Sediment Plume Modelling

(1)

Sediment Plume Modelling

Prepared for Trans-Tasman Resources Ltd October 2015

(2)

© All rights reserved. This publication may not be reproduced or copied in any form without the permission of the copyright owner(s). Such permission is only to be given in accordance with the terms of the client’s contract with NIWA. This copyright extends to all forms of copying and any storage of material in any kind of information retrieval system.

Whilst NIWA has used all reasonable endeavours to ensure that the information contained in this document is accurate, NIWA does not give any express or implied warranty as to the completeness of the information contained herein, or that it will be suitable for any purpose(s) other than those specifically contemplated during the Project or agreed by NIWA and the Client.

Prepared by:

Mark Hadfield Helen Macdonald

For any information regarding this report please contact:

Mark Hadfield

Marine Physics Modeller Marine Physics

+64-4-386 0363

[email protected]

National Institute of Water & Atmospheric Research Ltd Private Bag 14901

Kilbirnie

Wellington 6241 Phone +64 4 386 0300

NIWA CLIENT REPORT No: WLG2015-22 Report date: October 2015 NIWA Project: TTR16301

Quality Assurance Statement

Dr Graham Rickard Reviewed by:

P Allen Formatting checked by:

Dr Alison MacDiarmid Approved for release by:

(3)

Sediment Plume Modelling

Figure 2-3: Time-averaged, depth-average velocity from the Cook Strait model. 18 Figure 2-4: Time-averaged, depth-average velocity from the inner (sediment) model. 18 Figure 2-5: Bottom boundary layer & sediment bed time series at instrument site 7. 21 Figure 2-6: Daily average SSC vs flow for Whanganui River. 23 Figure 2-7: Near-bottom speeds (m s−1) on the inner model domain. 24 Figure 3-1: M2 tidal velocity comparison (ADCP Site 7 Deployment 2). 31 Figure 3-2: M2 tidal current profile comparison (ADCP Site 7 Deployment 2). 32 Figure 3-3: Tidal velocity comparison for S2, N2 and K1 (ADCP Site 7 Deployment 2). 33 Figure 3-4: Sub-tidal velocity comparison (ADCP Site 7 Deployment 2). 35 Figure 3-5: Comparison of detiding methods (ADCP Site 7 Deployment 2). 36 Figure 4-1: SSC time series at near-shore site 11 (Whanganui). 39 Figure 4-2: SSC time series at near-shore site 12 (Kai Iwi). 39 Figure 4-3: SSC time series at near-shore site 13 (Waitotara River). 39 Figure 4-4: SSC time series at near-shore site 14 (Patea). 40 Figure 4-5: SSC time series at near-shore site 15 (Manawapou). 40

(5)

Figure 4-6: SSC time series at near-shore site 16 (Ohawe). 40 Figure 4-7: Modelled and observed 5th percentile surface concentration of background

sediment. 42

Figure 4-8: Modelled and observed median (50th percentile) surface concentration of

background sediment. 43

Figure 4-9: Modelled and observed 95th percentile surface concentration of background

sediment. 43

Figure 4-10: Time series of near-bottom sand concentration at instrument site 6. 45 Figure 4-11: Time series of near-bottom sand concentration at instrument site 7. 45 Figure 4-12: Time series of near-bottom sand concentration at instrument site 8. 45 Figure 4-13: Time series of near-bottom sand concentration at instrument site 10. 46 Figure 5-1: Surface plume Case 1 (suspended source at location A). 48 Figure 5-2: Vertical structure of the plume in Case 1 (suspended source at location A). 49 Figure 5-3: Surface plume and vertical transect for Case 2 (suspended source at location

A). 50

Figure 5-4: Surface plume and vertical transect for Case 3 (suspended source at location

A). 51

A). 52

A). 53

A). 54

Figure 5-8: Median near-surface concentration of suspended sediment from mining at

source location A. 57

Figure 5-9: 99th percentile near-surface concentration of suspended sediment from

mining (50 Mt/a) at source location A. 58

Figure 5-10: Median near-bottom concentration of suspended sediment from mining

(50 Mt/a) at source location A. 59

Figure 5-11: 99th percentile near-bottom concentration of suspended sediment from

mining (50 Mt/a) at source location A. 60

Figure 5-12: Near-source statistics of mining derived sediment concentration from mining

source A. 61

Figure 5-13: Median summer (December–February) near-surface concentration of

suspended sediment from mining (50 Mt/a) at source location A. 63 Figure 5-14: Median winter (July–August) near-surface concentration of suspended

sediment from mining (50 Mt/a) at source location A. 64 Figure 5-15: Maximum 5-day increment in sediment bed thickness for suspended sediment

from mining source A. 66

Figure 5-16: Maximum 365-day increment in sediment bed thickness for suspended

sediment from mining source A. 67

Figure 5-17: Near-source maximum increment of mining derived sediment, source A. 68 Figure 5-18: 2-year increment in sediment bed thickness for suspended sediment from

mining source A. 69

B). 71

(6)

Sediment Plume Modelling Figure 5-20: Surface plume and vertical transect for Case 2 (suspended source at location

B). 72

B). 73

B). 74

B). 75

Figure 5-24: Surface plume and vertical transect for case61 (suspended source at location

B). 76

Figure 5-25: Median near-surface concentration of suspended sediment from mining at

source location B. 78

Figure 5-26: 99th percentile near-surface concentration of suspended sediment from

mining at source location B. 79

Figure 5-27: Median near-bottom concentration of suspended sediment from mining at

source location B. 80

Figure 5-28: 99th percentile near-bottom concentration of suspended sediment from

mining at source location B. 81

Figure 5-29: Near-source statistics of mining derived sediment concentration from mining

source B. 82

Figure 5-30: Maximum 5-day increment in sediment bed thickness for suspended sediment

from mining source B. 84

Figure 5-31: Maximum 365-day increment in sediment bed thickness for suspended

sediment from mining source B. 85

Figure 5-32: Near-source maximum increment of mining derived sediment, source B. 86 Figure 5-33: Thickness of fine sediment from the patch source. 88 Figure 5-34: 99th percentile of near-surface SSC of fine sediments from the patch. 89 Figure 5-35: 99th percentile of near-bottom SSC of fine sediments from the patch. 90

Figure A-1: ROMS vertical grid schematic. 96

Figure B-1: Tidal ellipse comparison (all ADCP deployments). 97 Figure B-2: Tidal profile comparison (all ADCP deployments). 98 Figure B-3: Sub-tidal variance ellipse comparison (all ADCP deployments). 100 Figure B-4: Sub-tidal along-axis velocity comparison (all ADCP deployments). 101 Figure B-5: Sub-tidal across-axis velocity comparison (all ADCP deployments). 102 Figure C-1: Comparison of median near-surface SSC due to mining at source location A. 106 Figure C-2: Comparison of 99th percentile near-surface SSC due to mining at source

location A. 107

Figure C-3: Comparison of median near-bottom SSC due to mining at source location

