Loading of phosphorus and nitrogen to Lake Waco, Texas

Abstract

Cyanobacterial blooms in Lake Waco are linked to nutrient loading from the watershed. Direct measurement of nutrient levels over a period of 8 years facilitates calculation of loading of water, phosphorus (P) and nitrogen (N) to Lake Waco. The North Bosque River (NBR) provides 61% of the flow to Lake Waco, while the Middle Bosque River (MBR) contributes 15.5%. The South Bosque River (SBR) and Hog Creek (HC) each contribute about 7.5% of the total water load to the lake, while other sources contribute less. The lake receives average annual total phosphorus (TP) loading of approximately 147,000 kg and an annual total nitrogen (TN) load of about 1,450,000 kg, with considerable variability expected among years. The NBR contributes 67% of the TP and 37% of the TN. The MBR provides slightly more than 17% of the TP and 41% of the TN. Inputs from SBR are estimated at about 8% of the TP and 10% of the TN. All combined, HC, direct drainage to the lake, atmospheric deposition, ground water inseepage, human inputs associated with recreation, waterfowl inputs and internal loading contribute <8% of the TP and about 12% of the TN. The NBR is clearly the dominant source of TP to the lake, while MBR and NBR contribute roughly equal amounts of TN and much more than any other source. Inputs from wastewater treatment facilities (WWTF) represent about 4% of the TP input and <2% of the TN load to Lake Waco. Despite greater availability of the forms of P in WWTF inputs, this is a minor source for the lake. Inputs from dairy operations, including areas of active animal use and waste application fields, are estimated to contribute 34–42% of the NBR TP load and 23–28% of the overall TP load to Lake Waco. Dairy operations are estimated to contribute about half of the biologically available P load to Lake Waco and generate low N:P ratios in the lake. Despite high loading and resultant blooms, Lake Waco exhibits lower concentrations of P than would be predicted by multiple models, and the actual concentrations stimulate less algal growth than would be expected. In the absence of major inputs related to human activities, the predicted P load to Lake Waco would be less than half the current load, and very little of it would be biologically available. The initial regulatory target of a 50% reduction in P loading to NBR set by the Texas Commission on Environmental Quality is appropriate. If dairy inputs are successfully controlled, measurable improvement can be expected. Supplemental materials are available for this article. Go to the publisher's online edition of Lake and Reservoir Management to view the free supplemental file.

Key words:

• dairy
• flow
• Lake Waco loading
• nitrogen
• phosphorus

Waco Reservoir, known as Lake Waco, is located in Waco, Texas (Fig. 1), and is both a major recreational resource and drinking water supply. For the purposes of this assessment, the surface area of Lake Waco (the reservoir) is approximately 29 km² with a volume of approximately 179 million m³ (McFarland et al. 2001). The maximum water depth is 24 m and the mean depth is 6.2 m; more than 75% of the reservoir bottom occurs at a depth of 9 m or less (Abraham et al. 1999). A permanent pool rise of 2 m in October of 2003 increased these values, but the data applied here are from 2001 and earlier.

Figure 1 Location of Lake Waco and in-lake sampling stations, Waco, TX.

The reservoir watershed (4267 km²) is drained to the north arm of Lake Waco by the North Bosque River (NBR), and to the south arm by the Middle (MBR) and South (SBR) Bosque Rivers and Hog Creek (HC; see Conry 2010, this issue). Several smaller tributaries and a number of storm water systems drain directly into the reservoir. The NBR drains about 75% of the watershed, with MBR draining just over 12%, SBR and HC draining about 5% each; direct drainage accounts for slightly more than 2%, and the reservoir itself covers <1% of the total system area. The watershed is 147 times the area of the reservoir for the timeframe addressed here, and 117 times the reservoir area after pool rise, both large watershed to lake area ratios. A watershed of such size, relative to the lake into which it drains, has the potential to dominate water quality in that lake through runoff processes.

Each of the tributaries and drainage systems discharges nutrients, sediment, possible pathogens and other substances both natural and human-derived. This analysis focuses on measurement of loading to the reservoir over a decadal period leading up to the rise in pool elevation, providing estimates of inputs from areas and defined sources, based on a large database formed from the efforts of multiple organizations. This analysis provides both an evaluation of loading and a comparison for model outputs (White et al. 2010a, 2010b) and has been conducted independently of the modeling exercise.

This effort seeks to assess the magnitude and relative size of water and nutrient contributions to Lake Waco from actual data collected over multiple years, allowing assessment of loading in relation to acceptable levels for maintaining desired conditions in the reservoir. This information can be used to establish priorities and set goals for specific load reductions to maximize in-lake water quality and minimize effects on the drinking water supply and recreational uses of Lake Waco.

Methods

Available data

Data were collected from multiple sources: (1) monitoring efforts by the City of Waco, (2) the Texas Institute of Applied Environmental Research (TIAER), (3) the Texas Commission on Environmental Quality (TCEQ; through its predecessor, the Texas Natural Resources Conservation Commission [TNRCC]), (4) the US Army Corps of Engineers, (5) the US Geological Survey (USGS), 6) the Brazos River Authority, and (7) Baylor University. Nearly half a million data points were entered and screened based on a US Environmental Protection Agency-approved QA/QC program. All data were uploaded to a Microsoft Access Data Base for use in this and other studies. Almost all data collected under the 2002–2005 City of Waco/Baylor University/ENSR study (Conry 2010) were excluded from this data base but were used separately for comparison. The exceptions include data for a more recent urban loading study, which were used to evaluate localized loading to Lake Waco in the absence of other data from the direct drainage area.

In the process of screening of these data from the various sources, gaps and inequities were found. Data from many stations were evaluated, but only a subset was chosen to represent the generation, routing and processing of nutrients in this watershed. Consequently, while a very large quantity of data was reviewed, not all was applied in this analysis.

Assessed time period

This analysis evaluated data from 1994 into 2001. The intent was to assess conditions over the most recent period that included enough data to ensure differences among the various sampling programs did not unduly influence the resultant calculations. While land use and management practices are always changing to some extent, this period was viewed as representative of current conditions in the watershed.

Precipitation drives the routing of water and pollutants to the reservoir (TNRCC 2001). The period applied, 1994–2001, captures sufficient variation to represent the range of loading Waco experienced by Lake Waco over a longer period of record (Fig. 2).

Figure 2 Variation in precipitation at Waco Airport from 1970 through 2005.

Assessed water quality variables

Phosphorus (P) and nitrogen (N) are of primary interest to conditions in Lake Waco. Because the load is the product of concentration and flow, the volume of water passing any point of interest per unit of time (i.e., the flow) is another critical variable in assessing loading to the reservoir. This analysis focuses on total phosphorus (TP), ortho-phosphorus (OP, or soluble reactive P) and total nitrogen (TN). Additional assessment of OP (the most available P form) is made because it provides a lower boundary for potential fertility. Ratios of TN:TP are also examined, based on mass; interconversion of forms during transport and data limitations prevented construction of ratios using different forms of N and P.

Assessed stations

More than 200 stations were sampled over the last 15 years as part of the various sampling programs. Stations chosen for further analysis in this study had adequate data for both wet and dry conditions, ranging from 50 to more than 100 samples for each condition. (Key stations in Table 1, complete list in Online Supplement). Additional stations were considered where data were not ideal to calculate a load, yet where a load calculation was considered essential (e.g., direct drainage area). Stations represent drainage areas or point sources within the watershed of Lake Waco and have been applied to a watershed model (White et al. 2010a). The watershed layout represents a complex network of inputs and stations (Fig. 3, detailed schematic in Online Supplement).

Table 1 Information for key sampling stations in the Lake Waco watershed.

	Station #
Description	Mainstem Station	Tributary to Mainstem	Area Represented (km2)	Latitude	Longitude
North Bosque River
North Bosque River at Stephenville (FM 8), upstream of WWTF	17226		227	32.23500	98.20389
Stephenville WWTF discharge		tp040	Point Source	32.19750	98.18361
North Bosque River at Erath CR 454, 3.3 KM downstream of US 377/67 in Stephenville, downstream of WWTF discharge	11963		296	32.19305	98.18389
North Bosque River at US 281, upstream of Duffau Ck confluence	11961		939	31.97722	98.03500
Hico WWTF discharge		lb010	Point Source	31.97917	98.02694
Duffau Creek at FM 927 West of Iredell		11810	224	31.98656	97.89208
Iredell WWTF discharge		lb020	Point Source	31.98806	97.86139
Meridian WWTF discharge		lb030	Point Source	31.92056	97.65444
North Bosque River at FM 219 NE of Clifton	11956		2549	31.78583	97.56778
Clifton WWTF discharge		lb040	Point Source	31.78417	97.56778
Neils Creek at SH 6, upstream of NBR confluence		11826	351	31.69400	97.53567
North Bosque River downstream of Neils Ck confluence	17605		3026	31.66842	97.46422
Valley Mills WWTF discharge		lb050	Point Source	31.66389	97.46361
South Bosque River
McGregor WWTF discharge		lb070	Point Source	31.41417	97.39611
South Bosque River upstream of Church Road	17229		216	31.47361	97.28056
Middle Bosque River
Middle Bosque River at FM 3047	17612		477	31.50897	97.36667
Hog Creek
Hog Creek at FM 185	17212		184	31.55528	97.35639

Figure 3 Major rivers, key sampling stations, dairy farms and wastewater discharge locations in the Lake Waco watershed. Station descriptions are in Table 1. See Online Supplement for a detailed sampling station schematic. NBR = North Bosque River, MBR = Middle Bosque River, SBR = South Bosque River, HC = Hog Creek.

