A combined watershed–water quality modeling analysis of the Lake Waco reservoir: II. Watershed and reservoir management options and outcomes

Abstract

In this study, calibrated watershed and reservoir models are used to explore a range of possible watershed conditions and potential management options to reduce available nutrients and algal growth in the Lake Waco reservoir. The management options are divided between watershed and reservoir options. The watershed management options include wetland construction, manure haul-off, agriculture conversion to pasture, absolute nutrient retention in the watershed and control of urban nutrient run-off. For the reservoir, management options of phosphorus inactivation and increased algal consumption by grazers were evaluated. For all individual management scenarios, only complete conversion of agricultural lands into rangeland decreased nutrient levels and algae growth significantly and achieved target levels for chlorophyll-a and total phosphorus. Combined management scenarios including wetland construction, manure haul-off from dairy operations and increased in-reservoir herbivory could further reduce chlorophyll-a and nutrient values, but with less efficiency than agricultural conversion alone. The management option study showed that decreasing nutrient inputs and water clarity were important factors for controlling algal growth in Lake Waco, and that substantial reduction in total phosphorus is needed to achieve target conditions.

Key words:

• chl-a
• CE-Qual-W2
• Lake Waco
• management
• reservoir
• SWAT
• TP
• water quality
• watershed

In Texas, as in many other states, water availability and quality have declined associated with population increases and shifts in allocation (Wurbs et al. 1994). In addition, future changes in precipitation and temperature associated with climate change will compound water deficits and quality decreases in this region (Wurbs et al. 2005). Management of watersheds associated with municipal water reservoirs is challenging due to the need to balance economic, environmental and public health needs. Quantifying the trade-offs between these factors is rarely estimated due to insufficient data and the tendency to address human needs over environmental consequences with regard to water issues (Kindler 1998).

Integrated, systematic information related to the function of a watershed and reservoir is necessary to develop management options based on the strength of causal relationships. Managing watersheds for improved reservoir water quality must identify and quantify all potential sources of landscape pollutants to develop site-specific strategies (Tim and Jolly 1994, Carpenter et al. 1998). However, sources of impaired water quality are often diffuse with indirect interrelationships. For example, nonpoint pollution and water quality decline are associated (Carpenter et al. 1998); however, the mechanisms for their relationships are based on ecological processes that may have varying levels of predictability at various spatial and temporal scales (Hunsaker and Levine 1995). Models can be used to approximate and test hypotheses regarding mechanisms related to watershed-water quality relationships, especially if the models are based on first principles (Wing et al. 2002). Deterministic models of linked watersheds and reservoirs are best suited to provide an experimental design framework for testing “what-if” questions in these large, complex ecosystems (Arnold et al. 1987, Soyupak et al. 1997, Gunduz et al. 1998, Staudenrausch and Flügel 2001, Arnold and Fohrer 2005).

Challenges to water management in the Lake Waco watershed and reservoir include balancing the needs of agricultural interests with water supply for roughly 200,000 people in central Texas. Of concern is the increase in cyanobacteria in the reservoir over the past 2 decades that has resulted in taste and odor problems for municipal drinking water treatment and supply (Di Luzio et al. 2004, Rodriguez and Matlock 2008). In this study, management options are tested for the Lake Waco watershed using calibrated versions of the Soil and Water Assessment Tool (SWAT) for the watershed and CE-Qual-W2 model for the reservoir (White et al. 2010) to address algal production and nutrient concentrations in the reservoir. Management scenarios are assessed to evaluate impacts on key water quality characteristics including chlorophyll-a (chl-a), dissolved oxygen (DO), and ratios of total nitrogen (TN) and total phosphorus (TP). These indicators were chosen to best represent impact of algal growth and carbon metabolism and potential for cyanobacterial growth in the reservoir. The management scenarios include changes in land use, dairy practices, stream nutrient constituents, in-reservoir inactivation of phosphorus and planktivory in the reservoir. Confirmation of simulated water quality data with observed data are presented as a check on model accuracy. Management options are assessed individually and in combination based on evaluated efficacy for improving water quality metrics.

Methods

Model simulations for SWAT and CE-Qual-W2 (ver. 3.2) were developed using various scenarios for the Lake Waco reservoir and contributing watersheds. For all scenarios presented here, a baseline condition was established, which is the model system derived from the calibration parameters described in a previous study (White et al. 2010). The baseline scenario included all concentrated animal feeding operation (CAFO) lagoons, CAFO dairy treatment fields (DTF), small flood-control ponds (referred to as PL-566 reservoirs) and 8 municipal waste water treatment plants (WWTP) within the Lake Waco basin. The constructed Waco wetlands also were included in these simulations but not included in the original calibration because the calibration period was for 1997–1998 and the wetlands were constructed in 2001.

Parameter values necessary for SWAT simulations for the Waco wetland were derived from average sampled values from Scott et al. (2005). Maximum surface and volume of the wetland were set at 132.2 ha and 1 × 10⁴ m³, respectively. Sediment concentration was initialized at a value of 125 mg/L. Settling rates for phosphorus (P) and nitrogen (N) were set to 5.0 and 10.0 m/d, respectively. Finally, the wetland bottom hydraulic conductivity was set at 2.3 mm/hr based on observed values. These same parameters were used as default values for additional wetland scenarios described in the Methods section.

Meteorological data acquired from 18 National Climatic Data Center (NCDC) weather stations in the watershed for 1999–2004 were used for both models. However, the simulation analysis period was set to 2001–2004 to allow for a 2-year period for model equilibration and is referred to as the “baseline” in the study. Confirmation of model values was assessed by comparing reservoir values of DO, chl-a, TP and TN measured by the Texas Institute for Applied Environmental Research (TIAER) during 2001–2002. For the observed data, both TP and TN were reported; for the model, however, TN has been calculated as the sum of predicted total Kjedahl N plus NO₂-NO₃-N_. To assess accuracy, the root mean squared error (RMSE) was calculated for these constituents:

where n is the number of observed (o) and predicted (p) pairs. These values were reported and compared to ranges of water quality threshold values described later.

Scenarios were arranged into watershed and reservoir management options. Watershed management options affected setup and simulation outcomes from the SWAT model. For the reservoir management scenarios, the baseline SWAT simulation outputs were used for CE-Qual-W2 inputs. The watershed management scenarios included (1) addition of 3 new wetlands in the North Bosque River subwatershed (w2); (2) removal of 50% of liquid and solid waste from the North Bosque diary CAFO (man); (3) conversion of all agricultural land to rangeland in all drainages (ag-man); (4) 30% reduction in orthophosphorus concentration from the North Bosque River by unspecified means (p-man); and (5) removal of all TP and TN for the urban watershed directly adjacent to the reservoir (urb).

Simulations of the SWAT model were used as inputs for baseline simulations of the reservoir model, CE-Qual-W2. For the wetland addition simulation (w2), 3 smaller wetlands were added at the towns of Stephenville, Hico and Clifton in the North Bosque watershed. The wetlands were one-third the size of the Waco wetlands and affected simulated flow and nutrient contribution from the North Bosque River. In the dairy management scenario (man), 50% reduction in waste application to DTF in the watershed was assumed. This is conceptually similar to the scenario presented by Santhi et al. (2001) in which they simulated 100% removal of dairy manure (liquid and solid) from the system. The conversion of all agricultural land to pasture (ag-man) was accomplished through simple reclassification of the land cover–landuse data used in the SWAT model. Phosphorus reduction (p-man) was accomplished by simply reducing PO₄-P by 30% from the simulated baseline output files from SWAT for the North Bosque watershed. The method for this removal was unspecified but represents a target value considered reasonable based on a mass–balance approach of past nutrient inflows into the reservoir. It addresses only dissolved P, however, not the potentially large pool of organic P carried by tributaries to Lake Waco. Finally, the urban watershed located along the southeastern shore of Lake Waco reservoir is a potential site for nutrient additions, equating to approximately 2% of total inputs for these constituents. Although not a large portion of the total load, nutrients from this area enter the reservoir quickly, and their control represents a management option under direct control of the City of Waco, based on jurisdiction.

The reservoir management scenarios required changes in CE-Qual-W2 parameters while using the baseline simulation inputs from the SWAT model. The reservoir management scenarios included (1) dynamic inactivation of phosphorus through treatment of elevated inflows (p-inact) and (2) increased phytoplanktivory (plank). The inactivation of phosphorus assumes treatment in which 90% inactivation of PO₄-P is achieved by addition of an inactivating agent (typically aluminum or calcium; Cooke et al. 2005). In the simulation, PO₄-P concentration values were reduced by 90% when inflow rates exceeded 15 cm³/s. Note that organic P entering the lake was not assumed to be reduced by this treatment. Increased grazing activity was assumed to be induced by increased numbers of zooplankton, brought on by reduction in the number of filter feeding fish in the reservoir. Grazing was considered to be nonselective, represented in the CE-Qual-W2 model by increasing daily mortality rates by 10% for all algal groups. For combined management effects, several selected scenarios were merged or modified and re-run through the models based on the results of the initial, individualistic simulations and whether management options were considered feasible.

Response of the reservoir to management scenarios was analyzed through assessment of predicted chl-a, TP and TN:TP, and threshold criteria were identified as targets for best management to control algal productivity and cyanobacteria prominence in the reservoir (K. Wagner, ENSR/ACECOM, June 2005, pers. comm.). These target values for the reservoir included (1) chl-a concentrations <20 μg/L, (2) TP concentrations <0.05 mg/L and (3) TN:TP ratio >10. Improvement from management actions was assessed by calculating the ratio of the number of days the simulated results were less than the target values divided by the total simulation days between 1 January 2001 and 31 December 2004 and presented as a percentage. The simulation data utilized for this analysis are from the major water withdrawal location for the City of Waco's water treatment plant in the main body of the reservoir.

Results

Inflow analysis

Simulated watershed inputs were assessed by first calculating the annual averages for all scenarios. The range of simulated values varies substantially for the baseline loading among years, with an interannual range of 0.80 × 10⁵ to 1.50 × 10⁵ kg TP/yr (Fig. 1a). For TN, the baseline scenario had TN loading ranging from 5.15 × 10⁵ to about 1.4 × 10⁶ kg TN/yr (Fig. 1b). Using the baseline simulation as a “control,” statistical analysis of nutrient loading under any scenario using the nonparametric Mann-Whitney U-test could be applied. This testing showed that only the projected natural condition and the ag-man scenario had significantly lower TP and TN values for the simulation period (P < 0.05). Loading values from the ag-man scenario ranged from 0.43 × 10⁵ to 0.63 × 10⁵ kg TP/yr and 1.50 × 10⁵ to 3.81 × 10⁵ kg TN/yr.

Figure 1 Phosphorus and nitrogen predicted by the SWAT model for the inflows into Lake Waco. The scenarios were determined ad hoc as possible alternative watershed conditions.

Reservoir simulation assessment

Before the impact assessment, water quality values predicted by CE-Qual-W2 were compared with observed values collected by TIAER at Tarleton State University, Stephenville, Texas, for the Lake Waco reservoir from 2001–2002. Surface values of dissolved oxygen (at 1 m depth) showed high correlation between simulated and observed values for this period (Fig. 2) with the calculated RMSE value of 0.88 mg/L. Both simulated and observed data show distinct seasonality in surface DO values with peak values occurring during winter months and lows occurring during autumn months. The average predicted DO value for the 2001–2004 period was 8.4 mg/Lcompared to the observed reservoir average of 8.7 mg/L. The reservoir showed no long-term stratification or anoxia with daily simulated summer DO values, consistent with observed conditions. Lake Waco is considered a polymictic reservoir, and the model showed a similar hydrodynamic regime represented by these simulations.

Figure 2 Dissolved oxygen values (mg/L) predicted by the CE-Qual-W2 model (baseline) compared with observed data from TIAER. Averages and standard deviation of surface values (at 1 m depth) are shown for Lake Waco reservoir.

Observed values of chl-a exhibit acceptable correlation in terms of timing of peaks and magnitude of values (Fig. 3) with an RMSE value of 8.3 μg/L. This error is less than half the threshold value set for the management scenarios. The average predicted and observed data also compare favorably with values of 19.8 and 19.9 μg/L (Table 1), respectively. As with DO, chl-a dynamics also have distinct seasonality with peak values occurring between spring and autumn. Note that some disparity exists between modeled and observed values, such as during October of 2001 and 2002.

Table 1 Summary data for the CE-Qual-W2 scenario simulations. Percentages of simulation days that achieve target values are shown. These target values for the reservoir included (1) chlorophyll-α concentrations <20 μg/L, (2) TP levels <0.05 mg/L, and (3) TN:TP ratio >10. The * indicates a significant different value (P < 0.05) for a scenario as compared with the baseline simulations based on the Mann-Whitney U-test.

	Mean Values and Percentage (%) of Days that Achieve Listed Target Values
	Chl- a (20 μ g/L)		TP (0.05 mg/L)		TN:TP ( > 10)
Scenarios	μ g/L	%	mg/L	%	mg/L	%
Observed	19.9	49.0	0.12	5.9	9.5	35.2
Baseline (Predicted without management)	19.8	58.5	0.13	5.4	6.8	2.9
Phosphorus inactivation treatment (p-inact)	12.3	89.0	0.08	9.5	9.4	24.7
Increased phytoplanktivory (plank)	13.0	81.3	0.11	5.4	6.7	3.1
Removal of all phosphorus and nitrogen for the urban watershed within Waco (urb)	14.1	77.6	0.11	5.4	6.8	3.1
Addition of new wetlands (w2)	15.9	76.1	0.11	5.4	7.0	8.0
Removal of 50% manure from the North Bosque diaries (man)	13.5	80.0	0.09	7.1	7.5	6.7
Conversion of all agricultural land to rangeland (ag-man)	6.9*	100.0	0.03	95.6	8.6	16.8
Reduction in orthophosphorus concentration in the North Bosque by 30% (p-man)	14.1	77.2	0.10	7.4	7.7	8.8
p-man+plank	11.0	92.4	0.08	9.1	9.2	23.4
p-man+plank+man	10.1	98.2	0.07	15.5	9.6	35.7

Figure 3 Chlorophyll-a values (μg/L) predicted by the CE-Qual-W2 model (baseline) compared with observed data from TIAER. Averages and standard deviation of surface values (at 1 m depth) are shown for Lake Waco reservoir.

The analysis of modeled and observed TP values showed acceptable temporal correlation (Fig. 4). The RMSE of these data was 0.03 mg/L, a relatively low value. Analysis of the surface TP values measured by TIAER showed high spatial variability indicated by the size of error bars (Fig. 4). Analysis of average values showed that the model slightly overestimated TP for 2001–2002 with values of 0.13 mg/L compared to observed values of 0.12 mg/L.

Figure 4 Total phosphorus values (mg/L) predicted by the CE-Qual-W2 model (baseline) compared with observed data from TIAER. Averages and standard deviation for the TIAER observed data are shown.

Observed values demonstrated pulses of nitrogen in the reservoir with large amplitudes, which compared favorably with predicted values (Fig. 5). The RMSE of these data was 0.34 mg/L. Predicted values from the CE-Qual-W2 model showed seasonal fluctuations similar to observed values, though the overall the magnitudes were lower., With considerable spatial variation in TN across the reservoir, however, the predicted values represent a reasonable reservoir average. In contrast to TP values, average modeled TN values were lower than observed values with values of 0.81 mg/L as compared to 0.97 mg/L. This difference in overestimation and underestimation by the model for TP and TN, respectively, resulted in lower predicted TN:TP ratios. The predicted average TN:TP for 2001–2002 was 6.8 compared to an observed value of 9.5 (Table 1).

Figure 5 Total nitrogen values (mg/L) predicted by the CE-Qual-W2 model (baseline) compared with observed data from TIAER. Averages and standard deviation for the TIAER observed data are shown.

Reservoir scenario impacts

Using inputs from the SWAT simulations coupled with parameters established for the CE-Qual-W2 model, the scenarios run for 2001–2004 generate the percentage of simulation days that meet previously established reservoir water quality criteria (Table 1). The observed data showed targets for chl-a, TP and TN:TP were achieved 56.6, 48.0 and 65.8% of the time, respectively, from observed data compared with respective baseline simulation values of 65.8, 5.4 and 72.0%. Simulations of the various scenarios with the CE-Qual-W2 model indicated very little difference when compared against the baseline values except for the ag-man scenario, which had significantly different annual average chl-a values based on the Mann-Whitney U-test (P < 0.05; Table 1).

Most scenarios showed a high number of days with chl-a less than the target value of 20 μg/L (Table 1). However, most scenarios had few days with TP values <0.05 mg/L or TN:TP values >10. For the ag-man scenario, 100% of days had chl-a values below the 20 μg/L threshold due to drastically reduced inflow of P and N under this scenario. However, the p-inact scenario only achieved 9.5% of simulations for 0.05 mg/L target TP values as compared with ag-man with >95% of simulations days meeting the TP target. Although both the ag-man and p-inact scenarios showed an increased TN:TP value, neither achieved the target for >25% of simulation days.

Combined management scenarios did improve water quality measures somewhat. Combined scenarios including p-inact and plank scenarios increased the number of simulation days meeting targets for chl-a, TP and TN:TP changed to 92.4, 9.1 and 23.4%, respectively. Also, simulation including combination of the man, p-inact, and plank management scenarios yielded even better results, with 98.2, 15.5 35.7% of simulation days meeting target values for chl-a, TP and TN:TP, respectively.

Discussion

Model performance

Integrated models of watershed and reservoir processes as shown in this study can be accurately used to represent potential land use activity impacts on specific aquatic responses (Arnold et al. 1987, Summer et al. 1990, Flowers et al. 2001, Xu et al. 2007). Process models such as those used here provide users mostly physically based parameters that are somewhat intuitive to managers and researchers. However, process models can be complex and have specific scales of applicability. Scale is particularly important as ecological processes in watersheds and reservoirs happen and are modeled at different spatial and temporal scales. For example, inflow of water and nutrients predicted by SWAT occur at a daily timestep whereas in CE-Qual-W2, movement of water and nutrients is modeled at minutes or seconds. These discontinuities are implicitly recognized in previous integrated modeling work by Flowers et al. (2001) and Xu et al. (2007) where investigators found difficulty in simultaneous calibration of both watershed and reservoir models. One solution to this scale issue was generated by Debele et al. (2006) where subhourly predictions from SWAT watershed predictions were derived before input into the CE-Qual-W2 reservoir simulations. No attempt was made to account for these model timing discrepancies due to computation limits; however, elevated RMSE values for modeled constituents are likely influenced by these scale discontinuity issues.

Confirmation of predicted water quality parameters showed that the models generated fairly accurate results when compared to a separate dataset not used in the original calibration with RMSE values mostly less than half of the range of the stipulated management thresholds. Confirmation is a separate step of model development and is essential for establishing confidence regarding expected success or failure from management actions. For example, stratification of the reservoir is a concern of managers because it could exacerbate current water quality problems. While the model underestimated observed values, especially during autumn months of 2002, no significant summer stratification as indicated by DO values was identified over continuous periods, consistent with observed data (Fig. 3). Differences between observed and predicted chl-a are associated with parameters used to convert predicted biomass to chl-a for the 4 different algal groups represented in the reservoir model. In CE-Qual-W2, this ratio is static per algal group, though in reality it is quite dynamic seasonally (Felip and Catalan 2000). Simpler models of nutrient uptake, retention, and release would likely be insufficient to characterize this polymictic reservoir due to fluctuating hydroperiod associated with nutrient inflows. This creates a dynamic resource environment in which nutrient cycling is best represented by a continuous succession of phytoplankton types of varying sizes, phytopigment composition, light response and nutrient stoichiometry.

Total phosphorus concentration varies daily within waterbodies as a function of hydrodynamics and environmental factors controlling organic matter decomposition, algal uptake, substrate adsorption and settling (Baldwin et al. 2003). Grab samples capture a snapshot of the total phosphorus levels at a given point in the day while the CE-Qual-W2 model represents the average values for a larger temporal and spatial unit; therefore, disparity between predicted and observed values is expected for this particular constituent. Bias in predicted and observed TP were found mostly under low flow conditions (Fig. 4), when TP values approached the laboratory detection limit. The model adequately represents the variability observed in the reservoir, and the difference in the predicted and observed TP indicates a potential knowledge gap where TP levels in the reservoir may be higher than observed due to low sampling intensity.

Nitrogen fixation is not directly modeled in this version of CE-Qual-W2. It is approximated by shifting affinity for nitrogen sources between NH₄ and NO₂-NO₃, allowing a cyanobacterial group to be more competitive under low nitrogen environments. However, no nitrogen is added to the water from a mass balance basis. Significant nitrogen fixation occurs in the Lake Waco reservoir at the inlet of the North Bosque River related to high phosphorus inflows and temperature (Scott et al. 2008, Doyle et al. 2010). A TN:TP ratio target was set to discourage conditions that favored N₂-fixing cyanobacteria because these are linked to taste and odor episodes. Modeling these dynamics spatially is challenging, and CE-Qual-W2 represents with some accuracy the timing and magnitude of nutrient fluxes at a timescale relevant for management purposes, perhaps weekly to monthly.

Management implications

The results of this study show that the watershed exerts considerable influence on the reservoir as demonstrated through the scenario outcomes. Predominance of agricultural activity within the watershed, primarily in the North Bosque River drainage area, is directly related to algal production and nutrient levels within the Lake Waco reservoir, consistent with previous research (Rodriguez and Matlock 2008). Reduction of the nutrients associated with the liquid storage lagoons associated with dairy farms decreased loading of TP and TN by approximately 5% (Fig. 1a and 1b). However, the dominance of agriculture in the watershed makes agriculture the most significant anthropogenic source of nutrients to Lake Waco.

Manure composting and haul-off is a current part of an agricultural management plan for North Bosque watershed dairies, with some observed reductions in phosphorus loading (Bekele et al. 2006). Combined effects of manure reduction, increased planktivory and phosphorus inactivation may produce measurable water quality improvement for the reservoir, but the results of scenario modeling suggest that this will not be sufficient to achieve water quality targets.

The modeling system presented here is complex, though blunt in terms of representing reality; however, it is more likely that existing loading of nutrients is simply too high to observe any effect of management until a threshold is reached, and most management scenarios tested fell short of sufficient reductions in nutrient loading. Lake Waco is likely to be fairly fertile as a function of impounding water from a large watershed, but baseline loading is substantially higher than projected “natural” loading (Fig. 1a and 1b). While the results lack high sensitivity, the trends are important and likely represent the range of expected responses to the external and internal drivers of reservoir function, consistent with findings by Jeppesen et al. (2005). In particular, it will take a major reduction in TP loading to meet TP and TN:TP targets.

Lesser reduction is necessary to meet the chl-a target, probably as a function of current light limitation from suspended sediments (Filstrup and Lind 2010). Current conditions, either from actual data or the baseline model results, indicate that the chl-a target is reached about half the time now, compared with <6% of the time for TP and <35% of the time for TN:TP (Table 1). Management of chl-a in the reservoir was affected by most scenarios, with ag-man and p-inact treatments providing distinctly more desirable results among single scenarios (Table 1). Differential reduction of TP yields lower TN:TP ratios, although the number of days reaching the threshold values was never more than one-quarter of the simulation time. This system may be naturally susceptible to low TN:TP ratios, necessitating control of cyanobacterial blooms by maintaining low TP or limiting light.

Although individually the ag-man scenario provided desirable results, it should be noted that soil retention is higher for natural rangelands than for agricultural operations. Implementation of ag-man could result in less sediment load to the reservoir with potentially greater water clarity for inflow water (Fig. 6). The effect of clearer water on production per unit of nutrients available may not be sufficiently modeled in this exercise, and productivity decreases may not be as great as projected if the sediment load declines. Management focus on reducing nutrient inflow to nutrient limitation levels while maintaining current turbidity levels until will control algal growth. However, the feasibility of implementing the complete ag-man scenario is low; it is unrealistic to expect that all agriculture in the watershed would be converted to pasture or native grasslands. Additionally, erosion of river banks is substantial in many areas, and sediment sources may not be diminished appreciably with reduced agriculture in this case.

Figure 6 Annual sediment loading (metric tonnes) predicted by the SWAT model for Lake Waco between 2001 and 2004 for baseline and the agricultural land conversion (ag-man) scenario.

The technology for implementing the p-inact scenario requires development. This inactivation system, where aluminum or calcium is added to the North Bosque River inflow at times of elevated stream flow, would require a large dosing system, chemical storage, detection of increasing flows and monitoring of water chemistry to avoid toxic or other unintended impacts. Yet such systems are in place at smaller scales in other systems to counteract stormwater inputs (K. Wagner, ENSR/AECOM, April 2010, pers. comm.).

Results from the plank scenario also showed reduction in algal biomass (Table 1). In this scenario, it was assumed that a grazing increase of 10% could be achieved and sustained. Results of analysis of habitat shifts associated with the 2003 pool-rise of the Lake Waco reservoir (White et al. 2006) projected an increase in bluegill sunfish with a decline in largemouth bass, the reverse of what would be desirable for greater zooplankton grazing. The presence of gizzard shad in Lake Waco (Tibbs and Baird 2004) will also make enhanced grazing difficult to implement; however, a stocking program with hybrid striped bass has been initiated at Lake Waco and may reduce enhanced grazing.

A major goal of management is to reduce TP loading to Lake Waco, thereby decreasing the TN:TP ratio and reducing chl-a, especially in cyanobacteria. A major effort will be necessary to achieve the level of reduction needed. At the same time, loss of capacity in the PL-566 reservoirs could raise TP inputs well above current levels (Fig. 1a); management to preserve that capacity is potentially as important as management to control existing TP sources.

Acknowledgments

We thank the City of Waco, particularly Mr. Tom Conry, for his assistance for providing the funding for this study through ENSR. We thank Dr. Ken Wagner from ENSR/AECOM for his leadership in managing this project in addition to his early comments on the design of this study.