Professional Documents
Culture Documents
Irreversible Phosphorus Sorption in Septic System Plumes?: by W.D. Robertson
Irreversible Phosphorus Sorption in Septic System Plumes?: by W.D. Robertson
Abstract
The mobility of phosphorus (P) in septic system plumes remains a topic of debate because of the considerable
reactivity of this constituent. In this study, a septic system plume in Ontario was monitored over a 16-year period
with detail that clearly shows the advancing frontal portion of the P plume. This monitoring record provides
insight into the extent of secondary P attenuation in the ground water zone beyond that available from previous
studies. A P plume 16 m in length developed over the monitoring period with PO4-P concentrations (3 to 6 mg/L)
that approached the concentrations present under the tile bed. Simulations using an analytical model showed that
when first-order solute decay was considered to account for the possibility of secondary P attenuation in the
ground water zone, field values could only be matched when decay was absent or occurred at an exceedingly slow
rate (half-life greater than 30 years). Thus, hypothesized secondary P attenuation mechanisms such as slow re-
cystallization of sorbed P into insoluble metal phosphate minerals, diffusion into microsites, or kinetically slow
direct precipitation of P minerals such as hydroxyapatite were inactive in the ground water zone at this site or
occurred at rates that were too slow to be observed in the context of the current 16-year study. Desorption tests on
sediment samples from below the tile bed indicated a PO4 distribution coefficient (Kd) of 4.8, which implies a P
retardation factor of 25, similar to the field apparent value of 37 determined from model calibrations. This exam-
ple of inactive secondary P attenuation in the ground water zone shows that phosphorus in some ground water
plumes can remain mobile and conservative for decades. This has important implications for septic systems
located in lakeshore environments when long-term usage scenarios are considered.
tracer testing techniques have been described previously extraction with concentrated HCl and HNO3 under heat.
(e.g., Robertson and Cherry 1992; Aravena and Rob- This method liberates cations associated with carbonate,
ertson 1998). Piezometer bundles each consist of 3 to 13 hydroxide, and sulfide minerals but not silicate minerals.
variable-depth sampling tubes. Section A to A9 through Sediment readily desorbable or ‘‘plant-available’’ P
the center of tile bed 2 contains approximately 170 moni- content was measured at the University of Guelph, Soil
toring points (Figure 1), including 47 that were installed and Nutrient Laboratory, by leaching with a 0.5 M NaH-
in 2003/2004 (bundles 123 to 138, section B to B9; Fig- CO3 solution (McBride 1994).
ure 2) targeting specifically the PO4 plume.
Ground water samples were collected using a peristal- Mineral Saturation Conditions
tic pump and were filtered (0.45 lm) in-line prior to An interactive version of the chemical speciation
atmospheric exposure. Samples were acidified to pH less model MINTEQA2 (Allison et al. 1990) using a standard
than 2 with HCl immediately after collection for cation, data base (visual MINTEQ version 2.51, Gustafsson
metals, PO4, and NH4 analyses, or were left untreated 2006) was used to establish equilibrium solute speciation
(other ions). Measurement of pH and Eh was completed and mineral saturation indices (SI).
in-line, prior to atmospheric exposure, using field porta-
ble meters. Eh readings were checked against Zobell’s Simulation of P Migration
solution (Nordstrom 1977) and pH readings were checked Ground water velocities within the local P plume
against buffers of pH 4 and 7. zone were likely to be relatively uniform. Under these
PO4 was analyzed at the University of Guelph, Ontario, conditions, an analytical model can be used to simulate
Soil and Nutrient Laboratory using a colorimetric technique solute transport and has the advantage of being computa-
(cadmium reduction) with a Cobas FaraTM infrared spec- tionally simple and free of numerical dispersion. The ana-
trometer, which provided a detection limit of 0.01 to 0.05 lytical model CMM (Sudicky et al. 1991) was used to
mg/L P. NO3 and Cl were analyzed by ion chromatogra- simulate 1D solute transport along the plume core flow-
phy in the Earth Sciences Department, University of path (streamtube model). The model solves the advection-
Waterloo, which provided a detection limit of less than dispersion equation and considers solute retardation and
0.05 mg/L for NO3-N and less than 0.5 mg/L for Cl. first-order decay. The modeling exercise involved assign-
Analyses of other ions were completed at Maxxam Ana- ing the field indicated velocity value (28 m/year, see sub-
lytics Inc., Mississauga, Ontario, using a variety of ana- sequently) and then calibrating for PO4 retardation (R)
lytical techniques. and hydrodynamic dispersion using the field data. Addi-
Sediment acid-extractable Ca, Fe, Al, and Mn con- tional simulations then considered the effect of slow sol-
tents were analyzed at Activation Laboratories, Ancaster, ute degradation. Degradation in this case is not process
Ontario, using an Aqua Regis technique, which involved specific but could be representative of any attenuation
W.D. Robertson GROUND WATER 46, no. 1: 51–60 53
Figure 2. Section B to B9, Long Point tile bed 2: (a) year 13 PO4 distribution (September 2003 to May 2004), (b) year 16 PO4
distribution (October 2006), and (c) year 16 NO3 distribution (October 2006).
