Evaluation of drought tolerance in a world collection of Eruca sativa L. based on selection indices and some agronomic traits

ABSTRACT Drought is the most destructive environmental limitation that jeopardizes agronomic production. Given the wide range of industrial and medicinal applications of arugula (Eruca sativa. L.), the present study evaluated the response of a world collection of 64 arugula genotypes to drought stress over a two-year study period (2019 and 2020) through selection indices and some agronomic traits. The main objective was to identify the best selection indices from Susceptibility Index (SSI), Yield Stability Index (YSI), Tolerance Index (TOL), Mean Productivity (MP), Geometric Mean Productivity (GMP), and Stress Tolerance Index (STI) that could be exploited in the genotypic selection of the crop’s drought-tolerant cultivars. The highest values of STI, MP, and YSI were recorded for the genotypes G58 and G35 in 2019 and 2020, respectively. Generally, the superior genotypes for seed yield in 2019 were G56 and G58 while G35 and G55 proved so in 2020. It was concluded that these genotypes could be utilized as donor parents in drought tolerance in arugula.


Introduction
Arugula (Eruca sativa L.), an annual and edible oilseed, has elite industrial and nutraceutical properties (Garg and Sharma 2014). Belonging to the family Brassicaceae (Bell and Wagstaff 2019) it is native to the Mediterranean region (e.g. Italy), North Africa, Middle East (e.g. Iraq, Iran and Syria), and eastern Asia (including Pakistan, Afghanistan, and India) (Sastry 2003). The crop is greatly adaptable to a wide range of climatic conditions such as those encountered in Iran (Golkar and Abdollahi Bakhtiari 2020).
Identified by such local names as Mandab, Kakesh, and Kakoj, the crop is used in Iran as a vegetable in traditional medicine (Mirzabe et al. 2021). The whole plant is rich in a wide range of bioactive compounds, minerals, vitamins, flavonoids, and antioxidants to which the plant owes its medicinal properties (Marwat et al. 2016;Bell and Wagstaff 2019). The medicinal applications include its use as a tonic, digestive, rubefacient, laxative, astringent, and stomachic (Jaafar and Jaafar 2019). Moreover, the pharmaceutical industry can benefit from this marginal oil crop in making different psychostimulants (Bell and Wagstaff 2019). In industry, it is utilized for producing biofuels, oil cake for animal feed, cosmetics, detergents, polymers, lubricants, and illuminating agents (Garg and Sharma 2014;Mumtaz et al. 2016). The fresh aerial parts are consumed as a vegetable or food ingredient in sauce and salads (Tassi et al. 2018).
Water deficit is a critical abiotic stress that limits plant growth and development, especially in arid and semi-arid zones all around the world (Mahmood et al. 2020;Takahashi et al. 2020), with the timing and duration of drought incidence playing the main roles in the intensity of adverse drought effects (Blum 2011); hence, the quest for drought-tolerant cultivars. Drought tolerance is a complex trait involving a multitude of genes that modify several physio-morphological and molecular responses (Mahmood et al. 2020). These biological signals need to be generated at suitable speeds and times in order for the plant to respond appropriately to stressed conditions for survival (Takahashi et al. 2020).
The global climate change has encouraged a myriad of breeding programs to combat drought stress in various species (Rauf et al. 2016). Generally speaking, the high variation in relative grain yield performance under diverse environments due to genotype × environment interaction in genotypes makes it extremely difficult to select the superior genotypes that guarantee breeding progress (Blum 2011;Rauf et al. 2016). It is, therefore, instructive to focus on selecting and/or breeding genotypes or varieties characterized not only by improved yields under drought stress but also by appropriate responses to water-limited conditions as beneficial drought-combat strategies (Raman et al. 2012). Unfortunately, insufficient sources of tolerant parents have led to only a meagre progress in breeding drought-tolerant genotypes in arugula. A useful measure that may be adopted to address this insufficiency, is genetic variability evaluation accompanied by comparisons of yield performance in different genotypes under both stressed and non-stressed conditions for breeding programs in drought-prone environments. In this regard, several selection indices have been suggested for screening drought-sensitive and drought-tolerant plants based on yield under both environments; these include drought-resistance index (DRI) (Fischer and Maurer 1978), geometric mean productivity (GMP) and stress tolerance index (STI) (Fernandez 1992), mean productivity (MP) and tolerance Index (TOL) (Rosielle and Hamblin 1981), stress susceptibility index (SSI) (Fischer and Maurer 1978), yield stability index (YSI) (Bouslama and Schapaugh 1984), and yield index (YI) (Gavuzzi et al. 1997).
A rather large number of studies have been reported on the use of drought indices for selecting superior genotypes in such oil plant species as safflower (Golkar et al. 2021), cumin (Karimi Afshar et al. 2021, corn (Naghavi et al. 2013), and rapeseed (Aliakbari et al. 2014).
Few studies have examined the physiological traits (de Oliveira Mangarotti Dp et al. 2020) and germination indices (Huang et al. 2017) of arugula in response to the drought stress. Furthermore, the evaluated drought responses of arugula were based on photosynthetic pigments and certain secondary metabolites (Ginzburg and Klein 2020). To the best of the present authors' knowledge, however, there have been no studies regarding the drought tolerance screening in arugula based on drought indices and agronomic traits.
This gap motivated a preliminary study into the drought-tolerant arugula genotypes with the highest seed yields and certain agronomic traits under both drought and non-stressed conditions. Therefore, the current study was designed and implemented to assess drought tolerance variation among arugula accessions, and to compare drought indices for identifying the superior accessions under both conditions.

