492

Chapter 12 Summary, Conclusions, and Recommendations for Future Research

A. Summary

The objectives of this study were to evaluate statistically the interrelationships among variables selected to represent the organic component of whole seam samples of United States' coals, to evaluate sources of variation within coals that have been grouped by different criteria (age, geographic source, rank, and stratigraphic position), and to group coals according to a multivariate classification. The methods employed included univariate analysis, bivariate analysis, partial correlation analysis, analysis of variance, principal components analy­sis, and cluster analysis.   The results of these analyses provided information about coalification processes, and allowed various models of coalification which have been proposed by different researchers to be evaluated and tested against the data.

The 5 strongest factors from the principal components analyses were used in the cluster analyses to divide the 277 coals of the data set into groups of greatest similarity. The best division was attained by 4 groups, with the high-rank group of coals being most strongly differen­tiated from all others. These groups approximately correspond to the 4 groups found in an earlier, similar study which was based on a data set of 119 coals (Waddell and others, 1978). The groups identified within the present study are as follows:

493

1. Group I is composed of the lowest rank coals of the data set (lignites through high volatile C bituminous). All 50 coals of this group are post-Pennsylvanian in age, and represent the Rocky Mountain, Northern Great Plains, Gulf, and Pacific Prov­inces. They are characterized by low organic sulfur content and high moisture, volatile matter, and oxygen values. The coalification trends which are most evident for this group of coals are the decrease in oxygen and moisture, and the increase in reflectance, carbon, and calorific value. Nitrogen enrichment appears to be more strongly related to increasing rank for this group of coals than for any other grouping employed in this study.

2. Group II coals occupy the rank range from high volatile B through medium volatile bituminous. The group contains coals from all 6 provinces and with ages that range from Pennsyl-vanian through Eocene. The principal characteristics of this group of 93 coals are their intermediate rank and low organic sulfur content.   The most sensitive rank indicators for this group are carbon and reflectance, which increase, and oxygen and volatile matter, which decrease.   Hydrogen displays a small increase with coalification for this rank range.

3. Group III coals essentially occupy the high volatile bitumi­nous rank range.   Their principal characteristics are their moderate rank and their high organic sulfur content. Although the Eastern, Gulf, Interior, and Rocky Mountain Provinces are represented, the Interior Province contributes 61 of the 82 coals of this group. The sensitive rank indicators for these

494

coals are calorific value and carbon, which increase, and oxy­gen and moisture, which decrease. Reflectance increases with rank, but not as sensitively as calorific value or carbon. Nitrogen enrichment with rank is also evident for these coals. 4. Group IV coals, with 1 exception, vary in rank from medium volatile   bituminous through anthracite.   Their principal characteristic is their high rank. They are composed primari­ly of Eastern Province coals, but about 25 percent are from the Interior Province and a single sample is from the Rocky Mountain Province. This 1 coal is the only non-Pennsylvanian age coal in the group. The sensitive rank indicators for this group are hydrogen and volatile matter, which decrease, and reflectance and carbon, which increase.    Nitrogen   also decreases with rank for this group, with about half of the variance of nitrogen related to the effects of rank. The lip-tinite maceral group was deleted from the data set because 42 of the 50 coals had zero "liptinite contents.

The interrelationships among the variables in coals grouped by rank yielded the following results:

1. The group of 22 lignite and subbituminous coals are all post-Pennsylvanian in age and represent the Northern Great Plains, Pacific, Gulf, and Rocky Mountain Provinces.   The sensitive rank indicators for these low-rank coals are the reflectance, which increases, and oxygen   content,   which   decreases. Although 9 of the 12 variables load on the rank factor, all loadings are less than 0.81. This indicates that at low lev­els of rank, much of the variance of these variables is due to

495

depos-itional sources which have not yet been obscured by the effects of rank.   Nitrogen increases with rank through this rank range. Within this low-rank range, vitrinite-rich coals have lower moisture contents than vitrinite-poor coals. The processes of coalification which have been identified for this group of coals are as follows: a change in the chemical compo­sition which affects the calorific value, a change in the chemical structure which affects the reflectance, a change in the porosity which affects the moisture content, and an alteration of the nitrogen content. In addition, the data may be interpreted to imply that bituminous substances may be gen­erated with increasing coalification in this range, possibly contributing to the observed decrease in porosity  with increasing rank.

2. The 57 high volatile C bituminous coals represent the Gulf, Interior, and Rocky Mountain Provinces. The most sensitive rank indicators for this group of coals are oxygen content, which decreases, and calorific value and carbon content, which increase. High-organic sulfur coals from the Interior Prov­ince tend to confuse the analysis of this group; they are lower in rank, as measured by carbon, than the remaining coals of the group, and are the probable cause of the association between the organic sulfur content and the rank-related vari­ables in the factor analysis. The organic sulfur distribution for the entire group is trimodal, indicating that the group is not composed of a homogeneous population.

496

3. The 43 high volatile B bituminous coals represent the Eastern, Interior, Gulf, and Rocky Mountain Provinces, and vary in age from Pennsylvanian through Eocene. The most sensitive rank indicator for the group is carbon content, which increases;

other rank indicators are calorific value and reflectance, which increase, and volatile matter yield and oxygen content, which decrease. As in the previous rank category, .high-organic sulfur coals from the Interior Province tend to con­fuse the factor analysis of this group because they are slightly lower in 'rank, as measured by carbon, than the remaining coals of the group.

4. The 97 high volatile A bituminous coals represent all 6 prov­inces and range in age from Pennsylvanian through Eocene. They have carbon and reflectance as their most sensitive rank indicators; ,both variables increase with increasing rank. The loss of volatile matter and hydrogen become significant rank effects at this rank level.

5. The 54 coals of medium volatile bituminous and higher rank are all from the Eastern and Interior Provinces and are all of Pennsylvanian age, except 1 sample which is from the Rocky Mountain Province and which is Cretaceous in age. The coals of this group have hydrogen loss as their dominant rank effect.    Three other variables -- reflectance, volatile matter, and carbon -- have loadings on the first factor that exceed 0.90, indicating that rank has accounted for almost all the variance of these variables. These high-rank coals exhi­bit a significant nitrogen loss with rank increase, an effect

497

which accounts for approximately 50 percent of the variance of this variable.   The liptinite maceral group was deleted from the data set because 42 samples had zero liptinite contents. Bivariate analyses revealed the following rank-related trends:

moisture and calorific value reverse the slope of their relationship in the high rank ranges, and hydrogen shows slight enrichment up to the high volatile A bituminous stage, then strong depletion; nitrogen increases in the low rank ranges and decreases within the high rank ranges; and the vitrinite maceral content decreases prior to the second coalification jump.

The interactions between the variables were studied for coals grouped by provinces. The results of the studies are as follows:

1. Eastern Province coals which were included in the study were all Pennsylvanian in age and ranged in rank from high volatile B bituminous through anthracite. The presence of relatively few anthracites, and their distinctively higher degree of metamorphism, tended to skew many of the distributions of the rank-related variables.   The most sensitive rank indicators were reflectance, which increases, and volatile matter and hydrogen, which decrease.   Nitrogen decreases with rank in this group of coals. Three stratigraphic groups -- the Potts-vine, Allegheny, and Monongahalea Groups -- were studied by analysis of variance; they differed significantly only in car­bon, volatile matter, hydrogen, and nitrogen.

2. Interior Province coals which were included in this study were all Pennsylvanian in age and ranged in rank from high volatile C to low volatile bituminous.   The most sensitive rank498

indicators for these coals are carbon and reflectance, which

increase, and oxygen and volatile matter, which decrease. The

distinguishing characteristics of the Interior Province coals

are their elevated organic sulfur contents, their depressed

reflectances, and, within the high volatile bituminous rank

range, their elevated moisture contents.

3. Northern Great Plains Province coals which were included in

this study are Cretaceous to Paleocene in age. Of a total of

only 15 samples, 13 are of subbituminous B or lower rank, and

the remaining 2 are high volatile A bituminous. The most sen­sitive rank indicators for this group are oxygen, which

decreases, and reflectance and calorific value, which increase. Nitrogen increases with rank for these samples.

4. Rocky Mountain Province coals which were included in this study are Cretaceous to Eocene in age, and, with the exception of 1 semianthracite, vary in rank from subbituminous A through high volatile A bituminous. The most sensitive rank indica­tors are oxygen, which decreases, and carbon, which increases. Nitrogen increases with rank in these coals.

5. Only 3 coals represented the Pacific Province in this study; all are Eocene in age and come from Washington State. They vary in rank from subbituminous B through high volatile A relative to other coals of ranges. No factor analysis

bituminous. Their properties, similar rank, fall within typical was attempted for this group.

499

6. Gulf Coast Province coals were represented by only 7 samples. Of these, 1 is Cretaceous in age and the remainder are Eocene. The ranks vary from subbituminous B through high volatile A bituminous. No factor analysis was attempted on this group. Their properties, relative to other coals of similar rank, fal,l within typical ranges.

An analysis of samples grouped by age for all coals younger than Pennsylvanian determined that the group of Paleocene coals (15 samples) differ at the 0.001 probability level from both the younger Eocene (10 samples) and the older Cretaceous coals (34 samples) in the following ways: they are higher in moisture and oxygen, and are lower in carbon, hydrogen, nitrogen, and calorific value.

Analysis of variance was applied to evaluate the variability that occurs within selected mines for which single seams were sampled more than once within the Eastern, Interior, and Rocky Mountain Provinces. The variables chosen for this portion of the study loaded independently in the principal components analysis: carbon, the percentage of the inertinite maceral group, and hydrogen. The analysis indicated that the variance at this level was inhomogeneous, precluding the pooling of the data to compare the variance at higher levels.

A reconnaissance study of 5 ash constituents -- the oxides of iron, sodium, potassium, calcium, and magnesium -- as well as a subdivision of the inertinite maceral group indicated that iron oxide content of the ash and organic sulfur content are directly related and are independent of rank; macrinite appears to be inversely related to organic sulfur, and calcium and magnesium oxides decrease and potassium oxide increases with rank. The rank-dependence of the oxides of calcium, magnesium, and

500

potassium was attributed to a geographic bias; Western coals are lower in rank than coals from the Interior and Eastern Provinces and contain high contents of calcium and magnesium oxide and low contents of potas­sium oxide. This portion of the study provided an opportunity to deter­mine the extent of the variability of some of the major ash constitu­ents, as well as some of their relationships to the constituents of the organic fraction of the coal.

B. Conclusions

1. Principal components analysis of the entire data set of coals

and on subsets grouped by ASTM rank category, by geographic

area, and by cluster analyses, indicated that rank is the most important factor in determining the variability of the United

States' coals. Maceral composition is usually the second most important factor, and organic sulfur content is frequently the third.

2. Cluster analyses, using the 5 strongest factors, indicated that the coals of the United States may be divided into groups of greatest similarity, with the best division achieved with 4 groups.

3. The principal components analyses of the 4 groups determined by the results of the cluster analysis were more successful in separating the factors than were similar analyses of the 5 groups which were classified according to only their ASTM rank category (lignites through subbituminous A, high volatile C, B, and A bituminous, and medium volatile bituminous through

501

anthracite). This indicates that, for the variables used in this study, the cluster analysis produces more homogeneous groups of coals than does the ASTM classification procedure. High-organic sulfur coals, usually from the Interior Province, are dissimilar from coals of equivalent rank which are low in organic sulfur; high- and low-organic sulfur coals which are grouped together in the same ASTM rank classification manifest a complex set of variable interactions. The ASTM rank system utilizes parameters for classifying coals which involve less expensive determinations than the chemical procedures which are employed for the ultimate (or elemental) analysis. The widely used and standardized procedures for the ASTM classifi­cation may serve the needs of most producers and consumers of coal, but as a basis of scientific investigation in which com­ponents of the petrographic and elemental analyses are also used as variables, the ASTM classification becomes less ade­quate. A system which produces more homogeneous groups of coals should use a measure of rank as the primary classifica­tion criterion and some other parameter, or set of parameters, as a secondary criterion.        The secondary criterion may be maceral content, the sulfur content,, or some other parameter, depending on the specific purpose for which the coals are being classified. This study suggests that some measure of the sulfur content would be an important secondary criterion for a scientific classification of the coals of the United States.

502

4. A principal components analysis of each of the 5 ASTM rank groups indicates that rank has affected the variables to,such an extent that even when grouped in categories of restricted rank range, these effects account for more variability of the data sets than any other factor. The effects of rank appear to have erased much variability due to pre-coalification con­ditions and processes. If the groups were composed of coals of a more limited rank range, the potential for reducing the effects of rank would be increased. The information which exists among the interrelationships of the variables that can be attributed to pre-coal ification processes may be so limited that the only way to adequately control the rank range of the data set would be to make intensive studies of individual seams, or sets of seams, over restricted geographic areas.

5. Principal components analysis of the data set formed by all but Group IV coals indicated that the vitrinite content decreases with rank up to the start of the medium volatile bituminous range. The decrease in vitrinite content with increasing rank was attributed by Waddell and others (1978) to a "pseudocorrelation" resulting from the relationship of each to age. The relationships between the 3 variables (vitrinite content, rank, and age) were tested in the present study by a partial correlation analysis; the results indicated that decreasing vitrinite content is associated with increasing rank, and not with increasing age. The independence between vitrinite content and rank in the principal components analy­sis for the entire data set results from a discontinuity in

503

the relationship at the second coalification jump (approxi­mately 88 percent carbon, dmmf),

macerals abruptly disappear and are microscopically indistin­guishable from vitrinite, causing the vitrinite content to increase in the medium volatile bituminous and higher rank coals. The observed relationship between vitrinite content and rank below the second coalification jump is contrary to generally held theories of coalification which consider rank and total vitrinite content to be independent over this range of rank.

6. Principal components analyses of the 4 groups classified by the cluster analyses, and also on the 5 groups classified by only their ASTM rank category, gave insights into the processes of coalification over various portions of the rank range. An unexpected result of this portion of the study was that the nitrogen content was observed to be dependent on rank: it increases with rank in the lignite through subbitumi­nous A range, and decreases with rank in the medium volatile bituminous through anthracite range. An increase in nitrogen had previously been reported for a group of low-rank coals from New Zealand (Suggate, 1959), but in general the total nitrogen content is taken to be independent of rank over the entire rank range. The observed increase in nitrogen for the coals in the present study is approximately equal to the observed decrease in nitrogen in the high ranks, so that when all the coals are analyzed together, the nitrogen appears to be independent of rank. The decrease in nitrogen with

which the liptinite

504

increasing rank in the high-rank coals may be an artifact due to an incomplete extraction of nitrogen during the chemical analytical procedure, since it has been reported that nitrogen is more difficult to extract in high-rank coals than in low­rank coals (Rees, 1966; Montgomery, 1978).

7. An analysis of samples grouped by age for all coals younger than Pennsylvanian determined that the group of Paleocene coals are of lower rank than either the younger Eocene or the older Cretaceous coals. This unexpected result is attributed to a sampling bias produced by the low rank of those Paleocene coals of the data set which were obtained from the Fort Union Region of the Northern Great Plains Province. The lack of a direct relationship within the post-Pennsylvanian coals between increasing age and increasing rank permitted an assessment of the effect of age in this relationship. The increase in nitrogen in this data set was found to be related to the increase in rank, but is unrelated to the age of the coal.

8. The inferential statistical techniques used in this study assume that random samples were drawn from populations in which the variates were normally distributed.         Sampling maps show the sample sites to be tightly clustered in some areas and widely spread in others: the Gulf Province is represented by samples which are clustered in a relatively small portion of southern Texas; except for 1 sample, all semianthracite and anthracite coals come from a very restricted geographic region of the Eastern Province; and the concentrated sampling of

505

coals from the Hanna Basin of the Rocky Mountain Province inherited an unusual relationship which appears to be unique to the Province -- older coals on the west side are lower in rank than younger coals on the east (Glass and Roberts, 1980). The effects of such sampling biases reside with the data; the extent to which they effect the analysis is difficult to esti­mate and even more difficult to remove.

9. Bivariate analyses indicate that none of the relationships between the plotted variables are linear over the entire range of rank represented by the coals of the entire data set.          The non-linearity of the relationships has a deleterious effect on the principal components analyses, making the loadings more complex and the interpretations, therefore, less secure. An improvement in the correlation coefficients between many of the plotted variables can be obtained by taking the logarithms of both variables.

10. Analysis of variance was applied to data from 3 stratigraphic groups from within the Eastern Province; from stratigraphical­ly lowest to highest, they are the Pottsville, Allegheny, and Monongahalea Groups.      The groups differed significantly only in carbon, volatile matter, hydrogen, and nitrogen.  Whereas the nitrogen difference is puzzling, the other 3 variables vary in such a way as to imply that the coals lower in the stratigraphic section have attained higher levels of rank. This result is consistent with the well-known Hilt's Law of coalification.

506

11. Analysis of variance indicated that the Interior Province coals of the high volatile bituminous rank range contain sig­nificantly higher moisture and lower reflectance values, for a given carbon content, than do the coals from the other prov­inces. When low volatile bituminous rank ranges are compared, the Interior Province coals do not differ significantly from other coals in either carbon or moisture, but do have signifi­cantly lower reflectance values. As rank increases from the start of the high volatile bituminous stage, the moisture pro­perties of the Interior Province coals reach a normal range in the high volatile A bituminous stage; the reflectance values, however, continue to be low through the low volatile bitumi­nous stage, but they more closely approach the normal range as rank increases. The effects of increasing rank tend to make the coals from the Interior Province more similar to all other coals.

12. Analysis of mines in which single seams of coal were sampled more than once over various intervals from within the Interi­or, Eastern, and Rocky Mountain Provinces revealed that the variance at the lowest level -- within mines -- is inhomogene­ous for the 3 variables which loaded independently on the fac­tor analysis: carbon, the percentage of the inertinite maceral group, and hydrogen. This prevents the data from being pooled to then compare variation between mines within each province, or the variation from province to province. Analysis of vari­ance of these 3 variables revealed that the variation within mines and the variation between mines was inconsistent from

507

variable to variable and from province to province.                          Thus no conclusions can be drawn about which province exhibits the most variation within mines and which exhibits the most varia­tion between mines. The lack of replicate analyses for any of the samples prevents the within mine variance from being further decomposed into the portion which is due to measure­ment error and that which is due to

samples.            The analysis emphasizes of variation within coal seams so plans and effective study designs can be formulated to take maximum advantage of this variation.

variability between the the need to assess sources that efficient sampling

C. Recommendations for Future Research

Based upon the results of this study, the following recommendations for further research are proposed:

1. This study has shown that rank accounts for the largest source of variation of the data set and, consequently, obscures sources of variation due to pre-coalification effects.                                                                                             The effects of rank could be controlled to allow the study of pre-coalification processes by sampling coals from very res­tricted rank ranges, and/or by selecting variables which are less rank dependent.

2. The variation of selected independent variables within mines from which more than 1 sample was taken was found to be inho­mogeneous, and an unknown portion of this variation is due to analysis error.  Duplicate analyses would allow a suitably

508

designed study to establish confidence intervals for the dif­ferent variables and would allow the within mine variation to be further decomposed into its constituent parts.

3. The present analysis has shown that many of the bivariate relationships among the variables are log/log linear and that much redundancy exists among the parameters which load upon the rank component. A factor analysis, using appropriate transformations, may be applied to test specific models of coalification which are presented in this study and elsewhere. 4. The decrease in nitrogen content of the high-rank coals may be due to an incomplete extraction of the nitrogen in the chemi­cal analysis process. A specific study could be designed to test this effect.

5. The cluster analysis procedure has produced groups of coals which are more homogeneous than the ASTM classification sys­tem. A specific study could be designed to establish a new set of criteria for the scientific classification of coals.