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


The following three sections describe spatial and temporal variations in instream transport of nutrients and sediment. The first section examines the relation of nutrient transport to the environmental variables: season, hydrologic condition at the time of sample collection, and physical location along the stream channel. The second section describes estimated nutrient and sediment instream loads. The third section compares nutrient yields to inputs from nutrient sources and other factors influencing transport.

Relation of Concentrations to Season, Streamflow, and Reach Location

Interpretation of the relation of nutrient and sediment concentrations to season and streamflow is confounded because season and streamflow variables generally do not operate independently of one another. For example, an observed seasonal pattern in nutrient concentration may be caused by correlation between concentration and streamflow, rather than directly by seasonal change in water-quality processes. A stratified statistical analysis, such as multiple linear regression, provides a way to identify the confounding effects of different variables. In this approach, the influence of each variable is interpreted independently by examining the statistical significance of each regression coefficient (Cohn and others, 1992). The results of seven-parameter log-linear regressions of nutrient and sediment concentration data are organized by constituent for each of the 20 water-quality monitoring sites (table 8). The period of record used in the regression analysis varied among sites and is indicated in table 8 along with the statistical significance and sign (positive or negative) of the regression coefficients.

Statistical significance of β1 indicates that streamflow, independent of other influences, is a good predictor of concentration. Statistical significance of both β5 and β6 indicates a significant seasonal pattern in concentration data (fitted to a simple sine wave function), independent of other influences. Temporal trends in nutrient and sediment concentration are interpreted based on the statistical significance of β3; these are discussed in the section "Trends of Nitrogen, Phosphorus, and Sediment."

Relation of Concentrations to Season

Seasonal variation in nutrient concentrations is commonly attributed to assimilation by algae and aquatic macrophytes. During the summer months, concentrations of bioavailable forms of nitrogen and phosphorus may decrease as a result of aquatic plant growth in water bodies with long hydraulic-residence time. Other possible causes of seasonal variation include seasonal fertilizer application, temperature-driven nitrification and volatilization, and seasonal flow variation. The extent to which seasonal processes influence nutrient transport in the LTEN River Basin is indicated by the significance of the seasonal regression coefficients (β5 and β6) for the bioavailable forms of nutrients (total nitrite plus nitrate and dissolved orthophosphorus, table 8). These results will indicate seasonal effects independent of flow variation, which is accounted for by the regression coefficient β1.

Statistically significant seasonal variation of total nitrite plus nitrate concentrations was observed at 9 of the 18 sites tested. Seasonal variation was closely correlated with site type: 7 of the 9 sites showing significant seasonal variation were in reservoirs or were influenced by reservoir tailwater, suggesting strong seasonal influence by aquatic plants in water bodies with long hydraulic-residence times. The seasonal pattern of total nitrite plus nitrate for most of the sites influenced by reservoirs was annual maximum concentrations in winter and annual minimum concentrations in summer.

Significant seasonal variation was observed at less than half of the sites for the other nutrient constituents, except for dissolved orthophosphorus. Of the eight sites with dissolved-orthophosphorus data, four showed statistically significant seasonal variation of this constituent. The detected pattern at the reservoir sites (sites 16 and 15, Tennessee River at Pickwick Landing Dam and at river mile 23, respectively) was slightly offset (annual maximum concentrations in fall rather than winter, and annual minimum concentrations in the spring) from the pattern for total nitrite plus nitrate, but may be caused by the same phenomenon of aquatic-plant assimilation.

Scatterplots of concentration and time of year of sampling (fig. 7) illustrate seasonal patterns in nitrate and total phosphorus at a few selected sites. Comparison of these patterns with the regression result (statistical significance of β5 and β6, also shown on fig. 7) illustrates how the regression analysis isolates the influences of different environmental variables. For example, concentrations of nitrite plus nitrate appear to indicate a seasonal pattern at site 2 (Buffalo River at Flat Woods); however, the regression analysis did not detect a statistically significant pattern. The apparent pattern may be caused instead by a seasonal bias in sampled streamflow conditions (five of the six highest streamflows sampled occurred during the month of January) combined with a strong correlation between concentration and streamflow (β1 is significant and positive, table 8).

