Mineralization distributions are very heterogeneous in nature. As large or superlarge mineral deposits are quite rare whether in time or in space, it is difficult to detect all the largest mineral deposits in a region in a limited period of time owing to the restriction of technology and exploration degrees-this is called "not all discovered". However, all discovered large, especially superlarge, mineral deposits generally have a complete census in the geological literatures. On the other hand, not all discovered small mineral deposits are recorded in the geological literatures because for economic reasons people have not much interest in them-this is called "not all recorded". This practice often results in the observation truncations, that is, the data points near the two ends in an observable population, which is obtained by fitting a power law size-frequency distribution to discovered mineral deposits in a given region, show concave-down departure from the correlation line fitted. The authors suggest that the size and number of undiscovered deposits may be forecast by fitting a fractal size distribution to discovered mineral deposit sizes between the upper and lower truncation observations and then extrapolating the scale-independent area to deposit sizes larger than the upper truncation limit. Based on the statistical results obtained by the fractal size-frequency distributions of 394 discovered gold deposits with sizes greater than 21 Au in China and 83 known gold deposits with sizes over 0.3 t Au in the Jiaodong area of China, the authors forecast according to the present commercial standards for gold ores that the total resources of undiscovered gold deposits ranging in tonnage from 50 to 2000 t Au are more than 4500 t in China, and that in the Jiaodong area of China the total resources of potential gold deposits with sizes in the range of 30 to 650t Au are about 700t.