A. 108

Figure C-4: Comparison of 99th percentile near-bottom SSC due to mining at source

location A. 109

Figure C-5: Comparison of median near-surface SSC due to mining at source location B. 110 Figure C-6: Comparison of 99th percentile near-surface SSC due to mining at source

location B. 111

(7)

Figure C-7: Comparison of median near-bottom SSC due to mining at source location

B. 112

Figure C-8: Comparison of 99th percentile near-bottom SSC due to mining at source

location B. 113

Figure C-9: Maximum 5-day increment in sediment bed thickness for suspended sediment

due to mining at source location A. 114

Figure C-10: Maximum 365-day increment in sediment bed thickness for suspended

sediment due to mining at source location A. 115

Figure C-11:: Maximum 5-day increment in sediment bed thickness for suspended sediment

due to mining at source location B. 116

Figure C-12: Maximum 365-day increment in sediment bed thickness for suspended

sediment due to mining at source location B. 117

(8)

8 Sediment Plume Modelling

Executive summary

Trans-Tasman Resources Ltd (TTR) proposes to extract titanomagnetite sand (ironsand) from an area in South Taranaki Bight. NIWA was commissioned by TTR to investigate potential environmental impacts of the proposed extraction operation. The operation will result in suspended-sediment plumes and sediment deposition on the seabed. It is recognised that this will have an impact on the ocean environment.

Following the refusal of consent for an earlier application in June 2014, TTR re-assessed their scientific case as background for a second application. A detailed review and a subsequent test program by HR Wallingford Ltd (HRW) has allowed for more accurate modelling of the plume by addressing the following aspects:

 Flocculation: The original plume model neglected flocculation, a process in which fine sediment particles combine into fast-sinking aggregates, called flocs;

 Sediment settling rates: The extent to which the fine suspended sediment would settle to the bottom and be trapped in the matrix of discharged sand has been reviewed by HR Wallingford and is predicted to occur to a greater extent than assumed previously.

 Sediment resuspension: The HR Wallingford tests found that the shear stress required for resuspension of freshly deposited material was in the range 0.2–0.3 Pa rather than the 0.1 Pa (minimum value) assumed by NIWA.

The output of sediment from the ironsand extraction operation is represented with two sources:

 The suspended source, representing fine sediment (grain size < 63 µm) introduced into suspension via two discharge streams: the overflow from the hydro-cyclone

dewatering system and the de-ored sand discharge;

 The patch source, an area of 3 × 2 km representing one year’s discharge of de-ored sand, including trapped fine material.

Of these, the suspended source has by far the greater impact in terms of the extent and magnitude of the concentrations in the sediment plume.

Two source locations are considered, at the inner end (A) and the outer end (B) of the project area.

These two points represent the limits of inshore and offshore mining locations within the proposed project area.

The suspended source was introduced in a simulation of 1000 days duration, with the source operating for the final 800 days (with 20% down-time) and statistics calculated over the final two years. The analysis of suspended sediment concentrations (SSC) focussed on the median and 99th percentile, comparing values for background sediments, mining-derived sediments and the combination of the two.

Plumes from the suspended source generally extend to the east-southeast as a result of the prevailing winds and residual currents. Occasionally the plume will pool around the mining areas or move towards the west or south in response to changed wind patterns which affect the prevailing currents. The envelope of the area predicted to be impacted by the plume over the course of the simulated release has been established. The plume envelope from the inshore source, location A, shows that the plume influences the coast between Patea and Whanganui with very low

(9)

Sediment Plume Modelling 9 concentrations, substantially (around 100 times) less than the naturally occurring background

concentrations, following the coast towards Kapiti. The highest surface concentrations associated with the plume occur at the source location and are 1.45 mg/L (median) and 8.2 mg/L (99^th

percentile). At 20 km downstream from the source the surface concentrations reduce to 0.35 mg/L (median) and 2.8 mg/L (99th percentile). The plume envelope for the offshore source location B is located further offshore but follows a similar path to the east-southeast, with the concentrations being significantly lower than for source location A. In both cases the plume of mining-derived sediment contributes noticeably to the total SSC within a few kilometres of the source but is insignificant relative to the background SSC near the coast.

An analysis of mining-derived and background SSCs for the suspended source at location A in summer (December–February) and winter (July–August) indicates that both mining-derived and background concentrations are lower in summer than winter. The net effect is that the mining- derived plume is somewhat more pronounced relative to the background in summer than in winter.

Deposition from the suspended source has been characterised by two statistics, the maximum 5-day deposition (i.e. the maximum amount of material predicted to accumulate over any 5-day interval) and the maximum 365-day deposition. As defined by these statistics, the deposition footprint of mining-derived sediment is widespread but at very low values of 0.01–0.05 mm, i.e. one-tenth the thickness of a human hair (typically 0.1 mm). The deposition of mining derived sediment would only be able to be distinguished from the background within a few kilometres of the source.

With the patch source, fine sediments trapped in the pit are eroded, transported and deposited, but only at a low rate, forming a deposition footprint (> 0.01 mm) that extends up to 10 km from the patch boundary after two years. Suspended sediment concentrations in the associated plume are small relative to background SSCs.

(10)

1 Introduction

Trans-Tasman Resources Ltd (TTR) proposes to extract titanomagnetite sand (ironsand) from an area in South Taranaki Bight (Figure 1-1). The plan defines a project area in a roughly triangular region at 20–40 m depth off the South Taranaki coast near Patea. On hydrographic charts parts of this area are labelled Whenuakura Spur, Graham Bank, Patea Banks and The Rolling Ground. Here the area as a whole will be called the Patea Shoals.

As input to the Environmental Impact Assessment (EIA) for the proposal NIWA was commissioned by TTR to investigate the potential impacts of the proposed extraction operation. The set of sediment plume model results presented to the Decision Making Committee (as revised in March 2014) will be called the March 2014 configuration.

Following the refusal of consent by the Decision Making Committee in June 2014, TTR re-assessed their scientific case as background for a second application. One issue that arose was the need to provide more certainty and accuracy in the sediment plume modelling studies and the interpretation of these results. In July 2014 HR Wallingford undertook a structured review of the NIWA sediment plume modelling work which included a detailed assessment of the assumptions and inputs.

