Genetic and Environmental Contributions to Cotton Yield and Fiber Quality in the Mid-South

Producers need to know the contributions of genotype, environment, and their interaction (G ́E) in determining cotton (Gossypium hirsutum L.) lint yield, lint percentage, and fiber quality. The recent introduction of longer upper half mean length (UHML) fiber, lower micronaire cultivars may alter previously defined contributions. The objectives of this research were to define the genotype, environment, and G ́E contributions to lint yield, lint percentage, and fiber quality from common cultivars evaluated within the US Mid-South and define shifts in these contributions caused by the introduction of a longer UHML, lower micronaire cultivar. Data from 102 large-plot trials within Alabama, Arkansas, Louisiana, Mississippi, Missouri, and Tennessee were compiled from the 2015 and 2016 seasons; 85 site-years contained three common cultivars, and 69 contained four common cultivars. Analysis of three common cultivars within the 69-site-year dataset indicated environment dominated factors of lint yield (85.8%), lint percentage (88.5%), micronaire (70.9%), length (70.5%), and uniformity (70.4%). Large increases in the contribution of genotype to micronaire (26.0%) and length (37.6%) were observed when the lower micronaire, longer UHML cultivar was included. The relatively minor role of cultivar in determining lint yield and the substantial role of cultivar in determining micronaire and length suggest that producers within the Mid-South should begin to place more importance on fiber quality data when selecting cultivars. T.B. Raper, Dep. of Plant Sciences, Univ. of Tennessee, 605 Airways Blvd., Jackson, TN 38301; J.L. Snider, Dep. of Crop and Soil Sciences, Univ. of Georgia, 115 Coastal Way, Tifton, GA 31793; D.M. Dodds, Dep. of Plant and Soil Sciences, Mississippi State Univ., 138 Dorman Hall, Mississippi State, MS 39762; A. Jones, previous address, Fisher Delta Research Center, Univ. of Missouri, PO Box 160, Portageville, MO 63873; B. Robertson, Dep. of Crop, Soil and Environmental Sciences, Univ. of Arkansas, 649 Jackson 917, Newport, AR 72112; D. Fromme, Louisiana State Univ. Ag Center, 8105 Tom Bowman Dr., Alexandria, LA 71302; T. Sandlin, Dep. of Crop, Soil and Environmental Sciences, Auburn Univ., PO Box 159, Belle Mina, AL 35615; T. Cutts, previous address, Dep. of Crop, Soil and Environmental Sciences, 202 Funchess Hall, Auburn Univ., AL 36849; R. Blair, Univ. of Tennessee Extension, Univ. of Tennessee, 605 Airways Blvd., Jackson, TN 38301. Received 3 Apr. 2018. Accepted 13 Sept. 2018. *Corresponding author (traper@utk.edu). Assigned to Associate Editors Emily Merewitz and Gustavo Slafer. Abbreviations: G ́E, genotype ́ environment interactions; RHQ, Regional High Quality; UHML, upper half mean length; USDAAMS, USDA Agricultural Marketing Service. Published in Crop Sci. 59:307–317 (2019). doi: 10.2135/cropsci2018.04.0222 © Crop Science Society of America | 5585 Guilford Rd., Madison, WI 53711 USA This is an open access article distributed under the CC BY license (https:// creativecommons.org/licenses/by/4.0/). Published November 8, 2018

