DB Doc


Sampling Methods in the NGS

A wide variety of methods were used to collect the samples in the NGS database. These are described below in nine categories that are referenced in the table of datasets composing the NGS.
  1. Inherited. These datasets were incorporated into the NGS directly from pre-existing studies. Either the sampling protocol and analysis-of-variance methodology are unknown, or they may be found in references cited in the dataset description.

    Datasets in this category:

  2. NURE-Other. Reanalyses of NURE samples done by USGS projects other than the NGS. Prior to the NGS, numerous USGS projects reanalyzed samples from the NURE archives. Other USGS projects have continued to reanalyze NURE samples in parallel with the NGS. In the majority of these cases, most or all of the NURE samples in an area were reanalyzed. Sample densities and collection methods are therefore mostly the same as in the original NURE program and are documented in the hundreds of NURE reports; see Averett (1984) for an overview and index to these reports.

    Datasets in this category:

  3. NURE-Systematic. Systematic reanalyses of NURE samples done by the NGS. An archive of stream sediment and soil samples collected by the NURE program is stored at the USGS in Denver, Colo. Original sampling strategies and field methods are documented in NURE reports; see Averett (1984) for an overview and index to these reports. For the purposes of the NGS, a subset of samples from the archives was selected for reanalysis. A sample density of ~1 per 289 km2 was adopted as a minimum standard across the US. This was achieved in two ways, depending on how the samples were physically archived. In the eastern US, most NURE samples were stored in boxes corresponding to single counties. A calculation based on land area was made to determine the number of samples needed from each county to achieve the minimum density. Samples were then chosen at random from each county-box until the desired number was reached. In the western US, samples were stored in boxes corresponding to 1:250,000 scale quadrangles. For each quadrangle, a grid was created dividing the area into cells of ~17x17 km size. For example, in the central US, 1x2 degree quadrangles were divided into 9 columns east-west, and 7 rows north-south, producing 63 total cells. A sample was then selected at random for reanalysis from each cell.  For either of these selection methods, if a selected sample had insufficient mass for reanalysis or the sample could not be located in the archive, another was chosen at random from the same cell.

    In selected study areas, samples were chosen to give higher total densities. In these areas, the same county or cell-base scheme was used, but more samples were selected from each unit. These study areas included New England, parts of the Atlantic and Gulf of Mexico Coastal Plain Provinces, and transects across Cretaceous basins in the western US and north slope of Alaska.

    Approximately 2% of reanalyzed samples were selected at random for analysis of variance (AOV) studies. These samples were split into two fractions and analyzed separately. In addition, 10 quadrangles from the western US (including Alaska) were subsampled twice using the same cell-based scheme, producing two sets of samples covering the entire quadrangle area.

    Datasets in this category:

  4. NURE-Targeted. Targeted reanalyses of NURE samples done by the NGS. Suites of NURE samples from a number of areas of the US were reanalyzed by the NGS for a variety of reasons. Pilot studies for the NGS were done in Alabama (NURE Alabama I dataset) and in the Atlantic coastal states (NURE 800 dataset). Later studies were done of the Alaska and Atlantic and Gulf of Mexico Coastal Plain Provinces (NURE Alaska 99, NURE Coastal 98, NURE Coastal 99, NURE South Carolina), and Great Plains (NURE Midwest). Sampling densities for these studies were variable, but all used a random-selection method and AOV strategy similar to that described for NURE-Systematic datasets.

    Datasets in this category:

  5. USGS-Resampling. Reanalyses of USGS samples done by the NGS. Archives of stream sediment and soil samples collected for previous studies are stored at the USGS. The archive includes most of the samples for which there are analytical data in the National Geochemical Database, including those collected by USGS programs. The original collection methodology for these samples is either unknown or documented in primary literature (see references cited in the dataset description). Samples were selected for reanalysis and AOV using the same method described above for NURE-Systematic datasets.

    Datasets in this category:

  6. SE US States. Collaborative sampling programs by the USGS and states in the southeastern US.

    In Florida, Mississippi, Georgia, Alabama, and South Carolina, collaborative programs between the USGS and state government agencies were established to collect geochemical samples at a scale of ~1 sample per 100 km2. In all of these states except South Carolina, the primary sample medium was stream sediments. Where streams suitable for sampling do not exist, including the Mississippi River alluvial plain ("The Delta") and coastal lowlands, soil samples or other media such as beach sands and canal sediments were collected. In South Carolina, the primary sample medium was surficial soils taken from the "plow zone" (see below), which complemented an existing collection of stream sediments available in the NURE archive.

    Sample collection was based on 10 x 10 km grids drawn in the UTM coordinate system. Each 10 x 10 km cell was divided into four quadrants, and one was selected at random for sampling.

    For stream-sediment collection, small streams within the quadrant were identified, and one was randomly selected for sampling. The drainage basin of the sampling site was largely within the quadrant. Sample locations were chosen to be away from obvious sources of contamination, including roads. Ideally, at each stream site, 6 to 10 depositional zones containing fine-grained particulate matter in a 10-20 m reach of stream were sampled for the <2 mm size fraction. The goal was to select depositional zones representing upstream influences and various flow regimes, including left bank, right bank, center channel, and different depths of water. Each depositional zone at a site was sampled several times, and these were composited with samples from other depositional zones at the same site. After homogenization, a subsample was submitted for chemical analysis. All samples were sieved to -100 mesh (<150 μm) in USGS laboratories prior to analysis.

    For soil collection sites, a minimum of three samples of the Ap horizon ("plow zone"), operationally defined to be the top 15 cm of soil, were taken about 10 meters apart in a triangular spacing, and then composited, mixed, and subsampled for chemical analysis. Soils that were clearly disturbed or that appeared to be contaminated by a point-source of pollution were avoided. Samples in farm fields, forests, meadows, and in "upland" locations were the most satisfactory. Samples from marshes, swamps, and floodplains were avoided. Again all samples were sieved to -100 mesh prior to analysis.

    Samples were also collected for analysis of variance (AOV) studies. For stream sediments, AOV sampling strategy was designed to measure differences of sediment chemistry between cells, within cells, within streams, and between chemical analyses. 5-10 percent of the cells were randomly selected for the AOV sampling. In these cells, the primary sample was designated with the label "D1". One AOV sample was collected from a different stream in a different quadrant than the primary sample; this was labeled "D2". A second AOV sample was collected 50-100 meters upstream from the D2 sample; this was labeled "D3". A third AOV sample, labeled "D4," was later split from the D3 sample in the laboratory. AOV sampling for soils was done in an analogous manner, except the D3 site was selected 50-100 m away from the D2 site.

    Sample collection equipment was made of contaminant-free stainless steel screens and sieves, augers, and polycarbonate plastic. Samples were allowed to air dry at low temperatures below 50°C to avoid volatilizing mercury.

    To provide a check on laboratory performance, one standard samples of known composition was included for analysis with every 10 to 20 unknowns. Two in-house USGS reference materials, SAR-L.1 and SAR-M.1, were used for this purpose. The composition of these standards is listed in Folger (2000).

    Datasets in this category:

  7. Special. Sampling in Special Study areas in the Southeastern US.

    High-density sampling in these areas was conducted on a 1 square mile grid (based on township/range/section). Soils and stream-sediment samples were collected. Urban areas were avoided (in Tallahassee, samples were only collected from the periphery of the city). Analysis of variance sampling was conducted in the same manner as described above for the "SE US States."

    Datasets in this category:

  8. State Surveys. Collaborative sampling programs by the USGS and state geological surveys.

    Sampling programs in Louisiana, Michigan, Illinois, Indiana, Hawai`i, and parts of Alaska used the same methods as described for the southeastern states, except the sampling density was lower. In these states, the same 17x17 km grid described for resampling the NURE archive was used as the basis for sample collection. AOV samples were collected in the same way, but labeled differently (usually AOV1, AOV2, AOV3, and AOV4, instead of D1, D2, D3, and D4, as described for the southeastern states. In most of the states represented in these datasets, the primary sample medium was stream sediments. However, several states in the 2003 and 2004 datasets collected only soils (North Dakota, Iowa).

    Datasets in this category: