area growth rate (0.4192). Meanwhile, the selection index in dynamic hydroponics consisted of object extent Y from the side (0.4516) and convex hull from the side (0.4177). The regression of the two-selection index has a good determination of 0.84. Hence, these results showed that rice screening based on image-based phenotyping can be recommended for rapid screening under drought stress.
Keywords : Drought stress, Image-based phenotyping, Multivariate analysis, RGB, Rice
INTRODUCTION

Rice is a commodity that plays an important role in Indonesia's economy and is generally consumed by approximately 90% of the population as a staple food (Donggulo et al. 2017). In Indonesia, rice production in 2020 has decreased from 59.2 million tons in the previous year to 54.6 million tons (Iswara et al. 2021). This decline in production was influenced by several factors, which includes drought stress. According to Bouman and Tuong (2001), approximately 80% of the cultivation area of rice and other cereals is affected by drought as a limiting factor for production. In addition, several studies have shown that drought reduces the rate of growth and rice yield (Supriyanto 2013; Sujinah and Jamil 2016; Rusmawan et al. 2018). Therefore, the development of drought-tolerant varieties becomes a solution for increase rice production.

The development of drought-tolerant varieties requires an effective and efficient screening process, which is carried out artificially or directly in the targeted environment. Moreover, the environment for artificial selection is considered easier to restrain, especially the control of stress concentration levels, which makes this method effective and efficient in the increase of the screening process (Anshori et al. 2019). In drought stress, artificial screening is carried out using a hydroponic concept with a polyethylene glycol (PEG) selection agent (Swapna and Shylaraj 2017; Mangansige et al. 2018; Osmolovskaya et al. 2018). A previous study on the use of this method under drought stress conducted by Ai et al. (2010), Wening and Susanto (2017), and Farid et al. (2021) stated that the use of PEG 6000 with certain concentrations in the early (vegetative) phase can function as an alternative for quick or early selection. This concept is very important in increasing the effectiveness of selection, however, the vegetative screening under salinity stress has to validate or correlate with the reproductive phase, such as the dynamic hydroponic. Meanwhile, this hydroponic makes it possible to measure the tolerance screening until the reproductive phase with a controlled environment is achieved (Farid et al. 2021). Therefore, the use of PEG in static and dynamic hydroponic systems can be a good alternative in the screening process of drought tolerance of rice, however, it requires an accurate and precise assessment approach.

Assessment of drought stress tolerance through screening is a common method that is often considered inaccurate. Based on the genetic constitution, drought tolerance nature is polygenic and is indicated by the QTL report that encodes these characteristics (Lanceras et al. 2004), therefore, qualitative screening is not involved in the assessment of drought. Observations based on morphology are generally considered to be highly biased (Brereton et al. 2015), while physiological and molecular are considered effective and have a relatively low bias, but these approaches are expensive (Nuraida 2012). Therefore, an accurate and precise approach with a relatively low cost is needed in identifying the nature of rice tolerance to drought stress. One method that is considered effective is a digital technology approach through image-based phenotyping.

Image-based phenotyping technology is currently being developed to characterize morphological and physiological responses of a plant using digital images. This technology makes it easy to calculate phenotypes by analyzing plants through images accurately in large numbers within a short time (Das Choudhury et al. 2018). Several studies have reported the use of an effective and efficient image-based phenotyping selection method in rice (Hairmansis et al. 2014; Siddiqui et al. 2014; Duan et al. 2018; Guimaraes et al. 2020; Kim et al. 2020; Vishal et al. 2020; Anshori et al. 2021a) and maize (Asaari et al. 2019; Li et al. 2021). However, there are still few studies that examined the development of image-based phenotyping based on selection index and multivariate analysis on the static hydroponic systems.

A selection index is a breeding approach that combines several criteria in one formula (Anshori et al. 2021b). This approach is commonly used in plant breeding because it considers all selection criteria with current economic value on each criterion in genotype selection (Rajamani et al. 2016; Islam et al. 2017). The selection index often combines with multivariate analysis to increase its effectiveness (Anshori et al. 2019; Fadhli et al. 2020; Akbar et al. 2021; Anshori et al. 2021b; Farid et al. 2021). This multivariate analysis is used to analyze or integrate the variance of large variable data into a simpler and more comprehensive manner for the results to be easily understood (Mattjik and Sumertajaya 2011). Furthermore, several studies have reported the effectiveness of selection with the use of multivariate analysis (Anshori et al. 2018; Kose et al. 2018; Anshori et al. 2019; Fadhli et al. 2020; Farid et al. 2020; Akbar et al. 2021; Anshori et al. 2021; Farid et al. 2021). Based on these analyzes, the use of image-based phenotyping with the selection index based on multivariate analysis in static and dynamic hydroponic screening is necessary. Therefore, this study aims to determine the selection index based on multivariate analysis and assess the effectiveness of using image-based phenotyping in drought rice screening.

MATERIALS AND METHODS

The two experiments conducted in this study include the destructive test, which was carried out on static hydroponic (floating raft system), and the second was the non-destructive test carried out on dynamic hydroponic (deep flow technique system). These studies were carried out simultaneously in the greenhouse, Unhas Lecturer Residence, Makassar City, South Sulawesi Province, Indonesia (22.4 m above sea level [asl]) from August to November 2020.

Experimental design and procedures

The static hydroponic study used a split-plot design with the levels of drought as the main plot and varieties as the subplots (Fig. 1a).

Figure 1. Images of static hydroponic system at 20 DAS (a) and dinamic hydroponic system at 33 DAS (b).

However, the dynamic hydroponic used a nested design, where the replicates were nested in the drought level treatments (Fig. 1b) and the drought level factors were PEG 0%, PEG 10%, and PEG 20%. Furthermore, the varieties used were Inpari 34, IR 20 (drought-sensitive check), Salumpikit (drought-resistant check), Ciherang, and Jeliteng.

