Integrating data from asymmetric multi-models can identify drought-resistant groundnut genotypes for drought hot-spot locations

Water/drought stress experiments are frequently conducted under imposed stress or rainout shelters, while natural drought hot-spot investigations are rare. The “drought hot spot” in Anantapur, Andhra Pradesh, India, is appropriate for drought stress evaluation due to its hot, arid environment, limited rainfall, with over 50% rainfall variability. According to reports, 30 out of 200 groundnut cultivars in India are supposed to possess drought-tolerant characteristics. However, these cultivars are yet to be evaluated in areas that are prone to drought. This study tested these drought-tolerant genotypes in naturally drought-prone areas of Anantapur under rainfed conditions from Kharif 2017 to 2019. Pod yield and rainfall-use-efficiency (RUE) were measured for these genotypes. Genotype and genotype*environment interactions affected pod yield and RUE (GEI). The AMMI model exhibits significant season-to-season variability within the same area with environmental vectors > 90° angles. GGE biplot suggested the 2018 wet season for drought-resistant cultivar identification. Kadiri5 and GPBD5 were the most drought-tolerant cultivars for cultivation in Anantapur and adjacent regions. These types could also be used to generate drought-tolerant groundnut variants for drought-prone regions.


Chikbal
The development of water-stressed genotypes will be aided by an understanding of how well genotypes survive in arid environments.In general, genotypes selected for adaptability and performance under high input conditions are poorly suited to low input environments 13 .In India, more than 200 varieties have been released for commercial cultivation by a central or state varietal release committee 14 , of which 30-35 varieties have been reported to possess drought resistance but this has not been validated in drought hot spot locations.
Thus, the purpose of this study was to determine a variety's potential to grow in Anantapur's sparse rainfall conditions while maintaining steady yields throughout the seasons.Understanding how well genotypes survive in low-rainfall environments will benefit in the development of drought-tolerant genotypes.Notwithstanding fluctuations in annual weather patterns, these drought-resistant genotypes will be able to provide yields that are much greater over time.To investigate varietal stability and considerable year-to-year yield fluctuations use of stability models such as Additive main effect and multiplicative interaction (AMMI) [15][16][17][18] and GGE biplots were utilised.

Material and methods
Experimental materials.The experimental material comprised of thirty cultivars that had been released for cultivation in different parts of India and were considered to be drought-tolerant 14 .Table 2 provides a list of these varieties, along with their release year and area of adoption.These materials were sourced from our own gene bank at ICAR-Directorate of Groundnut Research, Junagadh.

Experimental locations and meteorological data.
Experiments were done at the Regional Research Station (RRS) of the ICAR-Directorate of Groundnut Research in Anantapur (latitude: 14° 41′ N, longitude: 77° 67′ E) during the three consecutive rainy seasons of 2017, 2018, and 2019.Red sandy loam of shallow depth, low in organic carbon (0.35%) and available nitrogen (142 kg/ha), medium in accessible phosphorus (32 kg/ha) and potassium (226 kg/ha) was the soil at the experimental site.The climate in the region of study is semi-arid, therefore it is hot and dry for the majority of the year, with average highs about 37 °C.The average weekly low temperature ranged from 19.70 °C in November to 25.30 °C in July, and the average weekly high temperature ranged from 35.4 °C in July to 35.70 °C in October.Although though the groundnut crop was grown in the same  1S).In 2018, moisture stress has become the most significant limiting factor in the study area.
Experimental design and data collection.Seeds were sown directly in a randomised complete block design (RCBD) with two replications during the third week of July.Each genotype was planted in 3 m rows of single rows per genotype each replicate.Plant geometry of 30 cm row-to-row and 10 cm plant-to-plant distance was maintained in each plot.Crops were harvested when they reached maturity.After drying pods, pod yields were recorded on a plot-by-plot basis.According to Oweis 19 , Rainwater use efficiency (RUE) was computed by dividing groundnut pod yield by cumulative rainfall received from seeding to harvest as follows: Pod yield (g/ m 2 )/Rainfall (mm).RUE denotes the yield achieved by a genotype per milli metre of rain water received during the study period.Because the crop receives no irrigation other than rain water, RUE would also reflect a genotype's water productivity or water use efficiency under rainfed conditions.