A separate possible cause of seasonal variations in nutrient concentrations, the agricultural planting cycle, may cause episodic increases in concentrations of nutrients derived from fertilizer application and of suspended sediment following periods of soil preparation and fertilizer application and when storm runoff is frequent (during April through June). The relation between the planting cycle and instream nutrient transport is not examined in this report, however, because the data from most of the sites are from quarterly monitoring programs rather than from programs designed to detect episodic, storm-related increases in concentrations.

Relation of Concentrations to Streamflow

The variation of nutrient concentrations with streamflow generally reflects the dominant sources in the watershed (point versus nonpoint sources). Transport of nonpoint-source-derived constituents mainly occurs during periods of high surface runoff and high base flow; therefore, higher concentrations are expected during high streamflow in watersheds dominated by nonpoint sources. Point-source discharges are generally independent of runoff; consequently, instream concentrations of constituents from these sources are expected to decrease during high streamflow.

Scatterplots of concentration and streamflow are shown (fig. 8), along with regression results for β1 (coefficient on streamflow), for total nitrite plus nitrate and total phosphorus at a few selected monitoring sites. Scatterplots and model regression results are included for the full set of load-computation sites for total nitrite plus nitrate, total ammonia, total phosphorus, and dissolved orthophosphorus (Appendix D). Interpretation of the relation of concentration to streamflow at flow-regulated sites is of less interest than at riverine sites in this examination of dominant sources in the watershed because at flow-regulated sites the observed streamflow-concentration relation reflects controlled impoundment releases rather than hydrologic processes such as runoff. Therefore, regression results are not shown on the concentration-streamflow scatterplots for flow-regulated sites in Appendix D. The flow corresponding to maximum turbine capacity is indicated on the scatterplots for these sites as a cutoff below which a relation based on hydrologic conditions in the contributing watershed is unlikely; above this cutoff, at least some component of streamflow reflects hydrologic conditions upstream from the impoundment (E.A. Thornton and D.L. Meinert, Tennessee Valley Authority, written commun., 1998).

Comparison between the relative influence of nonpoint sources and point sources on instream water quality in the selected tributary basins can be examined by comparing the statistical significance of the streamflow regression coefficient, β1 (table 8 and fig. 8) with a ranking of these basins according to estimates of inputs from these sources. For total nitrite plus nitrate, β1 was significant and positive at 5 sites (sites 2, 6, 10, 11, and 20) of the 11 riverine sites tested, and was positive in all of these cases, suggesting that nonpoint sources were the dominant source of input for total nitrite plus nitrate during the periods spanned by the calibration data sets. These results match the ranking of basins by nitrogen input reasonably well; the largest nitrogen inputs from nonpoint sources (the sum of atmospheric deposition and fertilizer application and livestock waste) relative to point sources (wastewater discharge) were for sites 11, 20, 12, 13, 10, and 2. β1 was significant and positive for total phosphorus at three of the six riverine sites (sites 2, 5, and 6), suggesting that nonpoint sources are the dominant source of phosphorus input in these watersheds. The nonpoint source might be a natural source, as for sites 5 and 6; suspended sediment at these sites is naturally phosphate rich as a result of phosphatic limestone in the watershed. The regression results for total phosphorus might, therefore, be explained by movement of sediment during high flows.

Dominance of point sources, such as wastewater discharge, is indicated by a negative concentration-streamflow relation, that is, by a significant and negative value for β1. β1 was significant and negative for total phosphorus concentrations at site 1 (Clarks River at Almo), suggesting dominance of point-source discharges or some other runoff-independent source at this site. Clarks River at Almo was ranked second out of 11 based on the ratio of point-source input (wastewater discharge) to nonpoint-source inputs of phosphorus. Data for dissolved orthophosphorus, which might have provided additional information about the influence of wastewater sources, were not available for this site.

The relation between nutrient concentration and streamflow has important implications for resource management apart from the issue of the relative contributions of nonpoint and point sources to nutrient load. The occurrence of high concentrations during low streamflows and the sources that cause high concentrations during low streamflows (for example, wastewater discharges) may not be significant for characterizing annual nutrient transport. However, these occurrences and sources of high concentration are of particular concern for evaluating impairment of a water body caused by constituent concentrations at harmful levels for short periods, for example during critical low streamflow periods. Although the monitoring networks from which these data were collected were not designed to detect impairment during critical periods, some insight as to where these conditions might occur can be gained by comparing concentrations of nitrogen constituents during low streamflows.