Loading calculations

Loads were determined for each sampling date as a function of recorded flow and concentration for all stream monitoring stations. Flows are generally the average of multiple automated measures from USGS gauges, although some measurements were collected singly at the time of sampling. Concentrations represent the result of grab or composite samples at the stations, depending upon the monitoring program from which the data were obtained.

Because the data were often discontinuous, it is not possible to total daily load estimates. Load estimates were categorized by wet or dry weather conditions and scaled to estimate the portion of time represented by wet and dry conditions. Checks on precipitation data suggest that wet weather data involve precipitation on the day of sampling or within a day before that sampling, while dry weather data tend to represent periods proceeded by no precipitation for at least 3 days.

Some events were possibly misclassified, specifically days with light precipitation in which no runoff was created. These misclassifications could lead to inclusion of low flows and loads in the wet weather category when dry weather conditions are really represented. Underestimation of wet weather loading could be significant under such a situation.

To bracket the probable actual loading from wet and dry conditions, the precipitation record was examined in relation to the flow record. Precipitation occurred on 13.9% of days but runoff clearly occurred on only 6.3% of days, suggesting that dry weather conditions exist at least 86.1% of the time but possibly as much as 93.7%. Using these factors with the derived average daily wet and dry weather loading values, low and high composite daily load estimates were derived, the low end values attributing more time to dry conditions and the high end values assuming greater wet weather conditions.

For point sources (WWTF), mean flows were multiplied by mean concentrations to derive load estimates. Relatively low variability in the WWTF discharge monitoring reports suggested that this was a reasonable approach for these point sources. Loads derived from multiplication of mean flows and mean concentrations from stream stations strongly influenced by nonpoint sources are likely to be less accurate than an average of event-based loads, but were calculated for comparison.

The drainage area immediately surrounding Lake Waco has not been subject to the level of sampling that many upstream stations have experienced, requiring another approach to load estimation. Results of the urban runoff assessment performed as part of the Lake Waco Comprehensive Study (City of Waco, 2002–2005, unpubl. data) were applied to derive loads as a function of the median concentration and flow proportional to that of Hog Creek on an areal basis (relative watershed area).

Additional sources for which loads were calculated included direct ground water seepage, direct atmospheric inputs, recreational releases (mainly from people in boats), waterfowl contributions and internal generation (e.g., release from sediment, N fixation by algae). Ground water inputs were estimated as the expected range for seepage (in L/m²) times the expected range of the interface area (in m²) times the expected range of concentrations in the seepage, thus bracketing likely inputs. Atmospheric inputs were estimated as the quantity of precipitation falling directly on the lake, based on the precipitation record from the adjacent Waco Regional Airport, times a range of expected concentrations for N and P in that rainfall. This ignores dryfall, but the concentrations used were raised somewhat from typical expectations to adjust for this omission. Recreational inputs were assumed to be human wastes from boats and were estimated as a range of person-days per year times the typical output of N and P by a single person in a year. Waterfowl loading of N and P was estimated in the manner of recreational inputs, multiplying an expected range of bird-years at the lake by the average expected input of a bird over the course of a year. None of these estimates is very reliable, but all are extremely small relative to the main tributaries, and errors are expected to have minimal impact on the overall loading assessment.

Internal loading estimates could be set at zero based on the work of Esten and Wagner (2010), where releases of N and P from bottom sediment were negligible in field assessments. However, Doyle et al. (2010) demonstrated a significant input from N fixation, and that value was assumed as the annual internal load for N. For P, low end release rates (Nurnberg 1984) were multiplied by the likely range of contributing area times the expected duration of releases in most lakes during summer. Finally, in-lake data can be used in simple, empirically derived models from other studies (Kirchner and Dillon 1975, Vollenweider 1975, Jones and Bachmann 1976, Larsen and Mercier 1976, Reckhow 1977) to back-calculate the load of P that could have entered a lake with the features of Lake Waco to result in the observed in-lake concentration. Bachmann (1980) also provides an empirical model for N loading as well. Given a load, these models can also be used to estimate the in-lake concentration.

In-lake stations and data

Twelve in-lake stations were selected for data assessment as a function of location and data availability (Fig. 1, Online Supplement). Data from these stations, sampled by multiple research programs as noted previously, provided reality check values for loading analyses for the watershed and were also used in a series of empirical in-lake models to back-calculate the likely range of loads entering the lake.

Results

Individual station and basin results

Values for flow, P concentration, P load as calculated from the mean of event loads and mean concentration times mean flow, and OP to TP ratio were derived for each watershed station with adequate data (Table 2, Online Supplement). The same concentration and loading values were derived for N, plus the TN:TP ratio (Table 3, Online Supplement). Estimates are provided for dry or wet weather and combined values based on either the lower or higher wet weather frequency, bracketing the likely influence of wet weather on water quality. The WWTF point sources are not subject to the same level of variation in response to weather and are calculated from mean flow and concentration for all data. The drainage areas for each sampling station ranged from 4.7 to 3191 km² (Table 4, Online Supplement). These values were used to estimate the unit area flows and loads for TP and TN.

Table 2 Summary of phosphorus loading results for selected individual stations.

		Flow		TP Conc.		TP Load (Event)				TP Load (Means)				OP:TP
Description	Station #	Dry Station (CMS)	Wet Mean (CMS)	Dry Mean (mg/L)	Wet Mean (mg/L)	Dry Mean (kg/d)	Wet Mean (kg/d)	Best Est. Low (kg/d)	Best Est. High (kg/d)	Dry Mean (kg/d)	Wet Mean (kg/d)	Best Est. Low (kg/d)	Best Est. High (kg/d)	Median Dry Ratio	Median Wet Ratio
North Bosque River
North Bosque River at Stephenville (FM 8), upstream of WWTF	17226	0.30	1.55	0.440	0.657	11.9	134.2	19.6	28.9	11.3	88.0	16.1	22.0	0.63	0.58
Stephenville WWTF discharge	tp040	0.0666		2.645						15.2	15.2	15.2	15.2	0.89
North Bosque River at Erath CR 454, 3.3 KM downstream of US 377/67 in Stephenville, downstream of WWTF discharge	11963	0.48	2.47	1.700	1.127	46.4	223.9	57.5	71.1	70.3	240.3	80.9	93.9	0.86	0.71
North Bosque River at US 281, upstream of Duffau Ck confluence	11961	2.89	8.74	0.299	0.475	111.2	473.7	133.9	161.6	74.6	358.7	92.4	114.1	0.58	0.55
Hico WWTF discharge	lb010	0.0087		4.102						3.1	3.1	3.1	3.1	0.93
Duffau Creek at FM 927 West of Iredell	11810	0.63	4.98	0.125	0.290	9.9	243.4	24.5	42.3	6.8	124.9	14.2	23.2	0.14	0.36
North Bosque River at FM 216 in Iredell, upst of Iredell WWTF	11960	3.76		0.209	0.088	52.9				67.7				0.22
Iredell WWTF discharge	lb020	0.0013		4.797						0.5	0.5	0.5	0.5	0.91
Meridian WWTF discharge	lb030	0.0196		3.555						6.0	6.0	6.0	6.0	0.91
North Bosque River at FM 219 NE of Clifton	11956	4.02	26.48	0.097	0.215	66.2	568.2	97.6	135.9	33.7	491.8	62.3	97.3	0.13	0.22
Clifton WWTF discharge	lb040	0.0144		2.372						3.0	3.0	3.0	3.0	0.79
Neils Creek at SH 6, upstream of NBR confluence	11826	1.09	6.13	0.065	0.132	7.0	128.5	14.6	23.9	6.1	69.7	10.1	15.0	0.12	0.12
North Bosque River downstream of Neils Ck confluence	17605	6.47	34.51	0.104	0.310	136.5	2049.0	256.2	402.4	57.9	923.4	112.1	178.2	0.12	0.19
Valley Mills WWTF discharge	lb050	0.0022		3.144						0.6	0.6	0.6	0.6	0.92
South Bosque River
McGregor WWTF discharge	lb070	0.0292		2.200						5.6	5.6	5.6	5.6	0.82
South Bosque River upstream of Church Road	17229	0.21	7.07	0.086	0.365	2.9	461.1	31.6	66.6	1.5	222.6	15.4	32.3	0.28	0.12
Middle Bosque River
Middle Bosque River at FM 3047	17612	2.63	9.13	0.120	0.287	50.1	290.4	65.1	83.5	27.3	226.6	39.8	55.0	0.06	0.07
Hog Creek
Hog Creek at FM 185	17212	0.80	3.88	0.084	0.183	5.6	88.1	10.8	17.1	5.8	61.4	9.3	13.6	0.12	0.13

Table 3 Summary of nitrogen loading results for selected individual stations.