as-possible."Specific fiber quality parameters measured by the USDA Agricultural Marketing Service (USDA-AMS) include fiber upper half mean length (UHML), fiber strength, micronaire (measure of fiber fineness and maturity), uniformity (consistency in length), color, and several additional parameters based on the cleanliness of the fiber.These parameters determine yarn fineness, uniformity, strength, and dying properties, which ultimately affect the quality of the final textile product.Because of the relationship between fiber quality and quality of the final textile product, fiber quality premiums received by producers can be substantial.
Although lint yield throughout the Mid-South (Alabama, Arkansas, Louisiana, Mississippi, Missouri, and Tennessee) has increased substantially over the past 20 yr, substantial improvements in several fiber quality parameters over this same period have not been noted.However, a divergence in cultivar fiber quality has recently been noted within Mid-South cotton cultivar testing programs.In the 2016 University of Tennessee large-plot cultivar testing program, Phytogen (PHY) 444 WRF (Dow AgroSciences) was characterized by significantly longer UHML and lower micronaire than any of the 14 other cultivars tested (Raper et al., 2016).Furthermore, these significant numerical separations were substantial; whereas PHY 444 WRF was characterized by an average micronaire of 4.2 and UHML of 31.8 mm, the next closest cultivars were characterized by a micronaire of 4.6 and UHML of 30.7 mm.Similar results were noted within the 2016 Mississippi State University On-Farm Cultivar Trial Program; PHY 444 WRF and DeltaPine (DP) 1646 B2XF (Monsanto Company) were characterized by significantly longer UHML and greater lint yields than the other eight tested cultivars (Dodds et al., 2016b).Furthermore, micronaire of PHY 444 WRF was significantly and substantially lower than that of any other tested cultivar.Large-and small-plot data generated within cultivar testing programs at Auburn University, the University of Missouri, the University of Arkansas, and Louisiana State University also captured exceptional UHML and lower micronaire from these two cultivars (Cutts et al., 2016;Glass et al., 2016;Jones, 2016;Bourland et al., 2017;Fromme et al., 2017).Additionally, similar trends were noted within the 2015 cultivar testing programs conducted in Alabama, Arkansas, Louisiana, Missouri, Mississippi, and Tennessee (Dodds et al., 2015;Glass et al., 2015;Jones 2015;Raper et al., 2015;Bourland et al., 2016;Fromme et al., 2016).
Over the past 20 yr, the roles of genotype, environment, and the interaction of these terms (G´E) on lint yield, yield components, and/or seed traits have been explored numerous times.Trial types used to define genotype, environment, and G´E can be clearly split according to plot type; several recent approaches have analyzed smallplot, replicated trial data, whereas other approaches have analyzed large-plot data that may not include within-location replicates.Analyses based on small-plot data typically consist of fewer site-years (used to characterize environment) than cultivars, and the selected cultivars do not always represent commercially adopted germplasm.Data from the USDA-ARS Regional High Quality (RHQ) Trials has been used several times within the past 15 yr to define the effects of genotype, environment, and G´E on lint yield, lint percentage, fiber quality, and seed traits (Meredith, 2003;Meredith et al., 2012;Zheng et al., 2014;Zeng et al., 2015).The RHQ Trials, generally conducted across the cotton belt in tandem with the USDA-ARS National Cotton Variety Trials, differ from entries within the National Cotton Variety Trials due to the emphasis on fiber quality placed on USDA-ARS RHQ entry selection (Meredith et al., 2012).Other attempts to define genotype, environment, and G´E that have analyzed small-plot data include an evaluation of eight cultivars in 12 site-years within the South Carolina Official Variety Trials by Campbell and Jones (2005), an evaluation of seven transgenic and conventional cultivars across seven site-years by Blanche et al. (2006), and an evaluation of 82 cultivars from the Pee Dee breeding program across 14 site-years by Campbell et al. (2011).
In contrast, large-plot approaches typically consist of more site-years and fewer cultivars.Cultivars included in these trials represent commercially available germplasm or experimental germplasm that is nearing commercial release.One example is the evaluation of 12 cultivars across 16 states during the 1996 to 1998 growing seasons by Kerby et al. (2000).Results indicated that environment was the dominant factor governing lint yield, lint percentage, UHML, strength, and micronaire (uniformity was not reported; Kerby et al., 2000; Table 1).Although it is unclear how Kerby et al. (2000) derived the interaction between genotype and environment within the strip-trial dataset, this information was reported.More recently, Snider et al. (2013) defined the role of genotype, environment, and G´E with seven commercial cultivars grown in 33 site-years in Georgia during the period from 2010 to 2011 (Table 1).Lint yield and fiber parameters such as UHML, micronaire, and uniformity were primarily driven by environment, with <10% of the variability within these parameters explained by genotype.Genotype did, however, dominate lint percentage (51.5% genotype, 38.8% environment) and play a strong role in realized fiber strength (27.7% genotype, 47% environment).Snider et al. (2013) assessed G´E interactions by comparing slopes after regressing cultivar performance at a given site-year vs. an environmental index and found significant interaction terms for all parameters of interest.
Care must be taken in interpreting and applying defined roles of genotype, environment, and G´E from each trial type.Although genotype, environment, and quality parameters; (ii) to define the shifts in contributions of genotype and environment to lint yield, lint percentage, and fiber quality caused by the introduction of a longer UHML, lower micronaire cultivar; and (iii) to characterize the G´E interaction for lint yield, lint percentage, and fiber quality parameters.