The median concentration of total nitrogen from the set of samples collected when streamflow was below the 80th percentile (the flow exceeded 80 percent of the time at that site) was plotted against estimated input of total nitrogen from wastewater in the watershed (fig. 9a). The correlation between these variables is significant (r = 0.71, p = 0.01) and suggests that wastewater discharges are significant contributors to the amount of nitrogen in transport during periods of low streamflow. That wastewater-discharge input is large (almost equal to or exceeding) compared to the median of observed daily loads during low streamflow (calculated as the product of observed concentration and daily streamflow) at most sites supports this conclusion (fig. 9b).

Downstream Variations in Concentrations

Box plots of nutrient concentrations at different water-quality sites along a river reach illustrate how nutrient concentrations vary with distance. Downstream variations in nutrient concentrations are illustrated as truncated box plots (Helsel and Hirsch, 1992) for the main stem Tennessee River (fig. 10) and for the Duck River, the largest tributary within the LTEN River Basin (fig. 11). Data from eight of the load-computation sites on the main stem Tennessee and the Duck Rivers were included in these graphical displays, in addition to data from 29 additional long-term monitoring sites on these rivers (as described in the section "Approach and Methods").

Median total nitrogen concentrations were generally less than 0.7 mg/L throughout the main stem of the Tennessee River (fig. 10B) compared to median values greater than 1.0 mg/L at many sites on the Duck River (fig. 11B). Variation of nitrite plus nitrate concentrations throughout the main stem of the Tennessee River was closely correlated with reservoir forebay areas (immediately upstream from the dams, fig. 10C). The range in concentration at a site reflects the seasonal variation of nitrite plus nitrate in reservoirs; concentrations decreased to below the minimum reporting level (MRL) during the summer months in the forebay areas. In contrast, concentrations of ammonia and total phosphorus did not vary much longitudinally (figs. 10D and 10E); instead they fluctuated for short distances (for example, near Tennessee River miles 100 and 300), possibly in response to point- or nonpoint-source inputs.

Concentrations of all nutrient constituents varied widely with distance along the Duck River, and appear to be strongly influenced by Normandy Reservoir (from river mile 248.6 to 265.4, fig. 11). Median nitrite plus nitrate concentrations were less than 0.3 mg/L within the reservoir, compared with median values of 0.6 mg/L or greater upstream and downstream from the reservoir (fig. 11C); total nitrogen concentrations were somewhat lower in the reservoir with median values less than 0.7 mg/L, compared with median values of 1.0 mg/L or greater upstream and downstream of the reservoir. Median total phosphorus concentrations were as low as 0.01 mg/L within the reservoir, compared with a range in median values of 0.08 to 0.3 mg/L at sites upstream and downstream from the reservoir (fig. 11E). This pattern probably reflects depletion of these nutrients caused by assimilation by aquatic plants in the reservoir during the summer months and, in the case of total phosphorus, by settling associated with sediment. A study of the nutrient balance in Normandy Reservoir by Broach and others (1995, p. 70) found that the reservoir functions as a sink for nitrite plus nitrate and organic nitrogen, but that ammonia outflow from the reservoir exceeds inflow by about threefold. The illustration of variation in total nitrogen (fig. 11B), which shows tailwater concentrations about equal to upstream concentrations, suggests that the reservoir may not be a sink for total nitrogen. The high concentrations of ammonia within Normandy Reservoir and in the tailwater (median value as high as 0.1 mg/L, fig. 11D) might be due to nitrate reduction and ammonification in the oxygen-deficient layer of the reservoir during summer months (Broach and others, 1995).

Background levels for nitrate, ammonia, and total phosphorus for 20 NAWQA study units are included to compare water quality in the Duck River with other rivers in the Nation (fig. 11). Average concentration for "undeveloped" basins was summarized as 0.7 mg/L for nitrate and 0.1 mg/L for ammonia and total phosphorus (Mueller and others, 1995). These background levels were not compared with concentrations on the main stem Tennessee River (fig. 10) because the levels do not reflect the effects of instream processing of nutrients in large rivers.

