Application of a published methodology for evaluating within-bale repeatability of measurements obtained with standardized instruments for testing cotton: Gossypium hirsutum in Eastern and Southern Africa

Abstract

The possible application of conclusions from a published study concerning cottons from West and Central Africa (WCA), involving an evaluation of the within-bale variability of fiber Micronaire, Length, Uniformity, Strength, Reflectance and Yellowness in cottons from Eastern and Southern Africa (ESA) was investigated.

We took eight cotton samples per bale from 240 bales produced by 32 ginning mills in six ESA countries in two crop seasons. Our representative sample comprised 1920 fiber samples that were centrally analyzed under controlled conditions using standardized instruments for testing cotton (SITC). We evaluated within-bale variability levels for both saw- and roller-ginned cottons and checked the applicability of the published conclusions to ESA.

We found that (1) sampling variance levels were comparable in ESA and in WCA for saw-ginned cottons, (2) WCA recommendations for saw-ginned cottons would also apply in ESA for most fiber characteristics measured by SITC, and (3) for roller-ginned cottons, the higher within-bale variability of roller-ginned cotton fibers compared to saw-ginned cotton would require the definition of a specific sampling and testing method based on an experiment to be designed.

Keywords

cotton fiber within-bale variability sampling testing repeatability classification

Cotton Gossypium spp. is grown in more than 100 countries worldwide and more than 150 countries in the world are involved in cotton exports or imports. In Africa, cotton is a source of income, employment, food and medicine. In many countries, cotton exports are not only a vital contribution to foreign exchange earnings but also account for a significant proportion of gross domestic product and tax income. Cotton plays a major role in economic development in Africa: 37 of the 53 African countries produce cotton and 30 are exporters.¹ Cotton is also a political crop because of its importance in world trade and in the economies of many developing countries.²

Cotton fiber plays a major role in the textile industry. Cotton fiber markets worldwide greatly depend on the fiber characteristics of a cotton bale, such as staple length, grade, color and Micronaire, strength, uniformity, maturity, fineness, elongation, neps, short fiber content, spinning performance, dyeing ability and cleanness. Cotton fiber is a natural and seasonal product. Genetic, environmental, harvesting and ginning factors impact its intrinsic and extrinsic characteristics. These characteristics play a major role in processing performance, costs, quality and utilization throughout the entire cotton textile chain, from the farm to the end-product.³

African cotton seems relatively uniform in its fiber characteristics, due to similar growing conditions and the smaller number of varieties planted in most countries, which is a comparative advantage on the world market.⁴ In addition, African cotton is handpicked and should consequently be less contaminated with plant matter, despite the fact that cotton is grown without irrigation by small farmers. However, variability within bales may be greater than what is generally seen in the bales produced in developed countries. The final uniformity levels of fiber characteristics within African cotton bales thus need to be investigated.

As the first step, ginning allows the conversion of seed-cotton into bulk cottonseed and bales of lint. Ginning may either at best preserve or negatively affect fiber quality, which in the latter case can damage its marketability.⁵

As a second step, a classification operation is carried out to determine fiber quality; fibers are evaluated for some of their technological characteristics. Traditionally, this classification has been carried out manually and visually.⁶ However, worldwide, the trend is a move toward classification based on the results of standardized instruments testing for cotton (SITC). The key issue of the classification operation is to provide reliable data so as to classify cotton bales according to their respective SITC results. The only applied sampling and testing method for instrumentally testing cottons is the method of the United States Department of Agriculture (USDA-USA), the ASTM standard.⁷ Applying these methods in Africa may lead to litigation claims, as cotton production in the USA is different from that in Africa.

