Comparative Upland Cotton Fiber Length Measurement and the Relation to Fiber Maturity

Abstract

Cotton fiber length and maturity, two critical fiber qualities, are commonly determined in the U.S. by Uster high volume instrument (HVI) and advanced fiber information system (AFIS). The main objectives of this investigation were to compare how HVI lengths agree with AFIS lengths and to examine whether the fiber length is linked with fiber maturity between the Universal HVI length calibration cotton standards and diverse upland lint samples. HVI micronaire (MIC) and AFIS fineness showed insignificant differences from HVI length calibration cotton standards to lint samples. Although there were strong and significant correlations between HVI upper-half mean length (UHML) and either AFIS UQL (w) or AFIS L5% (n), the relationship between UHML and L5% (n) was better suited than between UHML and UQL (w) in scrutinizing fiber lengths. Meanwhile, analysis revealed a moderate correlation between AFIS L5% (n) length and AFIS maturity ratio (MR), indicating the possibility of improving AFIS L5% (n) length by regulating fiber MR development. Further, AFIS MR values were positive and moderate correlated with algorithmic M_IR values of attenuated total reflection Fourier transform infrared (ATR FT-IR) spectra. The results suggested the feasibility of the ATR FT-IR method along with M_IR analysis in estimating AFIS MR rapidly away from fiber testing laboratories.

1. Introduction

Cotton fiber length and closely related properties, such as short fiber content (SFC), fiber length uniformity, and distribution, are important fiber quality attributes, as they impact fiber processing performance and finished yarn and fabric quality substantially [1,2,3,4,5,6,7] and also affect the marketable value of cotton fiber greatly [8,9]. One of other essential fiber qualities is fiber maturity, which is a vital cotton yield index and has a direct impact on fiber breakage and entanglement (neps) during mechanical processing, dye uptake in yarn and fabric products, as well as yarn processing and textile performance [10,11,12,13,14].

Representatives of commercially available and accepted instruments for simultaneous fiber maturity and length measurements in the U.S. are the Uster high volume instrument (HVI) and advanced fiber information system (AFIS) [9,12,15,16]. HVI reports the fiber micronaire (MIC, an indicative of fiber maturity and fineness and also a substitute for fiber maturity in textile mills), upper-half mean length (UHML), length uniformity index (UI), strength, elongation, color (color grade, Rd, +b), short fiber index (SFI), trash, and maturity properties from testing a bundle of fibers. Comparatively, AFIS generates 20 fiber quality parameters that consist of fiber lengths and their distributions by weight and by number, trash content, nep content, fineness, and maturity by measuring the individualized fibers. Appendix A summarizes the description of HVI and AFIS properties briefly. Differences in the principle of measurements between HVI and AFIS leads to a report of average quality parameters by HVI in a sample, in contrast to a report of both average quality properties and their distributions by AFIS. The HVI measurements are calibrated using standard materials for specific fiber qualities (i.e., MIC, UHML, and color) in a sample, while the AFIS is developed as a direct measurement of individual fiber property and distribution. Compared to the HVI, the AFIS instrument necessitates the operator in preparing the fiber sliver in advance and has a slower testing speed. HVI and AFIS data have been utilized considerably in the cotton industry through cotton research and breeding programs.

Enormous and comprehensive investigations have been performed in fiber length measuring system and algorithm development as well as in measured length comparisons among different systems with different measurement principles [4,16,17,18,19,20]. For example, Kelly et al. [16] connected two primary length parameters on 10,084 commercial bales and reported a R² of 0.3461 between HVI UHML and AFIS L (n), and they contributed a lower R² to such factors as AFIS L (n), measuring all fibers while HVI did not measure short fibers. van der Sluijs et al. [20] compared three traditional cotton fiber length measurement methods (HVI, AFIS, and Suter-Webb array (SWA)) and their significance on a set of ten upland and saw-ginned cotton samples with a wide UHML ranging from 24.28 to 32.38 mm (or 0.956 to 1.275 in), and observed that there were statistically strong positive relationships (Pearson correlation coefficient (R) = 0.923 to 0.982) between HVI UHML and the three AFIS lengths (L (w), L (n), and L5% (n)), without the AFIS UQL (w) that was strong and positive (R = 0.971) but was not statistically significant. Distinctive differences in R² and R might be due to different cotton sample origins (i.e., commercial bales [16] vs. saw-ginned samples with a wide UHML range [20].

