Nonpoint Source Pollutant Monitering on the Ponaganset River Watershed

The Ponaganset River Basin consists of an area of 14.4 mi located in the town of Foster, Rhode Island. This area is located within the northwest region of the Scituate Watershed. The source of this river comes from the Ponaganset Reservoir with an area of 2.1 mi and a storage capacity of 742 MG. Water quality samples were collected at United States Geological Survey (USGS) site (01115187) which is approximately 5 miles down stream from the reservoir and 0.4 miles upstream from Barden Reservoir. The Ponaganset River has the largest mean daily discharge of all the sampling locations in the Scituate Reservoir watershed. The concept of this analysis originated with the 1995 Water Quality Protection Plan which sited a lack of wet weather data on the Scituate Reservoir Watershed. No wet weather data was collected from the watershed between 1995 and 2003. In 2003, the Water Quality Protection Plan again sited a lack of wet weather data on the Scituate Reservoir Watershed as one of the major weaknesses. The plan recommended the need to determine potential wet weather impacts as well as the potential sources of those impacts on the environment. The objective of this analysis was to determine non-point sources of pollutants which contribute to the river and establish a preliminary wet weather monitoring program to determine pollutant loads contributed by stormwater runoff. In addition, this analysis was intended to establish a procedure to extrapolate wet and dry weather data from a characteristic sub watershed to the entire Scituate Watershed. In this study, the Ponaganset River site was selected based upon preliminary research, historical water data, and range of flows for the selected site. The data collected during wet weather sampling provided insight into the behavior of the sources during various storm events as well as storm characteristics. The information acquired for use in this analysis was used to explore load characteristics using linear and multiple regression models to predict loads then apply them to monthly and annual parameter data to determine if the site is either influenced more with dry weather or wet weather. As more stringent water quality standards continue to increase, monitoring the health of the watershed will increase as well. Evaluating the water quality under dry and wet weather conditions seems fitting to answer some of these questions in addition to fulfilling the requirements of this thesis. In this study, water quality results, loads, and linear/multiple regression models are used to determine load characteristics that exist at this site and to relate this information to the entire watershed. The field data used to develop the statistical models was conducted solely by the investigator and all samples were tested by Premier Laboratory in Dayville, Connecticut. Sampling and monitoring for the analysis occurred for a period of approximately two years during the months from April to September in 2005 and 2006. Three wet weather events were successfully captured for the wet weather program: Storm 1 (May 2-4, 2006), Storm 2 (July 12-14, 2006), and Storm 3 (September 19-20, 2006). A total of twelve dry weather samples were collected between April through August 2005 and May through September 2006. The initial samples collected consisted of total suspended solids (TSS), biological oxygen demand (BOD), inorganic constituents, total trace metals, and nutrients. During sample collection the introduction of errors was always a concern and careful consideration was taken to avoid any contamination to the water samples. A strict regimen of water sample collection techniques, preservation, and laboratory analysis were carefully adhered to avoid any contamination. Concentration data and flow data were used to calculate the mass load. With the use of the water quality data collected at the site, it allowed for the development of empirical equations used to determine dry and wet weather loads. _ Linear regression models were developed for dry weather and multiple linear regression models were developed for wet weather conditions for selected constituents. The six primary constituents included barium, manganese, aluminum, iron, sodium, and chloride. A limited amount of total coliform bacteria data was also included in the analysis. The largest loads observed at the site included sodium and chloride during wet weather conditions. The equations were later applied to hydrograph data which had been generated for a period of a year that occurred from October 1, 2003 to September 30, 2004. Although the data set used to develop the models was limited to twelve dry weather samples and three storm events, the data showed that it could be applied to the monthly and annual parameter data used to describe dry and wet weather load characteristics for this sub-watershed. The application of the mathematical models indicates that the Ponaganset River watershed is both dry and wet weather influenced. Finally, the analysis provides a procedure to determine annual loads and provide recommendations for future wet weather assessment for the entire Scituate Reservoir. Further evaluations of wet weather monitoring within the Scituate Reservoir Complex will be needed to access the overall health of the watershed. A team effort is needed as planning is crucial in order to gather accurate data. The findings of this analysis may lead to a more extensive wet and dry weather analysis encompassing the entire watershed. ACKNOWLEDGEMENTS I would like to acknowledge my utmost gratitude to those people who have helped me to successfully achieve the completion of this thesis: Dean Raymond Wright has been instrumental in my quest for knowledge and the completion of my graduate degree. His support and guidance have given me the confidence to continue to strive to the best of my ability. His knowledge of his subject has surpassed anyone's I have ever met and I am truly grateful to have learned all that I have from him. I would also like to thank Dr. Vincent Rose and Dr. Tom Boving who have helped to me to complete this thesis and take it to another level. Just as I thought I couldn't learn anymore on this analysis I did. Richard Blodgett, has also been a tremendous help with his knowledge of the watershed, creating GIS Maps specific to the site, funding for samples tested by Premier Laboratory in cooperation with Providence Water Supply Board, and his confidence in my ability to complete this analysis. Others whom I work with at the Providence Water Supply Board have also greatly assisted me in endless questions I have had through the various stages of this thesis that include: Frank Healy, Steve Soito, Chris Labossiere, & Nancy Crosby. Frank Healy has been the one individual who has helped me through

The objective of this analysis was to determine non-point sources of pollutants which contribute to the river and establish a preliminary wet weather monitoring program to determine pollutant loads contributed by stormwater runoff. In addition, this analysis was intended to establish a procedure to extrapolate wet and dry weather data from a characteristic sub watershed to the entire Scituate Watershed. In this study, the Ponaganset River site was selected based upon preliminary research, historical water data, and range of flows for the selected site. The data collected during wet weather sampling provided insight into the behavior of the sources during various storm events as well as storm characteristics. The information acquired for use in this analysis was used to explore load characteristics using linear and multiple regression models to predict loads then apply them to monthly and annual parameter data to determine if the site is either influenced more with dry weather or wet weather.
As more stringent water quality standards continue to increase, monitoring the health of the watershed will increase as well. Evaluating the water quality under dry and wet weather conditions seems fitting to answer some of these questions in addition to fulfilling the requirements of this thesis. In this study, water quality results, loads, and linear/multiple regression models are used to determine load characteristics that exist at this site and to relate this information to the entire watershed.
The field data used to develop the statistical models was conducted solely by the investigator and all samples were tested by Premier Laboratory in Dayville, Connecticut. Sampling and monitoring for the analysis occurred for a period of approximately two years during the months from April to September in 2005 and 2006. Three wet weather events were successfully captured for the wet weather program: Storm 1 (May 2-4, 2006), Storm 2 (July 12-14, 2006), and Storm 3 (September 19-20, 2006). A total of twelve dry weather samples were collected between April through August 2005 and May through September 2006. The initial samples collected consisted of total suspended solids (TSS), biological oxygen demand (BOD), inorganic constituents, total trace metals, and nutrients.
During sample collection the introduction of errors was always a concern and careful consideration was taken to avoid any contamination to the water samples.
A strict regimen of water sample collection techniques, preservation, and laboratory analysis were carefully adhered to avoid any contamination.
Concentration data and flow data were used to calculate the mass load.
With the use of the water quality data collected at the site, it allowed for the development of empirical equations used to determine dry and wet weather loads. _ Linear regression models were developed for dry weather and multiple linear regression models were developed for wet weather conditions for selected constituents.
The six primary constituents included barium, manganese, aluminum, iron, sodium, and chloride. A limited amount of total coliform bacteria data was also included in the analysis. The largest loads observed at the site included sodium and chloride during wet weather conditions. The equations were later applied to hydrograph data which had been generated for a period of a year that occurred from October 1, 2003 to September 30, 2004. Although the data set used to develop the models was limited to twelve dry weather samples and three storm events, the data showed that it could be applied to the monthly and annual parameter data used to describe dry and wet weather load characteristics for this sub-watershed. The application of the mathematical models indicates that the Ponaganset River watershed is both dry and wet weather influenced. Finally, the analysis provides a procedure to determine annual loads and provide recommendations for future wet weather assessment for the entire Scituate Reservoir.
Further evaluations of wet weather monitoring within the Scituate Reservoir Complex will be needed to access the overall health of the watershed.
A team effort is needed as planning is crucial in order to gather accurate data.