Description of Field Trials
Data from large-plot cultivar trials conducted within the Mid-South were compiled into a single dataset.States that provided data for this analysis included Alabama, Arkansas, Louisiana, Missouri, Mississippi, and Tennessee.Specific trial design, data collection procedures, and data collection varied from state to state and are summarized in Table 2. Trials within each state were placed on producer farms and were managed by the producers in accordance with corresponding extension service recommendations.Plot size varied according to field and equipment size; plot width varied from 4 to 12 rows, row width varied from 76 to 102 cm, and plot length varied from 50 to 700 m.Cultivars were randomly placed within each location.Within-field replicates were not included in all locations.Spindle-machine-harvested seed cotton from each plot was either weighed by an onboard load cell contained within the picker basket, a boll buggy outfitted with load cells, or a portable trailer load cell.Additionally, 2 to 7 kg of seed cotton was collected from each plot to determine lint percentage and fiber quality.Within Alabama, Arkansas, and Louisiana, all seed cotton samples were ginned with a 10-saw tabletop gin without stick machines, incline cleaners, or lint cleaners.In contrast, the Mississippi and Missouri locations were ginned at the University of Missouri MicroGin in Portageville, MO, and the Tennessee samples were ginned by the University of Tennessee MicroGin in Jackson, TN.Both of these 20-saw MicroGins also include one or more stick machines, inclined cleaners, and lint cleaners.After ginning, lint samples for each treatment were submitted to either USDA Classing Offices or processed internally on high volume instrument systems to determine fiber UHML length (mm), fiber bundle strength (kN m kg −1 ), micronaire, and length uniformity (%).Locations of high volume instrument analysis varied by state and are summarized in Table 2.
Due to the variability in length of growing seasons and performance instability of cultivars across the Mid-South, G´E defined by RHQ trials provide tremendous value for breeders attempting to maximize performance stability across the US Cotton Belt, the defined roles of genotype, environment, and G´E from RHQ studies will likely not characterize commercial production scenarios.For example, the many cultivars evaluated in the RHQ trials do not necessarily represent widely planted germplasm, and the range of these cultivars' traits and characteristics may inappropriately overestimate the percentage variability in lint yield, lint percentage, and fiber quality determined by genotype.Furthermore, the low number of site-years used in RHQ trials limits broad applicability of results to a range of production scenarios, and extremes in productivity captured for a small number of site-years could provide a skewed estimate of the impact of environment on lint yield, lint percentage, and fiber quality.In contrast, assessments that evaluate a limited number of cultivars across a large number of site-years, particularly across variable production areas, may inappropriately overestimate the percentage variability in lint yield, lint percentage, and fiber quality determined by environment.
To provide estimates of the roles of genotype, environment, and G´E that producers can expect to observe within a given production region, it is critical that (i) included cultivars represent commercially adopted germplasm within the region of interest, and (ii) included environments capture the variation expected to be observed within the region of interest.Analyses that focus on commercially available, widely planted cultivars across an area large enough to capture variable environments characteristic of the region should provide applicable definitions of genotype, environment, and G´E that can be used to better understand the impact of cultivar selection on lint yield, lint percentage, and fiber quality.Due to the introduction of longer UHML and lower micronaire cultivars into the Mid-South and the lack of information available on the impacts of genotype, environment, and G´E for modern, commercial cotton cultivars commonly planted throughout the Mid-South, the objectives of this study were (i) to quantify the effect of genotype and environment on cotton lint yield, lint percentage, and fiber