Besides HVI and AFIS methods, other current-in-use fiber maturity determinations comprise the cross-sectional image analysis microscopy (IAM) [21,22,23] and Cottonscope [24,25,26,27]. The IAM method is developed to be a direct and reference fiber maturity measurement for individual cross-sectional fibers, and the maturity values were taken to validate indirect measurement systems, including AFIS and Cottonscope. The IAM method is relatively low cost and accessible at fiber research sites, but the whole process is labor-consuming and the result is influenced by operators’ experience, hence it is unsuitable for analyzing hundreds of fiber samples efficiently. Cottonscope is relatively rapid to acquire fiber maturity by imaging fiber snippets with polarized microscopy, but the system is limited to several fiber laboratories. To perform fiber maturity measurements by any methods (HVI, AFIS, IAM, and Cottonscope), the samples have to be shipped into the fiber testing laboratories. Hence, there have been considerable studies of indirect cotton fiber maturity assessment using near infrared (NIR) and attenuated total reflection Fourier transform infrared (ATR FT-IR) spectroscopy combined with partial least squares (PLS) and simple algorithmic analysis [28,29,30,31,32,33,34]. The rationale arises from cellulosic NIR intensities (750 to 2500 nm or 13,300 to 4000 cm^–1) that originate from the overtones and combinations of fundamental group vibrations [35], and also from FT-IR absorptions (2500 nm to 25,000 nm or 4000 to 400 cm^–1) that represent those fundamental group vibrations (i.e., C–H, O–H, C–O) in fiber cellulose (a major constituent in cotton fiber) [28,36].

Unlike the HVI lengths, which depend on the standard samples for calibrating the HVI length measurement, the AFIS measures individual fiber length directly. Previous investigations have shown the varying degree of correlations among diverse sample sets between HVI lengths and AFIS lengths [16,20]. The purposes of this study were (1) to evaluate the consistency of correlations between HVI and AFIS lengths for the Universal HVI length calibration cotton standards and actual upland cotton samples representing different crop years, grown locations, and cultivars; (2) to correlate the fiber length with fiber maturity information; and (3) to determine whether the M_IR values obtained by ATR FT-IR method can predict fiber maturity independent of instrumental calibration effects. Comparison of HVI length calibration standards with ginned regular cotton fibers might disclose the degree to which length variation among regular fibers may occur, and further, strengthen the usefulness and effectiveness of either HVI or AFIS in monitoring the fiber length and maturity variation between fiber samples. AFIS method is an option for limited amounts of fibers from cotton fiber researchers, since it requires a minimum 0.5 g fiber sample that could be accessible via a single cotton boll or locule sampling, whereas HVI requires a minimum 10.0 g sample that contains multiple cotton boll fibers. The novelty of this investigation exists in substantiating a reasonable relationship between HVI UHML and AFIS L5% (n) from two diverse fiber sets (HVI length calibration cotton standards vs. lint samples), verifying the importance of fiber maturity on fiber length for optimizing fiber quality improvement, and highlighting the potential of ATR FT-IR technique in sensing fiber maturity rapidly and non-destructively from the lint to seed cottons and also in cotton fiber phenotyping.

2. Materials and Methods

2.1. HVI Length Calibration Cotton Standards and Their HVI and AFIS Quality Determination

A total of 9 Universal HVI calibration cotton standards known as 8 × 8 samples (Sample ID 30 through 38) were produced by the USDA’s Agricultural Marketing Service (USDA-AMS). HVI qualities were tested by an HVI 1000 (USTER Technologies Inc., Knoxville, TN, USA) with five replications per sample, and AFIS values were determined by an AFIS Pro 2 (USTER Technologies Inc., Knoxville, TN, USA) with three replications of 5000 fibers per measurement. Mean HVI or AFIS qualities were calculated for each sample and used in the analysis. These tests were performed at the Southern Regional Research Center of USDA’s Agricultural Research Service (USDA-ARS-SRRC) routinely, following the samples’ conditioning at a standard environment of 21 ± 1 °C temperature and 65 ± 2% relative humidity for a minimum 24 h.

