USDA Lake Waco-Bosque River Initiative Strategies for Monitoring Nonpoint Source Runoff Anne McFarland and Larry Hauck TR0115 December 2001 TIAER Texas Institute for Applied Environmental Research Tarleton State University • Box T0410 • Tarleton Station • Stephenville, Texas 76402 254.968.9567 • Fax 254.968.9568 [email protected] Strategies for Monitoring Nonpoint Source Runoff Acknowledgments This study was supported financially by the United State Department of Agriculture Lake Waco-Bosque River Initiative with additional funding provided by the State of Texas. The authors wish to acknowledge the support of landowners who allowed access to their land for stream monitoring sites. The authors also wish to acknowledge the dedicated work of the many field personnel and laboratory chemists involved with TIAER’s water quality monitoring program. Particularly with regards to storm water monitoring as rain often falls on weekends requiring staff to be on call seven days a week. Mention of trade names or commercial products does not constitute their endorsement. For more information about this document or any other document TIAER produces, send email to [email protected]. Authors Anne McFarland, Research Scientist, TIAER, [email protected] Larry Hauck, Assistant Director of Environmental Sciences, TIAER, [email protected] USDA02.09 2 Strategies for Monitoring Nonpoint Source Runoff Abstract Water quality monitoring in most cases is labor intensive and expensive. Careful planning is needed to make sure monitoring efforts are designed to provide appropriate information with the level of confidence needed to address local watershed-planning goals and objectives. Monitoring nonpoint source pollution adds an extra level of complexity to most monitoring programs with the need to measure storm water as well as base flow conditions. In developing any type of monitoring plan, there are six primary points to consider: (1) what to monitor, (2) where to monitor, (3) what type of monitoring to implement, (4) how often to monitor, (5) how long to monitor, and (6) what resources are available for monitoring. The strategies used for monitoring generally consider a balancing of cost versus necessary confidence in the data collected. Unnecessary monitoring is wasteful; while insufficient monitoring may lead to false conclusions or an inability to meet desired monitoring objectives. The purpose of this report is to provide an overview of considerations for stream monitoring with emphasis on monitoring in areas where agricultural nonpoint source pollution is prominent. Examples are drawn from TIAER’s extensive monitoring program of stream sites within the North Bosque River watershed. 3 Strategies for Monitoring Nonpoint Source Runoff 4 Contents Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Chapter 2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Chapter 3 What to Monitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Defining Water Quality Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Monitoring and Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Concentrations and Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Chapter 4 Where to Monitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Spatial Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Determining Sources of Pollution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Evaluating Effectiveness of Control Practices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Surrogate Sampling Locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Location within a Water Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Chapter 5 What Type of Monitoring to Implement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Chapter 6 How Often to Monitor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Average Annual Concentrations from Routine Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Annual Average Concentrations for Storm Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Annual Loadings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Integration Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Regression Approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Results Comparing Loading Estimation Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Chapter 7 How Long to Monitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Chapter 8 What Resources are Available for Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Chapter 9 Summary & Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Appendix A Sampling Sites, Weather, and Monitoring Protocols . . . . . . . . . . . . . . . . . . . 63 Sampling Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Precipitation and Hydrologic Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Data Collection Methods and Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 Appendix B Average Daily Flow at Sampling Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 Appendix C Impact of Sampling Frequency on Annual Average Concentration . . . . . . 73 Appendix D Regression Estimation of Concentrations and Loadings . . . . . . . . . . . . . . . 77 Appendix E Comparison of Annual Loading Estimation Methods . . . . . . . . . . . . . . . . . . 85 5 Strategies for Monitoring Nonpoint Source Runoff 6 Tables Table 1 Characteristics of seven selected monitoring sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Table 2 Comparison of water quality for impacted and least impacted watersheds . . . . . . . . . . . . 24 Table 3 Annual average percent error in PO4-P loading estimates compared to “true” loadings . 45 Table 4 Annual average percent error in TP loading estimates compared to “true” loadings . . . . 45 Table 5 Annual average percent error in TSS loading estimates compared to “true” loadings . . . . 46 Table 6 Minimal detectable change in storm water quality with varying years of monitoring . . . . 48 7 Strategies for Monitoring Nonpoint Source Runoff 8 Figures Figure 1 North Bosque River watershed and selected sampling sites. . . . . . . . . . . . . . . . . . . . . . . . . . 14 Figure 2 Annual average chlorophyll-a as a function of annual average time-weighted PO4-P . . . . 18 Figure 3 Phosphorus concentrations, stream volume, and loadings along the North Bosque River. 20 Figure 4 Cumulative daily differences in PO4-P and TP loadings and discharge volume for 1998. . 23 Figure 5 Variation in constituent concentrations with discharge at North Bosque River . . . . . . . . . . 30 Figure 6 Variation in NO2-N+NO3-N concentrations over time during six storm events . . . . . . . . . 31 Figure 7 NO2-N+NO3-N concentration during selected storm events in relation to discharge . . . . . 32 Figure 8 TSS concentration during selected storm events in relation to discharge at site BO070. . . . 32 Figure 9 Variation in TSS concentrations over time during six storm events . . . . . . . . . . . . . . . . . . . . 33 Figure 10 Comparison of flow-weighted and time-weighted averaging of storm samples.. . . . . . . . . 35 Figure 11 Impact of sampling interval on average annual concentration at site BO070.. . . . . . . . . . . . 38 Figure 12 Estimates of annual storm event mean concentrations based on subsampling. . . . . . . . . . . 40 Figure 13 Number of storm events needed to obtain certain allowable percent errors in EMCs. . . . . 41 9 Strategies for Monitoring Nonpoint Source Runoff 10 CHAPTER 1 Introduction The purpose of this report is to provide insight and information for stream monitoring as part of a watershed planning process that must consider agricultural nonpoint source pollution. The information discussed herein does not directly address monitoring for broad state and national programmatic needs that involve numerous river basins, such as, monitoring required under sections 305(b) and 303(d) of the Clean Water Act to determine the water quality conditions of all water bodies of a state and to list impaired water bodies. Rather, the monitoring addressed herein focuses on individual watersheds where monitoring is needed in response to existing water quality impairment or is being proactively pursued as part of a watershed management plan to prevent a threatened water body from becoming impaired. Successful watershed management approaches will require environmental monitoring. Monitoring is a linchpin of the Planned Intervention Microwatershed Approach (PIMA), an approach to watershed management being refined and demonstrated under the USDANatural Resources Conservation Service funded Agriculture and the Environment: Lake Waco-Bosque River Watershed Project (USDA Initiative). PIMA is an assessment-based, community-led, and performance driven approach to ameliorate agricultural runoff problems. Other watershed management approaches will contain similar characteristics, but the point is that there is a reliance on monitoring to support the assessment-based and performance-driven aspects of this and similar approaches. Assessment monitoring is needed to define the water quality problem, determine the sources of the problem and their geographic extent, provide geographic focus on hot spots, and specify allowable pollutant loadings by source and geographic area. Performance-driven implies that success is measured not only in the implementation of agricultural best management practices, but ultimately in improvement of water quality in receiving waters. This report addresses several relevant issues concerning strategies to monitor agricultural nonpoint source pollution. In some instances monitoring, by itself, may provide the information needed to meet objectives, whereas in other instances monitoring may be in a support role, for example, providing data for the application of computer models. Some pertinent objectives that can be addressed and supported by monitoring include the following: • Identifying water quality problems—Is there impairment and if so, what pollutants are of concern? What are current loads or levels of pollutants? • Defining discrepancies in current conditions compared to desired conditions—What loads or levels of pollutants are desired or allowable to overcome impairment, and how do current conditions compare to desired conditions? • Determining sources of pollution—What sources are contributing pollutants and how much? In other words, where is the impairment coming from? 11 Strategies for Monitoring Nonpoint Source Runoff • Making waste load allocations—How should pollutant loads be allocated between sources to meet desired loads or concentrations? (Computer models are often used to meet this objective in conjunction with watershed specific monitoring data.) • Recommending the implementation of pollution controls—What controls should be implemented to reduce impairment from different sources? (Monitoring may also supplement the application of computer models in meeting this objective.) • Evaluating effectiveness of implemented control practices—Are load allocations being met with implemented control practices? If not, how effective are implemented control practices and what more may be needed? • Assessing improvement—Is water quality getting closer to the desired level over time? Agricultural nonpoint source pollution to surface water bodies is stochastic in nature, because its occurrence is determined by rainfall-runoff events. A successful nonpoint source monitoring plan must meet the challenges presented by the stochastic nature of the source and may need to address several different temporal (e.g., within storm variations of water quality or annual averages) and spatial (e.g., major rivers, first- or second-order streams, or even edge-of-field) scales. Based on the above objectives, there are six primary points to consider in developing a monitoring plan. • What to monitor? • Where to monitor? • What type of monitoring to implement? • How often to monitor? • How long to monitor? • What resources are available for monitoring? This report will be organized around these six points, which when considered in aggregate, define how a monitoring plan will obtain observational data that meets desired objectives. In many instances the general discussion will be embellished through specific examples taken from monitoring efforts in the North Bosque River watershed under the USDA Initiative. The streams and rivers in the North Bosque River watershed are characteristically intermittent in flow with agricultural nonpoint source contributions dominating storm water quality conditions. Despite the dominance of agricultural contributions during storm events, the North Bosque River is strongly influenced by effluent from municipal wastewater treatment plants at low flow conditions, which prevail during and following the typical hot, dry Texas summer. 12 CHAPTER 2 Background To provide the reader a proper setting for examples using data from the North Bosque River watershed, a brief overview is provided of the watershed and selected monitoring sites. More detail on monitoring within the North Bosque River watershed may be found in Appendix A and other documents on the monitoring efforts of the USDA Initiative (e.g., Easterling, 2000; Kiesling et al., 2001; McFarland and Hauck, 1999; McFarland et al., 2001). The North Bosque River watershed of north central Texas is typical of many watersheds in the region. The dominant land covers are rangeland and woods. Improved pasture and some row crop farming are found to some degree throughout the watershed, though row crop is most common in the southern portions of the watershed, particularly in the floodplain of the North Bosque River close to the city of Clifton, Texas (Figure 1). The headwaters of the North Bosque River originate in Erath County and contain a large number of dairy operations and the watershed’s largest city, Stephenville. Within the headwaters, fields receiving solid and liquid dairy manure are a significant land use, and henceforth these fields will be referred to as dairy waste application fields. The North Bosque River has been on every State of Texas section 303(d) list since 1992 for water quality impairments, of which the most relevant impairments are excessive nutrient and aquatic plant concentrations. Generally, the section 303(d) listing has attributed the source of nutrients and the resulting over abundance of aquatic plants as originating from animal feeding operations, and more specifically, dairy operations. Research for the USDA Initiative has determined that phosphorus is the nutrient that controls (or limits) aquatic plant growth in the North Bosque River (Kiesling et al., 2001). Dairy operations and municipal wastewater, in that order, are the major controllable sources of phosphorus to the river (McFarland and Hauck, 1999). Because the greatest density of dairy cows and the largest municipal wastewater treatment plant discharge, which is from the city of Stephenville, are in the headwaters of the North Bosque River, an anthropogenically driven nutrient gradient exists along the North Bosque River with highest nutrient concentrations in the headwaters and decreasing nutrient concentrations in the downstream direction. Climatically, annual rainfall averages a little over 76 cm (30 in) per year across the watershed. Rainfall typically follows a slightly bimodal pattern with peaks in the spring and fall. On average the wettest month is May and the driest month is January. Most tributaries to the North Bosque River are highly intermittent and frequently become dry soon after each rainfall runoff event. Winter rains corresponding with the low evapotranspiration of that time of year can, however, establish a base flow that persists well into the spring in some years. Ground water does not contribute significantly to a base flow component in either the North Bosque River or its tributaries, especially in the upper half of the watershed. Hence, the North Bosque River typically becomes dominated by municipal wastewater treatment plant effluent in the summer and early fall. 13 Strategies for Monitoring Nonpoint Source Runoff Figure 1 North Bosque River watershed and selected sampling sites. From the network of monitoring sites operated for the USDA Initiative, seven sites were selected to represent a variety of drainage area sizes and differences in land covers (Figure 1 and Table 1). The monitoring at each site consisted of storm event sampling performed with automatic sampling equipment and biweekly (once every two weeks) manual grab sampling taken whenever flow was present at a site. Samples collected at each site were analyzed for a number of water quality constituents; however, the constituents presented in this report focus specifically on nitrite-nitrogen plus nitrate-nitrogen (NO2-N+NO3-N), orthophosphate phosphorus (PO4-P; often referred to as soluble reactive phosphorus), total phosphorus (TP; comprised of both soluble and particulate phosphorus forms), and total suspended solids (TSS). These four water quality parameters were selected because they include two water soluble (dissolved) constituents (NO2-N+NO3-N and PO4-P), a water suspended constituent (TSS), and a constituent that contains suspended and soluble components (TP). The seven sites used in this report not only have a wide variation in drainage area size and land cover, but also water quality response. Some key characteristics of each site are provided in Table 1. 14 Chapter 2 Background Table 1 Characteristics of seven selected monitoring sites. (Source: Adams and McFarland, 2001) Drainage Area (ha) Dominate Land Use/Land Cover Water Quality and Hydrology SF020, South Fork of N. Bosque River 848 Wood/range (96%) Least impacted small watershed site; highly intermittent flow NF020, North Fork of N. Bosque River 791 Dairy waste application fields (45%), wood/range (39%) Water quality highly impacted by nonpoint sources; highly intermittent flow NC060, Neils Cr. 35,147 Wood/range (77%), pasture and cropland (23%) Ecoregion reference site (or least impacted site); intermittent flow GC100, Green Cr. . 26,165 Wood/range (71%), dairy waste application fields (7%) Water quality moderately impacted by nonpoint sources; intermittent flow BO040, N. Bosque River below Stephenville, TX 25,719 Wood/range (51%), dairy waste application fields (12%) Water quality impacted point and nonpoint sources; perennial flow BO070, N. Bosque River at Hico, TX 93,248 Wood/range (68%), dairy waste application fields (7%) Water quality moderately impacted by point and nonpoint sources; nearly perennial flow BO090, N. Bosque River near Clifton, TX 253,740 Wood/range (72%), pasture and cropland (23%) Low impacts from point and nonpoint sources; perennial flow Site/Location The intermittent flow patterns observed in the North Bosque River are typical of patterns observed throughout portions of the central and western United States, and therefore this system provides a relevant example for many watersheds. As a caution, when strong surface water and ground water interactions occur and the base flow component comprises a significant portion of total flow, this difference in hydrology from the North Bosque River watershed must be considered in light of comments and recommendations within this report, which are largely determined from a storm water driven surface water system. 15 Strategies for Monitoring Nonpoint Source Runoff 16 CHAPTER 3 What to Monitor General Considerations The Clean Water Act refers to maintaining the physical, chemical, and biological integrity of our Nation’s waters. For monitoring, this means there are a multitude of constituents that could be monitored. The key is to determine which constituents are of primary importance in a given watershed. Often there is a specific concern that drives water quality monitoring and defines what constituents should be analyzed. For point source discharges from industries or wastewater treatment plant facilities, specific chemicals must be monitored in the effluent discharged. A National Pollutant Discharge Elimination System (NPDES) permit will specify what must be monitored. If there is a question as to whether or not a point source is meeting its permitted discharge limit, the answer to the question of what to monitor and where is quite simple—at the end of pipe, for the permit specified constituents. Compliance monitoring is appropriate to this situation. With regard to nonpoint source pollution, what constituents to monitor may not be as clear. Baseline monitoring of a board spectrum of constituents may be needed to establish the general health or conditions of a stream. Baseline monitoring should consider not only chemical and physical constituents, but a general evaluation of the biological health of a stream including habitat assessment and monitoring of benthic invertebrates, fish, and algae populations (e.g., USEPA, 1990; Davis and Simon, 1994; Barbour et al., 1999). An assessment or reconnaissance of land uses in the watershed may also be key to defining potential pollutants for monitoring. For example, certain pesticides or herbicides are often associated with specific crops, such as atrazine with corn. Defining Water Quality Problems What to monitor is a very important consideration regardless of the monitoring objective. However, this consideration is paramount when the objective is to identify the pollutant(s) causing water quality impairment. For example, the water quality impairment or effect may be the death of fish, but the cause may be toxins, low dissolved oxygen, or other factors that need to be determined. Within the North Bosque River nutrients (the causative factor) are being intensively monitored due to their impact on algal growth (the response factor). Nutrients, generally either nitrogen or phosphorus, are often limiting factors to algal growth and generally much easier to monitor than algal abundance or growth rates. The nutrients themselves are not the problem per se, but the excessive growth of algae caused by nutrients may be. In the North Bosque River, phosphorus has been identified as the primary nutrient limiting algal growth along much of the River (Matlock and Rodriguez, 1999; Dávalos-Lind and Lind, 1998). By evaluating relationships between in-stream phosphorus concentrations and algal abundance 17 Strategies for Monitoring Nonpoint Source Runoff as represented by chlorophyll-a concentrations (Figure 2), a potential stream target based on PO4-P was developed for the North Bosque River (Kiesling et al., 2001). Development of nutrient targets for the North Bosque River are presented in detail in Kiesling et al. (2001). EPA has also recently published reports specifically outlining the development of nutrient criteria for rivers and streams (USEPA, 2000a) and lakes and reservoirs (USEPA, 2000b). Intensive well designed studies are needed to develop these relationships, but once developed, these relationships can aid greatly in directing the constituents to be evaluated in a monitoring program. Annual Average CHLA (µg/L) Figure 2 Annual average chlorophyll-a as a function of annual average time-weighted PO4-P concentrations for North Bosque River and tributary sites. Figure excludes PO4-P saturated sites. Adapted from Kiesling et al. (2001). 