LIST OF FIGURES
Of all the sampling located situated throughout the Scituate watershed, the Ponaganset River has been described as one of the principle streams in the basin which has been indicated in several USGS publications such as  and (Nimiroski, DeSirnrnone, and Waldron, 2008 • When greater than one inch of rain fell on the watershed, the Arithmetic and Thessian methods were not similar and indicated a greater deviation between values.
In comparing the standard practices of PWSB's arithmetic methodology, rainfall data is usually described on a monthly and yearly basis for their annual reporting, but to identify specific storm characteristics, real-time precipitation are very important.

Preliminary Research of the Ponaganset River Site
The initial evaluation entailed determining runoff characteristics for selected storms in conjunction with rainfall data provided by PWSB Later, further elimination was done based upon the shape, amount of precipitation, and if antecedent moisture conditions occurred before and after the storm event. After filtering the remaining hydrographs, a total of 15 storms (Table 2.1) were selected for computation of the actual unit hydro graph using methods described in . The direct runoff volume (V) was determined by subtracting the difference of the discharge minus base flow using the concave method of base flow separation then summing the direct runoff for the entire storm. The concave method (Figure 2.3) separates the base flow by extending the recession curve before the storm at the initial base flow where the discharge begins to increase to a point directly below the peak discharge (time to peak). From this point a straight line is then extended to a point on the discharge hydrograph at 41 hours after the peak in the river. The 41 hour duration (N) was determined by multiplying the drainage area (A) in square miles (14.4 mi 2 ) using the formula: N = aA 0 · 2 where a = 1 if the drainage area is determined in square miles. The runoff depth (P 0 ) of the storm event was divided by each ordinate of the direct runoff to finally obtain the actual unit hydrograph, since storm duration was unknown. In section 6.2, these parameters are described for the storms that were evaluated for the analysis in Table 6.1.    with the lack of gravel beneath the basin surface. This leads to the assumption of using the lowest value in the range of storage coefficients to accurately fit the characteristic description of the Ponaganset river basin. The idea of trying to simulate synthetic hydrographs was to achieve peak ranges from 300 to 400 cfs for a standard duration of net rainfall ranging from 2 to 8 hours then back fit them into the 15 unit hydrographs described in Table 2.1 which occurred from April through September 2002. This portion of the analysis determined that Snyder's Method was inconclusive with respect to the duration of the storm events selected for this report.

Conclusions of Preliminary Research
Initial evaluation of the Ponaganset River site indicated a range of recovery times for the river to return to base flow conditions, ranging from approximately two to three days. This established that a minimum two day antecedent dry period is required to clearly separate individual storm events. The rainfall analysis indicated that when ::::; 1 inch of rainfall fell on the entire

Description of Study Area
The Ponaganset River Basin is located in the upper north west portion of continuous fifteen minute interval monitoring for discharge, precipitation, gauge height, specific conductance, water temperature, and air temperature. In this analysis, it was critical to obtain the precise discharge measurements as well as precipitation amounts which were measured at fifteen minute intervals, which can be correlated with the exact time the water quality sample was collected. This data can then be used to determine instantaneous loads and total wet loads by separating out the base flow.

Site Selection Criteria
The analysis began with identifying all the water quality monitoring sites there is a wooden bridge that crosses over a narrow section of the Ponaganset River. Below this bridge there are probes with sensors that are mounted below the surface of the water. These probes extend across the river to measure its discharge, specific conductance, and water temperature which is shown in Figure   2.5: The real-time data that is recorded at this site is sent instantly through an antenna mounted on the top of the building to a satellite, and then to a server, which sends this information to a website published by USGS in real time format.

Water Quality Parameter Selection
The initial set of water quality samples proposed for testing at this site were selected based upon historic data obtained from Providence Water for records that dated back to 1995. In addition to previously tested sampled, some addition parameters were selected from the "Blackstone River Initiative" by  and other studies that have performed similar type analyses.
The following constituents were selected for initial testing as follows: acidity, alkalinity, turbidity, color, dissolved orthophosphate, nitrate, total phosphorous, sodium, chloride, ammonia, fecal coliform, total nitrogen, pH, biological oxygen demand (BOD), and total suspended solids (TSS). In addition, PWSB requested sixteen additional trace metals be included for testing during both dry and wet conditions until it could be determined that they were below detection. The following trace metals were included for testing as total metals: copper, cadmium, zinc, lead, aluminum, nickel, iron, chromium, selenium, arsenic, beryllium, silver, mercury, barium, manganese, and vanadium.
The wet weather data collected from the site was limited to three storm events having a total rainfall depth of greater than 0.1 inches and an antecedent dry period of a minimum of two days.  hrs.) due to the intensity (0.24 in./hr.) and duration of the storm (5 hrs.).
The resulting samples collected for the first storm event provided evidence that the majority of the initial constituents tested indicated results below detection limits, therefore these were eliminated from the sample set. The detectable constituents included barium, zinc, manganese, copper, aluminum, sodium, and iron. In addition, acidity, alkalinity, turbidity, color, chloride, pH, total suspended solids, and total coliform were monitored at the collection site. The water quality data was analyzed for consistency and instantaneous loads were calculated in lbs./day. Later in this report, the data will be used to predicted load equations using linear and multiple regression model equations. Based on these equations monthly and annual load estimates for this site will be determined.