Statistical analysis. AMMI and GGE biplot analysis. AMMI stability model calculates environment and
genotype main effect and, multiplicative effects of GEI.The AMMI stability analysis was performed using package 'agricolae' 20 in R 21 and the model is represented as where the response variable such as pod yield and RUE is represented by Y ijk , grand mean represented by μ, genotype deviation from μ represented by G i , environment deviation from μ represented by E j , eigen value of kth interactive principal component (IPCA) represented by ʎ k , IPCA score for ith genotype on kth IPCA represented by α ik , IPCA score of jth environment for kth IPCA represented by γ jk , residual GEI unexplained by model rep- resented by d ij and model error represented by e ijk .The sum of square (SS) due to GEI signal 22

was estimated as
The stable performance of genotypes in AMMI stability model was calculated as Modified AMMI Stability Index (MASI) 23 in R using the package 'ammistability' 24,25 as PCn are the scores of nth IPC; and θn is the percentage sum of squares explained by the nth principal component interaction effect.
Simultaneous stability index (SSI) was used to compare stability of high pod yielding and RUE genotypes and ranking the genotypes by combining both yield and stability parameters 26 .The SSI for each genotype was estimated in R using the package 'ammistability' 24,25 as where, rSP is the rank of MASI stability value and rY is the rank of adjusted mean pod yield and RUE of genotype across environments.GGE-biplot analysis was performed on pod yield separately for Spanish and Virginia groups using 'GGEbiplotGUI' package 34 in R 21 .
Handling plant materials.The collection and handling of plant were in accordance with all the relevant guidelines.

AMMI stability analysis for Pod yield and RWUE.
For pod yield and RUE, the AMMI ANOVA revealed the significance of the mean sum of squares due to drought environments, genotypes, and GEI (Table 3).The main effects of drought environment, genotype, and GEI accounted for 81.76, 4.29, and 12.80% of phenotypic variability in pod yield and 74.87, 6.18, and 17.50% of phenotypic variability in RUE, respectively.The variance caused by GEI was further subdivided into variance caused by signal and noise (Table 4).The variance caused by signal occurred due to known factors such as genotypes and phase of drought, whereas noise variation was attributed to mistake caused by unknown model factors 22 .In our experiments, pod yield and RUE recorded 94.29 and 94.76 percent of GEI owing to signal and the remainder due to noise, respectively.This indicates that the AMMI stability model is adequate for comprehending the GEI 27 .The multiplicative component of AMMI models consists of the singular value/multiplication factor of IPCA, the genotype eigenvector, and the environment eigenvector 17 .The IPCA1 and IPCA 2 were highly significant for pod yield and RUE, with IPCA1 accounting for 86.99 and 73.33 percent of the GEI for pod yield and RUE, respectively, whereas IPCA 2 accounted for 13.01 and 26.27 percent of the GEI, indicating a significant contribution of environment on genotype and trait expression performance.AMMI 1 biplot.The biplot abscissa and ordinate reflected the 1st principal component (IPCA1) term and the trait's substantial significance in additive main effects and multiplicative interaction 1 (AMMI 1). Figure 1 depicts the additive main effects and multiplicative interaction effect of 30 genotypes on pod yield and RUE throughout three rainy seasons in this study.The rainy season of 2018 had a IPCA1 score or vector closer to zero for pod yield (Fig. 1a) and RUE (Fig. 1b) compared to prior rainy seasons, showing a reduced interaction effect, which almost guaranteed the competitive advantage of genotypes in that season and considered suitable for genotype evaluation.Variety GPBD 5 (11) and TG 72 (29) received near-zero scores on the IPCA1 axis, indicating that it is less affected by the environment.PC 1 scores next to zero lines of biplot suggested that varieties ICGS 44, (13), GPBD 5 (11), and Girnar 2 (10), were suitable for all conditions.Varieties with PC1 vectors of the same sign and score but away from zero biplot lines suggested that they were suited to a specific environment.
Variety ICGV 86031 (15) was discovered to be suited for the rainy season of 2017 and variety AK 265 (2) for the rainy season of 2018, as both genotype and environment exhibited the same sign.According to Murphy et al. 28 , Mogale et al. 29 , and Oladosu et al. 30 , when the IPCA1 score for a genotype or environment is close to zero, there is a small interaction impact; on the other hand, if a genotype and environment achieve the same sign on the PCA axis, there is a positive interaction; otherwise, there is a negative interaction.