In this study, the medium used was ABmix with a concentration of 8 mL/L, and the PEG concentration treatments were gradually given to the hydroponic nutrient solution. The first stage was given 1/2 concentration of PEG treatment at the age of 13 DAS, while the second stage was given at the age of 16 DAS to prevent osmotic shock. The pH control was applied to keep it constant at the desired value, which from range 5.8 to 6.2. Meanwhile, the pH control was carried out by adding HCl or NaOH to lower or raise the pH. Data collection on the static hydroponic method was carried out 2 weeks after the treatment application or 30 DAS, while dynamic hydroponics data were collected at 68 DAS.

The observations for static hydroponic included shoot and root fresh weight (actual measurements), 2nd, 3rd, and 4th leaf length, shoot and green shoot areas from the side view, ratio of green shoot to shoot area from the side view, red, green, blue, ratio of red to green, and area growth rate. The traits measured in dynamic hydroponic included shoot and root fresh weight (actual measurements), convex hull from the top view, shoot area from the top view, green shoot area from the top view, the ratio of green shoot area to shoot area from the top view, number of leaves from the top view, red, green, blue, the ratio of red to green, the ratio of shoot area to the convex hull from the top view, object extent X from a side view, object extent Y from a side view, shoot area from a side view, green shoot area from a side view, a ratio of green shoot area to the shooting area from a side view, convex hull from a side view, and the ratio of shoot area to convex hull area from a side view (Table 1).

Table 1 . Phenotyping characters used to analyze growth and drought-related traits.

CharactersDescription
Shoot fresh weight (SFW)The weight of the fresh shoot at the end of observation.
Root fresh weight (RFW)The weight of the fresh root at the end of observation.
2nd leaf length (LL2)The length of the 2nd leaf from the tip of the plant.
3rd leaf length (LL3)The length of the 3rd leaf from the tip of the plant.
4th leaf length (LL4)The length of the 4thleaf from the tip of the plant.
Shoot area (SA)The number of pixels scaled in cm on the shoot of the plant.
Green shoot area (GSA)The number of pixels that have been scaled in cm in the shoot section has a hue value of 50.
Convex hull area (CT for top and CS for side)The smallest area is covered by theouter contour of an object.
Object extend X (XT for top and XS for side)The length of the X-axis of the rectangle covering the object is used to measure the width of the plant.
Object extend Y (YT for top and YS for side)The length of the Y-axis of the rectangle covering the object is used to measure the height of the plant.
Number of leaves (NL)The numberof leaf tips visible from above the plant.
Area growth rate (AGR)Dividing the difference in shoot area between the drought stress treatment intervals by drought stress treatment time (days).
Red (R)Color is measured by averaging the Red color value of all pixels in the shoot. RGB values are between 0 (no color) and 255 (maximum color).
Green (G)Color is measured by averaging the Green color value of all pixels in the shoot. RGB values are between 0 (no color) and 255 (maximum color).
Blue (B)Color is measured by averaging the Blue color value of all pixels in the shoot. RGB values are between 0 (no color) and 255 (maximum color).


Image analysis

The images were taken with a Canon EOS 1200D RGB camera in a portable photo studio with a 75 cm × 75 cm × 75 cm. A white background with two 8 watt white LED lights in the studio with camera settings (5.6 F-stops, 1/160 seconds exposure time, ISO 800, and without flash) was used. The images were taken from a top view (0°) and the side view (90°) of the plant. For static hydroponic, the images were taken twice at 13 DAS (early treatment) and 30 DAS (after treatment) as shown in Fig. 2. For dynamic hydroponic, images were taken at 13 DAS (early treatment) and 68 DAS (after treatment) (Fig. 3 and the images results were analyzed using the Fiji application. Furthermore, the plant color was analyzed using hue channels with a hue 50 (green area) and the area growth rate was calculated using the projection of shoot area over two time periods.

Figure 2. RGB images of IR 20 and Salumpikit under drought stress at 30 DAS in static hydroponic.
Figure 3. RGB images of IR 20 and Salumpikit under drought stress at 68 DAS in dynamic hydroponic.

Data analysis

The data were analyzed independently for variance and the characters that have significant interactions with each other were further analyzed using the Pearson correlation test and Principal Component Analysis (PCA) with STAR 2.01 software. After static hydroponic has been tested by correlation analysis, it was followed by path analysis. All further analysis based on the stress tolerance index (STI) value of each character on individual PEG concentration (0% PEG (normal) –10% PEG (drought stress) = STI 1 and 0% PEG (normal)–20% PEG (drought stress) = STI 2). This concept was reported by Anshori et al. 2019, as follows:

Stress Tolerance Index (STI) is calculated by the equation (Fernandez 1992):

STI=Yp×YsY¯p2

Note: Yp = The character value of each variety in normal / non-stressed conditions.

Ys = The character value of each variety in a stressed condition.

Ȳp = Average character values of all varieties in normal / non-stressed conditions.

The selected characters from the PCA analysis results were used to obtain the index value. The validation of static hydroponic was conducted by the regression test toward the dynamic hydroponic.

RESULTS

Static hydroponic system

In the static hydroponic method, the analysis of variance showed that the diversity of PEG concentrations had a significant effect on almost all the characters, except for the green, ratio of red to green, and the ratio of green shoot area to shoot area (Table 2).

Table 2 . Means square value of analysis of variance on the static hydroponic method.

CharactersDVD × V
SFW178.73**28.92**10.37**
RFW24.77**7.3**1.96**
LL21,486.95**326.19**14.68ns
LL31,471.64**466.95**49.37*
LL4669.49**188.37**28.07ns
SA65,556.88**13,117.85**3,438.49**
GSA61,237.72**9,849.18**3,261.87**
R3,460.07**86.95ns748.95ns
G879.98Ns87.35ns1,498.13ns
B1,680.60*96.75ns272.54ns
RRG0.6Ns0.8ns0.96ns
RGSS0.57Ns0.02ns0.03ns
AGR202.25**32.11**10.6**

D: Drought level, V: Variety, *: Significant effect at P ≤ 0.05, **: Significant effect at P ≤ 0.01, ns: Not significant, SFW: Shoot freshweight, RFW: Root fresh weight, LL2: 2nd leaf length, LL3: 3rd leaf length, LL4: 4th leaf length, SA: Shoot area, GSA: Green shoot area, R: Red, G: Green, B: Blue, RRG: Ratio of red to green, RGSS: Ratio of green shoot area to shoot area, AGR: Area growth rate.



The genotype variance analysis also showed a significant effect on the shoot and root fresh weight, 2nd, 3rd, and 4th leaf length, shoot area, shoot green area, and area growth rate. Meanwhile, the characters that were affected significantly by drought-genotype interactions included shoot and root fresh weight, 3rd leaf length, shoot area, green shoot area, and area growth rate. All characters that significantly affected the interaction effect were continued by deep analysis on the static hydroponic system. Meanwhile, the phenotype varieties on these characters are shown in Supplementary 1.

The correlation analysis on static hydroponic results was based on the static hydroponic Stress Tolerance Index (STI) value (Table 3). The results showed that the 3rd leaf length (0.769 and 0.745), shoot area (0.981 and 0.983), green shoot area (0.978 and 0.984), and area growth rate (0.981 and 0.961) had a significant correlation on the shoot and root fresh weights, respectively. These correlation analyzes were followed by independent path analysis on the shoot and root fresh weights, which were the main characters of actual morphology under artificial vegetative screening with different roles. Therefore, this independent concept analysis can determine the specific relationship among image-based phenotyping characters to the conventional morphology characters.

Table 3 . Correlation analysis of STI values on the static hydroponic method.

RFWLL3SAGSAAGRSFW
RFW111111
LL30.745**0.789**0.991**0.977**0.981**
SA0.983**0.748**0.984**0.978**
GSA0.984**0.802**0.981**
AGR0.963**0.769**
SFW0.982**

The numeric in table indicate the correlation value. *: Significant correlated at P ≤ 0.05, **: Significant correlated at P ≤ 0.01, SFW: Shoot freshweight, RFW: Root fresh weight, LL3: 3rd leaf length, SA: Shoot area, GSA: Green shoot area, AGR: Area growth rate.



Path analysis on the STI value of shoot fresh weight in static hydroponic showed representative results with a determinant coefficient value of 0.86 (Table 4).

Table 4 . Path analysis of the STI value on shoot fresh weight on the static hydroponic method.

CharactersDirect influenceIndirect influenceResidual
RFWLL3SAGSAAGR
RFW0.600**‒0.006‒0.108‒0.0590.5550.589
LL3‒0.0090.447‒0.086‒0.0450.461‒0.007
SA‒0.1090.59‒0.007‒0.0590.567‒0.107
GSA‒0.0600.59‒0.006‒0.1080.562‒0.059
AGR0.576**0.578‒0.007‒0.108‒0.0590.565

R2: 0.86, **: Significant direct effect at P ≤ 0.01, RFW: Root fresh weight, LL3: 3rd leaf length, SA: Shoot area, GSA: Green shoot area, AGR: Area growth rate.



Based on this analysis, root fresh weight is the character that has the highest direct effect (0.600) on shoot fresh weight. Similarly, based on the image-based phenotyping charac-ter, the area growth rate is the character that has the highest direct effect (0.576) on shoot fresh weight. Therefore, the area growth rate directly affects the variance of shoot fresh weight as one of characters selection.

Moreover, the path analysis on the STI value of root fresh weight in static hydroponic showed representative results with a determinant coefficient value of 0.87 (Table 5). Based on this analysis, shoot fresh weight was the character that had the highest direct effect (0.548) on root fresh weight. Based on image-based phenotyping charac-ters, the shooting area (0.493) and green shoot area (0.377) were the characters that had a significant direct effect on root fresh weight. Meanwhile, the area growth rate has a negative significant direct effect (‒0.417) on the root fresh weight. This showed that the high correlation of area growth rate to root fresh weight was due to the indirect effect of the shoot and green shoot areas. Therefore, shoot area and green shoot area as image-based phenotyping characters can represent the root fresh weight variance, while the area growth rate is unable to be a selection character to represent root fresh weight variance.

Table 5 . Path analysis of the STI value on root fresh weight on the static hydroponic method.

CharactersDirect influenceIndirect influenceResidual
LL3SAGSAAGRSFW
LL3‒0.0130.3890.282‒0.3350.421‒0.010
SA0.493**‒0.0100.374‒0.4110.5380.485
GSA0.377**‒0.0100.488‒0.4080.5360.371
AGR‒0.417**‒0.0100.4850.3690.537‒0.402
SFW0.548**‒0.0100.4840.369‒0.4090.538

R2: 0.87, **: Significant direct effect at P ≤ 0.01, SFW: Shoot fresh weight, LL3: 3rd leaf length, SA: Shoot area, GSA: Green shoot area, AGR: Area growth rate.



The principal components analysis showed that there was one main component that can be used as the basis for the selection index (Table 6),

Table 6 . Principle component analysis of the STI value on the static hydroponic method.

VariablesPC1PC2PC3PC4PC5PC6
RFW0.4162‒0.22670.6382‒0.1950.49020.2996
SFW0.4183‒0.1602‒0.1349‒0.7478‒0.4166‒0.2201
LL30.35270.92580.1169‒0.0058‒0.0530.0447
SA0.421‒0.11390.03570.38580.1809‒0.7918
GSA0.4177‒0.22240.10760.5005‒0.61920.3613
AGR0.4192‒0.0579‒0.74030.05850.40890.3198
CP0.92840.98810.99350.99750.99891
EV5.57070.35780.03240.02430.00830.0064

The numeric in table indicate the eigenvector value, CP: Cumulative proportion, EV: Eigenvalues, PC: Principal component, SFW: Shoot fresh weight, RFW: Root fresh weight, LL3: 3rd leaf length, SA: Shoot area, GSA: Green shoot area, AGR: Area growth rate.



namely the first principal component (PC1) with a cumulative proportion of 0.928 which is approximately 1. Based on PC1, shoot fresh weight (0.4183) and root fresh weight (0.4183) in drought stress were in the same direction with phenotyping characters of shoot area (0.4210), green shoot area (0.4177), and area growth rate (0.4192). These three image-based phenotyping characters were combined in a static hydroponic selection index using the formula as follows:

Static hydroponic index = 0.421 shoot area + 0.4177 shoot green area + 0.4192 growth rate area

(Equation 1)

Dynamic hydroponic system

Analysis of variance in the dynamic hydroponic method in Table 7 showed that the PEG treatment as drought stress has an impact on all image-based-phenotyping characters, except on the ratio of red to green from the top view and the ratio of green shoot area to shoot area from the top view. Variety variance has a significant effect on the number of leaves, red, and blue from the top view, the ratio of shoot area to the convex hull from the top view, object extends X, and Y from a side view, and convex hull from a side view. Meanwhile, the characters that were significantly affected by the interaction include red and blue from the top view, the ratio of shoot area to the convex hull from the top view, object extend Y from a side view, and convex hull from a side view. All characters that were significantly affected by the interaction effect in the dynamic hydroponic system were correlated with the selection index on static hydroponic. This occurred to detect the selected characters that have a relationship with the static hydroponic. The varieties phenotype on these characters are shown in Supplementary 2.

Table 7 . Means square value of analysis of variance on the dynamic hydroponic method.

CharactersDVD × V
SFW8,042.24**319.95ns376.25ns
RFW467.01**53.23ns14.64ns
CT81,129,688.30**4,224,903.16ns1,562,942.10ns
SAT5,571,152.36**474,034.51ns97,115.97ns
GSAT3,072,273.82**333,330.03ns80,272.07ns
NLT2,478.87**617.19*156.84ns
RT3,946.91**462.79**189.34*
GT4,734.46**241.02ns59ns
BT505.53**193.2**74.72*
RRGT0.0051ns0.0082ns0.0099ns
RGSST0.0083ns0.0674ns0.048ns
RSCT0.0226**0.0085**0.001*
XS300.57**174.52*66.56ns
YS1,636.09**147.66**110.49**
SAS3,696,517.43**280,606.15ns91,206.26ns
GSAS1,651,359.59**212,567.33ns68,283.65ns
CS48,116,645.30**6,716,755.96**947,650.84*
RGSSS0.0853*0.0141ns0.0082ns
RSCS0.0463**0.0063ns0.0024ns

D: Drought level, V: Varieties, *: Significant effect at P ≤ 0.05, **: Significant effect at P ≤ 0.01, ns: Not significant, SFW: Shoot fresh weight, RFW: Root fresh weight, CT: Convex hull from top view, SAT: Shoot area from top view, GSAT: Green shoot area from top view, NLT: Number of leaves from top view, RT: Red from top view, GT: Green from top view, BT: Blue from top view, RRGT: Ratio of red to green from top view, RGSST: Ratio of shoot green area to shoot area from top view, RSCT: Ratio of shoot area to convex hull from top view, XS: Object extend X from side view, YS: Object extend Y from side view, SAS: Shoot area from side view, GSAS: Green shoot area from side view, CS: Convex hull from side view, RGSSS: Ratio of green shoot area to shoot area from side view, RSCS: Ratio of shoot area to convex hull from side view.



Significant characters on dynamic hydroponic were correlated with the selection index on static hydroponic. Based on this analysis (Table 8),

Table 8 . Correlation analysis of STI values on the dynamic hydroponic method.

RTBTRSCTYSCSSH index
RT1
BT0.88**1
RSCT‒0.74*‒0.84**1
YS‒0.48ns‒0.24ns0.3ns1
CS‒0.30ns‒0.03ns0.14ns0.97**1
SH index‒0.16ns‒0.04ns0.19ns0.9**0.91**1

The numeric in table indicate the correlation value, *: Significant correlated at P ≤ 0.05, **: Significant correlated at P ≤ 0.01, ns: Not significant, RT: Red from top view, BT: Blue from top view, RSCT: Ratio of shoot area to convex hull from top view, YS: Object extend Y from side view, CS: Convex hull from side view, SH index: Static hydroponic selection index.



the object extends Y from the side view (0.91) and convex hull from the side view (0.91) have a significant correlation with the selection index of static hydroponic. Therefore, these characters can be continued in PCA analysis to develop a selection index of a dynamic hydroponic system.

The principal components analysis on dynamic hydroponic results was shown in Table 9, which showed that PC1 has the highest cumulative proportion with a value of 58.83% of the total diversity of the initial data. Therefore, PC1 can be the basis in the weighting of selection characters such as the static hydroponic index. However, red from the top view and blue from the top view have negative eigenvectors. In contrast, the characters of the ratio of shoot area to the convex hull from the top view, object extend Y and convex hull from the side view have positive eigenvectors. This showed that the YS and CS characters were combined into a selection index in dynamic hydroponic screening using the formula stated below:

Table 9 . Principle component analysis of the STI value on the dynamic hydroponic method.

PC1PC2PC3PC4PC5
RT‒0.50980.2535‒0.44040.4184‒0.5539
BT‒0.47940.4091‒0.3104‒0.31160.6399
RSCT0.4142‒0.3275‒0.8412‒0.09710.0643
YS0.45160.4988‒0.02290.66150.3305
CS0.36710.6422‒0.0396‒0.53‒0.4128
CP0.58830.85440.95360.98011
EV2.94141.33060.4960.13220.0997

The numeric in table indicate the eigenvector value, CP: Cumulative proportion, EV: Eigenvalues, PC: Principal component, RT: Red from top view, BT: Blue from top view, RSCT: Ratio of shoot area to convex hull from top view, YS: Object extend Y from side view, CS: Convex hull from side view.



Dynamic hydroponic index = 0.4516 object extend Y from the side view + 0.4177 Convex hull from the side view

(Equation 2)

Selection index

The index value of each variety on static and dynamic hydroponics was shown in Table 10. Salumpikit variety at a PEG concentration of 10% (2.03) showed the highest value, while IR 20 at a PEG concentration of 20% (0.11) showed the lowest value in static hydroponics. Based on dynamic hydroponic, salumpikit variety at a PEG concentration of 10% (0.53) remained the best, while Inpari 34 (0.22) and IR 20 (0.22) were the varieties with the lowest index value. Furthermore, the regression analysis results showed that the static hydroponic index has a significant linear regression to the dynamic hydroponic index as shown in Fig. 4. Similarly, this figure also showed that the Salumpikit as the tolerant check variety has a good value index than IR 20 as a sensitive check variety on both 10% and 20% PEG.

