Assessing Water Quality Management Plan Implementation in the Middle and South Bosque Rivers and Hog Creek Watersheds

Transcription

1 Assessing Water Quality Management Plan Implementation in the Middle and South Bosque Rivers and Hog Creek Watersheds Prepared for: Texas State Soil & Water Conservation Board Funded by US EPA Section 319(h) Nonpoint Source Program Project Prepared by: Anne McFarland Jimmy Millican Texas Institute for Applied Environmental Research Tarleton State University Stephenville, Texas PR1104 July 2012

2

3 Assessing Water Quality Management Plan Implementation in the Middle and South Bosque Rivers and Hog Creek Watersheds Prepared for: Texas State Soil & Water Conservation Board Funded by US EPA Section 319(h) Nonpoint Source Program Project Prepared by: Anne McFarland Jimmy Millican Texas Institute for Applied Environmental Research Tarleton State University Stephenville, Texas PR1104 July 2012

4 ACKNOWLEDGEMENTS Funding for this report was provided through the Clean Water Act Section 319(h) project, Assessing Water Quality Management Plan Implementation in the Middle and South Bosque River and Hog Creek Watersheds. This project was sponsored by the Texas State Soil and Water Conservation Board (TSSWCB) in cooperation with the United States Environmental Protection Agency, Region 6, as TSSWCB project Matching funds were provided by the State of Texas through the Texas Institute for Applied Environmental Research (TIAER) at Tarleton State University in Stephenville, Texas. The authors would like to recognize the dedication and long hours of hard work by field personnel and laboratory chemists who are on call seven days a week, since rainfall and stormwater runoff occurs throughout the week, not just Monday through Friday. 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 to info@tiaer.tarleton.edu. Authors Anne McFarland, research scientist, TIAER, mcfarla@tiaer.tarleton.edu Jimmy Millican, senior research associate, TIAER, jmillican@tiaer.tarleton.edu ii

5 ABSTRACT Within the South and Middle Bosque Rivers, concerns regarding elevated nitrite-nitrate nitrogen concentrations have lead to the development of Water Quality Management Plans (WQMPs) focused on practices to improve water quality. In 2006, the Texas State Soil and Water Conservation Board (TSSWCB) initiated a project to aid landowners in watersheds of the South and Middle Bosque Rivers and Hog Creek in the adoption of WQMPs. Through this project 25 WQMPs were implemented that impacted about 10 percent of the cropland and pasture in these largely rural watersheds. In 2008, a monitoring project was initiated to evaluate water quality improvements. Routine grab and stormwater runoff samples were collected from five sites; three representing mainstem sites on Hog Creek, the Middle Bosque River and the South Bosque River, and two smaller drainage areas (< 8,000 ac) or microwatersheds within the Middle Bosque watershed. Water quality data were analyzed for a suite of nutrient constituents focusing on nitrite-nitrogen+nitrate-nitrogen (NO 2 -N+NO 3 -N). Historical data from about 1996 through 2003 were compared to more recent data in a before/after analysis using analysis of covariance (ANCOVA) and nonparametric Wilcox rank sum (WRS) statistics. Routine grab and storm water runoff data were evaluated separately and flow-adjusted prior to analysis to account for differences in flow over time as an ancillary variable. Significant decreases in NO 2 -N+NO 3 -N concentrations were evaluated at all five stream sites with decreases ranging from 37 to 65 percent for routine grab samples and 23 to 71 percent for storm events. Decreases were also seen at most sites for ammonia and phosphorus constituents (orthophosphate and total phosphorus). While bacteria levels, monitored as E. coli, were another important water quality parameter, significant decreases in E. coli were monitored only at TC020 on Tonk Creek, one of the microwatershed sites. Improvements in water quality, particularly NO 2 -N+NO 3 -N, were discussed in relation to implementation practices associated with WQMPs as well as decreases in fertilizer sales and changes in land use, primarily from row crop to pasture or rangeland, throughout these watersheds. While significant decreases in NO 2 -N+NO 3 -N were indicated, some concerns may still exist in that concentrations at specific sites were often above Texas Commission on Environmental Quality (TCEQ) assessment screening levels. These findings of reduced NO 2 -N+NO 3 -N are encouraging and indicate that benefits are occurring from implemented best management practices and further improvements in water quality will likely occur if best management practices are continued and/or expanded. iii

6 iv

7 Table of Contents Introduction... 1 Watershed and Site Descriptions... 5 Methods... 7 Routine Grab Sampling... 7 Storm Sampling... 8 Laboratory Analyses... 9 Data Management and Statistical Analyses Censored Data Bacteria Data Definition of Before and After Periods Exploratory Data Analyses and Basic Statistics Calculation of Event Mean Concentrations (EMCs) Before and After Analysis Using ANCOVA Water Quality Before/After Results Routine Grab Samples Storm Events Discussion of Factors Influencing Water Quality Change Wastewater Treatment Facility Discharges Naval Weapons Industrial Reserve Plant - McGregor Implementation of WQMP Practices Fertilizer Sales Changes in Land Use Changes in Weather Conditions Impact of Lag Time on Current and Future Water Quality Conclusions References Appendix A Appendix B Appendix C Appendix D Appendix E v

