Abstract: 【Objective】To analyze the fiber quality, yield, and early maturity of Upland cotton germplasm, screen the evaluation indicators comprehensively, establish a reliable evaluation model, and provide theoretical support for developing new Upland cotton varieties.【Method】 A total of 630 Upland cotton accessions selected from various sources were used to investigate 17 traits in 8 environments in three major cotton-growing regions in China. The best linear unbiased prediction (BLUP) was estimated for phenotypic traits by utilizing R package "lme4" and used for further analysis, including correlation analysis, principal component analysis, affiliation function method, cluster analysis and stepwise regression to evaluate population characteristics.【Result】This upland cotton population had high genetic diversity with a diversity index ranging from 1.961 to 2.084, and significant regional specificity existed. The boll number, fiber elongation, spinning consistent index, and short fiber index of this population had considerable variation, while that of fiber length, fiber strength, and growth period traits was lower. Correlation analysis showed significant or highly significant correlations among most traits, and there was a strong correlation among some fiber quality traits. The 17 traits were converted into 6 independent composite indices through principal component analysis with a contribution range of 5.860%-31.044% and a cumulative contribution of 82.642%. Principal component analysis can classify the fiber quality traits, yield traits and agronomic traits in this upland cotton population. The comprehensive evaluation value (F value) of phenotypes was calculated using the affiliation function method, the phenotypes of 17 traits were all significantly correlated with F values. The result demonstrated that the accessions with high F values (mean value of 0.668) had significantly higher yield traits (boll number, boll weight, lint percentage, and seed index), fiber quality traits (fiber length, fiber length uniformity, fiber strength, spinning consistent index, and short fiber index), plant height and fertility traits than those with low F values (mean value of 0.396). Regression equations for eight traits (boll number, boll open date, boll weight, flowering date, lint percentage, plant height, sympodial brand node, and spinning consistent index) as independent variables were established using stepwise regression. Based on the F value, the 630 germplasm were clustered into four categories, the first category was the high-quality fiber type, containing 118 accessions; the second category was the high-yield type, comprising 250 accessions; the third category was the early-maturing type, comprising 51 accessions; the characteristics of the fourth category was between the second and third categories. Finally, 23 better fiber quality and 135 high-yield germplasms were selected for breeding and production.【Conclusion】The phenotypic traits of Upland cotton are geographically specific; it is feasible to use multivariate statistical analysis to comprehensively evaluate Upland cotton germplasm; the whole population can be classified into four categories (high-quality, high-yield, early maturity and other types).

Key words: Gossypium hirsutum , Linn., germplasm resources, principal component analysis, cluster analysis, comprehensive evaluation