1. Introduction
Eutrophication poses a severe threat to coastal waters on local, regional, and global scales. Coastal ecosystems are naturally nitrogen limited and runoff from terrestrial systems from agriculture, wastewater, and industrial practices contributes to eutrophic areas of low dissolved oxygen, known as dead zones [
1]. Dead zones have been documented as doubling in occurrence every decade since the mid 1900s and have been increasing in areal extent. Factors such as rising ocean temperatures, increasing ocean acidification, sea-level rise, and changing climate variables are acting synergistically to exacerbate the eutrophication problem [
2]. Often, eutrophication, leading to conditions of hypoxia and anoxia in coastal waters, is overlooked until broad scale detrimental effects are obvious [
3]. The breakdown of coastal fisheries and large fish kills garner the most attention to eutrophic conditions. Fisheries located in shallow coastal waters are the most at risk and instances of fishery and shellfish closures have increased in recent years.
Narragansett Bay, located in the state of Rhode Island (RI), USA is no exception to the problem of excess nitrogen (N) and resulting eutrophication of coastal waters. Areas of the bay have experienced N inputs for over the past 200 years and impacted areas have expanded in extent as the state’s population has grown and development has increased throughout the state. Rhode Island is the smallest state and is the second most densely populated with roughly one million residents [
4]. Over the past several decades, the bay has ranked as one of the most heavily fertilized estuaries in the USA with the majority of nutrients originating from point sources [
5,
6]. The largest point source contributor is sewage treatment plants (STPs) [
6] in the northern portion of the bay where the state capitol is located and the population density is highest. Through the years, the state has faced harmful algal blooms stemming from eutrophication, leading to fisheries and beach closures, most often during the peak summer tourism season. Both fisheries and tourism are vital to the state’s economy. In 2003, the state experienced a severe fish kill with over one million menhaden, and other finfish and shellfish, killed from anoxic conditions within Greenwich Bay, which is a smaller offshoot of Narragansett Bay [
7]. This event spurred initiative for action to clean up pollution within the bay and to combat excess N from entering the state’s coastal waters in order to reduce risks to the health of the coastal ecosystems and the communities that rely upon them. Targeting the N point sources, the state set a goal of 50% reduction of N discharge into the bay from 1995 to 1996 levels, limiting the allowable discharge volumes of N from 11 wastewater treatment facilities in upper Narragansett Bay. The targeted reduction was achieved through stages by 2012 [
8]. Today, as N inputs decline from point sources, nonpoint sources, such as urban runoff from septic systems, are becoming a proportionately larger contributor of N into the bay. In RI, 30% of homes are on septic systems, also known as onsite wastewater treatment systems (OWTSs) [
8,
9]. For coastal watersheds with households served predominantly by OWTSs, these systems can serve as significant sources of N which can contribute to detrimental eutrophication of coastal waters.
The Hunt River and two of its tributaries (Fry Brook and Scrabbletown Brook), lie within coastal watersheds served predominantly by OWTSs. These waterbodies were listed as 303 (d) Group 1, under the Clean Water Act for highest priority impacted waters in RI. The Hunt River and its tributaries have made the 1998 303 (d), under the Clean Water Act list of impaired waterbodies for several years for being impacted by fecal coliform. The majority of high bacteria counts were found to occur during wet weather conditions. The 2001 total maximum daily loads (TMDL) report determined the largest wet weather source of fecal coliform to be stormwater runoff. Fecal coliform monitoring is often used as indication of pathogen presence; however, modeling fecal coliform in freshwater systems holds a unique set of challenges. Bacteria fate and transport relies heavily on the nature of pollution events and sensitivity of fecal coliform to environmental conditions [
10]. Leaking sewer and septic infrastructure greatly contributes to nonpoint nitrogen pollution to urban watersheds
In order to protect the health of any coastal community, it is important to quantify the influence of STPs and OWTSs on groundwater flow and the contributions of nutrients from these systems to coastal waters. For mitigation purposes, it is helpful to model these inputs on local watershed scales to identify potential areas of high N inputs to target for reduction through policies and infrastructure development. Few studies have been conducted modeling nutrient inputs on the watershed scale using the Soil Water Assessment Tool (SWAT) due to complexities stemming from the uncertainty of differences in subsurface hydrology, soil and water dynamics, and the often lack of detailed information about OWTS [
11]. Increased densities of OWTS have been shown to increase baseflow within a watershed system [
12,
13,
14], which can have the effect of increasing nutrient loading into ground and surface waters, eventually entering into coastal waters. OWTSs have a biozone layered designed to filter out nutrients, preventing their leaching into the groundwater. However, these biozone layers can become ineffective over time, resulting in an increased discharge of nutrients into sensitive coastal waters. Nutrient loading into a waterbody depends on local surrounding conditions, thus it is helpful to conduct modeling on smaller scales to identify areas to target for N mitigation.
In this study, we modeled the amount of N entering into Narragansett Bay on the watershed scale, with focus on the Hunt River Watershed. The Soil Water Assessment Tool (SWAT) was used to model N inputs into Potowomut Cove, which is fed by the Hunt River, and connects to Narragansett Bay. The effects on receiving waters due to nonpoint source loads, including OWTS could be effectively modeled using SWAT. The development and application of hydrologic models such as SWAT could therefore be used for the Hunt River Watershed to determine Best Management Practices (BMPs). We used a gauged watershed to calibrate the SWAT model, and estimate streamflow and nutrient loading into the bay. Recent development in the SWAT model allows the users to account for presence of septic systems or OWTSs. SWAT modeling also is unique in its ability to account for both watershed and in-stream processes.
4. Discussion
Streamflow and TN load calibration and validation were found to be successful using the SWAT model for the Hunt River Watershed. The model effectively captured nonpoint nitrate inputs from OWTSs through hydrologic modeling of upstream and instream processes. By representing the area under the septic system, using available sewer data and TN loads data, a more accurate model of anthropogenic nitrogen inputs into the Hunt River Watershed was able to be developed. This study showed the importance of groundwater contributions to streamflow and nitrogen loading and the findings are corroborated by research conducted by Jeong et al. and Hoghooghi et al. [
11,
14].
According to Jeong et al. [
11] and Hoghooghi et al. [
14], the contribution of groundwater to streamflow is relatively high during dry years, similar to the research findings of this study.
Since the Hunt River Watershed is listed among the highest priority impacted waterbodies within the state of RI, the ability to effectively model nonpoint nitrogen pollution is essential. Accurate predictions of streamflow and nutrient simulations are important to effectively manage water quality in the receiving watershed and therefore play an important role in nitrogen mitigation and decision-making processes. The Hunt River Watershed’s proximity to Narragansett Bay emphasizes the need for analyzing the impacts of nitrogen loading since this estuary has historically faced threats from eutrophication [
5,
6]. While the Hunt River Watershed is predominantly forested, the urbanization concentrated near the outlet amplifies the impact of OWTS into the receiving waters. Since the state has significantly reduced the input of point source nitrogen pollution from sewage treatment plants, OWTSs are proportionally becoming a larger source of nitrogen into the bay [
8]. Despite reductions in point sources to the bay, eutrophication remains a significant threat to Rhode Island’s fisheries and tourism industries, thus managers and policy makers need accurate models for identifying areas to target nitrogen pollution mitigation.
The use of URIWW data to successfully model nutrients entering Narragansett Bay highlights the value of volunteer water quality monitoring programs for their long-term monitoring assessments. While Narragansett Bay is no stranger to pollution, Rhode Island lacks historical nutrient data at daily and statewide monitoring levels. The lack of data on these scales inhibits the ability for modelers to make predictions based on past trends. Successful calibration of the Hunt River Watershed model could largely be attributed to the availability and use of URIWW data.
While the Nash–Sutcliffe index reflects strong calibration and validation for the model, several changes could be made to improve model fit. Limited knowledge of other nitrogen inputs into the watershed prohibited higher calibration achievement. As listed in the Hunt River TMDL, multiple point and nonpoint sources exist as major pollution contributors to the watershed. This model could be expanded upon by evaluating other potential sources of total nitrogen (TN) including lawn fertilizers, animal waste, and agricultural pollution. Another source of pollution that was not considered in this study includes backflow from the Potowomut Cove, which has the potential to be a significant contribution to the overall nitrogen budget within the area [
33]. Additional nitrogen input sources need to be evaluated for higher model calibration. Another source of potential error is a lack of data on OWTS health, and how failing OWTS contribute to nitrogen loading within the watershed. Currently, little data exists on the overall health of Rhode Island’s OWTSs. Additionally, RI has begun implementing and requiring advanced nitrogen removal OWTS that remove nitrogen through bacterial processes. However, there is still uncertainty on their magnitude of effectiveness and there is limited information on how many of these septic systems are currently installed. In the future, this model could be improved upon as more data becomes available on the status and health of OWTSs and other sources of pollutants.
Overall, SWAT and SWAT-CUP can be used as a tool that supports land use management decisions. Understanding how OWTS alter nitrogen and streamflow inputs into a small-scale watershed allows for analysis of base restoration efforts on numerous user end-benefits. SWAT models are largely acclaimed for their scenario modeling systems [
34]. This research did not explore projected changes via modeling scenarios; however, further research efforts could include scenario modeling due to anticipated changes in climate patterns and urbanization. Overall, the model results imply that OWTS input provides a better estimate of nitrogen loading, resulting in more accurate simulations of pollutant loading into Potowomut Cove. Better prediction models are therefore important to provide informed decision-making tools for the watershed managers and regulators.