The median values of concentrations of total nitrite plus nitrate and total phosphorus observed for many of the riverine sites on the Duck River exceeded NAWQA background levels of 0.7 mg/L for nitrate and 0.1 mg/L for total phosphorus (figs. 11C and 11E). The high phosphorus concentrations (greater than 0.2 mg/L) relative to the NAWQA background level in the lower reach of the river (from near Duck River Mile 120 to the mouth) were probably caused by contribution of phosphate-rich sediment, from soils formed on the outcrop of phosphatic limestones in the brown-phosphate districts (figs. 11A and 11E) of middle Tennessee (Smith and Whitlatch, 1940). Ammonia data were sparse in the riverine sections downstream and upstream of the Normandy Reservoir; median values at the five riverine sites with data were below the NAWQA background level of 0.1 mg/L (fig. 11D).

Instream Loads

This section presents estimates of nutrient and sediment instream loads and yields for 18 sites in the LTEN River Basin (table 9). These estimates are used in interpretations of spatial patterns of instream loads and comparisons of instream loads with inputs. Instream loads were calculated for total nitrogen (16 sites), total ammonia (13 sites), total ammonia plus organic nitrogen (16 sites), total nitrite plus nitrate nitrogen (16 sites), total phosphorus (12 sites), dissolved orthophosphorus (7 sites), and suspended sediment (5 sites). Several estimates of annual load were produced for each site-constituent combination. The first set of estimates--mean, minimum, and maximum annual load for the available period of record for load computation (concurrent water-quality and streamflow record)--is provided to indicate long-term transport conditions at the site. Load estimates can vary significantly in years with low or high annual streamflows. At each of the 18 sites where loads were estimated, streamflow durations during the period of load estimation were compared with long-term streamflow duration (Appendix E). At most sites, streamflows during these two periods were similar.

Transport of nutrients into the LTEN River Basin near Chattanooga, Tennessee, is characterized by the instream load estimate at site 18 (Tennessee River below Raccoon Mountain). Transport out of the basin is characterized by the estimate at site 14 (Tennessee River at Highway 60 near Paducah), except that the watershed for this site does not include the Clarks River drainage. Estimates of mean annual instream load of total nitrogen at the inlet (site 18) and outlet (site 14) were 29,000 tons/yr (for the period 1981-94) and 60,000 tons/yr (for the period 1980-84), respectively (table 9), representing a gain of 31,000 tons/yr, on average, across the area (18,930 mi2) between these two sites. The sum of the mean annual instream load from gaged tributaries to the main stem within the study unit was 14,000 tons/yr (period variable); however, this number cannot be compared with the gain between the inlet and outlet sites because the gaged area represents only 30 percent of the total area and the period of record at many tributary sites did not correspond with the period of record at the inlet or outlet sites.

Estimates of mean annual instream load of total phosphorus at the inlet (site 17) and outlet (site 14) were 1,300 tons/yr (1980-86) and 5,000 tons/yr (1980-84), respectively, representing a gain of 3,700 tons/yr, on average, across the study unit (table 9). For this comparison, site 17 (Tennessee River at South Pittsburg) was substituted as the inlet site because total phosphorus load was not estimated at site 18. The sum of the gaged tributary load, representing only 28 percent of the area contributing to the main stem between sites 17 and 14, was 4,300 tons/yr (period variable). Although this number cannot be closely compared with the gain throughout the study unit, for the same reasons given for total nitrogen, a general comparison suggests that the main stem of the Tennessee River and the tributary embayments along the main stem function as a sink for total phosphorus, removing a substantial amount from the water column through assimilation or deposition.

Load and yield estimates for a single water year also are presented for each site (table 9) to provide a common period for better spatial comparisons among sites. Data from 1992 were used for comparison where possible because more data were available on instream loads and sources during 1992 than for any other year. Hydrologic conditions during 1992 were close to average at all sites (except for Clarks River at Almo, which had relatively low flow in 1992). For those sites without load estimates for 1992, the estimate from a hydrologically similar year was used.

The range between the upper and lower limits of the 95-percent confidence intervals (table 9) indicates precision of the model estimates and is an important consideration for interpreting these data. The load estimates are useful for evaluating broad spatial patterns of instream load, and comparisons of instream load to inputs, but may not be sufficiently accurate for local-scale evaluations of water quality. A discussion of limitations of the data and error in the model estimates is contained in Appendix B.