Statistical Analysis
Site-year and cultivar impacts on lint yield, lint percentage, and fiber quality were determined separately in SAS version 9.4 (SAS Institute, 2013) with the PROC GLIMMIX analysis.For analysis Although cultivar means and separations across 69 site-years containing the four common cultivars DP 1522 B2XF, PHY 312 WRF, PHY 444 WRF, and ST 4946 GLB2 were similar to the analysis of 85 site-years containing three common cultivars, the introduction of PHY 444 WRF resulted in a greater data range for several parameters (Table 3).Cultivars PHY 444 WRF and PHY 312 WRF had greater lint yields than DP 1522 B2XF, which had greater lint yields than ST 4946 GLB2 among these cultivars in the 69-site-year comparison.Similar trends in strength were observed; PHY 444 WRF had greater fiber strength than DP 1522 B2XF and PHY 312 WRF, but significantly lower fiber strength than ST 4946 GLB2.Cultivar PHY 444 WRF also had significantly greater uniformity than the other cultivars tested.As suspected, fiber UHML of PHY 444 WRF (32.1 mm) was substantially longer than for the other included cultivars (range from 29.8 to 30.6 mm), and micronaire of PHY 444 WRF (4.0) was substantially lower than for the other included cultivars (range from 4.5 to 4.9).
Historically, the highest-yielding cultivars have often been associated with some negative fiber quality traits.Data presented here suggest that PHY 444 WRF breaks this trend; although its mean lint yield was not different than PHY 312 WRF, PHY 444 WRF lint yields were greater than those of DP 1522 B2XF or ST 4946 GLB2.Concurrently, PHY 444 WRF was characterized by the longest UHML, lowest micronaire, and greatest uniformity.According to USDA-AMS data, the fiber quality properties observed from DP 1522 B2XF, PHY 312 WRF, and ST 4946 GLB2 in both the 69-site-year and 85-site-year dataset closely mirror mean bale quality data reported by USDA-AMS for the region (USDA-AMS Cotton and Tobacco Program, of site-year, data were pooled across cultivar and site-year was analyzed as a fixed effect.For analysis of cultivar, data were pooled across site-years and cultivar was analyzed as a fixed effect.Separations were calculated using a one-way ANOVA and conventional LSD post hoc analysis at a = 0.05.To determine the variability in each parameter of interest attributed to genotype and environment, the sum of squares for cultivar and site-year reported by a combined ANOVA were each divided by the total model sum of squares to calculate percentage sum of squares.The role of each main effect was then reported as a percentage.Each of these procedures (ANOVA, LSD, and percentage sum of squares) was calculated for the 85-site-year dataset containing three common cultivars, the 69-site-year dataset containing the three common cultivars plus the longer UHML, lower micronaire cultivar, and the same 69-site-year dataset containing all combinations of three cultivars.Results from the 85-and 69-site-year analyses excluding the longer UHML, lower micronaire cultivar were compared to provide insight into the stability of environment.Results from the 69-site-year analyses consisting of all three-cultivar combinations and the four-cultivar combination were compared to provide insight into the stability of genotype within the dataset.
Cultivar stability across environments was assessed by plotting cultivar performance at a given site-year against a generated environmental index, which equaled the average of each analyzed parameter within site-year minus the average of each analyzed parameter across all site-years (Eberhart and Russell, 1966).Slopes were compared through an analysis of covariance.Cultivar was considered the main effect and environmental index was the covariate.A p value <0.05 for the interaction term (cultivar ´ environmental index) indicated significant differences in slope between two cultivars.