Sample Collection Procedure
Two identical plastic buckets were purchased and used to collect the samples. These buckets are approximately 1.5 gallon capacity with a pouring spout. A rope was tied to the handle so the bucket could be lowered into the water below the center of the bridge. The bucket's spout assisted in pouring the river water into each of the prepared sample bottles to prevent any splashing.
Traveling to this location was also a very important consideration m selecting the Ponaganset River site. The approximate distance to the site is estimated at about 12 to 15 miles from the Providence Water Engineering Department office building. The shorter the distance required for travel to the collection site the better the response time for sampling prior to a storm event.
Initial sampling at the beginning of a wet weather event is vitally important and always requires manual sampling. The use of automated sampling devices was considered for this project. However, due to the various types of samples being tested (Table 3 .1 ), it would have required the purchase of four or five of these devices and would have added a greater expense to this project.
The site setup for dry weather sample collection was fairly simple to accomplish and required no temporary shelter, lighting, and five sample bottles, compared to fifty sample bottles used in wet weather. During dry weather sampling one set of constituents, which consisted of five sample bottles were collected on a bi-weekly basis. The approximate time frame required to prepare, collect, and process the samples for analysis was estimated to be two hours.
For each sample set collected at the Ponaganset River site there were five individual bottles. Later, non-detected constituents were eliminated which 25 brought the set down to four individual bottles. During dry weather sample collection one set of samples was collected. During wet weather sampling a total of eight or more sets of samples were taken to fully cover the initial, peak, and tail end of the hydro graph. Table 3.1 indicates the attached form which was utilized in conjunction with the chain of custody form provided by Premier Laboratory for the samples collected at the site.
The proceeding section is intended to explain the types of constituents that were detected at the Ponaganset River site. These characteristics include potential sources of infiltration into the watershed and their effects on the human body when ingested in excess amounts. A brief description of primary and secondary water quality standards for constituents that were detected at the Ponaganset River site will be described.
26 Though the years, there has been growing concern about the significance of trace elements in the environment. These elements are necessary for plant and animal growth in rivers such as the Ponaganset River. Excess exposure to trace elements, however, may disrupt the ecosystem and make the river unsuitable as a drinking water source. The United States Environmental Protection Agency (U.S. EPA) has developed standards to help protect the environment against harmful pollutants in natural waters. These standards are known as primary and secondary drinking water standards.

Primary Drinking Water Standards
"The National Primary Drinking Water Regulations (NPDWRs or primary standards) are legally enforceable standards that apply to public water systems. Primary standards protect public health by limiting the levels of contaminants in drinking water (U.S. EPA, 2005 NPDWR)".
These standards pertain to the finished water from a treatment plant to the furthest point in the distribution system.

Laboratory Selection
The samples for this project were tested by a certified laboratory (Premier Laboratory located in Dayville, Connecticut). The funding for testing was provided by Providence Water Supply Board (PWSB).
All testing followed the appropriate EPA standard methods. After the initial set of 32 samples was analyzed post wet weather event one, the data was reviewed. Constituents that were not detected were eliminated. The analytical data generated by the laboratory was compared to water quality standards to assess the conditions that existed at the site at the time of sampling. Results were correlated with discharge and precipitation data. The data was also used to identify in relation SMCL for corresponding constituents and identify constituent trends in both dry and wet weather conditions.
The original parameters evaluated in this analysis were selected based upon detected historical water quality data reviewed for this site with a total of 32 constituents (Table 4.2) tested at the Ponaganset River, only 15 (yellow highlighted Period was followed in order to eliminate accounting for leaf fall or snowmelt which influences the groundwater infiltration to the river. These constituents are a result of land use, geology, and nonpoint sources that are specific to the Ponaganset River in this region of the watershed.

Total Suspended Solids
Organic or inorganic particles that are carried by the runoff into receiving water are termed total suspended solids (TSS) .
As discharge increases during wet weather conditions, particulates rise to the surface of the water. The suspension and settling of these materials are a function of the physical characteristics of the river channel, base flow, and rainfall characteristics. The suspension and resuspension of trace metals could cause greater environmental impacts with regard to the potential toxicity caused by trace metals. "The EPA has established acute and chronic concentrations for trace metals using relationship based on hardness" . Under dry weather conditions, the river's baseflow is influenced by groundwater drainage particles that tend to settle to the bottom of the river as sediment. This sediment includes eroded soil and other organic suspended solids which may require an oxygen demand on the surface water. In addition, the transport of sediment will eventually deposit into reservoirs, which over time will add significantly to the dead storage and reduce its useful life.

Turbidity
Turbidity is a measure of the optical characteristic that causes the light to be scattered and absorbed, rather than transmitted, with no change in direction through the sample. The amount of turbidity in the water is caused by the suspended and colloidal matter such as clay, silt, finely divided organic and inorganic matter, plankton, and other microscopic organisms (Clesceri, Greenberg, and Eaton, 1998). Some of these particulates may mask the screening for pathogenic microorganisms.
Hazardous materials such as pesticides or heavy metals have the potential to be absorbed on suspended particulate matter. In water distribution systems the presence of turbidity may cause a decrease in the efficiency of the disinfection processes.

Color
The color of water is measured by PWSB using the method known as the platinum cobalt units or the tannin scale. The aesthetic property of water plays a role in a human's desire to drink, swim, bathe, or clean with it. The clearer the water, the more desirable it becomes to utilize it.
Factors that affect the color in rivers are dissolved organic material from decaying vegetation, some types of species of organic matter, and excess formations of algae. Additionally, the presence of iron and manganese also influences the color of the water. The presence of color may make the water appear objectionable and may require treatment.

4.4
Chemical Characteristics of Water

pH
The pH level is a measure of the acidity or alkalinity of a water sample. The symbol pH stands for potential for hydrogen and equals the negative log of the H+. The pH of water, on a scale of 0 to 14, is a measure of the free hydrogen ion concentration. Water contains both H+ ions and OH-ions. Pure distilled water contains equal number of H+ and OH-ions and is considered neutral (at pH 7), neither basic nor acidic. If water contains more H+ than OH-ions the water is considered acidic with a pH less than 7. If the water contains more OH-ions than H+ ions, the water is considered basic with a pH greater than 7 (U. S for the past ten years. The laboratory has indicated that at furthest points of the distribution system that pH is approximately 9.5 to 9.6. The pH of the water strongly influences the mobility of heavy metals in aqueous environments. Metal behavior in aquatic rivers is somewhat similar to that outside a water body. Sediment found on the streambed has the same characteristics that are associated in the normal soil environment. The results of this phenomena causes heavy metals to be sequestered at the bottom of the riverbed, while some become dissolved. The pH becomes the master variable in the whole process. In acidic like conditions, the H+ ions occupy most of the negatively charged surfaces of clay and organic material, although little room is left to bind metals which will remain in the soluble phase (Fairfax County Virginia, 2009). The aquatic organisms will be affected more due to extended contact with soluble metals.

Alkalinity
Alkalinity is an important measurement of the river's ability to neutralize acid rain, acid mine drainage, or wastewater. If the water in the river has a low alkalinity, it is prone to rapid changes in the pH level, although if it is high, it is able to resist major shifts in pH. Alkalinity not only helps regulate the pH of a water body, but also the metal content.
Bicarbonate and carbonate ions in water can remove toxic metals (such as lead, arsenic, and cadmium) by precipitating the metals out of solution.

(U.S. Environmental Protection Agency, 2006 Rivers and Streams Total
Alkalinity Status and Trends)

Acidity
The acidity levels of water are directly linked to "Acid Rain" or pollutant rainfall. Atmospheric water vapor reacts with carbon dioxide (C02) and sulfur dioxide (S0 2 ), to form weak acids, resulting in a pH ranging from 4.5 to 5. 6. In areas of high industrialization, the combustion of fossil fuels such as oil and coal emit sulfur dioxide (S0 2 ) and nitrogen oxides (NOx) in the atmosphere. Further transformation of these pollutants into gases such as sulfuric acid (H2S04) and nitric acid (HN03) cause further elevated levels of acidity in the water. From concentrations observed at the site values appear to be higher than the historical average for this region.