AMMI based stability indices for cultivar performance in groundnut. The variety with the lowest
Modified AMMI stability Index (MASI) calculated from the IPCA axis and IPCA scores is the most stable 1,23,24 (Table 5).Under drought stress, TG 72, TAG 24, and Mutant 3 had the lowest MASI for pod yield, with pod yields of 154.Simultaneous selection indices (SSI) for genotypes were calculated using the sum of MASI ranks and genotype ranks determined from pod yield and RUE (Table 5).The genotypes with the lowest SSI values are the most stable and function well across environments [23][24][25]31 . Kadri 5, ICGV 86325, ICGV 91114, and ICGS 1 had the

GGE biplot to visualize GEI for pod yield and RUE. 'Which Won
Where Biplots' aids in visualising mega environments and identifying superior genotypes for drought environments.These biplots are known as 'Which Won Where biplots' because they plot the genotypic mean against the first interaction main components 1,18,32 to identify winning genotype for each environment.The irregular polygons divide the biplots into the vector and aid in determining the appropriate genotype for each environmental sector.In our investiga- tions, GPBD 5 was the best genotype for pod yield and RUE in 2018, Spanish Improved was the best in 2019, and R 2001-3 was the best genotype in 2017 (Figs.2a, 3a).GGE Biplots were utilised to depict the effects of environment, genotype ranking, GEI pattern, and identification of stable environments 33 .The discriminativeness vs. representative perspectives of GGE Biplots aid in identifying the appropriate environments with the most discriminative capacity for genotype differentiation.The type-1, type-2, and type-3 environments can be seen using average environmental coordinates (AEC) and test environments.Type-1 environments are represented by short vectors with average discriminative power, which indicate genotype performance on average.Type-2 environments are depicted as the longest vectors with the best discriminative capacities, capable of discriminating genotype performance.Type-3 environments are indicated by the longest vector with big angles, suited to the negative effects of the environment.The best environments have the longest genotypic vector and are placed on or at acute angles to the AEC 32 .Rainy season of '2019' and '2018' has the longest environmental vector with narrow angles to AEC with the highest discriminative power in our experiments for pod yield and RUE respectively and is regarded the optimal environment to discriminate drought tolerant genotype (Figs.2b, 3b).The shortest environmental vector '2018' and '2019' detected for pod yield and RUE indicate an average or similar genotype performance.
The mean vs stability biplots aid in understanding genotype mean performance across drought situations.Genotypes that are positioned close or in the direction of AEC are regarded excellent and best performing genotypes 32 .Kadiri 5 was the most stable and better performing variety, for pod yield (Fig. 2c) and RUE (Fig. 3c), with AK 265 and Spanish Improved doing poorly.According to our findings, Kadiri 5 and TG 72, located on or near the AEC, are the most stable and high-performing genotypes, but Dh 3-30 and DSG 41, which are placed away from the AEC, are less stable.Varieties Mutant 3 and TAG 24 were located on AEC axis and hence are highly stable but poor performers for Pod yield and RUE.
The genotype ranking on biplots aids in visualising the ideal genotypes based on their positions in the concentric circle 1,18,32 .None of the genotypes were located in the center of concentric circle however, genotypes Anantha and R 2001-2 were located closer to the center of the concentric circle for pod yield and RUE respectively (Figs. 2d, 3d), indicating its stable performance across environments, whereas AK 265 and Spanish improved were located on the last concentric circle for pod yield and RUE and were considered less stable under drought stress.