RESULTS AND DISCUSSION
Cultivar Effect on Lint Yield, Lint Percentage, and Fiber Quality Cultivar means and separations across 85 site-years containing DP 1522 B2XF, PHY 312 WRF, and ST 4946 GLB2 indicated significant cultivar effects on lint percentage, length, and uniformity; however, cultivar means for each of these parameters varied only slightly (Table 3).Cultivars DP 1522 B2XF and PHY 312 WRF had significantly greater Environmental Effect on Lint Yield, Lint Percentage, and Fiber Quality Site-year means calculated across 85 site-years containing the three common cultivars DP 1522 B2XF, PHY 312 WRF, and ST 4946 GLB2 indicated that significant environmental effects existed for all parameters of interest, and the range for each of these parameters was large (Fig. 2).
For example, average lint yield ranged from 1665 kg ha −1 at the Marianna, AR, location in 2015 to 633 kg ha −1 at the Brownsville, TN, location in 2016.Average lint percentage, micronaire, and strength were also characterized by large value ranges.Distribution curves of lint yield, lint percentage, micronaire, length, strength, and uniformity can be found in Fig. 2.Although most data appear to follow the expected normal distribution trends, two clear peaks are noted within the lint percentage histogram.Although the Missouri, Mississippi, and Tennessee locations process their seed cotton samples on 20-saw microgins that contain seed cotton cleaning and lint cleaning equipment, Arkansas, Alabama, and Louisiana process their samples on small tabletop gins.
Although other fiber quality parameters correlate strongly across commercial, microgin, and tabletop gin types, lint percentage from tabletop gins is typically greater than from commercial or microgins; these trends can be observed in testing programs that process samples on both tabletop and microgin equipment, such as the Mississippi State On-Farm and Official Variety Trial programs (Dodds et al., 2016a(Dodds et al., , 2016b)).Due to the similarities noted within a given site-year between cultivar lint percentage, the ?5% spread between these peaks, and the similarities between the magnitudes of each peak relative to the number of site-years using microgins vs. tabletop gins, the split peaks noted within the lint percentage distribution curve are hypothesized to be a function of gin type.

Genotypic and Environmental Impacts
Noted from Three Cultivars within 85 and 69 Site-Years Analysis of 85 site-years containing three common cultivars-DP 1522 B2XF, PHY 312 WRF, and ST 4946 GLB2-indicated that environment was the dominant factor influencing lint yield (87.0%), lint percentage (90.7%),micronaire (69.7%), length (69.7%), strength (47.0%), and uniformity (70.1%) (Table 4).The role of genotype was important in determining micronaire (18.0%) and strength (23.8%).Analysis of 69 site-years consisting of the same three common cultivars-DP 1522 B2XF, PHY 312 WRF, and ST 4946 GLB2-resulted in very similar results; the largest percentage change when comparing the 85-site-year analysis with the 69-site-year analysis was a 2.2% reduction in the role of environment in explaining lint percentage.All other changes were small and ranged from 0.2 to 1.6% (Table 4).
The percentage sum of squares noted from analysis of both the 85-and 69-site-year datasets containing the three common cultivars, DP 1522 B2XF, PHY 312 WRF, and ST 4946 GLB2, generally match results previously reported by Snider et al. (2013) and Kerby et al. (2000) (Table 1).The one substantial difference between the three datasets can be found within the lint percentage column; Snider et al. (2013) noted that environment only contributed 38.8% of lint percentage and cultivar contributed 51.5%.In contrast, Kerby et al. (2000) reported that environment and cultivar described 82 and 6% of the variability in lint percentage, respectively.Although results from Kerby et al. (2000) closely match those calculated here, results from Snider et al. (2013) are different.Two factors are hypothesized to have driven these differences.First, the range of lint percentages of the cultivars analyzed by Snider et al. (2013) was larger than the range of lint percentages of the cultivars analyzed within this project.As the range of lint percentages among the cultivars increase, the variability in lint percentage captured by the genotype term will inherently increase.The second factor that likely contributed to the increased role of genotype and limited the role of environment on lint percentage was ginning procedures.Although all data analyzed by Snider et al. (2013) were produced by the University of Georgia MicroGin in Tifton, GA, data analyzed by Kerby et al. (2000) and this project were ginned across various microgins and numerous tabletop gins.Variability due to ginning type would be introduced into the environment term and has possibly reduced the role of genotype in determining lint percentage within this analysis.
Slight differences noted between the 69-and 85-siteyear analyses closely resemble observations by Kerby et al. (2000) when they reanalyzed a subset of their data to find no substantial shifts in calculated contributions.Similarities between the two datasets analyzed in this project suggest that the 69-site-year subset is large enough to capture desired variability in environment and is appropriate to evaluate the defined objectives.

Genotypic and Environmental Impacts Noted from Four Cultivars within 69 Site-Years
Across all 69-site-year, three-and four-cultivar datasets, the inclusion of the longer UHML, lower micronaire cultivar PHY 444 WRF resulted in substantial increases in the percentage sum of squares explained by genotype for length and micronaire (Table 5).Compared with the Table 4. Percentage sum of squares explained by genotype and environment for lint yield, lint percentage, micronaire, length, strength and uniformity.Sum of squares were calculated for each parameter by ANOVA of three cultivars across 85 site-years and three cultivars across 69 site-years.

Avg. percentage sum of squares Source
Lint yield Lint percentage Micronaire Length Strength Uniformity three-cultivar dataset excluding PHY 444 WRF, the four-cultivar dataset captured a decrease in the role of environment in determining micronaire and length of 23.3 and 31.4%,respectively.Subsequently, the role of genotype in determining micronaire and length increased by 26.0 and 37.6%, respectively.Changes in the roles of genotype and environment in determining lint yield, lint percentage, and uniformity were also observed, but these shifts were slight (ranging from 0.1 to 6.4%).
To determine if the shifts in percentage sum of squares explained by genotype and environment for length and micronaire could be partially attributed to an increase in the number of cultivars present in the four-cultivar analysis as opposed to the three-cultivar dataset excluding PHY 444 WRF, the percentage sum of squares for three additional datasets containing all possible combinations of PHY 444 WRF plus the other two cultivars were also calculated (Table 5).Differences between the four-cultivar percentage sum of squares and each three-cultivar percentage sum of squares are reported in parentheses in Table 5.All threecultivar analyses containing PHY 444 WRF resulted in very similar percentage sum of squares as the four-cultivar dataset; with the exception of the analysis excluding ST 4946 GLB2, the largest noted difference between any two percentage sum of squares was 6.8.A substantial increase in the role of environment and decrease in the role of genotype in determining strength was observed when ST 4946 GLB2 was excluded from the analysis (Table 5).Strength of ST 4946 GLB2 was significantly greater than that of any other tested cultivar, including PHY 444 WRF (Table 3).Subsequently, it is logical that the removal of ST 4946 GLB2 reduced the role of genotype and increased the role of environment in determining strength.
The large, consistent increases in the percentage sum of squares explained by genotype for length and micronaire noted from the three-cultivar dataset excluding PHY 444 WRF to all other datasets containing PHY 444 WRF (both three-and four-cultivar datasets) suggest that the differences observed are not due to an inherent increase in the number of cultivars analyzed and instead appear to be directly linked to the introduction of PHY 444 WRF.
The large influence of genotype on length was noted previously by Campbell and Jones (2005), Campbell et al. (2011), andMeredith (2003), but these analyses were not focused solely on commercial germplasm.Neither Campbell and Jones (2005), Campbell et al. (2011), nor Meredith (2003) noted a large influence of genotype on micronaire.When results from this analysis are compared with those of Snider et al. (2013) and Kerby et al. (2000), analyses that were focused on commercial germplasm, it is clear that the large role of genotype in determining micronaire and length has not been previously captured within an assessment evaluating a limited number of commercial cultivars across a production region.

Stability Analysis
Linear regressions of yield, lint percentage, and fiber quality across environmental index graphed by cultivar from the four-common-cultivar, 69-site-year dataset are displayed in Fig. 3. Statistical separations of cultivar slope were uncommon; no statistical differences in slope were noted between cultivars regressed across environmental Table 5. Percentage sum of squares explained by genotype and environment for lint yield, lint percentage, micronaire, length, strength, and uniformity.Sum of squares were calculated for each parameter by ANOVA of four cultivars across 69 site-years and every combination of three cultivars across 69 site-years.

Source
Lint yield Lint percentage Micronaire Length Strength Uniformity --------------------------------------% --------------------------------------   indices calculated for lint yield, lint percentage, micronaire, strength, or length (Table 6).In contrast, significant differences in slopes were noted for uniformity when cultivar means were regressed across environmental indices.It is possible that gin type may have played a role in this interaction term; the propensity of a fiber to break could potentially vary by cultivar and gin type.Failure to identify interactions in other parameters of interest is likely due to the limited number of cultivars evaluated within this analysis and the emphasis each seed company places on identifying and advancing widely adaptable cultivars.Although values for many parameters vary by cultivar at a given site-year (visualized by differences in intercepts in Fig. 3), these separations are generally consistent for all parameters but uniformity.
Results from this research contrast the results from Snider et al. (2013), who captured a significant G´E response within lint yield, lint percentage, and all included fiber quality parameters of interest.Similarly, results from this research also contrast results from Campbell et al. (2012), who also noted significant G´E terms for lint yield, lint percentage, and all included fiber quality parameters, in addition to boll number, boll size, and seed index.In comparison with Campbell et al. (2012) and Snider et al. (2013), lack of G´E significance noted within this analysis may be partially explained by the differences in size of the geographic regions and the subsequent impacts on included cultivars.Results noted from this analysis were generated by evaluating common cultivars that were entered into five different state cultivar testing programs across a relatively large geographic area.It is fair to assume the seed companies that entered these cultivars believed that these cultivars were widely adaptable.By selecting common cultivars across cultivar testing programs, the dataset analyzed here may have excluded "niche" cultivars-cultivars that perform very well within a specific environment but may not perform well across all environments-and reduced the impact of G´E.By simply selecting common cultivars across such a large region, the selected cultivars for this analysis were likely to have a limited response to G´E.In contrast, the Snider et al. (2013) dataset analyzed common cultivars that were included within the state of Georgia.The focus within one state likely resulted in inclusion of "niche" along with widely adaptable cultivars.Similarly, the analysis by Campbell et al. (2012) included a very diverse group of 82 individual cultivars and breeding lines, which also likely included "niche" cultivars.

CONCLUSIONS
Defining the roles of genotype, environment, and G´E on lint yield, lint percentage, and fiber quality from commercial germplasm has immediate, practical applications for producers attempting to maximize returns.Results from this analysis indicate that the introduction of a longer UHML, lower micronaire cultivar within the Mid-South will cause substantial increases in the role of cultivar in determining micronaire and length.Furthermore, when selecting modern, commercial cotton cultivars for production within the Mid-South, the relatively small role of cultivar in determining lint yield and turnout suggests that a strong emphasis should be placed on fiber quality data.Analysis of environmental stability data indicated that all tested cultivars were equally adaptable with one exception, since statistical separations were noted with uniformity.The relatively minor role of cultivar in determining lint yield and the substantial role of cultivar in determining micronaire and length suggest that producers within the Mid-South should begin to place more value on fiber quality data when selecting among high-yielding cultivars.

Fig. 3 .
Fig. 3. Linear regressions of four commercial cotton cultivars on an environmental index calculated from the mean of all four like cultivars at each of the 69 site-years minus the overall mean for (A) lint yield, (B) lint percentage, (C) micronaire, (D) fiber length, (E) fiber strength, and (F) uniformity index.

Table 1 .
Percentage sum of squares explained by genotype and environment for lint yield, lint percentage, and fiber quality parameters of micronaire, length, strength, and uniformity reported by previous studies evaluating large-plot data.

Table 2 .
Characteristics of large plot cultivar trials conducted within the Mid-South listed by state during the 2015 and 2016 seasons.
Fig. 1.Map of the US Mid-South.The 85 locations containing four common cultivars during the 2015 and 2016 seasons are marked with red stars.
B2XF, PHY 312 WRF, and ST 4946 GLB2 across 85 yield environments 1522 B2XF, PHY 312 WRF, PHY 444 WRF, and ST 4946 GLB2 across 69 yield environments Numbers in parentheses represent the change of percentage sum of squares from the four-cultivar dataset to each three-cultivar dataset.

Table 6 .
Environmental stability of agronomic and fiber quality parameters for four commercially available cotton cultivars.A given slope represents the responsiveness of each parameter to environment relative to the other cultivars included within the analysis.