Chlorides
Chlorides are is not harmful to humans unless consumed in excess in the form of sodium chloride or table salt, which could cause high blood pressure from extensive use over a long period of time. The taste of sodium chloride may be apparent at levels of 250 mg/L and magnesium or calcium chloride at 1000 mg/L.
Chloride may be introduced into nver systems through rocks containing chlorides, agricultural runoff, and road salting during the winter months. Concentrations of chloride observed at the site were identified historically and were observed throughout this analysis during dry and wet weather conditions at the site and had the largest concentrations and loads of all the constituents.

Metal Characteristics
Trace metals that were detected at the Ponaganset River site included: barium, zinc, manganese, copper, aluminum, sodium, and iron. These constituents are considered inorganic and are primarily influenced by non-point sources that exist in this basin. The impact of these pollutants on the water quality may influence the ecosystem, and possibly render a body of water useless 1 g P eriods of time. The purpose of this section describes the types of for on detected metals that exist at the Ponaganset River site.

4.5.l Barium
Barium levels may arise from the erosion of augen granite-gneiss with alkali-feldspar porphyroclasts from which it is derived. Low levels of barium on the Ponaganset River are likely to be from the weathering of these types of rocks along the rivers path. Increased levels of this constituent in amounts greater than the MCL of 2 mg!L may cause an increase in blood pressure.

Zinc
Agricultural runoff that contains pesticides or herbicides may also contain lead and zinc. Contributions of zinc may result from urban runoff in the form of tire wear from roadway systems. Zinc is also widely used in the auto industry as a protective coating for iron and steel. Galvanized pipe is also used in water distribution systems.

Manganese
Manganese is a common compound that can be found all over the world. In water distribution systems, it is noticed as a black color. This may cause stains on washed clothing and give beverages a medicine like taste.

Copper
Copper is a common metal found in roadway runoff from bearing wear, moving engine parts, and brake dust. Sources of copper indicate the erosion of natural deposits in raw surface water systems and corrosion of 38 household plumbing systems contribute to elevated copper levels in effluent treated distribution systems. The MCL of copper is 1.3 mg/L.
Excess of levels of 1.3 mg/L have the potential to cause gastrointestinal disease from short-term exposure and liver or kidney disease from chronic exposure.

Aluminum
Traces of aluminum found on the Ponaganset River are likely due to the erosion of natural deposits along the rivers path. Other potential sources may include rusting of vehicle body frames being washed into the river from roadways.

Sodium
Many people associate salty water with oceans or salt lakes although, all water includes some salt. High concentrations of sodium in river systems could be due to road salting during the spring snowmelt and areas where crop irrigation is used. Crop irrigation often picks up salt as it passes through the soil and returns back to the river. Excess sodium concentrations in rivers have the potential to affect the crop's soil if river water is used for irrigation. Sodium levels usually tend to increase during the winter and early spring months during wet weather condition due to roadway runoff from salting practices. Additional studies such as Runge and Wright, 1989 as well as Nimiroski and Waldron, 2002 have investigated sodium loads that contribute to the Scituate Reservoir based upon the number of roadway systems, surface area, and estimated amount of salt used on state and local roadways.

Iron
Possible contributions of iron entering into the river system occur from rusting automobile body frame or any type of rusting metal. In addition, iron can be released in small quantities by rock weathering. Iron and other trace metals have a low solubility so they bond to clay particles, which exist in large quantities in the soil. Small amounts of iron are necessary for plant, animal, and human health. Iron also releases a brownish color to the water which makes it unpleasant to look at, bathe in, or drink.

Total Coliform Bacteria
Coliform consists of two groups: total coliform and fecal coliform.

40
The EPA has four approved analytical methods for testing coliforms which are listed as follows:

Overview of Water Sample Collection Methodology
The Ponaganset River site was inspected to determine where the water samples should be collected prior to the first dry weather water sample collection.
The objective was to identify and mark the exact cross section where the samples will be collected. From field observations it was determined that water samples have to be collected at the center of the bridge which allows for thorough mixing of the samples. Rainfall and discharge data was obtained using real-time fifteenminute interval recording equipment at Ponaganset River site before, during, and after this investigation.

42
The objective during sample collection was to provide the most accurate representation of the water at the Ponaganset River site. Prior to gathering the analysis sample, all sampling equipment was rinsed with river water three times to assure that the river water collected was representative. The sample was obtained at the downstream side of the wooden bridge at Rams Tail Road ( Figure   2.7). The collected water was distributed into five sample bottles provided by the laboratory shown in Table 5.1:

s. 4 Laboratory Analysis
Premier Laboratory, located m Dayville, Connecticut conducted water sample analyses for this project. The laboratory was commissioned to pickup biweekly water samples during the collection of dry weather samples and multiple sample pickups during collection of wet weather events. The coordination efforts with the laboratory and the investigator were crucial during wet weather collection because these samples required various holding times.
The samples were transported to the laboratory in a chilled cooler filled with ice and or ice packs. Each sample bottle was labeled with a permanent marker and a numbering system for the bottles was determined prior to the initial sampling. For simplicity, labels were preprinted. Each bottle was labeled with the sample number, date, and time just prior to collecting the sample. In addition, a spreadsheet was prepared with the bottle number I.D. that corresponded to the sample bottle. The laboratory also requires a chain of custody form. The data received after testing was reviewed for consistency with historic sample results.
The testing methods used by Premier Laboratory for constituents detected at the site are identified in Table 5

6
Dry Weather Results The first set of dry weather results (DWR) received by the laboratory revealed that out of the 32 analyzed, only 16 were detectable. Samples shown in Table 5.3 identify the 16 parameters that were detected at during dry weather conditions at the Ponaganset River site. The photo shown in Figure   from Premier Laboratory were reviewed approximately one to two weeks after DW 1 was collected. Prior to collection and analysis of the second set of dry weather samples, it was decided to duplicate testing for the initial 32 samples.
This was done to verify that all the constituents measured would reappear as detectable or non-detectable and not just a coincidence. The second dry weather collection revealed 3 mg/L of total suspended solids (TSS) for which the first dry weather sample set did not detect. TSS was detected a total of 8 out of 12 samples and concentrations ranged from 1 to 6 mg/L and averaged 3 .6 mg/L.
Aluminum was detected on the DW 1, 3, 6, 7, 9, 10, 11, and 12 (or 8 out of 12 samples). Concentrations of aluminum ranged from 0.044 to 0.120 mg/L and averaged 0.090 mg/L. for the 8 dry weather samples (Table 5.3). The average acidity observed was 6.43 mg/L during dry weather for only 10 out 12 samples tested. Alkalinity was detected on DW 1, 4, 5, 9, 11, and 12 (or 6 out of 12 samples) Total copper was initially detected on DW 1 and 2, and then was nondetectable for the remaining 10 samples. Total copper was detected at discharge rates greater than 25.5 cfs during dry weather conditions. Total coliform bacteria was detected in 11 out of 12 samples. Table 5.3 identifies the dry weather data results of the constituents that were observed at the site, and Figure    The one primary goal associated with this project was to determine the wet weather characteristics at the Ponaganset River site. This involved accurate forecasting of weather conditions that would be suitable for field sampling. The criteria for wet weather sampling required two days of antecedent dry period, a minimum of 0.1 inches of total precipitation, and collection of enough samples to adequately cover the hydrograph. Weather forecasting sites such as local news web sites, weather underground, national weather service, and other sources were used to estimate the expected rainfall events. In addition, coordination of sample pickup from Premier Laboratory required that sample pick-ups be done on Monday through Friday during regular work hours to avoid excess project expenses.
The frequency of sampling under wet weather conditions was determined on site, based upon the projected forecast and the predicted response the river.
From previous research, discharge data indicated that in almost all cases the river returned to baseflow within forty eight hours.
If the projected storm extended for a period of time, perhaps one to two days, the samples would be taken at intervals of four hours dependent on the intensity of the storm. Storms of short duration, such as summer storms were considerably more difficult because tracking is difficult to due to uncertain movements, intensity, and short duration. Various attempts were made to sample during thunderstorms, although none were captured for use in this analysis.
During wet-weather, samples were collected at a frequency of every one to four hours, and post rainfall at a frequency of approximately eight to twelve hours apart for the last two sets of samples. This sample frequency was utilized for the Ponaganset River site based upon the small sub watershed drainage area characteristics.