Plant materials
The plant materials employed in the present study included 64 accessions of arugula. The seeds of non-Iranian accessions were procured from the Leibniz Institute of Plant Genetics and Crop Plant Research (IPK) collection. In addition, the seeds of arugula accessions belonging to Iranica flora were obtained from Pakan Bazr Co. (Isfahan, Iran). The genotypic origins of all the studied genotypes are reported in Table 1.

Experimental design and crop planting
To evaluate the 64 arugula accessions during the two winter cropping seasons (October to June of 2018-2019 and 2019-2020), the experimental design was arranged as a split plot design based on a randomized complete block design with three replications under two (non-stress and stress) water conditions. The field experiments were carried out at Lavark Experimental Farm, Agricultural Research Station of Isfahan University of Technology (32°32ʹ N, 51°23ʹ E, 1630 m above sea level). The physio-chemical properties of the soil are presented in Table S1. Monthly temperatures (max, min and mean), rainfall and evaporation at the experimental station are shown in Table S2. Three lines for each genotype were planted in plots 1.5 m long with 70 cm row spacing and a distance of 10 cm between each two plants. No pesticide was utilized during the trial, and weeds were manually controlled. Before sowing, the plot was fertilized using 200 Kg ha −1 P and 250 Kg ha −1 N ha. In about 35 days after planting, 100 Kg ha −1 N was top dressed.

Irrigation treatments
Prior to irrigation, soil samples had been collected from a soil depth of 0 to 60 cm from both environments to determine soil water content and to calculate the irrigation water content based on a 60-cm rooting depth. Irrigation depth was then determined using the following formula: where, I is irrigation depth (cm), FC (-0.03 MPa) is soil gravimetric moisture percentage at field capacity (22%), θ (-1.5 MPa) is soil gravimetric moisture percentage at irrigation time (10%), D is rootzone depth (50 cm), and B is soil bulk density at the root zone (1.3 g cm -3 ) (Carter and Gregorich 2007).
The plants under both experimental (non-drought and drought stress) conditions were irrigated at sowing time, after which irrigation took place every week until the plants reached their flowering stage. In the drought stressed treatment, irrigation was terminated before 10% of the plants in each plot reached the flowering stage. Moreover, irrigation in both non-drought and drought treatments was continued until 50% and 90%, respectively, of the field capacity (FC) had been depleted from the root zone. A pressurized irrigation system using polyethylene drip-irrigation tapes (16 mm diameter) was applied beside each planting row. To monitor the volume of each water application, use was made of Parshall flume (Id = I × p; where p is the fraction of I that may be depleted from the root zone). Moreover, Ig = (Id/ Ea) × 100, with Ea representing irrigation efficiency (%), was assumed to be 65% on average. The differences observed in mean temperatures during the growing seasons across the two years of study caused differences in available water. After removal of edge effects, agronomic traits including seed yield (SY), number of branches per plant (NB), number of capsules per plant (NC), number of seeds per capsule (NSC) and 1000 seed weight (SW) were calculated in 10 randomly selected plants of the middle row.

Assay of selection indices
Drought tolerance indices were calculated for each genotype using the following equations: where, Y s , Ȳ s , Y p , and Ȳ p denote grain yield and mean yield of all the genotypes under drought stress and non-drought conditions, respectively.

