Phosphorus loading from onsite wastewater systems to a lake (at long time scales)

Abstract

Schellenger FL, Hellweger FL. Phosphorus loading from onsite wastewater systems to a lake (at long time scales). Lake Reserv Manage. 35:90–101.

Phosphorus from onsite wastewater systems (OWS, septic systems, cesspools) enters the groundwater and may migrate to and contribute to eutrophication of down-gradient lakes. Since phosphorus sorbs relatively strongly to soil solids, this source may be assumed negligible in lake management. However, phosphorus adsorption is reversible, and given enough time, some phosphorus will eventually reach the lake. We investigated the magnitude and timing of phosphorus loading to a case study lake in southeastern Massachusetts, using a model that accounts for cesspools and septic systems installed over a 270-yr time frame, the 1994 laundry detergent phosphorus ban, loss of phosphorus in the vadose zone, and transport and adsorption/desorption (retardation) in the surficial aquifer. We parameterized the model using literature and local information and applied it to all the OWS in a the lake watershed to estimate annual wastewater total phosphorus loading over a long time. Model results were compared to previous phosphorus budgets based on limited field measurements. The model predicts that groundwater transport of phosphorus from OWS by itself can account for the current eutrophic state of the lake and will continue to increase. Our results show that the OWS source can be significant and should not be ignored. We performed parameter sensitivity and Monte Carlo uncertainty analyses and illustrated the effects of removal in the vadose zone and travel time in the aquifer. Our model suggests that past management actions (cesspool to septic system conversion, laundry detergent phosphorus ban) effectively reduced the 2018 load by 31.4%. A centralized wastewater treatment facility would eliminate the OWS loading, but the half-life due to the existing phosphorus reservoir in the aquifer is long (20 yr).

Keywords:

• Cesspool
• eutrophication
• groundwater
• onsite wastewater disposal
• phosphorus
• septic system

Cultural eutrophication of lakes is a worldwide problem that is associated with harmful algae blooms and other symptoms (Schindler 2006; Dodds et al. 2009; Paerl et al. 2011), and phosphorus inputs are often considered the primary cause (Schindler 2006; Lewis and Wurtsbaugh 2008). Phosphorus can enter a lake from a number of sources, including wastewater treatment plants, tributaries, and direct runoff. Internal loading (i.e., recycling from the sediment bed) can also be an important source (Cooke et al. 2005; Wagner et al. 2017).

Groundwater is another source of nutrients to surface waters (Lewandowski, et al. 2015; Meinikmann et al. 2015), and onsite wastewater systems (OWS) have long been recognized as a possible source of groundwater contamination (Beal et al. 2005; Humphrey and Driscoll 2011; Oosting and Joy 2011; Eveborn et al. 2012). Nutrients from OWS may travel to down-gradient surface waters and contribute to their eutrophication. Based on studies of chemical reactions in OWS effluent plumes, a consensus developed that nitrogen (as nitrate) has the potential to travel great distances down-gradient from an OWS (e.g., Dudley and Stephenson 1973; Jones and Lee 1979; Bicki et al. 1984). On the other hand, these studies have reported that the unsaturated (vadose) zones of OWS soil absorption systems are highly efficient at sequestering phosphorus, and have concluded that phosphorus will be sufficiently retarded by sorption or immobilized by precipitation so that movement is limited to a few meters at most (Reneau and Pettry 1976; Stolt and Reneau 1991). For some older systems, down-gradient transport of phosphorus was found (Dudley and Stephenson 1973; Reneau and Pettry 1976; Gilliom and Patmont 1983); these were viewed as isolated cases, and down-gradient movement from a well-designed and maintained OWS is generally not considered a threat to surface waters (Sikora and Corey 1976; Jones and Lee 1979; Reneau et al. 1989). Consequently, the contemporary nutrient management process is based on this paradigm. Remediation and restoration efforts for eutrophic lakes and other surface waters generally focus on controlling above-ground nutrient sources and suppression of internal loading (Cooke et al. 2005; Wagner et al. 2017).