Indeed, cotton production in the USA is ensured by large-scale and mechanized farms, while cotton is mainly produced on small-scale and manual farms in Africa. This may cause different levels of within-bale variability of the measured characteristics between these production schemes. Indeed, one sample taken from a bale may not adequately represent the overall quality of the bale. It is therefore essential to study within-bale variability in African countries in order to enable African countries to supply their cotton with reliable instrument-based quality information respecting internationally accepted test rules and procedures. This would require adaptation of the USDA-USA procedure to take into account the African production scheme, probably leading to specific levels of within-bale variability. A variability study of fiber technological characteristics is a crucial step forward in defining the conditions for good implementation of instrumental testing in order to limit the risk of litigation between cotton producers in Africa and their international customers. It also needs to be more specific and the current arbitral procedures need to be adapted to instrumental classification.

The first objective of this study was to quantify the level of within-bale variability for cotton fiber characteristics, as measured by SITC, produced in the six countries within Eastern and Southern Africa (ESA) involved in the study. For this, we chose to apply the methodology proposed by Aboé et al.⁸

The second objective was to check whether the sampling and testing method proposed by Aboé et al.⁸ for countries within West and Central Africa (WCA) was also applicable as it stood to saw- and roller-ginned cotton produced in these six ESA countries.

Material and methods

Two experiments measuring within-bale variability were conducted over two seasons (crop season 1 and crop season 2). Six countries (Mozambique, Sudan, Uganda, Tanzania, Zambia, Zimbabwe; Figure 1) were chosen as being representative and twenty-three physical ginning mills were randomly selected according to their seed-cotton supply areas, their ginning equipment (roller versus saw) and the presence or absence of lint cleaners (each time a physical ginning mill was sampled in this experiment, it was counted as one ginning mill; if the same physical gin was sampled for the two years, it was counted as two ginning mills in the following text). In crop season 1, samples were taken from bales in only 16 ginning mills (out of 23) even though half the season was already over. In addition, during crop season 2, samples were taken from bales in 16 ginning mills (out of 23). Nine ginning mills were sampled in both seasons to enable us to repeat the measurement in the same ginning mills, and others were added in the second season to extend the sample of ginning mills to the region (Table 1). For reasons of confidentiality, all countries and ginning mills were encoded.

Table 1.

Numbers of bales, samples and ginning mills involved in this experiment

Crop season 1: 2009		Crop season 2: 2010
Gin code	Nb bales	Nb Samples	Nb bales	Nb samples
C1G1	10	80	5	40
C1G2	10	80	5	40
C1G3	10	80	5	40
C2G1	5	40
C2G2	10	80
C2G3	10	80
C2G4	5	40
C2G5	10	80	5	40
C3G1	10	80	5	40
C3G2	10	80	5	40
C3G3	10	80
C3G4	5	40
C4G1	10	80
C4G2	10	80	5	40
C4G3	10	80	5	40
C4G4	5	40
C5G1	10	80	5	40
C6G1	10	80
C6G2	10	80
C6G3	10	80
C6G4	5	40
C6G5	5	40
C6G6	5	40
Total number of bales	160	80
Total number of samples	1280	640
Total number of gins	16	16

Methods and theories used by Aboé et al.⁸ were applied in this study. The following paragraphs summarize the most important features for an understanding of this document.

Bale sampling to characterize fiber properties

Sampling was carried out on the assumption that seed-cotton transported in different trucks came from different farms and different buying posts. Thus, this would induce different levels of variability when the seed-cotton differed from one village to another because of soil types, field management and seed-cotton handling between the field and the buying posts. Also, we considered that 18 fiber bales of 225 kg each could be produced from each seed-cotton truck. So, to ensure that each sampled bale came from a different village, we decided to select one out of every 20 bales in each ginning mill.

Eight samples per bale from eight equidistant layers were collected from each sampled bale. In each ginning mill, 10 bales were sampled in crop season 1 and 5 bales in crop season 2. Including all selected ginning mills, the total numbers of bales and samples collected and tested were 160 bales with 1280 samples in crop season 1, and 80 bales with 640 samples in crop season 2 (Table 1).

Sample analysis with replicates arranged in randomized blocks

The six technological characteristics recommended by the CSITC Task Force of the ICAC⁹ for testing were measured: Micronaire (Mic; Micronaire unit), Upper Half Mean Length (UHML, millimeter), Length Uniformity Index (Unif, percent), Strength (Str, gram/tex = 0.981 cNewton/tex), Reflectance (Rd, percent), Yellowness (+b, Yellowness unit).

