IN THE MATTER
of the Resource Management Act 1991
AND
IN THE MATTER
of a Board of Inquiry appointed under section
149J of the Resource Management Act 1991 to
consider The New Zealand King Salmon Co.
Limited's private plan change requests to the
Marlborough Sounds Management Plan and
resource consent applications for marine farming
at nine sites located in the Marlborough Sounds
EXECUTIVE SUMMARY OF MARK HADFIELD
INTRODUCTION
1.
My name is Mark Hadfield. I am employed by the National Institute of Water &
Atmospheric Research Ltd. (NIWA) in the Marine Physics Group. My
qualifications and experience are set out in full in the peer review report.
2.
In the Decision of the Board with respect to applications for an independent peer
review of the applicant’s water column report and related calibration report on 29
June 2012, I was appointed to peer review the "modelling of the fate of nutrients
from the proposed salmon farms."
3.
I prepared, in conjunction with Dr Niall Broekhuizen, a peer review report entitled
Review of evidence submitted by Mr Ben Knight on behalf of New Zealand King
Salmon Co. Ltd.dated 31 July 2012.
4.
I have attended expert conferencing and have read the evidence of Drs Longdill
and Hartstein, and the rebuttal evidence of Mr Knight.
5.
I have read the Code of Conduct for Expert Witnesses in the Environment Court
Practice Note 2011 and I agree to abide by it. I confirm that the issues I
addressed are within my field of expertise. I have not omitted to consider material
facts known to me that might alter or detract from the opinions expressed by me.
SUMMARY
6.
This executive summary is structured around the two questions directed by the
Board of Inquiry that I answered in the peer review report.
Q1(a): Is the replication of hydrodynamics within the Sounds marine
environment, as described in the Calibration and Methodology Report, suitable
for the subsequent purpose of simulating the cumulative dispersion of proposed
farm wastes?
7.
In my opinion, there are several significant limitations in the model’s replication of
1
the hydrodynamics in the Marlborough Sounds marine environment . These
limitations increase the degree of uncertainty that must be attached to any
predictions generated from the model. However, all models have an inherent
degree of uncertainty. Whether or not the replication of hydrodynamics is suitable
for the subsequent purpose of simulating the cumulative dispersion of proposed
farm wastes depends on what level of uncertainty is acceptable. That depends on
the specific issue being addressed and the tolerances on the final result.
8.
The judgement of whether the level of uncertainty in the replication of the
Marlborough Sounds hydrodynamics is acceptable and is suitable for the
subsequent purpose of simulating the cumulative dispersion of proposed farm
wastes is outside my area of expertise. This issue is addressed in the executive
summary of Dr Broekhuizen.
1
The review I undertook addressed the hydrodynamic model as described in Cawthron Report No. 2185, attached as an appendix
to Mr Knight’s evidence.( Knight, B. (2012). Validation Report for Updated Hydrodynamic Models of the Marlborough Sounds.
Cawthron Report No. 2185. 53 pp plus appendices).
9.
For outer Pelorus Sound (Mr Knight’s Figure 8, upper panel) and outer Queen
Charlotte Sound (Mr Knight’s Figure 9, upper panel), I consider that the actual
surface tracer concentration values (if the experiment of releasing this tracer were
being carried out in the real world) could be larger than modelled by a factor of
two, but could not be larger than modelled by a factor of five. In estimating this
magnitude of uncertainty I take particular account of:
9.1
limitations in the model’s representation of vertical exchange and residual
currents;
9.2
uncertainties in the true bay-wide flushing rates; and
9.3
the fact that the modelled period may not be representative of the longer
term.
10.
For Port Gore (Mr Knight’s Figure 8), I consider that the uncertainty is larger than
for outer Pelorus Sound and outer Queen Charlotte Sound because of my
concerns about whether the hydrodynamic processes in Port Gore are simulated
correctly. I therefore consider that the actual surface tracer concentration values
in Port Gore could be larger than modelled by a factor of up to four.
11.
th
Regarding the 95 percentile modelled surface tracer concentration in Pelorus
Sound (Mr Knight’s Figure 11, lower panel) and Queen Charlotte Sound (Mr
Knight’s Figure 12, lower panel), I consider that the actual surface concentration
values could be larger than modelled by a factor of greater than two. This is
because of the possibility that the model is excessively diffusive in the vertical or
horizontal directions and because of concerns about the model’s residual
currents. I am unable to estimate an upper bound to the ratio between the actual
th
and modelled values, as measures of the 95 percentile are sensitive to any
temporal averaging applied to the model output. Any temporal averaging applied
to this model has not been documented.
