Sources and Description of Data

The data used for analysis of trends and determination of loads were retrieved from the USGS NWIS and the USEPA STORET databases and include streamflow; physical properties (temperature, dissolved oxygen, pH, alkalinity, and specific conductance); nutrient concentrations (unfiltered nitrogen, nitrite plus nitrate, ammonia, unfiltered ammonia plus organic nitrogen, and unfiltered phosphorus); suspended-sediment concentrations; and organic carbon and major-ion concentrations (chloride, calcium, sulfate, sodium, and silica). Hereafter in this report, the terms “unfiltered” and “filtered” refer to “total” and “dissolved,” respectively.

U.S. Geological Survey National Water Information System

U.S. Environmental Protection Agency Storage and Retrieval Database

The STORET database is a USEPA repository for water-quality, biological, and physical data collected by Federal, State and local agencies, university researchers, and others. Data retrieved from STORET were limited to nitrogen and phosphorus concentrations for WYs 1973–2004. The water-quality data for sites in Mississippi, Alabama, and Virginia for the period 1999–2004 were retrieved directly from databases maintained by the respective individual State agencies.

The STORET data were retrieved as part of the development of the regional SPARROW (SPAtially Referenced Regressions On Watershed) attributes model that is used to support NAWQA Program activities (Anne B. Hoos and Gerard McMahon, U.S. Geological Survey, written commun., 2007). The STORET data retrieval, with parameter codes in brackets, included unfiltered nitrogen [00600], unfiltered organic plus ammonia nitrogen [00625 or 00635], unfiltered and filtered ammonia [00610 and 00608], unfiltered organic nitrogen [00605], and unfiltered and filtered nitrite plus nitrate [00630 and 00631]. Phosphorus data included unfiltered phosphorus [00665], filtered phosphorus [00666], and suspended phosphorus [00667]. The data were screened to remove anomalous values and values with remark codes other than “less than” (<) and “greater than” (>). When values for unfiltered nitrogen [00600] were missing, values for unfiltered organic plus ammonia nitrogen [00625, 00635, or the sum of 00605 and 00610] were combined with nitrite plus nitrate nitrogen [00630 or 00631] values to estimate values for unfiltered nitrogen. When values for unfiltered phosphorus were missing, filtered [00666] and suspended [00667] phosphorus values were combined to estimate values for unfiltered phosphorus. As a result, the unfiltered nitrogen parameter is designated “60000,” and the unfiltered phosphorus parameter is designated “66500” in this report.

Basin Characteristics

Regional landscape variables and agricultural nutrient sources were compiled for each drainage basin for both the NWIS and selected STORET sites. Ancillary basin characteristics include variables for population density, atmospheric deposition of nitrogen, agricultural nitrogen and phosphorus (fertilizer and manure), land cover, soil characteristics, surficial geology, hydrologic characteristics, and ecoregions. The time periods covered by these data vary but are between WYs 1974 and 2006. The data sources and methods used to compile basin ancillary data are reported in Ruddy and others (2006) and Falcone and others (2007). These data were compiled to provide variables to categorize the trends-analysis results and to investigate possible relations between annual variations in water quality and basin characteristics. These data are presented in the section, “Time Series of Basin Characteristics” in this report.

Data-Compilation Methods

The following sections describe the methods for assembling and modifying the datasets prior to analysis. Additional information is provided on the methods used to select the sites that were included in the analysis, modify the censoring levels to ensure consistency, combine related constituents to create more complete datasets, and include streamflow data with the water-quality data.

Site Selection

Sites were selected from the NWIS database for time-trend analysis and load estimation by using summary information for the period of record (POR) and reviewing nutrient-concentration time-series scatterplots. The initial NWIS search included sites in the study area (HRs 03 and 06) where at least 20 water-quality samples had been collected during the POR and at least 1 sample had been collected during WYs 1973–2005 for analysis of nutrient or suspended-sediment concentration. Sites with discharge records for WYs 1973–2005 were identified, and after a review of nutrient-data time-series plots, the NWIS sites were prioritized as (1) sites with continuous discharge record for WYs 1973–2005 and (2) sites with recent (2004) record and at least 10 years of record.

Selection of STORET sites for time-trend analysis was determined by the minimum data requirements of the USGS trend-analysis computer program S-ESTREND (Schertz and others, 1991; Slack and others, 2003). The initial retrieval of STORET sites numbered in the thousands and was reduced to 290 sites by automated constraints within the program, which is described in the section, “Detection of Trends.” Sites were included in the analysis if sufficient data were available for at least one of the following periods: 1975–1985, 1985–1995, 1993–2004, and 1975–2004.