2.2. Seed Cottons, Laboratory Saw Ginning, and Fiber HVI and AFIS Quality Determination

Twenty-five commercial upland seed cotton samples (all Gossypium hirsutum species), harvested by cotton mechanical harvesters, were from 5 crop years (2016–2022), two U.S. states (Mississippi and New Mexico) and 21 cotton cultivars. None of these 25 samples were from the same cultivar in the same year or the same location. Samples, selected based on their availability, were from five brands (i.e., 2 ALL-TEX/DYNA-GRO, 3 Americot, 5 BASF, 4 Deltapine, and 11 Phytogen) and 5 crop years (i.e., 7 in 2016, 2 in 2018, 6 in 2020, 6 in 2021, and 4 in 2022). They were kept at laboratory conditions for least 48 h prior to the ginning.

After separating evident non-lint cotton trash (for example, large sticks and burrs), seed cotton (250.0 g per sample) was ginned by utilizing a 10-saw Dennis laboratory saw gin (5 inch saw diameter, 137 teeth per saw, and 497 revolutions per minute (RPM) saw speed; Dennis Manufacturing, Athens, TX, USA) within 3 min. All lint samples were retained for HVI and AFIS test in the same procedure as HVI length calibration cotton standards.

2.3. Fiber ATR FT-IR Spectral Collection and Interpreation

A Nicolet iS20 FT-IR spectrometer (Thermo Electron North America LLC, Madison, WI, USA) equipped with a KBr/Ge coated beamsplitter, a deuterated triglycine sulfate (DTGS) detector, and a Smart iTX^TM attenuated total reflection (ATR) attachment was used to scan all fiber spectra. The use of the ATR device allows fiber sampling easier and simpler than the KBr-pellet method in FT-IR transmission mode; therefore, ATR method is time effective upon processing many fiber samples. The sampling depth of the ATR device ranges 2 through 15 µm [37,38], while the thickness of mature cotton fiber secondary cell wall (SCW) area fluctuates from 2 to 7 µm [39], so that the ATR method can reflect chemical characteristics inside mature cotton fibers by employing a low refractive index crystal (i.e., diamond). The spectral background was taken ahead of scanning the samples. The spectra were collected from the 4000 to 600 cm⁻¹ region at 4 cm⁻¹ and 16 scans in the absorbance mode. Attention was necessary to ensure the complete coverage of the ATR window (2 mm in diameter) by fiber samples only. Six ATR FT-IR spectra (given in Figure 1a) were taken for each sample at different sub-samplings. Subsequent spectral analyses for fiber infrared maturity (M_IR) values were conducted by the proposed simple algorithm [28,29], and the mean M_IR value of each sample was obtained. No spectral preprocessing (such as baseline correction and smoothing) was applied to these spectra in absorbance units, with identical steps to earlier ATR FT-IR investigation of cotton fibers for result consistency and comparison. In essence, after exporting the spectral data into Microsoft^® Excel^® for Office 365, both R₁ and M_IR equations were applied for calculating the R₁ value of an unknown sample and then for converting the R₁ value to the M_IR value:

R₁ = (I₉₅₆ − I₁₅₀₀)/(I₁₀₃₂ − I₁₅₀₀)

(1)

M_IR = (R₁ − 0.14)/0.45

(2)

where I₁₅₀₀, I₁₀₃₂, and I₉₅₆ are each an average of the band intensities in a narrow spectral region at a respective wavenumber. The I₁₀₃₂ represented the positive and large intensity variation in the bands near 1032 cm⁻¹ due to C-O stretching modes in cotton fibers, while the I₉₅₆ exhibited the negative and large intensity variation in the bands centered at 956 cm⁻¹ due to C-O stretching modes in cotton fibers in the difference spectrum between immature and mature cottons [28]. The I₁₅₀₀ was selected to offset two readings due to its minimum absorbance. The M_IR values were compared to fiber maturities derived from traditional image analysis (IA) and AFIS measurements [28].

2.4. Data Analyses

Regression analysis of any HVI or AFIS quality pair was carried out with the use of Microsoft^® Excel^® for Office 365. Statistical interpretation was executed by employing the analysis of variance (ANOVA) function under Data Analysis in Microsoft^® Excel^® for Office 365 with a 95% confidence level and also using analysis of covariance (ANCOVA) available from http://www.biostathandbook.com/ancova.html (accessed 16 June 2025). Other professional statistical software (e.g., R, SPSS, or SAS) is a consideration for large datasets with complex property traits.

3. Results and Discussion

3.1. HVI and AFIS Fiber Maturity and Length Variability Within HVI Length Calibration Cotton Standards and Lint Samples

Table 1. HVI and AFIS fiber maturity and length statistics between HVI length calibration cotton standards and lint samples.