Ignoring the phosphorus contribution of OWS to a lake may not be warranted at long time scales. Some studies of phosphorus behavior in the OWS vadose zone have found that sorption of phosphate anions to aluminum and iron hydroxide coatings on soil particles, with the subsequent precipitation of mineral phosphates, is enhanced at groundwater pH values below circumneutral (e.g., Robertson 2003, 2012). Precipitation may be a permanent loss process, but phosphorus sorption is reversible (Dudley and Stephenson 1973; Whelan and Barrow 1984; Walter et al. 1996). There are a number of studies that explored the long-term transport of phosphorus from an OWS in groundwater (e.g., Koerner and Haws 1979; LeBlanc 1984; Robertson 2003; Andres and Sims 2013; Jarosiewicz and Witek 2014). Findings from these studies suggest that, given sufficient time, phosphorus can move down-gradient for a significant distance.

From the standpoint of eutrophication, the total load to a lake has to be considered, so studies of individual OWS are of limited utility. A number of studies have estimated groundwater phosphorus loading from the watershed to a surface water body, finding the load to be significant and suggesting OWS as a potential source (Vanek 1991; Moore et al. 2003; Jarosiewicz and Witek 2014; Meinikmann et al. 2015). However, these studies do not provide a quantitative link to OWS. To our knowledge there have been no studies that combine the estimated loading from all the OWS in a watershed with the long-term dynamics of transport and retardation in the groundwater and quantify the relationship to the trophic state of the receiving lake.

We estimated the cumulative loading of phosphorus from all the OWS in a lake watershed at a long time scale, from the installation of the first OWS to when the present-day loadings from all the OWS in the watershed reach the lake. We used a simple model of phosphorus transport to predict the time-variable phosphorus load from each system to a down-gradient lake (Fig. 1), and then summed the loads to obtain the total loading. This loading was then compared to results from studies intended to assess overall phosphorus loading to the lake.

Materials and methods

Study site

Oldham Pond is a 93.89-ha (232-ac) lake with mean depth of 2.65 m (8.7 ft), located on the border between the towns of Pembroke and Hanson in Plymouth County in southeastern Massachusetts (Fig. 2). The watershed is small, comprising 308.78 ha (763 ac), with 4 small tributary streams to the lake (CEI 2011). The watershed is located in a coastal plain that slopes gently toward Massachusetts Bay. The local climate is mild temperate. The lake occupies about 25% of the watershed, and two-thirds of the land area is medium-density residential households, while the remaining one-third is forest, open land, and wetlands (BEC 1993; CEI 2011). Like most of the lakes in southeastern Massachusetts, including Cape Cod, Oldham Pond lies in glacially deposited sand and gravel and is in contact with an underlying aquifer (Carlson and Lyford 2005). Groundwater studies (BEC 1993; Carlson and Lyford 2005) have concluded that the lake receives much of its inflow from groundwater, and that some outflow from the lake is recharge to the groundwater. The general flow of groundwater in the region is from west to east, under the lake, toward Massachusetts Bay (Carlson and Lyford 2005). Locally, along the shoreline the groundwater gradient is toward the lake.

The lake is surrounded by residential development and serves as a major recreation venue for swimming, boating, and fishing. It is eutrophic and supports blooms of algae each summer. In recent years, these blooms have been dominated by cyanobacteria and have often forced town officials to close the lake for swimming. Two studies have been performed to characterize the nutrient enrichment to the lake and to identify potential remedial actions (BEC 1993; CEI 2011). These studies have concluded that the lake is receiving excess phosphorus input from a number of sources, including overland and point-source runoff and recycling from the lake-bottom sediment. Inputs from groundwater were considered, but longer term inputs were discounted in the belief that phosphorus is readily sorbed to the iron-rich soil and thereby immobilized (BEC 1993; CEI 2011). Our study only addresses the phosphorus loads transported to the lake from OWS via the groundwater, and does not address the phosphorus loads from all sources. Our estimate of the P load delivered to the lake from OWS is independent of any other sources and not affected by them.

