Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modeling herring and hake larval dispersal in the Salish Sea Snauffer, Evgeniya Lyubomirova 2013

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2013_spring_snauffer_evgeniya.pdf [ 1.72MB ]
Metadata
JSON: 24-1.0073588.json
JSON-LD: 24-1.0073588-ld.json
RDF/XML (Pretty): 24-1.0073588-rdf.xml
RDF/JSON: 24-1.0073588-rdf.json
Turtle: 24-1.0073588-turtle.txt
N-Triples: 24-1.0073588-rdf-ntriples.txt
Original Record: 24-1.0073588-source.json
Full Text
24-1.0073588-fulltext.txt
Citation
24-1.0073588.ris

Full Text

MODELING HERRING AND HAKE LARVAL DISPERSAL IN THE SALISH SEA  by  EVGENIYA LYUBOMIROVA SNAUFFER  M.Sc., Sofia University, Bulgaria, 1999  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE  in  The Faculty of Graduate Studies  (Oceanography)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  February 2013  © Evgeniya Lyubomirova Snauffer, 2013  Abstract The Salish Sea includes Juan De Fuca Strait, Puget Sound, and the Strait of Georgia (SoG), which separates Vancouver Island from mainland British Columbia. Hake and herring are commercially important fish and both species use SoG as larval rearing grounds. Drift tracks of larvae for these species were simulated using a regional circulation model and a particle-tracking model, for up to six weeks after they hatch. Larvae with different behaviors (such as surface drifters or performing diel vertical migration) are traced in the springs of each of the years 2007, 2008, and 2009. Since herring larvae stay in the top 12m, their distribution is heavily influenced by the wind storms. Strong winds to the north during the hatching period wash herring larvae out of SoG and lead to poor recruitment later. Alternatively, wind storms blowing to the south help retain herring larvae in the Salish Sea. Northern and southern parts of SoG are weakly connected for herring larvae. Hake larvae reside deeper in the water column (50-200m) and the distribution of the hake larvae released in the central SoG is shaped by a deep gyre with cross-strait currents. Behavior changes distribution for both types of larvae but there is no single pattern. Behavior may enhance retention in SoG for the northern herring larvae. This study helps to identify important herring larvae habitat in the Strait of Georgia.  ii  Preface This project was initiated by my adviser, Dr. Susan Allen. Dr. Diane Masson provided the configuration of ROMS model for the Salish Sea. I used her configuration to store ROMS output at time intervals suitable for this project. Dr. Allen provided consultation on LTRANS numerical considerations, analysis of the model results and help with programming larval behavior. I was in charge of configuring LTRANS for herring and hake larvae and running the models. Chapter two will be revised into a manuscript co-authored with Dr. Masson and Dr. Allen. I was also in charge of the literature review, data analysis and preparation of the manuscript with the guidance and corrections from Dr. Allen.  iii  Table of Contents Abstract.......................................................................................................................................... ii Preface........................................................................................................................................... iii Table of Contents ......................................................................................................................... iv List of Tables ............................................................................................................................... vii List of Figures............................................................................................................................. viii List of Supplementary Materials................................................................................................. x Acknowledgements ..................................................................................................................... xii Chapter 1: General Introduction ................................................................................................ 1 1.1 1.2 1.3 1.4 1.5 1.6  General Introduction ....................................................................................................................... 1 SoG herring life cycle ..................................................................................................................... 5 SoG hake life cycle ......................................................................................................................... 6 Oceanography of the Salish Sea...................................................................................................... 6 Biological oceanography of the Salish Sea ................................................................................... 10 Models used in this study .............................................................................................................. 11  Chapter 2: Modeling herring and hake larval dispersal in the Salish Sea ............................ 13 2.1 Introduction ................................................................................................................................... 13 2.2 Methods......................................................................................................................................... 13 2.2.1 The Regional Ocean Modeling System ................................................................................. 15 2.2.2 Lagrangian TRANSport Model (LTRANS) .......................................................................... 17 2.2.2.1 Overview ........................................................................................................................ 17 2.2.2.2 Concentration maps ........................................................................................................ 18 2.2.2.3 Particle average paths ..................................................................................................... 19 2.2.3 Herring larvae in LTRANS.................................................................................................... 19 2.2.3.1 Base and Sensitivity studies ........................................................................................... 19 2.2.3.2 Herring spawn locations and particle release locations .................................................. 20 2.2.3.3 Incubation period and particle release date .................................................................... 21 2.2.3.4 Vertical migration........................................................................................................... 22 2.2.3.5 Swimming speeds ........................................................................................................... 23 2.2.3.6 Depth maintaining .......................................................................................................... 24 2.2.4 Hake larvae in LTRANS........................................................................................................ 24 2.2.5 Numerical considerations....................................................................................................... 26 2.3 Results ........................................................................................................................................... 27 2.3.1 Herring particles..................................................................................................................... 27 2.3.1.1 Wind impact on herring particle distributions ................................................................ 27 2.3.1.2 Herring base study.......................................................................................................... 29 2.3.1.3 Comparison with field observations ............................................................................... 31 2.3.1.4 Sensitivity to start position and date............................................................................... 34 2.3.1.5 Sensitivity to vertical swimming behavior ..................................................................... 34 2.3.1.6 Interannual variability for sensitivity study.................................................................... 36 2.3.2 Hake particles......................................................................................................................... 39 2.3.2.1 Sensitivity to start position and date............................................................................... 39 2.3.2.2 Sensitivity to vertical swimming behavior ..................................................................... 40 2.3.2.3 Interannual variability..................................................................................................... 40 2.3.2.4 Comparison with field data............................................................................................. 41  iv  2.4 Discussion ..................................................................................................................................... 43 2.4.1 Herring base study ................................................................................................................. 43 2.4.2 Comparison with field observations ...................................................................................... 44 2.4.3 Sensitivity to exact start position and date............................................................................. 45 2.4.4 Sensitivity to vertical swimming behavior............................................................................. 45 2.4.5 Imapct of wind direction on retention for herring particles ............................................... 46 2.4.6 Interannual variability of hake particles ...................................................................... 47 2.5 Conclusions ................................................................................................................................... 48  Chapter 3: Conclusion................................................................................................................ 50 Bibliography ................................................................................................................................ 54 Appendix A: Regional Ocean Modeling ................................................................................... 61 A.1 A.2 A.3 A.4 A.5  Brief general description of the model .......................................................................................... 61 Vertical sigma levels in SoG......................................................................................................... 62 ROMS configuration for this study............................................................................................... 62 ROMS rho grid.............................................................................................................................. 66 Frequency of ROMS ouput ........................................................................................................... 67  Appendix B: Lagrangian TRANSport model (LTRANS)....................................................... 69 B.1 B.2 B.3 B.4 B.5 B.6  Interpolation of some water properties.......................................................................................... 69 Flow diagram ................................................................................................................................ 70 Determining LTRANS internal time step by comparing numerical and analytical solutions....... 71 LTRANS internal time step........................................................................................................... 77 Sufficient number of particles ....................................................................................................... 77 Changes in LTRANS (version 2) relevant for ths study ............................................................... 