The six technological characteristics were measured in a single controlled laboratory using a SITC device, USTER Technologies model HVI 1000. Each sample analysis was carried out according to ASTM 5867 requirements⁷ with one measurement of the Micronaire and two measurements of the Length/Uniformity Index, Strength, Color Rd and Yellowness. For each bale, the set of eight layers was analyzed twice using a randomized block design: in the first replicate the eight layers were analyzed in a random order; then in the second replicate the same eight layers were analyzed again with another randomization. The reference materials used for calibration were Universal Micronaire Calibration Cottons, Universal High Volume Instrument Calibration Cotton Standards for length and strength parameters and the color tiles supplied by the manufacturers. The reference material was also tested after each group of 16 samples, and the testing conditions were recorded. All test results were grouped together in a database; then a statistical analysis was carried out using R software version 2.11.1 and SAS Institute software version 9.2.

Computation of commercial sample precision

When two or more samples are taken from a bale for commercial classing, they can be analyzed separately, with one or more replicate analyses for each sample (Figures 2(b) and (d)): this is what we call cluster sampling. Or, the samples from the same bale can be mixed together and the resulting composite sample analyzed with one or more replicates (Figures 2(a) and (c)): this is what we call composite sampling.

Additive errors and variance components model

In what follows, we use the term “measurement” for a sample analysis performed according to ASTM 5867, that is, with several measurements for certain characteristics (see above).

Two assumptions are considered in the following.

Assumption 1: For quantitative variables when measuring one sample of a bale, two additive errors are expected:

Sampling error: the sample mean differs from the bale mean;

Measurement error: due to the re-sampling of a specimen within the sample, and to the imperfection of the instruments.

Assumption 2: A bale is the result of stacking successive layers in a continuous production process.

Therefore, within-bale variability results from differences between the layers. The variance of the two error components with a standard two-stage sampling method was estimated.

The model for exploring the acquired results was result = (bale fixed effect) + (layer in the bale random effect) + sampling error + measurement error.

As there was only one sample per layer, the sampling error within the layer was confounded with the layer effect. As each sample was measured twice, and the two replicated measurements were arranged in a block design (see above), the measurement error could be split into a block effect and a residual. The resulting model was

Y_{ijk} = m_{i} + A_{ij} + b_{ik} + E_{ijk}

(1)

where Y_ijk is the sample analysis result for bale i, layer j, replicate k, m_i is the mean of bale i, A_i,j is the random effect of layer j in bale i, assumed to be

N (0, σ_{A}^{2})

_, b_{i, k} _:is the effect of block k when measuring the samples of bale i, E_{i, j,}_k is the measurement error on the kth measurement of layer j of bale i, E_{i, j,}_k ∼

N (0, σ_{E}^{2})

, i is 1…I bales, j is 1…J layers in each bale and k is 1…K replicates for each bale.

The two random terms A and E were adopted as variability sources and were assumed to be independent. Their standard deviations were as follows:

$σ_{A}$ is the between-layer component;

$σ_{E}$ is the within-layer component.

Our goal was to estimate $σ_{A}$ and $σ_{E}$ for each chosen situation and then to compute the commercial bale sampling error standard deviation $σ_{M}$ as a function of any contemplated combination of sampling and analysis scheme. This can be obtained with J’ sampled layers in a bale and K’ measurements per sample, and then the sampling error variance follows Equation (2). Or, if one decides to mix cotton fibers from J’ layers to analyze them all together with N’ replicates of measurements, the sampling error follows Equation (3):

σ_{m}^{2} = \frac{σ_{A}^{2}}{J'} + \frac{σ_{E}^{2}}{J' K'}

(2)

σ_{m}^{2} = \frac{σ_{A}^{2}}{J'} + \frac{σ_{E}^{2}}{N'}

(3)

Figure 3 is a graphical representation of the relationship between $σ_{A}$ , $σ_{E}$ and $σ_{M}$ , the last giving information about gin performance.