The March 2014 NIWA plume modelling assumed that the material discharged into the sea would remain in its particulate (unflocculated) form. The HR Wallingford tests indicated that most of the fine sediment in the tailings would exist in the environment in flocculated form and would therefore settle from the upper part of the water column more quickly than assumed in the NIWA sediment plume modelling. The HR Wallingford tests also found that the shear stress for resuspension of freshly deposited material was in the range 0.2–0.3 Pa rather than the 0.1 Pa (minimum value) as assumed by NIWA. HR Wallingford also addressed the trapping of mining-derived fine sediment in the matrix of coarser tailings in the mining pit, using a high resolution 3D flow and sediment transport model.

The present report presents a more accurate sediment plume model incorporating the confirmed sediment settling parameters and source rates from the HR Wallingford work (HR Wallingford 2015).

(11)

Sediment Plume Modelling 11 Figure 1-1: Location map.The proposed mining area is shaded red. The solid black rectangle outlines the inner model domain (Cape Egmont to Kapiti) on which the sediment model was set up, as described in Section 2. The orange line indicates the boundary of the territorial sea. Also shown are the towns and villages and the mouths of the principal rivers along the Taranaki and Manawatu coasts. Stars indicate named offshore oil/gas production platforms.

1.1 Oceanographic conditions

The movement of sediments in the vicinity of the proposed mining area is heavily influenced by physical conditions. This section briefly describes the background physical conditions that influence water and sediment movement. More detailed descriptions can be found in earlier reports

(MacDiarmid et al. 2010; MacDonald et al. 2012).

Tidal currents are typically strong in this area, due to the difference in tidal phase between the western and eastern ends of Greater Cook Strait Figure 1-2 shows tidal velocities from the NIWA tidal model (Goring 2001; Goring et al. 1997; Walters et al. 2001). The tidal flow amplitude exceeds 1 m/s in Cook Strait Narrows. Over Patea Shoals the tidal flows are around 0.4 m/s and are largely parallel to the shore.

(12)

12 Sediment Plume Modelling Figure 1-2: Peak velocity of the net tidal current The maximum speed is shown by the colour scale, while maximum and minimum velocity vectors are shown by the longer and shorter of the crossed arrows,

respectively. Figure from MacDiarmid et al. 2010.

The wave energy flux across the 50 m isobath from a 20 year hindcast (Gorman et al. 2003a, Gorman et al. 2003b) is shown in Figure 1-3. The prevailing wave swell direction tends to be from the south- west; as such, the wave energy flux is typically higher in the north-west part of the domain than the more sheltered south-east. The wave energy fluxes are not normal to the coastline in the domain, with the shore-parallel flux typically toward the south-east along the northerly part of the shoreline, while the flux has a northerly component along the eastern shoreline. This flux influences the distribution of sediment. The significant wave height statistics from the hindcast show that the north of the region has a mean peak at around 1.5 m, but that heights in excess of 8 m also arise, especially in the winter months during storm events. In the relatively sheltered eastern part of the domain, the mean heights are around 1 m, and the maximum wave heights are generally less than 6 m or so.

Wave periods are typically in the range 10–14 s.

A wind rose for 8 years of observations at Hawera between January 2004 and July 2012 is shown in Figure 1-4. Winds are predominantly from the north, southeast and west. The mean wind speed is 5.3 m/s and the maximum over the period was 21.1 m/s on 15 February 2004.

(13)

Sediment Plume Modelling 13 Figure 1-3: Spatial distribution of mean wave energy flux. This is the distribution of the flux along the 50 m isobath averaged over the full 20-year hindcast record. The colour scale shows the mean of the magnitude of the energy flux, while the arrows show the vector averaged flux. Figure from MacDiarmid et al. 2010.

Figure 1-4: Wind rose from measurements at Hawera over a period of 8 years (January 2004 to July 2012).

Meteorological convention is used in expressing the direction that the wind blows from. Figure from MacDonald et al. 2012.

(14)

2 Model setup 2.1 Nested grids

Sediment plume behaviour was predicted using a modelling system comprising a set of nested domains. The outer domain covered Greater Cook Strait (Figure 2-1). Two different inner domains were defined and used in different simulations, a larger domain extending from Cape Egmont to just north of Kapiti Island and another covering a smaller area over Patea Shoals. The model was ROMS (Haidvogel et al. 2008), a widely accepted ocean/coastal model with optional embedded models of suspended-sediment and sediment-bed processes (Warner et al. 2008). In all cases the inner domain is the one on which sediment processes are simulated.

In the model nesting procedure, the outer domain model provides time-varying lateral boundary data (temperature, salinity, velocity, sea-surface height) for the inner domain models. The simulations are carried out separately, with output fields from the outer model saved to files and later post-processed to provide boundary-data files on the inner grid. This process is called one-way, off-line nesting.

The model grids are shown in Figure 2-1.The outer grid covers Greater Cook Strait at 2 km resolution.

The inner grids have been implemented at two different resolutions: 1 km and 500 m. The model runs described in this report were carried out on the 1 km grids, with the 500 m grids used in the past to investigate the sensitivity of the model results to the grid resolution.

The bathymetry for the model grids was constructed using several different datasets, combined and gridded with the GMT mapping tools¹. The primary dataset was a bathymetry on a 100 m grid generated by NIWA (A. Pallentin and R. Gorman, pers. comm. 13 June 2013) and incorporating data from Patea Shoals surveys conducted by TTR and NIWA. This dataset was supplemented with several lower-resolution regional datasets, namely coastline data, land elevation data, continental shelf bathymetric contour data and the GEBCO² gridded ocean bathymetry.

2.2 Outer (Cook Strait) model

The Cook Strait model requires lateral boundary data to generate a realistic flow from west to east through Cook Strait, with the inflowing water having realistic temperature and salinity. The east-west flow is called the D’Urville Current (“DC” in Figure 2-2) and is a robust feature of ocean models. It is driven by the difference in density (and consequently mean sea level) between Tasman Sea and Pacific Ocean waters.

1 http://gmt.soest.hawaii.edu/

2 http://www.gebco.net/

(15)

Sediment Plume Modelling 15 a)

b

Figure 2-1: ROMS model domains. a) Outer (Cook Strait) and inner (Cape Egmont to Kapiti; Patea Shoals) domains; b) Inner (Cape Egmont to Kapiti) domain with bathymetry (coloured surface), coastline (yellow), 22.2 km territorial limit (thin white line), project area (thick white line), ADCP sites (dark blue) as described in Section 3.1, river locations (blue) and towns (black).

For the present work two different sources of boundary data for the outer model were tested. One was an application of ROMS to the New Zealand EEZ. The other was a global ocean analysis and prediction system operated by the US Naval Research Laboratory, using the HYCOM³ ocean model.

(The specific dataset used here is called Glba08.) The HYCOM system provides daily snapshots of the three-dimensional state of the global ocean on a 1/12° grid; at NIWA we have archived a subset of this data around New Zealand. The tests indicated that estimates of currents, plume dispersion and transport in Cook Strait were not sensitive to the source of boundary data. All the simulations described in this model use HYCOM.