78  Appendix C: Lighthouse data .................................................................................................... 80 C.1 Lighthouse positions and herring particles release spots .............................................................. 80 C.2 Comparison between ROMS output and lighthouse temperature data.......................................... 80  Appendix D: Vertical profiles of horizontal velocities............................................................. 81 D.1 Surface currents............................................................................................................................. 81 D.2 Deep currents ................................................................................................................................ 83  Appendix E: Herring particles .................................................................................................. 83 E.1 Base study ..................................................................................................................................... 83 E.1.1 Herring particle distributions in 2007 .................................................................................... 83 E.1.2 Herring particle distributions in 2008 .................................................................................... 85 E.1.3 Herring particle distributions in 2009 .................................................................................... 86 E.2 Sensitive studies ............................................................................................................................ 88 E.2.1 Sensitivity to start position (Two grid cells apart study) ....................................................... 88 E.2.2 Sensitivity to start date (One day apart study) ....................................................................... 89 E.2.3 Sensitivity to vertical swimming behavior............................................................................. 90 E.2.4 Interannual variability for sensitivity study ........................................................................... 90 E.2.5 Particle average paths ............................................................................................................ 96  Appendix F: Hake particles ..................................................................................................... 103 F.1 F.2 F.3 F.4 F.5  Vertical behavior ......................................................................................................................... 103 Swimming speeds for some larval lengths .................................................................................. 103 Sensitivity to start position and date............................................................................................ 104 Sensitivity to vertical swimming behavior.................................................................................. 104 Interannual variability ................................................................................................................. 106  v  F.6 F.7 F.8  Particle average paths.................................................................................................................. 110 Water currents at different depth................................................................................................. 111 Deep currents in central SoG ...................................................................................................... 112  vi  List of Tables Table B1. Concentration differences between the analytical and numerical solutions for a batch of 6000 particles ........................................................................................................................... 76 Table B2. Concentration differences between the analytical and numerical solutions for a batch of 8000 particles............................................................................................................................ 76 Table B3. Concentration differences between the analytical and numerical solutions for a batch of 10 000 particles......................................................................................................................... 76 Table F1. Larval speeds calculated for different size larvae………………………………...…103  vii  List of Figures Figure 1. Average surface currents in the Salish Sea...................................................................... 9 Figure 2. Coastline of the Straits of Georgia (SoG) and Juan de Fuca. ........................................ 16 Figure 3. Analysis of model results .............................................................................................. 18 Figure 4. Actual herring spawn locations for each statistical area ............................................... 21 Figure 5. Particle vertical position for LTRANS runs.................................................................. 23 Figure 6. Correlation between winds and percentage of surface drifters washed out through Johnstone Strait............................................................................................................................. 28 Figure 7. Hourly distances that particles traveled for the first 26 hours after their release are correlated (r2-value = 0.63) to the squared hourly wind speed in y direction for the same time period. ........................................................................................................................................... 29 Figure 8. Base study distribution maps for 2007 and 2009 .......................................................... 30 Figure 9. Comparison between model results and field observations for herring particles.......... 33 Figure 10. Distribution maps for different behavior types ........................................................... 35 Figure 11. Paths of water parcels from different depths (2m, 6m, and 12m) as if the water parcels were observed from a fixed point ................................................................................................. 36 Figure 12. Concentration maps for northern particles with DVM ............................................... 37 Figure 13. Concentration maps for particles with DVM released in central SoG. ....................... 38 Figure 14. Concentration maps for southern particles with DVM................................................ 39 Figure 15. Herring and hake particle average paths ..................................................................... 41 Figure 16. Comparison between field data and model results for hake particles ......................... 42 Figure A1. Vertical profile with sigma levels............................................................................... 62 Figure A2. Four points for which time series of surface current velocities between D. Masson’s configuration and this study were constructed.............................................................................. 64 Figure A3. Comparison of the magnitudes of surface current velocities between D. Masson and this study ....................................................................................................................................... 64 Figure A4. Differences in surface currents velocities between ROMS results done for this study and D. Masson .............................................................................................................................. 65 Figure A5. Surface height comparison ......................................................................................... 65 Figure A6. Differences in surface height for the whole domain................................................... 66 Figure A7. ROMS staggered horizontal grid ................................................................................ 67 Figure A8. PDFs calculated for different frequency of ROMS output and different number of particles. ........................................................................................................................................ 68 Figure A9. PDFs calculated for 15min, 30min, and 1h ROMS output and 1000, 2000, and 4000 particles ......................................................................................................................................... 69 Figure B1. Boundaries and horizontal velocities interpolated from ROMS predictions.............. 70 Figure B2. Flow diagram of LTRANS ......................................................................................... 71 Figure B3. Particle locations after floating for 28 days ................................................................ 72 Figure B4. Comparison between the numerical and analytical solutions for particle concentrations for a batch of 6000 particles ................................................................................. 73 Figure B5. Comparison between the numerical and analytical solutions for particle concentrations for a batch of 8000 particles ................................................................................. 74 Figure B6. Comparison between the numerical and analytical solutions for particle concentrations for a batch of 10 000 particles .............................................................................. 75  viii  Figure B7. Particle vertical position for LTRANS runs with internal time step 60s and 300s internal time step .......................................................................................................................... 77 Figure B8. PDFs calculated for different number of particles...................................................... 78 Figure B9. Particle crossing into one grid cell channel ................................................................ 79 Figure C1. Map of particle release locations and lighthouses ...................................................... 80 Figure C2. ROMS versus lighthouse temperature ........................................................................ 81 Figure D1. Along-shore velocities near Comox ........................................................................... 82 Figure D2. Along-shore velocities near Nanaimo ........................................................................ 82 Figure D3. Along-shore velocities in Central SoG....................................................................... 82 Figure D4. Along-shore velocities near Montgomery Bank......................................................... 83 Figure E1. Base study distribution maps for 2007........................................................................ 84 Figure E2. Base study distribution maps for 2008........................................................................ 86 Figure E3. Base study distribution maps for 2009........................................................................ 87 Figure E4. Batches of particles released in pairs in different locations in SoG and Juan De Fuca Strait.............................................................................................................................................. 88 Figure E5. Estimate error for herring PDFs in “Two grid cells apart “study ............................... 89 Figure E6. Estimate error for herring PDFs in “One day apart “study ......................................... 89 Figure E7. Percentage of surface drifters and particles with DVM washed out through Johnstone Strait and/or mouth of Juan De Fuca Strait................................................................................... 90 Figure E8. Concentration maps for northern surface drifters ....................................................... 94 Figure E9. Concentration maps for surface drifters released in Central SoG............................... 95 Figure E10. Concentration maps for southern surface drifters ..................................................... 96 Figure E11. Average paths for northern surface drifters .............................................................. 