Qualities		HVI Length Calibration Cotton Standards				Lint Samples				ANOVA p-Value
		Range	Mean	SD	CV%	Range	Mean	SD	CV%
HVI	MIC	4.08~4.53	4.28	0.17	3.9	2.52~5.45	4.25	0.61	14.3	0.87
	UHML, in	0.92~1.16	1.06	0.08	7.9	1.11~1.26	1.19	0.04	3.7	<0.0001
	Mean length (ML), in	1.16~1.39	1.31	0.08	5.8	1.34~1.53	1.43	0.05	3.6	<0.0001
	SFI, %	6.09~12.2	11.1	3.3	30.2	4.0~7.9	6.2	0.9	14.5	<0.0001
AFIS	L (w), in	0.80~1.00	0.88	0.07	8.2	1.00~1.16	1.06	0.04	3.6	<0.0001
	UQL (w), in	0.97~1.22	1.1	0.09	8.2	1.21~1.38	1.28	0.05	3.7	<0.0001
	SFC (w), %	8.7~12.2	13	2.8	21.7	4.6~9.6	6.7	1.3	19.8	<0.0001
	L (n), in	0.61~0.79	0.69	0.06	8.5	0.78~0.96	0.86	0.04	5	<0.0001
	L5% (n), in	1.10~1.38	1.25	0.1	7.8	1.35~1.57	1.43	0.06	4	<0.0001
	SFC (n), %	26.6~41.6	33.4	4.8	14.4	17.3~28.6	23	3.3	14.5	<0.0001
	Fineness, mTex	151~167	159	5.8	3.7	131~185	161	12	7.5	0.62
	IFC, %	6.7~8.9	8.1	0.7	8.7	5.3~9.3	6.9	0.8	11.8	<0.001
	MR	0.79~0.84	0.83	0.03	3.2	0.79~0.94	0.88	0.03	3.4	<0.001

compares the descriptive statistics of HVI and AFIS maturity and length between HVI length calibration cotton standards and 25 lint samples representing 21 cotton cultivars grown in 5 crop years and two U.S. states. HVI mean length (ML), estimated from UHML and uniformity index (UI) in HVI testing report, was also included in Table 1. Since HVI length calibration cotton standards were selected to maximize a broad range of HVI UHML variation from commercial cotton bales and used as HVI length calibration cottons, their length properties varied more than maturity properties, as indicated by a large coefficient of variation (CV%). For example, CV% changed greatly from 5.8 to 30.2 for length properties (UHML, ML, SFI, L (w), UQL (w), SFC (w), L (n), L5% (n), and SFC (n)) but from 3.2 to 3.9 for three of four maturity properties (MIC, fineness, and MR) without the IFC (CV% = 8.7). In contrast, for lint samples, CV% were greater for three maturity properties (MIC, fineness, and IFC) ranging from 7.5 to 14.3 than for length properties of 3.6 to 5.0 excepting three short fiber values (SFI, SFC (w) and SFC (n); CV% = 14.5 to 19.8). From HVI length calibration cotton standards to lint samples, there were statistically significant increases for UHML, ML, L (w), UQL (w), L (n), L5% (n), and maturity ratio (MR) but also statistically significant decreases for SFI, SFC (w), SFC (n), and IFC. In other words, UHML, ML, L (w), UQL (w), L (n), L5% (n), and MR were smaller among HVI length calibration cotton standards than among lint samples, while SFI, SFC (w), SFC (n), and IFC were greater in HVI length calibration cotton standards than in lint samples. This is expected, since the longer and more mature the fibers, the less the short fiber values (SFI, SFC (w) and SFC (n)) and also the immature fibers (IFC). In particular, MIC and fineness did not change significantly from HVI length calibration cotton standards to lint samples (p-value = 0.62 to 0.87), hinting that MIC and fineness might not affect fiber length properties in general.

3.2. Correlations of Fiber HVI Against AFIS Quality Between HVI Length Calibration Cotton Standards and Lint Samples

Table 2. Statistics from linear correlations between HVI and AFIS quality for HVI length calibration cotton standards and lint samples.