Distinct Rainfall Characteristics
Rainfall is the driving force for the hydrologic cycle which controls our water supplies. By understanding the nature and characteristics of rainfall, determinations can be made on its effect in relation to runoff, infiltration, evapotranspiration, and annual yields. In this analysis the Ponaganset River site was utilized for its recording rain gauge which provides a record of accumulation as a function of time. Hence, allowing for the total precipitation, intensities, and duration to be determined for each of the three individual storm events described in this analysis, (Table 6.1) allowing for the determination of empirical equations 59 employed for the analysis. These rainfall characteristics are briefly described below: • Total Rainfall One of the most obvious characteristic of rainfall is the total amount of precipitation that falls during the course of the storm. This is an easy measure of determining the severity of the storm being described.
The total rainfall amount often raises or lowers the amount of sediment transported in the river. During this analysis, three storms were captured with total precipitation amounts described in In addition to the characteristics described above, the rainfall distribution Thunderstorms often change direction quickly, thus influencing the rainfalls distribution within the watershed and also affecting the runoff. In Table 6.1 rainfall characteristics for storms 1,2, and 3 indicate 71.45%, 17.19%, and 18.44 % of the rain that fell in this sub-basin entered the river in the form of runoff.
Past studies commissioned by PWSB have indicated that approximately 50% of the rain that falls within the limits of the entire 92.8 square mile watershed is runoff and is stored in the five tributaries, which all flow into the Scituate Ol ·r In the case of this analysis, the Ponaganset river watershed is the most Reserv · forested region which also largely influences the amount of runoff entering the . er Prior to the start of runoff the amount of rainfall required must satisfy nv .
infiltration demands, evaporation, interception, and surface storage. For each of the three storms observed at this site (storms 1, 2,and 3) required 0.39, 0.47, and 0.37 inch of total precipitation or an average of 0.41 inch to satisfy these requirements before runoff can occur. Once these requirements are satisfied, the runoff contributions can be used to determine factors that affect the transport of constituents that are distinct to the point measured on the river.
Rainfall characteristic phenomena can be used in an analysis of its own.
The objective in this section is to briefly describe factors that are influenced by these rainfall traits. On previous studies conducted on the entire Scituate Reservoir Watershed attempts were made to determine if the arithmetic mean of the five rain gauges measured by PWSB is as accurate as the Thessian approach.
The result of that study showed that both were equally as accurate when compared to long term historic results. In this analysis real-time rainfall, discharge, and water quality records were used to determine the pollutant loads for constituents identified consistently at the site for wet and dry weather conditions.

·3 Storm 1 Characteristics
The   The results of testing for WW 1 indicate that out of the ten sets of samples collected throughout the storm, eleven constituents were detected consistently which are listed as follows: barium, zinc, manganese, aluminum, sodium, iron, turbidity, color, chloride, pH, and total coliform bacteria.
In addition, four of the constituents measured had partial records of concentrations found during WW 1 are listed as follows: copper, acidity, alkalinity, and total suspended solids.
All of the other sixteen constituents measured during WW 1 were nondetectable, as they were during DW 1. Ammonia was not tested during wet 63 weather due to very low trace amounts found during dry weather samples 1 through 7. The intent of this analysis was geared to detected constituents that are particular to the Ponaganset River at site 01115187 during dry and wet weather conditions.
The results from the samples that were detected during WW 1 one showed various patterns of concentrations in relation to runoff from the river. Barium, sodium, chloride, and pH, levels appear to decrease during higher flows due to dilution. The results found that these measured samples were less than base flow conditions that were collected during dry weather. Concentrations of zinc illustrate a somewhat erratic response, although for the most part it tends to increase during higher flows. In addition, a number of detected constituents measured such as manganese, aluminum, iron, acidity, turbidity, and total coliform bacteria increased in concentration as the runoff increased on the Ponaganset River. Results of alkalinity and total suspended solids were inconclusive due to partial detection of the ten samples measured. Color did not change from 30 color units throughout the entire period of sampling.

Elimination of Non-Detected Samples
After the data from the first storm event was collected, the analysis results were reviewed for accuracy and consistency. The resulting data revealed out of 32 of the proposed samples analyzed only 15 were detected. If the proposed constituents did not appear in both dry and wet weather conditions then the sample was discarded for testing for the second storm event. There was a concern that some constituents could exist during other various size storm events.
Parameters that were questioned included copper, alkalinity, and total suspended solids, which had partial records of concentrations found during WW 1. All of the other 12 detected parameters measured during the first storm event were detected in all 10 samples aside from one missing concentration out of ten for acidity.
Prior to the start of collection of samples for the second wet weather event a decision was made to test the fifteen parameters that indicated ten out of ten results that were above the specified detection limits. These parameters will be the basis for subsequent field collection of wet and dry water quality samples tested at the site. under small-scale rainfall conditions. The resulting data for TSS levels ranged from 1 to 11 mg/L but are not in direct correlation to discharge. Constituents such as barium, sodium, and chloride concentrations appeared relatively constant. The pH levels decreased from 6.10 to 4.70 with a difference of 1.4 (Table 6.3). Total coliform bacteria results were obtained for only four out of nine samples because testing methods used by the laboratory did not dilute the sample enough to determine bacteria levels in excess of 1000 colonies. Adjustments were made to correct the dilution methodology for the subsequent storm events.  Results showed similar characteristics patterns of the first two storm events, which indicated zinc, copper, and alkalinity as non-detectable parameters.
Concentrations of barium, manganese, aluminum, and iron decreased as expected in relation to the total rainfall and runoff amount being the least of the three storms due to the low amount of total precipitation, interception, and depression storage. In addition, characteristics such as turbidity, color, acidity, and total suspended solids also decreased in comparison to the first two storm events. The concentrations of pH and sodium levels increased unexpectedly in storm three which are closer to representative levels under normal dry weather conditions.
Lastly, concentrations of copper stayed fairly close to storm event two. During storm three, the rivers response in relation to the samples collected is illustrated in   The data collected for these six constituents was used to determine the total wet load (lbs.) for each of the three storms and twelve dry weather loads (lhs./day) for this site. Real-time monitoring of discharge and precipitation data was used to correlate to the specific time the sample was collected. The total wet load for each of the six parameters will be predicted with the use of Multiple Linear Regression (MLR) models designed for this site. The dependent variables used for these equations include the total precipitation of the storm and the rivers maximum flow rate ( cfs) minus the baseflow using the concave method for predicting the total wet load. The loads used for the wet weather analysis will be converted from "lbs./day'' to "lbs." by dividing the number of hours between sampling by 24. This number will be multiplied by the difference of the second through last load minus the first load in "lbs./day". After multiplying these numbers for each increment measured during the storm these values will be summed to obtain the total wet load (lbs.) for each of the three storms collected for this analysis. The twelve dry weather samples will be used to generate Linear Regression (LR) models using the baseflow discharge at the time of sampling. These equations will be described in more detail in the following chapter where the analysis is more in depth with regard to the statistical analysis for this site.