12.
In my opinion, the model cannot be relied on to reproduce or forecast the day-today movement of wastes from the farm. This is because:
12.1 reproducing the day to day transport and vertical mixing of farm wastes is
a challenging task for a model; and
12.2 I have concerns about the model’s representation of residual currents
and vertical mixing.
Q1(b): Do you believe the model has adequately replicated relevant aspects
including tidal forcing, wind forcing, stratification, density currents, and residual
flows?
13.
In my opinion, there are several aspects of the hydrodynamics within the Sounds
marine environment that are not adequately replicated in the model:
13.1 The hydrodynamic model produces plausible density stratification and
time-mean flows, but only with an ad hoc adjustment to the vertical
mixing scheme. The need for this adjustment calls the model’s
representation of vertical exchange into question. This means that the
rate of vertical mixing in the model may be too large in some situations
and/or too small in others.
13.2 Conclusions cannot be drawn about the accuracy of the model’s tidal
currents from the harmonic tidal analysis presented in Mr Knight’s
evidence because that analysis has been done inappropriately. However,
given that the modelled surface elevations agree with LINZ data, I
consider it likely that the tidal currents in the model are approximately
correct.
13.3 In general, the velocity measurements against which the model has been
compared have been made in sites that are subject to effects from local
topography. This makes it difficult to interpret what the comparisons
between model and measurements mean for the accuracy of model’s
representation of the larger-scale flows in the Sounds.
13.4 The variability of depth-average residual currents on time scales of 1–5
days is not well reproduced. This suggests that the effect of wind on the
currents is not represented accurately in the model. This deficiency
reduces the accuracy of the model’s representation of the dispersion of
material from the marine farms for time scales of a few days or more and
distance scales of a few kilometres or more.
13.5 There is no assessment in the evidence relating to the variation of
residual currents with depth. In my opinion, this should have been done
because currents in the Marlborough Sounds are known to vary
significantly with depth, meaning that the direction in which material is
dispersed will depend on the depth at which it is released. The accuracy
with which the model represents these effects is unknown.
13.6 Results have been presented for 30-day period in winter and a 30-day
period in summer. The dispersion of material from the farm sites will be
affected by seasonal and inter-annual variations in wind and freshwater
input. Modelling just two 30-day periods insufficient to give a full picture of
the seasonal and inter-annual variations. This means that the extent to
which the modelled period is typical of the long-term mean is unknown.
14.
Despite these concerns about the accuracy of the spatially explicit model, it is still
likely to produce a substantially more accurate picture than the flushed aspatial
model.
Q2(a): Are the limitations of the hydrodynamic modelling adequately described
and quantified where appropriate?
15.
In general, the limitations of the hydrodynamic modelling are not adequately
described and quantified. Some of the limitations listed above are recognised in
Mr Knight’s evidence, but some are not. In general, very little attempt has been
made to quantify the limitations and their effects on the model results. In
particular:
15.1 The need for an ad hoc adjustment to the vertical mixing scheme is
mentioned but its implications are not discussed.
15.2 The issue with the harmonic tidal analysis was not recognised.
15.3 The need for an ad hoc adjustment to the wind speed is mentioned, but
the potential effect of this on the model’s residual currents is not
discussed. The fact that the depth-averaged residual currents are of
approximately the correct magnitude but do not have the observed
variation in time is recognised but its implications are not discussed.
15.4 It is noted that the model produces plausible time-mean surface and
bottom currents, but otherwise the accuracy of the model’s residual
currents is not addressed, except in the depth-averaged sense.
Q2(b): Can you detail any additional limitations which should be placed on its
subsequent use to determine and quantify water column impacts from the
proposed farms?
16.
In my comments above, I have identified what I consider to be the major
limitations of the hydrodynamic modelling and estimated the degree of
uncertainty that I consider they introduce into its predictions.
17.
These limitations could be reduced significantly, but not eliminated, in future
model applications through:
17.1 improved wind forcing (however this may be computationally
challenging);
17.2 improving the representation of vertical exchange by increasing the
number of layers;
17.3 validation against measurements of currents in the main channels;
17.4 validation against temperature, salinity and nutrient data, with a view to
assessing the model’s representation of vertical exchange and
stratification; and
17.5 simulations from different times of year.
Mark Hadfield
23 August 2012