Samples 1	HVI Quality	AFIS Quality	Slope	Adjusted Intercept 2	R2	ANOVA p-Value	ANCOVA p-Value
HVI length standards	UHML	UQL (w)	1.07	−0.02	0.96	<0.0001	0.59 (Slope)
Lint samples			1.02	0.05	0.89	<0.0001	0.03 (Intercept)
HVI length standards	UHML	L5% (n)	1.16	0.04	0.97	<0.0001	0.81 (Slope)
Lint samples			1.19	0.01	0.93	<0.0001	0.58 (Intercept)
HVI length standards	ML	L (w)	0.84	−0.19	0.76	<0.01	0.06 (Slope)
Lint samples			0.46	0.33	0.40	<0.001	<0.0001 (Intercept)
HVI length standards	ML	L (n)	0.51	0.02	0.44	0.053	0.16 (Slope)
Lint samples			0.15	0.56	0.03	0.39	<0.0001 (Intercept)
HVI length standards	SFI	SFC (w)	0.68	0.46	0.65	<0.01	0.64 (Slope)
Lint samples			0.52	0.54	0.14	0.069	<0.001 (Intercept)
HVI length standards	SFI	SFC (n)	0.97	1.03	0.43	0.054	0.81 (Slope)
Lint samples			1.14	1.09	0.10	0.12	<0.01 (Intercept)
HVI length standards	MIC	Fineness	15.60	1.13	0.20	0.23	0.79 (Slope)
Lint samples			18.33	1.01	0.86	<0.0001	0.15 (Intercept)
HVI length standards	MIC	IFC	−1.15	2.10	0.07	0.48	0.92 (Slope)
Lint samples			−1.04	2.04	0.60	<0.0001	<0.0001 (Intercept)
HVI length standards	MIC	MR	0.06	0.22	0.17	0.28	0.56 (Slope)
Lint samples			0.04	0.28	0.61	<0.0001	<0.0001 (Intercept)

¹ HVI length standards meant for HVI length calibration cotton standards. ² Adjusted intercept was calculated by dividing the intercept from each linear regression line with the average of mean x-axis and y-axis.

recaps the correlations between HVI and relevant AFIS qualities from HVI length calibration cotton standards to lint samples. It is of great interest to compare HVI UHML with AFIS lengths, because HVI UHML estimates the fiber length from the fibrogram by scanning a fiber beard where some short fibers may not be scanned, while AFIS lengths reflect the complete within-sample distribution of all single fibers by weight and by number. Like HVI UHML that characterizes the average length by number of the longer half of the fibers by weight, AFIS UQL (w) evaluates the fiber lengths of the longer 25% of all fibers by weights, and L5% (n) determines the fiber lengths of the longer 5% of all fiber by number. As anticipated, HVI UHML exhibited great and significant correlations with UQL (w) and L5% (n) for the HVI length calibration cotton standards (R² = 0.96 to 0.97, p-value < 0.0001), along with the slope value of 1.07 to 1.16 and the absolute adjusted intercept (or |adjusted intercept|) of <0.05. This pattern remained among the lint samples with the large R² of 0.89 to 0.93 (p-value < 0.0001), the slope of 1.02 to 1.19, and |adjusted intercept| of ≤0.05. Higher R² (=0.96 to 0.97) between UHML and UQL (w), as well as between UHML and L5% (n), for the HVI length calibration cotton standards agreed with R = 0.971 to 0.982 for 10 upland fibers with a broad UHML range of from 0.956 to 1.275 in [20].

Based on two correlation lines between the HVI length calibration cotton standards and the lint samples, analysis of covariance (ANCOVA) underlined the significant difference in intercept for the pair of UHML vs. UQL (w) (p-value = 0.03) but insignificant differences in both slope and intercept for the pair of UHML vs. L5% (n) (p-value = 0.58 to 0.81). As an example, Figure 2a,b depicts two correlation lines of HVI UHML vs. AFIS UQL (w) and of HVI UHML vs. AFIS L5% (n) between HVI length calibration cotton standards and lint samples, respectively. There were two samples with an UHML of 1.11 and 1.26 showing relatively great differences between UHML and UQL (w) in Figure 2a or between UHML and L5% (n) in Figure 2b, and an initial examination did not relate these two samples with specific crop year, location, or cultivar, and also HVI or AFIS measurement error. However, the differences estimated by subtracting UQL (w) from UHML (i.e., UHML–UQL (w)) were found to be correlated with AFIS MR negatively and moderately (R² = 0.42), as well as with IFC positively and moderately (R² = 0.36), but with fineness and MIC weakly (R² < 0.10). Apparently, more mature fibers (higher MR and lower IFC) tended to show lower UHML or higher UQL (w) values than less mature fibers. Linking the differences between UHML and L5% (n) values with MR, IFC, fineness, and MIC showed similar patterns, but the differences (as UHML–L5% (n)) presented a reduced R² with MR (R² = 0.23), IFC (R² = 0.17), and also fineness and MIC (R² < 0.05). Hence, fiber MR, IFC, fineness, and MIC might have a less impact on the differences (as UHML–L5% (n)) than the differences (as UHML–UQL (w)). Along with the consideration of ANCOVA test, fiber maturity might impact less on the UHML vs. L5% (n) relationship than on the UHML vs. UQL (w) relationship. This finding implied the validity of measuring fiber UHML or L5% (n) by two independent systems and also may enable the conversion of UHML to L5% (n) by the formula of L5% (n) = 1.20 × UHML − 0.00 from all samples as given in Figure 2b. Confidence intervals for the slope and intercept of this equation with 95% confidence level were from 1.11 to 1.29 and from −0.11 to 0.09, respectively.