		Flow		TN Conc.		TN Load (Event)				TN Load (Means)				TN:TP
Description	Station #	Dry Mean (CMS)	Wet Mean (CMS)	Dry Mean (mg/L)	Wet Mean (mg/L)	Dry Mean (kg/d)	Wet Mean (kg/d)	Best Est. Low (kg/d)	Best Est. High (kg/d)	Dry Mean (kg/d)	Wet Mean (kg/d)	Best Est. Low (kg/d)	Best Est. High (kg/d)	Median Dry Ratio	Median Wet Ratio
North Bosque River
North Bosque River at Stephenville (FM 8), upstream of WWTF	17226	0.30	1.55	2.22	2.91	61.4	523.7	90.4	125.7	56.9	390.2	77.7	103.2	4.53	4.22
Stephenville WWTF discharge	tp040	0.0666		6.99						40.2	40.2	40.2	40.2	2.65
North Bosque River at Erath CR 454, 3.3 KM downstream of US 377/67 in Stephenville, downstream of WWTF discharge	11963	0.48	2.47	4.77	4.87	184.3	912.9	229.9	285.6	197.2	1039.2	249.9	314.2	2.95	4.16
North Bosque River at US 281, upstream of Duffau Ck confluence	11961	2.89	8.74	1.34	2.00	468.4	2010.2	564.9	682.7	333.7	1512.5	407.5	497.5	4.15	4.25
Hico WWTF discharge	lb010	0.0087		13.20						9.9	9.9	9.9	9.9	3.63
Duffau Creek at FM 927 West of Iredell	11810	0.63	4.98	0.90	1.90	75.2	1326.6	153.5	249.1	48.7	818.2	96.8	155.6	8.67	6.45
Iredell WWTF discharge	lb020	0.0013		19.17						2.2	2.2	2.2	2.2	6.62
Meridian WWTF discharge	lb030	0.0196		20.89						35.4	35.4	35.4	35.4	6.55
North Bosque River at FM 219 NE of Clifton	11956	4.02	26.48	1.05	1.37	455.1	3259.7	630.7	844.9	364.6	3128.0	537.6	748.8	12.81	7.02
Clifton WWTF discharge	lb040	0.0144		10.12						12.6	12.6	12.6	12.6	5.26
Neils Creek at SH 6, upstream of NBR confluence	11826	1.09	6.13	0.70	1.03	72.5	858.8	121.7	181.8	66.3	544.0	96.2	132.7	13.64	11.63
North Bosque River downstream of Neils Ck confluence	17605	6.47	34.51	1.14	1.99	844.1	9654.5	1395.6	2068.7	635.6	5929.3	967.0	1371.5	13.74	7.36
Valley Mills WWTF discharge	lb050	0.0022		18.85						3.6	3.6	3.6	3.6	6.12
South Bosque River
McGregor WWTF discharge	lb070	0.0292		11.44						28.9	28.9	28.9	28.9	5.87
South Bosque River upstream of Church Road	17229	0.21	7.07	5.85	6.67	84.7	4567.7	365.3	707.8	105.2	4071.0	353.4	656.4	153.97	39.14
Middle Bosque River
Middle Bosque River at FM 3047	17612	2.63	9.13	3.57	3.52	1326.1	4370.9	1516.7	1749.4	813.4	2781.8	936.6	1087.0	51.56	27.00
Hog Creek
Hog Creek at FM 185	17212	0.80	3.88	1.38	1.86	118.7	816.6	162.4	215.7	95.5	623.7	128.5	168.9	18.53	13.85

Table 4 Export coefficients for water, phosphorus and nitrogen from selected drainage areas.

			Flow		Flow Yield		TP Load Yield (kg/ha/yr)				TN Load Yield (kg/ha/yr)
Description	Station #	Area Represented (km2)	Dry Mean (CMS)	Wet Mean (CMS)	CMS/km2 Dry	CMS/km2 Wet	Dry Mean	Wet Mean	Best Est. Low	Best Est. High	Dry Mean	Wet Mean	Best Est. Low	Best Est. High
North Bosque River
North Bosque River at Stephenville (FM 8), upstream of WWTF	17226	227	0.30	1.55	0.0013	0.0068	0.2	2.2	0.3	0.5	1.0	8.4	1.5	2.0
Stephenville WWTF discharge	tp040	Point Source	0.0666
North Bosque River at Erath CR 454, 3.3 KM downstream of US 377/67 in Stephenville, downstream of WWTF discharge	11963	296	0.48	2.47	0.0016	0.0083	0.6	2.8	0.7	0.9	2.3	11.3	2.8	3.5
North Bosque River at US 281, upstream of Duffau Ck confluence	11961	939	2.89	8.74	0.0031	0.0093	0.4	1.8	0.5	0.6	1.8	7.8	2.2	2.7
Hico WWTF discharge	lb010	Point Source	0.0087
Duffau Creek at FM 927 West of Iredell	11810	224	0.63	4.98	0.0028	0.0222	0.2	4.0	0.4	0.7	1.2	21.6	2.5	4.1
Iredell WWTF discharge	lb020	Point Source	0.0013
Meridian WWTF discharge	lb030	Point Source	0.0196
North Bosque River at FM 219 NE of Clifton	11956	2549	4.02	26.48	0.0016	0.0104	0.1	0.8	0.1	0.2	0.7	4.7	0.9	1.2
Clifton WWTF discharge	lb040	Point Source	0.0144
Neils Creek at SH 6, upstream of NBR confluence	11826	351	1.09	6.13	0.0031	0.0175	0.1	1.3	0.2	0.2	0.8	8.9	1.3	1.9
North Bosque River downstream of Neils Ck confluence	17605	3026	6.47	34.51	0.0021	0.0114	0.2	2.5	0.3	0.5	1.0	11.6	1.7	2.5
Valley Mills WWTF discharge	lb050	Point Source	0.0022
South Bosque River
McGregor WWTF discharge	lb070	Point Source	0.0292
South Bosque River upstream of Church Road	17229	216	0.21	7.07	0.0010	0.0327	0.0	7.8	0.5	1.1	1.4	77.2	6.2	12.0
Middle Bosque River
Middle Bosque River at FM 3047	17612	477	2.63	9.13	0.0055	0.0192	0.4	2.2	0.5	0.6	10.1	33.4	11.6	13.4
Hog Creek
Hog Creek at FM 185	17212	184	0.80	3.88	0.0044	0.0211	0.1	1.7	0.2	0.3	2.4	16.2	3.2	4.3

North Bosque River (NBR)

The NBR drains 75% of the watershed area of Lake Waco, with all currently active dairy farms located in this drainage area. The area upstream of Stephenville (represented by the first 9 stations in Tables 2 and 3) has a high concentration of dairy farms and little else in the way of water quality influences. The TP and TN levels are very high in the streams in this area, and the TN:TP ratio is low, indicative of inputs of animal wastes. Export coefficients for water are generally low and typical of this geographic area (Central Texas; Dunn and Leopold 1978), while export coefficients for TP and TN are quite high in most cases (Clark et al. 2000).

Station 17226, on the NBR just upstream of Stephenville, represents most of the area upstream of Stephenville, and about one-third of the dairy farms. The NBR estimated average daily loads of 19.6–28.9 kg of P and 90.4–125.7 kg of N, at this point based on event averages. The mean concentration and flow values are lower and not as reliable but are provided for comparison. While the discharge load is rather large, on a unit load area basis the annual export coefficients are moderate (0.3–0.5 kg/ha/yr, Table 5). Flows, concentrations and loads are all much higher during wet periods, indicating that wet weather is responsible for most of the loading. Export coefficients are much higher for upstream subwatersheds (Online Supplement).

Table 5 Mean dry and wet weather flows and nutrient concentrations at selected stations.

		Flow			TP		TN
Description	Station #	Dry Mean (CMS)	Wet Mean (CMS)	USGS Mean (CMS)	Dry Mean (mg/L)	Wet Mean (mg/L)	Dry Mean (mg/L)	Wet Mean(mg/L)
North Bosque River downstream of Neils Ck confluence (aka bo100)	17605	6.47	34.51	8.09	0.104	0.310	1.14	1.99
South Bosque River upstream of Church Road	17229	0.21	7.07	1.82	0.086	0.365	5.85	6.67
Middle Bosque River at FM 3047, well upstream of lake	17612	2.63	9.13	2.01	0.120	0.287	3.57	3.52
Hog Creek at FM 185, well upstream of lake	17212	0.80	3.88	0.93	0.084	0.183	1.38	1.86

The Stephenville WWTF adds an average of 15.2 kg P/d and 40.2 kg N/d to NBR, with estimated downstream river lows of 23.3–32.6 kg P/d and highs of 78.7–105.5 kg N/d. By the time NBR passes through Hico, most dairy farm inputs have been made, with the exception of the Duffau Creek drainage area. Adding the loads at stations 11961 and 11810, the loads carried by NBR at Hico are 158.4–203.9 kg P/d and 718.4–931.8 kg N/d. The Hico WWTF adds another 3.1 kg P/d and 9.9 kg N/d. The area of the total NBR watershed involved is about 1163 km², resulting in export coefficients of 0.4–0.7 kg/ha/yr for P and 2.2–4.1 kg/ha/yr for N.