Discussion
Groundnut (Arachis hypogaea L.) is one of the world's most widely cultivated food legumes, prized for its high protein and unsaturated oil content while, drought stress is one of the most significant restrictions affecting its production.The Rayalasema region (Anantapur, Kurnool, and Chittoor) of Andhra Pradesh and neighboring districts of Karnataka (Bellary, Chickballapur, Chitradurga, and Tumkur) account for roughly 20-25% of the country's groundnut area.These areas are in dry zones, where rainfall distribution is variable and drought is a common occurrence.The main crop farmed in these areas under rainfed circumstances is groundnut, which suffers from periodic dry spells.Some groundnut varieties have been identified as drought tolerant 14 , but they have yet to be confirmed in actual drought hot spots.As a result, the goal of this study was to test 30 drought tolerant varieties in Anantapur's natural drought hot spot conditions during rainy seasons without protective irrigation in order to determine the most stable drought tolerant variety.
The AMMI stability model is commonly used to comprehend the GEI pattern and to find stable cultivars from target environments [23][24][25]31 . TheAMMI ANOVA revealed significant contributions from environments, genotypes, and GEI for pod yield and RUE, which is consistent with prior findings 1,27 .The sum of squares due to GEI Signal is greater than the GEI Noise for pod yield and RUE due to a greater contribution from the additive main effect of genotype and drought stress environments.AMMI model, followed by MASI and SSI stability models, and AMMI 1 biplots, identified Kadiri 5 as the most stable high yielding variety under drought stress conditions for pod yield and RUE, and hence is considered as drought tolerant groundnut variety.
GGE biplots aid in elucidating the interrelationship between environments, rating genotypes, and identifying the best performing genotype in a given environment 27 .According to our findings, the discriminativeness vs representative biplot indicated that '2018' was the ideal year to discriminate genotype for pod yield and RUE because this season experienced severe dry spells from July 23rd to September 3rd, 2018 and October 1st to November 12th, 2018, with rainfall during cropping season being the lowest of the three seasons (Fig. 1S and Table 2S).As a result, the rainy season of '2018' is preferable for identifying drought tolerant genotypes.The genotypes Kadiri 5 was found at an acute angle with an optimal environment for pod yield and RUE and might be considered ideal drought tolerant genotypes for these traits.With the fluctuations in annual weather pattern, Kadiri 5 identified under ideal drought conditions will be able to provide highest yields over the period of time under drought stress conditions.

Conclusion
Drought significantly impacted varietal performance, as demonstrated by AMMI and GGE stability models in the present study.Rainy season of 2019 facilitated clear differentiation of drought-tolerant varieties identifying Kadiri 5 as the most suitable cultivars for drought-prone regions.

Table 2 .
List of groundnut varieties used in the study.

Table 4 .
Estimates of sum of square due to signals and noises by using AMMI model for drought stress in groundnut.GEI genotype × environment interaction, 7, 174.4,and 171.6 g/m 2 respectively.MASI values of 6.03, 6.38, and 5.53 were obtained for high pod yielding genotypes such as ICGV 86031, R 2001-2, and R 2001-3, respectively.Mutant 3, TG 72, and TAG 24 had the lowest MASI for RUE.The MASI for the high RUE genotypes ICGV 86031, R 2001-2, and GPBD 5 was 2.32, 2.24, and 1.41, respectively.

Table 5 .
Ranking of 30 groundnut varieties for pod yield, and rain water use efficiency (RUE).SSI for pod yield, indicating that these genotypes have the ability to combine high stability and high pod yield across environments.For RUE, Kadiri 5, ICGV 91114 and ICGS 1 had the lowest SSI, indicating that these genotypes had high stability and high RUE across environments.Kadiri 5 had the highest pod yield (211.0 g/ m 2 ) and high RUE (45.86 g/m 2 -mm) with low SSI values.As a result, in drought stress conditions, Kadiri 5 is regarded as the most stable drought tolerant high yielding variety.