Analysis of the discovery order of mercury deposits in California clearly shows that larger deposits tended to be discovered early in the exploration process. A direct consequence of this process is that the frequency distribution of sizes determined at an intermediate exploration stage overestimates the frequencies of large deposits and underestimates the frequencies of small deposits of the yet-to-be discovered deposits.That is, the size distribution based on incomplete exploration of a region is biased with respect to the size distribution of the undiscovered deposits. The maximum likelihood method (Chung, 1990) tested here provides reasonable estimates of the sizes of undiscovered mercury deposits in California. Al most all the metal is confined to the three largest size classes. Overestimates of the number of deposits in the largest size classes are a serious cause for concern, whereas biased estimates for the small size classes are unimportant. Although the results presented here are based on an analysis of the data available in 1900, similar results can be obtained by using any year between 1890 and 1940. Tests of the discovery order versus sizes of other kinds of deposits are clearly warranted;however,care must be exercised in these studies to insure that the analysis Is performed with in the same exploration and geologic settings. This is necessary to avoid cases where,for example, exploration of the exposed bedrock is followed later by exploration for deposits under cover.The exploration process should be considered to have started over because neither small nor larged eposits would have been searched for in this environment in the first stage. Where the discovery order of the known deposits can be obtained, the sizes of the deposits should be plotted with respect to the discovery order. If the decline is significant, the distribution function of the size of deposits should not be estimated directly from the discovered deposits. The observed sizes of discovered deposits are biased samples similar to oil and gas deposits. Therefore the parameters of the size distribution function of deposits should be estimated using the methods applied to oil and gas deposits (Kaufmann, 1986; Wang and Nair, 1988). If such methodologies are not available, then the method proposed here is recommended to estimate the parameters. Where the discovery order is unknown,the maximum likelihood method proposed here is the only method known to us that can be used to predict the size distribution of undiscovered deposits. To apply the proposed technique, it is necessary to assume that the total number of deposits (discovered and undiscovered) is known. Because that number is not known, we suggest that the estimation procedure be repeated several times by fixing the total number of deposits equal to plausible values. Although the analysis of the problem and the maximum likelihood method proposed here are discussed in terms of the lognormal distribution, the results and procedures can be easily extended to other types of distribution functions such as the exponential and gamma.