60 y = 13.14Ln(x) - 35.94 2 R = 0.71 50 40 30 20 10 0 0 50 100 150 200 250 300 350 Annual Average Time-Weighted PO4-P (µg/L) Monitoring and Modeling Baseline monitoring may also be used to obtain data for the calibration and validation of instream water quality or lake models. Water quality models are often used to provide information for watershed planning efforts that may not be easily determined through direct monitoring activities. Watershed loading models are tools for simulating the movement of precipitation and pollutants from the landscape to surface and/or ground waters. Hydrologic water quality models are tools for simulating fate and transport of constituents in receiving waters, i.e., streams, lakes, and estuaries. Some models contain both watershed loading and water quality components. Models are particularly useful in projecting the impact of different implementation practices and in determining needed load allocations to obtain water quality targets. Models may also be used to take the place of intensive monitoring efforts, but when used without monitoring data for calibration, they should be considered only a guide to what water quality might be like rather than an indicator of actual water quality. Many sophisticated water quality models exist, such as SWAT (Soil Water Assessment Tool; Arnold et al., 1998) and CE-QUAL-W2 (Cole and Buchak, 1995), which were applied for the USDA Initiative. These sophisticated models can be very helpful in providing information in the watershed planning process, but require extensive watershed specific information to operate. Monitoring data may be needed to calibrate specific model coefficients and to 18 Chapter 3 What to Monitor provide data for overall model calibration and verification. Each water quality model is unique in its purpose, characteristics, and data needs, thus, before using a given model, the model should be thoroughly reviewed to make sure it meets the objectives of the planning process and that the data needed to operate the model are available or can be collected. A review of the general principles behind most surface water quality models can be found in Thomann and Mueller (1987) and Chapra (1997), and a basic description of many models is contained in Shoemaker et al. (1997). Concentrations and Loading Water quality data are generally measured as instantaneous concentrations, but the loading or mass of a constituent added to a water body or transported by a river over time may be just as important. To obtain loading, pollutant concentration and water flow must be measured, because loading by definition is determined as water volume times pollutant concentration. The difference between concentrations and loadings can sometimes be confusing. For example, along the North Bosque River, a clear decrease in phosphorus concentrations occurs from sites near the headwaters to sites near the mouth of the river (Figure 3). In contrast, the loading of phosphorus increases from upstream to downstream, because the volume of discharge increases dramatically between the three sites in this example. Constituent concentrations provide information on the exposure level to aquatic plants and animals at the time of the sample. For toxicants, the exposure concentration is often expressed as an LD50 or the concentration at which 50 percent of test organisms are killed in a given time period, such as 24 hours (see Rand and Pertocelli, 1985). For constituents, such as nutrients, concentrations give us an idea of their availability for aquatic plant growth, although even nutrients in some forms, for example elevated NH3-N concentrations, can be toxic to fish and other aquatic life (see Boyd, 1990). Multiplying concentrations by flow lets us know how much of a given constituent or the load of that constituent that has passed a given point in a stream or been added to a lake or reservoir. In streams and rivers, where flow continually moves waterborne constituents downstream, concentration information may be more important to watershed planning processes than load information with regard to exposure levels. For lakes and reservoirs, loadings from tributary streams and rivers may be as important information as the actual concentration of these loadings. This is because lakes and reservoirs have longer retention times than streams and rivers. A longer retention time allows for inflows to become mixed or integrated with resident waters. It is, thus, important to know both the concentration and loading of inflows into a lake or reservoir to assess their potential impact on in-lake water quality. For example, a high constituent concentration inflow at low flow will often have less of an impact on a lake than the same or even lesser concentration at a high flow. Due to the longer residence time, lakes and reservoirs also generally have a greater natural capacity to assimilate constituent loadings than streams or rivers. Assimilation, in this context, refers to the retention of a constituent in a form that is not easily released to the water column or readily available for biotic uptake (Reddy et al., 1996). To calculate the loading of a constituent, concentration data must be combined with stream flow data. There are a number of different methods to calculate loadings. An integration approach is generally the most accurate, but also one of the most data intensive methods for calculating loadings (Roberston and Roerish, 1999). Various loading calculation methods and the associated sampling effort needed are discussed in more detail in Chapter 6, “How Often to Monitor. “ 19 Strategies for Monitoring Nonpoint Source Runoff Figure 3 Phosphorus concentrations, stream volume, and loadings along the North Bosque River for 1997. Concentrations represent annual flow-weighted averages. 20 CHAPTER 4 Where to Monitor Location of sampling sites is another consideration not only regarding spatial aspects within a watershed, but also in regard to spatial aspects within a water body at a given location. Within a watershed, the size of the drainage area monitored, how many tributaries are located in the watershed area, and critical areas or sources, may influence the location of sampling sites within a watershed. For a given sampling location spatial aspects include depth of sampling and, for streams, lateral position along the cross section. Spatial Considerations As pointed out in the NRCS National Handbook of Water Quality Monitoring (Clausen, 1996), the scale or size of the watershed to monitor is probably one of the most difficult decisions in a water quality monitoring project. Some factors that should be considered are: • Stream order—Should headwater streams be monitored or is the primary interest only along the main stem of a river? Headwater streams are associated with much of the loading of nutrients to downstream water bodies (Peterson et al., 2001), and main stem river sites provide an integration of the various tributaries flowing into a river. The purpose of monitoring, for example defining sources of a pollutant to a river versus monitoring loadings from a river into a lake, will help define the stream order to monitor. • Stream permanence—Is the stream intermittent or perennial? In the North Bosque River watershed most of the tributaries are intermittent with little or no flow between storm events. Even reaches of the North Bosque River can go dry during periods of low rainfall and high evaporation that often occur in the summer. Very little water appears to be contributed to the North Bosque River from groundwater seepage, while in other areas groundwater contributions may be a primary source of base flow contributions. Stream permanence is very important for biological monitoring as the biotic community may be controlled by periodic lack of flow. Hendon et al. (1998) found that macroinvertebrate communities in the North Bosque River watershed were more influenced by persistence in flow (i.e., intermittent versus perennial conditions) than variation in water quality between sampling sites with similar flow regimes. • Weather variability—For paired watershed studies, it is important that similar weather conditions occur within the two watersheds of interest for comparison purposes. Variability in weather conditions, particularly with regard to rainfall, also plays an important role in determining the likelihood of rainfall-runoff events which help determine the frequency of sampling that can be expected as well as the length of time that might be needed to collect the data to meet monitoring objectives. • Heterogeneity of land uses—The heterogeneity of land uses generally increases with watershed size. For determining source contributions of a constituent by land use, a fairly small watershed may need to be monitored to isolate the water quality impact associated with a specific land use. 21 Strategies for Monitoring Nonpoint Source Solution • Number of land owners—If the purpose of monitoring is to influence the management behavior of landowners within a given watershed, the area monitored may need to be small enough that landowners can directly associate what they do on the land with water quality measurements. If the area is too large in comparison to the number of landowners or management units in a watershed, individuals may more easily dismiss their contribution as being less significant compared to others. • Watershed geology—Background water quality can vary greatly with geology. For example, karst geology is associated with cave formations, which allow preferential flow of pollutants into groundwater. Areas with karst geology may need special considerations for groundwater monitoring. The geology of an area also defines the mineral make up of the soil, which in turn defines the soil’s ability to bind up potential pollutants such as phosphorus. For example, Harris et al. (1994) found that sandy soils in Florida had a low phosphorus retention capacity even in the clay-fraction which had a high cation exchange capacity (CEC). Although a high CEC is often associated with a higher ability for a soil to bind anions, such as phosphorus, the dominance of noncrystalline-silica clays in the clay fraction of these soils had a very low affinity for phosphorus. • Watershed topography—Steeper slopes will lead to more runoff and greater potential movement of pollutants than shallower slopes. Topography also determines the dendritic pattern of a the stream system. Determining Sources of Pollution When a monitoring objective is to determine sources of pollutants, the location of a pollutant source is important in considering the location of sampling sites. As a stream sampling site gets further downstream from the pollutant source, there is a greater chance for dilution from other inflows and for physical and biochemical processes to act upon waterborne constituents. During periods of low flow, there may even be incomplete transport of pollutants from one point to another along a stream. These stored pollutants will likely be transported with the next significant rainfall-runoff event. The phenomenon of pollutant storage is aptly shown in Figure 4 through the differences in cumulative loadings of PO4-P and TP and differences in cumulative flow volume between three North Bosque sites. The three sites, in an upstream to downstream direction, are BO040 (North Bosque River below Stephenville), BO070 (North Bosque River at Hico), and BO090 (North Bosque River near Clifton). Site BO040 is immediately below (400 m) the city of Stephenville Wastewater Treatment Plant discharge and consequently loadings at this site reflect this substantial source. For this example differences in loadings and flow volume were calculated on a daily basis by subtracting values at BO040 from BO070 and by subtracting values at BO070 from BO090. These differences were then accumulated daily. Each cumulated difference was divided by the drainage area between the two sites to normalize the data on a per hectare basis, so the two sets of differences (BO070 BO040 and BO090 - BO070) would be visually discernible on the same plot. For the reaches between BO040 and BO070 and between BO70 and BO090, a clear decrease in the cumulative mass of PO4-P occurred from June through December 1998. A somewhat similar decrease in cumulative discharge is noted between BO090 and BO070, but not between BO070 and BO040. These decreases in cumulative loadings of PO4-P indicate that soluble phosphorus, as represented by PO4-P, contributed at upstream points in the North Bosque River watershed are not being transported during this period to BO090 as PO4-P. Pooling of water along the stream or slower flows allow more time for the conversion of PO4-P to less soluble forms via binding with sediment particles and biological uptake by plants and algae during the extended period of low flow (see Froelich, 1988). For the North Bosque River, transformations of soluble phosphorus to less soluble forms appear to occur primarily during 22 Chapter 4 Where to Monitor Figure 4 Cumulative daily differences in PO4-P and TP loadings and discharge volume for 1998 at selected sites along the North Bosque River. From upstream to downstream, site BO040 is located about 34 river kilometers above site BO070, and site BO070 is located about 73 river kilometers above site BO090. 23 Strategies for Monitoring Nonpoint Source Solution extended low flow periods and not during storm events due to rapid transport associated with elevated flows (see McFarland et al., 2001). A similar decrease in TP was indicated for the reach between BO070 and BO090. The decrease in cumulative TP loadings shown between BO070 and BO090 indicates a settling or binding of TP loadings along the North Bosque River prior to reaching sampling site BO090. These settled loadings are generally resuspended and moved further downstream with subsequent elevated flows associated with storm events. In a heterogeneous land-use watershed such as the North Bosque River, isolating contributions of pollutants between nonpoint sources will require some forethought and planning. In some instances it may prove practical and expeditious to establish monitoring sites to isolate individual sources, e.g., placing a site on a first-order stream that originates within the boundaries of a specific agricultural enterprise of interest. Such an approach was not taken in the North Bosque River watershed for the following reasons. First, the water quality problems in the watershed were contentious and volatile, and water quality information collected from a specific operation would be subject to third party open records requests and that information could potentially be used by that third party in legal action against the cooperating operation—at best an undesirable outcome and possibly an outcome that undermines agricultural support for the entire watershed project. Second, monitoring that isolates a specific enterprise, thus, reflects the soils, slopes, facilities, management, and other factors unique to that operation, and therefore monitoring results may not be representative of typical water quality conditions from that land use. However such focused monitoring may be very appropriate for specific circumstances, and monitoring data from such sites may be crucial in calibrating field and landscape loading models. If land-use specific monitoring occurs, multiple sites representing each land-use should be monitored to obtain statistically, representative results. The approach in the North Bosque River watershed was to locate sites on streams with drainage areas including at least two and frequently even more dairy operations. Such site placement made the data much less usable in court against an individual operation and resulted in integrated water quality measurements across the variety of conditions and practices on all operations above the site. A strong negative to this approach in sample site selection, particularly in a heterogeneous watershed, is that other land uses and land covers will undoubtedly be a significant part of the site's drainage area. An example of some of the complicating aspects of drainage area size, site location, and heterogeneity of land uses is depicted by four sampling sites within the North Bosque River watershed (Table 2). The four sites presented in Table 2 may be roughly paired based on drainage area size. Within each pair one site represents a least impacted watershed and the other a watershed impacted by dairy operations. Table 2 Comparison of water quality for impacted and least impacted watersheds of different sizes (Source: Adams and McFarland, 2001). Site Drainage Area (ha) General Water Quality Description Median Storm Event PO4-P (mg/L)a SF020 NF020 848 791 Control, least impacted Impacted by dairy operations 0.030 2.020 NC060 GC100 35,147 26,165 Control, least impacted Impacted by dairy operations 0.010 0.090 a. Represents the median of storm samples collected from January 1, 1996 through December 31, 2000. For data collected between January 1996 and December 2000, the median storm event PO4-P concentration of the two least impacted sites was approximately the same, whereas the median storm event PO4-P concentration varied over one order-of-magnitude between the 24 Chapter 4 Where to Monitor two watersheds impacted by dairy operations. Several factors can help explain this large difference in median PO4-P concentrations between impacted stream sites: 1) the larger watershed will exhibit a greater assimilative capacity than the smaller watershed, 2) environmental conditions and individual dairy operation practices will vary between watersheds, and 3) most importantly, the land use within smaller watershed (NF020) comprises 45 percent dairy waste application fields, while the land use within larger watershed (GC100) comprises only 7 percent dairy waste application fields (Table 1). While dairy waste application fields occur on only seven percent of the land area of GC100, they contribute a disproportionate amount of the nutrients expected in stream flow. Above GC100, over 70 percent of the land area is comprised of wood-range, which is considered the background condition for the watershed, and about 22 percent is associated with pasture and cropland not used for dairy waste application. Estimated export coefficients by land use within the Bosque River watershed for the period November 1, 1995 through March 30, 1998 indicate that dairy waste application fields export 22 times more PO4-P than pasture-cropland and 44 times more PO4-P than wood-rangeland resulting in a relatively small areal extent associated with dairy waste application fields being the major controllable nonpoint source of nutrients above GC100 (see McFarland and Hauck, 1999). This finding indicates that in determining sources a small portion of a watershed may be an important contributing source that needs to be monitored. Evaluating Effectiveness of Control Practices Sampling site location is very important in evaluating the effectiveness of control practices. The upstream-downstream approach is one common statistical design recommended for evaluating nonpoint source management practices (Spooner et al., 1985). While the impact of upstream water quality on downstream water quality can be a complicating factor in evaluating water quality data, nesting, in this case, can greatly aid evaluating the effectiveness of control practices by isolating drainage area contributions along a stream. Differences between paired observations at the upstream and downstream sites are used to evaluate treatment impacts. Ideally, pre- and posttreatment observations are collected, so the treatment impact can be isolated from potential variations in rainfall and flow between the two monitoring periods. Even within a relatively small area, inherent differences in such factors as geology, soils, and slope can cause differences in water quality separate from the treatment effect, so if only post-treatment samples are collected, it is difficult to determine if observed differences are due to the treatment or some other inherent difference. Care must be taken in site selection for evaluating the effectiveness of control practices, because sites that are too far apart may have no relationship between them, while sites that are too close together may not show a treatment difference if the treatment contribution area is relatively small. A paired watershed approach is another recommended method that allows researchers to measure the responsiveness of water quality to land use changes (Spooner and Line, 1993). A paired watershed approach is similar to the upstream-downstream approach but involves monitoring two similar watersheds for a control period and then applying a treatment to one of the watersheds while continuing the same practices as during the “control” period on the second watershed. A comparison of changes in the relationship between the two watersheds for the control and treatment periods is used to determine changes in water quality due to the treatment (e.g., Clausen and Spooner, 1993). The paired watershed approach works best when watersheds are in close or immediate proximity to one another to minimize spatial variability in weather conditions, especially rainfall, between the two watersheds. 