Measurements in NBR at Clifton are limited, especially for storm water. Loads are higher at Clifton (station 11956) than at Hico (stations 11961 and 11810), but only 7–15%, despite more than doubling the total watershed area represented. The loading estimates for the Clifton area may not be reliable as a function of emphasizing dry weather. In particular, Duffau Creek experiences great impact from dairy operations during wet weather but has minimal flow during dry weather.

Neils Creek (station 11826) meets the NBR slightly downstream of Clifton. As an ecoregional reference stream, Neils Creek represents better conditions than most other streams in the NBR system. The loads from the Neils Creek drainage area are 14.6–23.9 kg P/d and 96.2–132.7 kg N/d, yielding export coefficients of slightly less than 0.2 kg P/ha/yr and 1.3–1.9 kg N/ha/yr for this 351 km² drainage area. These values are lower than for other assessed subwatersheds and suggest a TN:TP ratio slightly in excess of 10:1. While there are some anthropogenic influences on Neils Creek, this drainage area seems to represent the level of loading that might be experienced in the absence of human influences (development and agriculture) in the Lake Waco watershed.

Slightly downstream of the confluence of the NBR with Neils Creek is the last NBR station with sufficient data to construct reliable loading estimates. The watershed area represented is 3036 km², with another 166 km² downstream before Lake Waco. At station 17605 the average daily P load carried by NBR is between 256.2 and 402.4 kg, while the average daily N load is between 1395.6 and 2068.7 kg. The Iredell, Meridian and Clifton WWTF, all downstream of the last noted WWTF at Hico, add a combined load of 9.5 kg P/d and 50.2 kg N/d. The TN:TP load ratio is between 5.1 and 5.5. The export coefficients at station 17605 are 0.3–0.5 kg/ha/yr for P and 1.7–2.5 kg/ha/yr for N. Downstream of station 17605, the Valley Mills WWTF adds another 0.6 kg P/d and 3.6 kg N/d, and the land uses are largely residential and open and low density agriculture, so the inputs should not be higher than proportional to the area they represent, relative to the rest of the NBR drainage area.

South Bosque River (SBR)

There are fewer stations sampled on SBR (and MBR and HC) than for NBR, and SBR has a smaller watershed (226 km²). There were enough data for an appropriate loading analysis at only one river station (17229), which represents 96% of the drainage area. The load of P averages 15.4–32.3 kg/d, while the N load averages 353.4–656.4 kg/d,

with a TN:TP ratio of 20–23. Median ratios for dry and wet weather concentrations are even higher. Export coefficients are 0.5–1.1 kg P/ha/yr and 6.2–12.0 kg N/ha/yr. The McGregor WWTF adds an average P load of 5.6 kg/d and an average N load of 28.9 kg/d. This represents a substantial fraction of the P load, but not of the N load. The low TN:TP ratio in the WWTF effluent is counteracted by inputs of N from elsewhere in the subwatershed, which has a substantial amount of land in crop agriculture.

Middle Bosque River (MBR)

The MBR drains an area of 516.1 km², combining with the SBR near the inlet to Lake Waco. Station 17612 is the closest one to the confluence with adequate data for a loading analysis, draining over 92% of the total area. The loads are between 65.1 and 83.5 kg P/d and 1516.7 and 1749.4 kg N/d, with a TN:TP ratio of 21–23. As with SBR, median values for ratios based on measured concentrations are even higher. Export coefficients are 0.5–0.6 kg P/ha/yr and 11.6–13.4 kg N/ha/yr. The Crawford WWTF is within this watershed but has no discharge; water evaporates from holding ponds before any is released to the stream system. Crop agriculture is an important land use in this subwatershed.

Hog Creek (HC)

Hog Creek drains a very linear watershed with an area of 211.5 km². The closest station to its inlet to Lake Waco for which adequate data were available (17212) drains 87% of the watershed. Loads at station 17212 average 10.8–17.1 kg P/d and 128.5–168.9 kg N/d, with a TN:TP ratio of 10–12. Export coefficients are 0.2–0.3 kg P/ha/yr and 3.2–4.3 kg N/ha/yr. Current land uses are largely open and agricultural.

Orthophosphorus fraction

The ratio of OP to TP was reported as the median value for dry or wet conditions, providing a quick summary of the portion of the TP load that is OP (Table 2, Online Supplement). Values are not especially different for dry vs. wet conditions at the vast majority of stations, but there is considerable variability among stations. More than 80% of TP is OP for the WWTF, consistent with expectations based on treatment technology applied at these facilities. For the stations downstream of dairy farms, values ranged from 0.15 to 0.75. On the mainstem NBR, dry weather values tended to decline in the downstream direction. The OP:TP ratio rose from 0.63 just upstream of Stephenville to 0.86 just downstream of the Stephenville WWTF discharge (which adds nearly all of its P as OP), but then declined to 0.58 near Hico, 0.22 near Iredell, 0.13 at Clifton and 0.12 slightly upstream of Valley Mills. Uptake and conversion of OP to particulate P forms is a likely factor in this trend. For wet weather values, the pattern is the same, although the OP fraction tends to be slightly less in the upper watershed and slightly larger in the lower watershed, yielding less change overall from upstream to downstream. This would be expected because there is limited time for processes to act upon P in the river during storm events before the load is delivered to Lake Waco.

For SBR, OP:TP ratios in the river were 0.12–0.28, although the ratio for the McGregor WWTF was expectedly high at 0.82. In MBR, ratios were even lower, ranging from 0.06 to 0.22. Hog Creek produced only one value, a low 0.13. Most P in these rivers is in particulate matter, not dissolved forms.

Loading to Lake Waco from the 4 major tributaries

The stations that represent the 4 major tributaries are 17605 for NBR, 17229 for SBR, 17612 for MBR and 17212 for HC. The NBR is the dominant source of water under all conditions, The MBR is next most important under all conditions, although SBR contributes similarly at least during wet weather, according to a limited USGS record for that river. Hog Creek adds about half the water that SBR contributes, but note that HC actually provides more water during dry conditions than SBR. This is undoubtedly related to shape, topography and land use in those basins.

Phosphorus concentrations were similar among the NBR, SBR, MBR and HC under dry conditions, but the average for HC was substantially lower during wet weather (Table 5). Nitrogen concentrations were more variable among rivers but less variable among dry and wet weather conditions. Nitrogen levels were highest in SBR, followed by those for MBR, with values for NBR and HC similar and lower yet. Loads for both N and P increase with wet weather, given sharp rises in flow, but the concentration rises substantially only for P during wet weather, making storms more important for the loading of P than N and depressing the N:P ratio (Table 2 and 3, Online Supplement).

The load of P delivered to Lake Waco from each of the 4 tributaries addressed here is predominantly particulate phosphorus, but NBR and HC contribute higher percentages of OP relative to TP, at 19–25%, relative to values of 7–9% for SBR and MBR (Table 2, Online Supplement). Given much higher flows, the load from NBR is considerably higher than for the other tributaries, representing 70% of the TP load and 84% of the OP load in this assessment. The MBR, SBR and HC, in that order, provide lesser amounts of TP and OP.

The situation is different for N, however. Despite much higher flows, lower N concentrations in NBR result in approximately equal contributions of TN from NBR and MBR, with MBR contributing slightly more N during dry weather and NBR providing somewhat more N during wet conditions. The SBR provides a little less than half as much N as NBR or MBR, and HC provides less than half of what SBR provides (Table 6). The TN:TP ratios from this load calculation approach vary widely, with NBR having a very low ratio of 5.1–5.5 and MBR yielding the highest ratio at 21.0–23.3. The SBR and HC exhibited intermediate ratios of TN to TP, SBR with a ratio of 10.6–11.6 and HC with a ratio of 12.6–15.1.

Table 6 Mean loads and nutrient ratios at selected stations.

		TP Load (Event)		OP Load (Event)		OP:TP		TN Load (Event)		TN:TP
Description	Station #	Best Est. Low (kg/d)	Best Est. High (kg/d)	Best Est. Low (kg/d)	Best Est. High (kg/d)	Low End Load Ratio	High End Load Ratio	Best Est. Low (kg/d)	High End High (kg/d)	Low End Load Ratio	High End Load Ratio
North Bosque River downstream of Neils Ck confluence (aka bo100)	17605	256.2	402.4	54.5	76.7	0.21	0.19	1395.6	2068.7	5.45	5.14
South Bosque River upstream of Church Road	17229	31.6	66.6	2.8	5.9	0.09	0.09	365.3	707.8	11.56	10.63
Middle Bosque River at FM 3047, well upstream of lake	17612	65.1	83.5	4.7	6.0	0.07	0.07	1516.7	1749.4	23.29	20.96
Hog Creek at FM 185, well upstream of lake	17212	10.8	17.1	2.7	4.0	0.25	0.23	162.4	215.7	15.05	12.62

Loading from areas and sources other than the 4 major tributaries

Data for other inputs to Lake Waco are less abundant, preventing the kind of detailed analysis provided for the major watershed sources. However, some site-specific data and knowledge of relevant literature values allowed estimation of inputs from direct drainage, ground water seepage, atmospheric deposition, recreation, waterfowl and internal sources (Table 7). Where possible, the likely range of inputs was bracketed. These additional sources are minor compared to the load from the 4 major tributaries. The largest direct N and P loads to the lake are from the direct drainage watershed, and these loads are similar to those calculated for HC, the smallest of the major tributaries. All other direct loads are substantially lower than for the direct drainage area, except for the estimated internal load.