97 Figure E12. Average paths for northern particles with DVM ...................................................... 98 Figure E13. Average paths for southern surface drifters .............................................................. 99 Figure E14. Average paths for southern particles with DVM .................................................... 100 Figure E15. Average paths for surface drifters released in central SoG..................................... 101 Figure E16. Average paths for particles with DVM released in central SoG............................. 102 Figure F1. Vertical distribution of (a) larvae off the coast of California and (b) eggs and larvae in Puget Sound ................................................................................................................................ 103 Figure F2. Estimate error for (a) two grid cells apart study and (b) one day apart study........... 104 Figure F3. Concentration maps for particles released in 2009 near Halibut Bank ..................... 104 Figure F4. Concentration maps for particles released in 2009 near Montgomery Bank. Particles have 3 types of behavior – hake larvae like, floaters at 75m, floaters at 200m.......................... 105 Figure F5. Concentration maps for particles released in 2009 near Montgomery Bank ............ 106 Figure F6. Concentration maps for particles released in 2009 near Texada............................... 107 Figure F7. Concentration maps for particles released in 2009 near Halibut Bank ..................... 108 Figure F8. Concentration maps for particles released in 2009 near Nanaimo............................ 109 Figure F9. Average paths for particles released in 2009 ............................................................ 110 Figure F10. Water parcels paths at different depths (75m and 200m) as if the water parcels were observed from a fixed point ........................................................................................................ 111 Figure F11. Particle tracks and deep circulation......................................................................... 112  ix  List of Supplementary Materials Supplementary materials (movies) are available from cIRcle, UBC's online Information Repository. The purpose of the movies is to show that the movements of herring particles from the sensitivity study are driven by the winds. sens2007dvm.avi sens2008dvm_begin.avi sens2008dvm_end.avi sens2009dvm.avi  x  Acknowledgements I would like to thank my advisers, Dr. Susan Allen and Dr. Diane Masson for their guidance, patience and support throughout my Masters research. Special thank you to Susan for believing in me and giving me the chance to achieve my dream. And thanks to my committee member Dr. Tony Farrell for his constructive feedback on my thesis. I would like to especially acknowledge Dr. Doug Hay and Bruce McCarter from Pacific Biological Station in Nanaimo for providing the herring database and their help with the biological part of this project. I thank my family and friends for their encouragement and good times. Drew, your endless love and support have been invaluable for me throughout this adventure.  xi  Chapter 1: General Introduction 1.1  General Introduction Pacific herring and hake are harvested for human consumption. Both types of fish  transfer energy from lower trophic levels to the top predators. Herring are a food source for many large fish, marine mammals and seabirds. As a large predator, hake pray on Pacific herring, northern anchovy, and shrimp and they serve as a food source for marine mammals and large fishes. Herring larvae are dependent on ocean currents to bring them to areas with enough food to sustain them. Larval mortality is the greatest of any life stages and it determines abundance at an adult age. Since ocean currents are important for larval retention into good habitats, larval tracks were modeled to study herring and hake larval dispersal in the Salish Sea and identify the physical drivers for the dispersal. The objectives of this project1 are to simulate young herring and hake larvae trajectories in the Salish Sea, to map larval distributions, identify congregation / disaggregation areas, point to the physical and biological factors that influence larval dispersion, determine interannual variability and perform sensitivity analysis on start position, start date, and vertical swimming behavior. This study is a first look at the connectivity of the Strait of Georgia from the perspective of herring and hake larvae. Larval dispersal depends on biological and physical processes: biological in the sense of larval production, growth, development, and survival; physical in the sense of advection and 1  This thesis is written with the support of Canadian Healthy Oceans Network (CHONe). CHONe is an NSERC strategic network focused on biodiversity science for the sustainability of Canada's three oceans. The network includes researchers from 14 universities across Canada, the federal Department of Fisheries and Oceans (DFO), and seven other government laboratories. CHONe’s goal is to provide scientific guidelines for policy in conservation and sustainable use of marine biodiversity resources in Canada's three oceans. CHONE’s Population Connectivity theme (to which this project belongs) addresses how dispersal of marine organisms, typically by early life stages such as eggs and larvae, influences patterns of diversity, resilience, and source/sink dynamics (recruitment "hotspots" versus poor areas for new individuals) of species and biological communities. One of its targets is to study larval dispersal and to estimate interlinked population connectivity.  1  diffusion; and biophysical in the sense of interactions between certain larval features (e.g., vertical swimming behavior) and physical properties of the water column (e.g. shear currents) (Cowen and Sponaugle, 2009). This study only focuses on physical and biophysical drivers of larval dispersal. Advection depends on currents, which depend on winds (especially near the surface), tides, river runoff, and density gradients while diffusion is related to turbulence. Different types of behavior may affect the dispersal of larval fish. For example, in some studies, passively drifting larvae are washed offshore but larvae performing vertical migration are advected close to the shore (Sentchev and Korotenko, 2007, Fox et al., 2006). Understanding larval dispersal is important because dispersal and retention in areas favorable to feeding and growth are critical to abundance and distribution (Stevenson 1962). There is a difference between larval transport and larval dispersal (Pineda et al, 2007). Larval transport is defined as the horizontal displacement of a larva between two points. To move from one point to another, larva can swim horizontally, or they can be transported by diffusion and advection, or larva can swim vertically and “catch” different currents at different depths. In this work, larval dispersal relates to the spread of larvae for the first few weeks after hatching2. Larva dispersal is described by the probability density function characterizing the chance of a larva to be found in a certain area after being transported for a few weeks. Coupled biophysical models are tools that can answer questions about larval transport and dispersal. Recently, a number of reviews have been published showing the increased usage of these models (Peck, 2012; North et al, 2009; Gallego et al, 2007; Miller, 2007; Werner et al., 2007). Coupled biological–physical models apply current and turbulence predictions from threedimensional circulation models to determine the transport of fish eggs, larvae, and juveniles from  2  This definition slightly differs from Pineda et al, 2007. Pineda et al, 2007 defines larval dispersal as “the spread of larvae from a spawning source to a settlement site”  2  spawning to nursery areas. These models are used to study how planktonic dispersal, growth, and survival are influenced by physical and biological processes. Coupled biological–physical models contribute to understanding of fish population variability and stock structure. These models can be as simple as particle-tracking models simulating transport of surface floaters (Helbig and Pepin, 2002) to very complex - simulating marine animal’s development, survival and behavior (Megrey and Hinckley, 2001). In Miller’s (2007) classification, almost all of the bio-physical models are used to study larval transport, and some of the models also focus on larval growth and survival. Herring larval dispersal has been studied extensively in the North Sea (Bartsch, 1993; Fox and Aldridge, 2000; Dickey-Collas et al., 2009), and the Norwegian Sea (Vikebo et al, 2010), using a variety of ocean circulation and particle tracking models. Coupled bio-physical models have advanced to the point that real-time predictions of ichthyoplankton distributions are established (Vikebo et al, 2011). Particles in the models are released in places and on dates matching herring spawning grounds and hatching dates, determined from observations. In most studies, particles perform vertical migration. Fox and Aldridge (2000) used a grid with finer resolution close to the shore in order to better represent complex flow near coasts but the other studies did not. Larval distribution was found to be dependent on winds (Bartsch, 1993). When imposing a steady state wind larval distributions differ from those simulated with meteorological data (Fox and Aldridge, 2000). Larval distributions are also affected by mean currents moving eastward in the southern part of the North Sea (Bartsch, 1993; Dickey-Collas et al, 2009) and the Norwegian Atlantic Slope Current moving northwards in the Norwegian Sea (Vikebo et al, 2010). Model results are sensitive to hatching date in some years in the North Sea (DickeyCollas et al, 2009), and drift depth and hatching date in the Norwegian Sea (Vikebo et al, 2010).  3  Diel vertical movement was relatively unimportant in the transport of larvae in the southern part of the North Sea (Dickey-Collas et al, 2009). There is a large interannual variability in the transport of herring larvae in the North Sea (Dickey-Collas et al., 2009) and the Norwegian Sea (Vikebo et al, 2010). There is a 61-73% area overlap between model simulations and observations in Vikebo et al (2011) and model results are in broad agreement with field observations (Bartsch, 1993; Fox and Aldridge, 2000; Dickey-Collas et al., 2009; Vikebo et al, 2010). Coupled biophysical models have been used more extensively to study larval dispersal on the East Coast of Canada than on the West Coast. A simple 3D model was employed to determine the residence time of capelin larvae in Conception Bay, Newfoundland (de Young et al, 1994). Larval drift patterns (provided by 3D ocean circulation model) were combined with otolith analysis and environmental measurements to study the processes that influenced growth and survival of radiated shanny larvae in eastern Newfoundland (Baumann et al, 2003). Lobster larval transport and connectivity patterns were investigated in the Gulf of Maine (Incze et al, 2010). On the West Coast of Canada, numerical models were used to study Dungeness crab dispersal (Crawford et al, 1995) and connectivity among conservation areas on the north Pacific coast of Canada. (Robinson et al, 2005). In SoG, herring larval distribution has been investigated through extensive ichthyoplankton surveys conducted in April and May from 1989 to 1992 (Mason et al 1982a & b). Highest densities of herring larvae (mean age between 6-27 days) were found on the Vancouver Island side and in the north. Lower larval densities were found in the middle of SoG and mainland side. Almost no larvae were caught in the joining inlets draining into SoG. Very few larvae were found in the southern areas of SoG (Hay and McCarter, 1997). Although the  4  ichthyoplankton surveys revealed abundance and herring larval distribution in the Strait, they did not identify the origin of the larvae, which is important for identifying key spawning grounds. Knowing larval origin also allow us to answer questions about the locations where larvae from each spawning area could be found. To my knowledge, this is the first study to simulate herring and hake larval dispersal in the Salish Sea using a modeling approach.  