Model description

The model considers three stages of total phosphorus transport: (1) loading from the cesspool or septic tank, (2) loading to the groundwater, and (3) loading to the lake (Fig. 1). The equations for these three stages are described next. The focus of the model is on the long-term (10+ yr) dynamics resulting from development (i.e., installation of systems), cesspool to septic system conversion, laundry detergent phosphorus ban, and travel time (advection and retardation). Variability at shorter time scales (e.g., seasonality of hydrology or household occupancy) is not considered.

Figure 1. Model overview, illustrating the three stages of the model for a conventional septic system.

Figure 2. Oldham Pond Watershed, Pembroke and Hanson, Plymouth County, Massachusetts. All 859 OWS locations are shown as circles, and groundwater flow paths as lines. Stippled polygons are wetlands or cranberry bogs.

Stage 1: Loading from the cesspool or septic tank: Wastewater is conveyed to a cesspool or septic tank, where heavier solids sink to the bottom and lighter solids form a scum at the top (USEPA 2002). Solids and scum are periodically removed as part of regular system maintenance. The liquid between contains mostly dissolved species. In a conventional septic tank system, liquid effluent is piped to the soil adsorption system (SAS). Before the advent of regulations requiring the use of a septic tank and SAS, a cesspool was common. A cesspool serves the same purposes as the later septic tank, but is built with porous sides and bottom. Effluent is released primarily from the sides of the cesspool. With the advent of septic system regulations, many cesspools were replaced by septic systems and the cesspools were abandoned. In most states (including Massachusetts) replacement is not required until property is sold, at which time an inspection determines compliance with current regulations. The phosphorus loading from the cesspool or septic tank, W_eff (M/T), is given by (1) $$W_{eff}=C_{eff}\times Q_o$$(1) where C_eff is the concentration of phosphorus in the effluent leaving the cesspool or septic tank (M/L³), and Q_o is the effluent flow rate (L³/T). Averages of C_eff and Q₀ are considered constant, but C_eff can change due to the laundry detergent phosphorus ban.

Stage 2: Loading to the groundwater: The effluent from the conventional septic system is distributed to the SAS via perforated pipes. Below these pipes are several feet of unsaturated soil, also called the vadose zone. After leaving the cesspool or septic tank, the effluent travels vertically through this vadose zone to the groundwater table. Depending on the soil texture, pH, content of metal hydroxides, and depth to the aquifer, a fraction of the phosphorus is sorbed or precipitated, preventing it from reaching the groundwater.

Typically, system installation depths and soil properties are not consistently available, so we use a removal efficiency. The efficiency of the vadose zone in removing phosphorus, η, is defined as (2) $$\eta =\left(C_{eff}-C_o\right)/C_{eff}$$(2) where C_o is the concentration of phosphorus in the effluent as it reaches the water table (M/L³). The loading from the vadose zone to the groundwater, W_o (M/T), is given by (3) $$W_o=C_o\times Q_o$$(3)

W_o is assumed to be a constant step input to the water table/saturated zone, which assumes that the septic system is mature enough to have reached a steady state. “Steady state” exists when phosphorus sorption and precipitation reactions are retarding phosphorus in the vadose zone at a rate that allows the same load to reach the water table at all times (Robertson 1995; Beal et al. 2005). The efficiency is different for cesspools and conventional septic systems. For cesspools, the bottom is typically deeper than a conventional SAS and the travel depth through and removal in the vadose zone is less, which can be accounted for by using a lower removal efficiency.