Environmental Influences of Water Quality
When rain storms occur in the watershed, biological agents, rock weathering, and soil nutrients dissolve into the river water. Contributions from the pure water also carry very small amounts of acidic chemical substances such as carbonic acid (H2C03). This type of reaction takes place when carbon dioxide (COz), from the earth's atmosphere reacts with the water (H20), to form carbonic acid (H 2 C0 3 ). The water in the river also contains small diluted amounts of hydrochloric and sulfuric acids  Station, 1981 ). The natural breakdown of rock partly influences the water quality in the river in the form of detected trace amounts of metals such as barium, manganese, aluminum, iron, sodium, and chloride.
Soil types were also reviewed for the Ponaganset River site for the various soils within proximity of 1000 ft radius surrounding the point the samples were collected. These soil types for this area are listed in Table 7 .1 and illustrated in

Loading Procedure
The differences between concentration and load are very different.  The dry weather loads were calculated using W = Ku * C * Q for each of the six trace metals selected for analysis. Five of these metals were identified twelve out of twelve times for barium, manganese, sodium, iron, & chloride.

Dry Weather Loads
Aluminum was detected eight out of twelve times above the detection limit 0.05 rng!L although the last dry weather sample indicated 0.04 mg/L by Premier Laboratory. In Table 7 .2 barium showed the smallest load concentration while chloride showed the highest. Characteristic relationships were plotted in Figure   8.1 by plotting the load in (lbs. I day) versus the discharge (cfs  Other studies, such as phase 2 of the Blackstone River Initiative have conducted more extensive analyses to identify sources of pollutants using data collected for three wet weather events. A point indicated in this previous study identified that the combination of the increase and decrease of individual constituents cause a more significant environmental impact. "The EPA has established acute and chronic concentrations for trace metals using relationships based on hardness. When hardness decreases, the potential toxicity increases."   Concentrations of sodium and chloride will tend to be larger during the winters that have significant snowfall due to road salting along Rt. 6 I Danielson Pike, although large flows during the spring and fall will produce larger loads from the river.    The results show a consistent pattern of load based upon the total precipitation. As the total precipitation increases in direct correlation with the rivers discharge, the total pollutant load increases and vice versa. If the storm event is large enough to produce a wet weather event in excess of one inch or greater of total precipitation, the pollutant load will increase significantly. The other five constituent results show a more significant difference between storms 1 and 2 with regard to the individual storm characteristics. Each of the three storms collected for this analysis, resulted in the loads generally corresponding to the amount of total rainfall (1. 38, 0.57, and 0.45 in.), and the discharge response of the river at the time the sample was collected.
In the next chapter, the total loads for each of the three storm events will be used to predict the estimated load contributions through the use of developed linear and multiple linear regression models specific to the Ponaganset River site.
These statistical models will identify significant advantages over periodic manual sampling practices. The predicted chemical load transported at the Ponaganset River site will be used to determine the amount of pounds per year during both wet and dry weather conditions. During periods of high flows, the river has a substantial effect on the transport of these chemicals. Therefore, estimated loads are far more accurate using real-time water quality monitoring sites such as the Ponaganset River. The data described in Tables 7.3, 7.4, and 7.5 identifies the load (lbs.) during the time of sampling throughout each of the three captured stonn events. The summation of total load per storm event is summarized in Table 7.6 for each of the six constituents. These observed loads will be used in comparison to predicted loads for which statistical analysis will identify if there is 86 a correlation. This data can later be used to provide estimated event and annual based load estimates for the Ponaganset River site.

Regression Analysis Overview
The concept behind Multiple Linear Regression (MLR) analysis relates to one dependent variable that is conditioned by more than one independent variable.
In River site's real time monitoring capabilities provides more of an accurate determination of discharge and precipitation at a precise time than conventional water quality sites. Using data collected from the river, in conjunction with water quality results collected at a precise time, allowed this analysis to determine predicted estimates of load contributions for selected constituents.
The loads described in chapter 7 were used in conjunction with real time data to generate predictive linear and multiple linear regression equations. These predicted constituent loads can be used to determine the event and annual based pollutant loads during dry and wet weather conditions. Information generated from this analysis can be used to indicate the sub-basin load contributions as it relates to land-resource management practices within the Scituate Reservoir Complex. Finally, the use of these predictive equations can be used to estimate peak loads during extreme conditions. If changes do occur these equations could provide the water supplier with sufficient time to respond to any negative effects on the environment or allow them to make adjustments to the water treatment processes.

Linear Regression Analysis
Linear Regression models were developed to predict the estimated load contributions at the Ponaganset River site (01115187) during dry weather conditions. This was done by graphing the load (lbs./day) versus discharge (cfs) for the 12 samples collected during dry weather conditions. The Linear Regression equations were developed based upon data collected at the site for the 12 dry weather samples. These equations are valid for use as an estimate of the load during dry weather if the antecedent dry period is equivalent to two or more days during the spring and summer months at the precise location where the sample was collected.
The establishment of these Linear Regression models was then taken one step further by merging the historic PWSB water quality data with the twelve current analysis samples. To do this, data collected from PWSB was first . .

50
. . . .    . .    The These MLR equations use the total precipitation and peak discharge as the independent variables used to predict the load (lbs.) for a specific wet weather event. These independent variables were specifically chosen to easily obtain data collected from storms utilizing the real-time monitoring equipment at the site.
The equations described in equations 7 thru 12 were used to determine the predicted loads for each of six constituents during each of the three storms which were compared to the actual load summarized in Table 8. 3. In this comparison the actual versus predicted data are dependent because the actual values were used to derive the MLR models. These predicted MLR equations can be tested in the future by conducting more wet weather monitoring to identify if the predicted loady (lbs.) falls within an acceptable range of the actual load. The actual versus predicted loads (lbs.) shown in Table 8.3 were fairly close for the first wet weather event while storms two and three indicate more significant differences. The predicted loads during storm three indicate all positive loads except for aluminum and iron which indicate negative load amounts. These negative predicted loads obviously do not decrease below zero pounds of aluminum and iron. In these cases, the equations developed for this analysis cannot predict loads for storms characteristic of the third storm collected during wet weather monitoring. The data described in Table 8.3 identify that the predicted equations appear to have some significant error associated with storm 3.
This error could be due to the statistical fit of the data, or the equation has limitations to the size of the storm it can predict. From the comparison of this data it would appear that these MLR equations can be used for storms with a total precipitation of greater than 0.5'' of rain and an antecedent dry period of at least two or more days. In the following chapter these predicted MLR equations shown in Equations 7 thru 12 will be used to predict monthly and partial annual load estimate for one year worth of data supplied by the USGS.
In the case of MLR models generated for sodium and iron during wet weather, it was decided to eliminate iron for predicting monthly and annual load estimates due to variations of the actual (8.80 lbs.) versus predicted load (-1.26 lbs.) shown in  Runge and Wright (1990) comparing road density and median stream-sodium concentrations for the period of 1983 through 1989 and 1990 through 2000.
In reviewing the salting mixture CaCh and NaCl indicated by the weighted ratio 60:40 using atomic weight for the compounds, the breakdown is For CaCh there is an association with water or 6H 2 0 that must be included to account for the moisture content. For the purposes of load estimation for the analysis 40% of Na was used based upon the chloride load for this specific constituent. In the future, a more precise empirical model could be generated for sodium with the addition of data for one or two more collected storm events for this site for storms that range between 0.45 in and 1.38 in. of total precipitation and greater than 1.38 in ..