mechanism that is concentration dependent, such as those ‘‘rapid transformation zone.’’ Consequently, the 2.6-m-
suggested previously to account for secondary P attenua- deep piezometer at nest 123 was selected as the source
tion (e.g., molecular diffusion, slow recystallization of location for the modeling exercise. This piezometer is
sorbed P, or slow direct precipitation). Selection of the located at the downgradient edge of the tile bed (Fig-
source P concentration was considered carefully for this ures 1 and 2), 1.9 m below the infiltration pipes (~0.6 m
modeling exercise. The septic tank effluent was sampled below the water table), and samples plume water that is
on 15 occasions during 1989 to 2004 and PO4-P concen- fully nitrified. Thus, sewage oxidation reactions are
trations remained consistent (6.3 6 1.4 mg/L). At most largely completed at this depth. Additionally, elevated
septic system sites, however, some P attenuation occurs ground water PO4-P concentrations, greater than 1 mg/L,
abruptly in the first 1 to 2 m of subsurface flow in were measured at this location within 4 months after ini-
response to changing redox conditions and ensuing rapid tiation of sewage loading in 1990 (Figure 3). For the
mineral precipitation reactions (Zanini et al. 1998; modeling exercise, the mean PO4-P concentration mea-
Robertson et al. 1998). Thus, it would be inappropriate sured in this piezometer during the last three year-16 sam-
for a modeling exercise focused on slower secondary pro- pling events (6.4 6 0.7 mg/L, 08-11/06) was used as the
cesses to choose a starting point that included the initial P source concentration. This value was similar to the mean
54 W.D. Robertson GROUND WATER 46, no. 1: 51–60
the upper boundary of the plume to migrate progressively
deeper as the season progresses. With the initiation of
sewage loading in May, a new slug of waste water begins
to occupy the shallow water table zone near the tile bed
and then migrates laterally away from tile bed at a rate
reflecting the horizontal ground water velocity. Conse-
quently, the piezometer nests on section B to B9 each
have one to three shallow monitoring points that oscillate
seasonally between being in or out of the plume. The
background ground water (electrical conduction [EC] 300
to 600 lS) is easily distinguished from the plume water
(EC 800 to 1400 lS) and the boundary frequently occurs
over a vertical distance less than the piezometer spacing
(less than 0.5 m). In 2004, detailed EC monitoring was
undertaken on seven occasions, which revealed a rate of
advance of the plume of 40 m/year over the 140-d period
from May 1 to September 17. However, the ground water
velocity was likely higher than average during this period
Figure 3. Distribution of phosphorus in the source area
due to the added sewage loading. If an arbitrary assump-
underlying the tile bed: (a) depth profile of sediment ad-
sorbed P content, core 121 year 16 (08/06) and (b) depth pro- tion is made that the flow velocity for the remainder of
files of ground water PO4 content (years 0.4 to 7, nest 106; the year is 50% of the May 1 to September 17 rate (i.e.,
years 13 to 16.5, nest 123). 20 m/year), an average annual flow velocity of 28 m/year
is obtained. This estimate is considered more reliable
than previous estimates based on the Darcy equation
long-term effluent value of 6.3 6 1.4 mg/L, although (e.g., 30 m/year, Aravena and Robertson 1998), although
more recent effluent values have indicated a slightly high- values are similar. Additionally, the Br tracer test carried
er value of 7.2 6 1.2 mg/L (n ¼ 3, 1995 to 2004). The out in 1990 revealed the vertical migration rate of the
reason for the modest amount of P attenuation in the sewage plume below the tile bed (4 cm/d) during a period
vadose zone at this site compared to other sites (25% to of moderate sewage loading in the fall (Figure 1).