An additional consideration in spatial interpretation is the extent to which the set of monitoring sites represent conditions in the LTEN River Basin. The gaged tributary area represents a relatively small part of the study unit (less than 30 percent for most constituents); therefore, spatial extrapolations must be made with caution. In addition, the tributary sites in Alabama were monitored as part of special studies of known water-quality problems, compared to tributary sites in other states that were part of ambient monitoring networks, which may bias conclusions about basin-wide water quality.

To facilitate spatial comparisons of instream loads at sites draining watersheds of greatly differing size and streamflow characteristics, the load estimates at each site were normalized in two ways: by dividing each estimate by the watershed area (producing an estimate of yield in tons per square mile), and by dividing each estimate by the mean streamflow at the site for the load-computation period (producing the equivalent of flow-weighted mean of the model-estimated daily concentrations, in milligrams per liter). The model-estimated daily concentrations are used rather than the set of observed concentrations to derive the estimate of flow-weighted mean concentration because the former account for flux during unsampled periods and are considered to better represent average concentration during a specified period.

Estimates of flow-weighted mean concentration are useful for evaluating average water-quality conditions at the site and for comparisons (which follow) with national data sets and guidelines; estimates of yield are useful for comparisons with inputs in a mass balance analysis. Both yields and flow-weighted mean concentrations are reported in table 9 and displayed graphically in figures 12 and 13, and estimates of yield of total nitrogen and total phosphorus are displayed on maps in figures 14 and 15. Because runoff characteristics for site 1, Clarks River at Almo (median runoff for period of load computation is 3.0 in., table 1), differed from the other tributary sites (median runoff ranges from 8 to 13 in.), the ranking of this site for constituent transport estimated by flow-weighted mean concentration differs substantially from the ranking based on yield estimates (compare in figure 12).

Estimates of annual flow-weighted mean concentration of total nitrogen range from 0.53 to 2.8 mg/L as nitrogen (table 9), representing a fivefold difference among the watersheds. The smallest estimates were at site 2 (Buffalo River near Flat Woods) and site 4 (Duck River below Normandy Dam), which represent a minimally developed watershed (with 29 percent combined urban and agricultural land use, fig. 5) and tailwater from a tributary reservoir, respectively. The largest estimate, at site 1 (Clarks River at Almo), is from the watershed with the largest areal percentages of urban and agricultural land use (corresponding to 80 percent combined urban and agricultural land use, fig. 5), and with the largest amount of wastewater discharge (table 3), suggesting that human activity increases instream transport of total nitrogen by as much as fivefold. The difference in flow-weighted mean concentration at sites 1 and 2 might not be caused by human activity alone, however, because the natural basin characteristics also differ for these two sites (fig. 2).

Estimates of annual flow-weighted mean concentration of total phosphorus range from 0.02 to 0.73 mg/L as phosphorus (table 9), representing a fortyfold difference among the watersheds. The smallest estimate was for site 2 (Buffalo River near Flat Woods), which is a minimally-developed watershed. The largest three estimates, 0.73, 0.53, and 0.52 mg/L (from sites 6 and 5 on the Duck River and site 10 on the Elk River, respectively), probably represent a natural source: the phosphatic limestone formations of the brown-phosphate district outcrop in the watersheds for these three sites (figs. 11A and 11E). The outcrop pattern of these phosphatic limestones is an important factor to consider as regional boundaries are established for attainable region-specific water-quality criteria for total phosphorus (U.S. Environmental Protection Agency, 1998). The estimate of 0.35 mg/L for site 1 (Clarks River at Almo) is a twentyfold difference from site 2 and better indicates to what extent human activity has increased instream transport of total phosphorus in the LTEN River Basin.

The range of annual flow-weighted mean concentration for each constituent can be placed in national context by comparing with the statistical distribution of estimated values of annual flow-weighted mean concentration from about 200 basins monitored during 1993-94 as part of the NAWQA program (D. Mueller, U.S. Geological Survey, written commun., 1998). The distributions for total ammonia, dissolved orthophosphorus, and total nitrite plus nitrate at the LTEN River Basin sites match the national distributions reasonably well (fig. 16); median values for the LTEN River Basin sites fall within the 45 to 65 percentile range of the national distribution. Distributions for total nitrogen and total phosphorus at LTEN River Basin sites, however, depart more from the national distributions (fig. 16); median values for the LTEN River Basin sites are lower (corresponding to the 35 percentile in both cases). The maximum value of total phosphorus flow-weighted mean concentration (0.73 mg/L for site 6, Duck River above Hurricane Mills, table 9) falls at the high end of the national distribution. The nine sites in the national data base with higher values than for site 6 were predominantly from western States or were from small (drainage area less than 80 mi2), intensively cultivated, watersheds (D. Mueller, USGS, written commun., 1998). The flow-weighted mean concentration of total phosphorus for site 6 is an extreme value within the context of both the national and the regional (LTEN River Basin) distributions, and further emphasizes the significance of the phosphatic limestone formations in the basin for this site. Flow-weighted mean concentrations for site 5 (Duck River at Williamsport, 0.53 mg/L) and site 10 (Elk River near Prospect, 0.52 mg/L) also are extreme.