HVI ML had moderate correlations with AFIS L (w) and L (n) for the HVI length calibration cotton standards (R² = 0.44 to 0.76), accompanied by the slope range of 0.51 to 0.84 and the |adjusted intercept| of 0.02 to 0.19. The correlations became worse for the lint samples with reduced R² (0.03 to 0.40) and slope (0.15 to 0.46) as well as enlarged |adjusted intercept| (0.33 to 0.56). Regardless of either the HVI length calibration cotton standards or the lint samples, the ML showed significant correlations with L (w) (p-value < 0.01). The R² (=0.44 to 0.76) between ML and L (w) as well as between UHML and L (n) for the HVI length calibration cotton standards were slightly lower than the R = 0.823 to 0.923 for 10 upland fibers with a wide UHML range [20]. ANCOVA test suggested significant distinctions in intercept for both pairs (ML vs. L (w) and ML vs. L (n)). The correlation lines of HVI ML vs. AFIS L (w) between HVI length calibration cotton standards and lint samples are compared in Figure 3.

Unlike moderate relationships between HVI SFI and AFIS SFC (w) or SFC (n) for the HVI length calibration cotton standards (R² = 0.43 to 0.65), there were weak correlations for the lint samples (R² = 0.10 to 0.14). The slope ranges of SFI against SFC (w) or SFC (n) for the HVI length calibration cotton standards (0.68 to 0.97) were similar to those for the lint samples (0.52 to 1.14), and the |adjusted intercept| for the HVI length calibration cotton standards (0.46 to 1.03) resembled that for the lint samples (0.54 to 1.09). There existed a significant correlation between SFI and SFC (w) for the HVI length calibration cotton standards (p-value < 0.01). The R² (=0.43 to 0.65) between SFI and SFC (w) or SFC (n) for the HVI length calibration cotton standards concurred with R = 0.655 to 0.777 for 10 upland fibers varying in UHML [20]. ANCOVA analysis revealed statistically significant differences in intercept for both pairs of SFI vs. SFC (w) and SFI vs. SFC (n). One of the three short fiber parameters tended to decrease for higher MR samples, as fibers with greater maturity have larger secondary cell wall areas (or coarse fibers usually) and are not broken easily compared to lower MR samples during ginning and processing.

HVI MIC showed weaker and insignificant correlations with fineness, IFC, and MR for the HVI length calibration cotton standards (R² = 0.07 to 0.20, p-value = 0.23 to 0.48) than for the lint samples (R² = 0.60 to 0.86, p-value < 0.0001). This is likely due to less variation in fiber MIC for the HVI length calibration cotton standards (Table 1). Interestingly, there were insignificant differences in intercept and slope (p-value = 0.15 to 0.79) for the pair of MIC vs. fineness (Figure 4) and only in slope (p-value = 0.56 to 0.92) for the two pairs of MIC vs. IFC and MIC vs. MR between HVI length calibration cotton standards and lint samples.