Stage 3: Loading to the lake: Stage 3 considers transport in the aquifer. Once in the aquifer, the effluent plume is advected in the direction of groundwater flow, as well as dispersed in all three dimensions, horizontal, vertical, and transverse to flow. Depending on the soil properties, phosphorus will be retarded by sorption, which is modeled as an instantaneous and reversible process (i.e., local equilibrium). In this conceptual model, the groundwater is an unconfined aquifer of glacially derived sand and gravel typical of southeastern Massachusetts, including Cape Cod, and the aquifer is underlain by an aquaclude, usually bedrock, but sometimes dense silt, clay, or glacial till that is much less permeable than the sand. The focus of our study is on the long-term dynamics, and for that purpose the groundwater hydraulics are considered constant in time (i.e., constant flow velocity). These processes can be described using the advection–dispersion–sorption (ADS) equation (Domenico and Schwartz 1990; Schnoor 1996). Here, we are concerned with the total loading to the lake by groundwater. Although there will be some vertical and lateral dispersion of the plume, this will not affect the total load to the lake. We therefore use the one-dimensional (1-D) version of the ADS equation. For a continuous step input, and in terms of loading, W, the analytical solution to the 1-D ADS equation is (Ogata and Banks 1961): (4) $$W=\left(\frac{W_0}{2}\right)\mathrm{\{}\mathit{\text{erfc}}\frac{\left(R\mathrm{}x-u_x\mathrm{}t\right)}{2\mathrm{}{\left({\alpha }_x{\mathrm{}u}_x\mathrm{}R\mathrm{}t\right)}^{1/2}}+\hspace{0.17em}\text{exp}\hspace{0.17em}\left(\frac{x}{{\alpha }_x}\right)\mathrm{}\mathit{\text{erfc}}\frac{\left(R\mathrm{}x-u_x\mathrm{}t\right)}{2\mathrm{}{\left({\alpha }_x{\mathrm{}u}_x\mathrm{}R\mathrm{}t\right)}^{1/2}}\mathrm{\}}\text{where}R=1+\frac{K_d\mathrm{}{\rho }_b}{n}$$(4)

In this equation, erfc is the complimentary error function, t is time (T), and x is distance in the direction of groundwater flow (L); u_x is the average pore water velocity in the x direction (L/T²); α_x is the dispersion coefficient in the x direction (L); ρ_b is the soil bulk density (M/L³); n is the effective porosity (L³/L³); K_d is the linear sorption distribution coefficient (L³/M); and R is the retardation due to sorption (dimensionless). The model accounts for the time lag between the input to and output from the aquifer (to the lake) due to advection and adsorption/desorption (retardation).

The total phosphorus load to the lake, W, is calculated using equation 4(4) $$W=\left(\frac{W_0}{2}\right)\mathrm{\{}\mathit{\text{erfc}}\frac{\left(R\mathrm{}x-u_x\mathrm{}t\right)}{2\mathrm{}{\left({\alpha }_x{\mathrm{}u}_x\mathrm{}R\mathrm{}t\right)}^{1/2}}+\hspace{0.17em}\text{exp}\hspace{0.17em}\left(\frac{x}{{\alpha }_x}\right)\mathrm{}\mathit{\text{erfc}}\frac{\left(R\mathrm{}x-u_x\mathrm{}t\right)}{2\mathrm{}{\left({\alpha }_x{\mathrm{}u}_x\mathrm{}R\mathrm{}t\right)}^{1/2}}\mathrm{\}}\text{where}R=1+\frac{K_d\mathrm{}{\rho }_b}{n}$$(4) for each OWS in the watershed. Calculations were done in MS Excel, using a custom function written in Visual Basic. The input loading, W_o, can change in time for any OWS, by replacement of a cesspool with a conventional septic system and/or a switch to use of a nonphosphorus detergent. Even for those cases when a cesspool was replaced by a conventional septic system, the long time scale of phosphorus transport in groundwater necessitates explicit consideration; the cesspool may have been eliminated, but the plume from it is still “en route” to the down-gradient lake. The analytical solution assumes a continuous step input, but time-variable input can be modeled by combining multiple solutions using the principle of superposition (Freeze and Cherry 1979); see Supplement A for an illustration of this superposition calculation.