1.2  SoG Herring Life Cycle Most herring in the Salish Sea belong to migratory populations, with fish moving  seasonally between the west coast of Vancouver Island and the Salish Sea (Therriault et al, 2009). In some areas, however, such as Puget Sound and inlets on the eastern side of the Salish Sea, herring populations are nonmigratory and resident in approximately the same areas throughout the year (ibid). Migratory herring return to SoG in late Fall to spend the winter and spawn in March and April (Taylor, 1964). Herring spawn near shore and their eggs stick to rocky substrates and vegetation (Hay, 1985). Mainly seabirds (ducks and seagulls) feed on herring eggs (Hourston and Haegele, 1980). Larvae hatch after approximately 2 weeks (ibid). During the first 6 days after hatching larvae derive their nourishment from the yolk sack attached to their bodies (ibid). Throughout that time larvae develop their swimming ability and start to feed on small copepods and nauplii (Blaxter, 1962). Herring larvae spend their first few weeks of life in the surface 10m (Stevenson 1962). Since larvae cannot outswim the ocean currents, larvae depend on the currents to bring them to areas with adequate food supply to support them (Stevenson 1962). Young salmon, small perch, arrow worms and jellyfish prey on herring larvae (Hourston and Haegele, 1980). Larvae metamorphose into juveniles after about 10-13 weeks after hatching and grow quickly over the summer (ibid). Herring spend their first year in the SoG and start  5  migrating to their summer feeding grounds off the West Coast of Vancouver Island on their second year (Therriault et al, 2009).  1.3  SoG Hake Life Cycle Pacific hake are the most abundant resident fish in SoG (McFarlane and Beamish 1985).  During the summer hake aggregate at depth 50-100m and are found in association with a dense plankton layer (ibid). Hake feed predominately on euphausiids (King and McFarlane, 2006). By fall, hake aggregations move to Johnstone Strait and schools become dispersed (McFarlane and Beamish 1985). In early winter hake reappear in SoG and begin to swim towards the central strait in preparation for spawning (ibid). Pacific hake spawn in March and April and their eggs are found at depths 100-300m (ibid). There is a debate regarding the depths at which hake larvae are found. McFarlane and Beamish (1985) and Bailey (1982) consider that young hake larvae reside deeper in the water column - between 50 and 300m depth. On the other hand, Hay et al (1989) mention3 that hake larvae spend some time at the surface. For the purpose of this study, it is considered that hake larvae reside deeper - between 50 and 200m depth. Due to this debate, and the lack of data allowing to calculate precise hatching dates, the main focus of this study is on herring larvae distributions and hake larvae distributions are given for comparison only.  1.4  Oceanography of the Salish Sea Since water currents transport larvae of both species, a short overview of the  oceanography of the Salish Sea is given with attention on the surface and deeper currents. This project is mainly focused on the Strait of Georgia but the study domain includes the whole Salish  3  also, D.E. Hay and Bruce McCarter state in personal communication that they found many hake larvae at the surface during surveys in the late 1980s, early 1990s  6  Sea, which consists of the Straits of Georgia (SoG), Juan De Fuca, and Puget Sound (Fig.2). The Strait of Georgia separates Vancouver Island from British Columbia mainland. SoG is a semienclosed basin approximately 200 km long and 40 km wide, with a maximum depth 420 m in its central part (Thomson, 1981). SoG connects to the open ocean through Juan de Fuca Strait to the south, which is also the opening for the Puget Sound Basin. The Gulf and San Juan Islands (a group of small islands) and shallow sill areas constrict the currents between the two straits. To the north, the Strait of Georgia is linked to the Pacific Ocean via several narrow channels, notably Johnstone Strait. The Salish Sea is essentially a large estuary and its oceanography is influenced by its bottom topography, freshwater runoff, tides, and winds (Waldichuk, 1957; Thomson, 1981; LeBlond, 1983; Masson and Cummins, 2004). Approximately 80% of the freshwater runoff in the Strait of Georgia originates from the Fraser River (Tully and Dodimead, 1957) which has a large peak freshet discharge in June associated with snowmelt. During the freshet period, Strait of Georgia waters are strongly stratified near the surface, with a thin (2-10m thick), warmer and brackish layer at the surface. Brackish water (with Practical Salinity below 24, (Yin et al, 1997)) flows southward mostly through the passages between the southern Gulf Islands where it becomes mixed with deeper water by the tidal currents. These mixed waters flow seawards over the saltier waters of the Juan de Fuca Strait. The depth of the return flow in the estuarine circulation is between 50-200m (Pawlowicz et al, 2007). Tides are mainly semidiurnal through the whole system, except in the eastern end of Juan de Fuca Strait, where they are mainly diurnal. The tidal currents in Juan de Fuca Strait are strong (0.75-1.3m/s) (Thomson, 1981). The tidal currents in central SoG are moderate (to the order of ~0.5m/s). Currents in the narrow and shallow passages in the southern SoG, around the San Juan  7  Islands can reach up to 5-6m/s. Tidal currents in the northern part of SoG are weak (order of a 0.1m/s) (LeBlond, 1983). Generally, winds in the winter are dominated by the Aleutian Low atmospheric system and blow from the south-southeast. During the summer, usually Pacific High prevails - the winds are weaker and to the north-northwest (Thomson, 1981). Spring and Fall are shoulder seasons; the winds can be dominated by either of the two atmospheric systems. Surface currents in the central and northern part of SoG are primarily driven by the winds, tides have more impact on the currents in the southern SoG, between the San Juan and Gulf Islands. The depth of the wind driven circulation in SoG is determined by the depth of the halocline (2-10m, (LeBlond, 1983)). Mean surface currents (Fig. 1) are weak (~ 0.2m/s) implying no simple mean surface circulation, so it is hard to predict where drifters will be advected. Surface temperature in March – April in the northern part of the SoG varies between 7-11oC as measured at Chrome Is lighthouse, and in the southern SoG the temperature varies between 6-13oC.  8  Figure 1. Average surface currents in the Salish Sea. Currents, averaged over 15.5 days (in April 2007) to remove the tidal effect. PS denotes Puget Sound  Low frequency current fluctuations (~0.1m/s) in central SoG have been observed at 100, 200 and 290 m depth (Stacey et al., 1987). The fluctuations have a horizontal scale of about 10km and their amplitude e-folding time is about 4 days (Stacey et al, 1991). These currents have energy levels as high as those of the diurnal and semi-diurnal tides (Chang et al., 1976). The fluctuations are driven primarily by buoyancy forcing associated with the freshwater runoff and the effects of the tides and the wind stress are of secondary influence (Masson and Cummins, 2004). The observed currents were also reproduced with ROMS model used in this study (Fig  9  F11c). These are gyre like features with cross-strait currents associated with them. The recirculation pattern obtained with ROMS is close to the observations of Stacey et al. (1987).  1.5  Biological Oceanography of SoG The northern SoG is under the influence of weaker tidal mixing and receives freshwater  runoff from the northern inlets, resulting in a near-surface stratification, which is favorable for phytoplankton growth. Relatively high mean chlorophyll concentrations were measured over most of the northern SoG, especially between Texada Is and Vancouver Is (Masson and Pena, 2009). The central SoG is influenced by Fraser River plume, where the water column is strongly stratified. Higher chlorophyll concentrations were measured along the mainland coast north of the Fraser mouth and in the estuarine plume (ibid). Strong tidal mixing in Juan De Fuca Strait and around the Gulf and San Juan Islands leads to weak near-surface stratification. The measured phytoplankton biomass there is about half of the amount measured in SoG (ibid). Higher chlorophyll concentrations in SoG than in Juan De Fuca Strait are associated with relatively higher zooplankton abundance in SoG than in JDF (Mackas et al, 1980). The SoG zooplankton community is characterized as small copepods and nauplii (Harrison et al, 1983) which serve as food for herring and hake larvae (Blaxter, 1962; King et al, 2006). As there is more phytoplankton in the northern and central parts of SoG than in the southern part or Juan De Fuca Strait (Masson and Pena, 2009), it seems likely that there is more food available for the larvae in these regions.  10  1.6  Models Used in This Study There is a large variety of coupled bio-physical models that can be used to model larval  transport (Miller, 2007). At the time this project started, the Regional Ocean Modeling System (ROMS) had been implemented for the Salish Sea and it was convenient to use it. Lagrangian TRANSport model (LTRANS) was chosen because of its compatibility with ROMS. ROMS also has the ability for tracing particles but LTRANS uses less computational resources than ROMS. ROMS simulates the 3-D ocean circulation in the Salish Sea and LTRANS interpolates three components of water velocities from ROMS output and follows particles in three dimensions. A more detailed description of the models is given in Appendices A and B, but here, only the strong and weak aspects of the models are mentioned. A strong feature of ROMS is that it reproduces the estuarine circulation of the Salish Sea and the deep currents in the central SoG correctly. A weak side of ROMS (and subsequently LTRANS) is that ROMS does not resolve velocities near shore well because its 1 km grid has no enhanced resolution close to the coasts. Herring larvae studies using grid with enhanced resolution at the coasts (e.g. Fox and Aldridge, 2000) had particular area(s) along the shore of greater importance than other parts of their study domain but in this study we are more concerned with larval transport in the Salish Sea and not specifically near the coasts. Another disadvantage of ROMS is that it does not resolve surface waves or beach surf. LTRANS strong points are its computational inexpensiveness, and the ability of the model to assign different types of particle behavior such as floaters at different depths or particles performing vertical migration. A disadvantage of LTRANS is that it does not simulate larval mortality. Growth is implicitly modeled as particles start swimming faster as time progresses. In this study, we are interested in the transport of the surviving larvae by the ocean  11  currents and, for this reason, resolving velocities near the coast, or simulating surface waves, or larvae mortality are less important. Therefore the combination of the two models is perfectly suitable for our purposes.  