Definition of the litigation risk and computation methods

The risk of litigation starts when the provider delivers a material that is considered different from the customer’s requirements; this risk corresponds to the zone where the difference between customer and provider appreciations exceeds the tolerances (Table 2).

Table 2.

Tolerances used to calculate the litigation risk

Characteristic	Commercial tolerances
Micronaire	+/−0.1 unit
UHML	+/−0.508 mm
UI	+/−1%
STR	+/−1.5 cN/tex
Rd	+/−1%
+b (Yellowness)	+/−0.5 unit

When several measurements are performed on a sample, the width of the distribution will vary according to observed variability in the results. For numerical calculations, a litigation risk of 10% was chosen. The evaluation of the overall sampling standard deviation ( $σ_{M}$ , Equation (4)) as a function of the commercial tolerances is given by

σ_{M} = \frac{Tol}{(1.28 . \sqrt{2})}

(4)

Using the data collected in the gins, we calculated $σ_{A}$ , $σ_{E}$ and $σ_{M}$ for each gin. The values are displayed on the chart $σ_{A}$ versus $σ_{E}$ (Figure 3). According to this figure and Equations (2) and (3), the size of the envelope curves fullfilling the commercial tolerances varied in size as a function of the number of samples per bale and/or the number of measurement replicates per sample and the type of sampling (Figure 2). When the dot representing the performance of a gin was included in the envelope curve (Figure 4), then the commercial tolerance was respected. For economic reasons, only three ways of decreasing sampling variance exist: (1) increasing the number of replicates; (2) increasing the number of layers without increasing the total number of replicates; and (3) increasing the number of layers and replicates that are parts of the combined sample without increasing the total number of replicates.

Results and discussion

Preliminary check testing reference materials during the testing sequences

Testing the different bales in a single crop season, plus from one season to the next, took several days. The time needed to test the samples from the eight layers in a bale with two replicates never exceeded 20 minutes in a row. We did not detect any time effect or test day effect on the analysis of the acquired data using the reference material included in the sample tests. External conditions did not significantly affect the technological measurements with 20-minute intervals and it was not necessary to adjust data according to the results of the reference materials.

First objective: evaluation of within-bale critical standard deviations

The results for the first objective are given in Table 3 where values for

σ_{A}

and

σ_{E}

are given for all the gins over the two years of the experiment.

Table 3.

Variances for the characteristics measured for each ginning mill and year

Gin code	Crop	Sigma	Micronaire	UHML	Unif	Strength	Rd	Yellowness
C1G1	2009	A	0.0593	0.2424	0.1384	0.3773	0.3254	0.1647
	2009	E	0.0286	0.3559	0.7602	0.7926	0.5636	0.2084
	2010	A	0.0532	0.1813	0.4699	0.1014	0.4147	0.1424
	2010	E	0.0477	0.2311	0.5844	0.9687	0.6066	0.2182
C1G2	2009	A	0.0538	0.2389	0.3794	0.3429	0.4100	0.1081
	2009	E	0.0331	0.2795	0.5297	0.6761	0.5465	0.2080
	2010	A	0.0371	0.2493	0.3948	0.3202	0.0429	0.1226
	2010	E	0.0272	0.2979	0.6898	0.8947	0.6679	0.2855
C1G3	2009	A	0.0655	0.1850	0.2061	0.3862	0.8467	0.1765
	2009	E	0.0331	0.4269	0.6805	0.8068	0.4966	0.2231
	2010	A	0.2361	0.1317	0.3262	0.6481	1.1804	0.4807
	2010	E	0.1854	0.2952	0.5530	1.1561	0.6838	0.3424
C2G1	2010	A	0.0325	0.1758	0.4323	0.4190	0.1539	0.0736
C2G1	2010	E	0.0321	0.2331	0.5723	0.6895	0.3690	0.1757
C2G2	2009	A	0.0266	0.0000	0.1847	0.0000	0.3064	0.0989
C2G2	2009	E	0.0505	0.3956	0.6953	1.0352	0.4877	0.1892
C2G3	2009	A	0.0000	0.0410	0.2998	0.1777	0.0560	0.0793
C2G3	2009	E	0.0599	0.3705	0.6391	1.1239	0.5936	0.2279
C2G4	2010	A	0.0000	0.1881	0.0000	0.1403	0.4009	0.1206
C2G4	2010	E	0.0357	0.2507	0.6175	0.5649	0.3977	0.2037
C2G5	2009	A	0.0183	0.0000	0.1576	0.3453	0.3206	0.1203
	2009	E	0.0378	0.4437	1.0958	0.7782	0.4762	0.2490
	2010	A	0.0406	0.1798	0.0000	0.5317	0.3041	0.0623
	2010	E	0.0351	0.3328	0.7057	0.6394	0.3455	0.2031
C3G1	2009	A	0.0348	0.2128	0.0000	0.8991	0.1925	0.0000
	2009	E	0.0592	0.4575	0.9184	1.3795	0.5038	0.2516
	2010	A	0.0627	0.1834	0.0000	0.5004	1.7449	0.2025
	2010	E	0.0613	0.2949	0.7354	0.7838	0.6408	0.1861
C3G2	2009	A	0.0000	0.0882	0.3344	0.6410	0.0000	0.0000
	2009	E	0.0598	0.4767	0.8028	1.5406	0.6530	0.2367
	2010	A	0.0494	0.2695	0.4310	0.0000	0.2694	0.1872
	2010	E	0.0371	0.2364	0.6088	0.7241	0.5885	0.2074
C3G3	2009	A	0.0273	0.1263	0.2552	0.0000	0.2259	0.0000
C3G3	2009	E	0.0468	0.3531	0.8330	1.4980	0.5511	0.2764
C3G4	2010	A	0.0618	0.2523	0.2706	0.4244	0.3462	0.1428
C3G4	2010	E	0.0519	0.3366	0.7052	0.7192	0.5227	0.1981
C4G1	2009	A	0.0320	0.3307	0.4656	0.6752	0.5802	0.2756
C4G1	2009	E	0.1233	0.5645	1.0853	1.5309	0.9776	0.3865
C4G2	2009	A	0.0640	0.3657	0.4765	0.3406	0.4258	0.0000
	2009	E	0.0831	0.4402	0.7256	1.1768	0.7099	0.3865
	2010	A	0.0850	0.0969	0.5006	0.6891	0.4944	0.2120
	2010	E	0.0771	0.3579	0.8147	1.1150	0.6441	0.4153
C4G3	2009	A	0.0293	0.0000	0.0000	0.3985	0.3931	0.1124
	2009	E	0.0497	0.4690	0.9116	0.9752	0.5014	0.2034
	2010	A	0.0344	0.0000	0.4080	0.1857	0.3068	0.0427
	2010	E	0.0387	0.2817	0.5679	0.7332	0.3978	0.2092
C4G4	2010	A	0.1009	0.0000	0.3835	0.1792	0.7767	0.2449
C4G4	2010	E	0.0658	0.4123	0.9175	0.9879	0.5448	0.2864
C5G1	2009	A	0.1353	0.3752	0.8262	0.6453	0.9005	0.3004
	2009	E	0.0294	0.4470	0.8341	0.8611	0.8800	0.2169
	2010	A	0.0867	0.1806	0.6727	0.7048	1.3251	0.1976
	2010	E	0.0545	0.4030	0.6562	0.7283	0.7110	0.2676
C6G1	2009	A	0.1088	0.4177	0.2894	0.8062	0.8729	0.1937
C6G1	2009	E	0.0558	0.5506	0.9239	1.0545	0.7103	0.3559
C6G2	2009	A	0.0936	0.4327	0.0000	0.5146	0.6402	0.1989
C6G2	2009	E	0.0608	0.5121	1.1270	1.1823	0.6764	0.3036
C6G3	2009	A	0.0975	0.4781	0.5479	0.6186	0.5484	0.0000
C6G3	2009	E	0.0507	0.5080	0.9627	1.1159	0.6647	0.3198
C6G4	2010	A	0.1094	0.3513	0.5837	0.7300	0.6153	0.1254
C6G4	2010	E	0.0863	0.2718	0.8048	0.9693	0.7806	0.4020
C6G5	2010	A	0.1631	0.4028	0.7641	0.8988	0.8938	0.3109
C6G5	2010	E	0.0860	0.3508	0.7010	1.0483	0.8711	0.4603
C6G6	2010	A	0.0993	0.3368	0.4827	0.0000	0.3760	0.2087
C6G6	2010	E	0.0612	0.2395	0.7841	1.2161	0.8768	0.2784

Figures 5 –10 show the data given in Table 3, categorized according to the ginning technique: saw or roller. These figures also display the calculation results of Equations (2) and (3) for several combinations of sampling and testing conditions, one of them being from Aboé et al.⁸

The second objective was the checking of Aboé et al.’s⁸ operating method applied in ESA countries for saw-ginned cottons

The second objective involved counting the number of gins that were included in the envelope curves (Table 4) displayed in Figures 5 –10. Not all the sampling–testing combinations shown in Table 4 have been drawn on the figures, for easier reading.

Table 4.

For saw-ginned cottons, tabulation of the percentage of gins during two crop seasons (based on 23 gins) with a litigation risk under 10% depending on various combinations of J’, and of K’ or of N’

For saw-ginned cottons (percentages)
Sampling testing combinations	Micronaire	UHML	Unif	Strength	Rd	Yellowness
J'1 - K'1	30	0	83	22	17	70
J'1 - K'2	48	39	87	70	65	87
J'2 - K'1	70	57	91	83	74	91
J'2 - K'2	78	91^a	96^a	87^a	74	96
J'2 - N'1	52^a	13	87	35	30	78
J'2 - N'2	70	57	91	83	74^a	91^a

Percentage of gins in Eastern and Southern African countries included in the envelope curves defined in the recommendations made by Aboé et al.⁸ for sampling and testing bales of saw-ginned cotton in West and Central African countries.

To keep the litigation risk below 10%, while respecting agreed international tolerances, Aboé et al⁸ reported that J’ = 2 and N’ = 1 was required for Micronaire, that J’ = 2 and K’ = 2 was required for UHML, Unif and Strength, and that J’ = 2 and N’ = 2 was required for Rd and yellowness; most of the gins were included in the envelopes, but a few of them fell outside the envelope curves, as poor seed-cotton management practices play a large role in the within-bale variability of cotton fiber characteristics.

Respecting the recommendations made by Aboé et al.⁸ and counting the number of ESA gins within the envelope curves shown in Figures 5 –10, Table 4 was drawn up for saw-ginned cottons. The cells indicate the percentages of gins (Table 4) included in the envelope curves respecting the recommendations Aboé et al. first made for WCA countries.

For all characteristics but Micronaire, Aboé et al.’s⁸ recommendations could also be applied for saw-ginned cottons produced in ESA countries, as the number of gins included in the envelope curves was close to the maximum of gins tested in this experiment. For Micronaire, assuming that no more than two samples can be economically drawn from bales, the only better combination was J’ = 2 and K’ = 2 in place of J’ = 2 and N’ = 2, moving from composite testing toward cluster testing (18 gins out of 23, rather than 12). In this case, only the cost of testing increased. If J’ = 2 and N’ = 1 were to be kept, it would correspond to taking a greater litigation risk than the 10% set at the beginning; in fact, under these conditions, the litigation risk rose to a level of 33% for the gins displaying a poor performance, meaning that a quality claim would arise for the Micronaire value one out of three times.

To conclude this section, the recommendations made by Aboé et al.⁸ for sampling and testing methods in WCA countries can also be applied to ESA countries for five out of six of the measured fiber characteristics: UHML, Uniformity Index, Strength, Rd and Yellowness. By extension, we deduced that within-bale variances were comparable in WCA and in ESA. However, some small changes are needed for Micronaire by modifying the testing conditions in order to stay within a 10% litigation risk and by opting for cluster testing rather than composite testing, or by better seed-cotton management at the gin. It is important to note that a study measuring between-bale variability within lots of bales may also lead to further modifications to this proposed operating method.

