The NURE (National Uranium Resource Evaluation) geochemical database (http://greenwood.cr.usgs.gov/pub/open-file-reports/ofr-97-0492/) is characterized by two important attributes: its massive size, and the dubious quality of some of the data. We believe that this database is too valuable to disregard but too problematic to be taken entirely at face value. Previous studies using NURE data have concentrated chiefly upon cartography and spatial analysis, while giving less attention to geochemical aspects of the data.

We are examining the NURE stream-sediment and soil data for a region of the Southwest that is centered on AZ and includes parts of adjacent CA, NV, UT, CO, and NM. We are emphasizing geochemical criteria for evaluating the data, which comprise some 39,400 samples from 31 2°-by-1° (1:250,000) quadrangles. Initially, we have focused upon the Nogales and Tucson quads (NTQ, 2960 samples) and the Mesa quad (MQ), southern Arizona.

Evaluation methods include searching for unrealistic values or patterns; study of histograms; comparison with average upper continental crust (UCC); and comparison of data for pairs of geochemically similar elements, such as Zr-Hf, La-Ce, Th-U, and V-Fe. Some examples: (1) Nearly all of the 710 determinations of Cr in the MQ are >100 µg/g, median 170 µg/g. These values are too high and too homogeneous, and inconsistent with more reasonable data in adjacent quads. The MQ data have other problems as well. (2) In the NTQ, at least 10 elements, with a wide range of geochemical properties, have median concentrations essentially the same as UCC: Ba, Ti, V, Mn, Fe, Zn, La, Ce, Th, and B. This agreement, surely not coincidental, strongly argues against any assertion that these NURE data are simply invalid. (3) La-Ce ratios in virtually all terrestrial igneous rocks (and therefore in derivative sedimentary and metamorphic rocks) should be bounded by those of MORB (La/Ce ª 0.3) and lamproites (La/Ce ª 0.6), the silicate igneous rocks with least and greatest LREE enrichment. Nearly all of the NTQ data satisfy these criteria. The La-Ce plot reveals a tight linear array, encompassing > 90 % of the data, concentrated along La/Ce ª 0.5 and bracketed by 0.4 £ La/Ce £ 0.6. Virtually no samples have sub-MORB La/Ce. Less than 5 % of the data have apparent La/Ce > 0.6; these few analyses are suspect.