Model inputs

The Oldham Pond watershed contains 859 OWS serving residences and a few small commercial facilities. We assumed that this number will remain constant in the future, as the watershed is nearly completely developed. A build-out analysis by CEI (2011) indicates that infill development of less than 2% on an area basis (about 13 residences) can be accommodated in the watershed.

The model inputs include, for each of the 859 OWS, the installation time and cesspool to septic system conversion date (if applicable) and the location in the watershed, which was used to derive the flow path using GIS functions (see Supplement B for details). The flow rate (Q_o) was based on census data on household occupancy. Groundwater water elevations were based on observations or estimated from ground level and used to calculate the hydraulic slope. The cesspool and septic tank effluent concentrations (C_eff) were based on literature values reported by McCray et al. (2005). Vadose zone efficiency (η) was based on a review of a number of literature sources and local expertise (see Supplement B). Hydraulic conductivity (K_s), effective porosity (n), and longitudinal dispersivity (α_x) were based on reported literature values. The distribution coefficient (K_d) was determined from data on the phosphorus plume from the Otis Air Force Base infiltration beds (LeBlanc 1984), using a curve-fitting approach. Model inputs are provided (Table 1) with literature references, and further described in Supplement B.

Table 1. Parameter values.

Parameter	Symbol	Units	Value used (range)	Literature values	Sources/notes
Effluent concentration	Ceff	mg/L
CP, before 1994 ban			15.0 (6.3–24.9)	6.3–24.9	a, e, 5
CP, after 1994 ban			9.6 (1.4–16.2)	1.4–16.2	a, e, 5
SS, before 1994 ban			15.0 (6.3–24.9)	7.2, 6.3–24.9	a, e, 1, 5
SS, after 1994 ban			9.6 (1.4–15.2)	7.2, 1.4–16.2	a, e, 1, 5
SAS efficiency	η	—
CP			0.50 (0.45–0.55)
SS			0.75 (0.675–0.825)	0, 0.27, 0.47, 0.75–0.80, 0.99	a, 10, 1, 6, 12, 11
Saturated hydraulic conductivity	KS	m/d	13.7 (8.64–86.4)*	91.5, 21.9, 13.7, 8.64–86.4	a, 3, 2, 7, 8
Effective soil porosity	n	—	0.39 (0.30–0.42)	0.35, 0.39, 0.30–0.42	a, 2, 3, 8, c
Longitudinal dispersivity	αx	m	1.0 (0.1–100)*	1.0, 0.96, 0.1–100	a, 2, 4, 9, d
Soil bulk density	ρb	g/cm^3	1.617		b
Linear sorption coefficient	Kd	cm^3/g	5.0 (4.6–7.0)*	1.4–478	a, 5,6
Average distance to Oldham Pond	x	m	180.9 [5.0–853.0]
Average groundwater velocity	ux	m/yr	217.2
Average effluent loading rate	Qo	m^3/yr/household	218.45
Retardation coefficient	R	—	21.72

CP: cesspool; SS: septic system; SAS: soil absorption system.

Sources: 1. Robertson (1995); 2. Robertson et al. (1991); 3. Walter et al. (1996); 4. LeBlanc (1984); 5. McCray et al. (2005); 6. Heufelder and Mrockza (2006); 7. Carlson and Lyford (2005); 8. Freeze and Cherry (1979); 9. Garabedian et al. (1991); 10. Robertson (2008); 11. Robertson et al. (1998); K. Wagner (personal communication).