99%, Robertson et al. 1998) is uncertain but could be
related to the high sewage loading rate (10 cm/d during Plume Development
peak use). In any case, the high PO4 concentrations pres- Figure 1 shows the current position of the 16-m-long
ent should only increase the propensity for secondary re- PO4 zone, within the larger septic system plume. Figure 3
actions to occur along the plume flowpath. shows depth profiles of PO4 distribution below the tile
bed over the 16-year period beginning with commission-
ing of the tile bed in 1990. During the last decade, PO4-P
Results concentrations in the source area have remained consis-
Aquifer Characteristics tently in the 4 to 7 mg/L range. Figure 2 shows detailed
Coring undertaken previously determined that the PO4 distribution within the P plume zone (section B to
surficial aquifer is 5 to 6 m thick and consists of silt-free B9) in years 13 (2003) and 16 (2006). By year 13, the P
fine to coarse sand, underlain by a flat lying aquitard of plume had advanced to approximately 15 m from the
lacustrine clayey silt. The mean hydraulic conductivity of edge of the tile bed and by year 16, further advancement
the aquifer in the P plume zone is 0.02 cm/s (Table 1). of the P plume was evident. Although denitrification is
The aquifer sands are calcareous (6.3 6 0.8 wt % Ca) active at depth (Aravena and Robertson 1998), Figure 2c
and contain substantial amounts of acid-extractable shows that the P plume remains entirely within the
Al (0.13 6 0.02 wt %), Fe (2.5 6 1.2 wt %), and Mn nitrate-rich zone (NO3-N greater than 10 mg/L).
(0.21 6 0.03; Table 1). This presumably reflects the pres- The absence of NH4-N in the plume water (less than
ence of hydroxide minerals, which should contribute to 0.1 mg/L) and low dissolved organic carbon (DOC) val-
PO4 sorption. The aquifer also contains trace quantities of ues (3 mg/L; Table 2) indicate that the effluent is well
organic carbon (0.15 wt %) and reduced S (0.02 wt %, oxidized. Acidity generated from the oxidation reactions
probably as biogenic pyrite), both of which are suffi- has been buffered by carbonate mineral dissolution as
ciently labile to promote denitrification in deeper parts of indicated by increases in Ca and Mg concentrations in the
the aquifer (Robertson and Cherry 1992; Aravena and plume water compared to the effluent (Table 2). Other
Robertson 1998). plume constituents such as K (17 to 21 mg/L) and Na (34
to 46 mg/L) remain consistent along the plume and are
Ground Water Velocity similar to the effluent values. Although these cations are
The seasonal nature of sewage loading at the site also subject to sorption reactions, their retardation factors
provided an opportunity to directly measure the horizon- in sands are normally small (e.g., R less than 5; Dance
tal ground water velocity. During the 6-month period and Reardon 1983; Ceazan et al. 1989) compared to the
when no sewage loading occurs (November to April), large number of pore volumes (~30) that have migrated
precipitation continues to recharge the aquifer, causing through the plume zone of interest here.
W.D. Robertson GROUND WATER 46, no. 1: 51–60 55
Table 1
Sediment Properties: Hydraulic Conductivity Determined from Grain-Size Analyses of Core 137, 0.8 to 2.5 m
Depth (n ¼ 6); Acid-Extractable Cations (Al, Ca, Fe, and Mn) from Core 128, 1.0 to 1.8 m Depth (n ¼ 5), OC
Content (Aravena and Robertson [1998]), and CEC (Robertson and Cherry [1992])
Hydraulic Conductivity (cm/s) Al (wt %) Ca (wt %) Fe (wt %) Mn (wt %) OC (wt %) CEC (meq/100 g)
Aquifer 0.02 6 0.01 0.13 6 0.02 6.3 6 0.8 2.5 6 1.2 0.21 6 0.03 0.15 5.2
Simulation of PO4 Migration (1995/1996) when no sewage loading occurred to tile bed
Figure 4 shows model calibrations for the P retarda- 2, a reduced ground water velocity of 20 m/year was
tion factor (R, ground water velocity/solute migration assumed (see previously). This value is 71% of the nor-
velocity) and aquifer longitudinal dispersivity (aL) along mal velocity; thus, this 2-year period is represented as
the plume core flowpath. Note that for the 2-year period a shorter interval of only 1.4 years in the modeling exer-
cise. A good fit with the field data is obtained using a P
retardation factor of 37 and an aL value 0.5 m. This
Table 2
Geochemistry of the Septic System Plume with
Increasing Distance from the Tile Bed and SI for
Several Minerals That Could Influence
P Mobility