Applicability of Statistical Models
The linear and multiple linear regression models were generated for barium, manganese, aluminum, iron, sodium, and chloride. These water quality parameters were selected based upon historically detectable trace elements observed in the sub-basin. The statistical models were based upon data that was collected during months of April through September. The reason for sampling during these months was to capture the period of highest concentrations due to the vegetation, soil drainage, and temperature for the majority of constituents which typically occur in the spring and summer months. The statistical models are only applicable to this period. Loads estimated using the equations will diminish if they are used outside the parameters of the analysis. Future collection of data for the fall and winter may require the additional development of seasonal equations.
Therefore, water quality monitoring, particularly during wet weather should continue on the watershed. Since PWSB performs routine monitoring during dry weather, that data may be used as a base line for steady state flow in the rivers.
The purpose of these statistical models is to develop a tool to determine long term load estimates. These long term estimates will indicate mass load for constituents having the largest impact to Barden Reservoir and its ultimate contribution to the Scituate Reservoir. Many of these constituents are dissolved and pass through the filtration process which then ultimately enters the distribution system.
During the fall period, leaf litter changes the runoff. Leaves begin to fall from October through December. At this time, the surface runoff will decrease as a result of leaf coverage. In addition to the coverage of the leaves, the vegetation begins to die which, reduces the interception, and frost begins which also reduces 106 the soil drainage. If the models are applied to the fall load, estimates will be larger than what the actual load will be. In the winter, the direct runoff increases even more. This is due to the lack of interception, depression storage, and accumulation of snow pack. In addition, in the winter the precipitation occurs in the form of snow which covers the surface area. The snow pack continues until the temperature increases in the early spring then precipitation suddenly begins to wash the snow pack into the river as well as the precipitation that falls. Also in the winter the river level is primarily affected by the ground water discharge from below the frost line much like dry weather but lacking nutrients that are present during spring and summer months. Sodium and chloride will begin to increase after the first snowfall and particularly when temperatures start to increase and wet weather occurs which will wash a significant amount of road salt into the river.

Annual Parameter Objectives
The

Annual Data Analysis
The final objective of this analysis entailed determining a partial estimated data was used in correlation with linear and multiple linear regression equations to predict the estimated annual loads for selected constituents measured at this site.
After the initial data was sorted, filtered, and reviewed for consistency and accuracy individual storm events were extracted from the data for a years worth of record. A total of 42 initial hydrographs were constructed and evaluated to detennine the area under the hydrograph by separating the base flow and calculating the percentage of effective runoff contributions. Later, the hydrographs were reevaluated and a majority of them were eliminated due to seasonal criteria, antecedent dry periods less than two days, irregular precipitation patterns associated with summer thunderstorms, or storms less than or larger than those collected for the purposes of this analysis. Watershed which was intended to lower the sodium concentration in the water supply (Rhode Island Department of Administration, 1988) and .
Review of the annual water quality records collected from PWSB from Nov. 1995to Nov. 2008 (13 years of data) indicate an increase in chloride concentrations illustrated in Figure 9 .1, which therefore increases the sodium in the water. The power trend line shown in Figure 9.1 for chloride indicates an accuracy of R 2 = 55. 1%. Results from previous analyses done by (Runge) and (Wright) identify a significant increase in sodium and chloride trends in 1986 and 1987 at 22 sub-basins throughout the entire watershed. Results indicated that both roadway and residential density were high in the eastern portion of the watershed with Moswansicut sub-basin having the highest (1.74% and 20.33%, respectively). Kent sub-basin had the lowest roadway and residential densities (0.15% and 0.82%, respectively) . USGS has previously predicted chloride concentrations using their predicted MLR equations for the Little Arkansas River in terms of specific conductance and discharge using logarithmic transformation . Earlier analysis attempted to use these predictive parameters for use in this analysis for estimating loads during wet weather at the Ponaganset River, although a strong correlation could not be made this way. Iron also could not be accurately determined during wet weather conditions due to results received for wet weather event number two.
These results skewed the results of the predicted MLR equations used to predict loads during wet weather.   (Breault, Waldron, Barlow, and Dickerman 2000), all of which can cause disease in humans.

111
Historical PWSB data (Appendix E) for total coliform bacteria was reviewed for this site for the period from November 1995 through November 113 2 008. The historic PWSB data cannot be compared to the analysis data due to Loads and compansons were made with total coliform bacteria to determine the basin characteristics, although data for this parameter was not complete. Of the twelve dry weather samples tested, eleven produced positive results. For the wet weather sampling, only storms one and three had sufficient data to determine wet weather loads. With only two data points, an equation could not be established for wet weather conditions. A partial data set of the dry weather loads was compared to the actual loads for WW 1 and WW3. The loads observed for WW 1 indicated 6.1 x 10 9 cfu (colony forming units) and for WW 3 was determined at 2.1 x 10 8 cfu for the entire storm. These actual loads were divided by the duration of the storms in days compared to eleven dry weather loads measured in billion colonies per day. The comparisons are described in Figure 9.2 and indicate that total coliform bacteria is wet weather driven in this sub-basin. Potential sources of bacteria could be runoff from roadways and other impervious areas, and leaching of aged septic systems.  The Multiple Linear Regression (MLR) equations were solved using the least squares methodology which could only be partially predicted for storms evaluated during the course of this investigation. The project criteria indicated that water quality samples would be collected during the period of April through September, with at least a two day antecedent dry period, and at least 0.10 in. of total precipitation. After inserting the predicted MLR equations for all storms it was found that these equations could be applied to only a small portion of storms due to the number of storms collected and the project criteria for this analysis.
The dry weather Linear Regression equations used to estimate the load for the entire year were summed for each constituent during base flow condition for the entire fiscal year is described in Appendix A.

9.5
Summary of Monthly Dry and Wet Weather Loads The empirical equations described in chapter eight were applied to daily flow and precipitation data for a six month period occurring from April through September 2004. The discharge that occurred during this period is shown in Figure 9.2 and 9.3. Monthly load estimates shown in Table 9.2 identify that during periods of low precipitation amounts at the Ponaganset River sub-basin are primarily dry weather driven for May through July aside from manganese and aluminum in July 2004 which indicated loads being more wet weather driven during those months. In Table 9.2 the predicted load estimates for April, August, and September 2004 suggest that it is wet weather driven with the exception of  in. of total rainfall that occurred during that month. The "Blackstone River Wet Weather Initiative" water quality program was written by Makam in 1989-1990, and Table 9.4 based on these storms: For dry weather conditions annual loads were determined for the entire year. To do this, hydrograph data had to be separated from the base flow, while the average daily discharge was used during dry days. The results of the annual dry weather loads estimated for this analysis are identified in Table 9.5 below from smallest to largest load in lbs./yr.:

Summary of Results
The results of the annual analysis data indicate estimated loads in (lbs. I day) during dry weather conditions and (lbs.) for partial wet weather contributions over a period of one year's worth data from October 1, 2003 to September 30, 2004. The MLR equations applied to the wet weather conditions at this site were found to be the most difficult and complex portion of this analysis. The MLR models used for this period of data indicated negative loads, although it is obvious that they cannot decrease below zero. Typically, loads increase accordingly until reaching a maximum load then either level out or begin to dilute with increasing discharge rates. The summary of load characteristics associated with these six constituents for this period of record are described in Figure 9.4 for both dry and wet weather conditions. The predicted dry weather loads are based upon 189 days of record using average daily discharge (cfs). The ranges of results predicted during dry weather for this 189 day period are summarized in Table 9.3 for the six constituents reviewed for the dry weather analysis. In addition, graphs indicated in Figure 9.2 may be used to determine expected loads, based upon the average daily discharge measurement at the site used in conjunction with the predicted dry weather linear regression models. The wet weather load characteristics are based on seven storms that occurred between April through September 2004. These seven storms were the only ones that could be compared to the analysis storms collected in 2006 for the analysis. In addition, MLR equations for sodium and iron could not be used due to irregularities in the analysis of Storm 2 data which appeared to only affect these two constituents. The predicted loads for sodium was estimated in terms of chloride or 40% of what was predicted for chloride based upon the fact that NaCl and CaCh enter the watershed from road salting during the winter months. The characteristics associated with the independent variables used for the wet weather MLR equations (Total precipitation (in.) and QMAX. -Base Flow) were compared for the three analysis storms and the seven storms taken from the annual parameter' data. In these comparisons shown in Figure 9.5 both indicate a high R 2 correlation (92 % (2004) and 99% (2006) A total of thirty two parameters were initially tested, although only fifteen were observed at the site. Metals that were consistently observed include: barium, manganese, aluminum, iron, and sodium. Other parameters include: copper, acidity, alkalinity, turbidity, color, pH, total suspended solids, ammonia, chloride, and total coliform bacteria. Total coliform bacteria was also considered for this evaluation but the data for this parameter was insufficient to correlate a predictive load equation.
The results of this investigation suggested six predicted dry weather linear regression models that were generated from twelve sample events collected at the site. The linear regression models use instantaneous loads to develop a 130 relationship using the river's discharge to predict the load in pounds per day. Wet weather loads were predicted using six multiple linear regression models using maximum discharge minus the baseflow and total precipitation as the independent variables. Other independent variables attempted in the MLR models (i.e., specific conductance, intensity) were not successful. The precipitation characteristics of the three storms that were evaluated ranged from 0.45 in. to 1.38 in. (Table 6.1 ). Storms that are larger or smaller can not be accurately predicted without further collections of storm data. Load estimates generated using developed MLR equations proved to predict loads for barium, manganese, aluminum, and chloride. Sodium and iron could not be accurately modeled due to skewed results from storm two. Predicted MLR equations applied to the annual parameter data indicated negative loads for storms less than 0.45 in. of total precipitation and extremely large predicted values for storms greater than 1.38 in.
of total precipitation. Sodium load estimates were determined based upon forty percent of the predicted chloride load. The percentage of sodium was derived based upon the molar concentration in road salt used on roadways throughout the Scituate Reservoir watershed.
Other factors which influence the water quality conditions at this site are bedrock geology, land use, land slope, road salting, and leaching from septic systems. The bedrock geology consists of augen granite-gneiss with alkalifeldspar porphyroclasts from which barium, manganese, and iron can be derived.
The iron detected at the site may be present from natural deposits, sanding of roadways, moving engine parts, and auto body rust. The USGS has indicated that "the drainage basin characteristics explain at least 50 percent of the variability in 131 concentrations of water quality constituents" (Breault, Waldron, Barlow, and Dickerman, 2000). Barium levels are relatively consistent with little to no variation under wet or dry weather conditions. Concentrations ranged from 0.013 to 0.021 mg/L during dry weather conditions. The largest loads observed at the site indicated 3.48 lbs./day during dry weather and 9.2 lbs. for wet weather.
Barium is considered a minor pollutant although the USEP A considers this parameter important and requires testing for drinking water and providing this data in the Consumer Confidence Report (CCR) which is distributed to PWSB wet weather. This is due to dilution of these constituents because sodium and chloride dissolve in the river. Aside from the fact that concentrations decreased during wet weather, loads increased due to large river flow rates. Throughout this entire analysis the concentrations sodium and chloride were less than the USEP A SMCL.
In the future, areas of denser land use within the watershed may require additional sampling and more frequent testing. It is recommended that PWSB should consider the installation of real-time monitoring equipment at a few important site locations that currently have only a staff gauge. In addition, realtime precipitation gauges would be ideal to replace the five existing rain gauges situated throughout the entire watershed and install one new rain gauge at the Providence Water Treatment Plant in Scituate Rhode Island.

I
The following recommendations are offered to establish long term patterns of instantaneous loads for PWSB: • Record staff gauge height or discharge at the time the sample is collected.
• Determine precise date and time samples are collected.
• Record antecedent dry period.
• Specific constituents such as sodium, iron, manganese, and aluminum should be collected and tested on a monthly instead of quarterly basis in addition to currently tested water quality parameters.
• Analysis of barium and other minor pollutants could be limited in extent.
• Consider testing for both dissolved and total metal concentrations.
Since 1945 PWSB has been testing for pH, color, turbidity, total coliform bacteria, acidity, alkalinity, iron, and manganese on the watershed. In 2009, PWSB continues to collect and test samples on a monthly and or quarterly basis at 35 sites throughout the Scituate Reservoir watershed. In the future, water quality will be of greater concern with more stringent regulations. As additional commercial buildings are constructed and vehicle transport increases throughout the watershed, water quality monitoring should continue to increase as well. This analysis identified how loads estimation can be used to develop long term characteristics. These water quality characteristics can be used to determine the contribution of dead storage to tributaries such as Barden Reservoir in the case of this analysis or be used to develop a full scale wet weather analyses on the entire Scituate Reservoir Watershed.

Procedure for Future Development of Annual Loads
The following suggestions are being offered from the results of this analysis to be able to determine annual load contributions associated with the Ponaganset River site. • Develop or use hydro graph data for a minimum of one year.
• Separate dry days and individual storm events for that year.
• Using the independent variables QA VE. for dry weather and QMAX.-BF and PT for wet weather, determine loads using developed linear and multiple linear regression models for constituents identified in the analysis.
This will provide an annual percentage of loads contributed for both dry and weather deposition for a years worth of data for this site.

Future Wet Weather Monitoring on the Scituate Reservoir Watershed
In the future, wet weather monitoring on the Scituate Reservoir Watershed should consist of the following: