Yield-based drought tolerance index evaluates the drought tolerance of cotton germplasm lines in the interaction of genotype-by-environment

Data analysis

High-yield classification is the best indicator for assessing drought tolerance (Ramirez & Kelly, 1998). Therefore, the yield and drought tolerance of the tested germplasm lines were evaluated by determining their changes in yield under water stress. The yields of crop species under control were reduced by environmental factors other than water stress (Golestani-Araghi & Assad, 1998; Zhao et al., 2019). Four drought tolerance indices, GMP, MP, STI, SSI and Yr were calculated as: ${\displaystyle \mathrm{GMP}=\sqrt{\left(Ys\ast Yp\right)}}$ ${\displaystyle \mathrm{MP}=\frac{\left(Yp+Ys\right)}{2}}$ ${\displaystyle \mathrm{STI}=\frac{Yp\ast Ys}{\overline{\left(Yp\right)2}}}$ ${\displaystyle \mathrm{SSI}=\frac{1-\left(Yp+Ys\right)}{1-\overline{Yp}+\overline{Ys}}}$ ${\displaystyle \mathrm{Y\; r}=1-\left(\mathrm{Y\; s}/\mathrm{Y\; p}\right)}$ where Ys is the yield under water stress and Yp is the yield under normal irrigation conditions.

An analysis of variance was performed on cotton yield and the drought tolerance indices using SSPS 21.0. Correlation analysis was used to determine the correlation between the seven screening indicators and cotton yield under stress conditions. Three-dimensional (3D) maps were draw with R software to identify the genotypes with strong drought resistance and high yield. The PCA and GGE biplots based on the indices and yield of the three environments were drawn using R. The best linear unbiased prediction (BLUP) value of the cotton yield genotype effect under different drought stress treatments was evaluated with R (lme4 package), and the mixed linear model of each cultivar was as follows (He et al., 2018): ${\displaystyle yi=\mu +Gi+Ei+ei}$ where yi = phenotypic value; µ= total mean value of the total yield in all environments; Gi = genotype effect; Ei = environment effect; and ei is the random error. The genotype effect and environmental effect were random effects, and the others were assumed to be fixed (He et al., 2018).

The GGE biplots model is shown below (Yan et al., 2000): ${\displaystyle Yij-\mu -\beta j={\lambda }_1{\xi }_{\mathrm{i1}}{ŋ}_{\mathrm{j1}}+{\lambda }_2{\xi }_{\mathrm{i2}}{ŋ}_{\mathrm{j2}}+\varepsilon ij}$ where Yij is the expected yield (i is the genotype, j is the environment); µis the grand mean; βj is the environmental main effect; μ + βj is the mean yield in the environment; λ₁ and λ₂ are the singular values (SV) of PC1 and PC2, respectively; ξ_i1 and ξ_i2 represent the eigenvectors of PC1 and PC2, respectively; ŋ_j1 and ŋ_j2 are the eigenvectors of PC1 and PC2 in different environments; ɛij refers to residuals (i refers to genotype, j refers to environment) (Yan et al., 2000).

Phenotype, yield per plant and values of the eight screening indicators for 103 natural cotton populations under normal and stress treatments

The results for the different screening indicators calculated as the BLUP value for cotton plant yield are shown in Table 1. Compared with that under normal conditions, the yield per plant of all cotton germplasm lines under drought stress decreased to varying degrees (Table 1, Table S6). GMP, MP, STI, and SSI confirmed that the genotypes Bellsno, Shi yuan 321, ND359-5 and Xin lu zao 36 were the most tolerant. The genotypes Shuo feng 1 hao, Ji zha 81, and Hong ye mian were identified as having the least sensitivity in terms of GMP and STI (Table 1, Table S6). The results of the analysis of variance of the combined data for the yield per plant and screening indices for 103 cotton germplasm lines are presented in Table 2. We found significant differences among the three growing years and different genotypes. There was a significant difference in yield per plant and all screening indicators among all genotypes (p < 0.01), which indicated that the different germplasm lines performed differently regarding yield per plant and all screening indicators. The differences among all genotypes under the stress treatment provided the basic screening indicators.

Correlation analysis between seven indices across germplasm lines and yield per plant after drought stress

The most effective or best indicators under drought conditions were identified as a method for evaluating ideal drought-resistant genotypes. The BLUP values of 103 germplasm lines under drought and normal conditions were calculated, and the correlation among the 7 yield indicators was analyzed. The correlation analysis results showed that Yp had a very significant positive correlation with GMP (r = 0.78), MP (r = 0.87), STI (r = 0.73), SSI (r = 0.87) and that Ys had a very significant positive correlation with GMP (r = 0.83), MP (r = 0.75), STI (r = 0.84), and SSI (r = 0.75). These results show that GMP, MP, STI, and SSI can be used to identify stable and high-productivity genotypes under normal and drought stress conditions. The several production indicators were divided into two main categories, among which GMP, MP, and STI were used to identify high-yielding lines, and SSI was used to identify stable-yielding lines. The GGE biplot was drawn in R, and the results showed that GMP, MP, STI and SSI were positively correlated with Yp and Ys (Fig. 1). This is consistent with the results of the correlation analysis.

Principal component analysis of five evaluation indices and yield indices under drought stress

The principal components of each index were analyzed according to the cotton yield amounts and yield indices. The principal components were able to explain 97.66% of the total variation in cotton yield under the different moisture conditions (Table S5, Fig. 2). GGE biplots were drawn based on the BLUP values for Yp, Ys, GMP, MP, STI, and SSI, and the eigenvalues of the first two components were >1. The first principal component explained 74.74% of the variation in yield, and Yp, Ys, GMP, MP, STI, and SSI were closely related (Fig. 2). This indicates that these indices can be used to identify genotypes that will produce high yields under drought stress and normal conditions. The second principal component explained 22.92% of the variation and was correlated with Yr under normal conditions. Five germplasm lines CQJ-5, Xin lu zao 45, Bellsno, Zhong R 2016 and ND 359-5 had high values of PC1 (Fig. 2), which reflected their performance under drought conditions and indicated that they have high drought tolerance.

Yield and 3D STI graphs of 103 germplasm lines under normal conditions and drought stress conditions

Of the indices considered, STI had very significant positive correlations with Yp and Ys, and the cotton varieties exhibited stable STI values in production. Based on STI, Ys and Yp, 3D maps were drawn to identify the characteristics of 103 germplasm lines (Fig. 3). According to the 3D graph analysis, CQJ-5, Xin lu zao 45, Bellsno, Zhong R 2016 and ND 359-5 were determined to have stable yield in terms of the STI indicator. These genotypes had higher STI values than the other genotypes and were located in the first quadrant of the 3D graph, indicating that they produce higher yields under drought conditions and normal conditions. These results show that these genotypes have a more stable, higher yield capacity under stress and normal conditions. These results demonstrate the function of the screening index for evaluating cotton genotypes.

GGE biplots for the seven indices and yield of 103 germplasm lines under drought and normal conditions

The yield stability of the 103 cotton germplasm lines is shown in the figure. The Yp, Ys and all indicators for the two conditions (stress and normal) were used to draw the GGE biplots. The genotypes CQJ-5, Zhong R 2016, Liao 5856 and Xin lu zao 45 are located in front of the arrow in the figure, indicating that they had high yields. The genotypes Hongyemian and Ji zha 81 had lower yields, as indicated by their position behind the arrow. The stable genotypes included KK1543, Tianyun 10 and Zao 19, which showed relatively high yield and stability. Tu 76-94, Xi bu 50, Zhongmian 49 and Zhong R 2069 were identified as drought-sensitive genotypes. CQJ-5, Xin lu zao 45, Bellsno, Zhong R 2016 and ND 359-5 had high yields and stability (Fig. 4). This analysis was able to distinguish high-yielding and stable genotypes that can be used as high-yielding germplasm lines or for breeding ideal stable and high-yielding hybrids for cultivation under drought stress conditions. The results from the GGE biplots and the screening index analysis were basically the same, indicating the accuracy and effectiveness of the screening index.