12  Chapter 2: Modeling herring and hake larval dispersal in the Salish Sea  2.1  Introduction Pacific herring and hake are not only commercially important fish in SoG (DFO 2011,  Hake Assessment 2012) but they are also key components of the SoG ecosystem. Herring provide food source for a variety of species, including halibut, flounder, and dogfish (Schweigert et al., 2010, Ware and Thomson, 2005). They are also believed to be an important part of the diet of seabirds and marine mammals, such as: sea lions, seals, porpoises, dolphins, and humpback whales. Hake feed predominately on euphausiids (King and McFarlane, 2006) and are a food source for seals (Olesiuk et al 1990), and other marine mammals and large fish. Most herring come into SoG in October-November to spend the winter and spawn in March and April. In early summer, they migrate to the west coast of Vancouver Island (Taylor, 1964). Herring spawn on vegetation in sheltered bays, just below the intertidal zone (Hay, 1985). The eggs adhere to the rocky substrate to which the vegetation is attached (Haegele and Schweigert 1985b). General spawning locations are roughly the same from year to year but within a general location, the precise site may vary between years. Incubation time is about 2 weeks (Hourston and Haegele, 1980). The larvae float at the surface for the first 6 days after hatching and then start performing diel vertical migration (dvm) (Hourston and Haegele, 1980). Diel vertical migration (type I) is a pattern of movement during which larvae move up to the surface at night and return to deeper waters during the day. The herring larvae dvm is a behavioral response to active pursuit of prey (Ferreira 2012). Herring mortality is greatest during the larval stage (Hourston and Haegele, 1980).  13  Mortality at this stage determines the abundance of the group at an adult age (ibid). In addition to predation, other factors, such as currents, may lead to loss of large numbers of larvae. Herring larvae depend on the ocean currents for retention into areas favorable to feeding and growth (Stevenson 1962), such as SoG. Higher chlorophyll concentrations in SoG than in Juan De Fuca Strait (JDF) (Mason and Pena, 2009) are associated with relatively higher zooplankton abundance in SoG than in JDF (Mackas et al, 1980). The SoG zooplankton shallow community is characterized as small copepods and nauplii (Harrison et al, 1983), which serves as food for young herring larvae (Blaxter, 1962). On the other hand, Pacific hake are recognized as the most abundant resident fish in SoG. They also spawn in March and April, all throughout SoG (McFarlane and Beamish, 1985) but the young larvae reside deeper in the water column - between 50 and 300m depth (Bailey, 1982). In SoG, herring larval distribution has been studied through extensive ichthyoplankton surveys conducted in April and May from 1989 to 1992 (Mason et al 1982a & b). Highest densities of herring larvae (mean age between 6-27 days) were found on the Vancouver Island side and in the north (Hay and McCarter, 1997). Lower densities were in the middle of SoG and mainland side (ibid). Almost no larvae were caught in the joining inlets draining into SoG (ibid). Very few larvae were found in the southern areas of SoG (ibid). Although, the ichthyoplankton surveys showed abundance and herring larval distribution in the Strait, they could not recognize the origin of the larvae, which is important for identifying key spawning grounds. Knowing larvae origins also allow us to determine where larvae from each spawning area are dispersed. The objectives of this project are to study herring and hake larval dispersal in the Salish Sea and identify the physical drivers for it. For this purpose, larvae were tracked for a few weeks in the Salish Sea using a regional circulation model and a particle tracking model. A similar  14  approach has been used to study, for example, oyster larvae in Chesapeake Bay (North et al, 2008), and harmful algal blooms in the Gulf of Maine (He at al, 2008). The particle tracking model does not simulate biological processes (such as growth and mortality), it only focuses on following the 3D trajectories of particles which imitate larval swimming. From this point, particles traced with the model will be referred as particles and not as larvae to avoid confusion.  2.2  Methods First, details on the two models employed in this project are given, including how the  models were configured, and the parameters used. Then the configuration of the particle tracking model is described for the two different types of larvae. 2.2.1  The Regional Ocean Modeling System (ROMS) The Regional Ocean Modeling System (Shchepetkin & McWilliams, 2005, 2008) is used  to model the three-dimensional circulation in SoG. It is a primitive equation, terrain-following model that has wide usage in regional and coastal circulation studies. (She and Klinck, 2000; Foreman et al, 2008; Di Lorenzo 2005). A brief general description of ROMS is included in Appendix A, section 1. A detailed description of the ROMS model implementation for the entire coast of British Columbia (including the Salish Sea) is given in Masson and Fine (2012). Here only a short summary is included. The model domain for the present study includes the Straits of Georgia and Juan de Fuca and Puget Sound (Fig.2). The model has open boundaries at the mouth of the Juan De Fuca Strait (on the southwest boundary) and the Johnstone Strait (to the north) and free-slip boundary conditions along the coast.  15  The model grid has 1km horizontal resolution (184370 cells). Bathymetry was extracted from a 500 m resolution hydrographic dataset and smoothed appropriately in order to minimize pressure gradient error. The grids axis is rotated counterclockwise approximately 38o so that the y-axis is aligned along the length of the SoG. There are 31 sigma vertical levels with finer resolution in the top 50m (Appendix A, section 3). The minimum model depth was set to 7m to avoid creation of mudflats during low tide.  Figure 2. Coastline of the Straits of Georgia (SoG) and Juan de Fuca, and release locations for herring and hake particles. GI indicates Gulf Islands, SJI – San Juan Islands, PS – Puget Sound. Triangles show particle release locations for herring sensitivity study and dots for the hake study. WV stands for Westview, CM – Comox, PH – Pender Harbor, HS – Howe Sound, NA – Nanaimo, FR – Fraser, CW – Cowichan, SA – Saanich, JDF – Juan  16  de Fuca. Hake release spots: MB – Montgomery Bank, TX – Texada (next to Texada Island), NA – Nanaimo, HB – Halibut Bank. The model domain is from the mouth of Juan De Fuca to 50o N near the top of the figure. Herring spawn near the shore whereas hake spawn along the center of the Strait. The model was forced with daily short wave radiation and hourly wind data from observations gathered from multiple sites. Monthly atmospheric observations were used to compute the heat flux. Daily discharge and temperature from 20 rivers (including Fraser and Campbell Rivers) were used. Eight tidal constituents (K1, O1, P1, Q1, M2, S2, N2, and K2) are used to force the model at the lateral boundaries. Seasonally varying temperature and salinity profiles are used at the lateral boundaries of the model. The circulation model was initialized with uniform initial conditions (Practical Salinity of 30, 10oC temperature) and then run for the year of 2007. The conditions obtained after one year of spin up time were used as the initial conditions for a continuous 3-year run (2007-2009). The hydrographic data from ROMS was stored once every day at midnight. The conditions in late January of each year from the 3-year run were used as the initial state for the larvae drift study. More details about ROMS configuration specifically to this project are given in Appendix A, section 4. Although ROMS has the option to track particles, it is computationally very expensive to run the model multiple times. We have chosen the offline Lagrangian TRANSport model (LTRANS) as a more practical alternative for tracking particles.  2.2.2 2.2.2.1  Lagrangian TRANSport Model (LTRANS) Overview The Lagrangian TRANSport model (LTRANS) (North et al, 2011) is an offline particle-  tracking model which is driven by ROMS model output. LTRANS follows the path of particles  17  in three dimensions using advection, mixing, and behavior. Starting from ROMS grid locations, LTRANS interpolates water properties, such as: sea surface height, salinity, temperature, diffusivities and the three components of velocities to particle positions (Fig. B1). A 4th order Runge-Kutta scheme is used for particle advection. Particle motion due to subgrid turbulence is replicated by a random displacement model component. The Lagrangian transport model has the ability to incorporate different types of particle behavior such as drifting or performing diel vertical migration. 2.2.2.2  Concentration Maps Often in this study, LTRANS results are analyzed using particle concentration maps.  These maps are used to discern particle transport both across the strait (east-west) and along the length of the strait (north-south). In these maps, the ROMS domain is divided in 12 study areas (Fig. 3a). The number of particles contained in each area was divided by the total number of released particles in order to determine the probability density function (PDF) at any given time (Fig. 3b).  18  Figure 3. Analysis of model results. Results are analyzed based on the number of particles in each red box in a). Red areas in b) have more particles than the blue ones. a) Particle positions (black dots) after they have been floating for 4 weeks. Green color represents water and red boxes represent PDF areas. Particles were released in Comox. b) Concentration map for the same particle positions as in a), color represents PDFs. The ratio in the top right corner denotes the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca Strait.  2.2.2.3  Particle Average Paths Another way of analyzing LTRANS results is to plot weekly particle average paths as  follows:  xn =  1 N  N  ∑x i =1  i  y  n  =  1 N  N  ∑y i =1  i  (1)  where N is the total number of particles, ( x n , y ) show the weekly averaged position, n  and (xi,yi) are particle positions at week n. Circles representing weekly standard deviation or the dispersion from the average position were also calculated.  2.2.3 2.2.3.1  Herring Larvae in LTRANS Base and Sensitivity Studies In the Base study, actual spawn dates and locations for the years of 2007-2009 were  extracted from the herring database (DFO - Herring Database) (Fig. 4). The most herring eggs were laid in 2007, 39% of the total amount of eggs laid during the three years (DFO - Herring Database), and presumably with the most larvae hatched. The least amount of eggs was laid in 2008 (28%). In all three years, most of the larvae (78-91%) hatched in Comox (SA14). Nanaimo (SA17) contributed 6-15% of the larvae in SoG. 0-1% of the larvae came from Cowichan (SA18) and 3-7% from Saanich (SA19). The first eggs hatch in Comox (SA14) in all three years. Egg  19  hatching started about 2 weeks earlier in 2008 than in the other two years and 2008 is also the year with the longest hatching season (March 10 – May 10). In the Base study, particles were released as close as possible to the actual spawning locations recorded in herring database (Fig. 4). In one statistical area (SA) there is more than one spawning location. Each herring spawning location is represented by a particle release location but one particle release location may represent more than one spawning location. Incubation times are calculated as described in section 2.2.3.3 for each spawning location and date. For this study, particles were assigned only one type of behavior – herring DVM (section 2.2.3.4). After the particles were tracked for 6 weeks, relative contributions of larvae particles coming from different statistical areas (Comox – SA 14, Nanaimo – SA 17, Cowichan – SA 18, and Saanich – SA 19) are calculated based on the recorded number of eggs for each spawning location. Particles “older” than 6 weeks are not considered in the distribution maps. In the Sensitivity study, the dependence of particle distribution on release location, start date, vertical swimming behavior, and interannual variability is explored. The particle release spots are the averaged spawning locations for each of the 9 statistical areas (DFO - Herring spawn tables) (Fig.2). The release date is 14 days after the weighted spawn date4, which is different for each area. Three types of behavior are studied: surface drifters – particles with this behavior drift at the surface; floaters at 10m – particles reside between 8m and 12m depth; and herring DVM – described below in Section 2.2.3.4. All particles in this study are tracked for 4 weeks. 