Critique of Predictions/No Followup Monitoring

More details on Sandy Lake re: section 4.2 Critique of the AECOM (2014) Predictions/No Followup Monitoring: 

(i) As cited  by AECOM (2014), modelled phosphorus concentrations differed by far more than 20% of the measured concentrations (it was 67% higher),  indicating the model is not valid, but they did not follow recommended procedures to revise the model.

From AECOM (2014) Appendix E, p.13:

The predicted phosphorus concentration in Sandy Lake under current conditions is 20 µg/L. This predicted value is significantly greater than the median measured phosphorus concentration of 12 µg/L. Brylinsky (2004) and MOE (2010) indicate that a model is not valid if the modeled phosphorus concentrations differ by more than 20% of the measured concentrations, as is the case with the Sandy Lake results.

As discussed in Section 3, the model assumptions may over-predict phosphorus concentrations by not factoring in phosphorus retention as water travels within the watershed before it reaches the lakes. Predicted phosphorus inputs to the lakes can be reduced by adjusting the lake phosphorus retention coefficient, which in turn reduces the predicted lake phosphorus concentration.

In this study, the phosphorus retention coefficient was adjusted from 0.33 to 0.6 in the LCM to reduce the predicted phosphorus concentration to match the measured phosphorus. 

Brylinski 2004, p. 38ff,  provides a detailed example of how model re-evaluation was conducted for a case  – Lake George in Kings Co. – in which “The model under predicts the lake’s phosphorus concentration by 21.9 % which is above the 20% difference generally considered acceptable for model validation.”  None of the steps for re-evaluation advised by Brylinski 2004 were conducted by AECOM 2014 even though the predicted value exceeded the observed value by a much larger factor (67%); rather AECOM 2014 made the model work by “adjusting the lake phosphorus retention coefficient… from 0.33 to 0.6 in the LCM to reduce the predicted phosphorus concentration to match the measured phosphorus.” This is not a procedure cited by Brylinski 2004.

The  MOE 2010 document cited by AECOM likewise adopts a maximum of  20% difference between observed and predicted P values as acceptable. As in Brylinski (2004), the MOE (2010) document makes no mention of a one-stop book-keeping fix for the model as applied by AECOM 2014. So the predicted impacts of development  and of mitigative measures on lake Total P are highly hypothetical.

(ii) Setting the Water Quality Objective (WQO)  for Total P in Sandy Lake at 50% above the current value is not justified.

AECOM (2014) provides the following ratioanle for setting the WQO for Total P at 50% above the current value:

7.1 Development of Total Phosphorus Water Quality Objectives (WQO)

For the Sandy Lake watershed AECOM recommends the use of Environment Canada’s trophic status classification to set WQOs for total phosphorus.

As noted in section 1.2.1, an objective of the 2006 HRM Regional Plan is to “maintain the existing trophic status of our lakes and waterways”. This suggests that both Sandy and Marsh Lakes should be maintained in their current mesotrophic state and so the WQO (water Quality Objective) for total phosphorus should be the upper limit of the mesotrophic range, or 20 g/L. However, since both lakes are currently at the lower end of the mesotrophic range, considerable water quality degradation could occur before the lakes were at risk of exceeding such a WQO.

If the objective is to “maintain the existing trophic status of our lakes and waterways”, how can the  WQO be set at 50% higher than the current (circa 2012)  value?  According to this  interpretation of the statement “maintain the existing trophic status of our lakes and waterways”, no efforts would be made to reduce phosphorus levels in lakes that currently have phosphorus levels highly elevated above recorded historic levels such as Banook Lake. All that is being maintained is the classification of the lake as mesotrophic.

AECOM 2014  cites evidence showing that in its current state (at 12 ug P/L ), Sandy Lake is already seriously degraded. In 1979, Total P was  7 ug/L or less, well under the upper limit (10 ug/L ) for oligotrophy (the ‘clean’ state of a lake); the transition to mesotrophic status occurred circa 2001 (re AECOM Fig 9). AECOM (2014) cites data showing that  by 2008-2011 deep water Total P on some samplings was much higher than  surface water P, which they attributed to oxygen deprivation. From AECOM 2014:

Table 5 compares the phosphorus concentrations of shallow (epilimnion) to deep (hypolimnion) samples from three sampling events. Total phosphorus concentrations in the shallow surface (epilimnion) samples are less than in the deep (hypolimnion) samples in two of the three examples. Although the data are limited, this suggests that the deeper portions of Sandy Lake may be fully or partially oxygen- deprived during certain times of the year, a situation that may arise when decomposing organic matter consumes available oxygen at depth. This in turn promotes the release of phosphorus from lake sediments, which is recorded in the water samples.

In commenting that “that both Sandy and Marsh Lakes should be maintained in their current mesotrophic state and so the WQO for total phosphorus should be the upper limit of the mesotrophic range, or 20 g/L”, AECOM (2014) seems to have taken a cue from the MOE (Ontario Minister of Environment) 2010 document in which it is suggested that “If the model fails….a total phosphorus concentration of 20 µg/L will be used as the upper limit to protect against nuisance algal blooms.”