Statistical analyses
The data collected over the two consecutive years were subjected to combined analysis of variance and correlation analysis using the SAS statistical program (version 9.4; SAS Institute Inc., Cary, NC, USA). Also, a correlation analysis was performed using different stress tolerance indicators to describe patterns of associations. The results of PCA analysis were used to construct biplots based on yield-based indices of drought tolerance by appropriate factor loadings (Bahrami et al. 2014). Cluster analysis was also performed based on all the studied traits using the Ward's method to identify similarities.

Analysis of variance and means comparisons
Analysis of variance revealed the highly significant effects of genotype, year, and genotype × year interaction for seed yield under non-drought conditions (Y p ) and stress conditions (Y s ) for all the agronomic traits and tolerance indices of the genotypes studied ( Table 2). The significant effects of year and year × genotype interaction indicating that genotype behaved differently in view point of seed yield and agronomic traits in the two study years. The ten minimum and maximum mean values of Y p , Y s , and tolerance indices for seed yield in 2019-2020 are reported in Table 3 and Table 4). The results of means comparisons for all the studied traits showed that the values obtained in 2020 were higher than those obtained in the first study year.

Agronomic traits
The mean comparison of different agronomic traits is presented in Table S4. When compared with the non-stress results, the mean NB, NC and SW values decreased significantly under drought-stress in both 2019 (Table S4a) and 2020 (Table S4b). According to the results, the highest NB values were recorded for G 58 (14.3) in 2019 (Table S4a) and G 55 (7.33) in 2020 (Table S4b). The highest NC values were reported for G 56 (166.9) in 2019 (Table 5) and G 35 (302.2) in 2020 ( Table S4b). The highest SW values were seen in the G 5 (1.52 g) and G 3 (1.48 g) genotypes in 2019 and 2020, respectively (Table  S4). The mean NSC values showed a significant increase from 42.48 in 2019 and 42.32 in 2020 under non-stress to 46.96 in 2019 and 46.97 in 2020 under drought-stress (Table S4 a-b). The G 24 (50.3) and G 62 (49.7) genotypes had the highest NSC values under drought stress in 2019 and 2020, respectively (Table S4).
In the non-drought treatment, the highest seed yield values were recorded for G 24 (1501.6 kg ha −1 ) and G 52 (2576 kg ha −1 ) in 2019 and 2020, respectively, while the lowest seed yields under non-stress conditions were recorded for G 7 (790 and 1360 kg ha −1 ) in 2019 and 2020, respectively. This is while G 58 (1216.6 kg ha −1 ) and G 35 (2130 kg ha −1 ) had the highest seed yields under drought stress conditions in 2019 and 2020, respectively, but G 32 (608.3 kg ha −1 ) and G 21 (1050 kg ha −1 ) showed the lowest values for this trait in 2019 and 2020, respectively (Table 3).      Table 1.