Table 7 Loads from additional areas and sources to Lake Waco.

Source	Low End Estimate of Flow (CMS)	High End Estimate of Flow (CMS)	Low End TP Conc. (mg/L)	Low End TP Conc. (mg/L)	Low End Load TP (kg/d)	High End TP Load (kg/d)	Low End TN Conc. (mg/L)	High End TN Conc. (mg/L)	Low End Load TN (kg/d)	High End TN Load (kg/d)
Direct Drainage Area	0.462	0.462	0.21	0.28	9.1	12.2	2.33	3.11	274.9	366.5
Ground Water Inseepage	0.02	0.07	0.05	0.15	0.1	0.9	0.6	3	0.9	17.4
Atmospheric Inputs	0.75	0.75	0.02	0.05	1.3	3.3	0.2	0.5	13.0	32.6
	Units/Year	Units/Year	kg/unit/d	kg/unit/d			kg/unit/d	kg/unit/d
Recreation	50	100	0.0041096	0.0041096	0.21	0.41	0.0126301	0.0126301	0.63	1.26
Waterfowl	500	1000	0.0005	0.0005	0.27	0.55	0.0026	0.0026	1.30	2.60
	Area (ha)	Duration (d)	mg/m2/d	mg/m2/d					From Doyle et al. 2010
Internal Load	1450	120	0.5	1.0	2.4	4.8			142	338

The internal P load as estimated was minor (Table 7). It is most likely generated in the summer and fall, when it would be proportionally more important during the time of greatest concern Compared to watershed loads internal P loading is small. The internal N load is derived from direct N fixation measurements (Doyle et al. 2010) and is a function of cyanobacterial activity. The low end average value of 142 kg/d is the result of dividing the total measured N fixation load by 365 days. Because N fixation occurs mainly over about 5 months of the year, however, the high end value of 338 kg/d is more appropriate when considering the important summer–fall period of problematic algal blooms.

Recreation and waterfowl input calculations relied on assumptions of the contributing number of units (humans or birds; Table 7) without substantial data upon which to base those assumptions. However, even increasing the number of units 10-fold, the resultant loads would still be small relative to watershed inputs.

Overall load to Lake Waco

Combining the load estimates for all itemized sources, the best available estimate of loads to Lake Waco was derived (Fig. 4, Online Supplement). Estimates represent anticipated long-term averages, based on best professional judgment regarding applied assumptions (Online Supplement). Application of alternative assumptions suggests that these values will change somewhat, but most load estimates and all the larger ones co-vary, such that the order of sources does not change and the ratio among estimates changes little.

Figure 4 Apportionment of loads among sources for Lake Waco: A. Flow; B. TP; C. OP; D. TN.

Water enters Lake Waco at a long-term average rate of 14.1 cms, 62% of which is delivered by NBR. Another 15.5% is delivered by MBR, with no other remaining source suppling even half as much. While NBR is by far the greatest source of water to Lake Waco, it has a lower yield per unit area than most other sources, owing to greater evaporative losses over the longer trip to the lake in the much larger NBR drainage area.

The TP load to Lake Waco averages 147,093 kg/yr, two-thirds of which enters via NBR (Fig. 4). The MBR supplies 17.4%, while SBR provides less than half of that (8.2%), and each remaining source delivers less than half the SBR load. Sources such as internal load, waterfowl and recreation inputs may have slightly greater importance than their percentages of the TP load indicate because they are concentrated in the growing season for problem algae, but the percentages are so low as to be inconsequential even if magnified 10-fold.

Most TP enters the lake in particulate form; the OP load is estimated to average 28,331 kg/yr, about 19% of the TP load. The OP is immediately available for algal uptake and would represent the minimum effective load to the lake; however, the remaining 81% of the TP load is not necessarily inert. Some non-OP may be readily available as well, and organic particles can decay and convert particulate P into OP. For OP, however, NBR is again the largest single source, at almost 70% of the overall OP load (Fig. 4). The NBR is also expected to contain the greatest fraction of particulate P that could be converted into available P, having an expected high organic content relative to other tributaries, but data for additional P fractions is not available for tributary loads.

The TN load to Lake Waco averages 1,447,383 kg/yr, with MBR as the largest source (41.2%), followed closely by NBR (37.0%). The SBR is the next largest source at 9.6%, with no other source contributing even half as much. The difference between TN and TP load distributions is striking; the load from NBR is not as large as that from the much smaller MBR basin. The unit load coefficient of TN for NBR is considerably smaller than for the other 3 tributaries and the direct drainage area.

The TN:TP ratio for the complete load to Lake Waco is close to 10:1, a threshold of concern with regard to promotion of cyanobacterial blooms. The ratio can be expected to decline during dry periods (summer and fall in most years), when NBR becomes proportionally even more important as a water and nutrient source than it is on average. Higher flow periods will not necessarily raise the TN:TP ratio much beyond the 10:1 threshold, however, given the importance of flow from NBR, but the ratio can be expected to reach a maximum when there are greater inputs from MBR.

Effective load to Lake Waco

The “effective” load to Lake Waco is the amount of nutrient loading reflected in the reservoir's condition. A considerable amount of P, and possibly much N as well, enter the lake in unavailable forms that may not become available for algal uptake, at least not in the short-term. Application of in-lake concentration data and basic physical and hydrologic features of the lake in a number of empirical models allows back-calculation of the load necessary to achieve the observed in-lake concentrations. Entering a load into the models on a unit area basis allows prediction of in-lake concentrations as well.

Running the empirical models with the total estimated TP input to Lake Waco of 5.07 g/m²/yr and an average input concentration of 290 μg/L (flow weighted average) suggests that the average in-lake concentration should be 222 μg/L, more than twice the actual average for 1994–2001 (Table 8). Application of the actual average in-lake concentration of 104 μg/L yields a predicted load of only 2.46 g/m²/yr (average of the 5 empirical model results), slightly less than half the estimated load from actual data. Lake Waco responds as though it is getting slightly less than half the phosphorus load that it has been estimated to receive. Using the average TN concentration in the lake (1130 μg/L) and the estimated load (49.9 g/m²/yr), the result is roughly the same; models predict that the lake should receive about half that load or have an in-lake concentration about twice what it exhibits.

Table 8 In-lake concentrations for phosphorus and nitrogen at selected stations.

Description	Station #	Mean TP (mg/L)	Median OP:TP Ratio	Mean TN (mg/L)	Median TN:TP Ratio
North Arm
Lake Waco North Bosque Arm at FM 185 bridge	11946	0.130	0.17	1.07	8.05
Lake Waco/NBR near 17204, almost 2 mi from 11946	1	0.117	0.25	1.03	11.19
Lake Waco/NBR near Speegleville Park	17205	0.100	0.05	0.96	10.18
Lake Waco/NBR approaching airport area	11945	0.095	0.06	0.98	14.50
South Arm
Lake Waco/SBR just downstream of confluence of SBR/MBR	11949	0.115	0.11	2.54	16.88
Lake Waco just upstream of SH 6 Bridge	11948	0.093	0.07	1.30	13.25
Lake Waco near Koehne Park	17210	0.095	0.07	1.16	13.03
Main Body
Lake Waco about 30 yds out from structure of dam, near intake	11942	0.087	0.06	1.04	13.84

Examination of the actual concentrations observed in the lake and their distribution over space indicates that there is a decline in TP moving from the major NBR and MBR/SBR inlets out into the main body of the lake. The closest lake station to the NBR inlet (11946) has a mean TP concentration of 130 μg/L, and the station closest to the MBR/NBR inlet (11949) has a mean TP level of 115 μg/L, while the station at the dam (11942) has a mean value of 87 μg/L. Additionally, the mean TP concentrations for the stations closest to the major inlets are considerably lower than the values for the upstream river stations from which input loads were derived; much P may be lost quickly in this system.

The OP:TP ratio also declines markedly over that space. There is a corresponding decline for TN as well, but this decrease is less dramatic for the north arm of the lake and more dramatic for the south arm, contrary to the relative change in TP as water moves through those arms and into the main body of the lake. As a result, the TN:TP ratio increases from the NBR inlet to the main body of the lake, while it decreases between the MBR/SBR inlet and the main body of the lake. Much of the load to Lake Waco is apparently unavailable, such that application of total loads or concentrations will yield predictions much higher than reality for this reservoir.

Although some OP may be inactivated in the lake, most likely by calcium (Esten and Wagner 2010), most will be available for uptake by algae and represents a lower boundary on the expected effective load to the lake. Applying the OP load as determined by the loading analysis in the empirical models, the concentration expected from that load is 37 μg/L, a little more than one-third of the measured average in the lake. To get the observed average TP concentration in Lake Waco of 104 μg/L, a load equivalent to all of the OP load and about one-third of the remaining particulate P load is needed, equating to the 2.5 g/m²/yr predicted to yield that value.

Relative importance of loading from wastewater treatment facilities

Records for WWTF in the Lake Waco watershed are sufficient to make reliable estimates of total nutrient loads from those facilities (Table 9). Given limited variability over time and no expected correlation between flow and concentration, calculations were based on mean or median flows and do not vary substantially by calculation method. Although actual discharge rates are lower, the design flows are applied to represent the maximum allowable discharge from these WWTF. The TP load from WWTF during the period of record applied in this analysis is 12,551 kg/yr; the corresponding load for TN is 49,030 kg/yr. As mentioned previously, the Crawford WWTF has no actual discharge; effluent evaporates in holding ponds with no overflow to MBR recorded since operation of this facility began.

Table 9 Wastewater inputs to tributaries of Lake Waco.

WWTF	Station #	NPDES #	Design Flow (mgd)	Avg Flow (mgd)	Max Flow (mgd)	Mean TP Conc. (mg/L)	Median TP Conc. (mg/L)	Min. TP Conc. (mg/L)	Max. TPConc. (mg/L)	TP Load (by means) (kg/yr)	TP Load (by median) (kg/yr)	Mean TN Conc. (mg/L)	Conc. (mg/L) Conc. (mg/L)	Conc. (mg/L) Conc. (mg/L)	Max. TN Conc. (mg/L)	TN Load (by means) (kg/yr)	TN Load (by median) (kg/yr)
Hico	lb010	TX0026590	0.20	Unknown	Unknown	4.10	3.67	0.53	39.50	1141	1021	13.20	12.25	0.84	155.13	3672	3407
Iredell	lb020	TX0024848	0.05	0.03	0.06	4.80	2.91	0.05	184.00	200	121	19.17	17.67	1.39	101.33	800	737
Meridian	lb030	TX0053678	0.45	Unknown	Unknown	3.55	3.43	1.02	21.50	2225	2146	20.89	21.25	1.87	36.58	13075	13296
Clifton	lb040	TX0033936	0.65	0.33	0.60	2.37	2.23	0.09	10.20	1088	1023	10.12	6.65	0.93	49.50	4642	3052
Valley Mills	lb050	TX0075647	0.36	0.05	0.10	3.14	3.16	0.17	6.39	219	220	18.85	18.90	0.61	29.99	1311	1314
Crawford	lb060	TX0054666	0.03	0.00	0.00	1.52	0.80	0.11	5.38	0	0	7.56	5.17	1.94	21.45	0	0
McGregor	lb070	TX0023914	1.10	0.67	2.81	2.20	1.72	0.42	19.80	2050	1603	11.44	10.89	1.42	23.70	10658	10146
Stephenville	lb080/tp040	TX0024228	1.85	1.53	4.76	2.65	2.59	0.11	15.00	5628	5500	6.99	5.96	1.50	20.00	14872	12681
	Total		4.69	2.61						12551	11634					49030	44634

The TP load from the 7 WWTF with discharges, using design flows (Table 9) and assuming no attenuation of the load, represents 8.5% of the TP load entering Lake Waco. The TN load from WWTF represents 3.4% of the TN load to the lake. If we factor in an estimated 33% attenuation (loss) of these loads on the way to the lake, the WWTF contribute 5.7% of the entire TP load to Lake Waco, and 2.3% of the whole TN load to the lake. At the current average discharge rates (Table 10), the load from WWTF is 35% less, suggesting that the TP load to Lake Waco attributable to WWTF is on the order of 3.7% while the TN load would be 1.5%.

Table 10 Nutrient export coefficients and loading for land in subwatersheds including dairy operations.

Sampling Site	Station #	Drainage Area (ha)	P Export Coefficient (kg/ha/yr)	P Export minus Bkgrd (0.15 kg/ha/yr)	P Load Beyond Bkgrd (kg/yr)	P Load Adjusted for Bkgrd and WWTF (kg/yr)	N Export Coefficient (kg/ha/yr)	N Export Bkgrd (1.4 kg/ha/yr)	N Load Beyond Bkgrd (kg/yr)	N Load Adjusted for Bkgrd and WWTF (kg/yr)
North Bosque River
Scarborough Creek at CR 428	17221	860	4.63	4.48	3850	3850	11.57	10.22	8790	8790
Scarborough Creek at CR 423	17222	530	12.63	12.47	6611	6611	35.45	34.10	18070	18070
Unnamed Tributary of Scarborough Creek at CR423	17223	471	5.04	4.89	2302	2302	20.45	19.10	8994	8994
North Fork North Bosque River @ SH 108	17413	7830	0.73	0.58	4563	4563	2.62	1.27	9925	9925
South Fork North Bosque River 1km upstream FM 21 9	17218	817	0.48	0.33	267	267	3.28	1.93	1577	1577
Goose Branch downstream of FM8	17215	606	3.28	3.13	1896	1896	19.24	17.88	10837	10837
South Fork North Bosque River 2 KM Upstream of SH 108 in Stephenville	14382	12400	0.68	0.53	6584	6584	2.70	1.35	16693	16693
North Bosque River at Stephenville (FM 8), upstream of WWTF discharge	17226	22700	0.63	0.48	10791	10791	2.38	1.03	23280	23280
North Bosque River at Erath CR 454, 0.6 KM West of US 281 and 3.3 KM downstream of US 377/67 in Stephenville, downstream of WWTF discharge	11963	29600	0.71	0.56	16493	10941	2.83	1.48	43867	29197
Indian Creek near US 281	17235	688	1.29	1.13	780	780	3.80	2.45	1685	1685
Alarm Creek 2.7km east FM914	17237	5420	0.67	0.52	2804	2804	1.97	0.62	3346	3346
Sims Creek Upstream of US 281	17240	1710	0.85	0.70	1197	1197	4.60	3.25	5557	5557
Green Creek at unnamed road 1 .8 KM upstream of the confluence with the North Bosque River	13486	25800	0.37	0.22	5704	5704	2.72	1.36	35178	35178
Spring Creek at CR271	17242	1350	0.29	0.14	183	183	1.91	0.56	750	750
North Bosque River at US 281 near Hico, upstream of Duffau Ck confluence	11961	93900	0.52	0.37	34629	30471	2.20	0.84	79141	68150
Duffau Creek at FM 927 West of Iredell	11810	22400	0.40	0.25	5546	5546	2.50	1.15	25719	25719
North Bosque River at FM 216 in Iredell, upst of Iredell WWTF	11960	140200	0.42	0.27	37561	33194	1.93	0.57	80583	68651
North Bosque River downstream of Neils Ck confluence(akabo100)	17605	302550	0.31	0.16	47643	41114	1.68	0.33	100096	76852

Since the time that most data were collected for this assessment, both the Stephenville and Clifton WWTF have improved treatment technology to achieve a discharge concentration of 1.0 mg/L for TP. Using this lower TP concentration to estimate the loads from these 2 facilities, the total load from WWTF is reduced by 33% and represents 2.5% of the estimated total P input to Lake Waco. In terms of total load, WWTF represent a minor influence on the Lake Waco system.

Relative importance of loading from dairy operations

Dairy operations in the Lake Waco watershed are concentrated upstream of Iredell, placing them in the upstream half of the watershed, most in the upstream third of the watershed (Fig. 3). The permit records for dairies in this watershed (TCEQ 2002) indicate 68,334 head of dairy cows, although not all permitted operations may be active, and this may not account for replacement animals being raised in the same area but not yet milked. However, the most recent tally by TIAER researchers (TIAER 2007) suggests that the number of active dairy cows in the watershed is 40,350.

Measurement of specific dairy operation inputs is complicated by the physical location of operations and management practices that include on-site lagoons, various manure storage options and off-site waste application fields (WAF). Some subwatersheds have many more dairy farms than others, and a few subwatersheds have no dairy farms, thereby providing a reference condition. Background contribution for watersheds without dairy farms, WAF and WWTF discharges was determined from data for the Neils Creek and Meridian Creek subwatersheds. The estimated annual export of P is 0.15 kg/ha/yr, while for N it is 1.4 kg/ha/yr. These values are consistent with expectations from the literature (Reckhow et al. 1980, Clark et al. 2000). Subtracting this background load from total loads for subwatersheds with dairy farms and/or WAF but no WWTF provides an indication of dairy operation inputs (Table 10). For subwatersheds with WWTF, that contribution can also be subtracted to generate a dairy input estimate.

Corrected export coefficients (with background and WWTF contributions removed) for P from subwatersheds of NBR that include dairy operations range from 0.14 to 12.5 kg/ha/yr. Corresponding export coefficients for N range from 0.3 to 34.1 kg/ha/yr. The wide range of export coefficients attributed to dairy operations reflects several factors, including the area of the subwatershed devoted to dairy operations, the proximity of those operations to watercourses, attenuation as the load moves downstream and possible current management practices. The lowest corrected export coefficients come from Spring Creek, which has few dairy operations and none close to the stream, while the highest values are associated with the Scarborough Creek system and Goose Creek, having notably high concentrations of farms in small drainage areas (which places the sampling point closer to the actual sources). Dairy operations occupy only a fraction of watershed land overall (<5%); export coefficients provided here are not for specific farms or even dairy operations alone, but for subwatersheds containing dairy operations with estimated background loads and influence from wastewater treatment discharges removed. Dairies can contribute at much higher levels per unit of actual operational area (Reckhow et al. 1980).

At the 2 downstream mainstem NBR stations for which nondairy source loads were subtracted, estimates of 33,194 (Iredell) and 41,114 (Valley Mills) kg/yr are derived. The dairy-related load represents 56 and 44% of the total P load at those points, respectively. There may be some minor urban or agricultural inputs not being subtracted from the load at these points, but these influences should be adequately covered by the background estimate derived for this system. There will be some additional attenuation of dairy loads between these points and the reservoir. Comparing the TP loads attributed to dairy inputs in NBR at Iredell and Valley Mills to the TP load to Lake Waco from the entire NBR drainage area, dairy operations are estimated to account for 34–42% of the NBR TP load. Comparing the estimated dairy load to the TP load to the reservoir from all sources, dairy operations account for 23–28% of the total.

The soluble portion of the P load from subwatersheds dominated by dairy operations ranges from 0.44 to 0.75, with an average of 0.56 (Online Supplement). This is consistent with literature values for runoff and leachate from dairy operations (Sharpley et al. 1984), higher than any other known source in this watershed except for WWTF inputs. Additionally, much of the particulate P will be in a degradable organic form, which may form soluble P as the load moves downstream and particularly in the lake itself. Background loads will be largely inorganic P bound to soil particles and considerably less available for algal uptake over time. This means that the dairy inputs are likely to represent proportionally more of the OP load and the biologically available load to Lake Waco.

Scaling the loads based on the fraction of TP represented by OP suggests that dairy inputs account for 46% of the OP loaded to the lake. Considering the biodegradability of most dairy inputs, relative to background loads, dairies could provide as much as 47% of the particulate P load that degrades in the lake. This suggests that dairy inputs could account almost half of the biologically available P load to Lake Waco (46% of all OP; 47% of particulate P that eventually becomes OP in the lake), about twice the dairy contribution to the TP load in terms of percentage.

The same analysis for N reveals dairy-related loads at Iredell and Valley Mills of 68,651 and 76,852 kg/yr, respectively, representing 13–14% of the NBR TN load to Lake Waco and 4.7–5.3% of the overall TN load to the reservoir. The relative contribution of N from dairy operations to NBR and to Lake Waco is much smaller than that for P, resulting in low N:P ratios. Load ratios of N to P for MBR, SBR and HC are much higher, consistent with expected ratios for lands dominated by crops (Uttormark et al. 1974).

As a check for this overall approach to estimating loads from dairy operations, the general production of P and N by dairy cattle can be calculated for the existing herds. According to several older estimates of loading per cow (Uttormark et al. 1974, Omernik 1976), dairy cows can produce 20–25 kg P/1000 lb animal/yr. Dairy cows are larger now than 30 years ago when this research was done, and feed mixes may elevate the P output, but this assessment assumes an output of 25 kg per animal per year. Assuming 40,350 cows (TIAER 2007) at the 25 kg/yr P export rate, in excess of 1.0 million kg of P are generated in the Lake Waco watershed each year by dairy cows. If all granted permits were filled to the 2002 limit, the value would be in excess of 1.7 million kg/yr. The low end manure-based TP generation estimate is at least 6.8 times the estimated annual TP load to the reservoir. The high end estimate is 11.8 times the annual TP load to the lake. Even with major losses of TP to soil adsorption in WAF or through settling in PL566 reservoirs as the load moves downstream, it is not difficult to envision the estimated load from dairy cows reaching Lake Waco.

The same analysis for N from dairy cows (38 kg/animal/yr) indicates that at least 1.5 million kg and possibly as much as 2.6 million kg of N are generated by dairy cows in the watershed. Compared to the potential TP generation, this represents a ratio of <2:1 for N:P, but N is more mobile than P and higher ratios in runoff would be expected. High N levels in dry weather samples from the NBR drainage area are consistent with a high N burden in groundwater induced by dairy-related loading.

Further evidence of dairy influence on NBR water quality comes from tracking OP concentration during storms (Online Supplement). Flows are available for Hico and Valley Mills USGS stations, and the water from upstream of Hico arrives at Valley Mills no sooner than the peak for the downstream Valley Mills station and passes by Valley Mills for several days afterward. With all but a few dairies upstream of Hico, the load from the dairies will peak no earlier than the peak flow at Hico. The relationship between flow and OP concentration and load at Valley Mills is consistent with the scenario of major inputs far upstream in the watershed, the only significant source of which is dairy operations. Elevated concentrations and loading are often prolonged after the peak at Valley Mills, such that a disproportionate amount of the total load associated with a storm passes Valley Mills after the peak flow, coincident with the runoff from dairy operations. Consequently, the last water to enter Lake Waco after a storm contains the most P. The flushing value of storms may be offset by reloading the lake with new P at the end of storm events.

Projected background conditions in Lake Waco

The export coefficient derived for land in the watershed with limited effects from human activities is assumed to represent the “natural” level of P loading, removing dairy, other major agriculture (e.g., extensive cropland, other livestock operations) and WWTF influence from the watershed. It does not eliminate all human influence, or even the impact of all agriculture, and certainly incorporates some development and related activities, which are present in the watersheds used to derive the background export coefficient. The corresponding export might be considered to represent a lower expected limit on loading from this watershed.

Considering an area of 426,700 ha and applying an export coefficient of 0.15 kg/ha/yr suggests a watershed load of 64,000 kg/yr for TP. Adding 2572 kg/yr for atmospheric, ground water, recreation, waterfowl and internal loads yields an estimated total load of 66,572 kg/yr, less than half the current load. Flows and nonwatershed inputs are assumed to remain the same, although reduced human activities on land could reduce these as well; the values are not large enough to make a major difference in this calculation. This load equates to 2.3 g/m²/yr for TP; however, processes on the way to the reservoir and inside it currently limit the effective load to the OP load (0.2 g/m²/yr) plus 33% of the remaining particulate P load (0.7 g/m²/yr), or about 0.9 g/m²/yr. From empirical models, this suggests a “background” TP level in Lake Waco of 38 μg/L.

The expected background TP concentration for Lake Waco is therefore about 36% of that experienced by the reservoir prior to pool rise (38 vs. 104 μg/L). Applying the post-pool rise area and volume in the empirical models, which adjust for changed area, depth and detention time, the predicted in-lake TP value would be 43 μg/L under the current reservoir size. Ecoregional criteria have been established to represent expected conditions in the absence of major human influence, with values potentially applicable to Lake Waco ranging from 26 to 60 μg/L (TCEQ 2004, OKOSE 2004). The seemingly most appropriate value is 37 μg/L. Estimated background conditions for Lake Waco would therefore be expected to approximate corresponding ecoregional values.

Discussion

Overall state of Lake Waco

Although many factors can affect algal production, phosphorus is widely recognized as the most influential nutrient in freshwaters (Holdren et al. 2001, Kalff 2002), with elevated values fostering algal blooms and related water quality problems. Thresholds have long been recognized based on surveys of many lakes (NAS/NAE 1973, USEPA 1974, Wetzel 1983), with most researchers in agreement that values <10 μg/L rarely sustain enough algae to impair uses, while values >100 μg/L almost invariably cause elevated productivity and related use impairment. Work by Walker (2004) for Texas reservoirs supports this general framework for reference but notes issues with nonalgal turbidity weakening the link between P and algae.

Local and regional factors affect the progression from minimum to maximum effects, with many lakes showing signs of impairment at P levels >20 μg/L and relatively few lakes avoiding impairment with P levels >50 μg/L. For Texas reservoirs, where high nonalgal turbidity and low bioavailability of P is common, values in the range of 26–60 μg/L may be considered natural, with a value of 37 μg/L suggested for the ecoregion most appropriate to Lake Waco (TCEQ 2004, OKOSE 2004). Work by Kiesling et al. (2001) suggested that the appropriate range for Lake Waco was 15–50 μg/L, with 30 μg/L set as an appropriate target for management. Research by Smith et al. (2001) and Robertson et al. (2007) suggests high calcium loads can double acceptable P loads; TP levels between 20 and 40 μg/L might be tolerable in Lake Waco.

Elevated P concentrations are known to favor blue-green algae (cyanobacteria), which tend to become the dominant form of phytoplankton in lakes at phosphorus concentrations greater than about 50 μg/L (Watson et al. 1997). At phosphorus values above 100 μg/L cyanobacteria may represent nearly all of the phytoplankton biomass. Because cyanobacteria are a major cause of use impairment in many lakes, including Lake Waco, these observations are consistent with the phosphorus-impairment relationship previously discussed. The N:P ratio remains important, but N:P ratios are typically lower when P is high.

Nitrogen has also been evaluated by many researchers over time, with a resulting transitional impact range of roughly 0.30–2.0 mg/L. The form of N is important to its impact, and the ability of some cyanobacteria to fix dissolved nitrogen gas (Graham and Wilcox 2000) constrains the potential for N to limit overall algal production. However, the ratio of N to P remains very important in determining the types of algae that will be present. Given that many of the cyanobacteria favored by low N:P ratios (Smith 1983) are also taste and odor and/or toxin producers (Rashash et al. 1996, Chorus and Bartram 1999, Carmichael 2001), there may be concern over low N as well as high N, depending on P availability. Low N:P ratios would be most prevalent when P levels are high, reinforcing the observation by Watson et al. (1997) that increasing P leads to increasing cyanobacterial dominance.

The P and N concentrations in Lake Waco are excessive in comparison to commonly applied thresholds, yet it is important to consider the forms of N and P present when predicting impact on the lake. If the biological availability of N or P is low, it will affect the quantity and types of algae present.

Studies have suggested impairment of uses at chlorophyll a levels as low as 4 μg/L (Welch 1989). Current work by Walker (2004) for Texas reservoirs indicates impairment of recreational uses occurs at chlorophyll a levels of 10–20 μg/L. Impairment for water supply purposes is often observed at lower chlorophyll a levels, simply as a function of filter clogging, and is exacerbated by pH fluctuations, disinfection byproduct precursors, taste and odor, and toxins at higher chlorophyll a levels.

For the period covered by the loading analysis, Lake Waco chlorophyll a values exhibit respective means and medians of 15.6 and 14.2 μg/L in the main body of the reservoir (including near the intake), 20.2 and 19.1 μg/L in the NBR arm of the lake, and 16.4 and 14.0 μg/L in the SBR arm (City of Waco, unpubl. data). These fall into the range indicating impairment of recreational uses, and are certainly above desirable levels for a water supply. The range of individual values is wide, however, with many values <10 μg/L and some values in excess of 100 μg/L. A shift from assemblages with very few cyanobacteria seems to have occurred in the 1980s to one with much more of these algae since the 1990s (City of Waco, unpubl. data). Cyanobacteria are not consistently dominant, however; other factors besides N and P levels are affecting the algal assemblage to a significant degree in Lake Waco.

Amount and variation of nutrient loading

It is apparent from working with the Lake Waco system data that prediction of day to day, week to week, and month to month variation in conditions will be difficult without accurate weather forecasting. For the purpose of comparing loads from areas or source types, analysis of an extended time period is essential. Models constructed by researchers from TIAER (McFarland and Hauck 1999) and Baylor (White et al. 2010a) predict higher loads for the calibration periods applied, which were wetter than the longer term average. The importance of precipitation and runoff to loading in this system leads to those higher estimates, which may very well be correct for the periods represented. Most interesting is the relative magnitude of sources as derived from those estimates.

The NBR is the primary source of TP and OP to Lake Waco. The models indicate that NBR is also the main contributor of TN, while the results of this assessment suggest that MBR and NBR provide roughly equal amounts of TN. This balance may shift between wet and dry weather periods. Modeling of denitrification may also be insufficient in the models, or the sampling and testing may be missing key N fractions in NBR some of the time. Both this assessment and TIAER modeling suggest that actual inputs from WWTF accounted for about 4% of the TP load to Lake Waco, prior to any enhancement of the Stephenville and Clifton WWTF. The portion of the OP load to Lake Waco from WWTF was estimated at 7–13% in this assessment, while the TIAER analysis estimates the WWTF OP contribution at 9%. There seems to be good agreement that WWTF are a minor source of P to Lake Waco.

Dairy operations were estimated to contribute 23–28% of the TP and 46% of the OP entering Lake Waco in this loading analysis, while the TIAER analysis yielded estimates of 21% of the TP and 35% of the OP entering the lake attributable to dairy operations. The TIAER effort concentrated on WAF, which are undoubtedly the main source, but may have underestimated other dairy inputs. The export coefficients for dairy operations derived by TIAER are consistent with the range calculated here from the available data, but any small difference at these high input levels could affect the estimated load appreciably.

The Baylor model (White et al. 2010b) demonstrates that loading of organic matter and associated P is a driving force in the condition of Lake Waco, and it seems that the dairy operations are a major source of that organic P. The availability of inorganically bound P seems low in this system. Although sediment loading and resuspension in Lake Waco are substantial (Filstrup and Lind 2010), it seems that most of the load to the reservoir is never “active” in the production of planktonic algae. Loading from dairies is estimated to represent almost half the biologically available P in Lake Waco, yet dairies represent only about 25% of the TP input and <5% of the land use in the watershed.

All loading assessments predict a low TN:TP ratio in the lake, and this may be the most important finding with regard to N in Lake Waco. All the analyses suggest enough P loading to support major algal blooms, even with a low biological availability of those loads, but the low TN:TP ratio is a major factor in promoting cyanobacterial blooms.

Impact of loading on algae in Lake Waco

Water clarity is low in Lake Waco but not as low as suggested by the models. Because algal standing crop is also lower than expected, the low transparency cannot be a function of algal particles. Rather, other organic matter and inorganic sediments are responsible for the observed low clarity of Lake Waco, as described by Filstrup and Lind (2010). This also favors cyanobacteria, many of which are buoyant.

The influence of inputs with low TN:TP ratios is a major concern with regard to cyanobacterial blooms, and the loading analysis has shown that TN:TP ratios in the rivers are depressed by WWTF and dairy inputs. Inputs from WWTF are a minor and localized influence, while the dairy operations in the drainage area of NBR clearly depress the TN:TP ratio of the overall nutrient load to Lake Waco to a point that favors N-fixing cyanobacteria from late spring into autumn of many years.

Potential for improvement

The body of data and analysis compiled by TIAER (Flowers et al. 2001, Kiesling et al. 2001) makes a compelling case for reducing the P load to NBR by about 50%, and the TNRCC/TCEQ promulgated a TMDL (TNRCC 2001) and adopted an implementation plan (TCEQ 2002) based on that recommendation. It has been noted that even a 50% reduction in P loading to the upper NBR will not achieve a desirable P concentration in the upper NBR, but improvement would be expected and targets could be achieved in the lower NBR and in Lake Waco. This analysis suggests that the 50% reduction is a logical and appropriate initial target, to be revisited as progress is made and data are collected to evaluate system response.

The TIAER and TNRCC/TCEQ work was focused on OP, but the empirical models run as part of this analysis suggest that the reservoir responds as though it is getting an active P load equivalent to the OP load plus about one-third of the remaining particulate P load. Consequently, reducing the TP load by more than 50% may be necessary to achieve truly P-limited conditions at an algal production level considered appropriate for a drinking water supply and major recreational resource. However, the current N:P ratio situation suggests that every increment of P reduction achieved has the potential to shift that N:P ratio toward higher values and P limitation, potentially altering the composition of the phytoplankton before any appreciable decrease in actual productivity is attained. This could be beneficial to Lake Waco because cyanobacteria dominance seems to be causing the greatest problems for water supply and recreation. Therefore, while a P load reduction on the order of 50% is desirable, some benefit may result from lesser reductions, if P is reduced without any commensurate reduction in N load.

From empirical models, eliminating most agricultural and waste management sources in the watershed would yield an in-lake TP concentration of 43 μg/L at current pool level. This concentration is above the 30 μg/L level recommended as a target by Kiesling et al. (2001) but within the suggested range of 15–50 μg/L. However, if it is possible to emphasize reduction of OP and the biologically available fraction of the remaining particulate P in the management program, the beneficial effect on the lake could be substantially increased.

If a 50% reduction in P load is desired, a number of sources must be managed, but the choice of key target sources is limited. Given the importance of N:P ratios and the biological availability of P as well as the actual quantity of P loaded to the lake, large source contributors of OP or organic P at low N:P ratios, such as dairy operations and WWTF, are the key targets of control. The WWTF have attracted attention because both the regulatory framework and the technology to lower P outputs already exist and have been applied successfully elsewhere. Because WWTF serve a rate-paying public, it is economically possible to implement P load reductions for those WWTF, albeit potentially unpopular. Improvements made at the Stephenville and the Clifton WWTF have reduced the P input from all WWTF to Lake Waco by 34%. However, because WWTF represent a minor fraction of the TP and OP loads to the lake, the benefit of additional reductions from WWTF will be limited.

Dairy operations represent the largest contributor by source type. Inputs from dairy operations are diffuse in many instances, but these inputs have been declared point sources under the Clean Water Act and are subject to regulatory controls. Rules are in place to govern the use of WAF, the primary source of dairy P to the aquatic system, and half of the generated manure is supposed to be hauled out of the watershed. Recent data indicate that improved dairy waste management could be making some difference in the P concentration in NBR (City of Waco, unpubl. data), but it is too soon to tell if the improvement is real and attributable to dairy management.

The magnitude of potential P inputs from manure generated and retained in the watershed is high (much greater than the current estimate of loading to Lake Waco from all sources). The bioavailability of P in dairy wastes is greater than most other sources in this watershed. The low TN:TP ratio associated with dairy inputs favors blooms of objectionable algae. Consequently, dairy operations offer the greatest potential for making a difference in Lake Waco through watershed management. Based on all analyses to date, it is not possible to achieve the desired conditions in Lake Waco without reducing inputs from dairy operations, but it may be possible to detectably improve conditions by addressing only dairy-related inputs.

Acknowledgments

This assessment was made possible by all those contributing to the study of Lake Waco over the last 2 decades. The efforts of researchers at TIAER (at Tarleton State University) and Baylor University are especially acknowledged, along with their willingness to share data and insights despite controversies among watershed stakeholders and associated political pressures. The City of Waco is to be commended for undertaking the study; the foresight exemplified by City officials and willingness to expend the necessary funds is impressive. Careful review by Tom James helped condense a very large amount of information into a more concise paper.