8 List of Tables Table 1. Estimated land use and drainage area above sampling sites Table 2. Site locations and monitoring history of samples collected by TIAER through September Table 3. Abbreviations, units and analysis methods for water quality constituents Table 4. Constituent and maximum reporting limit used to left-censor data prior to data analysis Table 5. Summary statistics for routine grab samples for before and after monitoring periods Table 6. P-values from analysis of covariance (ANCOVA) and Wilcoxon rank sum (WRS) statistics comparing routine grab sample concentrations between before and after periods Table 7. Estimated change in flow-adjusted, grab concentrations before and after Table 8. Summary statistics for routine grab samples for before and after monitoring periods Table 9. P-values from analysis of covariance (ANCOVA) and Wilcoxon rank sum (WRS) statistics comparing EMCs between before and after periods Table 10. Estimated change in flow-adjusted, EMCs concentrations before and after Table 11. Management practices implemented with WQMPs within the South and Middle Bosque River and Hog Creek watersheds Table 12. Management practice and expected effects on surface water quality Table 13. Land use change between mid-1990s (old) and early 2000s (new) Table 14. Percent of routine grab samples exceeding nutrient screening levels or the single sample bacteria criterion for E. coli List of Figures Figure 1. Map showing Bosque River watershed with the North Bosque, Middle Bosque, South Bosque Rivers and Hog Creek watersheds delineated... 2 Figure 2. Location of project sampling sites Figure 3. Map showing the segments and subsegments within the Middle and South Bosque Rivers and Hog Creek watersheds... 4 Figure 4. Annual fertilizer sales for McLennan County for Figure 5. Annual precipitation as measured at TC020 and the National Weather Service station on the Waco Dam vi

9 Introduction The Middle and South Bosque Rivers and Hog Creek are located in central Texas. These three major streams flow into Lake Waco, an impoundment that also captures streamflow from the North Bosque River (Figure 1). The North Bosque River comprises the majority of the drainage area for Lake Waco (about 74 percent) with the Middle and South Bosque Rivers, Hog Creek and other minor tributaries representing the balance (McFarland and Hauck, 1999). Much of the land area within the drainage areas of the Middle and South Bosque Rivers and Hog Creek is used for row crop production of corn, grain sorghum, and forage sorghum and to a lesser degree crops such as soybeans and cotton. Improved pasture for hay or grazing is also common in conjunction with small-scale beef cattle operations. The City of Crawford (estimated population 808) resides within the Middle Bosque River watershed and the City of McGregor (estimated population 4,930) is located within the South Bosque River watershed (Figure 2; Texas State Data Center, 2010). Another prominent feature within the South Bosque River watershed is the former Naval Weapons Industrial Reserve Plant (NWIRP) near McGregor. The NWIRP was in operation during World War II and provided a site for the production and storage of explosive materials over a number of years (Easterling et al., 2001). The majority of land area associated with the NWIRP has now been transferred to the City of McGregor after extensive remediation and is now used primarily for light industry, pasture, and crop production (USEPA, 2009). Assessment monitoring starting in the spring of 2008 was conducted within the Middle and South Bosque Rivers and Hog Creek watersheds to evaluate implementation of water quality management plans (WQMPs) in controlling nonpoint source pollution (Figure 2). According to the 2004 Texas Water Quality Inventory, the Middle and South Bosque Rivers (Segment 1246) and Hog Creek (Segment 1225A) are fully supporting of their designated uses; however, there are nutrient enrichment concerns due to elevated nitrite-nitrate nitrogen concentrations in Segment 1246 (Figure 3). Subsegments 1246D (Tonk Creek) and 1246E (Wasp Creek), which flow into the Middle Bosque River, are also listed for concerns with regard to elevated nitritenitrate nitrogen. In addition, Wasp Creek (1246E) shows nonsupport for the use of contact recreation based on elevated bacteria concentrations, although the 2004 assessment indicates that further data are needed before Wasp Creek is considered for scheduling a total maximum daily load. Lake Waco, the receiving water body for the Middle and South Bosque Rivers and Hog Creek, also indicates concerns for elevated nitrite-nitrate nitrogen concentrations. Agriculture is considered one of the primary sources of nitrite-nitrate nitrogen within these watersheds. Within the Middle and South Bosque Rivers and Hog Creek watersheds, technical and financial assistance was provided to landowners to aid in the development of WQMPs through the McLennan County Soil and Water Conservation District (SWCD) starting in This technical assistance occurred under a 319(h) project (06-9), WQMP Implementation in the Middle and South Bosque River Watersheds. A WQMP is a site-specific plan, which includes appropriate land treatment practices, production practices, technologies and combinations thereof, and an implementation schedule. As the lead agency for abating agricultural nonpoint source pollution, the TSSWCB works closely with local SWCDs to reduce nonpoint source pollution from various agricultural activities. 1

10 Figure 1. Map showing Bosque River watershed with the North Bosque, Middle Bosque, South Bosque Rivers and Hog Creek watersheds delineated. 2