3 http://hycom.org/

(16)

16 Sediment Plume Modelling Figure 2-2: Surface circulation around New Zealand. Excerpt from “Ocean Circulation New Zealand” (Carter et al. 1998).

Model times are expressed in days relative to a reference time of 00:00 UTC on 1 January 2005. The model was initialised at 2000 days and run to 3000 days. Analyses of statistical quantities (means, percentiles) in the remainder of this report are generally based on the last 730 days of the simulation (21 March 2011 to 20 March 2013), allowing the first 270 days of the simulation for the model to approach equilibrium. (However, note that different aspects of the model approach equilibrium at different rates. Currents adjust within days to weeks, and temperature and salinity within months, but deposited sediments move slowly through the system and do not reach equilibrium within a period of several years.)

In both models the heat flux through the sea surface was calculated using data (6-hourly averages) from a global atmospheric analysis system called the NCEP Reanalysis (Kalnay et al. 1996), with a heat flux correction term that causes the model sea surface temperature (SST) to be nudged towards observed SST (the NOAA Optimum Interpolation 1/4° daily SST dataset, Reynolds et al. 2007). The heat flux correction prevents the modelled SST from departing too far from reality due to any biases in the surface fluxes, but has a negligible effect on day-to-day variability. The surface stress was calculated from 3-hourly winds from the NZLAM 12 km regional atmospheric model⁴. The standard formula relating wind speed to surface stress involves a wind-speed-dependent term called the drag coefficient. For the present work this was calculated by the method of Smith (1988), however a comparison of preliminary model results with measurements indicated that wind-driven variability in

4 NZLAM is part of the NIWA Ecoconnect environmental forecasting system: http://EcoConnect.niwa.co.nz/

(17)

Sediment Plume Modelling 17 the model was generally too low. The drag coefficient was therefore multiplied by a factor that was adjusted to optimise agreement: the final value chosen was 1.4. A similar adjustment has been found to be necessary in previous coastal modelling exercises around New Zealand by us (Hadfield and Zeldis 2012) and others (e.g. P. McComb pers. comm.). The need for this adjustment may indicate that the NZLAM wind speeds are biased low and/or that the Smith (1988) drag coefficient formula gives results that are too low for the wind and wave conditions in coastal areas.

The purpose of the Cook Strait model is to support a reasonably accurate description of the bathymetry of Cook Strait and, with it, the paths of currents through the strait.

To illustrate this point Figure 2-3 shows depth-average velocity vectors averaged over two years of a Cook Strait model run and Figure 2-4 shows similar data from the inner model. A continuous current can be seen entering Cook Strait from the south along the Kahurangi coast, then crossing the strait at its shallower, western end. The path of the current then follows the 50–100 m depth band to the South Taranaki coast, skirting the Patea Shoals. From there it follows the coast south past Manawatu, Horowhenua and Kapiti, through the Narrows and then northward along the Wairarapa coast. The existence of this current system has been known for several decades, but the details of its spatial pattern and temporal variability were not previously well described. An accurate representation of this current and its variability is important because it is expected to have a major influence on sediment plumes from the proposed ironsand extraction operation.

(18)

18 Sediment Plume Modelling Figure 2-3: Time-averaged, depth-average velocity from the Cook Strait model. Velocity vectors are averaged over 730 days and shown at every 4th grid point. Depth contours are at 50, 100, 250, 500, 1000 and 2000 m.

Figure 2-4: Time-averaged, depth-average velocity from the inner (sediment) model. Velocity vectors are averaged over 730 days and shown at every 4th grid point. Depth contours are at 10, 25, 50, 75 and 100 m.

2.3 Inner (sediment) model

The inner model of the Cape Egmont to Kapiti domain was used to generate the majority of the results in this report. It was forced at the lateral boundaries by data from the Cook Strait model at an interval of 3 hours. The inner model had similar surface forcing and dynamics to the Cook Strait model, but with the addition of several processes:

(19)

Sediment Plume Modelling 19

 The ROMS sediment module was activated, representing both natural sediments and plumes generated by the project. More detailed information about the sediment module and associated parameterisations is given in Section 2.4 and the sediment properties are described in Section 2.7.

 Tidal forcing was applied at the boundaries. Amplitude and phase data for 13 tidal constituents were interpolated from the output of the NIWA EEZ tidal model (Walters et al. 2001). The ROMS tidal forcing scheme then calculated tidal sea surface height and depth-averaged velocity at each time step and added them to the lateral boundary data from the Cook Strait model. Applying tides in addition to the lateral boundary data from an outer model in this way means that the outer model fields do not need to be stored at a high temporal resolution.

 A bottom boundary layer scheme was activated (the Sherwood/Signell/Warner variant, see Warner et al. 2008). This scheme takes account of waves in calculating bottom drag and sediment erosion/deposition and requires data on the wave orbital velocity near the bottom.

 The larger rivers that drain into South Taranaki Bight were represented as point sources of freshwater and suspended sediment, as described in Section2.5.

The inner model of the Patea Shoals domain was used in simulations where the emphasis was on short-range dispersion (Sections 5.3). In its setup and forcing it was very similar to the Cape Egmont to Kapiti model, except that there was no river input.

2.4 Sediment model setup

The ROMS parameterisations for sediment and associated quantities (eg. bottom boundary layer, wave forcing) that were implemented in the inner model are as described by Warner (2008). The model accepts a series of user-defined sediment classes, each characterised by several properties such as grain size, grain density and settling velocity. Typically a continuous distribution of sediment particle sizes in reality is represented in the model by several sediment classes, each with a different grain size. A description of the sediment classes and their properties as used for the present report is given in Sections 2.8 and 2.9.

At the bottom of the water column there is a multi-layer sediment bed, each layer being composed of a mixture of the sediment classes. The top bed layer exchanges sediment with vertically with the lowest level in the water column and horizontally with its immediate neighbours (a process called bedload transport).

The thickness of the bed layers is adjusted at each model time step according to the scheme described by Warner (2008), within any changes resulting in mass-conservative exchange between the layers. At the beginning of each time step an active layer thickness is calculated (Warner et al.

2008, Equation 21; Harris and Wiberg 1997, Equation 1). The active layer thickness sets a minimum for the top bed layer thickness, meaning that sediment is immediately mixed over this depth, and this is important in regulating the availability of fine sediment in the bed for erosion. It is important to note that the Harris and Wiberg relationship for active layer thickness was developed for a location dominated by biogenic roughness rather than transport-induced bed forms.

(20)

20 Sediment Plume Modelling For the simulations described below the total sediment bed thickness was set initially to 1 m, with eight layers. Initial layer thickness was 0.125 m, but this adjusts after the first time step. Incidentally, the ROMS model optionally allows for the depth at the base of the water column to be adjusted as the total sediment bed thickness changes, but this facility was turned off for the present simulations.

Vertical exchange of sediment between the top bed layer and the water column is the sum of two terms (Warner 2008, Equation 22). The first is deposition: it occurs continuously and is calculated separately for each sediment class as the product of the near-bottom concentration and the settling velocity. The second is erosion (Warner 2008, Equation 23; Ariathurai and Arulanandan 1978): it occurs only when the bottom stress exceeds a critical value, user-specified for each class.

Bedload transport (Warner et al. 2008 Section 3.4) is optionally calculated with the Soulsby and Damgaard (2005) formulation. This process has been included in only one of the simulations described below (the patch source simulation, Section 5.3) where it results in a modest increase, around 20%, in the rate at which medium sands are transported out of the patch area, relative to a simulation with bedload transport excluded.

Bottom stress is calculated with the Sherwood, Signell and Warner bottom boundary layer

formulation (Warner et al. 2008, Section 3.7). This requires data on the height, period and direction of surface wind waves. Two sources were considered: the NZWAVE wave forecasting model⁵ and dedicated runs of the SWAN wave model (R. Gorman, pers. comm). Another choice to be made in the model setup was the bottom roughness formulation. A bottom roughness length is calculated from the median grain density of the sediment bed (Warner et al. 2008, Equation 44). This changes as the composition of the sediment bed evolves but for the simulations below it was typically ~0.4 mm.

Additional terms account for sediment bedload transport (Warner et al. 2008, Equation 45) and for bedform ripples (Warner et al. 2008, Equation 46). The roughness length associated with bedload transport was negligible in these simulations. The bedform roughness length varied in space and time but was generally largest at around 20–30 m depth, where it was 0.5–2.5 mm. All these terms can be enabled or disabled in the ROMS code. As part of the calibration process, several simulations were carried out with different combinations of forcings and processes, the calibration target in this case being measured SSC data (Section 4). The model configuration finally selected used NZWAVE wave data and neglected the bedform roughness length. The main problem with the latter was that, when it was included, the model was not able to reproduce the isolated spikes that are seen by the ABS in near-bottom SSC (Section 4.3).

Figure 2-5 presents several time series relevant to erosion and deposition processes. The location is in a water depth of 31 m depth on outer Patea Shoals and appears elsewhere in this report as mining site A. It is characterised by strong tidal currents, moderately large wave amplitudes and complex bedforms (Section 4.3 this report; MacDonald et al. 2012, Figure 3-56). It is one of 24 locations in the model where hourly data have been saved to allow evaluation of short-timescale phenomena like tides.

5 NZWAVE is part of the NIWA Ecoconnect environmental forecasting system: http://EcoConnect.niwa.co.nz/

(21)

Sediment Plume Modelling 21 a)

b)

c)

Figure 2-5: Bottom boundary layer & sediment bed time series at instrument site 7. a) Near-bottom current speed; b) bottom wave orbital speed; c) sediment bed active layer thickness.

Panel a shows the near-bottom current speed. Values are generally less than 0.4 m/s, with occasional excursions as high as 0.75 m/s. Panel b shows the near-bottom wave orbital velocity, calculated within ROMS from NZWAVE wane height and period data. The wave orbital velocity is frequently higher than the current speed (note the different vertical scale) with several excursions above 1 m/s.

This indicates that waves will frequently be more effective in lifting sediment from the seabed than

(22)

22 Sediment Plume Modelling currents. Panel c shows the active layer thickness calculated by ROMS. It is generally less than

10 mm, but occasionally exceeds 40 mm, with all the peaks coinciding with wave events. The figure suggests that erosion of bottom sediments here (and in shallower water) tends to occur in occasional high-wave events and that in the model these mix the upper sediment bed at this location to a depth of at least 40 mm.

2.5 River inputs

Data for some 40 rivers were extracted from the NIWA Water Resources Explorer (WRENZ) website⁶ (Hicks et al. 2011) and 11 were selected for inclusion in the model based on their mean flow rate and sediment input rate (Table 2-1). The rivers that were selected comprised the ten with the highest flow rate, plus the Tangahoe River, which ranks 13th for flow but 9th for sediment input.

The model was supplied with a time series of daily-average flow rate and sediment input for each river. The flow rate was estimated from gauging station data collected by NIWA, Taranaki Regional Council and Horizons Regional Council. The gauging station data was available for all rivers except the Tangahoe, for which the WRENZ mean flow was used throughout. Where the gauging station was well upstream from the coast, the data were scaled by the ratio between the catchment area above the gauging station and the total catchment area. For several rivers, data were not available after Dec 2012; these gaps, plus a few short gaps elsewhere in the record, were filled with the WRENZ mean flow.

Table 2-1: Rivers represented in the inner model, with mean freshwater and sediment input rates from WRENZ.

Name Flow rate

(m³/s)

Sediment rate (kg/s)

Whanganui River 229.0 149.03

Manawatu River 129.5 118.46

Rangitikei River 76.4 35.04

Whangaehu River 47.2 21.82

Patea River 30.4 9.85

Waitotara River 23.3 15.08

Otaki River 30.1 5.46

Whenuakura River 9.9 8.75

Kaupokonui Stream 8.6 0.31

Turakina River 8.4 9.54

Tangahoe 4.2 1.39

Total 593 373

For the Whanganui River, a time series of SSC was available up to December 2012 from Horizons Regional Council. A relationship between SSC and flow was established from the data before this time (Figure 2-6) and used to fill the remainder of the SSC time series. (No filling of the flow time series was required, as a complete flow dataset was available.)

6 http://wrenz.niwa.co.nz

(23)

Sediment Plume Modelling 23 Figure 2-6: Daily average SSC vs flow for Whanganui River. Linear regression fit indicated by a red line.

For all rivers other than the Whanganui, SSC data was unavailable. Instead an SSC time series was constructed on the assumption that SSC is directly proportional to flow (as is approximately the case for the Whanganui), with a constant of proportionality adjusted so that the mean sediment input matched the WRENZ value.

The Patea Shoals is clearly a high-energy environment. This point is illustrated in Figure 2-7 by plots of near-bottom speeds associated with three different processes: non-tidal currents, tidal currents and surface wind waves. The non-tidal speed (Figure 2-7a) is evaluated from 90 days of a simulation with no tidal forcing and is largest (with a peak of ~ 0.15 m s⁻¹), along the path of current skirting the south of Patea Shoals. The tidal speed (Figure 2-7b) is evaluated by a harmonic analysis of a run with tidal forcing and represents the speed at maximum ebb or flood at a tidal amplitude halfway

between spring and neap; this is largest (with a peak of ~ 0.35 m s⁻¹), on the top of Patea Shoals; the wave orbital velocity is a mean from 90 days of the winter simulation; its highest values (greater than 1 m s⁻¹) occur in the shallow water near the shore and on Patea Shoals it is ~ 0.3–0.45 m s⁻¹,

depending mainly on depth. Storm events typically generate wave orbital speeds several times the mean.

(24)

24 Sediment Plume Modelling Figure 2-7: Near-bottom speeds (m s−1) on the Cape Egmont to Kapiti inner model domain. (a) Mean non- tidal speed. (b) Maximum tidal speed (semi-major axis) of the main lunar semi-diurnal constituent, M2. (c) Mean bottom wave orbital speed.

(25)

Sediment Plume Modelling 25

2.6 Sediment model setup

Sediment calculations were carried out on the two inner domains (which were used for different purposes), each nested within a larger-scale Greater Cook Strait model (Figure 2-1). Here we use Greater Cook Strait to refer to the waters between the Tasman Sea (to the west) and the Pacific Ocean (to the east). Cook Strait itself is shown in Figure 1-1, being the narrowest stretch of water between the North and South Islands of New Zealand.

2.7 Sediment properties and release parameters

The sediment scheme in the ROMS modelling system requires the modeller to define a set of sediment classes and, for each one, to specify several properties, notably: median grain size; grain density; porosity (when in the sediment bed); settling velocity (when in suspension); critical bed shear stress for erosion and deposition; and a rate parameter used in the formula relating erosion rate to bed shear stress. The number of sediment classes is unlimited, but the computational expense increases for each additional class.

The results presented in this report were produced with the ROMS sediment model on one of the inner domains. The following sub-sections describe the different sediment classes

2.8 Background sediments

The base simulation represents background sediment processes using 7 sediment classes:

 The river-derived sediments that are injected by the rivers. There are two classes: a coarse silt (16–63 µm) and a fine silt/clay (< 16 µm).

 The seabed-derived sediments comprise the seabed at the beginning of the simulation.

They range from a coarse sand (500–1000 µm) to a fine silt (4–16 µm).

There were two reasons for introducing the background sediments:

 Primarily, to support a realistic interaction between the mining-derived sediment plumes and the seabed;

 Secondarily, to produce estimates of background suspended sediment concentrations that are approximately correct and can be compared with predictions of sediment concentrations resulting from the ironsand extraction operation.

The seabed is initially populated with a combination of coarse sand (500–1000 µm, 20%), fine–

medium sand (128–500 µm, 72%), very fine sand (63–128 µm, 6%), coarse silt (16–63 µm, 1.5%) and fine silt (4–16 µm, 0.5%). The proportions were initially based on seabed particle size distribution (PSD) data from the extraction area, and the fine sediment fractions were then adjusted so that the model produces surface SSCs of approximately the correct magnitude in the near-shore area (Sections 4.2 and 4.1). The seabed composition was assumed to be uniform over the model domain.

For the present calculations, the background sediment parameters differed from those used in the March 2014 calculation in two ways:

 The minimum critical stress was increased from 0.1 to 0.2 Pa. in line with the HR Wallingford findings.

(26)

 We abandoned the imposition of a minimum settling velocity of 0.1 mm/s to accurately represent flocculation for background sediments. While the process of flocculation does tend to increase the settling velocity of the bulk of fine sediments, the laboratory results suggest that it leaves a residual level of very slow-settling material. Therefore the minimum settling velocity for both riverine and seabed background sediments was reduced from 0.1 mm/s to 0.01 mm/s.

The background sediment parameters are listed in Table 2-2, with changes from the March 2014 configuration highlighted in a bold font. Other properties required by ROMS include an erosion rate parameter (2 × 10⁻⁴ kg m⁻² s⁻¹ for all classes) and a porosity (0.4 for all classes).

Table 2-2: Background sediment parameters. Values that differ from the March 2014 configuration are shown in a bold font.

Label Source Nominal

size range (µm)

Settling velocity (mm/s)

Critical stress

(Pa)

Fraction initially present

in seabed

sand_01 River 16–63 0.63 0.200

sand_02 River 4–16 0.01 0.200

sand_03 Seabed 500–1000 103 0.431 20%

sand_04 Seabed 128–500 38 0.219 72%

sand_05 Seabed 63–128 6.3 0.200 6%

sand_06 Seabed 16–63 0.76 0.200 1.5%

sand_07 Seabed 4–16 0.01 0.200 0.5%

Note that we also considered using a non-zero external concentration (~0.1–0.5 mg/L) for the finest seabed class (sand_07) to provide an input of fine sediment from outside the model domain. This was intended to offset the tendency of the model to underestimate the mean SSC south of Patea Shoals in comparison with remote sensed data (Section 4.2). After subsequent discussions with Dr Matt Pinkerton it was decided to represent this material by an adjustment in the optical post- processing of the model output (Pinkerton 2015c) and the external concentration was set to zero.

The model of background sediment processes was initialised at 2000 days (relative to a reference time of 00:00 UTC on 1 January 2005) and run to 3000 days, with statistical analyses based on the last 730 days of the simulation. The sediment bed adjusts within the first 100–200 days after

initialisation as the finer classes are stripped out of the uppermost bed layer in high-energy areas like Patea Shoals but remain in low-energy areas. A slower adjustment process occurs throughout the duration of the simulation as seabed material continues to be eroded from the high-energy areas and deposited in the low-energy areas.

2.9 Mining-derived sediments

Two main sediment streams from the ironsand extraction are considered: the hydro-cyclone overflow discharge and the de-ored sand discharge. The hydro-cyclone overflow results from dewatering of the de-ored sand before it is pumped to the bottom. It is a discharge of mostly fine sediment with a large flow (8.8 m³/s) of water. The de-ored sand discharge involves de-watered, de- ored sand being released from a pipe with a view to depositing it as compactly as possible, usually into a pit that has been excavated earlier. The de-ored sand is predominantly fine–medium sand (125–500 µm) with some finer material. Both discharges are no more than 4 m or so above the

(27)

Sediment Plume Modelling 27 bottom and in the current proposal are released close to each other, with a view to maximising the trapping of fine sediment in the pit with the coarser sands.

These combined sediment stream is represented in the model by two different mechanisms, treated in different simulations. The first, the suspended source, is a continuous source of fine suspended sediment. The second, the patch source, is a 3 × 2 km rectangular patch of sand representing the area filled by one year’s mining and containing all the material that has not been released in the suspended source. The set-up of the simulations that implement these two sources is described further below.

2.9.1 Suspended source

The suspended-source sediment parameters in the current simulations are based on laboratory and model results outlined in the HRW report (HRW 2015). The resulting classification is presented in Table 2-3.

Table 2-3: Suspended source sediment parameters. The discharge rate is for a plant throughput of 50 Mt/a. Values that differ from the March 2014 configuration are shown in a bold font.

Class Source Settling

velocity (mm/s)

Critical stress

(Pa)

Discharge rate (kg/s)

sand_08 Overflow 1.0 0.200 1.45

sand_09 Overflow 0.10 0.200 12.55

sand_10 Overflow 0.01 0.200 6.00

sand_11 Underflow 1.0 0.200 0.25

sand_12 Underflow 0.10 0.200 1.80

sand_13 Underflow 0.01 0.200 0.85

The movement from a grain-size-based classification (March 2014) to a settling-velocity-based classification has not changed the settling velocity of the finest sediment classes, which remain at 0.1 mm/s and 0.01 mm/s. These are the sediment types that are most readily mixed to the surface and they are the most optically active, so they are the sediment types that are most important in determining the near-surface SSC and optical effects. The changes in the discharge rates for these sediment classes are summarised in Table 2-4.

Table 2-4: Summary of changes in discharge rates for the finer mining-derived sediments. . Settling velocity

(mm/s)

Discharge rate (kg/s) Ratio (Mar 2015/

Mar 2014) Mar 2014 Mar 2015

0.1 14.5 14.35 0.99

0.01 14.7 6.85 0.47

all < 0.1 29.2 22.8 0.73

Another difference from the March 2014 simulations is the depth at which the tailings discharge is introduced into the model. For the March 2014 simulations this release height was assumed to be

(28)

28 Sediment Plume Modelling 15 m below the surface, based on the plant design information that was available when the

simulations were originally set up (in June/July 2013). Since then it has become clear that:

 The discharge point for the tailings will be lower than originally indicated, nominally about 4–6 m above the bottom.

 Subsequent to the discharge the plume will descend to the bottom and form a bottom- attached plume of a few metres thickness (Hadfield 2014c).

2.9.2 Patch source

Of the fine sediments (< 63 μm) that are released with the tailings, a fraction will settle to the bottom in the mining pit and then be buried with the de-ored sand. This fraction, which depends strongly on settling velocity, was quantified by HR Wallingford (2015) and taken into account in setting the suspended source discharge rates (Section 2.9.1). The trapped fine sediment is then available for resuspension at a rate controlled by the erosion of the sand and the concentration of the fine sediment in the sand matrix.

To model this process we consider a rectangular patch representing one year’s worth of ironsand extraction and populate this patch with material that reflects the composition of the combined hydro-cyclone and de-ored sand discharge streams, minus all the material that was released in the suspended source.

The area of this source is calculated as follows: Assuming a volume extraction rate of 1.195 m³/s at full operation (mass extraction rate 8000 tonne/h with a bulk density of 1860 kg/m³) and an up-time of 80%, the annual volume extracted is 30.15 × 10⁶ m³. At a mean mined depth of 5 m, this indicates that an area of 6.05 km² would be mined in one year. This area is represented in the model as a 3 × 2 km rectangular patch centred on the mining site, which is taken to be mining site A.

The patch source was implemented in the inner model on the Patea Shoals domain. The sediment model included the five seabed sediment classes, but not the riverine sediment classes (as the major rivers are outside this domain and are not important for the predominantly short-range, near-bottom transport involved). In addition there were three “patch” sediment classes representing the trapped fine sediment. The model was initialised at 2000 days with seabed sediments filling the domain, then at 2200 days the seabed in the 3 × 2 km patch area was repopulated with a mixture of the three coarsest seabed sediment classes and the three patch classes. The simulation was continued to 3000 days.

The composition to which the patch was set at 2200 days (Table 2-5) was calculated from the same discharge rate data for fine sediments (including flocculated fine sediments) as was used in deriving the suspended source parameters (Table 2-3) but this time considering the fraction trapped on the bottom rather than the fraction remaining in suspension. The total rate at which fine sediment is discharged and trapped in the pit is 71 kg/s and this material mixes with the de-ored sand, which is discharged at 1910 kg/s. The trapped fine sediment (classes sand_06 to sand_08) therefore

comprises 3.7% of the patch. For simplicity, the remainder of the patch is assumed to be composed of the three coarsest classes in the original seabed, with the proportions adjusted slightly to ensure the total is 100%.

(29)

Sediment Plume Modelling 29 Table 2-5: Patch sediment parameters. The initial mass fraction column specifies the seabed composition at the start of the simulation. The patch mass fraction column specifies the seabed composition imposed inside the 3 × 2 km patch at 2200 days.

Class Nominal size range (µm)

Settling velocity (mm/s)

Mass fraction in seabed

Initial Patch

sand_01 500–1000 103 20% 19.7%

sand_02 128–500 38 72% 70.8%

sand_03 63–128 6.3 6% 5.9%

sand_04 16–63 0.76 1.5%

sand_05 4–16 0.01 0.5%

sand_06 10 2.6%

sand_07 1 0.8%

sand_08 0.01 0.3%

(30)

3 Hydrodynamic model evaluation 3.1 Field measurements

A programme of field measurements was carried out in the area from Patea Shoals to Whanganui, involving instrument deployments at several sites for three periods between September 2011 and July 2012. The programme is described in detail in a dedicated report (MacDonald et al. 2012). The data that are most relevant for the sediment plume modelling are:

 Vertical velocity profiles from acoustic Doppler current profilers (ADCPs) at five sites, shown in Figure 2-1b.

 Profiles of temperature and salinity from moored sensors at several sites.

 Measurements of suspended sediment from optical and acoustic backscatter sensors (OBS and ABS).

Below, in the present section, modelled velocities are compared with the ADCP measurements. In Section 4 modelled near-bottom suspended sediment (sand) concentrations are compared with ABS data to give an approximate check on the model’s representation of bed resuspension processes.

The times and locations at which ADCP data are available are listed in Table 3-1.

Table 3-1: ADCP data availability from the field measurements. For more information see MacDonald et al.

(2012), specifically Section 3.1 and Tables 3-1 & 3-2. The sites are shown in Figure 2-1b.

Deployment Site 5 Site 6 Site 7 Site 8 Site 10

D1, 06/09/2011 to 01/12/2011 X X X

D2, 08/12/2011 to 09/02/2012 X X X

D3, 24/04/2012 to 01/07/2012 X X

In all the deployments the ADCP instruments were mounted on the bottom, pointing vertically upwards and measuring horizontal velocities every 10 minutes at a series of levels (or “bins”) above the instrument. The vertical spacing between the bins varied between instruments but was typically 0.5 m. The model was set up to store velocity profile data at all the ADCP locations at an interval of 30 minutes. For comparisons between ADCP and model, the model profiles were interpolated vertically and in time to match a specified ADCP level; the analyses below concentrate on one that will be labelled “mid-depth”, i.e. the one nearest to halfway between the surface and the bottom.