3.3. Relationship Between Fiber Length and Maturity

Since the L5% (n) length manifested an excellent correlation with UHML from the HVI length calibration cotton standards to the lint samples in Figure 2b, L5% (n) length was further examined with AFIS maturity parameters (fineness, IFC, and MR) and HVI MIC (a common substitute for fiber maturity by textile manufacturers). Contrary to insignificant correlations between L5% (n) and fineness or MIC (R² = 0.02 to 0.04, p-value = 0.26 to 0.45) for combined HVI length calibration cotton standards and lint sample set, there were weak but statistically obvious relationships between L5% (n) and IFC (R² = 0.18, p-value = 0.01) as well as between L5% (n) and MR (R² = 0.22, p-value < 0.01). Figure 5 and Figure 6 show the relationships for the pairs of L5% (n) vs. IFC and L5% (n) vs. MR, respectively. A R² of 0.18 to 0.22 revealed that AFIS IFC and MR explained about 18 to 22% of the variation in AFIS L5% (n), which is reasonable since fiber length is largely determined by the genotype, growth environment, and mechanical operations in harvesting and ginning that might break longer fibers into short ones [4], whereas fiber IFC and MR are chiefly affected by the environmental factor.

In Figure 5, an encircled two-lint sample cluster (L5% (n) = 1.49) and another encircled two-lint sample group (L5% (n) = 1.35 to 1.40) are apparently away from a general regression line. These four samples were not any one of the two samples showing relatively great differences between UHML and UQL (w) or L5% (n) in Figure 2a,b. The first two-lint sample cluster with L5% (n) of 1.49 was expected to have an IFC < 8.0, but their measured IFC was 8.83 in average. Further analysis showed that these 2 samples had the smallest or close to the smallest fineness and MR within the 25 lint samples. In contradiction, the second two-lint sample group with L5% (n) of 1.35 to 1.40 was assumed to have an IFC around 7.0, but their actual mean IFC was 5.47. These two samples had the largest or close to the largest fineness and MR among the 25 lint samples. If these four samples were removed from the data set due to their evident deviations from the regression line in Figure 5, the regression line was improved to be R² = 0.51. This observation indicated that both more mature fibers (with greater fineness and MR) and less mature fibers (with lower fineness and MR) were prone to impact fiber length.

Similarly to Figure 5, Figure 6 reveals four identical samples being apart from a general tendency. The two-lint sample cluster with L5% (n) of 1.49 was projected to have a MR close to 0.85, but their actual mean MR was 0.81. As inserted in Figure 5, these two samples had the largest or close to the largest IFC and also the smallest or close to the smallest fineness and MR among the 25 lint samples. Meanwhile, the two-lint sample group with L5% (n) of 1.35 to 1.40 was forecasted to possess a MR around 0.85, but their real mean MR was 0.92. These two samples had the largest or close to the largest fineness and MR but also the smallest or close to the IFC among the 25 lint samples. If these four samples were not considered from the data set, the regression line was improved (R² = 0.51). Overall results in Figure 5 and Figure 6 suggested that either MR increase or IFC decrease is beneficial to L5% (n) length moderately, in addition to the influence of a complex fiber maturity and handling interactions on fiber length.