11 Figure 2. Location of project sampling sites. Sites HC060, MB063, SB065, TC020 and WC020 were used for direct and non-direct data collection, while sites SS001 and SB050 were used only for non-direct data collected during the before period. Note: Land area for NWIRP boundary was estimated based on changes in city boundary for McGregor, which acquired most of the NWIRP land area in The TSSWCB addresses the prevention or abatement of nonpoint source pollution through the WQMP program. The WQMP program provides agricultural producers in priority areas, such as the Middle and South Bosque Rivers and Hog Creek watersheds, an opportunity to comply with State water quality laws through traditional, voluntary, incentive-based programs. The TSSWCB oversees and is responsible for the financial assistance component of the program. Local SWCDs provide or arrange for technical assistance for applicants to develop WQMPs. Based on information provided by the TSSWCB, 25 WQMPs were implemented impacting about 10 percent of the crop and pasture land within these watersheds as part of the technical assistance project (06-9). 3

12 Figure 3. Map showing the segments and subsegments within the Middle and South Bosque Rivers and Hog Creek watersheds. Through this water quality monitoring project, data were collected to help assess the efficacy of implemented WQMPs and other best management practices in the Middle and South Bosque Rivers and Hog Creek watersheds in reducing nonpoint source pollution, specifically nitritenitrate nitrogen. Historical water quality data collected by TIAER between 1995 and 2003 were compared with data collected under the current project between 2008 and The goal was to evaluate whether changes in water quality were occurring and if so, could these changes be associated with implementation of WQMPs or other factors occurring within the South and Middle Bosque Rivers and Hog Creek watersheds. Because a known change in the watershed occurred between the historical and current monitoring data (i.e., implementation of WQMPs) and there was a large gap between the two monitoring periods, a before/after analysis approach was used. Historical water quality monitoring data from 1995 to 2003 was defined as the before period, while data from 2008 through 2010 was defined as the after period. 4

13 Watershed and Site Descriptions Monitoring activities during the after period focused on five stream sites within the Middle and South Bosque Rivers and Hog Creek watersheds (Figure 2). These five sampling stations were selected as representative of land uses activities within this region (Table 1). Information on the general land use of these drainage areas was based on a classification of satellite imagery from 2001 through 2003 conducted by the Spatial Sciences Laboratory of the Texas Agricultural Experiment Station (now, Texas AgriLife; Narasimhan et al., 2005). Drainage areas above sampling sites were delineated from 30-meter digital elevation models created from United States Geological Survey (USGS) 1:24,000 topographic maps using the AVSWAT 2000 extension within ArcView. TIAER Site ID Table 1. Estimated land use and drainage area above sampling sites. Wood & Range (%) Pasture (%) Cropland (%) Urban (%) Other (%) Total Area (Aces) HC ,900 MB ,700 SB ,700 TC ,400 WC ,300 Historical water quality data collected by TIAER were also available for the same or nearby sites for comparison (Table 2). Historical or before data were collected by TIAER between September 1995 and March 2003 and largely represent projects that focused on water quality throughout the Bosque River watershed. These historical or non-direct data were collected and analyzed by TIAER under QAPPs for the following projects: 1. Data collected by the Brazos River Authority (BRA) and TIAER, as a subcontractor, under the TCEQ Clean Rivers Program. The QAPP is the BRA document entitled Quality Assurance Project Plan for the Bosque River Watershed Pilot Project, which encompasses data collected from September 1, 1995 through May 31, Data collected by TIAER under the United State Department of Agriculture (USDA) Lake Waco-Bosque River Initiative. The QAPP is TIAER document entitled Quality Assurance Project Plan for the Lake Waco-Bosque River Initiative, which encompass data collected from September 1, 1996 through September 1, Data collected by TIAER under the Clean Water Act Section 319(h) Nonpoint Source Pollution Control Program Middle and South Bosque Assessment Project. This project includes data collected from February 2000 through February 2001 under a TSSWCB and EPA approved QAPP. 5

14 TIAER Site ID Table 2. Site locations and monitoring history of samples collected by TIAER through September TCEQ ID Watershed and General Location Monitoring Period Date of First Grab Sample Date of Last Grab Sample Date of First Storm Sample Date of Last Storm Sample HC MB SB050 a SS001 c SB065 d TC Hog Creek at FM 185 Middle Bosque at FM 3047 South Bosque off Church Road South Bosque at U.S. Hwy 84 South Bosque at Old Lorena Road Tonk Creek at FM 938 Before 26-Sep Mar Nov Mar 2003 After 22 Apr Sep May Sep 2010 Before 07 Dec Mar Feb Mar 2003 After 22 Apr Sep May Sep 2010 Before NA b NA 23 Jan Mar 2003 After NA NA NA NA Before 05 Nov Mar 2003 NA NA After NA NA NA NA Before NA NA NA NA After 22 Apr Sep May Sep 2010 Before 26 Sep Mar Apr Mar 2003 After 22 Apr Sep May Sep 2010 WC Wasp Creek at FM 938 Before 26 Sep Mar Jul Mar 2003 After 22 Apr Sep May Sep 2010 a. Initial dates include data from a more downstream site (SB060, TCEQ ID South Bosque at FM 2837) that was moved to SB050 in July 1997 due to backwater impacts from Lake Waco. b. NA indicates not applicable, because no monitoring occurred at this site during the period indicated. c. Initial dates include grab data from at a more upstream site (SB060). Routine grab sampling at SB060 was shifted to SS001 in March d. Site SB065 was established instead of using SS001 and SB050 due to problems with access and changes in the area due to bridge and road construction. Nondirect data from SS001 and SB050 were combined and used to represent the before water quality for the South Bosque River. 6