Selection indices
Based on the stress tolerance index (STI), G 58 (

Correlation coefficient and principal component analysis
The data obtained from the correlation analysis of seed yield and drought tolerance indices can be exploited to screen the superior genotypes. The correlation coefficients obtained among Y p , Y s , and all the drought tolerance indices are reported in Table 5. Clearly, seed yield under the non-drought treatment showed positive and significant correlations (0.56** and 0.55**) with those of the stress treatment in 2019 and 2020. Results revealed that all selection indices included in the study, except for GMP, had significant and positive correlations with seed yield in the non-drought treatment. Moreover, MP, STI, and YSI established the highest correlations with Y p and Y s in both years. Negative correlations (r = −0.23 and r = −0.44) were established between TOL and seed yield under the stressed condition (Y s ) in the two study years, respectively. It was concluded that selection for tolerance would lead to a decline in yield in drought-stressed environments but to an increase in seed yield under non-drought conditions, as evidenced by r = 0.66** and r = 0.51** in the two years of 2019 and 2020, respectively (Table 5). This is while a perfect correlation was established for YSI to Y S (r = 1**) in both years, with values close to 1 indicating that the two elements are behaving almost identically. Principal component analysis (PCA) was performed on the mean values of indices obtained for the two study years (Y p and Y s ) to discover the weight of each trait towards the observed variation. The first two components explained 87.5% and 87.6% of the total variation in 2019 and 2020 (Table 6). In 2019, STI, MP, YSI, Y P , and Y S were positively correlated with PC 1 , which accounted for 57.9% of the total variation. Thus, PC 1 may be identified as the yield potential and drought tolerance. The second component (PC 2 ) that justified 29.6% of total variation had high positive correlations with TOL and SSI but a low one with GMP (Table 6). This PC can, therefore, be termed 'the stress-tolerant dimension'. Regarding the results of this biplot (Figure 1a), the genotypes G 15 , G 59 , G 62 , G 50 , G 60 , G 61 , and G 52 had relatively high PC 1 values but low PC 2 ones so that they may be identified as the superior genotypes for breeding in both non-drought and drought-stress conditions (red color). Genotypes G 32 , G 5 , and G 31 were recognized as the ones with relatively high susceptibility but low productivity while G 23 , G 25 , G 42 , and G 7 were characterized by both low productivity and susceptibility (green color). Genotypes G 13 , G 19 , G 22 , and G 48 (blue color) were found more suitable for nondrought than stress environments due to their high PC 2 values recorded in 2019 while G 23 , G 25 , G 7 , and G 42 proved more desirable in 2020 for stressed conditions rather than non-drought environments.
The results of principal component analysis for the second year showed that the first and second principal components explained 87.6% of the total variability (Figure 1b). PC 1 exhibited high correlations with Ys, MP, YSI, and STI while the second (PC 2 ) had positive correlations only with  the susceptibility indices of TOL, SSI, and Y p . Based on the biplot of PC 1 vs. PC 2 (the coordinates representing genotypes and vectors representing indices), genotypes G 35 , G 14 , G 55 , G 42 , and G 52 (green color) had the highest STI, YSI, and MP values (Figure 1b). Based on the current findings, genotypes number G 61 , G 37 , G 7 , G 11 and G 12 (blue color) in 2020 are more suitable for stressed conditions than for non-drought environments while the genotypes with high PC 2 values are more likely to gain superior results in non-drought environments than under stress conditions. Because of the positive and high significant correlation observed over the two years of study between STI and seed yield under non-drought (0.88 and 0.83) and stressed (0.88 and 0.91) condition, three-dimensional graphs were drawn based on the STI index ( Figure 2). The two graphs divided the genotypes into four groups, each representing one combination of high yield in the nondrought environment (Group A), genotypes with high yields under both environments (Group B), low yield genotypes under both environmental conditions (Group C), and high yield ones in a stressful environment (Group D). The genotypes included in group A based on the data from the second year are G 13 , G 59 , G 33 , and G 47 with higher seed yields in non-drought environments but low yields under drought conditions. Group B comprised the genotypes including G 1 , G 7 , G 9 , G 11 , and G 37 characterized by superior performance under both stress and non-stress conditions. The genotypes with low yields under both environmental conditions were classified under group C (e.g. G 4 , G 24 , G 46 , and G 44 ). No genotype was detected with a high yield in a stressful environment (Group D). Thus, it is possible to screen and identify the genotypes with the most reliably steady performance in all environments.
Based on this analysis, the arugula accessions studied in 2020 were classified into three groups. Group 1 consisted of G 1 , G 7 , G 11 , G 37 , G 9 , G 6 , G 20 , G 12 , G 8 , G 28 , G 30 , G 34 , G 40 , G 10 , G 31 , G 53 , G 16 , G 32 , and G 64 (blue color) that exhibited the lowest seed yields under either environmental conditions as well as MP, STI, and YSI indices (Figure 3b). The genotypes G 13 , G 59 , G 33 , G 47 , G 54 , G 58 , G 21 , G 29 , and G 57 (red color) grouped in the second cluster for which high values of Y p , TOL, and SSI were recorded. Finally, the genotypes in the third cluster exhibited high values of Ys, MP, STI, and YSI.

