USGS Home
Contact USGS
Search USGS

 Go to:      NAWQA Home     National-Synthesis Assessments

Nutrients National Synthesis Project

Summary of nutrient concentrations for streams and ground water - July 2000

This data was provided for use by the Heinz Center for the report "The State of the Nation's Ecosystems" published in 2002.

Explanation for national analysis of nutrient concentrations in streams and rivers
by David K. Mueller

Flow-weighted mean concentrations for 5 nutrient species are presented for 372 fixed sites in the 1991 and 1994 NAWQA study units. The nutrient species are ammonia (NH₄ - includes dissolved ammonium plus unionized ammonia), nitrite plus nitrate (NO₃), total nitrogen (TN), orthophosphate (PO₄), and total phosphorus (TP). A regression model was calibrated to estimate the daily load of each constituent at each site. The calibration data included periodic sample concentrations and daily streamflow measurements. The models were used to estimate daily loads of each species based on daily streamflow for a period of 1 or 2 water years. The daily loads and daily streamflow volumes were summed for the entire period, and the ratio of the sums was used to calculate the flow-weighted mean concentration for each species.

The concentrations of TN used in model calibration were calculated from measured concentrations of total Kjeldahl nitrogen (TKN - ammonia plus organic N) and NO₃. If NO₃ was less than detection (0.05 mg/L), TN was assumed to be the concentration of TKN. If TKN was less than detection (generally 0.2 mg/L), TN was assumed to be <0.2 or the concentration of NO₃ if this was >= 0.2.

The model considers concentrations that are less than detection, but requires a minimum of 7 values greater than detection for calibration. At many sites, one or more species failed to meet this criterion. In these cases, the flow-weighted mean concentration was estimated to be less than the highest detection limit during the sampling period. In some other cases, where a model estimate was possible, the flow-weighted mean concentration was less than this detection limit. For non-biased comparison among sites, these estimates were censored at the highest detection limit for each species, which was considered the minimum reporting level (MRL). Mean-annual concentrations computed less than the MRL are shown in the spreadsheet in blue as negative values of the MRL.

The flow-weighted mean concentrations were checked by comparing TN to the sum of NH₄ plus NO₃ and by comparing TP to PO₄. Discrepancies were evaluated and rectified by replacing the concentration that seemed to be erroneous. In most cases, TN values were replaced by the sum of NH₄ and NO₃ and TP values were replaced by PO₄. In a few cases, NO₃ values were replaced by the difference of TN and NH₄. These replacements were made under the assumption that particulate N or P was negligible at the affected sites. The replacement concentrations are not necessarily accurate, but are reasonable for use in relative ranking among sites. All replacements are shown in red in the spreadsheet.

In a few cases, the model could not be calibrated or substantial errors were indicated. In these cases, no estimated concentration was included in further analyses, and the value in the spreadsheet is a blank.

The spreadsheet also lists selected ancillary data for each site. Basin areas, land use percentages, population, and crop group are from national synthesis analyses based on national GIS coverages and study-unit submitted basin boundaries. Land use classifications generally were based on land-use percentages, but in some cases were modified based on information provided by the study units.

Explanation for statistical analysis of nutrient concentrations in ground water
by Bernard T. Nolan

Statistical analysis of nutrient concentrations is based on data retrieved in July 2000 from the NAWQA National-Synthesis Data Warehouse. These data represent land-use studies and subunit surveys conducted by study units that began in 1991 and 1994. These results supersede those posted in April 2000, which were based on an earlier data aggregation.

Sites selected for data analysis are designed to minimize sampling bias from different sampling strategies. For this reason, the following sites are excluded from this data subset:

• springs and drains, because of uncertainties in the source of water and contributing land-use area;
• networks with less than 10 wells (unless in undeveloped areas, such as forest land);
• spatially co-located (nested) wells, to preclude redundant sampling of an area for purposes of national comparison;
• and networks sampled densely within a small area, to preclude over-representation of areas.

For example, albelusag2 was excluded because it consists of domestic wells located next to installed wells comprising albelusag1; and trinlusur2 and trinlusur3 were not used because they represent nested wells sampling deeper ground water beneath network trinlusur1. One well each was retained from densely sampled networks acfbluscr1 and acfbluscr2 and added to the larger overlapping network acfblusag1. Networks ucollusrc1a-e were combined to yield an aggregate land use study with greater than 10 wells.

Nutrient analyses reported here consist of dissolved concentrations of ammonia (P00608), nitrite-plus-nitrate (P00631) (designated "nitrate" because nitrite contribution generally is negligible), and orthophosphate (P00671). To preclude undue influence on results by wells that were sampled several times, only the most recent sample per well was used in statistical analyses.

Nutrient concentrations were statistically analyzed first by determining median concentrations in ground water for each land-use study and subunit survey, then by using these medians as input to additional analyses to compute national statistics such as percentiles and rankings. Ranks were determined for networks grouped by each type of ground water study (agricultural, urban, subunit survey, undeveloped land, and mining). Using network medians as a basis for national statistics negates sampling bias introduced by network size. Some networks have more than 100 wells, but most have 30 wells or less. Medians were used as a measure of central tendency because they are resistant to the effects of outliers typical of skewed data sets. Censored values were set to one-half the detection limit before calculating network medians. In the case of ammonia, which has two reporting levels, the higher one (0.015 mg/L) was used. Exceedance percentages of the nitrate maximum contaminant level (MCL) were determined for each land-use study and subunit survey based on the number of wells in the network with nitrate concentration greater than 10 mg/L.

Table 1 in the data spreadsheet shows the number of observations, median concentrations, and ranks of ammonia, nitrate, and orthophosphate for each land use study and subunit survey. Network medians that are less than the method reporting level are shown in tables 1 and 2 as the negative value of the reporting level. For example, a network median nitrate concentration of 0.025 mg/L (indicating that more than half the samples in the network had nondetectable nitrate concentration) is shown as -0.050.

The percentage of wells with nitrate exceeding the MCL of 10 mg/L is shown in table 1 for each network, but we recommend interpreting exceedance percentages for subunit surveys only. Subunit surveys are assumed to sample deeper ground water ("major aquifers") commonly used for drinking. In contrast, land use studies sample shallow, recently recharged ground water that might or might not be used for drinking.

Table 2 in the spreadsheet shows quartiles and other percentiles for each nutrient, based on median concentrations in agricultural and urban land-use studies. For example, the highest median nitrate concentration associated with any agricultural or urban land-use study is 13 mg/L, and the overall median nitrate concentration for these land uses is 1.95 mg/L. Agricultural and urban network medians were lumped in this analysis.

Table 3 in the spreadsheet shows selected percentiles of nitrate exceedance percentages associated with subunit-surveys. For example, 90 percent of the subunit surveys had a nitrate MCL exceedance percentage of about 10 percent.