15 Because of differences in project objectives and changes in stream access between the before and after periods, the exact same sites were not monitored consistently for South Bosque River. For the South Bosque River during the before period, storm sampling occurred at site SB050, and for safety reasons, routine grab sampling occurred about a mile downstream at site SS001. During the after period, access to sites SS001 and SB050 was no longer feasible. Site SB050 was located on private land and due to a change in landowners was no longer accessible. Road and bridge work along the South Bosque River changed the access to site SS001 also making it impractical to monitor during the after period. Instead of sampling at sites SS001 and SB050, sampling occurred at a new location (SB065) near the bridge where the South Bosque crosses Old Lorena Road. Site SB065 is located about a quarter mile (half a kilometer) downstream of SB050. For evaluating changes in water quality over time, data from sites SB050 and SS001 were compared to data from site SB065 and are jointly referred to as site SB065 throughout the rest of this report. Methods Unless specifically noted, sampling and laboratory analysis methods outlined below apply to both the before and after periods. While some differences are noted, impacts of these differences on resulting water quality data were considered minimal for the purpose of comparing data between the before and after periods. Routine Grab Sampling Routine grab sampling at all sites was performed on a biweekly basis when flow was present. Samples were not collected when sites were dry or pooled. Samples were collected at a depth of about 0.08 to 0.15 meters (0.25 to 0.5 ft). Of note, for non-direct data collected prior to October 2003, filtration and preservation other than temperature reduction by placing samples in coolers with ice was performed in the laboratory. Beginning in October 2003, sample collection procedures were changed to follow TCEQ procedures prescribing filtration and acid preservation in the field for routine grab samples (TCEQ, 2003). Routine samples for nutrients and total suspended solids (TSS) were collected in a one-liter plastic bottle. Prior to October 2003, this one-liter bottle was turned into the lab for analyses without any further field preparation besides transport on ice. All acidification and filtration steps occurred in the lab. Starting in October 2003 acidification and filtration activities for grab samples were implemented in the field following TCEQ procedures (TCEQ, 2003). Aliquots for analytes requiring filtration and/or acidification were taken from the 1-L bottle after it had been agitated thoroughly to ensure total mixing of sediments. Samples that required field filtration were filtered through a micron filter using a 50 CC (or larger) syringe or using a filtration flask and pump. An aliquot for nitrite-nitrogen plus nitrate-nitrogen (NO 2 -N+NO 3 -N) and ammonia-nitrogen (NH 3 -N) was filtered and transferred to an acidified 60-mL plastic bottle, labeled, capped, and shaken to disperse the acid in the sample. A filtered aliquot for orthophosphate-phosphorus (PO 4 -P) analysis was iced and submitted to the lab in the syringe or in a bottle. An aliquot for total phosphorus (total P) and total Kjeldahl nitrogen (TKN) was poured from the 1-L bottle into a labeled and acidified 250-mL plastic bottle, which was capped and shaken to disperse the acid. The remaining sample (about 500 ml) was submitted to the lab for TSS analysis. Of note, if samples were too turbid preventing field filtration in a reasonable 7

16 amount of time, a comment was added to the chain of custody form and aliquots associated with constituents requiring filtration were kept in the one-liter bottle for filtration and acidification. Therefore in situations when the samples were too turbid, the filtration and acidification steps were performed in the lab. In addition to nutrients and TSS, chlorophyll-a (CHLA) and bacteria (either fecal coliform or Escherichia coli) were analyzed with routine grab samples. Chlorophyll-a samples were collected in an amber plastic 1-L bottle. During the after period, bacteria samples were collected in sterile, disposable plastic 250 ml bottles that were factory autoclaved and sealed and included an addition of 10 percent sodium thiosulfate to minimize the impact of potential chlorine residuals. For samples collected during the before period (prior to April 2003), bacteria samples were collected in sterile, disposable-plastic bottles; the addition of thiosulfate was done by the lab if a positive residual for chlorine was tested. All samples for bacteria were screened in the laboratory for the presence of chlorine residuals. Bacteria and CHLA samples were iced immediately in the field, and transported to the laboratory. While routine grab samples for lab analysis were being collected, measurements were taken insitu and recorded for water temperature, dissolved oxygen, ph, and specific conductance (conductivity) using a multiprobe instrument. Storm Sampling Storm sampling was conducted using an Isco 4230 or 3230 bubbler-type flow meter in conjunction with an Isco 3700 sampler; both enclosed in a sheet-metal shelter. Electrical power was provided by marine deep-cycle batteries recharged by solar cells located near or on top of each shelter. Each flow meter recorded water level at five-minute intervals by measuring the pressure required to force an air bubble through a 3 mm (0.125 inch) polypropylene tube. The Isco flow meters were programmed to initiate sample retrieval by the Isco 3700 samplers upon a set rise in water level. Sampling by the Isco 3700 samplers was generally triggered by a rise in water level of about 4 to 8 cm (0.12 to 0.24 ft) above baseflow. Once activated the sampler would retrieve one-liter sequential samples. The typical sampling sequence for the smaller watershed sites (TC020 and WC020) was: An initial sample Three samples taken at one-hour intervals Four samples taken at two-hour intervals All remaining samples taken at six-hour intervals For the larger watershed areas captured by stations HC060, MB063 and SB065, the sampling sequence was modified to allow for a more extended hydrograph. The sampling sequence at these sites was as follows: An initial sample One sample taken at a one-hour interval One sample taken at a two-hour interval One sample taken at a three-hour interval One sample taken at a four hour interval 8

17 One sample taken at a six-hour interval All remaining samples taken at eight-hour intervals Storm samples were iced immediately in the field and transported to the lab for analysis. For storm samples collected during the before period prior to mid-june 2000, each individual sample was analyzed by the lab. After mid-june 2000, storm samples from a given site were composited on about a daily basis using a flow-weighting strategy. A flow-weighting strategy was implemented to decrease the overall number of storm samples submitted for laboratory analysis. The flow-weighting strategy used stage data recorded during a storm, the rating curve developed for each site, and a TIAER-developed computer program. During sample collection, stage data were uploaded from data loggers to portable computers, and then downloaded at TIAER headquarters for use with the computer program. The program reads the water level or stage associated with the time interval for each sample collected at a site, correlates the stage to flow using the site's rating curve, and calculates the amount of flow associated with each water sample taken during the storm event. For a group of bottles, the program would then designate the amount to be taken from each bottle to compose a one-liter composite based on the relative volume of flow associated with each bottle within the group. This flow-weighting strategy allowed a reduction in sample load without compromising the intended use of the data in determining storm loadings of waterborne constituents and storm-event mean concentrations. Site specific stage-discharge relationships were developed from manual wading-type flow measurements taken at various water levels following United State Geological Survey (USGS) methods (Buchanan and Somers, 1969). Stage-discharge relationships were extrapolated for water levels too deep to permit safe wading using the cross-sectional area and a least-squares relationship of average stream velocity to the log of water level. In addition at sites HC060 and MB063, information from corresponding USGS gauging stations ( on Hog Creek near Crawford, Texas and on Middle Bosque River near McGregor, Texas) were used to aid in the development of stage-discharge relationships. At site WC020 on Wasp Creek, the rating curve was based on a culvert equation using the dimensions and slope of the culvert at the corresponding road crossing with stage-discharge measurements used for verification. If for some reason (e.g., equipment failure), the automated sampler failed to collect samples, a storm grab sample was collected for analysis. If samples could not be flow-weighted because stage data were missing or could not be electronically downloaded at the time samples were retrieved, storm samples were analyzed individually. Routine laboratory analyses of storm samples included nitrogen and phosphorus constituents as well as TSS. Laboratory Analyses Routine grab and storm samples were evaluated for NH 3 -N, NO 2 -N+NO 3 -N, TKN, PO 4 -P, total P, and TSS (Table 3). Routine grab samples also included the analysis of CHLA and bacteria. In the before period, fecal coliform (FC) was generally analyzed with routine grab samples for evaluating bacteria concentrations. From April 2002 through March 2003 with selected grab samples, TIAER analyzed both FC and E. coli using plating methods, because TCEQ was in the process of changing the water quality criteria for bacteria from FC to E. coli (TNRCC, 2000). During the after period, all bacteria samples were analyzed for E. coli using the IDEXX Colilert method. 9

18 Table 3. Abbreviations, units and analysis methods for water quality constituents. Constituent Abbreviation Units Method a Field Measurements b Specific Conductance Conductivity µs/cm EPA c or SM d 2510B Dissolved oxygen DO mg/l EPA ph ph standard units EPA Water temperature Water temp. ºC EPA Laboratory Measurements Chlorophyll-a CHLA µg/l SM 10200H Escherichia coli E. coli colonies/100 ml e SM 9222G/9223B Fecal coliform FC colonies/100 ml SM 9222D Ammonia-nitrogen f NH3-N mg/l SM 4500-NH3-G or EPA Nitrite-nitrogen+nitrate-nitrogen f NO2-N+NO3-N mg/l SM 4500-NO3-F or EPA Total Kjeldahl nitrogen TKN mg/l SM 4500-NH3-GEPA Orthophosphate-phosphorus f PO4-P mg/l SM 4500P-E or EPA Total phosphorus TP mg/l EPA Total suspended solids TSS mg/l SM 2540-D or EPA a. Although more than one method may be listed for a constituent, the methods listed are comparable and indicate only that a different reference (either EPA or SM) was used depending when the analysis was completed. b. All field activities follow guidelines as outlined in the most recent version of TCEQ s Surface Water Quality Monitoring Procedures Manure (i.e., TCEQ, 2003; 2008). c. EPA refers to Methods for Chemical Analysis of Water and Wastes (USEPA, 1983). d. SM refers to the Standard Methods for the Examination of Water and Wastewater, 18th (APHA, 1992) or most recent online edition. e. The IDEXX method for E. coli was implemented in April Results from the IDEXX method are reported as most probable number (MPN)/100 ml whereas plating technique results are reported as colonies/100 ml. In this report, data for all E. coli results are presented in units of colonies/100 ml. f. Field-filtering for NH 3 -N, NO 2 -N+NO 3 -N, and PO 4 -P began in October Censored Data Data Management and Statistical Analyses Left censored data, indicated as below the reporting limit (RL), were entered into the database as one-half the RL following recommendations by Gilliom and Helsel (1986) and Ward et al. (1988). Prior to 2003, method detection limits (MDLs) were used as the reporting limit. Starting in 2003, some TIAER projects, but not all, started to require ambient water reporting limits (AWRLs) set by TCEQ as data reporting limits. TIAER has continued to evaluate MDLs as part of good laboratory practice, but has shifted to using AWRLs for most projects unless another reporting limit is specified by the project sponsor for a constituent. For data evaluation purposes, the highest RL noted during the before and after periods for each constituent was determined (Table 4) and the half-value was used to left-censor data for each constituent, except bacteria. For bacteria, the maximum RL of 1 colony/100 ml was used to left-censor values. This standardization of the RL across the before and after periods was done to avoid detection of a change between the two periods in relation to changes in the RL. 10