25 Strategies for Monitoring Nonpoint Source Solution Surrogate Sampling Locations Sometimes it is not practical to directly monitor the location of interest due to difficulty in accessing the site or expense of monitoring. In such cases, a surrogate for the location may be appropriate depending on the objectives of monitoring. For example, in the upper portion of the North Bosque River watershed, 40 Public-Law 566 (PL-566) reservoirs exist. These PL-566 reservoirs were built in the 1950s and 1960s to reduce and retard high flows related to flooding. The streams flowing into most of these reservoirs are intermittent and are characterized by episodic runoff events1. Although not their intended purpose, these small reservoirs act as integrators of water quality for the drainage above them. During small storm events, these reservoirs may catch all the water that comes into them (McFarland and Hauck, 1995a). Water quality within these small reservoirs is directly related to the quality of the water flowing into them (McFarland and Hauck, 1995b). Therefore, under certain monitoring situations it may be appropriate to use these reservoirs as surrogate sampling sites of the water quality flowing into them, thus, avoiding the expense of storm water monitoring that would be necessary to characterize water quality associated with intermittent stream flow. Location within a Water Body Determining where to sample within a water body may be just as important as determining which water bodies to sample within a watershed. Klemm et al., (1990) outlined a number of considerations for biological sampling, such as making sure stream depth, substrates, and velocity are similar between sampling locations. Several references go into detail on selecting appropriate stream and lake or reservoir sampling sites (e.g., Sanders et al., 1990; Clausen, 1996; TNRCC, 1999). General site selection criteria includes such factors as making sure appropriate habitat and land uses are represented, that the stream section monitored has a straight uniform cross section, and that a stream site is not located at obstructions. The vertical and horizontal velocity profiles of a stream should also be considered in placement of stream sampling sites. In general, the larger the stream system, the more important variations in the vertical and horizontal profile become. Velocities will be highest at the top and in the middle of a stream and slowest toward the bottom and outside edges of a stream. These variations in velocity can influence the transport of constituents and determine the type of sampling needed for a given site. Velocity is also an important consideration in defining mixing zones associated with point source contributions. Mixing zones represent limited and defined areas where initial dilution of permitted or authorized discharges may occur. Water quality within mixing zones will be changing and may not be representative throughout the entire stream profile as the discharge volume for the point source is diluted by stream water. Mixing zones will vary depending on the size and rate of the discharged contribution compared to stream flow and volume. Suggested procedures on selecting site locations with regard to mixing zones for stream water quality sampling are provided by Montgomery and Hart (1972) and Sanders et al. (1977). For macroinvertebrate monitoring, Klemm et al. (1990) suggest including samples along both stream banks as well as midchannel, if a site is directly below a discharge that is not well mixed. 1 26 Sites SF020 and NF020 represent two such intermittent stream sites. The charts of average daily flow for sites NF020 and SF020 show extended periods of low or no flow punctuated by dramatic rises and falls in flow in response to rainfall (see Appendix B for details). Chapter 4 Where to Monitor Where two streams debouch or merge, the complete mixing of their waters does not occur instantaneously. A downstream reach is required for waters of the streams to completely mix. The length of this reach will be a function of many variables including geomorphology, stream flow, and sinuosity. Sampling within this mixing zone or reach can result in erroneous characterization of the combined water quality of the two streams and should be avoided for the majority of monitoring purposes. When locating monitoring sites below two debouching streams, keep in mind that under some circumstances the distance required for complete mixing may be surprisingly long. The depth at which a sample is taken is another consideration in determining sampling site location. Although differences with depth are more often associated with lake and reservoir systems, depth can also be a factor is stream systems, particularly in larger streams. Sediment and sediment bound constituents can be expected to show a distinct gradient with depth, because lighter particles, such as clay and silt, can stay suspended in the current, while larger particles, such as sand and rocks, will move along the stream bottom. A couple of other important constituents that may show distinct differences with depth include dissolved oxygen and water temperature. Water temperatures and dissolved oxygen concentrations generally decrease with increasing depth in lakes and reservoirs, particularly during summer months. Stratification of dissolved oxygen and temperature in streams is less common, but can occur and is most likely associated with base flow conditions in deeper pools or within slow moving, deep streams. 27 Strategies for Monitoring Nonpoint Source Solution 28 CHAPTER 5 What Type of Monitoring to Implement The type of sampling to implement depends on the purpose of monitoring as well as the variability in constituent concentrations with flow and depth at a given site. In dealing with vertical and horizontal variability in water quality at a site, samples may need to be collected at varying depths and widths. Sampling at increments by depth is often used for constituents that are known or expected to have potential changes with depth, such as temperature and dissolved oxygen in lakes or reservoirs, which exhibit a high potential for seasonal stratification. Within streams, monitoring vertical and horizontal differences in the concentration of different sediment fractions may be important. The USGS water sampling handbook (USGS, 1999) covers in detail methods for collecting depth and width integrated samples from streams to obtain a composite, discharge-weighted sample that is proportional to total stream flow. For stream sites, integrated sampling can be very intensive and impractical at high flows when a sampler cannot be lowered properly through the vertical or horizontal profile. For most nonpoint source sampling of stream sites, discrete samples are collected at one point within the stream profile using grab or automated samplers. However, modifying the intake of an automatic sample to include a float to raise a perforated intake line can allow depth-integrated sampling during elevated flows (e.g. Clausen, 1996). While depth and width integrated sampling is sometimes necessary, most stream sampling with regard to nonpoint source contributions occurs as discrete samples collected at one point within the stream profile using manual grab sampling techniques or automated samplers. Manual sampling is often used in determining baseline conditions within a water body and is generally used for compliance and assessment monitoring (sometimes called ambient monitoring) by regulatory agencies. Manual sampling generally consists of dip or point sample. A dip sample commonly involves wadding into a stream and dipping a narrowmouthed bottle into the water or lowering a bottle with a weight into the water from a bridge or other structure. Generally, dip samples are collected below the water surface at a depth of about 0.3 meters (1 ft). A point sample involves specialized equipment and involves lowering the sampler to a specified depth for sample collection or using a single-stage sampler that collects a sample when the water level rises to a specified height. Automated or pump samplers are most often used for monitoring storm runoff when changes in discharge occur rapidly making it difficult and potentially dangerous for individuals to manually collect samples. Flow varies greatly over a storm event as well as the concentration of different constituents. For example at North Bosque River site BO070 near Hico, Texas, a typical storm event may show a rapid rise in the concentration of TSS and TP peaking near the time of maximum discharge. Whereas PO4-P and NO2-N+NO3-N concentrations actually decrease initially and then increase on the falling portion of the hydrograph (Figure 5). 29 Strategies for Monitoring Nonpoint Runoff Figure 5 Variation in constituent concentrations with discharge at North Bosque River site BO070 from a March 1998 storm. 30 Chapter 5 What Type of Monitoring to Implement The pattern of concentrations for a given constituent can also vary greatly between events. In the storm event presented in Figure 5, NO2-N+NO3-N indicates a fairly constant increase in concentration with increased elapsed time during the event. In Figure 6, NO2-N+NO3-N concentrations are compared for six different storm events with storm 1 representing the same storm event as shown in Figure 5. The other five events presented show varying, often complex patterns in NO2-N+NO3-N concentrations (Figure 6a). When NO2-N+NO3-N concentrations are normalized for the maximum concentration in each event and elapsed time is also normalized (Figure 6b), there is a clear indication that the maximum NO2-N+NO3-N concentration for a storm can occur at any time during an event. There does appear to be a weak pattern of decreasing NO2-N+NO3-N concentrations at the beginning of several of the storm events, which appears to be related to instantaneous discharge (Figure 7). For dissolved constituents, such as NO2-N+NO3-N, a decrease in concentration during storm events with increased flow is not uncommon and is attributed to a dilution effect associated with initial runoff volumes generally containing lower concentrations of soluble constituents than later in the event (see Reckhow et al., 1980). Figure 6 Variation in NO2-N+NO3-N concentrations over time during six storm events for a) actual concentrations and elapsed time and b) normalized concentrations and elapsed time. Data for North Bosque River site BO070. 31 Strategies for Monitoring Nonpoint Runoff NO2-N+NO3-N (mg/L) Figure 7 NO2-N+NO3-N concentration during selected storm events in relation to discharge at site BO070. 10.00 1.00 0.10 0.01 1 10 100 1,000 Instantaneous Discharge (m3/s) Storm 1 Storm 2 Storm 3 Storm 4 Storm 5 Storm 6 For particulate or water suspended constituents such as TSS, a very different response pattern for storm event concentrations is observed. With TSS, the maximum TSS concentration is generally a function of flow energy with peak TSS concentrations occurring near peak flow in the hydrograph (Figure 8). Peak TSS concentrations generally occur early in a storm event (Figure 9a & 9b). These phenomena are most likely the result of dislodgement and transport of sediment particles from the land surface during the initial stages of runoff and the resuspension of sediment within the stream bed with increasing stream flow energy. For water suspended constituents, such as TSS, concentrations generally decrease as flow energy decreases during the falling portion of the hydrograph. Figure 8 TSS concentration during selected storm events in relation to discharge at site BO070. TSS (mg/L) 10,000 1,000 100 10 1 1 10 100 1,000 Instantaneous Discharge (m3/s) Storm 1 32 Storm 2 Storm 3 Storm 4 Storm 5 Storm 6 Chapter 5 What Type of Monitoring to Implement Figure 9 Variation in TSS concentrations over time during six storm events for a) actual concentrations and elapsed time and b) normalized concentrations and elapsed time. Data from North Bosque River site BO070. Because concentration patterns can vary greatly between constituents within storm events and for the same constituent between events, storm sampling must be performed to allow for this variability. It may be necessary to sample several events and represent several points in time during each event to properly characterize storm water contributions of a given constituent or set of constituents. While manual grab sampling can be used to sample storm events, if intensive storm sampling is necessary to represent several points during an event, the use of automated samplers is recommended. Automated samplers allow samples to be collected at prescribed time or flow intervals and are generally more reliable than manual sample collection when specific sample intervals are required. Automated samplers also help minimize potential safety risks to individuals conducting wet-weather sampling. In some cases, probes for continuous, real-time sampling may be used. Although real-time, continuous monitoring is most often conducted for research purposes, general applications are becoming more frequent. Stable and chemically sensitive probes for 33 Strategies for Monitoring Nonpoint Runoff continuous, real-time sampling are commonly available for constituents, such as water temperature, specific conductance, dissolved oxygen, chlorophyll-a, turbidity, and pH. Probes are also available for constituents, such as ammonium, nitrate, and soluble phosphorus, but the sensitivity of these probes should be considered before using them for monitoring. The chemical sensitivity currently associated with probes for constituents such as ammonium may allow monitoring in enriched systems such as wastewater or industrial discharges, but in many natural waters, these probes may not be sensitive enough to provide useful data. The reliability and sensitivity of equipment for continuous real-time monitoring is improving rapidly, as well as the type of constituents that can be monitored. Efforts are also being made by several researchers to use constituents that can be reliably monitored using continuous, monitoring probes as surrogates to predict concentrations of other constituents, such as chloride and sulfate from measurements of specific conductance and fecal coliform and total phosphorus from measurements of turbidity (e.g., Christensen et al., 2000). While automated samplers allow the collection of multiple samples over a storm event, it may be desirable to composite these samples. A composite sample is basically a collection of single samples combined into one sample by some factor, generally time or flow in the case of storm events. Depth composited samples are also common when evaluating the vertical profile of a water body. Compositing is often desirable to decrease the number of samples needed for laboratory analyses while attempting to maximize the amount of information obtained from the sample. While time-compositing of storm samples is sometimes used because of the simplicity with which it can be implemented, time-composited samples will typically provide a skewed characterization of water quality during a storm event. Further, a time-weighted concentration has no intrinsic relationship to the stream hydrograph precluding the use of such concentrations to calculate loadings, unless flow or concentration is constant over the period of sampling. However, time-weighted sampling may be totally appropriate for instances when monitoring is addressing time-related issues, such as exposure time of organisms to acute or chronic concentrations of toxins or average nutrient concentrations available to periphyton. Flow weighting of storm samples is generally preferred if storm samples need to be composited. Flow compositing of storm sample is more complicated than time compositing, because flow compositing requires measurement of stream stage or level throughout the event and knowledge of the stage-discharge relationship at the sampling site. The following example shows how time compositing and flow compositing of samples can lead to different interpretations of average storm event concentrations for constituents. In evaluating sequentially collected data from six storm events at BO070, fairly large differences in event mean concentrations (EMCs) occurred, particularly for TP and TSS, depending on how data were averaged (Figure 10; storms 1 through 6 match the storms presented in Figures 5-9). For most storms, flow weighting substantially increased the EMC for TP and TSS compared to time weighting, because the largest TP and TSS concentrations typically occurred with the highest discharges. In contrast, for NO2-N+NO3-N and PO4-P, the greatest EMC for most storms occurred when data were time weighted rather than flow weighted, however, trends in EMCs across storms were not as consistent for these constituents as for TP and TSS. Generally, when the flow-weighted EMC is greater than the time-weighted EMC, the higher constituent concentrations are occurring with higher flows, and the converse is true when the flow-weighted EMC is less than the time-weighted EMC. As indicated in Figures 6 and 9, different storm events can lead to different patterns of concentration response for the same constituent, and, thus, differences when averaged if either flow weighting or time weighting is used. If compositing must be done to allow estimation of constituent loadings, flow weighting is preferred, because by definition loading is the integration of waterborne constituent 34 Chapter 5 What Type of Monitoring to Implement Figure 10 Comparison of flow-weighted and time-weighted averaging of storm samples. 35 Strategies for Monitoring Nonpoint Runoff concentration with flow over time. Using storm 1 as an example, the total storm volume discharge was about 35,970,000 m3 (1,271,000,000 ft3). This translates into a mass loading for TSS of 37,200 metric-tons (41,000 tons) using the flow-weighted EMC of 1,033 mg/L TSS and 12,200 metric-tons (13,400 tons) using the time-weighted EMC of 339 mg/L TSS. Using an integration approach of discharge with water quality for the individual sequential samples, a TSS loading of 37,200 metric-tons (41,000 tons) was calculated. The integrated loading equals the loading estimated using the flow-weighted EMC and provides the best estimate of actual storm loading. The time-weighted average EMC greatly underestimated TSS loadings in comparison. While the differences were most dramatic for TSS, similar differences in loading estimates could be expected for the other constituents depending on the weighting of composited samples. 36 CHAPTER 6 How Often to Monitor Data collection efforts generally represent a compromise between confidence in the data and available resources. Defining an appropriate frequency for sampling, particularly when monitoring nonpoint source contributions is a formidable task. To illustrate this problem, we present a comparison of several common sampling strategies that employ different sampling frequencies to estimate annual concentrations and loadings of PO4-P, TP, and TSS by subsampling data from seven sampling sites within the North Bosque River watershed. As presented previously (Table 1), the seven sites selected for these comparisons provide a variety of water quality and hydrologic conditions, which may be summarized as follows: • BO040 - water quality highly impacted by point and nonpoint sources; perennial flow • BO070 - water quality moderately impacted by point and nonpoint sources; nearly perennial flow • BO090 - water quality slightly impacted by point and nonpoint sources; perennial flow • NC060 - good water quality; intermittent flow • GC100 - water quality moderately impacted by nonpoint sources; intermittent flow • NF020 - water quality highly impacted by nonpoint sources; highly intermittent flow • SF020 - good water quality; highly intermittent flow. Note that the above water quality characterizations are relative and not an indication of water quality condition relative to state and federal surface water standards. Average Annual Concentrations from Routine Sampling Estimation of annual average concentrations are often an objective of routine monitoring when assessing ambient water quality conditions. Routine monitoring generally involves collecting near surface grab samples at set time intervals throughout the year. Common time intervals used are biweekly, monthly, or quarterly. To evaluate how changing the time interval might influence annual average concentrations, a data set containing routine biweekly grab sampling data was subsampled on a monthly and quarterly basis for each of seven sites. Two monthly subsamplings were evaluated. One using the first biweekly grab sample collected in a month, and the second using the last biweekly grab sample collected in a month. For those months in which three biweekly grab samples were collected, the middle sample collected in a month was excluded. To represent the potential variability in quarterly sampling, six subsampling options were evaluated assuming at least six biweekly samples collected in a quarter. Q1 represented the first biweekly sample in a quarter, while Q6 represented the last biweekly sample in a quarter. Subsampling options Q2 through Q5 represented the biweekly sample closest to increments of 32, 48, 64, and 80 percent elapsed time in a given quarter. Results comparing different grab sampling intervals are presented for North Bosque River site BO070 in Figure 11. 37 Stategies for Monitoring Nonpoint Source Runoff Figure 11 Impact of sampling interval on average annual concentration at site BO070. B represents biweekly sampling, M1 and M2 represent monthly sampling schemes, and Q1 through Q6 represent quarterly sampling schemes. 38 Chapter 6 How Often to Monitor Annual average stream concentrations of PO4-P, TP, and TSS at site BO070 clearly show an increase in statistical confidence with increasing sampling frequency when using the annual average of biweekly sampling as the “true” concentration. To summarize the results in Figure 11, a percent error from the annual average concentration based on biweekly sampling was calculated for the monthly and quarterly sampling schemes. For PO4-P, percent error for monthly sampling ranged from plus 18 percent to minus 17 percent, and percent error for quarterly data was greater than plus or minus 50 percent. The percent error for TP was fairly similar to that of PO4-P, but for TSS, the difference between sampling schemes increased dramatically. For TSS, percent error for monthly sampling ranged from plus 81 percent to minus 83 percent. For quarterly TSS samples, the disparity increased to plus 90 percent to minus 419 percent that of average annual values based on biweekly sampling. For site BO070, monthly sampling might be appropriate for measuring PO4-P or TP concentrations if a 20 percent error compared to biweekly sampling is acceptable. For TSS, monthly sampling was insufficient to maintain this same 20 percent margin of error compared to biweekly sampling. Similar results were found for the other six sites evaluated, and results for all seven sites are presented in Appendix C. Annual Average Concentrations for Storm Events Storm events will generally have concentrations for constituents that differ from base flow. Because many routine sampling programs focus on base flow conditions, special monitoring will be necessary to evaluate constituent concentrations associated with storm events. It is generally infeasible, due to limitations in money and personnel, to monitor every storm event. Therefore, it is important to know how many events need to be monitored to obtain a reasonable estimate of annual average concentration of constituents associated with storm events. This question was evaluated for seven selected stream sites within the North Bosque River watershed by using successive storm events as subsamples to represent estimates of the annual average storm event concentration. Flow-weighted mean concentrations of successive storm events monitored during the year were compared with the annual average flowweighted concentration of all storm events in a given year. The subsampled mean concentration was calculated for PO4-P, TP, and TSS by adding the mass associated with each additional storm event and dividing by the cumulative storm volume to obtain a flowweighted mean value. Sampling procedures used for storm events are outlined in Appendix A under the section, Data Collection Methods and Procedures. Masses for each constituent were calculated by combining flow and water quality data using a midpoint, rectangular integration method between water quality samples (Stein, 1977). Figure 12 shows storm event mean concentrations for PO4-P, TP, and TSS at North Bosque River site BO070 for each successive subsampling of events. In general, there is a stabilization in the mean concentration for storm events as the number of storm events monitored increases indicating that not all storms in a given year need to be monitored to obtain an estimate of the annual average storm concentration for a constituent. To estimate the number of storm events needed to obtain a reasonable estimate of annual average storm concentrations for PO4-P, TP, and TSS, a percent error was calculated for each subsample estimate in comparison to the annual average flow-weighted concentration using all events. The number of storm events needed to obtain accuracy within a prescribed percent error in a given year was evaluated using allowable percent errors of 10, 20, 30, 40, and 50 percent. Relatively little difference was noted in the number of storm events needed to meet these allowable percent errors between sites within a given year, so annual data for all seven sites were combined and averaged (Figure 13). As expected, the number of storm events that should be monitored increases as the allowable percent error decreases. For PO4-P, at least 39 Stategies for Monitoring Nonpoint Source Runoff Figure 12 Estimates of annual storm event mean concentrations based on subsampling successive storm events at North Bosque River site BO070. 40 Chapter 6 How Often to Monitor Figure 13 Number of storm events needed to obtain certain allowable percent errors in EMCs for seven sampling sites in the North Bosque River watershed. Number of Storms Monitored 10 9 8 7 6 5 4 3 2 1 0 10 20 30 40 50 Allowable Percent Error PO4-P TP TSS eight storms should be monitored or about 50 percent of all events, based on an average of 16 storms per year, to stay within an allowable percent error of 10 percent. Only three storms need to be monitored if the allowable percent error is 40 percent. Storm event TP concentrations appeared to be more stable than PO4-P or TSS concentrations in the North Bosque River watershed as indicated by the need to monitor fewer storm events to obtain the same percent error. Annual Loadings A further step in determining how often to monitor depends on whether or not estimates of annual loadings are necessary to meet monitoring objectives. Estimating annual loadings generally requires a combination of routine and storm event monitoring. As discussed in the section above, few monitoring programs can afford to sample every storm event, because of constraints in the availability of labor, equipment, and money. It may not be feasible to invest in expensive automated sampling equipment for intensive storm sampling, so alternative sampling strategies may need to be considered to obtain estimates of storm loadings. There are also a number of different approaches available for estimating annual loadings (for examples see Cohn, et al., 1989; Preston et al. 1989; and Robertson and Roerish, 1999). To provide insight into the estimation of loadings and the impact of different sampling strategies, two common loading calculation approaches, an integration and a regression approach, were compared with different sampling strategies representing different levels of monitoring intensity by subsampling from data available for seven sites within the North Bosque River Watershed. An integration approach using all available monitoring data was assumed to be the “true” estimate of loading for comparison purposes, because the monitoring program within the North Bosque River watershed is highly intensive and tries to capture every storm event (see McFarland and Hauck, 1999 for details). 41 Stategies for Monitoring Nonpoint Source Runoff Integration Approach In estimating loads, an integration approach of combining storm and grab sampling is generally recommended (Robertson and Roerish, 1999). The integration approach assumes a constant water quality between samples, although differences are expected between base flow and storm flow in an area with large nonpoint or point source contributions. While it will typically be infeasible to sample all storm events, it may be possible to sample a select number of storm events and associate these data with elevated flows in the calculation of loadings. Three subsets of data were used with the integration approach to estimate loadings. The first data subset used only biweekly grab sampling data. The second and third data subsets included biweekly grab sampling data but associated a volume-weighted event mean concentration (EMC) with elevated flows. In the second subset, EMCs for constituents were set annually based on an overall EMC of the first four storms in a given year. The first four storm events were selected based on previous analyses indicating a plus or minus 20 to 30 percent error using four storm events compared to the annual average storm event concentration for PO4-P, TP, and TSS (see Figure 13). In the third data set, EMCs for constituents were set quarterly based on EMCs associated with the first storm in each quarter. Data used in these evaluations were collected from January 1, 1995 to December 31, 1999. Regression Approach The second loading estimation method used a regression approach as presented by Cohn et al. (1992) using a minimum variance unbiased estimator (Bradu and Mundlak, 1970). The regression approach was based on the following seven parameter model: 1) ln[C] = bo + b1ln[Qa] + b2{ln[Qa]}2 + b3[Ta] + b4[Ta]2 + b5sin[2pT] + b6cos[2pT] +e where ln [ ] denotes the natural logarithm function, C is the instantaneous constituent concentration on a given day, Qa is the average daily discharge adjusted with a centering variable, T is time over the monitoring period measured in fractions of a year, and Ta is time adjusted with a centering variable. The sine and cosine functions are designed to account for periodic or cyclic variations in water quality, such as annual variations with season. The values b0 through b6 represent coefficients estimated through the regression process. The error or residual of the model is represented by e. The centering variables used for adjusting discharge and time were calculated following procedures outlined in Cohn et al. (1992). Centering adjustments of Q and T were made to avoid multicollinearity among the regression equation coefficients. One of the potential advantages of the regression approach over the integration approach to estimating loadings is that the regression approach provides continuous estimates of daily concentrations with discharge and time rather than assuming set values for concentrations. Daily concentration estimates from the regression model were adjusted using a minimum variance unbiased estimator (MVUE) and multiplied by daily discharge and summed by year to estimate annual loadings. The estimator modifies the bias introduced in the retransformation from “log space” to “real space” of the regression model estimates of concentration (Cohn et al., 1989). The MVUE is generally considered applicable only if the residuals of the equation are normally distributed. A Smearing bias estimator (Duan, 1983) was also used and compared to results using the MVUE. The Smearing estimator is recommended as a retransformation bias corrector when residuals are not normally distributed (Hirshch et al., 1993). Consistently, the MVUE performed equal to or better than the Smearing estimator even when residuals from the regression model were not normally 42 Chapter 6 How Often to Monitor distributed based on a Shapiro-Wilks test of normality (a = 0.05). The results using the MVUE are presented in the body of this report. The results using the Smearing estimator may be referenced in Appendix D. Four subsets of data were evaluated using the regression approach. The first included only biweekly grab data. Biweekly grab data were then supplemented to include storm data. It was assumed that intensive monitoring of individual storms was not an option and only one sample was collected per storm event. Although TIAER’s storm sampling program is quite intensive, these samples still represent only set time periods within an event (see Appendix A for details on TIAER’s storm sampling protocols). Within a selected storm event, the sample closest in time to the criteria outlined below was used as the representative sample. Storms included in the different scenarios were selected to allow monitoring during normal working hours (8 am to 5 pm), if manual sampling was involved. On average between 1995 and 1999 about 17 storms per year occurred at each of the 7 sampling sites evaluated. Assuming only a limited number of storms could be monitored per year, sampling schemes were further limited to sample only about three to eight storms per year focusing on larger events. This was done by allowing sampling to occur only when instantaneous discharge levels exceeded a specific limit in relation to “historical” average daily volume following procedures similar to those outlined by Cohn et al. (1992). Storm sampling included the following sampling scenarios: • Peak flow—Storm samples used in this scenario represented concentrations associated with the peak discharge of storm events. Storm events were limited to those in which peak flow was greater than the 50th percentile of average daily flow at a site as calculated from available discharge data from January 1, 1995 to December 31, 1999. Storm events were also constrained to have peak flow occurring between 9 am and 4 pm or normal working hours allowing one hour of travel time at the beginning and end of the day. This scenario, while fairly unrealistic in reality, was evaluated to determine how important it was in the regression approach to include samples representative of the peak discharge. • Storm chasing—The events used for the storm chasing scenario were chosen based on the constraints described above for the peak flow scenario, except a two-hour delay was assumed in obtaining samples after peak discharge occurred. Peak discharge was, thus, allowed to occur between 7 am and 2 pm for storm events to keep sampling within normal working hours. Samples selected were, thus, representative of the hydrograph two-hours after peak discharge rather than at peak discharge. • Set stage—The set stage scenario assumes a sampling device is available that will collect a sample when stream flow rises above a certain level. Samples used in this scenario represented the rising limb of storm events. To limit the number of storm events sampled, only storms with instantaneous discharges exceeding the 95 percentile of average daily flow for a given site were included in this scenario. No time of day restriction was imposed, because it was assumed that samples could be retrieved during normal working hours and still meet holding time restrictions for laboratory analysis. These three scenarios used with the regression loading approach assume a knowledge of historical and current flow conditions at each site. Historical data were used to set flow thresholds, so monitoring efforts were concentrated on larger storm events. The underlying assumption being that the purpose of storm monitoring was to best represent concentrations associated with higher flows for input into the regression model for loading estimates as a supplement to routine grab sampling data. It was assumed that most monitoring sites will have some history of at least average daily flows which can be used as an indicator of larger storm events based on instantaneous flows. For the data sets evaluated, the distribution of average daily flows for each site was based on discharge data for the timeframe from which data subsets were selected (January 1, 1995 through December 31, 1999). While not presented 43 Stategies for Monitoring Nonpoint Source Runoff in the body of this report, the regression approach was also evaluated using the full data set of storm and grab samples available for each site to see if increasing the number of samples would improve the results of this approach. The results using the full data set are presented in Appendix D. These storm sampling scenarios used with the regression approach assumed that a single grab sample was collected to represent each sampled storm event. While actually collecting a sample during peak discharge of a storm event is somewhat unrealistic, storm chasing and set stage sampling are reasonable approaches given adequate measurements of stream stage. All loading estimates assumed the availability of an adequate and fairly accurate estimate of discharge on at least a daily basis with real-time availability of stream level data. Results Comparing Loading Estimation Approaches In this section, the biweekly, storm chasing, and set stage scenarios using the regression method for estimating loadings are presented in comparison to the integration method scenarios. The results of the peak flow scenario with the regression method generally did not perform well in comparison to the “true” loading estimates, and, thus, are presented only in Appendix D. Loadings for the integration and regression scenarios were calculated annually and compared to the “true” loading estimates as a percent error by year. For most sites, years 1995 through 1999 are presented. For sites BO090 and NC060, only results for years 1996 through 1999 are presented, since sampling at these two sites did not begin until October 1995. Results of estimated annual loads were somewhat mixed as to what might be the best sampling method at a given site. For PO4-P, measuring the first four storm events along with biweekly grab samples with the integration approach gave the best overall results (Table 3). Using the regression method for estimating loadings, a set stage at 95 percent of average daily flow along with biweekly grab samples gave somewhat reasonable average results, although with much more variability between years than the integration method. At sites BO040 and BO070, all six methods seemed to give fairly reasonable results, while at BO090, the three regression method approaches greatly overestimated loadings. For PO4-P and TP (Tables 3 and 4), use of the integration method with just biweekly grab samples generally underestimated annual loadings, except at site BO040 where loadings were slightly overestimated. The close proximity of the Stephenville wastewater treatment plant discharge to site BO040 makes water quality at this site slightly higher in concentration of PO4-P and TP at low flow, when routine grab samples are more likely to be collected, than during storm events (Adams and McFarland, 2001). Results for TP were similar to the results for PO4-P (Table 4). Overall the best loading estimation method appeared to use the integration approach with biweekly grab samples supplemented with flow-weighted EMC data from the first four storm events in a year to represent the concentration for elevated flows. On average across sites, using the first four storm events generally underestimated or overestimated loadings within 20 percent of “true loadings”, although variability in percent error for specific years and sites ranged greatly (see Appendix E). The regression approach using set stage or storm chasing to represent concentrations at elevated flows showed some promise as potential loading estimation methods, although both these scenarios greatly overestimated loadings at site BO090 compared to “true” loads, and the variability in loadings between years at NF020 was probably outside the margin of acceptable error. 44 Chapter 6 How Often to Monitor Table 3 Annual average percent error in PO4-P loading estimates compared to “true” loadings using regression and integration approaches. Integration Approach Regression Approach First Four Storms + Quarterly Storms Biweekly Grabs Biweekly Grabs + Biweekly Grabs Biweekly Grabs Set Stage + Biweekly Grabs Storm Chasing + Biweekly Grabs 51% 66% -1% 26% 3% 17% 17% 29% -3% 24% -8% 25% 3% 22% 25% 23% 14% 18% -24% 20% -16% 46% -28% 44% 285% 297% 111% 150% 204% 223% Average Std. Dev. -32% 27% -3% 8% -9% 46% 5% 57% 14% 65% 18% 72% GC100 95-99 Average Std. Dev. -18% 37% 8% 92% -24% 36% 213% 132% 23% 62% 128% 118% NF020 95-99 Average Std. Dev. -42% 47% -11% 12% -17% 19% 69% 111% -27% 11% 28% 70% SF020 95-99 Average Std. Dev. -24% 65% -15% 44% 22% 57% 164% 163% 65% 48% 18% 38% Site & Years PO4-P BO040 95-99 Average Std. Dev. 26% 28% -7% 25% BO070 95-99 Average Std. Dev. 0% 18% BO090 96-99 Average Std. Dev. NC060 96-99 Table 4 Annual average percent error in TP loading estimates compared to “true” loadings using regression and integration approaches. Integration Method Regression Method Biweekly Grabs First Four Storms + Quarterly Storms Biweekly Grabs + Biweekly Grabs Biweekly Grabs Set Stage + Biweekly Grabs Storm Chasing + Biweekly Grabs 17% 26% -1% 22% -3% 13% 18% 22% -10% 32% -24% 13% -17% 9% 15% 20% 9% 18% -49% 33% 2% 28% -36% 43% 36% 150% 123% 254% 118% 228% Average Std. Dev. -71% 13% -14% 30% -40% 45% -72% 9% -5% 32% 6% 45% GC100 95-99 Average Std. Dev. -52% 20% -18% 52% -21% 39% -7% 24% -4% 23% 3% 23% NF020 95-99 Average Std. Dev. -48% 39% 5% 13% -10% 21% 27% 87% -2% 49% 23% 74% SF020 95-99 Average Std. Dev. -57% 34% -22% 32% -20% 18% 22% 73% 39% 82% 3% 52% Site & Years TP BO040 95-99 Average Std. Dev. 2% 21% -11% 27% BO070 95-99 Average Std. Dev. -22% 19% BO090 96-99 Average Std. Dev. NC060 96-99 45 Stategies for Monitoring Nonpoint Source Runoff For TSS, no one method consistently estimated loadings with reasonable accuracy, although as with PO4-P and TP, integration using biweekly grab samples supplemented with an EMC from the first four storm events gave the most promising results (Table 5). These results generally underestimated TSS loadings, except at NF020 where TSS loadings were generally overestimated. Under- or overestimation of TSS loadings by 50 percent or more often occurred for individual years. These loading results emphasize the great variability in TSS concentrations that can occur as illustrated in Figure 11. If an accurate estimate of TSS loadings is necessary, very intensive monitoring of storm events and base flow maybe necessary. Factors complicating the estimation of TSS loadings may be contributions from bank sluffing and internal contributions, which may be relatively large in some streams or stream reaches. Such complications occurred during large storm events at the North Bosque River site BO090 near Clifton (McFarland et al., 2001). Table 5 Annual average percent error in TSS loading estimates compared to “true” loadings using regression and integration approaches. Integration Method Regression Method First Four Storms + Quarterly Storms Biweekly Grabs Biweekly Grabs + Biweekly Grabs Biweekly Grabs Set Stage + Biweekly Grabs Storm Chasing + Biweekly Grabs -52% 41% -16% 29% 20% 80% 137% 172% -32% 49% -73% 12% -32% 39% -19% 9% 157% 157% -76% 30% 10% 42% -54% 64% 502% 1110% 2379% 4512% 1540% 3011% Average Std. Dev. -95% 6% -21% 59% -53% 80% -95% 2% 211% 341% 123% 227% GC100 95-99 Average Std. Dev. -91% 8% -37% 57% -42% 53% -87% 2% -54% 11% -78% 3% NF020 95-99 Average Std. Dev. -93% 11% 83% 143% 76% 269% -94% 5% 762% 1154% 22% 126% SF020 95-99 Average Std. Dev. -86% 17% -41% 47% -61% 25% 41% 214% 1832% 2538% 175% 375% Site & Years TSS BO040 95-99 Average Std. Dev. -74% 19% -42% 37% BO070 95-99 Average Std. Dev. -70% 35% BO090 96-99 Average Std. Dev. NC060 96-99 For loadings, storm sampling using automated samplers is recommended to complement manual sampling. PO4-P, TP, and TSS loadings were generally underestimated by manual sampling alone for sites where PO4-P, TP, and TSS was transported predominately by nonpoint source runoff. An integration approach using biweekly grab sampling and a flowweighted EMC from the first four storm events at a site for elevated flows appeared to give the best loading estimations on an annual basis compared to “true” loadings of the scenarios evaluated. For reference, a comparison of these six methods by year is presented in Appendix D. While annual loadings and discharge varied greatly among years and sites, these factors did not seem to impose a pattern on the percent error indicated by the various loading estimation methods. 