2.2.3.2  Herring Spawn Locations and Particle Release Locations Although herring eggs occur near shore in or just below the intertidal zone (Hourston and  4  Spawn dates are weighted by Spawn Habitat Index (SHI). SHI represents the combined, long-term frequency and magnitude of spawns along each kilometre of coastline over time. The index is a measure of shoreline utilization by spawning herring. (Definition is taken from DFO website)  20  Haegele 1980; Stevenson, 1962), some of the spawning locations appear on land when plotted on the ROMS grid (Fig. 4) because the minimum depth in ROMS model was set to 7m and places shallower than 7m occur as land. Actual spawning locations which appear in a wet grid cell next to land also cannot be used as a particle release location, because (a) LTRANS considers half of the wet grid cell next to land as land (Appendix B, Fig. B1), (b) ROMS and consequently LTRANS velocities near shore do not represent real velocities near the coast correctly, and (c) neither of the two applications model surface waves and near shore surf. In such cases, the particles are released in the closest (to the actual spawn location) wet grid cell surrounded by water.  Figure 4. Actual herring spawn locations for each Statistical Area (SA). Circles indicate spawning locations for SA14 (Comox), diamonds – SA17 (Nanaimo), squares – SA18 (Cowichan), and triangles - SA19 (Saanich). Color indicates year: black – 2007, blue – 2008, red – 2009. Spawn locations may vary from year to year. Sensitivity studies investigated herring released from all sites shown in Figure 2.  2.2.3.3  Incubation Period and Particle Release Date All spawning records in the herring database are in March and April. Egg incubation  periods have been determined in the laboratory (Alderdice and Velsen 1971). Days to hatching (HT) is given by Hay (1990):  21  HT = 100/(0.765 + 0.437T + 0.024T2)  (2)  where T is water temperature. Incubation times were calculated for each spawning event and each site using the 15-day average of the temperature measured at the closest lighthouse (DFO – Lighthouse data; Appendix C). For the spawning site (Saanich) for which there is no lighthouse data available, the mean temperatures are taken from the ROMS model. On average, the difference between lighthouse data and model output is about 1oC (Fig. C2 in Appendix C) which leads to a maximum 3-day difference in incubation time. This method is used to determine larval incubation times in the base study (section 2.2.3.1) but for the sensitivity study (section 2.2.3.1) the incubation period is considered to be 14 days. 2.2.3.4  Vertical Migration Hatching occurs in the hours of darkness between 5pm and 7am (Alderdice and Velsen  (1971)) on a high tide and it is approximately simultaneous for eggs deposited in a single spawning. (Hourston and Haegele, 1980). Young herring larvae are found in the near-surface waters (0-10m) (Stevenson, 1962). The newly hatched larvae (6 days or younger) appeared to be strongly attracted to both daylight and faint light in the surface waters at night and their concentration is observed to be highest in the top 2m. Older larvae could be found in the surface 2m during the night but not during the day. After 6 days, the larvae start performing diel vertical migration (DVM), during which the animals swim downwards during the day and migrate back up to the surface at night (Hourston and Haegele, 1980). LTRANS was configured so that particles float at the surface between day 1 (hatching) and day 6, then they start performing DVM (Fig. 5). Particles float in the surface 2m at night and swim between 8m and 12m depth during the day. Herring DVM behavior is based on times of sunset and sunrise averaged over the tracking period. Effects on the intensity of light due to  22  cloud cover and water turbidity are neglected. This was added to LTRANS as behavior 104.  Figure 5. Particle vertical positions for LTRANS runs. 50 particles float at the surface for 6 days after they were released and then start performing diel vertical migration. During that time strong winds blew, induced strong mixing of the water column so that particles could not maintain their preferred depth, given their swimming speed. In addition to dvm, sensitivity studies investigated float trajectories of particles fixed at 10m depth and the surface.  2.2.3.5  Swimming Speeds In the laboratory, Atlantic herring larvae swim at speeds of 5-20mm/s for larvae about  10mm long (1 week old) (Batty, 1984). Swimming speed increased with length, larvae 22mm long (4-5 weeks old) have swimming speeds in the range 5-60mm/s. SoG herring larvae in the wild have cruising swimming speeds of 5-30mm/s, which does not seem to change with age (Von Westerhagen and Rosenthal, 1979). Larvae moved at a higher speed over several minutes: 10.5mm larvae moved at 27mm/s and 13.5mm larvae swam at 43mm/s (ibid). In both, sensitivity and base studies, LTRANS was configured so that particles initially swim slowly (< 0.02m/s), representing young larvae. Then the particle swimming speed increases linearly up to 0.06m/s, simulating larvae growing up.  23  2.2.3.6  Depth Maintaining Particles can be vertically displaced to unfavorable depths due to advection or diffusion.  However, the particles need to maintain their depths for all types of behavior (section 2.2.3.1). During each internal time step, if the particle does not float in its predefined depth range, described by the preferable depth (zp) ± the tolerance depth (ztol), LTRANS will move the particles into the desirable depth range.  2.2.4  Hake Larvae in LTRANS Hake eggs and larvae are abundant from March to May at depths of 150-300 m all  throughout the SoG (McFarlane et al, 1985). However, the densest congregations of eggs are located in south central SoG (Halibut Bank), northwest of Texada Island (Montgomery bank), and close to Nanaimo (ibid). There is also a suspected spawning location southeast of Texada Island (Iwamoto et al, 2004). Accordingly, in LTRANS, particles were released at four locations: Halibut Bank, Montgomery Bank, Texada and Nanaimo (Fig.2). Hake larvae are rare5 in the surface waters of the open Strait (McFarlane et al, 1985). Larvae were found as deep as 300m but the highest density was in the interval 200-250m during April; age was not specified (ibid). The vertical distribution of larvae < 5mm in lenght is similar in Puget Sound and off the coast of California (Bailey, 1982) (Appendix F, Fig. F1). However, the older (>5mm) Puget Sound animals move up in the water column, because the minimum depth there is only about 110m (Bailey, 1982). Since the highest density of SoG larvae is between 200-250m (McFarlane et al, 1985), it is considered in LTRANS that the particle vertical distribution resembles the vertical distribution of California larvae (Appendix F, Fig. F1), i.e.  5  There is a debate where hake larvae reside. According to D.E. Hay and B. McCarter from Pacific Biological Station in Nanaimo, BC, hake larvae are the second most abundant species (after herring) in the surface waters of SoG (pers. comm.). However, for the purposes of this study, hake larvae are considered to live deeper.  24  most of the larvae < 8mm reside between 51-100m depth, larvae between 8 and 12mm swim between 101 – 200m depth. In order to simulate hake behavior in LTRANS, the larval age in days (D) is necessary. It can be calculated from larval length in mm (L) using the following formulae (Bailey, 1982): L = 2.75 + 0.16D  (3)  for larvae less than 20 days old (or ~6mm long), and L = 1.72*exp(3.15*(1 – exp(-0.02624*D)))  (4)  for larvae older than 20 days (or longer than ~6mm) The swimming speeds for different larva sizes are estimated as (Miller et al, 1988): log10 Va = 1.07 log10 L – 1.11  (5)  where Va is the average speed, and L is the length of the larva. Thus estimated speeds of herring larvae were compared to the measured speeds (Batty, 1984; Von Westerhagen & Rosenthal, 1979). The estimated average speeds for herring larva are within the range of speeds discussed in section 2.2.3.5. Based on a good correlation between measured and calculated average swimming speeds for herring larvae, the average swimming speeds for hake larva are calculated using the formulae (3) and (4) to obtain the larval length based on its age, and then substitute the length in (5) to estimate the average swimming speed. Hake particles in LTRANS are released on March 1, March 15, April 1, April 15, May 1, and May 15 in each of the years 2007-2009 and they are tracked for up to 4 weeks. Particles with hake behavior (type 200) stay between 51m and 100m for the first 26 days, and then they sink to depths 100-300m until day 28. Particle swimming speed increases linearly from 0.02 m/s at the beginning to 0.08 m/s at the end of their tracking period.  25  2.2.5  Numerical Considerations Since the current velocities are interpolated, LTRANS does not resolve velocities near  the coast well. LTRANS considers the half of each grid cell next to land to be land (Fig.B1). If a particle crosses a horizontal or vertical boundary because of turbulence or advection, it is reflected off the boundary with an angle of reflection that equals the angle of approach to the boundary. The distance that the particle is reflected is equal to the distance that the particle surpassed the boundary. If the particle goes beyond the horizontal boundary due to behavior, it is placed just below the surface or close to the bottom (i.e. it is placed near the boundary). At open ocean lateral boundaries, particle movement is stopped and it is not tracked further. The length of LTRANS internal time step is determined as balance between a reasonable representation of larvae particle movements (shorter time step) and fewer computations (longer time step). Tests with particles performing DVM showed the highest sensitivity to the time step. These tests were performed at time when strong winds blew and induced mixing of the water column. Runs with time steps of 60s (Fig B7a) and 120s (Fig. 5) were similar whereas the 300s runs (Fig. B7b) showed less mixing of the particles. Therefore time step of 120s was used in this study. The frequency of ROMS output used by LTRANS is chosen based on two considerations: (a) particle concentrations converge as ROMS results are stored more and more often, and (b) computational cost is reduced when the output is less frequent. Different sets of ROMS output files (results stored at every 15min, 30min, 60min, and 120min intervals) were used to run LTRANS, for 4 weeks. The probability density functions (PDF) was calculated for each area. PDFs calculated for the 15min and 30min ROMS output are really close to each other  26  (i.e. PDFs converge) but the PDFs for 60min and 120min output are different. (Appendix A, Fig A9, A10) Therefore ROMS output stored every 30min was used. The average density of herring eggs in the spawning grounds is of the order of 105 eggs/ m2 (Hourston et al, 1980). However, tracking so many particles released at the same location is computationally very expensive and impractical. The PDF error is determined as the variance σ of a finite population of size N and it is given by:  σ= where  x  b  x (1 − x ) b  b  N  (6)  = K N is the PDF function; K is the number of particles inside the area; N is  the total number of particles. With 95% confidence, the PDF is within the range (PDF –2σ, PDF + 2σ). Maximum error occurs at a PDF of 0.5 and, with 6000 particles, the error is 0.0013. Since the PDF error for 6000 particles is small, this number of particles was released in each spawning location. 