19 Table 4. Constituent and maximum reporting limit used to left-censor data prior to data analysis. Constituent CHLA Bacteria (E. coli or FC) NH3-N NO2-N+NO3-N TKN PO4-P TP TSS Maximum Reporting Limit 5.4 µg/l 1 colony/100 ml 0.10 mg/l 0.05 mg/l 0.30 mg/l mg/l 0.11 mg/l 10 mg/l For bacteria data, it is possible that there may be right censored data or values reported as greater than a given number. The datasets evaluated were screened for right-censored data and no rightcensored values were found. Bacteria Data Of note, grab samples during the before period were analyzed almost exclusively for FC, while due to a change in the bacteria water quality standard established by TCEQ, samples during the after period were analyzed for E. coli. To compare bacteria concentrations between the before and after periods, FC data were converted to E. coli equivalents using a regression relationship developed comparing between the two methods. This regression relationship was based on samples collected at sites throughout the Bosque River watershed between November 2000 and March 2004 for which both E. coli and FC were analyzed (McFarland and Millican, 2010). This period of overlap was used to determine if FC could be adjusted to comparable E. coli values using accepted statistical methods for comparing different analytical methods (Bland and Altman, 1986). The agreement between these two methods of bacterial analysis was evaluated by McFarland and Millican (2010) and produced the following regression relationship based on a comparison of 1075-paired observations: ln(e. coli) = * ln(fc) R 2 = 0.93 McFarland and Millican (2010) observed that this regression relationship did not meet all the assumptions associated with regression analysis. That is the distribution of residuals was peaked and, thus, not normally distributed even after data were log-normally transformed. It was assumed that the regression model developed was robust enough that this violation to the statistical assumption of normally distributed residuals would have only a minor impact on the conversion of FC concentrations to equivalent E. coli concentrations for comparison between the before and after periods. Also of note, the regression relationship developed was based on plating techniques for both FC and E. coli, while E. coli concentrations during the after period were evaluated using IDEXX, a chromogenic method (Noble et al., 2003). While no overlapping data were collected to compare the plating versus the IDEXX method for E. coli, several studies have concluded that 11

20 results from these two methods are comparable (Noble et al., 2004; 2003; Hargett and Goyn, 2004; Eckner, 1998) and TCEQ accepts E. coli data using either method. In lieu of alternatives, the regression relationship between FC and E. coli was applied for comparisons of bacteria concentrations as the best available information. Definition of Before and After Periods The before period was defined as data collected prior to April 2003 (generally between September 1995 and March 2003); while the after period was defined as data collected between April 2008 and September 2010 (see Table 2). Exploratory Data Analyses and Basic Statistics As an initial review of the data, exploratory data analysis (EDA) was performed for routine grab data and storm-event mean concentrations (EMCs). The EDA approach is a graphical technique used to characterize distributional properties, identify potential outliers, and evaluate patterns in the data for defining appropriate statistical tests (Tukey, 1977). The EDA included histograms, time series plots, and box and whisker plots focusing on the before and after time periods. The potential for seasonality was evaluated separately for each period using monthly correlograms (Reckhow et al., 1993). No seasonal patterns were observed for any of the constituents for routine grab or EMCs. Evaluations also included plots of flow versus water quality concentrations due to the anticipated impact of flow on most nonpoint source parameters (Helsel and Hirsch, 1992). Basic statistics (mean, median, standard deviation, minimum and maximum) were calculated and are presented in Appendix A by site for the before and after periods separately for routine grab samples and event mean concentrations (EMCs). Time series plots of NO 2 -N+NO 3 -N as the primary constituent of concern are presented in Appendix B. Calculation of Event Mean Concentrations (EMCs) Event mean concentrations for storm events were calculated by accumulating the mass via rectangular integration using a midpoint rule to associate concentration with streamflow for each storm hydrograph (Stein, 1977). Instantaneous 5-minute stage readings were used as the minimum measurement interval to indicate flow in cubic feet per second (cfs) and multiplied by 300 seconds to obtain flow for each 5-minute interval. The flow associated with each 5-minute interval was multiplied by the associated water quality concentration and summed across the event to calculate the total constituent loadings. Total constituent loadings were divided by total storm volume to calculate EMCs. In determining the use of a storm event for data analysis purposes, hydrographs of each storm event were graphically reviewed for anomalies and the availability of water quality data was assessed. Storms with large data gaps or questionable hydrographs were excluded from the datasets used for statistical analyses. Before and After Analysis Using ANCOVA To evaluate the impact of WQMP efforts and other activities in the South and Middle Bosque Rivers and Hog Creek watersheds, before and after analyses were conducted on routine grab and storm data using both parametric and nonparametric statistics as a step trend (Hirsch, 1988). Step trend procedures were used because there were gaps in the data record for all five sites breaking the data into two distinct time periods (see Table 2) and because there was a known 12

21 event (WQMP implementation during the post period) that was expected to result in a change in water quality (Helsel and Hirsch, 1992). Data collected prior to April 2003 were designated as the before period while data collected after April 2003 were designated as the after period (Table 2). The data were analyzed as a before/after monitoring design (Grabow et al., 1999; Smith, 2002; Spooner et al., 1985) using analysis of covariance (ANCOVA) and the nonparametric Wilcoxon rank sum (WRS) procedures using SAS statistical software (SAS, 2000). In the ANCOVA procedure, average flow for each storm event was used as the covariate and two regression lines were developed relating concentration to flow, one for each period. To satisfy assumptions of homogeneity of variance and homogeneity of regression, ANCOVA was performed on the data using a natural-log transformation (Littell et al., 1996). For the ANCOVA procedure to clearly indicate significant differences between the before and after periods, three steps need to be considered (Littell et al., 1996; NRCS, 1997). First, the linear regression equations relating streamflow and concentration for each monitoring period should be significant; second, the slopes of these two regression lines should be equal; and third, the intercepts of these two regression line should be significantly different. The ANCOVA evaluates differences among treatment level means ( before and after periods) that would occur if all concentrations had the same streamflow (Keppel, 1991), and estimated means from the ANCOVA represent flow-adjusted means based on this average streamflow. In the WRS analysis, routine grab samples and EMCs were flow adjusted prior to analysis using locally weighted regression and smoothing scatterplots (LOWESS) with a smoothing coefficient of 0.5 (Helsel and Hirsh, 1992; Bekele and McFarland, 2004). The residuals from LOWESS regression were then used in the WRS test. Both parametric and nonparametric procedures were implemented in the before and after analysis, because assumptions associated with the ANCOVA could not be fully met for all constituents at all sites. In addition, the application of both parametric and nonparametric methods on the same dataset is considered useful because it provides assurance in the interpretation of results (NRCS, 1997). A step trend confirmed by both analyses is considered more meaningful than one indicated by only one test. Statistical significance was evaluated at an = 0.05 probability level. Routine Grab Samples Water Quality Before/After Results As an overview of water quality conditions during the before and after periods, summary statistics of the flow adjusted and natural log (ln) transformed data were back transformed into original units (Table 5). Because the standard deviation of ln transformed data is not symmetrical about the mean when back transformed, a lower and upper error bound was calculated representing the range about the mean plus or minus one standard deviation. 13

22 Generally, lower mean concentrations were found during the after compared to the before period for most constituents, although the standard error bounds often overlapped making it difficult without further evaluation to determine if these differences were statistically meaningful. The ANCOVA and WRS results indicated significant decreases for several constituents across the various sites (Table 6). Most notably all five sites showed highly significant decreases for NO 2 -N+NO 3 -N. For other constituents, decreases between the before and after periods varied by site. For NH 3 -N, significant decreases were noted at all sites, while TKN indicated decreases only at sites HC060 and TC020. For the P constituents, decreases in PO 4 -P and total P were indicated at TC020 and WC020. A decrease in PO 4 -P was also indicated at HC060 but only based on the ANCOVA test, and an increase in total P was indicated at SB065 but based only on the WRS test. Site SB065 was the only of the five sites to indicate decreases in CHLA, while TC020 was the only site to indicate decreases in E. coli and TSS. For the most part, similar findings occurred with the ANCOVA and WRS tests, but notable exceptions occurred for PO 4 -P at HC060 and total P for SB0605. These two tests evaluated the data in different ways. The ANCOVA, a parametric test, evaluated mean values assuming a linear relationship in adjusting for flow, while the WRS, a nonparametric test, evaluated median values of data that were flow-adjusted using a nonlinear method (LOWESS). Because these statistical methods use different approaches to evaluate these data, the exact same findings were not expected in all instances, but when both tests indicated the same finding, more strength was given to that result. To give an idea of the level of change occurring in water quality, the absolute and relative differences in mean concentrations based on the ANCOVA were calculated (Table 7). Of note, not all changes presented were significant. The changes associated with differences that were not significant are presented to give an idea of the level of change that might be needed before significant changes occur. For NO 2 -N+NO 3 -N, the largest absolute decrease was mg/l at WC020 on Wasp Creek. Fairly large absolute decreases in NO 2 -N+NO 3 -N were also noted at TC020 on Tonk Creek (-3.58 mg/l) and SB065 along the South Bosque River (-3.98 mg/l). These three sites also had the highest NO 2 -N+NO 3 -N concentrations during the before period. At all five sites, NO 2 -N+NO 3 -N indicated relative decreases from 37 to 65 percent. Of note, when the assumption of equal slopes was not met in the ANCOVA (Table 6), the absolute and relative changes presented in Table 7 are only estimates based on average measured flow conditions. This means that the amount of change varied with flow and was not consistent across all flow conditions. As a secondary focus of the project, significant decreases in bacteria, as presented by E. coli, occurred at TC020 with a relative decrease of just over 50 percent. At WC020, a decrease of about 24 percent was noted, but due to the large variability associated with bacteria concentrations, a much larger decrease will be needed at WC020 before statistically significant differences can be shown. 14

23 Site Table 5. Summary statistics for routine grab samples for before and after monitoring periods. Data were transformed using a natural log transformation and flow adjusted as per ANCOVA procedures and then back transformed into original units. Constituent Number of Events Mean Lower Standard Error Bound Upper Standard Error Bound Before After Before After Before After Before After HC060 CHLA (µg/l) E. coli (colonies/100 ml) NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l) MB063 CHLA (µg/l) E. coli (colonies/100 ml) NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l) SB065 CHLA (µg/l) E. coli (colonies/100 ml) NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l) TC020 CHLA (µg/l) E. coli (colonies/100 ml) NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l) WC020 CHLA (µg/l) E. coli (colonies/100 ml) NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l)

24 Table 6. P-values from analysis of covariance (ANCOVA) and Wilcoxon rank sum (WRS) statistics comparing routine grab sample concentrations between before and after periods. Arrows indicate significant increases and decreases ( = 0.05) in constituent concentrations. Site Statistic CHLA E. coli NH3-N NO2-N +NO3-N TKN PO4-P Total P TSS HC060 ANCOVA NA a ( ) ( ) NA ( ) WRS < ( ) ( ) ( ) MB063 ANCOVA ( ) ( ) WRS ( ) < ( ) SB065 ANCOVA ( ) NA NA WRS ( ) < ( ) ( ) ( ) TC020 ANCOVA NA ( ) < ( ) < ( ) NA ( ) ( ) ( ) WRS ( ) ( ) < ( ) ( ) ( ) ( ) ( ) WC020 ANCOVA ( ) < ( ) < ( ) < ( ) WRS ( ) < ( ) < ( ) ( ) a. NA indicates not applicable because the assumption of equal slopes for ANCOVA was not met between the pre and post periods at =

25 Table 7. Estimated change in flow-adjusted, grab concentrations before and after. Site Constituent Before a After a ANCOVA b WRS b 17 Absolute Change Relative Change (%) c HC060 CHLA (µg/l) ns E. coli (colonies/100 ml) ns ns NH3-N (mg/l) ** ** NO2-N+NO3-N (mg/l) ** ** TKN (mg/l) ** PO4-P(mg/L) * ns Total P (mg/l) ns ns TSS (mg/l) 7 6 ns ns MB063 CHLA (µg/l) ns ns E. coli (colonies/100 ml) ns ns NH3-N (mg/l) ** ** NO2-N+NO3-N (mg/l) ** ** TKN (mg/l) ns ns PO4-P(mg/L) ns ns Total P (mg/l) ns ns TSS (mg/l) 7 6 ns ns SB065 CHLA (µg/l) * ** E. coli (colonies/100 ml) ns ns NH3-N (mg/l) ** NO2-N+NO3-N (mg/l) ** TKN (mg/l) ns ns PO4-P(mg/L) ns ns Total P (mg/l) ns * TSS (mg/l) ns ns TC020 CHLA (µg/l) ns E. coli (colonies/100 ml) ** ** NH3-N (mg/l) ** ** NO2-N+NO3-N (mg/l) ** ** TKN (mg/l) ** PO4-P(mg/L) ** ** Total P (mg/l) * * TSS (mg/l) 7 6 * * WC020 CHLA (µg/l) ns ns E. coli (colonies/100 ml) ns ns NH3-N (mg/l) ** * NO2-N+NO3-N (mg/l) ** ** TKN (mg/l) ns ns PO4-P(mg/L) ** ** Total P (mg/l) ** ** TSS (mg/l) ns ns a. Back transformed from natural log into original linear scale, where before and after represent mean concentrations for routine grab samples adjusted for the covariate flow on the natural log scale from the ANCOVA procedure. b. For the ANCOVA and WRS, indicates not applicable, ns indicates not significant, * indicates significant at = 0.05 and ** indicates significant at = c. Percent change on a linear scale calculated as ([After Before]/Before)*100.

26 Storm Events In general, higher average concentrations occurred for storm events than for routine grab samples (Table 5 and 8). An exception occurred for NO 2 -N+NO 3 -N at sites SB065, TC020 and WC020 where average routine grab concentrations were higher than average EMCs. As with routine grab samples, most EMCs decreased between the before and after periods (Table 8). Table 8. Summary statistics for routine grab samples for before and after monitoring periods. Data were flow-adjusted and transformed using and natural log transformation and then back transformed into original units. Site Constituent Number of Events Mean Lower Standard Error Bound Upper Standard Error Bound Before After Before After Before After Before After HC060 NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l) MB063 NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l) SB065 NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l) TC020 NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l) WC020 NH3-N (mg/l) NO2-N+NO3-N (mg/l) TKN (mg/l) PO4-P (mg/l) Total P (mg/l) TSS (mg/l)