Second objective: checking of Aboé et al.’s operating method applied in ESA countries: roller-ginned cottons

Similar to the previous way of interpreting the collected data, the percentages of roller gins within the envelope curves traced with various combinations of J’, K’ and/or N’ are reported in Table 5. If the recommendations made by Aboé et al.⁸ for sampling and testing saw-ginned cotton fibers were used for roller-ginned cotton fibers, few gins would be able to fulfill the 10% litigation risk, as many gins fell outside the corresponding envelopes. Roller-ginned cottons are known to be more variable than saw-ginned cotton.³ As a consequence, a specific sampling and testing method is used to class US Pima cottons that are roller ginned.³

Table 5.

For roller-ginned cottons, tabulation of the percentage of gins during two crop seasons (based on nine gins) with a litigation risk under 10% depending on various combinations of J’, and of K’ or of N’

For roller-ginned cottons (percentages)
Sampling testing combinations	Micronaire	UHML	Unif	Strength	Rd	Yellowness
J′1 - K′1	0	0	22	0	0	0
J′1 - K′2	0	11	44	11	0	22
J′2 - K′1	0	11	89	11	0	56
J′2 - K′2	0	33^a	89^a	100^a	44	89
J′2 - N′1	0^a	0	56	0	0	0
J′2 - N′2	0	11	89	11	0^a	56^a
J′3 - K′1	0	44	100	100	44	89
J′3 - K′2	11	67	100	100	78	100
J′3 - N′1	0	0	67	0	0	0
J′3 - N′2	0	33	100	22	11	67
J′4 - K′1	22	56	100	100	78	89
J′4 - K′2	56	67	100	100	100	100
J′4 - N′1	0	0	78	0	0	0
J′4 - N′2	0	33	100	22	44	67
J′5 - K′1	78	67	100	100	89	100
J′5 - K′2	78	100	100	100	100	100
J′5 - N′1	0	0	78	0	0	0
J′5 - N′2	0	33	100	22	44	67

In the same way, a sampling and testing method has to be designed for African roller-ginned cottons. It was only when J’ = 5 and K’ = 2 that a large percentage of the gins were able to respect the 10% litigation risk, while it became technically and economically unfeasible. However, Gourlot and Drieling¹⁰ reported in their Figure 74 that the range of distributions for within-bale standard deviations $σ_{M}$ of roller-ginned cottons among all African cottons are smaller than for African saw-ginned cottons, even though the median levels of those distributions are generally at a higher level for roller-ginned cottons. This tends to conclude that seed-cotton management and ginning practices in roller ginning are more even – albeit at a higher level – than for saw ginning, given the 63 gins in Africa.

At this point, pending the conclusion of another study on between-bale variability in roller-ginned cottons, we would conclude that taking two samples per bale and taking two replicate measurements per sub-sample of the combined sample would enable the 10% litigation risk level to be respected for the following characteristics for most gins: Unif, Strength and Yellowness (Table 5). The results would appear less acceptable for UHML and Rd, while strong efforts need to be made for Micronaire.

Indeed, for both saw- and roller-ginned cottons, the existence of variations may be due to poor production practices and to mismanagement of seed-cotton. Other factors, such as continued exposure to weather and/or the action of microorganisms and/or frost, drought or other adverse weather conditions, may cause a deterioration in several fiber characteristics in terms of mean and within-bale variability. Similarly, the moisture content, storage time, amount of high-moisture foreign matter, variation in moisture content throughout the stored mass, initial temperature of the seed-cotton, temperature of the seed-cotton during storage, climatic factors during storage (temperature, relative humidity, rainfall) and protection of the cotton from rain and wet ground affect fiber quality.

There are some limitations to this study: (1) we did not consider reproducibility conditions that might appear when results may differ from one classing laboratory to another; (2) it will also be necessary to periodically quantify the within-bale variability for each situation, in order to ensure the litigation risk for any given situation; (3) finally, we limited the litigation risk to 10% for any single bale while commercial agreements and contracts generally concern lots of several bales and the General Rules of Cotton Associations; the lot litigation risk will have to be evaluated too.

Conclusion

The evaluation of within-bale variability levels for six fiber characteristics (Micronaire, UHML, Uniformity, Strength, Reflectance and Yellowness) was made possible using a broad sample of cotton bales produced in Eastern and Southern Africa. With this data, we were able to verify that (1) sampling variances are on the same scale in ESA countries as in WCA countries for saw-ginned cottons, and (2) that sampling and testing conditions proposed earlier for saw-ginned cottons produced in WCA countries also apply in ESA countries for respecting agreed international tolerances and a litigation risk under 10%; an adjustment is, however, required for seed-cotton management practices, to improve the Micronaire situation in ESA countries.

For roller-ginned cottons, samples were also taken from bales to measure their within-bale variability. As expected, the characteristics of roller-ginned cotton fibers were more variable within the bales than those of saw-ginned cotton fibers. These higher within-bale variability levels require the definition of a specific sampling and testing method devoted to roller-ginned cottons, or improvements in seed-cotton management practices.

Footnotes

Disclaimer

This report was prepared as part of the CFC/ICAC/33 project. The views expressed are not necessarily shared by the Common Fund for Commodities and/or the European Commission and/or the International Cotton Advisory Committee. The designations employed and the representation of materials in this report do not imply the expression of any opinion whatsoever on the part of the Common Fund for Commodities and/or the European Commission or the International Cotton Advisory Committee concerning the legal status of any country, territory, city or area, or its authorities or concerning the delineation of its frontier or boundaries.

Funding

This work was undertaken as part of the CFC/ICAC/33 Commercial Standardization of Instrument Testing of Cotton project, which was funded by the Common Fund for Commodities, an intergovernmental financial institution established within the framework of the United Nations, headquartered in Amsterdam, the Netherlands, and by the European Union in the framework of its “All ACP Agricultural Commodities Programme” under the sponsorship of the International Cotton Advisory Committee (ICAC) Washington (USA) and implemented by the Faserinstitut Bremen (FIBRE), Germany.

Acknowledgments

The authors would like to thank the personnel of the Regional Technical Center of Dar Es Salaam, formed by the association of the Tanzania Bureau of Standards and the Tanzania Cotton Board in Dar Es Salaam, Tanzania, who performed all the fiber characterizations for the present study.

The authors want to thank the cotton companies who allowed sampling of their production in their facilities in order to proceed with this experiment.

References

ICAC. Cotton: World statistics, Washington, DC: International Cotton Advisory Committee, 2010, pp. 295–295.

UNCTAD. UNCTAD Handbook of Statistics, New York and Geneva: United Nations Conference on Trade and Development, 2008, pp. 512–512.

Estur G and Gergely N. The economics of roller ginning technology and implications for african cotton sectors. World Bank, 2010, p.75.

Townsend

. The importance of cotton in Africa, Washington, DC: The Next Steps for Africa, 2006, pp. 6–6.

ICAC. Report of an expert panel on ginning methods, Washington, DC: ICAC, 2001, pp. 30–30.

Estur G. Quality and marketing of cotton lint in Africa. Africa Region Working Paper Series No 121, 2008, p.82.

ASTM. Standard test methods for measurement of physical properties of cotton fibers by High Volume Instruments. Annual Book of ASTM Standards. Philadelphia, PA, 1996, pp.886-893.

Aboé

Gourlot

Gozé

. New findings on within bale repeatability of measurements obtained with standardized instruments for testing cotton (SITC) in Gossypium Hirsutum Fiber produced in West and Central Africa. Textil Res J 2012; 82: 1469–1484.

ICAC-Secretariat. Instrument Testing of Cotton at the Producer Level for trading purposes. Washington, DC, 2011.

10.