Sources, Instream Transport, and Trends of Nitrogen, Phosphorus, and Sediment in the Lower Tennessee River Basin, 1980-96


During the period 1980-96, Federal, State, and local agencies and universities collected water-quality samples at more than 700 stream and reservoir sites in the LTEN River Basin. Most of the water-quality data analyzed in this report were collected as part of ambient monitoring programs conducted by TVA, TDEC, ADEM, KDEP, ORSANCO, and USGS. Some additional data were collected as part of special studies conducted by ADEM and TVA (Sweatt, 1996), the FCWP (1996), and as part of a sediment data-collection network conducted by USGS.

Data from the ambient monitoring programs of TVA, ORSANCO, and State agencies were obtained from the STOrage and RETrieval (STORET) data base of the U.S. Environmental Protection Agency (U.S. EPA). Data from USGS monitoring programs were obtained from the WATer data STOrage and REtrieval system (WATSTORE) data base of USGS.

Instream loads of nutrients and sediment were estimated for water-quality monitoring sites for which samples were collected at least quarterly for a period of 5 consecutive years during which daily streamflow data also were collected (or could be adapted from a nearby gage). Sites that met these criteria are shown in figures 2 and 3 and are listed as sites 1-18 in Appendix A. Most of these sites were sampled quarterly, but four sites (sites 1, 11, 12, and 18) were sampled monthly.

Instream nutrient loads were estimated at 16 of the sites (all sites except 3 and 8, where only sediment data were collected). Streamflow data for 11 of the 18 sites were collected by USGS; TVA provided streamflow data for the remaining sites (E.A. Thornton, Tennessee Valley Authority, written commun., March 1998). Additional information for the streamflow-gaging sites is provided in table 1. Instream loads of sediment were estimated at five of the USGS sites (sites 2, 3, 8, 16, and 17).

Of the 18 sites for which loads were computed, 11 are in free-flowing, or riverine, reaches of tributaries to the Tennessee River, 2 are on flow-regulated sections of tributaries (sites 4 and 9), and 5 are on flow-regulated sections along the main stem of the Tennessee River (sites 14-18). Each drainage basin contributing to the tributary sites (sites 1-13) is shown in relation to environmental setting (figs. 2 and 3) to illustrate the composition of each basin with respect to subunits and land use/land cover. Summaries of subunit and land use/land cover for each tributary basin are shown in figure 5. The combined percentage in pasture, cultivated, and urban land use/land cover in this set of basins ranges from 29 percent (site 2, Buffalo River near Flat Woods) to 80 percent (site 1, Clarks River at Almo). Land use and subunit summaries are not shown for the sites along the main stem of the Tennessee River (sites 14-18) because the many large impoundments along the main stem in both the upper and lower parts of the Tennessee River alter the instream transport of nutrients and sediment, confounding comparison between basin characteristics and water quality for those sites.

Eight of the 18 water-quality monitoring sites were not colocated with a corresponding streamflow-gaging station, but were located some distance upstream or downstream from a gaging station (drainage areas listed in table 1 indicate a difference for the eight sites). To adjust for these differences in drainage areas, the estimates of instream load at these eight water-quality monitoring sites were multiplied by the ratios between the drainage areas of the paired streamflow and water-quality monitoring sites. A ratio within 0.8 to 1.2 was considered to be acceptable. Two of the sites, however, fell outside of this criterion: Sequatchie River at Valley Road (site 13) and Shoal Creek at Highway 43 (site 7) had ratios of 1.4 and 0.5, respectively. These sites were included in the set of instream-load computation sites to provide representation for a particular combination of land use and subunit for which no other sites were available (as shown by combinations in fig. 5), but the load estimates are presented with the qualifier that the estimation error may be significantly larger than for other sites.

Trends in concentration were estimated at sites for which samples were collected at least quarterly for 5 consecutive years, and for which streamflow data were available for each sample (either from a nearby continuous-recording streamflow gage or from measurements of instantaneous streamflow concurrent with sample collection) so that concentrations could be adjusted based on streamflow. Flow adjustment of concentration data eliminates variation in concentration related to streamflow, allowing for more accurate detection of time trends in water quality. The sites that met the data requirements for this analysis were the 18 instream-load computation sites (table 1), and sites 19 and 20 (figs. 2 and 3 and Appendix A) sampled as part of a special study conducted by ADEM and TVA in the Cumberland Plateau area in Alabama (Sweatt, 1996).

Downstream variations in nutrient concentrations along the Tennessee and Duck Rivers were evaluated by summarizing data from numerous monitoring sites along the rivers where at least 20 samples had been collected during 1980-96 and with record extending past 1989. (The last criterion was relaxed to 1985 in screening for sites on the Duck River, where only a few sites had data extending past 1985.) Of the more than 400 monitoring sites located on the main stem Tennessee River and Duck River (based on nutrient data in the STORET or WATSTORE data bases), 37 sites met this criterion and are listed in Appendix A.

Data were reviewed to ensure comparability between data from the different monitoring networks, despite variations in analytical methods and data-reporting levels among agencies. Analytical data derived from various analytical procedures were grouped when appropriate (table 2) to construct the most complete nutrient data sets possible. Differences in minimum reporting levels (MRL) among sites tend to confound spatial comparisons of instream loads, thus load estimates for sites and constituents with high MRL's (compared with other sites) are reported with a qualifier.

Instream load of nitrogen, phosphorus, and sediment was calculated as the product of daily streamflow and estimated daily concentration using the Cohn's Estimator model (Cohn and others, 1989; Cohn and others, 1992; Gilroy and others, 1990). This model includes a seven-parameter log-linear regression analysis of constituent concentrations against measured environmental variables:


ln[C] = β0 + β1(ln[Q/Q′] + β2(ln[Q/Q′]) 2               (1)
             + β3[T-T′] + β4[T-T′]2 + β5sine[2πT]
             + β6cosine[2πT] + e


ln[ ] is natural logarithm function;

C is estimated daily concentration, in milligrams per liter;

Q is daily streamflow, in cubic feet per second;

T is time, in decimal years;

π is 3.14169;

β0 - β6 are calibration coefficients of the regression model;

e is model error;

Q′ is centering variable defined so that β1 and β2 are statistically independent; and

T′ is centering variable defined so that β3 and β4 are statistically independent.

The regression analysis assumes that model errors (e) are independent and normally distributed, with zero mean and variance. The Minimum Variance Unbiased Estimator (Bradu and Mundlak, 1970) is included in the model to correct for the retransformation bias associated with log-linear regression models; the model also employs the Adjusted Maximum Likelihood Estimator (Cohn, 1988), which statistically addresses censored data and multiple reporting limits. Additional information about the data requirements of Cohn's Estimator model is given in Appendix B.

Variations in concentrations of nitrogen, phosphorus, and sediment with season, streamflow, and time were displayed for sites 1 to 20 with scatterplots, and quantified with the multivariate log-linear regression component of Cohn's Estimator model. A significance level of 0.05 was used as the criterion for statistical significance to interpret regression analyses and correlations.

Next Title Page Table of Contents List of Illustrations List of Tables Conversion Factors and Vertical DatumGlossary

Last modified: Thu Mar 30 13:02:45 CST 2000

Tennessee home page