Table 10 . Selection index on static and dynamic hydroponics.

VarietiesTreatmentsSH indexDH index
Inpari 34PEG 10%0.610.31
IR 20PEG 10%0.270.23
SalumpikitPEG 10%2.030.53
CiherangPEG 10%0.590.35
JelitengPEG 10%0.780.34
Inpari 34PEG 20%0.120.22
IR 20PEG 20%0.110.22
SalumpikitPEG 20%0.390.36
CiherangPEG 20%0.250.27
JelitengPEG 20%0.160.3

SH index: Static hydroponic selection index, DH index, Dynamic hydroponic selection index.


Figure 4. Analysis of static hydroponic selection index regression on dynamic hydroponic.
DISCUSSION

The significant interaction in ANOVA is one of the early indicators in stress screening. Meanwhile, a previous study by Akҫura and Ҫeri (2011); Anshori et al. (2019); Farid et al. (2020) showed that characters that are influenced by interactions had different patterns of decreased response between genotypes on normal and stressful environments. This difference can be used as a parameter to identify tolerance traits between genotypes in stress (Ali et al. 2014; Safitri et al. 2016; Fadhli et al. 2020). Based on this analysis, the image-based phenotyping of 3rd leaf length, shoot area, shoot green area, and area growth rate on static hydroponic screening as well as red, blue, and the ratio of shoot area to the convex hull from the top view, object extend Y, and convex hull from the side view in dynamic hydroponic screening are used as a candidate to selection criteria in stress tolerance. These characters were also significantly influenced by the variance of PEG concentrations and varieties, which indicated that the variations in interactions between genotypes on environmental differences are shown directly (Farid et al. 2020; Farid et al. 2021). Therefore, in-depth analysis with several series of multivariate analyzes was required to obtain the best selection criteria for each drought screening process.

The assessment of varieties' response to stress needs to be carried out using a tolerance index. Meanwhile, a study by Mau et al. (2014), Aboughadareh et al. (2019), Anshori et al. (2019); Aboughadareh et al. (2020), and Farid et al. (2021) stated that the Stress Tolerance Index (STI) is appropriate in characterizing the most tolerant of drought stress. This index is classified as dynamic because it considers the general response of the population means, even though the general response on this index focused on normal conditions (Anshori et al. 2019). Therefore, the STI value becomes the basis for deeper analysis as shown in previous studies conducted by Anshori et al. (2019) on rice against salinity stress, Fadhli et al. (2020) on maize against drought stress, and Farid et al. (2021) on wheat against drought stress. Hence, the use of STI as the basis for a more in-depth analysis was also carried out for static hydroponic and dynamic screenings.

The correlation and path analysis on the static hydroponic were focused on the shoot and root fresh weight characters. A previous study showed that these fresh weights have a close relationship with drought tolerance (Saha et al. 2019). Moreover, the relationship between image-based phenotyping characters on these two characters can be the basis to determine the best selection criteria for drought stress. Similarly, the concepts of correlation and path analysis are often combined to determine the selection criteria (Safitri et al. 2011). This correlation still cannot describe the magnitude of the importance of a character to the diversity of the main characters (Singh and Chaudhary 2007; Rohaeni and Permadi 2012; Anshori et al. 2018). The use of this analysis has been widely reported by several studies (Aman et al. 2020; Ashraf et al. 2020; Saleh et al. 2020), therefore, it needs to be carried out on the character of the image-based phenotyping on the two fresh weights.

The correlation and path analyses results on static hydroponic results showed that shoot and green shoot areas, as well as area growth rate can be used as selection criteria. This is in line with a study by Hairmansis et al. (2014) on the potential of these image-based phenotyping characters on rice under salinity stress and barley under drought stress, while Kim et al. (2020) were on rice under drought stress. Based on the path analysis results, the area growth rate has an inverse orientation to the shoot and the green shoot areas. This difference in orientation is shown in the direct and indirect effect on the character of the shoot and root fresh weights (Table 4 and 5). This showed that the shoot and root fresh weights become the middle of the three image-based phenotyping characters. However, the combination of the three characters can make the selection easier because each character has specific potential. This is in line with a study by Acquaah (2007) and Anshori et al. (2021) which showed that the selection index formed by combining characters with opposite orientations has the potential to increase the selectivity of a selection. Therefore, these three characters are still used as selection criteria. However, the crucial thing in the formation of the selection index of the three characters is the weighting value of the selection, which can be determined using the principal component analysis.

The principal component is a multivariate analysis that aims to extract important information from large data into a new simpler set of orthogonal variables (Ilmaniati and Putro 2019). This analysis has been reported as a weighting indicator for the selection index by Anshori et al. (2019), Alsabah et al. (2019), Fadhli et al. (2020), Akbar et al. (2021) and Farid et al. (2021). Based on the PCA results in Table 6, the characteristics of shoot area, green shoot area, and area growth rate have the same eigenvector direction and variance with the shoot and root fresh weights. This showed that the three characters generally have the same variance direction, although, in the broad cross-section, the growth of the shoot has a different orientation from the other two characters. However, due to the inclusion of the two fresh weight characters, the three characters generally have the same range direction that can be combined into the selection index in Equation 1.

Validation of image-based phenotyping characters in static hydroponics was carried out by identifying the relationship between these characters and dynamic hydroponic characters. Relatively, dynamic hydroponic has a complex growth rate than static hydroponic (Sagita et al. 2020). This was because dynamic hydroponic observations were carried out in the flowering phase, while static hydroponic were in the vegetative phase. Since the application of the selection agent in the two methods was carried out at the same time in the 21 days after sowing, therefore, the static hydroponic selection index assessment can be related to the growth character of dynamic hydroponic as part of its validation.

The dynamic hydroponic ANOVA showed that red, blue, and ratio of shoot area to the convex hull from the top view, object extend Y and convex hull from a side view were the characters that were significantly affected by the interaction of drought level–genotype treatments. Meanwhile, shoot and root fresh weights showed an insignificant effect on the genotype and interaction of drought level–genotype. This was due to the high error of these characters and the variance domination from drought level treatment, hence, the genotype and its interaction are difficult to be determined on both fresh characters. Therefore, the shoot and root fresh weights are inappropriate as the main characters in detecting the selection characters from image-based phenotyping in this dynamic hydroponic system.

Moreover, the correlation analysis of image-based phenotyping characters on dynamic hydroponic that have significant interactions is a method of validating static hydroponic selection indexes. Based on the dynamic hydroponic screening correlation results, object extend Y and convex hull from the side view can be used as selection characters. This is in line with a study by Duan et al. (2018) and Kim et al. (2020) which stated that the leaves on plants that are stricken with drought turn yellow and bend, which caused a reduction in the calculated convex hull. However, the results of this study also need to be supported by other analyzes such as PCA and regression between selection indices (Fig. 4) as stated by Anshori et al. (2018) in determining the selection criteria.

Based on PCA results, the object extends Y and the convex hull from the side view has the same eigenvector direction. Meanwhile, the eigenvectors that are relatively similar in both dimensions showed the closeness of the variance between the two variables (Anshori et al. 2018). This showed that both characters can be used as validation characters through the selection index. Furthermore, the dynamic hydroponic selection index can be used as the Y factor in a regression. Based on the regression results, the static hydroponic index has good adherence to the dynamic hydroponic index. The differences in tolerance groups between the Salumpikit as the tolerant check variety and IR 20 varieties as the sensitive check variety had a significant distance, which is in line with a study by Hairmansis et al. (2020) and Kartina et al. (2019). Meanwhile, the use of control varieties can be one of the proofs to detect effective selection. This concept had reported in previous studies by Ali et al. (2014), Safitri et al. (2016) and Anshori et al. (2020) on salinity stress, Anshori et al. (2019) and Wening et al. (2019) on submergence, Akbar et al. (2018) and Akbar et al. (2021) on drought stress. Furthermore, the use of Salumpikit and IR 20 have been reported as a check or control of rice drought tolerance by Akbar et al. (2018) and Akbar et al. (2021). Based on these results, the selection index on static hydroponic screening has the potential to be used to identify the tolerance properties of rice to drought stress.

Therefore, the use of multivariate analysis on image-based phenotyping characters has the potential to increase the effectiveness of selection in drought stress tolerance screening. The selection index for static hydroponic is 0.421 shoot area + 0.4177 green shoot area + 0.4192 area growth rate. Also, the character selection index for dynamic hydroponic is 0.4516 objects extend Y from the side view + 0.4177 convex hull from the side view. These results showed that rice screening based on image-based phenotyping can be recommended as a more effective and efficient method for the rapid screening of drought stress.

Supplemental Materials
pbb-9-4-272-supple.docx
ACKNOWLEDGEMENTS

We are grateful to Hasanuddin University for funding this research through the Penelitian Dasar UNHAS Scheme with contract number 2649/UN4.1/KEP/2020.

References
  1. Aboughadareh AP, Omidi M, Naghavi MR, Etminan A, Mehrabi AA, Poczai P, et al. 2019. Effect of water deficit stress on seedling biomass and physio-chemical characteristics in different species of wheat processing the D genome. Agron. 9: 522.
    CrossRef
  2. Aboughadareh AP, Mohammadi R, Etminan A, Shooshtari L, Maleki-Tabrizi N, Poczai P. 2020. Effect of drought stress on some agronomic and morpho-physiological traits in durum wheat genotypes. Sustainability 12: 5610.
    CrossRef
  3. Acquaah G. 2009. Principles of plant genetics and breeding. Blackwell Publishing, Oxford, UK.
  4. Akbar MR, Purwoko Bs, Dewi IS, Suwarno WB, Sugiyanta. 2018. Penentuan indeks seleksi toleransi kekeringan galur dihaploid padi sawah tadah hujan pada fase perkecambahan. J. Agron. Indonese. 46:133-139.
    CrossRef
  5. Akbar MR, Purwoko BS, Dewi IS, Suwarno WB, Sugiyanta, Anshori MF. 2021. Agronomic and yield selection of doubled haploid lines of rainfed lowland rice in advanced yield trials. Biodiversitas. 22: 3006-3012.
    CrossRef
  6. Akҫura M, Ҫeri S. 2011. Evaluation of drought tolerance indices for selection of turkish oat (Avena sativa L.) landraces under various enviromental conditions. Zemdirbyste. 98:157-166.
  7. Ali MN, Yeasmin L, Gantait S, Goswami R, Chakraborty S. 2014. Screening of rice landraces for salinity tolerance at seedling stage through morphological and molecular markers. Physiol. Mol. Biol. Plants 20: 411-423.
    Pubmed KoreaMed CrossRef
  8. Alsabah R, Purwoko BS, Dewi IS, Wahyu. 2019. Selection index for selecting promising doubled haploid lines of black rice. SABRAO J. Breed. Genet. 51: 430-441.
  9. Aman J, Bantte K, Alamerew S, Sbhatu DB. 2020. Correlation and path coefficient analysis of yield and yield components of quality protein maize (Zea mays L.) hybrids at Jimma, Western Ethiopia. Int. J. Agron. 2020: 9651537.
    CrossRef
  10. Anshori MF, Purwoko BS, Dewi IS, Ardie SW, Suwarno WB, Safitri H. 2018. Determination of selection criteria for screening of rice genotypes for salinity tolerance. SABRAO J. Breed. Genet. 50: 279-294.
  11. Anshori MF, Purwoko BS, Dewi IS, Ardie SW, Suwarno WB. 2019. Selection index based on multivariate analysis for selecting doubled-haploid rice lines in lowland saline prone area. SABRAO J. Breed. Genet. 51: 161-174.
  12. Anshori MF, Purwoko BS, Dewi IS, Suwarno WB, Ardie SW. 2020. Cluster heatmap for detection of good tolerance trait on doubled-haploid rice lines under hydroponic salinity screening. IOP Conf. Ser. Earth Environ. Sci. 484: 012001.
    CrossRef
  13. Anshori MF, Farid M, Nasaruddin, Musa Y, Iswoyo H , Sakinah AI, et al. 2021. Development of image-based phenotyping for selection characters of rice adaptability on the seedling salinity screening. IOP Conf. Ser. Earth Environ. Sci. 807: 032022.
    CrossRef
  14. Anshori MF, Purwoko BS, Dewi IS, Ardie SW, Suwarno WB. 2021. A new approach to select doubled haploid rice lines under salinity stress using indirect selection index. Rice Sci. 28: 368-378.
    CrossRef
  15. Asaari MSM, Mertens S, Dhondt S, Inzé D, Wuyts N, Scheunders P. 2019. Analysis of hyperspectral images for detection of drought stress and recovery in maize plants in a high-throughput phenotyping platform. Comput. Electron. Agric. 162: 749-758.
    CrossRef
  16. Ashraf ATMH, Rahman MM, Hossain MM, Sarker U. 2020. Study of correlation and path analysis in the selected okra genotypes. Asian Research Journal of Agriculture 12: 1-11.
    CrossRef
  17. Bouman BAM, Tuong TP. 2001. Field water management to save water and increase its productivity in irrigated lowland rice. Agric. Water Manag. 49: 11-30.
    CrossRef
  18. Brereton M, Salle BDL, Ardern J, Hyde K, Burthem J. 2015. Do we know why we make errors in morphological diagnosis? An analysis of approach and decision-making in haematological morphology. EBioMedicine 2: 1224- 1234.
    Pubmed KoreaMed CrossRef
  19. Das Choudhury S, Bashyam S, Qiu Y, Samal A, Awada T. 2018. Holistic and component plant phenotyping using temporal image sequence. Plant Methods 14: 1-21.
    Pubmed KoreaMed CrossRef
  20. Donggulo CV, Lapanjang IM, Made U. 2017. Growth and yield of rice (Oryza sativa L.) under different jajar legowo system and planting space. J. Agroland 24: 27-35.
  21. Duan L, Han J, Guo Z, Tu H, Yang P, Zhang D, et al. 2018. Novel digital features discriminate between drought resistant and drought sensitive rice under controlled and field conditions. Front. Plant Sci. 9: 492.
    Pubmed KoreaMed CrossRef
  22. Fadhli N, Farid M, Rafiuddin, Effendi R, Azrai M, Anshori MF. 2020. Multivariate analysis to determine secondary trait in selecting adaptive hybrid corn lines under drought stress. Biodiversitas. 21: 3617-3624.
    CrossRef
  23. Farid M, Nasaruddin, Musa Y, Anshori MF, Ridwan I, Hendra J, et al. 2020. Genetic parameters and multi-variate analysis to determine secondary traits in selecting wheat mutant adaptive on tropical lowlands. Plant Breed. Biotech. 8: 368-377.
    CrossRef
  24. Farid M, Nasaruddin, Musa Y, Ridwan I, Anshori MF. 2021. Effective screening of tropical wheat mutant lines under hydroponically induced drought stress using multivariate analysis approach. Asian J. Plant Sci. 20: 172-182.
    CrossRef
  25. Fernandez GCJ. 1992. Effective selection criteria for assessing stress tolerance. Proc. of the International Symposium on Adaptation of Vegetables and Other Food Crops in Temperature and Water Stress. Shanhua, Taiwan. pp. 115-121.
  26. Guimaraes PHR, Lima IP, Castro AP, Lanna AC, Melo PGS, Raissac M. 2020. Phenotyping root systems in a set of japonica rice accessions: Can structural traits predict the response to drought. Rice 13: 1-19.
    Pubmed KoreaMed CrossRef
  27. Hairmansis A, Berger B, Tester M, Roy SJ. 2014. Image-based phenotyping for non-destructive screening of different salinity tolerance traits in rice. Rice 7: 1-10.
    Pubmed KoreaMed CrossRef
  28. Hairmansis A, Yullianida, Hermanasari R, Lestari AP, Sasmita P, Suwarno. 2020. Drought tolerant rice breeding lines developed for rainfed lowland areas. IOP Conf. Ser. Earth Environ. Sci. 423: 012019.