The spatial distribution of tributary annual flow-weighted mean concentration indicates which areas of the basin contribute more nutrients, on a discharge-weighted basis, to downstream receiving waters and, therefore, which water bodies may be at greatest risk for eutrophication and consequent ecological disruption. Although the trophic status of a water body relates not only to nutrient influx but also to assimilative capacity (which includes factors such as light attenuation and residence time), nutrient influx is a major factor. Watersheds contributing the largest amounts of total nitrogen on a discharge-weighted basis are sites 1, 12, and 10 (Clarks River at Almo, Town Creek near Geraldine, and Elk River near Prospect, respectively, table 9). Watersheds contributing the largest amounts of total phosphorus, on a discharge-weighted basis, are sites 6, 5, and 10 (Duck River above Hurricane Mills and at Williamsport and Elk River near Prospect, respectively). These sites also have small nitrogen-to-phosphorus (N:P) ratios (less than 4:1, compared to a range of 8:1 to 29:1 for the other tributary sites, table 4; values greater than 7:1 indicate that phosphorus is the limiting nutrient). Furthermore, inflow from these tributaries appears to cause a decreasing trend of N:P ratio along the main stem of the Tennessee River (27:1 at site 17, Tennessee River at South Pittsburg, compared with 12:1 at site 15 near the mouth), towards a less phosphorus-limited system.

Although water-quality criteria limiting nutrient influx have not been established, the National Technical Advisory Committee (1968) recommendation of 0.05 mg/L phosphorus for waters entering impoundments can be compared with the annual estimate of phosphorus flow-weighted mean concentrations at each monitoring site (table 9). The annual flow-weighted mean concentrations exceeded the recommended value at five of the eight tributary sites and three of the four main stem Tennessee River sites. Furthermore, these phosphorus concentrations exceeded the value of 0.1 mg/L, recommended by Mackenthun (1969) to prevent algal blooms in streams, by threefold or more at four of the eight tributary sites. The phosphorus influx to consider for evaluating ecological risk to water bodies may not be the annual mean influx, however, but rather the influx during spring and summer, which is the season of algae and macrophyte growth. Mean rates of influx for this period are different from the mean annual rate, due to seasonal variation in loading rate (illustrated in fig. 17).

Estimates of flow-weighted mean concentration of nutrients during the period March-July are reported in table 9 along with the annual estimates to illustrate differences in annual influx and seasonal influx. Estimates of seasonal flow-weighted mean concentration were about 50 percent lower than annual estimates; the range for the tributary sites was from 0.28 mg/L (site 2, Buffalo River near Flat Woods) to 1.2 mg/L (site 1, Clarks River at Almo) of total nitrogen, and from less than 0.01 mg/L (site 9, Elk River below Tims Ford Dam) to 0.18 mg/L (site 5, Duck River at Williamsport) of total phosphorus. The seasonal estimate of total phosphorus flow-weighted mean concentration was above the value of 0.05 mg/L, recommended for water entering impoundments, at four of the eight tributary sites, but was much closer to this value than the corresponding annual estimates.

Whereas water-quality impairment from eutrophication is related to excessive nutrient influx over a period of weeks or months, other types of impairment (for example, acute toxicity of ammonia) result from short-term fluctuations of concentrations. The estimates of annual or seasonal yield and flow-weighted mean concentration are not useful for addressing this latter type of water-quality impairment. Information about the spatial distribution of ammonia concentration during prolonged periods of low streamflow, similar to the information presented in figure 9A, would be more useful in assessing where impairment caused by ammonia toxicity is likely to occur.