Notes:

a. Values in parentheses are ranges used in the parameter sensitivity and uncertainty analyses. For ranges marked with asterisk, the Monte Carlo selection was done in log space.

b. Saturated soil density, ρ_s, is assumed to be 2.65 g/cm³, and soil bulk density is then given by ρ_b = (1 – n) × ρ_s (Freeze and Cherry, 1979).

c. The effective porosity, n, of soils around Oldham Pond is taken as the same value reported by LeBlanc (1984), as typical of sands in the region.

d. Longitudinal dispersivity, α_x, for soils around Oldham Pond is taken as the same value reported by LeBlanc (1984) and by Robertson et al. (1991). Range for sensitivity per Garabedian et al. (1991).

e. Values are total P, converted from literature ortho-P values assuming 87.5% of total P is ortho-P, based on 85–90% from McCray et al. (2005).

Results and discussion

Time series of P loading to the lake

Phosphorus loading calculations were performed for the 859 OWS in the Oldham Pond watershed at a time step of 1 year and summed to give the total phosphorus loading to the lake for each year from 1750 to 2250 (Fig. 3); see Supplement C for sample loading time series for 5 OWS. This time period starts when the first OWS was installed and ends when the present-day loading (i.e., assuming no additional development in the watershed going forward) from all OWS reaches the lake. The loading increases with time as the phosphorus from more systems reaches the lake. After the 1994 laundry detergent ban, the predicted loading declined briefly, but then resumed an increasing trend by 2010. This increase is not due to an increase in OWS after 2010 (see earlier discussion), but to a continuous increase in the number of OWS plumes reaching the lake. After about 2070 all the OWS plumes are contributing and the total loading is relatively constant. The model predicts that phosphorus loading from OWS will continue to increase in the future, even with no further development in the watershed.

Figure 3. Estimated total phosphorus loading to Oldham Pond from OWS versus time, 1750 through 2250. Circles are loading estimates by Baystate Environmental Consultants, Inc. (BEC 1993; K. Wagner 2018, personal communication), and Comprehensive Environmental, Inc. (CEI 2011), respectively, as described in the text.

Our model estimates a relatively steep increase in loading from 1930 to 1994. The slope of this increase is consistent with observations of in-lake phosphorus concentration. Specifically, the model slope corresponds to a loading increase by about 15.6% per decade for 1930 to 1996. Observations of average in-lake total phosphorus concentration of 40 μg/L for 1987–1988 (BEC 1993) and 66 μg/L for 2006 through 2010 (CEI 2011) show that the in-lake total phosphorus concentration increased by about 17.9% per decade from 1988 to 2010. There is a temporal mismatch: For the years corresponding to the observations, the model already predicts a decrease due to the phosphorus detergent ban (discussed further in the next section).

Relation to trophic state and other loading estimates

Trophic status bands for “permissible” and “dangerous” loading (Vollenweider 1975) put these loads into the context of lake eutrophication (Fig. 3). Oldham Pond phosphorus loading predicted by the model from OWS alone passed the “dangerous” level into the eutrophic zone by about 1965 and continues to be sufficient to maintain a eutrophic state, without considering any other source of phosphorus to the lake.

We compared our results to an estimate using the NPSLAKE procedure of Mattson and Isaac (1999), that is, 0.5 kg/yr for each household within 100 m of the lake. This procedure is recommended by the Massachusetts Department of Environmental Protection to estimate phosphorus loading when establishing total maximum daily loads (TMDLs) according to the Clean Water Act (CWA 2002). Similar procedures based on export coefficients are used elsewhere in the United States and Canada (e.g., Paterson et al. 2006; Dillon et al. 1994). In 2016, there were 241 households within 100 m of Oldham Pond in the watershed; the NPSLAKE procedure yields an estimated phosphorus loading substantially below our estimate (Fig. 3). Loading estimates based on static export coefficients do not account for variation over time, and the NPSLAKE model does not account for OWS beyond 100 m.

Our estimate of phosphorus loading from OWS via groundwater can be compared to independent estimates based on nutrient budgets. For 1988, our model estimate is 605 kg/yr and considerably above the 302 kg/yr from a study by Baystate Environmental Consultants (BEC 1993, revised using improved methods by K. Wagner, original author of the BEC 1993 study, 2018, personal communication).

Comprehensive Environmental, Inc. (CEI), estimated the phosphorus input by groundwater in 2010 (CEI 2011). It based its estimate on runoff coefficient models and the concentration measurements by BEC (1993). When CEI compared its estimated total load (from all sources) to a value back-calculated from the in-lake phosphorus concentration (using the method of Reckhow (1979)), it concluded that there is a large unaccounted-for source. We attribute that source to groundwater, and the resulting phosphorus loading by groundwater is substantially higher, 604 kg/yr, and comparable to our model estimate for 2010, 556 kg/yr.

Comparison of the model-estimated load to those of the two studies (and their P concentrations, mentioned in the previous section) shows that there is a timing mismatch. The observations suggest a later rise in OWS loading, which may be due to retention in or on the sediment bed, a mechanism not accounted for in the model. Specifically, P entering the lake via groundwater may sorb onto iron oxide on the sediment bed surface or iron oxides precipitated from dissolved iron also carried by the groundwater. In that case the P would enter the lake as part of the “internal loading” at a later time, consistent with the observed mismatch.

Parameter sensitivity analysis and uncertainty (Monte Carlo) analysis

There are a number of sources of uncertainty in our model, including assumptions in the model (e.g., constant vadose zone removal efficiency), spatial input data (number of contributing OWS and their flowpaths), and parameter values. To quantify the sensitivity of the model to the input parameters we performed a parameter sensitivity analysis (see Supplement E.1). The results show that the magnitude of the loading is most sensitive to the effluent concentration (C_eff), and to a lesser extent, to the vadose removal efficiency (η). The timing of loading is sensitive to the hydraulic conductivity (K_s) and distribution coefficient (K_d). To quantify the overall uncertainty in the model results we performed a Monte Carlo analysis (Fig. 3). The results show uncertainty in timing and magnitude. For example, the upper and lower 90% confidence intervals have final loading magnitudes about a factor of 2.0 and 3.1 different from the median or base values, respectively. This level of uncertainty is relatively high, and illustrates the need for more data specific to the Oldham Pond watershed. However, the model still provides important information for management. In climate science, models also have a high level of uncertainty, but it is recognized that they represent our best estimate, and their predictions are used for management (Maslin and Austin 2012). We also performed the Monte Carlo analysis using the full literature ranges of parameters (vs. the local ranges used in the preceding). The resulting uncertainty is substantially higher, further highlighting the need for site specific parameter estimates in these types of studies (see Supplement E.3).

Effect of vadose zone immobilization and travel time

To illustrate the effect of phosphorus immobilization in the vadose/unsaturated zone and retardation in the aquifer/saturated zone, we compared the loading to the lake to that leaving the OWS and entering the groundwater (Fig. 4). The total load leaving all the OWS increased with development of the watershed and exhibits an abrupt drop in 1994 due to the detergent phosphorus ban. The load entering the groundwater is approximately proportional to this, although there are temporal differences in the vadose zone efficiency as cesspools are replaced by septic systems. The load to the lake reflects the retardation and travel time in the groundwater, and shows that the plumes of a substantial fraction of the OWS are still en route to the lake.

Figure 4. Comparison of phosphorus loading by model stage.

Effectiveness of past and potential future management actions

We explored the effect of past management actions, including the 1994 detergent phosphorus ban and the conversion of cesspools to septic systems. If these management actions had not been implemented, phosphorus loading to the lake in 2018 would be 31.4% higher with no detergent phosphorus ban, and 89.5% higher with no ban and continued use of cesspools (Fig. 5). These differences increase for future years. In addition, the model is used to estimate the impact of replacing all OWS by a sewer system and wastewater treatment facility (WWTF). Because design, site selection, licensing, and construction would take time, the year 2030 is used as a facility start date for illustration. We assume the effluent from the WWTF is discharged to the lake. Based on tertiary treatment for phosphorus removal that achieves 100 μg/L P in the effluent (USEPA 2007; MADEP 2004), the phosphorus loading from the facility could be as low as 18.73 kg/yr. Although the loading to the lake decreases rapidly at first, residual loading continues for many years due to the present reservoir of phosphorus in the aquifer (Fig. 5). The time from installation to when the loading from OWS and WWTF to the lake is cut in half is 20 yr. The loading is estimated to decrease to Vollenweider’s “permissible” load by 2069.