46 CHAPTER 7 How Long to Monitor The necessary duration of sampling can vary greatly depending on the purpose of the monitoring program. For reconnaissance monitoring or synoptic monitoring where the purpose is to provide a “snapshot” of current conditions for comparison between a number of sites, monitoring may occur in one day. To evaluate increasing or decreasing trends in water quality, many years of data are needed along with ancillary information on discharge and weather conditions. When monitoring has the purpose of measuring water quality improvement after a change in management practices, establishment of baseline (or impaired) conditions should first occur, so the effectiveness of management control practices can be determined. The duration of sampling for water quality improvement may take many years depending on the level of impairment, the ability of the water body to recycle assimilated pollutants, the effectiveness of control measures implemented, and the source of the pollutants. In determining the duration of a monitoring program, some fairly simple formulas are available, although these formulas assume a fairly sophisticated knowledge of the system being monitored. A key to these formulas is an estimate of the variance expected in water quality data for a given constituent without the variance due to any change or trend. Preliminary monitoring may be necessary to obtain an estimate of this variance. These formulas allow an estimation of the time needed for sampling to statistically detect potential trends or changes in water quality. Minimizing both the time and sampling effort involved can be very important, not only in minimizing cost of sampling, but also in determining whether or not implemented control practices are sufficient to meet target water quality goals. If implemented control practices are insufficient, it is better to find out earlier in an implementation program than later. For example using a paired watershed design with sampling sites SF020 and NF020, it might be desirable to know how much improvement in PO4-P and TP concentrations at site NF020 would be needed to detect a statistical change after the implementation of phosphorus control practices over a three-year period. Sites NF020 and SF020 have intermittent flow and are generally dry between storm events, so storm event mean concentrations (EMCs) are used for comparison. Between 1993 and 1999, 42 paired events were monitored at NF020 and SF020 or an average of 7 storms per year. Using the following formula as presented by Clausen and Spooner (1993), we can obtain an estimate of the smallest difference that would be statistically meaningful based on one, two, and three years of monitoring data. ì 2) ü n1 n2 ï ï MSE 1 ------------- = ----------------- ´ í --------------------------------------------ý 2 n1 + n2 ï æ F d F 1 + --------------------------ö ï î è n 1 + n 2 – 2ø þ 47 Strategies for Monitoring Nonpoint Source Runoff where MSE equals the mean square error from the regression analysis between paired watershed sites prior to treatment, d2 equals the square of the smallest worthwhile difference between pre- and posttreatment dependent mean concentrations, n1 equals the number of storms monitored during the pretreatment period, n2 equals the number of storms monitored during the posttreatment period, and F equals the F-table value for n1 and n2 using a = 0.05. The equation can be rearranged to solve for d to indicate the percent decrease needed to see a significant difference in water quality given only a certain length of monitoring, i.e., one, two, or three years. With EMCs from SF020 as the independent variable and EMCs from NF020 as the dependent variable, our regression analysis for the pretreatment period indicates a MSE of 0.470 and a mean for the dependent variable of 0.05 for PO4-P concentrations based on a natural log transformation of the data. For TP concentrations, a MSE of 0.294 and a mean of the dependent variable of 0.64 was obtained also using a natural log transformation. Assuming 7 storm events per year, in year one a 41 percent change in PO4-P concentrations would be needed to measure a statistically meaningful difference in water quality due to improvement practices (Table 6). By year three, only a 23 percent change in EMCs of PO4-P would be needed, because as the number of posttreatment storm events monitored increases, our ability to measure smaller statistical differences increases, assuming that the inherent variability in the data does not change. For TP, only a 19 percent change in EMCs at NF020 would be needed to statistically measure improvement after three years of storm data collection. Based on this formula, fairly large improvements in water quality would be necessary to show statistically significant improvement in storm event phosphorus concentrations with less than three years of monitoring data. Table 6 Minimal detectable change in storm water quality with varying years of monitoring using equation 2. Analysis based on an a = 0.05. Number of Pre-BMP Number of Post-BMP Storms Sampled Storms Sampled Geometric Mean Pretreatment Concentration (mg/L) Minimum Mean P Posttreatment Concentration Needed to Detect Statistical Differences on Normal Scale (mg/L) Percent Decrease Needed Constituent Years of Monitoring PO4-P 1 2 3 42 42 42 7 14 21 1.06 1.06 1.06 0.62 0.76 0.81 41% 28% 23% TP 1 2 3 42 42 42 7 14 21 1.89 1.89 1.89 1.24 1.46 1.54 34% 23% 19% The above example assumes an average of seven paired storm events per year. From 1993 through 1999, the following number of paired events actually occurred: 48 • 1993 - 6 events • 1994 - 11 events • 1995 - 4 events • 1996 - 3 events • 1997 - 10 events • 1998 - 5 events • 1999 - 4 events Chapter 7 How Long to Monitor The year-to-year variability in paired storm events indicates that more than three years of monitoring may be necessary to obtain an acceptable degree of statistical confidence that a change in water quality has occurred. Drought or wet weather conditions can dramatically influence the duration of monitoring needed to evaluate a change in water quality. Equation 2, and others like it, should, thus, be considered only as a guide in estimating the number of samples needed (or the duration of sampling) for evaluating trends or changes in water quality. An additional factor that should be considered in evaluating duration of sampling with regard to monitoring a change in water quality between paired sites is the range in size of storm events and EMCs measured during the pre- and post-BMP periods. Ideally, the same range of EMCs and storm event sizes should be represented in the posttreatment period as measured in the pretreatment period. This factor may extend the duration of monitoring needed if drought conditions or very wet conditions, which impact runoff volume and concentrations, dominate either the pretreatment or posttreatment period and not the other. For evaluating a trend at a single site, the problem of estimating the duration of sampling needed actually becomes even more difficult. Potential impacts on trend analysis as identified by Gilbert (1987) include the following: • Cyclic patterns—Various cyclic patterns, such as seasonal patterns, can easily be mistaken for trends depending on what time frame is used in the data analysis. • Changes in field or laboratory procedures—These changes may indicate a false step trend in water quality data taken over time, if values are not adjusted for these changes. • Correlated data—Data taken close in time are likely to be correlated and most statistical tests require uncorrelated data. • Relationships of water quality with flow—Adjustments to the data may be needed before assessing trends if flow and water quality are related. • Missing values—Some statistical tests for trend or time series analysis work best if values are evenly spaced. If there are missing values in a data set representing routine monitoring, the performance of the statistical test will depend on the number and location of missing values. • Method detection limits (MDLs)—Values below the method detection limit and changing MDLs throughout a time series of water quality data may mask potential trends. Care should be taken that the MDL used by the laboratory is low enough so that actual trends in the data can be measured. While it is best to consider the above factors prior to starting out on a monitoring program with the goal of measuring long-term trends, often historical data sets must be used that may have been collected for other reasons. Particular care should be used in evaluating historical data, particularly if field or laboratory methods for the data are not clearly indicated. Reviews of statistical methods for evaluating trends are presented by Gilbert (1987) and Esterby (1996), which consider adjustments to help account for most of the factors outlined above. 49 Strategies for Monitoring Nonpoint Source Runoff 50 CHAPTER 8 What Resources are Available for Monitoring The strategies used for monitoring generally involve a comparison of cost with confidence in the data collected. Unnecessary monitoring is wasteful, while insufficient monitoring may lead to false conclusions or an inability to meet desired monitoring objectives. Sampling has fixed and variable costs. Fixed costs are generally associated with such things as sampling equipment. Variable costs are commonly a direct function of the intensity of the monitoring effort. The more samples collected means more trips to a sampling site, an increased consumption of field and laboratory supplies, increased labor costs for sample collection, and an increase in the number of laboratory analyses. A direct relationship often exists between sampling frequency and the confidence level in the information obtained from the data (as an example see Figure 11 on p. 38). As more samples are collected, our confidence in the data and the analysis results increases, and the sampling program is more likely to reflect actual conditions, but this increased confidence must be balanced by the reality of costs and budgets. Questions to answer in weighing the cost of sampling should relate directly to the objectives of the monitoring program. For example: • Will the frequency of sampling be sufficient to meet monitoring objectives within expected confidence limits? • Will the precision or method detection limit used by the laboratory allow monitoring objectives to be met? • Will the constituents and ancillary variables being measured meet monitoring objectives? • Will fixed temporal sampling be sufficient or is automated sampling necessary to meet monitoring objectives? This last question is very important, because automated sampling has its own set of expenses. The inclusion of automated sampling in a monitoring program greatly enhances the chances of collecting samples at elevated flows, which is often necessary to adequately characterize wet weather or storm water contributions associated with nonpoint source pollution. It is often difficult or unsafe to have field personnel manually collecting stream samples during elevated flows. Automated sampling also allows more control of when and at what stream level samples are collected. As shown in Tables 3-5 on pp. 45-46, annual loading estimates are generally underestimated by strictly grab sampling programs. Even when set on a strict biweekly schedule, routine grab samples in the North Bosque River watershed were more likely to represent base flow rather than storm flow conditions. If evaluating the contributions from nonpoint sources is an important part of a monitoring program, then supplementing a routine grab sampling program with storm sampling at some level is recommended.Tables 3-5 indicate that it may not be necessary to monitor every storm event. For sites in the North Bosque River watershed, intensively monitoring the first four 51 Strategies for Monitoring storm events in a year appeared to give a fairly good estimate of the event mean concentration for elevated flows throughout a year. While leading to a larger margin of error in calculated loadings (Tables 3-5), storm chasing may also be an acceptable option for some monitoring programs, if automated sampling equipment is outside the range of the program’s budget. Additional costs associated with automated sampling can include: • Automated sampler and housing for equipment • Level meter for stream stage recordings • Solar recharge panel and battery • Training of personnel on sampler programming and sample retrieval • Increased labor in equipment installation and maintenance • Increased laboratory costs for analyses on an increased number of samples (compositing, preferably by flow-weighting, may alleviate part of the increased laboratory costs) • Increased data storage and data analysis costs with increased data collection Another important consideration is whether a stage-discharge relationship must be developed for a site. A stage-discharge relationship provides information on stream flow which is necessary for calculating loadings. If the sampling site is located at a culvert or another control structure, establishing a stage-discharge relationship may be fairly simple task using established hydrologic equations. If manual measurements are needed to develop a stagedischarge relationship, this can be a very costly and time consuming effort that is often at the whimsy of the weather. Ideally, a number of manual measurements need to be obtained across the range of potential stream stage levels. For highly elevated flows, this may be unsafe or impossible, unless the site is at a bridge. When determining sampling site locations for which flow measurements are necessary, placing a site near or at the location of U.S. Geological Survey (USGS) stream flow station may be an important consideration. With cost saving measures being implemented by almost all government agencies, the USGS has had to reduce the number of sites it monitors and the types of data collected at some sampling sites. For example, at some sites, flow at all stream levels is recorded, while at some stations only “peak” flows are recorded. The USGS is very willing to partner with cooperators and can be approached to assist in obtaining flow data for specific stream sites. Another sampling cost consideration includes site location within the watershed. In practical terms, as the location of sampling sites becomes more spread out and further from headquarters and analytical laboratories, sampling costs increase and sample holding times may become a constraint. Travel times from sampling crew headquarters to sites, travel times between sites, and travel times from sites to delivery at the analytical laboratory need to be considered in relation to project budgets and acceptable holding times for the analysis of different constituents. Shorter holding times and/or increased travel distances can increase the number of personnel needed to retrieve samples in a timely fashion, and also increase the potential need to pay overtime for extended work days or work on weekends or holidays. As the number of field personnel increase, overall salaries and equipment associated with each field crew member also increase as well as the need for more vehicles to transport these people and the maintenance of these vehicles. Wet weather sampling can be quite hard on vehicles, particularly when they need to be taken off road to retrieve samples from more secluded areas. Further cost considerations in developing a monitoring strategy include the resources needed to properly document, store, and analyze collected data. Data management can be a very expensive part of a monitoring program, particularly if very intensive data are collected. Careful attention must be paid to data entry or electronic data transfer from data loggers to 52 Chapter 8 What Resources are Available for Monitoring make sure the collected measurements are as close to error free as possible. If data does not get into the database correctly, subsequent analyses will be flawed and wrong conclusions may be reached. It is important to have properly trained people on staff for data entry, data management, and data analysis. The objectives of most monitoring projects do not end with data collection and laboratory analysis. If data are appropriately collected, properly analyzed in the laboratory, and carefully stored in a database, but then forgotten, they are of little use to anyone. Considering what data are needed to answer the questions associated with the objectives monitoring should be one of the first considerations in developing a monitoring program. Data must be processed into information which hopefully answers the objectives of the monitoring program. Often that information must be published and distributed. Data storage, data analysis, report writing, and publication all have costs that must be considered. The costs for data storage and analysis, again, generally increase with the intensity of the monitoring program and the amount of data collected. The costs associated with report writing, publication, and distribution will depend largely on the audience that must be reached with the water quality information. In addition, many state and federal programs require a quality assurance project plan (QAPP). A QAPP specifies information on standard operating procedures from data collection to data analysis and management. A detailed QAPP is an invaluable document, because it provides clear and concise documentation of procedures used to control the quality of the data collected and helps ensure that monitoring will meet the stated objectives. This documentation can be used to ensure consistency in methods over time, and also allows others to evaluate and use the data with a thorough knowledge of what the data represents. Although the QAPP document should be considered indispensable, it represents a cost in time and money that must be considered up front in the planning process of a monitoring project. 53 Strategies for Monitoring 54 CHAPTER 9 Summary & Conclusions Defining appropriate water quality monitoring strategies for nonpoint source pollution can be quite challenging. Nonpoint source pollution is inherently dynamic occurring largely during rainfall-runoff events and will generally only be quantifiable through intensive storm water (or wet-weather) monitoring. Routine grab sampling, which is often used for assessing ambient or base flow conditions, will be inadequate to characterize nonpoint source pollution in most instances. Wet-weather monitoring will be a critical aspect of most watershed approaches concerned with water quality impairments from nonpoint sources and is, most certainly, a linchpin of the Planned Intervention Microwatershed Approach that was refined and demonstrated under this project’s funding. Under such an approach, monitoring program objectives will typically involve assessing present water quality conditions in a watershed (e.g., levels of impairment and sources) and then measuring progress towards remediation of impairments once control actions are implemented. Other objectives may also be pertinent to a particular watershed, such as obtaining data for testing of computer models. The monitoring plan used to meet these objectives can be defined through six primary points of consideration: 1) what to monitor, 2) where to monitor, 3) what type of monitoring to implement, 4) how often to monitor, 5) how long to monitor, and 6) what resources are available for monitoring. This report reviewed these six considerations in developing a monitoring plan with particular emphasis on wet-weather monitoring of nonpoint source contributions. Major points associated with each consideration are summarized below: • What to monitor? ¤ In defining what to monitor, it is important to consider not only the primary variable of interest, say the abundance of algae in an area with algal bloom problems, but also ancillary variables, such as nutrients, that may be influencing the variable of interest. Often water quality problems are in response to an imbalance caused by some other factor, and these cause and effect relationships may need to be defined. ¤ Stream flow or discharge is a critical parameter for consideration in development of a monitoring plan involved in measuring nonpoint source contributions. Accurate stream flow measurement when obtaining a water quality sample, however, can significantly increase the time and instrumentation required for sample collection. Continuous stream flow measurement requires continuous measurement of water level and either the development of a stage-discharge relationship or presence of a control point (e.g., a flume or weir) for which the stage-discharge relationship is known. Continuous stream flow data are necessary to calculate loadings for a waterborne constituent. Stream flow at the time of water quality sampling also adds important information for subsequent data interpretation, because concentrations of most waterborne constituents respond to variations in stream flow. 55 Strategies for Monitoring ¤ • 56 If mathematical models are to be used to help meet objectives, model selection should precede definition of the variables to be monitored. Many sophisticated water quality models are available that can greatly aid in providing information to the watershed planning process, but these models often require extensive watershed specific data. Monitoring data may be needed to tune specific model coefficients and to provide data for model calibration and verification. Where to monitor? ¤ In determining sampling site locations, it is important to consider stream transport and distance factors that may influence transformations of pollutants. The proximity of a sampling site to a pollutant source will greatly influence the magnitude of concentrations measured. As more distance occurs between the pollutant source and the sampling location, not only is dilution of the pollutant expected by mixing with other waters, but also there is a greater travel time, which allows for physical and biochemical processes to transform pollutants. ¤ To define sources of nonpoint source pollution it may be impractical to monitor individual sources, because even small watersheds generally represent a mix of different land uses. When monitoring heterogeneous land-use watersheds care should be taken in attributing sources, because relatively small land areas may disproportionately contribute pollutants. Also, individual sources will typically be indistinguishable once they have reached a stream. Edge of field monitoring may be required in some situations and in other situations statistical regression analysis of results from sampling several heterogeneous land-use watersheds may be appropriate wherein land uses above each sampling site become the independent variables (e.g., McFarland and Hauck, 1999). Uncertainty and statistical validity should be considered, and mathematical models may be needed to support findings. Though not discussed in this report, pollutant source identification can make use of technologies such as nitrogen isotope analyses, DNA techniques for microorganisms, and monitoring of surrogate constituents, such as various antibiotics uniquely associated with animal feeding operations. ¤ Adequate distance should be allowed for complete mixing when locating sampling sites below point source discharges and in a similar fashion, below the merging of two streams. In these situations, a mixing zone will exist for some distance downstream before the separate waters completely mix, and sampling in a mixing zone should generally be avoided. ¤ The possibility of spatial variations in constituent concentrations within a water body should be contemplated when deciding where to sample. A few relevant examples of when a simple grab sample may not prove to be representative are provided. During storm events, constituents associated with sediment will often be unequally distributed through the stream cross section with greater concentrations found at greater depths and near the center of flow. Estuarine and tidal rivers often experience salinity variations with depth, which can also cause sufficient density stratification to induce vertical variation in other water quality constituents. Deeper lakes and reservoirs will experience seasonal temperature stratification, which also induces vertical stratification and water quality variation. Lakes often have a longitudinal water quality gradient from inlet of streams to the main body of the lake. Chapter 9 ¤ • • Summary & Conclusions The upstream-downstream approach and paired watershed approach are two statistical designs recommended to evaluate effectiveness of nonpoint source management practices. Both approaches use a control-sampling site and a treatmentsampling site in conjunction with pretreatment and posttreatment sampling periods to isolate water quality changes from an area receiving improvements in management practices. Proper location of sampling sites is critical to the success of both approaches as described in guidance documents (e.g., Spooner et al., 1985; Spooner and Line, 1993). What type of monitoring to implement? ¤ In many stream systems flow is dominated by surface flow originating from rainfall and runoff, and groundwater contributions are of secondary importance. To characterize nonpoint source pollution in these streams, wet weather sampling will typically be a necessity. The North Bosque River exemplifies these surface runoff dominated systems found in many parts of our nation. ¤ The use of automated sampling equipment is recommended for characterizing storm events. Water quality can be highly variable both within an individual storm event and between different storm events. Consequently, accurate characterization of water quality from storm events will require multiple samples collected during each event and the sampling of more than one event. Automated sampling is also more reliable and safer than manual grab sampling during periods of elevated flows. ¤ If it is necessary to composite storm samples, flow weighting rather than time weighting should be used to obtain a more accurate characterization of overall storm concentrations. Compositing samples is a systematic method of combining a number of individual samples into one representative sample and thereby reducing analytical (laboratory) costs. Flow-weighted samples can also be used in loading calculations, while time-weighted samples are not related to stream discharge and, thus, have no intrinsic relationship to loading. ¤ The equipment for real-time continuous monitoring is rapidly improving and may be an important consideration for aiding wet-weather sampling or for evaluating changes in water quality over time. How often to monitor? ¤ In general, confidence in results will increase as the frequency of sampling increases. ¤ In comparison to biweekly sampling, percent errors in the calculation of annual average stream concentrations increased greatly when monthly or quarterly sampling was considered. ¤ For storm sampling, it may be possible to monitor a subset of storm events during the year rather than every storm event. For selected sampling sites in the North Bosque River watershed where on average about 16 storm events occur per year, monitoring of 4 to 6 events per year provided a percent error of about 20 percent in the calculation of annual storm concentrations of PO4-P, TP, and TSS. ¤ For calculating annual loadings, some storm monitoring will be necessary in areas where nonpoint source contributions are important. An integration approach using an overall flow-weighted mean of the first four storm events to complement biweekly grab sampling appeared to give the best loading estimations, on an annual basis, compared to "true" loadings of the sampling schemes evaluated for sites in the North Bosque River watershed. 57 Strategies for Monitoring • • 58 How long to monitor? ¤ To reasonably characterize the water quality of a system with important nonpoint source pollution contributions will usually require a minimum of two to three years of monitoring and perhaps even more time if the period sampled is atypically wet or dry. The inherent variability of rainfall runoff will typically necessitate such a commitment if meaningful results are to be obtained. ¤ It may take several years to measure change or improvement in water quality associated with implementation of control practices. Estimates of the needed duration of sampling can be calculated if an estimate of the expected variance in the constituent of interest can be obtained. This variance may be estimated from preliminary or pretreatment monitoring or obtained from other monitoring programs. ¤ The duration of posttreatment monitoring is dependent on the percent change in water quality expected from control practices. In general, the larger the impact of the control practice, the shorter the duration of monitoring that should be needed to measure significant improvements in water quality. ¤ Year-to-year variability in weather conditions can greatly influence the duration of monitoring needed to statistically evaluate significant trends or improvements in water quality with control practices. Care should be taken to make sure that changes in water quality are actually associated with the control practices being implemented and are not due to changes in weather conditions or other factors between the pre and posttreatment monitoring periods. What resources are available for monitoring? ¤ The resources needed to meet monitoring objectives will need to be balanced with constraints imposed by budgets and available equipment, instruments, and personnel. Monitoring resources are directly related to the frequency of sampling needed, the precision with which data must be analyzed, the number of ancillary variables that need to be monitored, and whether manual or automated sampling needs to be implemented. Automated sampling, while expensive, greatly enhances the chances of collecting samples at elevated flows, which may be necessary to properly characterize storm water contributions associated with nonpoint source pollution. ¤ For ease in obtaining stage-discharge information, efforts should be made to locate sampling sites at control structures (e.g., road culverts) or established USGS monitoring sites. If this is not possible, time and resources will need to be allocated for manual measurements of flow to establish site-specific, stage-discharge relationships, if discharge data are required. ¤ Travel times between sampling sites as well as from headquarters to sampling sites and to the analytical laboratory should be considered in relation to holding times for the constituents of interest. ¤ Resources available for monitoring should consider data storage, analysis, and publication as well as the resources needed for sample collection and laboratory analysis. While project dependent, these items may represent significant investment of resources. ¤ A quality assurance project plan (QAPP) should be a mandatory resource available as part of all monitoring programs to help ensure that the sampling strategies used will meet the objectives of monitoring and aid in the watershed planning process. References Adams, T. and A. McFarland. 2001. Semiannual Water Quality Report for the Bosque River Watershed (Monitoring Period: January 1, 1996-December 31, 2000). Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas, TR0105 (July 2001). APHA, American Public Health Association, American Water Works Association, and Water Environment Federation. 1995. Standard Methods for the Examination of Water and Wastewater, 19th edition. APHA, Washington, D.C. Arnold, J.G., R. Srinivasan, R.S. Muttiah, and J.R. Williams. 1998. Large area hydrologic modeling and assessment. Part I: Model development. Journal of the American Water Resources Association 34:73-89. Barbour, M.T., J. Gerritsen, B.D. Snyer, and J.B. Stribling. Rapid Bioassessment Protocols for Use in Streams and Wadeable Rivers: Periphyton, Bethic Macroinvertebrates and Fish, 2nd Ed. U.S. Environmental Protection Agency, Office of Water, Washington, D.C., EPA 841-B-99-002. Boyd, C.E. 1990. Water Quality in Ponds for Aquaculture. Birmingham Publishing Co., Birmingham, Alabama. Bradu, D. and Y. Mundlak. 1970. Estimation in lognormal linear models. Journal of the American Statistical Association 65:198-211. Chapra, S.C. 1997. Surface Water Quality Modeling. McGraw-Hill, New York, New York. Christensen, V.G., Jian, Xiaodong, and Ziegler, A.C. 2000. Regression analysis and real-time water-quality monitoring to estimate constituent concentrations, loads, and yields in the Little Arkansas River, southcentral Kansas, 1995-99. U.S. Geological Survey Water-Resources Investigations, Report 00-4126. Clausen, J.C. 1996. National Handbook of Water Quality Monitoring. Part 600, National Water Quality Handbook, United States Department of Agriculture Natural Resources Conservation Service, Washington, D.C. Clausen, J.C. and J. Spooner. 1993. Paired Watershed Study Design. U.S. Environmental Protection Agency, Office of Water, Washington, D.C., EPA 841-F-93-009. Coan, T. and L. Hauck. 1996. Interim Biological Data Analysis Report: Upper North Bosque River Watershed Agricultural Nonpoint Source Studies September 1, 1992 - January 31, 1994. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas, WP 96-03 (October 1996). Cohn, T.A., D.L. Caulder, E.J. Gilroy, L.D. Zynjuk, and R.M. Summers. 1992. The validity of a simple statistical model for estimating fluvial constituent loads: An empirical study involving nutrient loads entering Chesapeake Bay. Water Resources Research 28:2353-2363. Cohn, T.A., L.L. DeLong, E.J. Gilroy, R.M. Hirsch, and D. Wells. 1989. Estimating constituent loads. Water Resources Research 25:937-942. Cole, T.M. and E.M. Buchak. 1995. CE-QUAL-W2: A Two-Dimensional, Laterally Averaged, Hydrodynamic and Water Quality Model, Version 2.0, User Manual. Instruction Report EL-95-1. U.S. Army Corps of Engineers, Waterways Experiment Station, Vicksburg, Mississippi. Dávalos-Lind, L.D. and O.T. Lind. 1998. The Algal Growth Potential of and Growth-Limiting Nutrients in Lake Waco and its Tributary Waters. An Interim report to TIAER. Limnology Laboratory Dept. of Biology, Baylor University, Waco, Texas. 59 Strategies for Monitoring Nonpoint Source Runoff Davis, W.S. and T.P. Simon (eds). 1994. Biological Assessment and Criteria: Tools for Water Resource Planning and Decision Making. Lewis Publishers, Boca Raton, Florida. Duan, N. 1983. Smearing estimate: A nonparametric retransformation method. Journal of the American Statistical Association 87:605-610. Easterling, N. 2000. Future Growth Projections for the Lake Waco/Bosque River Watershed. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas, WP0005 (May 2000). Esterby, S.R. 1996. Review of methods for the detection and estimation of trends with emphasis on water quality applications. Hydrological Processes 10:127-149. Froelich, P.H. 1988. Kinetic control of dissolved phosphorus in natural rivers and estuaries: A primer on the phosphate buffer mechanism. Limnology & Oceanography 33:649-668. Gilbert, R. O. 1987. Statistical Methods for Environmental Pollution Monitoring. Van Nostrand Reinhold Company, New York, NY. Gilliom, R.J. and D.R. Helsel. 1986. Estimation of distributional parameters for censored trace level water quality data. 1. Estimation techniques. Water Resources Research 22:135-126. Hach. 1991. DR/2000 Spectorphotometer Procedures Manual. Hach Company, Ames, IA. Harris, W.G., H.D. Wang, and K.R. Reddy. 1994. Dairy manure influence on soil and sediment composition: Implications for phosphorus retention. Journal of Environmental Quality 23:1071-1081. Hendon, T., F. Mitchell, and A. McFarland. 1998. Biological Data Analysis Report Volume 1: Upper North Bosque River Watershed Agricultural Nonpoint Sources Studies, September 1992-November 1995. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas (PR 99-04, August 1998). Hirsch, R.M., D.R. Helsel, T.A. Cohn, and E.J. Gilroy. 1993. Statistical Analysis of Hydrologic Data, pp. 17.1-17.55. In: Handbook of Hydrology (ed. D.R. Maidment). McGraw-Hill, Inc., New York, New York. Kiesling, R.L., A.M.S. McFarland, and L.M. Hauck. 2001. Nutrient Targets in the Bosque River Watershed: Developing Ecosystem Restoration Criteria for a Eutrophic Watershed. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas, TR0107 (October 2001). Klemm, D.J., P.A. Lewis, F. Fulk, and J.M. Lazorchak. 1990. Macroinvertebrate Field and Laboratory Methods for Evaluating the Biological Integrity of Surface Waters. U.S. Environmental Protection Agency, Environmental Monitoring and Support Laboratory, Cincinnati, Ohio, EPA-600-4-90-030. Matlock, M.D. and A.D. Rodriguez. 1999. Preliminary Report of Findings: Periphytometer Study on Streams in the Lake Waco/Bosque River Watershed. Prepared for the Texas Institute for Applied Environmental Research, Stephenville, Texas. Dept. Agricultural Engineering, Texas A&M University, College Station, Texas (February 1999). McFarland, A. and L. Hauck. 1999. Existing Nutrient Sources and Contributions to the Bosque River Watershed. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas, PR9911 (September 1999). McFarland, A.M.S. and L.Hauck. 1995a. Evaluating the Effectiveness of Flood Water Retention Reservoirs as Integrators of Water Quality, pp. 231-238. In: Animal Waste and the Land-Water Interface (ed. K. Steele). Lewis Publishers, Boca Raton, FL. McFarland, A. and L. Hauck. 1995b. Scientific Underpinnings for Policy Analysis: Analysis of Agricultural Nonpoint Pollution Sources and Land Characteristics. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas (September 1995). 60 References McFarland, A., L. Hauck, and R. Kiesling. 2001. Fate and Transport of Soluble Reactive Phosphorus in the North Bosque River of Central Texas. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas (TR0101, February 2001). Montgomery, H.A.C. and I.C. Hart. 1974. The design of sampling programs for rivers and effluents. Journal of the Institute of Water Pollution Control 33:77-101. Peterson, B.J., W.M. Wolheim, P.J. Mulholland, J.R. Webster, J.L Meyer, J.L Tank, E. Marti, W.B. Bowden, H.M. Valett, A.E. Hershey, W.H. McDowell, W.K. Dodds, S.K. Hamilton, S. Gregory, and D.D. Morrall. 2001. Control of nitorgen export from watersheds by headwater streams. Science 292:86-90. Petrick, D. 2001. Personal communication. Surface Water Quality Monitoring Program, TNRCC, Austin, TX. Preston, S.D, V.J. Bierman, Jr., and S.E. Silliman. 1989. An evaluation of methods for the estimation of tributary mass loads. Water Resources Research 25:1379-1389. Rand, G.M. and S.R. Petrocelli (eds). 1985. Fundamentals of Aquatic Toxicology: Methods and Applications. Hemisphere Publishing Corporation, Washington, D.C. Reckhow, K.H., M.N. Beaulac, and J.T. Simpson. 1980. Modeling Phosphorus Loading and Lake Response under Uncertainty: A Manual and Compilation of Export Coefficients. U.S. Environmental Protection Agency, Clean Lakes Section, Washington, D.C., EPA-440/5-80-011. Reddy, K.R., E. Flaig, L.J. Scinto, O. Diaz, and T.A. DeBusk. 1996. Phosphorus assimilation in a stream system of the Lake Okeechobee basin. Water Resources Bulletin 32:901-915. Richards, R.P. and J. Holloway. 1987. Monte Carlo studies of sampling strategies for estimating tributary loads. Water Resources Research 23:1939-1948. Robertson, D.M., and E.D. Roerish. 1999. Influence of various water quality sampling strategies on loads estimates for small streams. Water Resources Research 35:3747-3759. Rottler, C., R. Jones, and J. McNitt. 1999. The Planned Intervention Microwatershed Approach (PIMA) in TMDLs: Reducing Agricultural Nonpoint Source Pollutants. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas, PR9907. Sanders, T.G., D. Adrain, and J.M. Joyce. 1977. Mixing length for representative water quality sampling. Journal of the Water Pollution Control Federation 49:2467-2478. Sanders, T.G., R.C. Ward, J.C. Loftis, T.D. Steele, D.D. Adrain, and V. Yevjevich. 1990. Design of Networks for Monitoring Water Quality, 3rd Ed. Water Resources Publications, Littleton, CO. Shoemaker, L, M. Lahlou, M. Bryer, D. Kumar, and K. Kratt. 1997. Compendium of Tools for Watershed Assessment and TMDL Development. Office of Water, Washington, D.C., EPA841-B-97-006. Spooner, J. and D.E. Line. 1993. Effective monitoring strategies for demonstrating water quality changes from nonpoint source controls on a watershed scale. Water Science and Technology 28:143148. Spooner, J., R.P. Maas, S.A. Dressing, M.D. Smolen, and F.J. Humenif. 1985. Appropriate designs for documenting water quality improvements from agricultural NPS control programs, pp. 30-34. In: Perspectives on Nonpoint Source Pollution. U.S. Environmental Protection Agency, EPA 440-5-85-001. Stein, S.K. 1977. Calculus and Analytic Geometry, second edition. McGraw-Hill Book Company, New York, New York, pp. 421-427. TIAER, Texas Institute for Applied Environmental Research. 1998. Quality Assurance Project Plan for the United States Department of Agriculture Bosque River Initiative. Texas Institute for Applied Environmental Research, Tarleton State University, Stephenville, Texas. 61 Strategies for Monitoring Nonpoint Source Runoff Thomann, R.V. and J.A. Mueller. 1987. Principles of Surface Water Quality Modeling and Control. Harper and Row Publishers, Inc., New York, New York. TNRCC, Texas Natural Resource Conservation Commission. 2001. Two Total Maximum Daily Loads for Phosphorus in the North Bosque River for Segments 1226 and 1255. Strategic Assessment Division, TMDL Team, TNRCC, Austin, Texas (February 2001). TNRCC, Texas Natural Resource Conservation Commission. 1999. Surface Water Quality Monitoring Procedures Manual. TNRCC, Austin, Texas, GI-252 (June 1999). USEPA, United States Environmental Protection Agency. 2000a. Nutrient Criteria Technical Guidance Manual: Rivers and Streams. Office of Water and Office of Science and Technology, USEPA, Washington, D.C., EPA-822-B-00-002 (July 2000). USEPA, United States Environmental Protection Agency. 2000b. Nutrient Criteria Technical Guidance Manual: Lakes and Reservoirs. Office of Water and Office of Science and Technology, USEPA, Washington, D.C., EPA-822-B-00-001 (April 2000). USEPA, United States Environmental Protection Agency. 1990. Biological Criteria: National Program Guidance for Surface Waters. Office of Water, Washington, D.C. EPA-440-5-90-004. USEPA, United States Environmental Protection Agency. 1983. Methods for Chemical Analysis of Water and Wastes. Environmental Monitoring and Support Laboratory, Office of Research and Development, USEPA, Cincinnati, OH. EPA-600/4-79-020, Revised March 1983. USGS, United States Geological Survey. 1999. National Field Manual for the Collection of Water-Quality Data. U.S. Geological Survey Techniques of Water-Resources Investigations. Ward, R.C., J.C. Loftis, H.P. DeLong and H.F. Bell. 1988. Groundwater quality: A data analysis protocol. Journal of the Water Pollution Control Federation 60:1938-1945. 62 APPENDIX A Sampling Sites, Weather, and Monitoring Protocols Sampling Sites For this report, data analyses evaluating monitoring strategies focus specifically on seven stream sites representing the diversity of land uses and flow conditions within the watershed of the North Bosque River (Tables A-1 and A-2). These sites are BO040 on the North Bosque River below Stephenville, BO070 on the North Bosque River above Hico, BO090 on the North Bosque River near Clifton, GC100 on Green Creek, NC060 on Neils Creek, NF020 on the North Fork of the North Bosque River, and SF020 on the South Fork of the North Bosque River (see Figure 1, p. 14). Table A–1 Site Estimated land uses for the drainage area above selected sampling sites. Wood/Range (%) Pasture (%) Dairy Waste Application Fields (%) Urban (%) Other (%) Area (hectares) 8.4 6.5 8.9 11.7 7.2 3.7 3.8 1.7 1.5 1.2 1.0 0.2 25,719 93,248 253,740 7.2 12.8 6.9 0.0 0.7 0.0 0.7 0.2 26,165 35,147 6.8 1.0 45.4 0.0 0.0 0.0 0.5 0.2 791 848 Cropland (%) Sites on the North Bosque River BO040 51.1 23.8 BO070 68.2 15.4 BO090 71.9 13.8 Site on Major Tributaries to the North Bosque River GC100 71.2 13.3 NC060 76.5 10.5 Sites on Microwatershed Stream Segments NF020 38.8 8.5 SF020 96.1 2.7 Sites BO040, BO070, and BO090 represent sites with predominately perennial flow along the North Bosque River. Site BO040 is located about 0.4 kilometers (0.25 miles) below the Stephenville wastewater treatment plant (WWTP) discharge. The water quality at BO040, thus, reflects the point source discharge from the WWTP at base flow mixed with nonpoint sources during storm events. Site BO070, near Hico, is located above the Hico WWTP discharge, thus, making the Stephenville WWTP discharge the only point source discharge permitted above this site. Site BO090 is located above the Clifton WWTP discharge, but integrates permitted point source discharges from the WWTPs of the cities of Stephenville, Hico, Iredell, and Meridian. Particularly during the dry summer when evaporation is high, flows at BO040, BO070, and BO090, if they occur, are dominated by discharge contributions from the WWTPs located along the river. Neils Creek and Green Creek are major tributaries to the North Bosque River. Site NC060 on Neils Creek is considered a least impacted site and used as a reference site within the 63 Strategies for Monitoring Nonpoint Source Runoff Table A–2 Daily streamflow statistics by year. Site Year Average (m3/s) Median (m3/s) Minimum (m3/s) Maximum (m3/s) Total Volume (m3) BO040 BO040 BO040 BO040 BO040 1995 1996 1997 1998 1999 0.89 1.02 1.24 0.80 0.23 0.22 0.32 0.15 0.18 0.13 0.04 0.04 0.02 0.05 0.04 31.5 13.8 55.7 50.8 2.85 28,100,000 32,200,000 39,200,00 25,100,000 7,300,000 BO070 BO070 BO070 BO070 BO070 1995 1996 1997 1998 1999 4.53 2.44 7.46 4.08 0.91 2.40 0.63 2.65 2.17 1.10 0.10 0.00 1.19 0.38 0.00 116 74.9 197 159 9.03 143,000,000 77,000,000 235,000,000 129,000,000 29,700,000 BO090 BO090 BO090 BO090 BO090 1995 1996 1997 1998 1999 14.31 4.22 19.95 8.95 1.45 5.26 0.93 4.41 1.47 0.38 0.01 0.07 0.39 0.02 0.01 461 183 677 1420 186 451,000,000 133,000,000 629,000,000 282,000,000 45,800,000 GC100 GC100 GC100 GC100 GC100 1995 1996 1997 1998 1999 1.27 0.48 2.55 0.69 0.02 0.45 0.17 0.36 0.00 0.00 0.00 0.00 0.00 0.00 0.00 35.0 29.1 84.6 44.9 4.15 39,900,000 15,200,000 80,400,000 21,600,000 713,000 NC060 NC060 NC060 NC060 NC060 1995 1996 1997 1998 1999 naa 0.94 5.46 2.73 0.29 na 0.14 1.06 0.74 0.20 na 0.00 0.20 0.00 0.00 na 100 252 474 11.9 na 29,900,000 172,000,000 86,200,000 9,050,000 NF020 NF020 NF020 NF020 NF020 1995 1996 1997 1998 1999 0.03 0.03 0.07 0.04 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 2.83 1.15 6.49 5.40 0.07 809,000 852,000 2,100,000 1,110,000 14,800 SF020 SF020 SF020 SF020 SF020 1995 1996 1997 1998 1999 0.03 0.04 0.05 0.02 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 2.97 2.34 4.68 3.05 0.08 1,080,000 1,250,000 1,610,000 552,000 14,000 a. na, not applicable. Discharge data collection was not initiated at site NC060 until October 5, 1995. watershed (Coan and Hauck, 1996), while the drainage area above GC100 contains a moderate number of dairies and land area associated with dairy waste application. Green Creek is considered intermittent with perennial pools, while Neils Creek is considered to have perennial flow (D. Petrick, personal communication, 2001). The North Fork and South Fork of the North Bosque River are considered microwatershed sites. A microwatershed, in this context, is defined as a drainage area small enough to deal with all landowners within the drainage area when intervention is needed for water quality improvement (Rottler et al., 1999). The drainage area above NF020 is heavily impacted by dairy operations, while the drainage area above SF020 has very limited impacts associated with intensive agriculture and is primarily wood and range land. Both NF020 and SF020 are intermittent and often dry between rainfall runoff events. 64 Appendix A Sampling Sites, Weather, and Monitoring Protocols Precipitation and Hydrologic Conditions Most of the monitoring data presented were collected between January 1, 1995 and December 31, 1999. During that time weather conditions and, thus, hydrologic conditions varied considerably from month to month (Figure A–1). Precipitation was punctuated by several months of well above average rainfall often followed by periods of drought. Notably high precipitation occurred in August 1995, August 1996, and February 1997 at both the Stephenville and Waco Dam sites, indicating that above average precipitation was probably occurring throughout the watershed (Figure A–1). The period December 1997 through March 1998 was also abnormally wet, although extended periods of below average precipitation occurred October 1995 through June 1996, July 1997 through November 1997, April 1998 through August 1998, and through much of 1999. Figure A–1 Monthly precipitation and departure from normal for Stephenville and Waco Dam National Weather Service observer sites. Long-term average or normal monthly precipitation calculated for 1967-1996. 65 Strategies for Monitoring Nonpoint Source Runoff Annual stream-flow statistics for sampling sites BO040, BO070, BO090, GC100, NC060, NF020, and SF020 indicate the lowest flow occurring in 1999 across sites and the highest flows occurring in 1997 (Table A-2). Distinct wet and dry periods occur between 1995 and 2000, which should be considered in evaluating any water quality data set. Average daily discharge for each site is, thus, graphically presented in Appendix B for reference. Data Collection Methods and Procedures Data collection at sites BO040, BO070, BO090, GC100, NC060, NF020, and SF020 consists of routine grab samples collected on a biweekly basis and storm samples collected by a programmed automated sampler. Routine grab samples consist of a single, representative sample taken at 0.1 to 0.3 meters (0.25 to 1.0 ft.) below the surface of the water. Routine sampling occurs on a set biweekly schedule and, thus, does not necessarily represent only base flow conditions. A grab sample is not taken during routine monitoring if a site is dry or if water is pooled at a site and not flowing. During low flow conditions, riffle areas near each sampling site are observed to determine the presence or absence of flow. Storm samples are collected using automated sampling equipment consisting of an ISCO 3230 or 4230 bubbler-type flow meter and an ISCO 3700 automated sampler enclosed in a sheet metal shelter. The flow meter measures the pressure required to force an air bubble through a 3-mm (0.125-in) polypropylene tube (bubbler line) and records this pressure as water level. The flow meters are programmed to record stream water level (stage) and are used to initiate sample retrieval by the automated samplers when a certain stream level is obtained. Each flow meter is programmed to record water level at five-minute intervals and typically actuates storm sampling when a stream rise of 4-cm (1.5 in) above the bubbler datum is registered. The actuation level was selected by trial and error as the lowest level at which the sampler would actuate for rainfall-runoff events while avoiding undesired actuation from nonrainfall event causes such as wave action. Electrical power is provided to the flow meter and sampler by marine deep-cycle batteries with recharge provided by solar cells. Most storm samples were collected using a sequential sampling sequence. The sampling sequence for sites NF020 and SF020 included an initial sample, three samples taken at onehour intervals, two samples taken at two-hour intervals, and all remaining samples taken at eight-hour intervals. At sites BO040, BO070, BO090, GC100, and NC060 the sampling sequence was set to retrieve an initial sample, then a single sample at one-hour, two-hour, three-hour, four-hour, and six-hour intervals with all remaining samples taken at eight-hour intervals. A more frequent collection sequence was used at sites NF020 and SF020, because these two sites are in the headwaters and respond much more quickly to rainfall-runoff events. These sampling sequences allow for more frequent sample collection during the typical rapid hydrograph rise and peak periods following sampler actuation and less frequent sample collection during the longer, receding portion of a storm hydrograph. In the laboratory, most samples collected between 1995 and 1999 were analyzed separately as individual samples. In January 1995 through July 1995, flow-weighting at the sampler was implemented at sites BO040, BO070, and GC100 by programming the sampler to take a sample after a given amount of flow had passed by. This procedure was problematic in determining a flow volume for sampling that would capture both small and large storm events. If the flow volume between samples was sufficiently small to capture small runoff events, there would be inadequate sample bottles (24 bottles in a sampler) to capture large events, If the flow volume was set to allow characterization of large events, no sample would be collected for small events. In mid-1997, daily flow-weighting of sequential samples was initiated again at sites BO040, GC100, and NF020 to help decrease the number of samples 66 Appendix A Sampling Sites, Weather, and Monitoring Protocols analyzed. This time samples were collected using the sequential sequence, but before being processed in the laboratory, samples collected on a given day were combined into a single flow-weighted sample based on continuous five-minute discharge information using a midpoint rectangular integration method between collection times. Water level data was converted to discharge using site specific stage-discharge relationships. For stream sites BO040, GC100, NC060, NF020, and SF020, stage-discharge relationships were developed from manual measurements of flow at various stream levels. At site BO070, the USGS stage-discharge relationship for site 08094800 was used in conjunction with measured level data to calculate flow. At site BO090, fifteen-minute discharge data from corresponding USGS site 08095000 were obtained for use in this report. Routine grab and storm water samples were analyzed for soluble ammonia-nitrogen (NH3N), soluble nitrite-nitrogen plus nitrate-nitrogen (NO2-N+NO3-N), total Kjeldahl nitrogen (TKN), soluble orthophosphate phosphorus (PO4-P), total phosphorus (TP), and total suspended solids (TSS). In addition, conductivity, dissolved oxygen (DO), pH, and water temperature values were measured about one-foot below the surface at stream sites when routine grab samples were collected using a Hydrolab RecorderTM Water Quality Multiprobe Logger connected to a Scout®2 Display (Hydrolab Corporation, Austin, TX). Samples analyzed for chlorophyll-a (CHLA) were also collected with routine grab samples but on a monthly rather than biweekly basis. Data analyses for this report focus primarily on PO4-P, TP, and TSS, because excessive phosphorus is considered the major water quality problem in the North Bosque River watershed associated with excessive algal growth (TNRCC, 2001). All sampling and analyses were conducted under EPA or TNRCC approved quality assurance project plans (e.g., TIAER, 1998). Methods of laboratory analysis are outlined in Table A-3. Table A–3 Analysis methods for water quality constituents. Constituent Field Measurements Conductivity Dissolved Oxygen pH Water Temperature Laboratory Measurements Ammonia-Nitrogen Chemical Oxygen Demand Chlorophyll-a Fecal Coliform Nitrite-Nitrogen + Nitrate-Nitrogen Total Kjeldahl Nitrogen Orthophosphate-Phosphorus Total Phosphorus Total Suspended Solids Method Source SM 2510B EPA 360.1 EPA 150.1 EPA 170.1 APHA (1995) USEPA (1983) USEPA (1983) USEPA (1983) EPA 350.1 Hach 8000 SM 10200H SM 9222D EPA 353.2 EPA 351.2 EPA 365.2 EPA 365.4 EPA 160.2 USEPA (1983) Hach (1991) APHA (1995) APHA (1995) USEPA (1983) USEPA (1983) USEPA (1983) USEPA (1983) USEPA (1983) In data management, left censored data (values measured below the laboratory method detection limit) were entered into the water quality database as one-half the method detection limit (MDL) as recommended by Gilliom and Helsel (1986) and Ward et al. (1988). MDLs were evaluated about every six months by TIAER’s water quality laboratory (Table A-4). 67 Strategies for Monitoring Nonpoint Source Runoff Table A–4 68 Laboratory method detection limits effective March 1996 through December 1999. Effective Dates NH3-N (mg/L) NO2-N+NO3-N (mg/L) PO4-P (mg/L) Total-P (mg/L) TKN (mg/L) COD (mg/L) TSS (mg/L) CHLA (µg/L) 19 Sep 95 - 20 Mar 96 21 Mar 96–27 Oct 96 28 Oct 96–04 May 97 05 May 97–30 Nov 97 01 Dec 97–16 Dec 97 17 Dec 97–14 May 98 15 May 98–30 Nov 98 01 Dec 98–01 Jun 99 02 Jun 99–10 Dec 99 11 Dec 99–31 Dec 99 0.016 0.015 0.037 0.022 0.022 0.022 0.009 0.015 0.030 0.037 0.004 0.015 0.006 0.016 0.008 0.008 0.016 0.016 0.010 0.013 0.009 0.003 0.010 0.008 0.011 0.011 0.006 0.006 0.003 0.006 0.074 0.110 0.101 0.077 0.153 0.024 0.048 0.052 0.053 0.086 0.116 0.299 0.194 0.173 0.195 0.195 0.113 0.165 0.150 0.122 6 5 6 4 4 4 6 4 4 4 10 10 10 10 3 3 7 6 4 6 5.00 5.36 8.63 12.20 3.30 3.30 8.84 1.14 0.99 2.46 APPENDIX B Average Daily Flow at Sampling Sites Figure B–1 Average daily flow at site BO040 on the North Bosque River below Stephenville. BO040 10.00 1.00 0.10 Jan-00 Sep-99 May-99 Jan-99 Sep-98 May-98 Jan-98 Sep-97 May-97 Jan-97 Sep-96 May-96 Jan-96 Sep-95 May-95 0.01 Jan-95 Average Daily Flow (m3 /s) 100.00 69 Strategies for Monitoring Figure B–2 Average daily flow at site BO070 on the North Bosque River at Hico. Average Daily Flow (m3 /s) 1,000.00 BO070 100.00 10.00 1.00 0.10 Figure B–3 Jan-00 Sep-99 May-99 Jan-99 Sep-98 May-98 Jan-98 Sep-97 May-97 Jan-97 Sep-96 May-96 Jan-96 Sep-95 May-95 Jan-95 0.01 Average daily flow at site BO090 on the North Bosque River near Clifton. Average Daily Flow (m3 /s) 10,000.00 BO090 1,000.00 100.00 10.00 1.00 0.10 70 Jan-00 Sep-99 May-99 Jan-99 Sep-98 May-98 Jan-98 Sep-97 May-97 Jan-97 Sep-96 May-96 Jan-96 Sep-95 May-95 Jan-95 0.01 Appendix B Figure B–4 Average Daily Flow at Sampling Sites Average daily flow at site GC100 on Green Creek. Average Daily Flow (m3 /s) 100.00 GC100 10.00 1.00 0.10 May-98 Sep-98 Jan-99 May-99 Sep-99 Jan-00 Sep-98 Jan-99 May-99 Sep-99 Jan-00 Jan-98 Sep-97 May-97 Jan-97 Sep-96 May-96 Jan-96 May-98 Figure B–5 Sep-95 Jan-95 May-95 0.01 Average daily flow at site NC060 on Neils Creek. NC060 100.00 10.00 1.00 0.10 Start date: 05Oct95 Jan-98 Sep-97 May-97 Jan-97 Sep-96 May-96 Jan-96 Sep-95 May-95 0.01 Jan-95 Average Daily Flow (m3 /s) 1,000.00 71 Strategies for Monitoring Figure B–6 Average daily flow at site NF020 on the North Fork of the North Bosque River. Average Daily Flow (m3 /s) 10.00 NF020 1.00 0.10 Figure B–7 Jan-00 Sep-99 May-99 Jan-99 Sep-98 May-98 Jan-98 Sep-97 May-97 Jan-97 Sep-96 May-96 Jan-96 Sep-95 May-95 Jan-95 0.01 Average daily flow at site SF020 on the South Fork of the North Bosque River. Average Daily Flow (m3 /s) 10.00 SF020 1.00 0.10 72 Jan-00 Sep-99 May-99 Jan-99 Sep-98 May-98 Jan-98 Sep-97 May-97 Jan-97 Sep-96 May-96 Jan-96 Sep-95 May-95 Jan-95 0.01 APPENDIX C Impact of Sampling Frequency on Annual Average Concentration Table C–1 Impact of sampling interval on average annual concentration at site BO040. B represents biweekly sampling, M1 and M2 represent monthly sampling schemes, and Q1 through Q6 represent quarterly sampling schemes. Year B M1 PO4-P (mg/L) 1995 1.247 1.380 1996 1.703 1.626 1997 0.889 1.013 1998 1.362 1.505 1999 1.861 1.879 TP (mg/L) 1995 1.50 1.54 1996 1.87 1.82 1997 0.98 1.15 1998 2.16 1.76 1999 2.02 2.04 TSS (mg/L) 1995 35 9 1996 15 15 1997 20 23 1998 11 12 1999 11 10 Number of Observations 1995 29 12 1996 26 12 1997 28 12 1998 25 12 1999 26 12 M2 Q1 Q2 Q3 Q4 Q5 Q6 1.201 1.649 0.772 1.270 1.794 0.963 1.483 0.940 1.678 2.005 1.913 1.830 0.383 1.443 1.773 1.045 1.553 0.438 0.988 1.903 1.280 2.003 1.740 1.380 1.908 1.070 1.195 1.215 1.190 1.340 1.108 2.220 0.625 1.475 2.068 1.53 1.77 0.85 2.70 1.93 1.05 1.68 1.20 1.91 2.12 2.12 2.05 0.50 1.81 2.00 1.61 1.79 0.62 1.17 2.06 1.69 2.09 1.75 1.54 2.16 1.30 1.32 1.20 4.90 1.54 1.24 2.35 0.72 1.71 2.13 61 16 22 9 12 13 21 19 13 14 5 5 8 11 9 157 20 43 6 12 10 12 40 12 7 40 27 12 9 11 15 7 11 11 8 12 12 12 12 12 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 73 Strategies for Monitoring Table C–2 Impact of sampling interval on average annual concentration at site BO070. B represents biweekly sampling, M1 and M2 represent monthly sampling schemes, and Q1 through Q6 represent quarterly sampling schemes. Year B M1 PO4-P (mg/L) 1995 0.22 0.21 1996 0.31 0.32 1997 0.17 0.19 1998 0.14 0.14 1999 0.30 0.28 TP (mg/L) 1995 0.37 0.29 1996 0.40 0.41 1997 0.30 0.30 1998 0.30 0.31 1999 0.41 0.33 TSS (mg/L) 1995 71 14 1996 19 13 1997 16 23 1998 10 13 1999 12 6 Number of Observations 1995 27 12 1996 26 12 1997 27 12 1998 25 12 1999 17 8 M2 Q1 Q2 Q3 Q4 Q5 Q6 0.26 0.33 0.15 0.11 0.33 0.30 0.29 0.18 0.16 0.29 0.16 0.22 0.15 0.08 0.38 0.26 0.40 0.10 0.06 0.41 0.12 0.30 0.24 0.17 0.39 0.35 0.36 0.22 0.18 0.24 0.19 0.26 0.14 0.14 0.36 0.44 0.43 0.29 0.30 0.49 0.41 0.35 0.28 0.35 0.34 0.21 0.31 0.29 0.18 0.53 0.65 0.48 0.24 0.26 0.63 0.30 0.36 0.50 0.41 0.57 0.51 0.52 0.33 0.26 0.36 0.25 0.36 0.19 0.30 0.53 131 27 11 9 19 23 21 13 9 4 7 5 11 4 13 371 11 8 8 9 18 23 25 25 10 50 45 10 5 3 10 10 15 12 10 12 12 12 12 8 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 Table C–3 Impact of sampling interval on average annual concentration at site BO090. B represents biweekly sampling, M1 and M2 represent monthly sampling schemes, and Q1 through Q6 represent quarterly sampling schemes. Year B M1 PO4-P (mg/L) 1995 0.008 0.011 1996 0.033 0.038 1997 0.040 0.050 1998 0.034 0.043 1999 0.005 0.004 TP (mg/L) 1995 0.06 0.06 1996 0.08 0.07 1997 0.13 0.12 1998 0.13 0.17 1999 0.11 0.11 TSS (mg/L) 1995 16 19 1996 14 16 1997 40 40 1998 20 23 1999 9 9 Number of Observations 1995 8 4 1996 26 12 1997 27 12 1998 26 12 1999 27 12 74 M2 Q1 Q2 Q3 Q4 Q5 Q6 0.006 0.032 0.029 0.025 0.006 0.010 0.071 0.061 0.040 0.005 0.008 0.007 0.028 0.013 0.003 0.008 0.032 0.021 0.012 0.004 0.008 0.013 0.037 0.078 0.004 0.008 0.039 0.029 0.016 0.004 0.008 0.032 0.044 0.048 0.010 0.06 0.09 0.14 0.10 0.12 0.04 0.05 0.12 0.20 0.13 0.04 0.05 0.09 0.13 0.19 0.04 0.08 0.16 0.04 0.06 0.09 0.08 0.12 0.19 0.06 0.04 0.13 0.13 0.07 0.08 0.04 0.07 0.14 0.19 0.12 21 12 49 18 9 28 27 12 26 10 23 15 8 14 9 23 8 12 15 6 25 12 22 29 9 26 11 100 16 8 31 13 31 21 11 4 12 12 12 12 2 4 4 4 4 2 4 4 4 4 2 4 4 4 4 2 4 4 4 4 2 4 4 4 4 2 4 4 4 4 Appendix C Impact of Sampling Frequency on Annual Average Concentration Table C–4 Impact of sampling interval on average annual concentration at site GC100. B represents biweekly sampling, M1 and M2 represent monthly sampling schemes, and Q1 through Q6 represent quarterly sampling scheme. Year B M1 PO4-P (mg/L) 1995 0.073 0.051 1996 0.038 0.024 1997 0.079 0.084 1998 0.120 0.126 1999 0.015 0.017 TP (mg/L) 1995 0.13 0.10 1996 0.09 0.10 1997 0.20 0.24 1998 0.28 0.30 1999 0.14 0.17 TSS (mg/L) 1995 21 17 1996 9 9 1997 17 12 1998 15 16 1999 11 11 Number of Observations 1995 26 12 1996 23 12 1997 27 12 1998 14 7 1999 9 5 M2 Q1 Q2 Q3 Q4 Q5 Q6 0.089 0.051 0.085 0.141 0.013 0.057 0.029 0.090 0.130 0.030 0.060 0.016 0.028 0.181 0.009 0.085 0.033 0.081 0.076 0.006 0.068 0.027 0.140 0.123 0.014 0.126 0.054 0.078 0.142 0.011 0.055 0.065 0.091 0.223 0.015 0.14 0.10 0.21 0.30 0.15 0.12 0.07 0.35 0.36 0.14 0.10 0.10 0.25 0.25 0.11 0.20 0.12 0.22 0.18 0.12 0.20 0.09 0.23 0.33 0.28 0.10 0.15 0.13 0.28 0.12 0.12 0.08 0.14 0.42 0.09 28 12 24 18 12 19 7 8 23 14 25 9 10 18 10 54 12 47 20 9 6 7 20 25 11 15 13 15 24 10 19 10 8 27 14 11 12 12 7 5 4 4 4 3 2 4 4 4 3 2 4 4 4 3 2 4 4 4 3 2 3 4 4 3 2 4 4 4 3 2 Table C–5 Impact of sampling interval on average annual concentration at site NC060. B represents biweekly sampling, M1 and M2 represent monthly sampling schemes, and Q1 through Q6 represent quarterly sampling scheme. Year B M1 PO4-P (mg/L) 1995 0.005 0.005 1996 0.010 0.008 1997 0.024 0.031 1998 0.012 0.013 1999 0.005 0.003 TP (mg/L) 1995 0.05 0.04 1996 0.07 0.09 1997 0.07 0.09 1998 0.08 0.10 1999 0.07 0.06 TSS (mg/L) 1995 5 5 1996 5 6 1997 12 9 1998 3 3 1999 3 3 Number of Observations 1995 8 4 1996 21 11 1997 27 12 1998 21 10 1999 16 8 M2 Q1 Q2 Q3 Q4 Q5 Q6 0.005 0.012 0.018 0.013 0.007 0.005 0.011 0.035 0.014 0.003 0.005 0.008 0.022 0.010 0.004 0.005 0.004 0.023 0.010 0.005 0.005 0.004 0.025 0.016 0.003 0.005 0.007 0.017 0.014 0.004 0.005 0.023 0.021 0.017 0.008 0.05 0.05 0.06 0.08 0.09 0.04 0.15 0.10 0.10 0.08 0.06 0.05 0.07 0.14 0.06 0.04 0.05 0.04 0.06 0.08 0.06 0.05 0.13 0.10 0.06 0.04 0.05 0.08 0.05 0.10 0.04 0.05 0.05 0.13 0.12 5 5 17 3 3 5 5 4 3 4 5 5 5 3 3 5 5 5 3 3 5 5 5 3 3 5 5 40 3 3 5 5 5 3 3 4 11 12 11 8 2 4 4 3 4 2 4 4 4 4 2 4 4 4 4 2 4 4 4 4 2 4 4 4 4 2 4 4 4 4 75 Strategies for Monitoring Table C–6 Impact of sampling interval on average annual concentration at site NF020. B represents biweekly sampling, M1 and M2 represent monthly sampling schemes, and Q1 through Q6 represent quarterly sampling scheme. Year B M1 PO4-P (mg/L) 1995 0.819 0.686 1996 1.848 2.072 1997 1.002 1.151 1998 0.861 0.712 1999 na na TP (mg/L) 1995 1.39 1.03 1996 2.49 2.95 1997 1.40 1.41 1998 1.14 1.03 1999 na na TSS (mg/L) 1995 85 49 1996 35 31 1997 24 11 1998 17 12 1999 na na Number of Observations 1995 16 8 1996 10 6 1997 17 8 1998 9 5 1999 0 0 M2 Q1 Q2 Q3 Q4 Q5 Q6 0.789 1.683 1.090 0.906 na 1.290 1.890 1.555 0.285 na 0.300 2.203 0.995 0.495 na 0.457 2.033 1.575 0.540 na 1.043 1.493 1.338 1.015 na 1.670 1.603 1.113 0.455 na 0.650 1.397 1.073 1.200 na 1.70 1.77 1.48 1.14 na 1.62 2.29 1.65 0.57 na 0.76 3.59 1.52 0.69 na 1.26 1.92 2.27 0.65 na 1.47 1.81 2.12 1.42 na 2.53 1.94 1.02 0.62 na 1.26 1.20 1.52 1.59 na 146 31 39 20 na 23 35 8 9 na 89 31 7 40 na 310 12 64 37 na 38 16 10 16 na 37 25 8 10 na 139 37 9 38 na 8 6 8 5 0 3 3 4 2 0 3 3 4 2 0 3 3 4 2 0 3 3 4 2 0 3 3 4 2 0 3 3 4 2 0 Table C–7 Impact of sampling interval on average annual concentration at site SF020. B represents biweekly sampling, M1 and M2 represent monthly sampling schemes, and Q1 through Q6 represent quarterly sampling scheme. Year B M1 PO4-P (mg/L) 1995 0.041 0.031 1996 0.022 0.014 1997 0.044 0.046 1998 0.029 0.033 1999 na na TP (mg/L) 1995 0.14 0.11 1996 0.07 0.06 1997 0.13 0.15 1998 0.12 0.16 1999 na na TSS (mg/L) 1995 75 56 1996 9 5 1997 35 9 1998 9 6 1999 na na Number of Observations 1995 17 9 1996 10 6 1997 16 7 1998 6 4 1999 0 0 76 M2 Q1 Q2 Q3 Q4 Q5 Q6 0.050 0.028 0.049 0.024 na 0.033 0.019 0.057 0.030 na 0.012 0.008 0.020 0.030 na 0.073 0.014 0.019 0.030 na 0.021 0.009 0.057 0.028 na 0.057 0.013 0.040 0.028 na 0.011 0.041 0.035 0.030 na 0.10 0.08 0.13 0.09 na 0.11 0.07 0.16 0.17 na 0.06 0.06 0.09 0.08 na 0.13 0.05 0.14 0.08 na 0.15 0.05 0.20 0.11 na 0.27 0.07 0.13 0.11 na 0.02 0.08 0.09 0.08 na 82 11 70 11 na 60 5 11 6 na 22 17 5 16 na 160 17 5 16 na 52 17 139 7 na 59 5 23 7 na 24 17 5 16 na 9 6 7 4 0 3 3 3 2 0 3 3 3 2 0 3 3 3 2 0 3 3 3 2 0 3 3 3 2 0 3 3 3 2 0 APPENDIX D Regression Estimation of Concentrations and Loadings Table D–1 Regression method results for site BO040. Results based on data collected between January 1, 1995 to December 31, 1999. Concentration Regression Sampling Scheme Monthly Loadings - MVUE Residuals Normala R2 n R2 1055 0.61 **b no 0.93 168 0.58 ** yes 155 0.03 ns 158 135 0.69 0.50 995 Monthly Loadings - Smearing Slope Std Error R2 ** 1.02 0.04 0.93 0.91 ** 0.81 0.03 yes 0.86 ** 0.04 ** ** yes yes 0.88 0.91 ** ** 0.18 ** no 0.93 167 0.21 ** yes 156 0.29 ** 155 134 0.24 0.25 1071 Slope Std Error ** 0.41 0.01 0.91 ** 0.87 0.04 0.00 0.87 ** 0.04 0.00 4.21 1.11 0.20 0.05 0.88 0.90 ** ** 4.44 1.27 0.22 0.05 ** 1.05 0.04 0.93 ** 0.46 0.02 0.88 ** 1.01 0.05 0.88 ** 1.02 0.05 yes 0.89 ** 0.71 0.03 0.89 ** 0.71 0.03 ** ** yes yes 0.89 0.83 ** ** 0.96 0.77 0.04 0.04 0.89 0.84 ** ** 0.97 0.79 0.04 0.05 0.23 ** no 0.91 ** 1.06 0.04 0.91 ** 0.42 0.02 167 0.29 ** yes 0.89 ** 1.17 0.05 0.89 ** 1.18 0.06 156 0.22 ** yes 0.85 ** 1.55 0.09 0.85 ** 1.49 0.08 157 133 0.32 0.28 ** ** yes yes 0.89 0.85 ** ** 1.04 0.98 0.05 0.05 0.89 0.86 ** ** 1.03 0.99 0.05 0.05 TSS Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs TP Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs PO4-P Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs a. Normality of residuals based on a Shapiro-Wilks test with an a = 0.05. b. ** indicates significance at a = 0.01. 77 Strategies for Monitoring Table D–2 Regression method results for site BO070. Results based on data collected between January 1, 1995 to December 31, 1999. Concentration Regression Sampling Scheme Monthly Loadings - MVUE Residuals Normala R2 n R2 1425 0.55 **b no 0.92 156 0.57 ** no 151 0.09 *c 151 123 0.58 0.50 1378 Monthly Loadings - Smearing Slope Std Error R2 ** 1.80 0.07 0.92 0.88 ** 1.71 0.08 no 0.71 ** 0.03 ** ** no no 0.88 0.92 ** ** 0.10 ** no 0.98 154 0.08 ns no 151 0.03 ns 149 122 0.16 0.05 1434 Slope Std Error ** 0.62 0.02 0.88 ** 2.20 0.11 0.00 0.73 ** 0.04 0.00 4.55 1.07 0.22 0.04 0.87 0.91 ** ** 6.32 1.44 0.32 0.06 ** 1.13 0.02 0.98 ** 0.41 0.01 0.98 ** 0.96 0.02 0.98 ** 0.94 0.02 no 0.95 ** 0.54 0.02 0.95 ** 0.51 0.02 ** ns yes no 0.97 0.98 ** ** 1.43 0.79 0.03 0.02 0.97 0.98 ** ** 1.40 0.80 0.03 0.02 0.03 ** no 0.96 ** 1.18 0.03 0.96 ** 0.41 0.01 155 0.08 ns no 0.96 ** 0.92 0.03 0.96 ** 0.87 0.02 151 0.12 ** no 0.86 ** 2.01 0.10 0.86 ** 1.77 0.09 149 122 0.08 0.09 * ns no no 0.94 0.93 ** ** 1.05 0.87 0.04 0.03 0.94 0.94 ** ** 0.98 0.88 0.03 0.03 TSS Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs TP Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs PO4-P Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs a. Normality of residuals based on a Shapiro-Wilks test with an a = 0.05. b. ** indicates significance at a = 0.01. c. * indicates significance at a = 0.05. 78 APPENDIX D Regression Estimation of Concentrations and Loadings Table D–3 Regression method results for site BO090. Results based on data collected between October 10, 1995 to December 31, 1999. Concentration Regression Sampling Scheme Monthly Loadings - MVUE Residuals Normala R2 n R2 998 0.56 **b no 0.57 136 0.63 ** no 134 0.05 ns 130 113 0.49 0.44 958 Monthly Loadings - Smearing Slope Std Error R2 ** 1.51 0.19 0.57 0.54 ** 2.15 0.28 no 0.66 ** 0.02 ** ** yes no 0.56 0.65 ** ** 0.31 ** no 0.91 137 0.27 ** yes 134 0.10 *c 130 113 0.19 0.13 1011 Slope Std Error ** 0.57 0.07 0.53 ** 2.64 0.35 0.00 0.78 ** 0.01 0.00 1.28 0.43 0.16 0.04 0.53 0.61 ** ** 1.59 0.52 0.21 0.06 ** 1.06 0.05 0.91 ** 0.45 0.02 0.90 ** 1.03 0.05 0.89 ** 1.10 0.06 no 0.94 ** 0.31 0.01 0.94 ** 0.32 0.01 ** * yes yes 0.90 0.94 ** ** 1.17 0.62 0.06 0.02 0.86 0.91 ** ** 1.39 0.76 0.08 0.03 0.52 ** no 0.71 ** 1.97 0.18 0.71 ** 0.80 0.07 137 0.60 ** yes 0.72 ** 1.20 0.11 0.69 ** 1.26 0.12 134 0.41 ** no 0.69 ** 5.46 0.52 0.57 ** 9.04 1.10 130 113 0.53 0.58 ** ** yes yes 0.65 0.61 ** ** 1.96 2.30 0.21 0.26 0.57 0.53 ** ** 2.45 2.90 0.30 0.39 TSS Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs TP Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs PO4-P Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs a. Normality of residuals based on a Shapiro-Wilks test with an a = 0.05. b. ** indicates significance at a = 0.01. c. * indicates significance at a = 0.05. 79 Strategies for Monitoring Table D–4 Regression method results for site GC100. Results based on data collected from January 1, 1995 to December 31, 1999. Concentration Regression Sampling Scheme Monthly Loadings - MVUE Residuals Normala R2 n R2 734 0.38 **b no 0.81 121 0.55 ** yes 114 0.08 ns 114 99 0.41 0.35 677 Monthly Loadings - Smearing Slope Std Error R2 ** 0.45 0.03 0.82 0.81 ** 0.43 0.03 no 0.65 ** 0.03 ** ** yes yes 0.89 0.85 ** ** 0.28 ** yes 0.93 120 0.34 ** yes 114 0.15 ** 114 100 0.29 0.24 741 Slope Std Error ** 0.15 0.01 0.81 ** 0.45 0.03 0.00 0.66 ** 0.02 0.00 0.21 0.11 0.01 0.01 0.89 0.85 ** ** 0.24 0.12 0.01 0.01 ** 0.92 0.03 0.93 ** 0.42 0.02 0.94 ** 1.01 0.03 0.94 ** 0.99 0.03 yes 0.90 ** 0.56 0.02 0.90 ** 0.53 0.02 ** ** yes yes 0.93 0.91 ** ** 1.13 1.04 0.04 0.04 0.93 0.91 ** ** 1.13 1.06 0.04 0.04 0.41 ** no 0.90 ** 1.16 0.05 0.90 ** 0.53 0.02 121 0.26 ** yes 0.87 ** 0.87 0.04 0.87 ** 0.84 0.04 115 0.45 ** yes 0.80 ** 4.67 0.30 0.80 ** 4.76 0.31 115 100 0.39 0.39 ** ** no yes 0.85 0.83 ** ** 1.64 2.65 0.09 0.16 0.85 0.82 ** ** 1.64 2.79 0.09 0.17 TSS Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs TP Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs PO4-P Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs a. Normality of residuals based on a Shapiro-Wilks test with an a = 0.05. b. ** indicates significance at a = 0.01. 80 APPENDIX D Regression Estimation of Concentrations and Loadings Table D–5 Regression method results for site NC060. Results based on data collected from October 5, 1995 to December 31, 1999. Concentration Regression Sampling Scheme Monthly Loadings - MVUE Residuals Normala R2 n R2 847 0.47 **b no 0.73 119 0.65 ** no 114 0.17 ** 111 92 0.67 0.41 809 Monthly Loadings - Smearing Slope Std Error R2 ** 1.53 0.13 0.72 0.65 ** 4.51 0.46 no 0.47 ** 0.00 ** ** no no 0.70 0.78 ** ** 0.14 ** no 0.97 121 0.30 ** no 115 0.13 * 113 93 0.30 0.06 862 Slope Std Error ** 0.52 0.04 0.65 ** 6.20 0.64 0.00 0.49 ** 0.00 0.00 3.23 0.05 0.30 0.00 0.69 0.78 ** ** 3.98 0.05 0.37 0.00 ** 0.93 0.02 0.97 ** 0.38 0.01 0.98 ** 1.29 0.02 0.98 ** 1.34 0.03 no 0.77 ** 0.12 0.01 0.78 ** 0.12 0.01 ** ns no no 0.98 0.89 ** ** 1.56 0.21 0.03 0.01 0.98 0.91 ** ** 1.62 0.24 0.03 0.01 0.35 ** no 0.77 ** 1.01 0.08 0.77 ** 0.36 0.03 121 0.46 ** yes 0.80 ** 0.66 0.05 0.79 ** 0.66 0.05 115 0.49 ** no 0.42 ** 17.69 2.91 0.41 ** 20.22 3.45 113 93 0.48 0.46 ** ** yes yes 0.78 0.79 ** ** 0.69 0.57 0.05 0.04 0.78 0.77 ** ** 0.69 0.61 0.05 0.05 TSS Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs TP Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs PO4-P Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs a. Normality of residuals based on a Shapiro-Wilks test with an a = 0.05. b. ** indicates significance at a = 0.01. 81 Strategies for Monitoring Table D–6 Regression method results for site NF020. Results based on data collected from January 1, 1995 to December 31, 1999. Concentration Regression Sampling Scheme Monthly Loadings - MVUE Residuals Normala R2 n R2 519 0.36 **b no 0.69 85 0.56 ** yes 63 0.21 * 68 53 0.49 0.22 440 Monthly Loadings - Smearing Slope Std Error R2 ** 1.92 0.17 0.69 0.36 ** 9.30 1.61 yes 0.63 ** 0.01 ** ns yes yes 0.49 0.68 ** ** 0.14 ** no 0.95 84 0.27 ** yes 61 0.14 ns 66 51 0.33 0.24 520 Slope Std Error ** 1.43 0.13 0.37 ** 9.52 1.63 0.00 0.64 ** 0.01 0.00 1.33 0.00 0.18 0.00 0.46 0.65 ** ** 1.52 0.08 0.21 0.01 ** 0.85 0.02 0.95 ** 0.32 0.01 0.93 ** 0.90 0.03 0.93 ** 0.89 0.03 yes 0.89 ** 0.59 0.03 0.89 ** 0.59 0.03 ** *c yes yes 0.89 0.82 ** ** 1.10 1.01 0.05 0.06 0.88 0.81 ** ** 1.12 1.16 0.05 0.07 0.14 ** no 0.96 ** 1.09 0.03 0.96 ** 0.33 0.01 85 0.10 ns no 0.97 ** 0.66 0.02 0.97 ** 0.57 0.01 62 0.14 ns no 0.88 ** 0.81 0.04 0.88 ** 0.74 0.03 67 52 0.25 0.20 ** ns no no 0.88 0.82 ** ** 1.31 1.61 0.06 0.10 0.88 0.81 ** ** 1.20 1.85 0.06 0.12 TSS Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs TP Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs PO4-P Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs a. Normality of residuals based on a Shapiro-Wilks test with an a = 0.05. b. ** indicates significance at a = 0.01. c. * indicates significance at a = 0.05. 82 APPENDIX D Regression Estimation of Concentrations and Loadings Table D–7 Regression method results for site SF020. Results based on data collected from January 1, 1995 to December 31, 1999. Concentration Regression Sampling Scheme Monthly Loadings - MVUE Residuals Normala R2 n R2 868 0.22 **b no 0.65 89 0.45 ** yes 74 0.13 ns 71 50 0.46 0.39 775 Monthly Loadings - Smearing Slope Std Error R2 ** 1.37 0.13 0.64 0.37 ** 11.57 1.98 yes 0.41 ** 0.02 ** ** no no 0.34 0.28 ** ** 0.08 ** no 0.93 88 0.40 ** yes 73 0.04 ns 70 49 0.16 0.30 876 Slope Std Error ** 0.32 0.03 0.34 ** 11.60 2.11 0.00 0.44 ** 0.02 0.00 1.52 0.70 0.27 0.14 0.31 0.24 ** ** 1.84 1.12 0.36 0.26 ** 0.89 0.03 0.93 ** 0.25 0.01 0.79 ** 1.33 0.09 0.79 ** 1.35 0.09 yes 0.87 ** 0.35 0.02 0.90 ** 0.34 0.01 ns *c no yes 0.93 0.86 ** ** 1.10 1.26 0.04 0.07 0.92 0.86 ** ** 1.53 1.72 0.06 0.09 0.12 ** no 0.68 ** 1.31 0.12 0.68 ** 0.31 0.03 88 0.13 ns yes 0.74 ** 1.74 0.13 0.72 ** 1.76 0.14 73 0.25 ** yes 0.72 ** 5.66 0.46 0.71 ** 6.84 0.57 70 49 0.14 0.27 ns * yes yes 0.78 0.71 ** ** 1.22 3.02 0.08 0.25 0.78 0.67 ** ** 1.24 4.64 0.09 0.43 TSS Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs TP Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs PO4-P Full data set Set stage + biweekly grabs Peak flow + biweekly grabs Storm chasing + biweekly grabs Biweekly grabs a. Normality of residuals based on a Shapiro-Wilks test with an a = 0.05. b. ** indicates significance at a = 0.01. c. * indicates significance at a = 0.05. 83 Strategies for Monitoring 84 APPENDIX E Comparison of Annual Loading Estimation Methods 85 Strategies for Monitoring Table E–1 Annual percent error in PO4-P loading estimates compared to “true” loadings using regression and integration approaches. Integration Method 86 Regression Method Biweekly Grabs First Four Storms + Biweekly Grabs Quarterly Storms + Biweekly Grabs Biweekly Grabs Storm Set Stage + Chasing + Biweekly Biweekly Grabs Grabs Measured Annual Discharge (m3) "True" Annual PO4-P Loading (kg) 1995 1996 1997 1998 1999 28,088,179 32,234,746 39,158,673 25,072,194 7,031,493 Average Std 17,231 24,899 21,292 17,475 10,250 44% 64% -4% 22% 5% 26% 28% 16% -48% -2% 9% -11% -7% 25% 160% 47% 2% 50% -5% 51% 66% 43% -12% -26% -10% 2% -1% 26% 27% -5% -10% 16% -13% 3% 18% 65% -3% -2% 25% -1% 17% 29% BO070 1995 1996 1997 1998 1999 142,731,373 77,026,469 235,271,714 128,608,684 28,700,086 Average Std 27,031 18,900 59,019 26,930 8,447 8% 26% -7% -22% -3% 0% 18% -32% 28% 13% -12% -14% -3% 24% -31% 26% -31% 7% -10% -8% 25% 38% -13% -15% 12% -5% 3% 22% 66% 14% 6% 19% 20% 25% 24% 42% 1% -4% 11% 18% 14% 18% BO090 1996 1997 1998 1999 133,342,938 629,252,789 282,320,839 45,781,829 Average Std 13,919 87,735 22,049 225 -35% -44% 3% -21% -24% 20% -83% 21% 3% -5% -16% 46% -70% -61% 9% 12% -28% 44% -1% 58% 543% 539% 285% 297% -22% -7% 192% 283% 111% 150% -14% 40% 422% 368% 204% 223% NC060 1996 1997 1998 1999 29,871,838 172,206,270 86,236,844 9,048,835 Average Std 697 9,644 1,533 77 -21% -65% -41% -2% -32% 27% -13% 0% 5% -5% -3% 8% -45% -53% 37% 25% -9% 47% -14% -43% 87% -12% 5% 57% -10% -35% 110% -8% 14% 65% -5% -33% 125% -14% 18% 72% GC100 1995 1996 1997 1998 1999 39,920,190 15,212,549 80,386,270 21,645,543 677,454 Average Std 4,795 1,507 13,059 6,328 28 27% -63% 7% -45% -18% -18% 37% -55% 169% -12% -39% -25% 8% 92% -37% 21% -4% -27% -75% -24% 36% 159% 259% 213% 38% 398% 213% 132% 1% 38% 9% -50% 118% 23% 62% 104% 151% 94% -16% 309% 128% 118% NF020 1995 1996 1997 1998 1999 809,418 852,261 2,095,411 1,098,122 14,849 Average Std 987 1,657 3,908 1,646 6 -37% 29% -38% -65% -100% -42% 47% -3% -29% 1% -6% -17% -11% 12% -13% -46% -19% 8% -17% -17% 19% 191% 175% 45% -2% -64% 69% 111% -40% -10% -30% -31% -25% -27% 11% 104% 93% 21% -18% -60% 28% 71% SF020 1995 1996 1997 1998 1999 1,081,900 1,246,557 1,605,089 551,534 14,046 Average Std 35 51 65 22 1 20% -76% 57% -18% -100% -24% 65% -63% 53% -35% -27% 0% -15% 44% 0% -10% -40% 59% 100% 22% 57% 412% 166% 136% 151% -43% 164% 162% 121% 53% 70% 90% -7% 65% 48% 50% 29% 38% 22% -48% 18% 38% Site Year BO040 APPENDIX E Comparison of Annual Loading Estimation Methods Table E–2 Annual percent error in TP loading estimates compared to “true” loadings using regression and integration approaches. Integration Method Regression Method Biweekly Grabs First Four Storms + Biweekly Grabs Quarterly Storms + Biweekly Grabs Biweekly Grabs Storm Set Stage + Chasing + Biweekly Biweekly Grabs Grabs Measured Annual Discharge (m3) "True" Annual TP Loading (kg) 1995 1996 1997 1998 1999 28,088,179 32,234,746 39,158,673 25,072,194 7,031,493 Average Std 31,012 36,907 36,446 30,867 11,519 15% 25% -31% -4% 3% 2% 21% 10% -57% 1% 1% -11% -11% 27% 60% 14% 0% 16% -4% 17% 25% 10% -1% -26% -18% 29% -1% 22% 14% -20% 5% -11% -1% -3% 13% 29% -11% 45% 22% 4% 18% 22% BO070 1995 1996 1997 1998 1999 142,731,373 77,026,469 235,271,714 128,608,684 28,700,086 Average Std 74,423 43,476 124,058 74,397 12,293 6% -29% -34% -41% -12% -22% 19% -59% 22% 13% -10% -14% -10% 32% -41% -21% -34% -13% -11% -24% 13% -13% -30% -12% -21% -6% -17% 9% 40% 13% -5% -4% 29% 15% 20% 29% 4% -8% -6% 26% 9% 18% BO090 1996 1997 1998 1999 133,342,938 629,252,789 282,320,839 45,781,829 Average Std 56,128 295,036 197,197 2,878 -71% -53% -72% -2% -49% 33% -10% 28% -32% 21% 2% 28% -73% -54% -41% 26% -36% 43% -53% -35% -27% 261% 36% 150% -33% -8% 30% 503% 123% 255% -35% 2% 48% 456% 118% 228% NC060 1996 1997 1998 1999 29,871,838 172,206,270 86,236,844 9,048,835 Average Std 9,904 38,937 30,707 1,315 -76% -66% -85% -55% -71% 13% -27% 4% -50% 16% -14% 30% -76% -47% -62% 24% -40% 44% -78% -67% -80% -61% -72% 9% -21% 25% 18% -43% -5% 32% -22% 47% 43% -43% 6% 46% GC100 1995 1996 1997 1998 1999 39,920,190 15,212,549 80,386,270 21,645,543 677,454 Average Std 17,065 6,085 29,869 12,533 382 -39% -79% -33% -41% -68% -52% 20% -68% 58% -11% -5% -65% -18% 52% -47% 30% -6% -12% -70% -21% 39% -36% -19% 26% -10% 5% -7% 24% -24% -18% 20% -18% 22% -4% 23% -21% -12% 33% -8% 23% 3% 24% NF020 1995 1996 1997 1998 1999 809,418 852,261 2,095,411 1,098,122 14,849 Average Std 1,623 2,535 7,402 2,660 16 -22% 1% -54% -66% -100% -48% 39% 10% -3% 0% -8% 25% 5% 13% -26% -4% -36% 18% 0% -10% 22% 156% 71% -9% -16% -65% 27% 87% 61% 29% -14% -20% -65% -2% 48% 130% 59% -1% -9% -65% 23% 74% SF020 1995 1996 1997 1998 1999 1,081,900 1,246,557 1,605,089 551,534 14,046 Average Std 199 308 298 93 6 -52% -83% -30% -22% -100% -57% 34% -60% 0% 15% -15% -50% -22% 32% -44% 7% -27% -23% -17% -20% 18% 123% 34% 21% 14% -82% 22% 73% 162% 68% 28% -6% -54% 39% 82% 30% -1% 38% 33% -85% 3% 52% Site Year BO040 87 Strategies for Monitoring Table E–3 Annual percent error in TSS loading estimates compared to “true” loadings using regression and integration approaches. Integration Method 88 Regression Method Biweekly Grabs First Four Storms + Biweekly Grabs Quarterly Storms + Biweekly Grabs Biweekly Grabs Storm Set Stage + Chasing + Biweekly Biweekly Grabs Grabs Measured Annual Discharge (m3) "True" Annual TSS Loading (kg) 1995 1996 1997 1998 1999 28,088,179 32,234,746 39,158,673 25,072,194 7,031,493 Average Std 10,179,417 6,305,332 12,675,136 7,133,848 181,164 -46% -86% -81% -93% -62% -74% 19% -74% -73% 16% -38% -42% -42% 37% -93% -85% 9% -47% -42% -52% 41% 5% -49% -5% 12% -45% -16% 29% 132% -31% 73% -17% -57% 20% 79% 205% -7% 370% 172% -55% 137% 172% BO070 1995 1996 1997 1998 1999 142,731,373 77,026,469 235,271,714 128,608,684 28,700,086 Average Std 53,364,308 19,072,767 81,007,695 50,255,136 783,833 -10% -89% -88% -95% -70% -70% 35% -95% 5% 28% -51% -48% -32% 49% -79% -90% -71% -66% -58% -73% 12% 4% -63% 13% -41% -73% -32% 39% -6% -29% -25% -20% -13% -19% 9% 224% 16% 247% 330% -32% 157% 157% BO090 1996 1997 1998 1999 133,342,938 629,252,789 282,320,839 45,781,829 Average Std 68,023,763 507,143,617 266,662,904 429,205 -95% -80% -96% -32% -76% 30% 29% 34% -53% 28% 10% 42% -94% -85% -80% 42% -54% 64% -70% -73% -15% 2166% 502% 1110% -1% 0% 376% 9142% 2379% 4512% -32% -38% 176% 6055% 1540% 3011% NC060 1996 1997 1998 1999 29,871,838 172,206,270 86,236,844 9,048,835 Average Std 38,062,104 60,771,744 38,026,491 692,694 -100% -87% -99% -96% -95% 6% -65% 6% -74% 50% -21% 59% -99% -88% -92% 67% -53% 80% -98% -94% -94% -95% -95% 2% 6% 225% 686% -73% 211% 341% 8% 126% 439% -81% 123% 227% GC100 1995 1996 1997 1998 1999 39,920,190 15,212,549 80,386,270 21,645,543 677,454 Average Std 17,065 6,085 29,869 6,443,191 317,707 -77% -95% -90% -95% -97% -91% 8% -92% 36% -6% -26% -95% -37% 57% -92% 28% -11% -41% -95% -42% 53% -84% -89% -86% -88% -87% -87% 2% -36% -60% -53% -65% -54% -54% 11% -74% -80% -75% -78% -80% -78% 3% NF020 1995 1996 1997 1998 1999 809,418 852,261 2,095,411 1,098,122 14,849 Average Std 688,232 1,072,651 1,590,056 511,007 10,667 -73% -98% -96% -97% -100% -93% 11% 41% 334% 1% -13% 51% 83% 143% -56% 556% -27% -63% -28% 76% 269% -85% -95% -94% -95% -99% -94% 5% 2772% 271% 649% 220% -100% 762% 1154% 234% -37% 8% 6% -99% 22% 126% SF020 1995 1996 1997 1998 1999 1,081,900 1,246,557 1,605,089 551,534 14,046 Average Std 185,252 656,897 259,883 102,074 3,263 -66% -98% -68% -98% -100% -86% 17% -92% -33% 32% -71% -39% -41% 47% -54% -43% -76% -98% -38% -61% 25% 420% -38% -29% -50% -98% 41% 214% 6306% 934% 975% 973% -29% 1832% 2538% 838% 29% 48% 45% -86% 175% 375% Site Year BO040
© Copyright 2026 Paperzz