Total P is a well validated and widely accepted predictor of the general condition of temperate lakes, but it is also well recognized it does not predict the precise condition of a particular lake. For a set of Ontario lakes, Malot et al., 1992 observed that “lake morphometry exerts a large influence on  profiles and this influence is particularly evident in shallow (<20 m maximum depth) oligotrophic lakes” and that “Predictions of O2 profiles are sensitive to changes in TP concentrations, with all study lakes predicted to have severely O2-depleted hypolimnions by the end of summer at an epilimnetic TP of only 15 ug.L-1.”  At 21 m maximum depth, Sandy Lake could likewise be expected to very sensitive to Total P concentrations and indeed the evidence cited by AECOM 2014 suggests significant deterioration in hypolimetic oxygen at Total P of 12 ug.L-1.

(iii) Varying the phosphorus export coefficient, rather than increasing the lake retention coefficient would be a more realistic “fix” to make the model work and would likely increase the predicted impacts of development on Total P

The failure of the Phosphorus Model to meet the criterion of Brylinski (2004) and MOE (2010) that the predicted Total P values for existing conditions be within 20% of actual values and by  a large factor   means that the model as applied by AECOM 2014 cannot be regarded as a good description of how the system is working.

The model overestimated  Total P under existing conditions (when the report was written) by 67%, yet it is making predictions of changes of Total P  under different management scenarios of 1-4 ug/l, i.e of 8-33%.  The model was made to work by  increasing the lake P Retention Coefficient from 0.33 to 0.6, i.e. by assuming that more P is being trapped in sediments than initially assumed. However, as cited above there is actually evidence   that lake phosphorus retention has probably been reduced because of lowered oxygen in the hypolimnion in some years.

AECOM (2014) does not mention that evidence in relation to  the lake P Retention coefficient, but it does cite two possible sources of error (p 10 of Appendix E) that could have resulted in the overestimation of Total P  under existing conditions: (i) removal of phosphorous en route from the sources to the lake in the ground for systems on septic, and in watercourses or overland flow; and (ii) dilution by groundwater; these factors  are not included in the standard LCM model as presented by  Brylinski (2004).

The extent of dilution by groundwater would depend on the concentration of phosphorus in the ground water (likely to be less and possibly zero or much less than in surface waters), and the flow of groundwater relative to surface waters. They estimated the flow of groundwater to be 11% of total flow, so the error associated with this factor is likely to be low in comparison to the difference between predicted Total P values for existing conditions and actual values.

The absorption of septic P en route to the lake is also likely to be small compared to the overestimate of  Total P under existing conditions, as diverting both the Uplands Park Waste Water Treatment Facility effluent (or the input to the Facility) reduces lake total P by only 1 ug/L according to the AECOM 2014 model.

Removal of phosphorus as water flows through watercourses however could be very significant. From Brylinski (2004):

The model makes no allowance for the assimilation of phosphorus within upstream rivers or streams entering a lake, or for tributaries contained within a lake’s drainage basin. This is a potentially serious limitation if the model is used to determine the permissible level of development within the watershed of a lake that has effluents entering lakes located downstream. If a downstream lake exceeds a phosphorus objective, no upstream development would be allowed.

The retention of phosphorus in streams and rivers can result from settling of particulate phosphorus, sorption of dissolved phosphorus to stream sediments, chemical precipitation of phosphorus, and uptake of phosphorus by benthic algae and macrophytes (Wagner et al. 1996). Behrendt and Opitz (2000) carried out a number of studies in which it was found that as much as 20 to 40 % of the phosphorus load was retained within streams before reaching the receiving water body.

This limitation is acknowledged in Appendix E of the AECOM 2014 Report, p. 10 and it is commented that “The retention of phosphorus from water as it travels through the watershed can be approximated in the LCM by varying the phosphorus export coefficient as needed”.  However, that was not done, rather “the lake phosphorus retention factor (a variable that approximates the amount of phosphorus removed through sedimentation within a lake) can be used to approximate both phosphorus retention during transport and dilution by groundwater.”

There could be a large difference between the effects of development on lake Total P so estimated, and the effects if they were estimated with consideration of P removal in watercourses.  Currently, phosphorus from settled areas goes into watercourses fairly high up in the watercourses; much of the proposed development would be closer to the lake and thus bypass a lot of the watercourse, including the large wetland NIA1. This could result in a much higher proportion of the phosphorus from settled areas going into the lake than currently assumed.

(iv) Further monitoring, as strongly advised by AECOM 2014, has not been conducted.

One route Brylinski (2004) and MOE (2010) advised to revising an invalid model is to conduct more monitoring; AECOM 2014 was  explicit on the need for more monitoring. From page iii of the Executive Summary:

“The predictions from the phosphorus load model are consistent with observations of urbanization in other watersheds. However, the degree of influence of urbanization on water quality in Sandy Lake can only be approximated using the phosphorus load model because of limitations arising from assumptions and uncertainty in the application of the model. Therefore a robust water quality monitoring plan is proposed for the Sandy Lake watershed to provide a further assessment of current conditions and to evaluate the impacts of development on the water quality.”

Section 9 (p 42) in  AECOM (2014) provides specific recommendations for robust Water Quality Monitoring.

The AECOM Report was submitted  in 2014. In 2020, such  monitoring has still not been initiated, and, I am told,  a request for Secondary Planning in the study area of the AECOM report was submitted to HRM in the summer of 2020.

So we could be proceeding to secondary planning without any validation  of the AECOM model as advised by AECOM 2014, or revision of the model as advised by Brylinksi (2004).