Geophysics and Geochemistry - Basic Statistical Measures Used in Geochemical Investigations of Colorado Plateau Uranium Deposits

Distribution of minor elements in sandstone-type uranium deposits of the Colorado Plateau is being studied by the USGS on behalf of the A EC. A primary objective has been to provide basic data which could be useful in geo-chemical exploration for new deposits. The basic statistical measures — similar to those used in sedimentary petrology—are described. The underlying principle is that the frequency distributions fit or approach the log-normal form. Since 1952 the U. S. Geological Survey on behalf of the U. S. Atomic Energy Commission has conducted a study of the distribution of minor elements in sands tone-type uranium deposits of the Colorado Plateau. A primary objective in this study has been to provide basic data which could be useful in geo-chemical exploration for new deposits. Among the basic data considered important are estimates of the abundance and habit of each element in the ore and in the host rocks. It has been found that a few basic statistical techniques, and variations of these techniques, are useful in this type of study. Some of the basic techniques have been used to advantage by researchers in geochemistry and in other fields of geology, especially sedimentary petrology; others have been used to only a limited extent. Among the latter is the linear correlation coefficient, which has proved useful in assessing the element associations in a particular rock type without analysing individual mineral separates. In computing the basic statistical measures a scale has been used which is, in some ways, similar to the phi ($) scale used by sedimentary petrol-ogists in grain-size studies.' FREQUENCY DISTRIBUTIONS It has been recognized for some time that the results of analyses for a given minor element in a suite of samples of rocks or ores yield frequency distribution curves which are approximately normal when the curves are plotted on logarithmic or geometric scales. In other words, the analytical results give an approximate lognormal frequency distribution. A classical example is given by D. G. Krige2 in the study of the distribution of gold values from a mine on the Witwatersrand. Another example is shown in Fig. 1, the distribution of barium concentrations in samples of uranium ore from the Chinle formation. The geochemical data are analogous to grain-size data gathered by the sedimentary petrologist. Both types of data, after simple logarithmic transformations, may be treated by conventional statistical methods based on the normal distribution. This is not to say that other types of transformations may not serve equally well or better for particular element distributions.3 Spectrographers and other analysts are quite aware of the dependence of their precision on the concentration of the element present.4 For example, in the part per million range the variation may