Discussion
Drought stress is recognized as a limiting factor adversely affecting multiple aspects of plant growth and development (Kapoor et al. 2020). Breeding drought tolerant cultivars seems to be a promising solution but it poses difficulties due to the polygenic nature, low heritability, and genotype × environment interaction of each species (Blum 2011). A remedy entertained by scholars screening for tolerance, which has forced plant breeders to use reliable indices for this purpose (Blum 2011). Regarding arugula, few studies have been reported in the literature dealing with the effects of drought stress on this species (Huang et al. 2019). Despite the advances in conventional breeding and transformation technology, developing arugula cultivars for drought tolerance remains a major gap in efforts to cultivate this important industrial crop.
Investigating different aspects (biochemical, physiological, and agro-morphological) of plant response to drought stress enhances our knowledge of the capacity of crops to adapt properly to drought-prone environments (Kapoor et al. 2020;Pour-Aboughadareh et al. 2020). To improve this knowledge, the present study was implemented to evaluate the responses of 64 arugula genotypes to drought stress over a two-year study. The studied genotypes showed a highly diversified array of responses to both non-stress and stress conditions regarding all the traits investigated. Analysis of variance revealed a large genetic variation in drought tolerance as an unpredictable factor affecting seed yield and agronomic traits in genotypes collected from different geographical regions. The vast variety in responses to drought enhances the array of choices available for selecting droughttolerant genotypes.
Similar to previous reports for rapeseed (Shahsavari et al. 2014), the mean values of the studied genotypes over the 2019-2020 period showed a significant decreasing trend in NB, NC and SW under drought stress conditions. It could be concluded that the lessening of seed yield components (NB, NC and SW) led to the decreased seed yield of arugula genotypes under stress conditions. In a different trend, the NSC showed an increasing trend in response to drought stress, which may be due to the compensatory effects of the arugula seed yield components in dealing with drought stress. The trend of changes (ascending or descending) in agronomic traits were similar in both study years, but superior genotypes were not similar over the two-year study period. The superior genotypes for agronomic traits could be used for the improvement of drought tolerance in arugula through indirect selection methods. In this study, seed yield was significantly reduced due to the drought stress in both 2019 and 2020. Based on the present investigations, high seed yield values were recorded for the G 56 and G 55 genotypes under non-stress conditions in 2019 and 2020, respectively, whereas G 58 and G 35 were identified as high seed yield genotypes under stress conditions in 2019 and 2020, respectively. Thus, the G 49 , G 58 , G 59 , G 60 , G 35 , G 36 and G 15 genotypes may be considered as the ones with economic yields in drought affected years. On the other hand, the G 56 , G 55 , G 59 , G 24 , G 18 and G 52 genotypes were considered as high-performance under optimal rainfall conditions. High values of STI, MP, and YSI were recorded for G 58 , G 56 in 2019 but for G 55 and G 35 in 2020. A previous study has shown that the best indicator of yield potential is the index that establishes a significant correlation with seed yield in both normal and stressful environments (Bahrami et al. 2014;Golkar et al. 2021). Thus, genotypes with high values of YSI, STI, and MP and low values of SSI (namely, G 59 and G 15 in 2019 and G 14 and G 49 in 2020) will expectedly have higher yields under stress conditions. Mashilo et al. (2017) also maintained that screening for drought tolerance must be based on genotypes with high yields under both non-drought and stressed conditions. The present observation regrading reduced seed yield under water scarcity in all the genotypes studied confirmed previous reports on different oil species such as safflower (Bahrami et al. 2014;Golkar et al. 2021), cumin (Karimi Afshar et al. 2021, and bottle guard (Mashilo et al. 2017). The significant genotype × year interaction explained the different responses of arugula genotypes observed over the two years of the experiment. From among the ten-high seed-yielding genotypes under nondrought conditions, the highest seed-yielding ones over the two agronomic years were G 56 , G 58 , and G 13 .
Regarding the drought indices, G 13 and G 56 , respectively, were identified as the two genotypes with the highest values for TOL and MP indices over the two consecutive years. Also, high values of GMP were recorded for G 6 and G 8 in both years. It may, therefore, be concluded that he performance of these genotypes with regard to the studied indices is year-independent. This is in contrast to the different responses exhibited by other genotypes over the two study years under water limited conditions, indicating the significant effect of year on these genotypes.