Sediment yield estimates ranged from 65 (tons/mi2)/yr (site 2, Buffalo River near Flat Woods) to 260 (tons/mi2)/yr (site 8, Shoal Creek at Iron City) for the three tributary monitoring basins for which data were available, and from 17 to 26 (tons/mi2)/yr for the two main stem sites (Tennessee River at South Pittsburg and Tennessee River at Pickwick Landing Dam, respectively, table 9). Lower sediment yields for the main stem sites compared with the tributary sites is probably due to sediment deposition in the main stem of the Tennessee River and tributary embayments along the main stem. The sediment yield estimates for the tributary sites are lower than the estimate from Trimble and Carey (1984) of 800 (tons/mi2)/yr for basins in central and eastern Tennessee.

Comparison of Inputs from Nutrient Sources with Nutrient Yields

The estimates of nitrogen and phosphorus loading from major sources varied widely among the 11 tributary basins for which estimates were prepared, and are expected to represent the variability in these sources across the LTEN River Basin (figs. 14 and 15). Of the sources of land-phase nitrogen inputs (atmospheric deposition, fertilizer application, and livestock waste), livestock waste contributed the largest input in about two thirds (7 out of 11) of the load-computation basins (table 3 and fig. 14), and fertilizer application contributed the largest input in the remaining four basins (sites 1, 2, 7, and 9). Estimates of nitrogen input from fertilizer application were the most variable spatially among the land-phase nitrogen inputs, ranging from 1.5 to 23 (tons/mi2)/yr. Atmospheric deposition estimates varied the least from basin to basin, ranging from 1.6 to 2.0 (tons/mi2)/yr. The balance of land-phase nitrogen inputs on agricultural lands (fertilizer application plus nitrogen fixation plus livestock waste minus harvest) ranged from 2.4 to 15 (tons/mi2)/yr of nitrogen (table 3). Wastewater discharge contributed between 0 and 0.61 (tons/mi2)/yr of nitrogen (table 3); wastewater input of nitrogen is therefore equivalent to a relatively small part (less than 27 percent) of the annual instream nitrogen load, whereas the contribution of wastewater discharge during low flow is much more significant (fig. 9B).

The estimates of inputs can be compared and correlated with export yields; significant correlations between estimates of inputs and exports might be useful as predictive tools for instream water quality where monitoring data are not available. Export of nitrogen correlated moderately well with the balance of land-phase inputs to agricultural lands for the tributary sites (fig. 18). For example, nitrogen export was highest [3.5 (tons/mi2)/yr] for site 12 (Town Creek near Geraldine), for which the balance of agricultural land-phase input was also the highest [15 (tons/mi2)/yr]. Nitrogen export was low [1.0 (tons/mi2)/yr] for site 2 (Buffalo River at Flat Woods), for which the balance of agricultural land-phase input was correspondingly low [3.2 (tons/mi2)/yr, the second lowest]. The matrix of Pearson correlation for nutrient input, percentage land use and land cover, and export is shown in table 10. Among all the estimated nitrogen inputs and land-use types, there were significant correlations between the balance of agricultural land-phase input and export (r = 0.65, p = 0.03, shown in fig. 18) and between percentage pasture land and export (r = 0.67, p = 0.03). Nonparametric rank correlation analysis showed similar results, addressing the concern that the large values of export, livestock-waste input, and percentage pasture land at site 12 (Town Creek near Geraldine) might skew the correlation results. Correlation of wastewater discharge with nitrogen export was poor (r = 0.06), and contrasts with the significant correlation (r = 0.71, p = 0.01, fig. 9A) between wastewater discharge and nitrogen concentration during low streamflow. The poor correlation between wastewater discharge and annual export was expected, however, as wastewater discharge is a small fraction compared with annual instream nitrogen load.

The relation between total nitrogen export and the balance of agricultural land-phase input is shown in figure 18, with a line fitted to the data points using a simple linear regression. The y-intercept of the fitted line, 0.95 (ton/mi2)/yr, could be interpreted as the expected nitrogen export from a watershed without agricultural inputs. The results from sites 4 and 9 (Duck River below Normandy Dam and Elk River below Tims Ford Dam, respectively) represent the largest residuals from the regression line, and may reflect the effects of the reservoirs on instream delivery processes. Large residuals for sites 7 and 10 (Shoal Creek at Highway 43 and Elk River near Prospect, respectively) could be due to a difference in the land-water delivery process (such as different soil-drainage characteristics) in the watersheds for these sites, as compared with the other sites in the data set.