Also, we consider the tidal and sub-tidal components separately, the former being estimated by fitting harmonics of specified frequency to the data and the latter by applying a low-pass temporal filter to the data, an operation known as detiding.

3.2 Tidal current comparison

This section considers the accuracy of the model’s representation of tidal currents. As is the case elsewhere around New Zealand, the dominant tidal constituent in the area is the lunar, semi-diurnal constituent (M2), with a period of 12.42 hours. Figure 3-1 compares measured and modelled M2 tidal ellipses for one of the ADCP datasets, namely Site 7 (outer Patea Shoals), Deployment 2.

(31)

Sediment Plume Modelling 31 Figure 3-1: M2 tidal velocity comparison (ADCP Site 7 Deployment 2). Mid-depth M2 tidal ellipses from ADCP (blue) and model (red). The axes correspond to the velocity components towards due east (u) and due north (v). The ellipses represent the magnitude and orientation of the tidal velocity variations (see text) and the straight line from the origin to the ellipse represents the phase.

The tidal ellipse shows tidal velocity variations: it represents the path taken by the tip of a tidal current vector, rotating at a constant angular frequency and changing in length (current speed) through a tidal cycle. The tidal ellipse is defined by four parameters:

 Semi-major amplitude (m/s): The semi-major axes are lines from the origin to the two most distant points on the ellipse perimeter. The two axes are equal in length, and this length represents the amplitude of the velocity along the semi-major direction, or the maximum current speed during a tidal cycle.

 Eccentricity: At right angles to the semi-major axes are the semi-minor axes, representing the minimum current speed during a tidal cycle. The eccentricity, or

“fatness”, of the ellipse is the ratio of semi-minor to semi-major axis lengths. The eccentricity can be positive (vector rotates anti-clockwise) or negative (clockwise).

 Inclination (°T): The inclination is the orientation of one of the semi-major axes. The choice between the two is arbitrary: here we take the semi-major axis directed towards the northeastern or southeastern quadrant and express the inclination as the orientation in degrees clockwise from true north.

 Phase (°): The phase relates to the time at which the rotating tidal current vector passes through the semi-major axis. A phase difference of 1° corresponds to a time difference of 1/360^th of the tidal period.

(32)

32 Sediment Plume Modelling The tidal ellipses shown in Figure 3-1 clearly match reasonably well in magnitude and orientation.

Similar ellipses for the other ADCP deployments are shown in Appendix B, Figure B-1. Overall agreement is very good. The semi-major amplitudes agree to within ±6%; the eccentricities agree within ±0.04, the inclinations within ±8°, and the phases within ±2.4° (5.0 minutes). This is a very good match.

The tidal current amplitude normally decreases towards the bottom due to friction. Figure 3-2 compares measured and modelled vertical profiles of the M2 semi-major amplitude for the same ADCP dataset as Figure 3-1. The agreement in the middle and upper water column is very good (as was evident from Figure 3-1). Measured and modelled amplitudes both decrease towards the bottom, as expected. The lowest level at which data are available from the ADCP instrument at this site is 2.05 m above the bottom: here the modelled amplitude exceeds the measured amplitude by 10%. Similar profiles for the other ADCP deployments are shown in Figure B-2. Agreement is

generally good, with a tendency for modelled amplitude to exceed measured amplitude by a modest amount in the lowest few metres. This may indicate that the effective bottom roughness is

somewhat lower in the model than in reality.

Figure 3-2: M2 tidal current profile comparison (ADCP Site 7 Deployment 2). Semi-major amplitude of the M2 tide versus height above the bottom from ADCP (blue) and model (red).

The M2 constituent dominates the tidal velocity but several smaller-amplitude constituents also contribute. The larger ones in this area are S2 (12 hours), N2 (12.66 hours), and K1 (23.93 hours), for which mid-depth tidal ellipses are compared in Figure 3-3. Note the different velocity scale in this figure relative to Figure 3-1. Agreement is good, though not as close as for M2. The largest mismatch evident in these graphs is a difference of 31°, or 2.1 hours, in the phase of the K1 constituent. This difference probably results from a bias in the lateral boundary data taken from the EEZ tidal model, which is known to have significant errors for this constituent in this region (Stanton et al. 2001).

(33)

Sediment Plume Modelling 33 Figure 3-3: Tidal velocity comparison for S2, N2 and K1 (ADCP Site 7 Deployment 2). Mid-depth tidal ellipses from ADCP (blue) and model (red).

3.3 Sub-tidal current comparison

A comparison of sub-tidal currents is shown in Figure 3-4. The upper panel in this figure shows a scatter plot of velocities, with the ellipse in this case being a variance ellipse, a conventional representation of the magnitudes of variability in velocity data. A variance ellipse can be characterised by its semi-major axis (in this context called a principal axis), eccentricity and

inclination, like a tidal ellipse. However a variance ellipse does not have a phase (since it says nothing about the timing of the variability) and its eccentricity has no sign (since it says nothing about the rotation of velocity vectors). Also, the centres of the variance ellipses in Figure 3-4 are offset from the origin by an amount representing the mean current over the period of the deployment. The lower two panels in the figure indicate how well the fluctuations in sub-tidal velocity match up in time. The centre panel shows fluctuations along the principal axis of maximum variability (this being a compromise between the principal axes of the observed and modelled variance ellipses) and the bottom panel shows fluctuations perpendicular to this axis.

(34)

34 Sediment Plume Modelling Figure 3-4 indicates good agreement between model and measurements for the ADCP deployment in question. The variance ellipses have a similar shape and orientation, there is a high correlation (r = 0.849) between measured and modelled fluctuations along the principal axis and a somewhat lower correlation (r = 0.555) perpendicular to this. The mean current (the offset of the centre of the ellipse) is small relative to the dimensions of the ellipse but appears to be directed to the east in both cases.

On a technical matter, the time series in Figure 3-4 are quite smooth, as a result of the application of a low-pass filter in detiding the data. (The filter is the 51G113 filter from Thompson, 1983, applied to hourly values; see Figure 2 of that article for the filter’s frequency response.) To investigate the effect of this smoothing, the model-measurement comparison has been repeated with an alternative method, namely, removal of the tides by analysing a set of several tidal constituents (3 semi-diurnal and 2 diurnal) from the data and then constructing and subtracting the tidal time series. A

comparison of the results of the two methods (Figure 3-5, along with other output not shown) indicates that the latter method of detiding does leave more high-frequency information in the data, but for the purposes of comparison between model and measurements, both methods give similar results, as long as the model and measurement time series are detided in the same way.