Figure 5. Comparison of model estimated phosphorus loading to estimates (1) if there were no laundry detergent phosphorus ban (blue dashed curve), (2) if all OWS were built as cesspools (red dashed curve), and (3) if all OWS were replaced by a sewer system and WWTP in 2030 (green dashed curve).

Relation to other studies

Our results add to the growing body of evidence that suggests phosphorus from OWS can be a substantial contributor to eutrophication of downstream lakes. Our study is unique in that it explicitly links the OWS source to the lake loading and accounts for the dynamics of system installation, cesspool to septic system conversion, laundry detergent phosphorus ban, and transport and retardation in the groundwater. Past studies have established that groundwater input is significant and have implicated OWS (Vanek 1991; Jarosiewicz and Witek 2014; Meinikmann et al. 2015), or have estimated loadings from OWS, but have not considered the dynamics of transport to the downstream lake (Valiela and Costa 1988; Weiskel and Howes 1992; Dudley and May 2007). Our results suggest that long-term dynamics are important. For example, the model predicts the loading to the lake will continue to increase for several hundred years, even without additional OWS, so long-term dynamics need to be considered. Even with potential overestimation, it is clear that loading from OWS will not decline and represents an ongoing and significant threat to water quality in Oldham Pond.

Management perspective

Our results have serious implications for lakes with similar hydrology, watershed development, and wastewater disposal methods. They suggest that the assumption that OWS are not an important P source due to sorption and precipitation is not justified. This study suggests that current lake management methods to prevent or remediate algae blooms and nuisance plants are not likely to be successful in the long term until the OWS source is addressed. A feature of the OWS groundwater load important for management is the time lag, due to the travel time in the groundwater plus possible temporary incorporation into the sediment bed. That means any management action will affect the lake water column with a substantial time delay. Also, if the groundwater load enters the sediment bed and becomes part of the internal loading, this ongoing input may compromise in-lake management actions (e.g., alum treatment).

If continued use of OWS is elected in lieu of a sewer system and wastewater treatment facility, regulations for OWS should be changed to ensure vadose zone efficiency in immobilizing phosphorus is as high as possible. Additional and continued education of OWS users should emphasize decreased water use to lower the phosphorus load to the OWS, as well.

Further research is needed to quantify the impact of OWS at the watershed level over extended times. More field research is also needed to establish local soil parameters at a watershed scale, especially local average values for K_s, K_d, and vadose zone efficiency, η, to reduce the uncertainty inherent in using average literature values from other watersheds.

Outlook

The model we have presented here can be applied to other lake watersheds and, with some modifications, to a larger scale—regional and national—to characterize the long-term impact of OWS. Many of the required data are available in national databases—for example, population and household data from the U.S. Census; water use data from the U.S. Geological Survey; watershed and surface water body data from the National Hydrography Dataset (USGS); soils data from the USDA-NRCS Soils Survey Manual; and elevation data from the National Elevation Dataset (USGS). These data must be supplemented by local and regional data regarding groundwater aquifers. As in the Oldham Pond study, there will be substantial uncertainty associated with the estimated OWS loading in such a study. However, even “best estimate” or “worst case” loadings will be quite useful for lake managers.

Acknowledgments

The authors thank Donna Tremontana and Theresa Cocio of the Hanson Health Department and Lisa Cullity of the Pembroke Health Department, who provided data. The lead author thanks Eileen M. Penney for financial support. This article was improved by review and comment by Dr. Ken Wagner and two anonymous reviewers. This research did not receive funding from agencies in the public, commercial, or not-for profit sectors.