Modification of Censoring Levels

Trend-test methods and the choice of an appropriate test can be influenced by reported chemical values that have been censored (noted as nondetection or less-than (<) values) by the analyzing laboratory. Large amounts of censoring or censoring at different reporting levels are two important factors in choosing an appropriate trend test for a dataset. Thus, efforts were made to ensure that reporting levels, also referred to as censoring levels, were used in a consistent manner and that re-censoring techniques were applied to account for variability in reporting level caused by inconsistent application. Also, because analytical methods and performance can change over time, reporting levels can change as well. No effort was made to re-censor data to a common reporting level when the cause of variability was the result of changes in method or analytical performance, because this type of reporting-level variability accurately reflects the methods and performance at the time of analysis.

Modifications to censoring levels were performed only on the NWIS dataset because of the availability of reporting-level history and usage. It was not feasible to modify censoring levels for the STORET dataset because of the multiple sources of data, greater variation of laboratory methodologies and quality-control procedures, and the lack of reporting-level history and usage.

Historically, the USGS National Water-Quality Laboratory (NWQL) has used the minimum reporting level (MRL) for reporting detections; MRL is defined as the smallest concentration of a substance that can be measured reliably using a given analytical method (Timme, 1995). The reliability of the measurement can be determined by statistical methods or subjective criteria, including an analyst’s judgment (Oblinger and others, 1999). Since the MRL definition is not numerically specific, the NWQL began censoring data in 1996 at the laboratory reporting level (LRL). The LRL is established by using a consistent statistical method that reduces the chances of reporting false-negative results. The LRL generally is twice the method detection level (MDL), which is described as the minimum concentration of a substance that can be measured and reported with 99-percent confidence that the analyte concentration is greater than zero (Oblinger and others, 1999). This change in reporting level by the NWQL can create an artificial upward trend, especially in heavily censored datasets; therefore, all NWIS censored data reported with an LRL were re-censored to the MDL by dividing by 2. The letters “RC” appended to a parameter code indicate that the concentration has been re-censored (for example, P00625RC).

As with some other constituents, ammonia [00608] concentration data includes various reporting levels over the last 15 years because of changing analytical methods. The NWQL evaluated historical ammonia data, however, and recommended re-censoring to a common level; thus, all censored ammonia data were re-censored to < 0.02 mg/L (Mueller and Spahr, 2005).

Combining Related Properties and Related Constituents

In an effort to create complete datasets, related properties or constituents in the NWIS dataset were combined. Field and laboratory values for pH [PH] and specific conductance [SPCOND] were combined for each property by preferentially selecting the field measurement when available and using the laboratory measurement when field values were not available. Field alkalinity [39086] and acid-neutralizing capacity [00410] were used to create the single parameter ALK, preferentially using alkalinity (39086). In a similar manner, the flow [FLOW] dataset was populated first by using instantaneous discharge [00061] or, when unavailable, daily mean discharge [00060].

Throughout the period for which data were analyzed (1973–2005), laboratories used various reporting units and analytical methods for specific constituents. In addition, the nutrient species that were analyzed varied by project. Combined constituents were created, therefore, to consolidate multiple reporting units, analytical methods, and nutrient species in order to develop more complete and extensive datasets for specific constituents.

The calculation of combined constituents requires a specific conversion factor and a priority for each NWIS candidate parameter code used in the calculation. For example, specifying “00671, 1.0, 1; 00660, 0.3261, 2” for the calculation of filtered orthophosphate in milligrams per liter (mg/L) as phosphorus (P) means that priority 1 is given to parameter 00671 (orthophosphate, filtered, in mg/L as P) and the value is multiplied by 1.0. If parameter 00671 (orthophosphate, filtered, in mg/L as P) is not available, the next step in this example is to check for availability of the parameter with priority 2 (00660, orthophosphate, filtered, in mg/L as PO₄), and the value is multiplied by 0.3261 to obtain the value for orthophosphate in mg/L as P. In the case of the combined constituent unfiltered nitrogen [50000], nitrogen [70000] was combined with a preferred nitrate species (C.R. Kratzer, U.S. Geological Survey, written commun., 2004). The calculations used for the combined constituents are presented in table 2.

Streamflow Data

For each water-quality sample in the NWIS and STORET datasets, a mean value for streamflow on the day of sampling was retrieved from the surface-water data in the NWIS database (U.S. Geological Survey, 2007) and referenced to the water-quality sample if an instantaneous measurement was not available. For the water-quality datasets retrieved from STORET for which the water-quality sampling sites are not explicitly paired with a streamflow gaging site, a nearby USGS streamgaging station had to be identified. The National Hydrography Dataset Plus (NHD-Plus) was used to match ungaged water-quality sampling sites with nearby USGS streamgaging stations. If a USGS streamgaging station was within the upstream or downstream catchment areas of the ungaged water-quality site, drainage areas of the two sites were compared. The criterion for a goodness of match between sites was a drainage-area pairing ratio between 0.75 and 1.25 (Cassingham and Terziotti, 2006). When a matched-flow site was found, daily mean streamflow values on the sampling days were added to the water-quality dataset.