On Oct. 21, 1976, Congress enacted the Resource Conservation and Recovery Act (RCRA). In the course of promulgating regulations for managing hazardous wastes pursuant to Subtitle C of RCRA, the U.S. Environmental Protection Agency (EPA) deferred promulgating regulations for six "special wastes" until appropriate regulations could be determined. Cement kiln dust was one of these wastes. When the Solid Waste Disposal Act Amendments of 1980 was enacted, dust was temporarily exempt pursuant to a study of the adverse effects to human health and the environment from the disposal of the dust.

Since 1980, three studies of trace metals in dust have been conducted and reported because of this new interest in whether or not dust may be a hazardous waste. The first was Characterization of U.S. Cement Kiln Dust (Haynes and Kramer, 1982). The second was An Analysis of Selected Trace Metals in Cement and Kiln Dust (Portland Cement Association, 1992), and the third was Report to Congress on Cement Kiln Dust (U.S. Environmental Protection Agency, 1993).

Limitations of existing studies All of these studies were similar in that only one sample was analyzed from each plant and conclusions were drawn about the concentrations of trace metals in dust based upon the means of these data. The combination of these data in this manner is only statistically proper if all the samples were drawn from a population with mean = m and variance = s[superscript]2 (Miller and Miller, 1988 and Sokal and Rohlf, 1969). There is no evidence that any of the studies tested these null hypotheses before combining these data. It is more likely that it was assumed that the mean = m and variance = s[superscript]2 for every kiln and plant in these studies, because all the dust sampled came from portland cement plants. Therefore, all the kilns and plants were assumed to be alike. However, the independent variables of plants that affect dependent/variables' parameters can be grouped: raw materials, equipment, fuel, and control-room operations.

The sources for the raw feed that is converted to clinker in the rotary kiln by the pyroprocess are legion. Although a major component is limestone, constituents of limestone other than calcium carbonate vary according to geological formation and geographical occurrence (Hill et al., 1967). The other sources of the raw feed may be naturally occurring clays (Degans et al., 1957, 1958; Eckle, 1904) and sands (Hill et al., 1967) as well as anthropogenic sources such as fly ash, foundry sands, municipal sludge, spent catalyst, metallurgical slag, fuel ash, and other sources of oxides of silicon, aluminum, and iron. The distribution of constituents of the raw feed varies according to the physical and chemical properties of its sources and the equipment used for crushing, blending, and grinding.

The configuration of the equipment of a portland cement plant is determined by many factors such as site, age, sources of raw materials, engineering, and cost. These variations are expressed in wet kiln systems and dry kiln systems. The dry kiln systems may be long dry kilns, two-stage preheater kilns, four-stage preheater kilns, four-stage preheater-precalciner kilns, and six-stage preheater-precalciner kilns to name a few. In addition, there are structural and operational differences in the kilns, preheaters, and precalciner as determined by the manufacturer.

The differences in the fuel are due to its source and state. Common fuels are coal, petroleum coke, natural gas, No. 6 fuel oil, and hazardous waste derived fuel. Hazardous waste derived fuel is generally a blend of numerous waste streams. Some hazardous waste derived fuel is in solid form and burned in kiln systems equipped to handle solid material. All of these fuels are going to vary in their concentrations of constituents as based upon their sources, blends, and BTU values.

The operations of a kiln system are affected by the raw mix, equipment configuration, fuel, and operators. Consequently, kiln systems are operated at different temperatures, burn different types of fuel, produce different amounts of dust, and require different quantities of energy in order to produce clinker.

Although these differences among portland cement plants is only general, it is patently obvious that it is incorrect to conclude that statistics derived from one dust sample from each of many portland cement plants are estimations of parameters of all portland cement plants. All three studies suffer from inadequate data to draw conclusions about the distribution of trace metals in dust of the portland cement plants in the United States. If the assumption that all plants are alike were correct, then the number of samples reported in the studies of the BOM (113) and the PCA (97) would be adequate for the statistical estimation of the parameters of this population.

On the other hand, the EPA, distinguishing the plants sampled in its study to those burning conventional fuels (seven samples) versus those burning hazardous waste derived fuels (five samples), did not take enough samples to estimate these parameters with any significant amount of confidence, even if the assumption that all portland cement plants belong to the same population.

The Haynes and Kramer (1982) study was adequate to have served as a preliminary investigation to so as to aid the design of future sampling projects to obtain descriptive statistics of trace-metal concentrations in dust. To the extent that its purpose was to replicate the Haynes and Kramer (1982) study, the PCA (1992) study met that goal. Any other use of these data is inappropriate.

A new study Until 1996 there was neither reported interest nor data available to test the null hypothesis that mkiln 1 = mkiln 2 for concentrations of trace metals in dust produced by different kilns. Fortuitously, Region VII of the EPA made an all inclusive RCRA section 3007 request for dust data from Region VII portland cement plants burning hazardous waste derived fuels. A remarkable discovery was made while perusing this data. Ash Grove Cement Co. included in its data concentrations of total metals in CKD generated while burning hazardous waste derived fuels and while burning conventional fuels.

Results This request obtained the results of the analysis of more than 20,000 samples consisting of some 200,000 datum points collected at two kilns at each of two portland cement plants located in Chanute, Kan. and Louisville, Neb. Two types of fuel (coal and hazardous waste) were burned in each of these four kilns. Samples taken hourly were composited daily for analysis for trace metals. The results were reported daily according to fuel types.

In order to demonstrate that there is a significant difference in the means of concentrations of trace metals from different sources, a series of Student's t-tests were performed to determine if it was appropriate to combine the data from multiple kilns at the same cement plant. The results of Student's t-tests for the concentrations of barium, cadmium, chromium, and lead in dust reveal that these kilns operate as independent entities and that they produce dust that is different.

* Significant differences were observed in 53% of the tests between kilns at the same plant while burning the same type of fuel. Compilation of the results of Table 1 are:

* Significant differences between kilns were observed in 20% of the t-tests for the Chanute, Kan. plant while burning conventional fuels;

* Significant differences between kilns were observed in 60% of the t-tests for the Louisville, Neb. plant while burning conventional fuels;

* Significant differences between kilns were observed in 33% of the t-tests for the Chanute, Kan. plant while burning hazardous waste;

* Significant differences between kilns were observed in 58% of the t-tests for the Louisville, Neb. plant while burning hazardous waste.

Conclusions The conclusion from the above results is that the statistical estimation of parameters of the concentrations of trace metals in dust cannot be determined from the statistical analysis of the results of single samples or even multiple samples taken from each of many different kilns. Each kiln is a population unto itself. Consequently, its parameters are unique and they can only be statistically estimated from the analyses of an adequate number of samples taken from dust generated from that kiln while its variables remain unchanged. Therefore, the use of the measurements of the Haynes and Kramer (1982), PCA (1992), and the EPA (1993) to estimate parametric means and variance was statistically incorrect and consequently, nullifies their conclusions.

The EPA (1993) also drew erroneous conclusions regarding the affects of burning hazardous waste derived fuel and coal on the concentrations of metals in dust.