2.3  Results  2.3.1 2.3.1.1  Herring Particles Wind Impact on Herring Particle Distributions Particle distribution heavily depends on the wind storms (see videos; Fig. 6 and 7). Since  the herring particles spend the first 6 days at the surface, strong winds during those first few days of the tracking period influence particle distribution more than the wind storms later. Strong winds lasting few days have more impact on particle distribution than weeks of weak winds. Wind storms drive the significant variations in herring distributions from year to year for particles released at the same spot.  27  Figure 6. Correlation between winds and percentage of surface drifters washed out through Johnstone Strait. Hourly wind speed and direction is shown with black sticks on plots A) and C). Thick red sticks indicate squared wind speed averaged over 24 hours, which is proportional to the wind drag. Note the different scales. Plots B) and D) represent washed out particles in 2007 and 2009. Particles on both plots were released in Comox. Strong winds in 2007 blew out of SoG almost all Comox particles. In 2009 wind storms to the south helped particles to be retained in the Strait.  28  Figure 7. Particle positions are strongly influenced by the winds. There is a good correlation (with r2-value = 0.63) between hourly distances (in y direction) that particles traveled for the first 26 hours after their release, and the squared hourly wind speed in y direction for the same time period. Positive direction means north-west (along SoG). Particles were released near Nanaimo, and the winds were taken near Nanaimo in 2007, 2008 and 2009. Other effects on near surface currents are estuarine flow, mean flow, and tides. The latter are largely removed by the 26 h period. 2.3.1.2  Herring Base Study Particle distributions from the Base study in the beginning of May, 2007 and 2009 (Fig.  8) show that in 2007, most of the particles were washed out of the Salish Sea, whereas in 2009 the majority of the particles are retained in SoG. In 2007, the remaining particles are more uniformly dispersed in SoG than in 2009, when they congregate between Texada Is and Vancouver Is, and around the Gulf Is. In 2009, particles from Comox dominate in SoG but in 2007 particles from Nanaimo and Saanich are important, especially in the Southern SoG, around the San Juan Is, and in Puget Sound. In the beginning of May, 2008, particles are mostly concentrated around the Gulf and San Juan Is (Appendix E, Fig E2). Particles come from Nanaimo (mostly), Saanich and Cowichan. There are almost no particles from Comox in the central and south SoG. For more details see Appendix E, section 1.  29  Figure 8. Base study distribution maps for 2007 and 2009. Most 2007 particles were washed out of SoG and most of 2009 particles were retained in the Strait. In 2007, particles were more uniformly distributed throughout SoG and San Juan Islands and Puget Sound while in 2009 particles gathered between Texada and Vancouver Islands and around the Gulf Islands. Date in the title denotes 6 weeks after the first larvae hatched. Jet colors show PDFs. Black, white and grey colors indicate statistical areas in which particles were released. Each PDF area is designated a pie, specifying the relative number of particles originating from different statistical areas. The bottom left pie shows the relative amount of particles released in different statistical areas for the whole Salish Sea. The ratio in the title indicates the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca.  30  2.3.1.3  Comparison with Field Observations Herring larvae were sampled in SoG in two consecutive cruises during the spring of 2009  (Guan et al, In Prep). Field data is compared to model results on Fig. 9. Colored dots represent particle positions from the Base study. One red particle represents approximately 23 times more larvae particles than a yellow particle, and approximately 46 times more larvae than a blue one. Field data are plotted with triangles and the color scheme here is similar: blue triangles represent less larvae than green, green less than yellow and yellow less than red. Particle positions are plotted at noon on each of cruise days. The model results are in reasonable agreement with the field data. During day 1 of cruise 1, very little larvae were found in central SoG (dark blue triangles) and very few were sampled close to the shores (light blue triangles), which confirms models results showing almost no dots (particles) in the middle of the Strait, and a few particles near the shores. On day 2 of cruise 1, samples show relatively more larvae in the central part of Strait (yellow and green and light blue triangles), which contradicts model results – almost no particle positions were plotted there. On day 3 of cruise 1, model results and observation are in a very good agreement. The most larvae were sampled near Vancouver Island shores (red and green triangles), and a lot of particles were plotted there. On the other side, very little particles appear in the area with the blue triangles. On day 1 of cruise 2, the southern sampling locations (dark blue triangles) surrounded the patches of larvae positions shown by the model. Relatively more larvae were found in the central Strait (light blue triangles) which confirms model results showing some particles in that area. During day 2 of cruise 2, some larvae were found near the shore of Vancouver Island and in the central part of the strait (light blue triangles), these observations agree with the model results showing some particles in that areas and almost no particles around the dark blue triangles. On day 3 of  31  cruise 2, very few larvae were found (all triangles are blue) but according to the model, particles were concentrated in the sampled locations. Overall, model results and field observations show agreement in 4 out of 6 days. On the last day of cruise 2, the model shows larvae about 5 weeks old, where the observations show none. This disagreement is perhaps, due to mortality, not modeled here. However on day 2 of cruise 1, the model shows low concentrations of larvae where high concentrations were observed. The nearest high concentrations of larvae are approximately 30km to the northwest. The model simulations are not perfect, but do appear to be giving reasonable solutions within the accuracy expected by the sensitivity calculations.  32  Figure 9. Comparison between model results and field observations for herring particles. There is a good agreement between model results and field observations in 4 out of 6 days. The days in which model simulations and field data disagree are Apr 26, 2009 and May 1, 2009, see text. Particle positions (dots) from 2009 base study runs are compared to field observations (triangles). Dot color shows number of larvae represented by a particle in the model, and the color of the triangles displays number of herring larvae sampled at that location. Note that the range of the color scales is different for the field observations and model results.  33  2.3.1.4  Sensitivity to Start Position and Date Since particles in the herring larvae study are often released two grid cells apart from  their original spawn locations it is necessary to quantify the concentration error due to start position. For this purpose, batches of 6000 particles were released in pairs two grid cells apart in different parts of SoG (Appendix E, Fig E4), and tracked for 4 weeks. The PDF error due to particles being released 2 grid cells apart is 0.06 (Appendix E, Fig E5). In order to calculate error due to releasing particles 1 day apart, batches of particles were released in pairs in the same location but just one day apart. The maximum error was determined as 0.13 (Appendix E, Fig E6). Therefore PDFs are very sensitive to exact date and location. 2.3.1.5  Sensitivity to Vertical Swimming Behavior Behavior changes the particle distribution, but there is no single pattern (Fig. 10). Fewer  DVM particles in Comox were washed out through Johnstone Strait than particles at the surface or 10m depth. Concentrations of surface drifters and floaters at 10m in Nanaimo are somewhat opposite: most of the particles at the surface drifted southwards towards Gulf and San Juan Islands but most of the particles at 10m floated northwards. DVM distributions for particles in Nanaimo and Cowichan look like a combination of the distributions of the surface drifters and 10m floaters plus some dispersion. It seems that DVM behavior makes the particles disperse more (Comox and Cowichan) but that is not always the case (Nanaimo). In all three years, fewer particles with DVM from Comox, Nanaimo and Westview got washed out than surface drifters from the same areas (Fig. E7).The opposite is true for particles from Cowichan, Saanich, and JDF: less surface drifters got washed out than particles with DVM.  34  Figure 10. Distribution maps for different behavior types. Different behaviour of particles released in the same area, and at the same time, changes their distributions but there are no uniform patterns. Color represents PDFs. Title starts with the name of the release location (also shown as a red dot on the plot), followed by the behavior type in parenthesis. Tracking end date is on the second line in the title. Particles were tracked for 28 days. The ratio in the top right corner denotes the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca Strait.  Currents at different depths help to understand the variations of particle distributions for different types of behavior. Water velocities at different depths (2m, 6m, and 12m) display predominant northward currents near Comox (Fig. 11a). Currents near Nanaimo mostly flow southwest (Fig. 11b). The path of a 12m water parcel has bigger excursions, suggesting stronger  35  currents at that depth. The flow near Comox and Nanaimo is mostly uniform (Fig. 11 a, b) and Appendix D) while the flow near central SoG has shear: surface currents move south whereas 12m flow is eastwards. Flow at 6m depth seems to move south and then back north (Fig. 11c and Appendix D).  Figure 11. Progressive vector diagram of currents at different depths (2m, 6m, and 12m) near Comox (a), Nanaimo (b), and central SoG (c). Paths are plotted for the tracking period of Comox particles (a) and for the tracking period of Nanaimo particles (b) and (c). Currents near Comox flow mostly north, Nanaimo currents – mostly southwest, and there is shear in the currents in the top 12m of the central SoG. These diagrams explain the variation in the distributions in Fig. 10.  2.3.1.6  Interannual Variability for Sensitivity Study Surface drifters and particles with DVM were released in 9 different locations (Comox,  Westview, Pender Harbor, Nanaimo, Cowichan, Saanich, Juan De Fuca, Howe Sound, and Fraser), 14 days after the weighted spawn date in each of the years 2007 to 2009. Particles were tracked for 4 weeks (Fig.12-14, also Fig E8 – E10). The following trends were found: (1) In all years, particles from Comox, Westview, and Pender Harbor mostly disperse in the northern part of SoG or are washed out through Johnstone Strait, regardless of their behavior (Fig. 12, E8). (2) In all years, particles from Cowichan, Saanich and Juan De Fuca stay in the southern part or are washed out through the mouth of Juan De Fuca Strait (Fig.14, E10). (3) Particles from Nanaimo, 36  Howe Sound and Fraser can disperse either way (Fig.13, E9). (4) In general, there is cross-strait particle transport, i.e. particles from the east side get transported to the west side and vise versa. (5) Gulf and San Juan Islands seem to attract particles from Cowichan, Saanich, Fraser, and JDF (Fig.13, 14, E9, and E10).  Figure 12. Concentration maps for northern particles with DVM. Most of the northern particles stay in the northern part of the Strait or are washed out of SoG. Color shows PDFs. Plot title denotes the release location and tracking end date. Particles were tracked for 28 days. The ratio in the top right corner denotes the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca Strait.  