3.4. Inference of Fiber Maturity Sesnsing at Developmental Stage

As fiber maturity revealed a positive and moderate relationship with fiber length in Figure 6, there is a strong desire to determine fiber maturity well ahead of the application of harvesting aid for ideal fiber quality [40]. One of the approaches was to apply NIR instruments for fiber maturity in and outside of the laboratory [30,31,32,33]. In most of these studies, researchers mainly analyzed well-prepared and clean lint fibers with different fiber maturity references determined by the Fineness and Maturity Tester (FMT) method, cross-sectional IAM testing, and Cottonscope measurement [30,32,33]. Besides the lint fiber, Rodgers and colleagues acquired NIR spectra of individual clean and seed cotton boll fibers using portable NIR devices, before combining the boll fibers for maturity references [32,33]. Their studies addressed the potential of fiber maturity prediction on individual seed cotton bolls or on ginned lint from the seed cotton only after new samples were added to the original calibration set for re-developing new NIR calibrations models. Alternatively, the ATR FT-IR technique was reported for fiber maturity from seed cottons to ginned lint, as it is able to measure as little as 0.5 mg fiber directly (i.e., no need to remove any visible trash pieces and cotton seeds or to prepare fiber samples beforehand), to collect an individual sub-sample (or locule) rapidly within ~2 min, and further to reveal fiber maturity distribution [29,34]. As a concept of the study, Figure 7 shows a synchronous M_IR variation with AFIS MR increasing and also suggested the relationship of M_IR = 1.24 × MR − 0.26 with R² = 0.49 for the combined samples. A relatively low R² (=0.49) is consistent with earlier reports [25,28] because of such considerations as (i) AFIS MR is presented as the percentages of both mature fiber content (θ ≥ 0.50) and immature fiber content (θ < 0.25) from measuring the fiber shape at two different angels, whereas FT-IR M_IR is estimated by the algorithm of reflecting spectral intensity difference, and (ii) MR is sufficient for well-defined fibers but may be ineffective for regular cotton fibers after relating AFIS MR to Cottonscope or ATR FT-IR maturities. For example, Kim et al. [25] observed a weak correlation between Cottonscope and AFIS maturities on 550 multiparent advanced generation intercross (MAGIC) population samples and further revealed that the fiber blending process (or homogeneity) impacted the AFIS MR more than Cottonscope maturity. Liu et al. [28] reported a strong R² of 0.920 between AFIS MR and M_IR on six International Cotton Calibration (ICC) standards for MIC calibration but a much lower R² of 0.403 among 99 regular fibers, and they ascribed the low correlation to such factors as maturity variable fibers and sampling disparity between the two measurements. Meanwhile, ANCOVA suggested an insignificant difference in slope (p-value = 0.44) but a significant difference in intercept (p-value < 0.001) between the two sample sets. Strikingly, a conversion between AFIS MR and M_IR (MR = 0.39 × M_IR + 0.54) in Figure 7 is nearly the same as the equation between AFIS MR of ginned fibers and M_IR of 11 seed cottons (MR = 0.34 × M_IR + 0.62, R² = 0.30) described in a recent study by Liu and Delhom [29], despite different FT-IR instruments and cotton fiber sets being used between the two studies. Also, ranges of AFIS MR (0.80 to 0.94) and M_IR (0.676 to 0.900) from these seed cottons [29] were similar to those of AFIS MR (0.79 to 0.94) and M_IR (0.58 to 0.93) from lint fibers in Figure 7. A brief comparison between this study and a previous study [29] is summarized in

Table 3. Comparison of AFIS MR vs. FT-IR M_IR relationship between this study and a previous study [29].

Samples	This Study	Previous Study [29]
Number	34	11 (50 locule/sample)
MR range	0.79~0.94	0.80~0.94
MIR range	0.58~0.93	0.676~0.900
MIR vs. MR	MR = 0.39 × MIR + 0.54	MR = 0.34 × MIR + 0.62
R2	0.49	0.30

. Meanwhile, standard deviation (SD) in Figure 8 had a tendency to decrease, with the M_IR average increasing insignificantly for all samples (p-value = 0.12), or for HVI length calibration cotton standards (p-value = 0.09) and for lint samples (p-value = 0.13), indicating a larger variation in M_IR value within lower M_IR fibers than within higher M_IR fibers. This observation implied the effect of cotton genotype, environment (weather, location, and irrigation), and their interactions on fiber maturity. Overall, the observation in Figure 7 implied the capability of the algorithmic M_IR approach for an indirect AFIS MR estimation rapidly and non-destructively at the fiber developmental stage in remote/breeding locations away from a fiber testing laboratory. One application could unravel how agronomic practices or environmental stresses (e.g., waterlogging, nutrient management) impact fiber length and maturity. For example, a recent meta-analysis study indicated that waterlogging impacts cotton yield and fiber quality adversely [41]. Further investigation is necessary to understand how such environmental stresses interact with the length relationships reported by HVI and AFIS.

4. Conclusions

Being commercially available and accepted instruments for determining fiber length and maturity properties, both HVI and AFIS results were compared between universal HVI length calibration cotton standards and 25 lint samples of diverse upland cultivars. Compared to discernable differences in HVI and AFIS length and maturity from HVI length calibration cotton standards to lint samples, there existed insignificant differences in HVI MIC and AFIS fineness between the two fiber sets. One significant relationship between HVI qualities and relevant AFIS qualities was that HVI UHML was related with AFIS L5% (n) better than AFIS UQL (w), highlighting the consistency and effectiveness of measuring fiber length by HVI or AFIS. The analysis also revealed a moderate correlation between L5% (n) and MR and subsequently implied the feasibility of improving L5% (n) even a little by controlling MR development. Moreover, AFIS MR responded positively and moderately to algorithmic M_IR values from ATR FT-IR measurement of all samples, affirming the potential of the ATR FT-IR technique, in conjunction with the M_IR strategy, for a rapid AFIS MR estimation in remote/breeding locations. Since upland cotton has different genotypes and is grown in different environments across the world, the generalizability of this study might be expanded by extending to upland cottons produced from other geographical regions in the world.