Among the sources of land-phase phosphorus inputs (fertilizer application and livestock waste), livestock waste contributed most of the total input in 8 out of the 11 tributary basins (fig. 15 and table 4), and fertilizer application contributed most in the basins for sites 1, 2, and 7 (Clarks River at Almo, Buffalo River near Flat Woods, and Shoal Creek at Highway 43, respectively). The balance of agricultural land-phase inputs of phosphorus ranged from 0.87 (tons/mi2)/yr (site 13, Sequatchie River at Valley Road) to 4.7 (tons/mi2)/yr (site 12, Town Creek near Geraldine). Wastewater discharge contributed from 0 to 0.14 (tons/mi2)/yr, equal to as much as 1.2 times (site 9, Elk River below Tims Ford Dam) the corresponding phosphorus export from the basin.

In contrast with nitrogen, phosphorus export did not correlate well with any estimated inputs or land-use types (table 10). Phosphorus export was highest [1.1 and 0.93 (tons/mi2)/yr] for sites 6 and 5 (on the Duck River) and at site 10 [0.89 (tons/mi2)/yr, Elk River near Prospect]; however, estimates of inputs and percentage of each land-use type at these sites were not in the high end of respective ranges (table 4). The influence of a known natural source, outcrop of phosphatic limestone formations of the brown phosphate district, in the lower Duck and lower Elk River Basins (figs. 11A and 11E) was examined by removing sites 5, 6, and 10 from the correlation data set. The significant correlations between phosphorus export and wastewater discharge (r = 0.97) and between phosphorus export and fertilizer application (r = 0.94) for this trimmed data set (table 10) should be interpreted with caution because of the small number of sites (n = 5). Two related conclusions are suggested: (1) inputs from wastewater discharge and fertilizer application are strongly linked with instream transport of phosphorus in watersheds where the natural phosphorus source is not present and (2) the natural source, where it is present, might be the largest contributor to instream transport of phosphorus. A regression equation was not developed between total phosphorus export and any of these sources due to the small number of sites in the trimmed data set.

That the correlation between phosphorus export and percentage of pasture land (r = 0.85) is very different from the correlation between export and livestock waste (r = -0.06) is difficult to explain because the estimate for livestock waste is partly derived from distribution of pasture land; however, the estimate for livestock waste also accounts for distribution of animal populations, and, because of feedlot operations, this distribution may differ substantially from the distribution of pasture land. The apparent contradiction in correlation results might indicate that pasture land influences instream loads of phosphorus through processes apart from runoff from land areas with livestock waste, or that some controlling factor is coincidentally correlated with percentage pasture land.

The spatial pattern of estimated inputs and exports may be influenced by several factors other than sources and transport processes. These other factors, some of which are listed below, may confound meaningful interpretation of the correlation results:

  1. Inaccuracy in estimates of export caused by sparseness of monitoring data and lack of flow-stratified sampling for load estimation (discussed in Appendix B).
  2. Low variability in export, which reduces the ability to detect spatial patterns. For example, the relatively low variability in nitrogen export (only one half of an order of magnitude) may reflect conditions throughout the basin, or may result from network bias (lack of representation of the full range of conditions in the basin or lack of sufficiently homogeneous basins). Phosphorus export, ranging through almost two orders of magnitude, was more variable than was nitrogen export.
  3. Other sources of nutrients not quantified in this analysis (such as urban runoff, failing septic systems, or natural sources) might be responsible for much of the observed variability in export.
  4. The pairwise correlation analysis (table 10) allows examination of correlations of export with individual input variables, but does not test the correlation with co-occurrence of variables. The small size of the data set prevented testing combinations of variables using stepwise multiple regression analysis.
  5. Regional variability in soil and other geologic factors causes variability in land-water delivery processes and therefore affects the relation between sources (particularly land-phase loads) and export. The influence of regional differences in natural environmental setting on interactions between sources and exports could be examined with a larger set of monitoring sites, provided that the monitoring network was sufficiently stratified by environmental setting (that is, included several watersheds representing each type of environmental setting).

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

Last modified: Thu Mar 30 12:52:29 CST 2000

Tennessee home page