What pattern of sampling sites provides the best regional picture of trace elements in soils, for a given cost? Data from the NURE (National Uranium Resource Evaluation) Survey are sufficiently dense in the Sacramento Valley to Lake Tahoe, CA area to test several approaches. Subsets of the existing data were selected to simulate three sampling approaches. First, an areally unbiased set of 30 samples was selected, with each sample randomly selected from within equal-sized polygons. Second, five samples were randomly selected within each of the six major rock types of the area. Third, 30 samples were randomly selected from within the major soil orders of the area, an equal number from each soil order. Comparisons of each approach were performed by computing the overall mean and its 95% confidence interval for each of the three approaches. For approaches two and three, these statistics were weighted according to the geographic area of each rock type or soil order represented. Accuracy of the approaches was determined based on how closely the computed mean of 30 samples came to the mean of the entire data set for the area. Precision was evaluated by how wide were the confidence intervals of each approach. The most cost-effective approach, given that each approach used 30 samples, is the one with the smallest confidence interval. It provides the greatest precision of the estimated mean for the given cost. This evaluation was performed for several soil trace elements, including those whose primary sources are geologic or anthropogenic. Preliminary results for chromium, whose source is predominantly geologic, show that sampling based on rock type categories was most cost-effective.