Based on the diverse reactions of the studied genotypes during the 2019-2020 study period, it is highly probable that differences in climatic conditions (such as temperature, rainfall before the onset of drought stress, and relative humidity) led to significant changes in the seed yields and agronomic traits of the genotypes over the two-year study period. The considerable association established between both (stress and non-stress) treatments with respect to grain yield indicate the possibility for the selection of favorable genotypes (Karimi Afshar et al. 2021). Fernandez (1992) maintained that the best index for selecting stress-tolerant genotypes is one that has a relatively powerful correlation with seed yield under stress and non-stress conditions. It follows that selection of arugula based on higher STI, MP, and YSI values observed in the current study should result in drought tolerance and yield improvement, which agrees with the findings reported in Golkar et al. (2021), Hao et al. (2011), andEbrahimiyan et al. (2012). It has been suggested that STI and GMP might serve as the most powerful indices for screening drought-tolerant and high-yielding genotypes in both droughtstressed and irrigated conditions (Akcura and Ceri 2011;Ebrahimiyan et al. 2012;Golkar et al. 2021). Fernandez (1992) in a study of mung bean, and Farshadfar and Sutka (2002), in a study of maize, also found that Ys was strongly correlated with MP and STI while Y p was correlated with MP and STI. In contrast, high values of SSI and TOL are indicators of genotypes sensitive to drought stress, pointing out that tolerant genotypes should be selected based on low TOL values under stress conditions (Golkar et al. 2021).
Based on the SSI and TOL values obtained in the current study, G 61 (Portugal) and G 14 (Afghanistan3) were identified as drought-tolerant genotypes that performed desirably under the stress conditions in 2019 and 2020, respectively (Table 3). GMP and STI have been claimed as better predictors of Y p and Y s than other indices under control and stress conditions (Bahrami et al. 2014;Sardouie-Nasab et al. 2014;Golkar et al. 2021). Akcura and Ceri (2011) indicated that genotypes with a high STI are usually associated with highly different yields under the two different conditions. Moreover, selection based on an aggregate index as a combination of different indices might prove more useful for improving upon drought tolerance in genotypes (Golkar et al. 2021).
One advantage of principal component analysis (PCA) is that it allows enhanced interpretability of data by decreasing the size of data sets (Jolliffe and Cadima 2016). In the present analysis, the impacts of the different indices in each PC indicated that PC1 and PC2 could be identified as yield potential and stress susceptibility groups, respectively. The STI, MP, YSI, Y p , and Y s measured in 2019 were found to correlate positively with PC 1 that accounted for 57.9% of the total variation. Moreover, GMP and YSI negatively correlated with PC 2 while SSI and TOL exhibited a high and positive correlation with PC 2 (Table 6). PC 1 can be, therefore, considered to affect drought tolerance and yield potential (Bahrami et al. 2014). Nevertheless, the genotypes with low PC 2 and high PC 1 values are useful in both stressed and non-stressed conditions (Mashilo et al. 2017). Based on the mean comparisons of the indices calculated for arugula yield, G 56 was found the most drought tolerant genotype due its highest mean values of MP, YSI, and STI measured in 2019. The 3-D plot may be employed to identify and select genotypes with the most stable yields across all the environments (non-stress and drought stress) to arrive at a classification similar to that reported by Fernandez (1992).
Cluster analysis has also been widely used not only to discriminate high distance genotypes but also to determine genetic diversity based on similar traits under drought stress conditions (Mohammadi et al. 2011;Naghavi et al. 2013). Some scholars used cluster analysis to explore genetic diversity and group different plant species based on their responses to stress conditions (Bahrami et al. 2014;El-Rawy and Hassan 2014;Golkar and Abdollahi Bakhtiari 2020).
The cluster analysis performed in the current study assigned the genotypes to three groups of most tolerant genotypes for use as parents in drought-tolerant arugula breeding programs. The genotypes in Groups 2 and 3 were identified as the most drought tolerant genotypes in 2019 and 2020, respectively. However, the dendrograms did not support the division of the genotypes into distinct groups based on their geographical origin. It was, therefore, suggested that selection of parents for hybridization would not need be based on geographic diversity in these arugula germplasms. Moreover, a good agreement was observed between the results of the present cluster analysis and those of the PCA. Based on both analyses, the drought tolerant genotypes with high PC 1 and PC 2 values as well as those assigned to Groups of 1 and 3 in the cluster analysis may be used as extreme parental genotypes with the highest genetic distance to develop new hybrid varieties in arugula aimed at production of drought-tolerant cultivars. However, further evaluation of genotypes based on drought tolerance indices across multiple locations is required to confirm their stability for developing improved arugula genotypes.

Conclusion
In this study, drought stress significantly affected the seed yield and agronomic traits of arugula genotypes. The findings of the present study showed that the selection indices of STI, MP, and YSI were the superior ones for use in identifying genotypes capable of enhanced performance in both non-drought and stressed environments. The genotypes identified over the two years of study as the most tolerant (G 58 and G 35 ) and the most susceptible (G 24 and G 21 ) might be exploited to produce mapping populations for drought tolerance in arugula. In 2019, G 56 (i.e. Turkey 1) with the highest MP mean value exhibited the highest drought tolerance. Also, the highest values of YSI and STI were recorded for G 58 (Belgium) in the same year. This is while the highest values of MP, STI, and YSI were recorded for G 35 in 2020. Based on the data collected during the two-year study period, the G 58 , G 56, G 55 , and G 35 genotypes may be suggested as the most tolerant, not only for cultivation in arid and hot climates, but also as donor parents in hybridization programs.