    CrossRef
  29. Honsdorf N, March TJ, Berger B, Tester M, Pillen K. 2014. High-throghput phenotyping to detect drought tolerance QTL in wild barley introgression lines. PLoS One 9: 5. e97047.
    Pubmed KoreaMed CrossRef
  30. Ilmaniati A, Putro BE. 2019. Analisis komponen utama faktor-faktor pendahulu (Antecendents) berbagai pengetahuan pada usaha mikro, kecil, dan menengah (UMKM) di Indonesia. J. Teknol. 11: 67-78.
  31. Islam MR, Kayess MO, Hasanuzzaman M, Rahman MW, Uddin MJ, Zaman MR. 2016. Selection index for genetic improvement of wheat (Triticum aestivum L.). J. Chem. Bio. Phys. Sci. 7:1-8.
    Pubmed CrossRef
  32. Iswara RA, Lestari E, Rusdiyana E. 2021. Persepsi pemuda mengenai alih fungsi lahan pertanian padi sawah di Kecamatan Kebakkramat Kabupaten Karanganyar. Jurnal Indonesia Sosial Sains. 2: 1766-1776. https://doi.org/10.36418/jiss.v2i10.436.
  33. Kartina N, Purwoko BS, Dewi IS, Wirnas D, Nindita A. 2019. Skrining Awal Toleransi Galur-galur Dihaploid Gogo terhadap Cekaman Kekeringan pada Stadia Bibit. J. Agron. Indones. 47: 1-8.
    CrossRef
  34. Kim SL, Kim N, Lee H, Lee E, Cheon KS, Kim M, et al. 2020. High-throughput phenotyping platform for analyzing drought tolerance in rice. Planta 252.38: 1-18
    Pubmed KoreaMed CrossRef
  35. Kose A, Onder O, Bilir O, Kosar F. 2018. Application of multivariate statistical analysis for breeding strategies of spring safflower (Carthamus tinctorius L.). Turk. J. Field Crops 23: 12-19.
    CrossRef
  36. Lanceras JC, Pantuwan G, Jongdee B, Toojinda T. 2004. Quantitative trait loci associated with drought tolerance at reproductive stage in rice. Plant Physiol. 135: 384-399.
    Pubmed KoreaMed CrossRef
  37. Li Y, Wen W, Guo X, Yu Z, Gu S, Yan H, et al. 2021. High-throughput phenotyping analysis of maize at the seedling stage using end-to-end segmentation network. PLoS One 16: 1. e0241528.
    Pubmed KoreaMed CrossRef
  38. Mangansige C, Ai NS, Siahaan P. 2018. Panjang dan volume akar tanaman padi lokal Sulawesi Utara saat kekeringan yang diinduksi dengan polietilen glikol 8000. Jurnal MIPA 7: 12-15.
    CrossRef
  39. Mattjik AA, Sumertajaya IM. 2011. Sidik peubah ganda dengan menggunakan SAS. IPB Press, Bogor, ID.
  40. Mau YS, Ndiwa ASS, Arsa IGBA. 2015. Drought tolerance of local rote and check varieties of groundnut (Arachis hypogeae L.) under dry season in two locations in East Nusa Tenggara. AJAS 36: 268-277.
    CrossRef
  41. Nio SA, Tondais SM, Butarbutar R. 2010. Evaluasi indikator toleransi cekatman kekeringan pada fase perkecambahan padi (Oryza sativa L.). J. Biol. 14: 50-54.
  42. Nuraida D. 2012. Pemuliaan tanaman cepat dan tepat melalui pendekatan marka molekuler. El-Hayah Jurnal Biologi 2: 97-103.
    CrossRef
  43. Osmolovskaya N, Shumilina J, Kim A, Didio A, Grishina T, Bilova T, et al. 2018. Methodology of drought stress research: Experimental setup and physiological characterization. Int. J. Mol. Sci. 19: 4089.
    Pubmed KoreaMed CrossRef
  44. Rajamani S, Sreekanth M, Naik VS, Ratnam M. 2016. Selection indices for yield attributing characters improvement in pigeon pea (Cajanus cajan L. Millspugh). Int. J. Life Sci. Scienti. Res. 2:127-129.
  45. Rohaeni WR, Permadi K. 2012. Analisis sidik intas beberapa karakter komponen hasil terhadap daya hasil padi sawah pada aplikasi agrisimba. AJoAS. 2: 185-190.
  46. Rusmawan D, Ahmadi, Muzammil. 2018. Pengaruh ketersediaan air terhadap produksi padi sawah. Proc. Sem. Nas. Hari Air Dunia 2018. Palembang, Indonesia. pp.208-214.
  47. Safitri H, Purwoko BS, Dewi IS, Abdullah B. 2011. Correlation and path analysis on phenotypic characters of doubled haploid rice lines. Widyariset 14: 295-304.
  48. Safitri H, Purwoko BS, Dewi IS, Ardie SW. 2016. Morpho-physiological response of rice genotypes grown under saline conditions. Journal of ISSAAS. 22: 52-63.
  49. Sagita YA, Aini N, Azizah N. 2020. Pengaruh Beberapa Sistem Hidroponik Kultur Air dan Jumlah Tanaman per Netpot Pada Pertumbuhan dan Hasil Tanaman Selada (Lactuca sativa L.). Jurnal Produksi Tanaman 8: 594-600.
  50. Saha S, Begum HH, Nasrin S. 2019. Effect of drought stress on growth and accumulation of proline in five rice varieties (Oriza sativa L.). J. Asiat. Soc. Bangladesh, Sci. 45: 241-247.
    CrossRef
  51. Saleh MM, Salem KFM, Elabd AB. 2020. Definition of selection criterion using correlation and path coefficient analysis in rice (Oryza sativa L.) genotypes. Bull. Natl. Res. Cent. 44: 143. 1-6.
    CrossRef
  52. Siddiqui ZS, Cho JI, Park SH, Kwon TR, Ahn BO, Lee GS, et al. 2014. Phenotyping of rice in salt stress environment using high-throughput infrared imaging. Acta Bot. Croat. 73: 149-158.
    CrossRef
  53. Singh RK, Chaudhary BD. 1977. Biometrical methods in quantittative genetic analysis. Kalyani Publisher, New Delhi, IN.
  54. Sujinah, Jamil A. 2016. Mekanisme respon tanaman padi terhadap cekaman kekeringan dan varietas toleran. Iptek Tanaman Pangan 11: 1-7.
  55. Supriyanto B. 2013. Pengaruh ccekaman kekeringan terhadap pertumbuhan dan hasil padi gogo lokal, kultivar jambu. Agrifor : Jurnal llmu Pertanian dan Kehutanan 12: 77-82.
  56. Swapna S, Shylaraj KS. 2017. Screening for osmotic stress responses in rice varieties under drought condition. Rice Science 24: 253-263.
    CrossRef
  57. Vishal MK, Tamboli D, Patil A, Saluja R, Banerjee B, Sethi A, et al. 2020. Image-based phenotyping of diverse rice (Oryza sativa L.) genotypes. arXiv preprint arXiv: 2004.02498.
  58. Wening RH, Susanto U. 2017. Seleksi cepat galur-galur padi terhadap cekaman kekeringan. Proc. PERIPI 2017. Bogor, Indonesia. pp.326-337.
  59. Wening RH, Rumanti IA, Purwoko BS, Suwarno WB, Khumaida N. 2019. Screening of submergence tolerant rice under artificial condition based on multiple selection indices. Plant Breed. Biotech. 7: 360-374.
    CrossRef


December 2021, 9 (4)
Full Text(PDF) Free
Supplementary File

Cited By Articles
  • CrossRef (0)

Funding Information

Social Network Service
Services
  • Science Central