Identification of yield tolerance indices based on drought conditions

At present, water stress is the main abiotic stress factor that restricts crop growth (Boyer, 1982). In the cotton flowering and boll stages, water stress had a serious impact on cotton production (Kumar et al., 2008; Wang, Isoda & Wang, 2004). Therefore, cotton germplasm selection for varieties that require less water or tolerate water stress improves cotton adaptability to drought stress environments (Kumar et al., 2008). In the course of this study, seven screening indicators, MP, GMP, STI, SSI, Yr, Yp and Ys were used to identify the tolerance of 103 cotton germplasm lines to water stress (Table S6). The GMP indicates the average productivity of cotton material under water stress and normal conditions (Kristin et al., 1997). The MP indicates the average yield of cotton under water stress and normal conditions (Zhao et al., 2019). These two indicators reflect the average yield of cotton under different moisture conditions, but one of them provides a higher yield value and the other provides a lower yield value; this discrepancy may result in an incorrect assessment of the genotype (Table 1, Table S6). Therefore, using only the GMP and MP for production selection will lead to evaluation errors (Table S6). The results of Fernandez’s research under normal and stress conditions indicate that the STI can be used to evaluate high-yield germplasm lines (Fernandez, 1992). This indicator reflects the average yield of cotton material under stress and normal conditions (Table S6). Therefore, it was necessary to carry out a comprehensive appraisal and evaluation that included this index (Fig. 2). The genotype Xin lu zao 45 had lower MP and GMP values than the other tested varieties due to its lower yield, but it had a higher STI value, indicating that Xin lu zao 45 is a stable-yielding line (Table 1). There was a very significant positive correlation between the GMP, MP, and STI and cotton yield under normal and stress conditions, indicating that the GMP, MP and STI can collectively be used to distinguish varieties that produce high and stable cotton yields (Fig. 1). Other researchers have obtained similar research results in crops such as wheat and corn (Zhao et al., 2019). The results of other studies have indicated that there is a significant negative correlation between Yr and yield under water stress conditions (Fig. 1). Previous studies on low nitrogen tolerance have focused mainly on physiological or agronomic traits, including leaf area, sucrose content, biomass, etc (Duan et al., 2018). Different evaluation systems and processes have been proposed for use in different orders. The articles published on this topic thus far have proposed many screening indicators but have not reached a consensus on which indicators should be used.

Analysis of the cotton drought tolerance evaluation system and selection of drought-tolerant germplasm

In this study, the combination of the cotton yield BLUP values and PCA was used to develop a drought tolerance index to evaluate cotton drought tolerance in terms of cotton yield, and five drought-tolerant cotton germplasm lines were screened (Table S6, Fig. 3). GGE biplots were used to analyze the genotypes for high yield and stable yield, and the results were consistent with the screening results based on yield screening indicators (Zhao et al., 2019). GGE biplots screen out adaptive germplasm lines under different conditions. At present, GGE biplots are rarely used in evaluations of cotton yield combined with drought resistance (Fig. 2). We bred the tested varieties under drought or water stress and selected germplasm lines with high and stable yields (Fig. 4). With the yield-based drought tolerance index, GGE biplots and 3D plots, the genotypes CQJ-5, Xin lu zao 45, Bellsno, Zhong R 2016 and ND 359-5 were identified as producing high yields under drought conditions (Fig. 3). These germplasm lines can provide a basis for our future development of high-yielding drought-tolerant cotton. Some researchers have identified the drought resistance of some cotton germplasm lines through multiple indicators, including plant height, number of fruit branches, and fiber quality. Among them, Xin lu zao 45 and Zhong R 2016 were identified as drought-resistant materials (Li et al., 2019). These two drought-resistant germplasm lines were also similarly identified by the yield traits in this study, indicating that these two germplasm lines do have good drought resistance.