37  Figure 13. Concentration maps for particles with DVM released in central SoG. Particles released in the central Strait can disperse in either direction: north or south along SoG. Color shows PDFs. Plot title denotes the release location and tracking end date. Particles were tracked for 28 days. The ratio in the top right corner denotes the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca Strait.  38  Figure 14. Concentration maps for southern particles with DVM. Most of the southern particles stay in the southern part of the Strait or are washed out through the mouth of Juan De Fuca Strait. Color shows PDFs. Plot title denotes the release location and tracking end date. Particles were tracked for 28 days. The ratio in the top right corner denotes the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca Strait.  2.3.2 2.3.2.1  Hake Particles Sensitivity to Start Position and Date Sensitivity tests for start position and date for hake particles were performed similarly to  herring particles (Section 2.3.1.4). The error due to releasing particles two grid cells apart was  39  found to be 0.17 (Fig. F2a). The error from releasing the particles one day apart was found to be 0.2 (Fig. F2b). 2.3.2.2  Sensitivity to Vertical Swimming Behavior Distributions for different behaviors are different (Fig. F3, F4): 200m floaters are most  concentrated and particles performing hake larval dvm are most dispersed. Also, hake larval particle distribution is more similar to 75m floater distributions than to 200m distribution. 2.3.2.3  Interannual Variability Hake particles do not show much interannual variability in comparison to the herring  results. Generally, the majority of particles released near Montgomery Bank and Texada, stay in the northern part of SoG (Fig. F5, F6). Montgomery Bank particles released early in the season (Mar 1-15, 2007, and Mar 1, 2008/09) can be dispersed as far south as the Gulf Is. The bulk of the particles released near Halibut Bank and Nanaimo (Fig. F7, F8) tend to stay in the Central SoG. Some particles from Nanaimo disperse in the northern Strait. Particles from all areas released at the beginning of the season disperse farther than particles released later in the season. No particles were washed out of the Strait during the four weeks of tracking time. Hake particle average paths are compared to herring paths to show that herring particles disperse further away from their release spot than hake particles (Fig. 15). Hake circles are much tighter and there is no significant change in their paths from year to year. For example, herring particles released near Nanaimo travel on average 71km, with standard deviation 48km. On the other side, hake particles released near Nanaimo traveled average distance of 18km with standard deviation 22km.  40  Figure 15. Herring and hake particle average paths. Circles represent the dispersion from the average position after week 1, 2, 3, and 4. Particle release date is shown. Particles were tracked for 28 days. Hake particles disperse less and travel less than herring particles, because hake are deeper and deeper currents are weaker than near surface currents.  2.3.2.4  Comparison with Field Observations Observational data from two cruises conducted in the spring of 2009 (Guan et al, in prep)  is compared to model runs (Fig. 16). During cruise 1 (April 25-27, 2009) most of the hake larvae were found near the west bank of the central SoG which disagrees with model results. However during the second cruise (April 29 - May 3, 2009) most hake larvae were found in the central strait and near the east bank of the strait, which agrees with number of particle simulated in central SoG for that period of time.  41  Figure 16 Comparison between field data and model results for hake particles. Particle positions (red and blue dots) are compared with larvae, sampled in two cruises. Triangles show cruise stations. Less hake larvae was found at the blue triangles, and most larvae at the red triangles. Red particles (dots) are 15 days older than the blue ones. Color scale show the number of hake larvae found in the field samples at each station. Not surprisingly, the comparison is not successful as there is a lack of information on exact hake larval spawn locations, and dates, and larval depth preferences.  42  2.4  Discussion This study is one of the first studies to simulate transport of herring and hake larvae in the  Salish Sea using hydrodynamic forcing from observations and incorporating larval behavior. Model results differ for both species because different, species-specific behavior and spawning grounds were incorporated into the model. In the base study, the timing of the hatching and the release location of herring larvae were simulated as close as possible to reality. In the sensitivity study, it was explored how herring distributions depend on particle start position, start time, and vertical swimming behavior. Interannual variability and congregation /disaggregation areas are discussed below. Distributions of hake and herring particles are also compared.  2.4.1  Herring Base Study The base study showed strong connection between the direction of wind storms during  the early life of herring larvae and their retention in the SoG, (discussed also in section 2.4.5 below). In 2007, the hatching period was short, and strong winds to the north washed most of the particles out of the Salish Sea. The percentage of particles washed out of the model domain might be slightly over-estimated, because LTRANS does not consider particles advected back into the Salish Sea. In 2008, the hatching period was longer than in the other two years, and wind storms to the north early in the season washed some of the particles out of the Strait. Later, winds shifted to the south, and particles released after week 5 were retained in the strait. In 2009, winds blew predominantly to the south and most of the particles stayed in SoG. Although Comox was the most productive area in all three years, strong winds washed out most of the Comox particles in 2007 and 2008, and particles from Nanaimo dominated the Strait. In 2009, the majority of Comox particles remained in the system, and they dominated everywhere.  43  The model results showing that the wind washed out most of the larvae in 2007 is consistent with the fact that the year of 2007 had the lowest catch rate for age 0+ herring (Thompson and Schweigert, 2009) and subsequently poor recruitment of the 2007 year class in 2010 (Cleary and Schweigert, 2011). Juvenile herring surveys in 2008 found the most herring from the three years (Thompson et al, 2009). Model results showing that in 2009 larvae stayed in the Salish Sea, were also verified by the juvenile herring surveys (Thompson et al, 2010). Since herring populations in SoG consist of several year-classes, periodical population fluctuations are not likely to cause starvation issues for species feeding on adult herring. On the other side, although several other forage fish species are present in SoG, none approach the abundance levels of herring, and species that feed on age 0+ juveniles may have been substantially affected by the 2007 small cohort (Therriault et al, 2009). The ichthyoplankton surveys conducted in SoG from 1989 to 1992 showed highest densities of herring larvae on the Vancouver Island side and in the north, lower densities were in the middle of SoG and mainland side (Hay and McCarter, 1997). Very few larvae were found in the southern areas of SoG (ibid). However, model results illustrate that the herring larval distributions are highly variable and there are no distinctive areas where larvae gather.  2.4.2  Comparison with Field Observations Herring base study for 2009 was used to compare model results to field observations and  there is a reasonable agreement between the two (Fig. 9). However, particles plotted in red happened to be on average “older” than the blue particles. A lot of larvae die during their first few weeks of life but mortality is not simulated in the particle tracking model. Therefore, the difference in the numbers of larvae that the red and blue dots represent may not be as high.  44  A similar comparison between model results and sampled hake larvae was not very successful. The plotted hake particles were released on April 1 and April 15, 2009 (Fig. 16). However, in reality, larvae hatch every day in March, April and the beginning of May (McFarlane et al, 1985). Also, particles were released in 4 points but in reality, larvae are abundant all throughout the central SoG (ibid). It was impractical to do more model runs in order to simulate the realistic scenarios, because the exact hatching dates and locations are not known, and the depth at which hake larvae reside is not agreed on either. Also, there is not enough data to calculate the number of hake larvae that each particle represents, as was done for herring particles.  2.4.3  Sensitivity to Exact Start Position and Date Particle distributions (for both larvae) are sensitive at the 10-20% level to exact release  date and position. Hake PDFs are more sensitive than herring distributions. However the conclusions of this thesis on retention and broad distributions are robust at this level of uncertainty.  2.4.4  Sensitivity to Vertical Swimming Behavior For both types of larvae, PDFs are sensitive to exact swimming depth. Currents at  different depths explain variation in distribution for particles with different behavior.  Herring Particles Herring distributions are very sensitive to the exact depth at which particles swim. The  45  majority of the Comox particles, (regardless of their swimming depth) are advected northwards (Fig. 10a,b,c), which is in agreement with the water parcel paths (Fig. 11a) showing prevailing northward currents in the top 12m of the water column. Southward currents near Nanaimo (Fig. 11b) carry surface particles south (Fig. 8d). However, some of the Nanaimo 10m floaters (Fig. 8e) probably “catch” flows in the central strait moving north and east (Fig. 11c). Most of the particles from Cowichan stay near the Gulf and San Juan Islands (Fig. 10 g,h,i), around which the flow is more complex and the distributions cannot be explained by looking at velocities at one point only. Also, DVM behavior helps the herring particles released in Comox, Nanaimo, and Westview to be retained in the Strait. (Fig.E7) The percentage of washed out particles from those areas drops significantly when the behavior changes from surface to DVM. DVM behavior does not affect the Cowichan, Saanich and JDF particles significantly. For example, 6% of dvm particles from Cowichan in 2008 were washed out compared to 3% of surface particles.  Hake Particles Water parcels at 75m and 200m depths were traced near Montgomery Bank (northern SoG) (Fig. F10). The 75m water parcels traveled much further than the 200m water parcels suggesting shear currents. The distribution of particles with hake behavior is more similar to the distribution of floaters at 75m depth than to the distribution of floaters at 200m depth (Fig. F3, F4).  2.4.5  Impact of Wind Direction on Retention for Herring Particles If the wind storms blow predominantly to the north during the spring season, it could be  46  expected that most of the larvae from Comox, Nanaimo, Westview, and possibly Howe Sound will get washed out through the Johnstone Strait. Winds to the north push larvae from Juan De Fuca towards the Canadian coast of the strait, where the ebb tides are stronger (Thomson, 1981), and they will get washed out through the mouth of Juan De Fuca Strait. Alternatively, if the wind storms blow predominantly to the south as in 2009, then most of the larvae will stay in the system. Winds to the south drive larvae in Juan De Fuca Strait towards the US side of strait where the flood tides are stronger (Thomson, 1981) and carry the larvae towards Puget Sound.  2.4.6  Interannual Variability of Hake Particles Hake larvae do not show much interannual variability compared to herring larvae because  hake larvae reside deeper in the water column where inter annual changes in circulation are weaker than in the wind driven surface layer. Since SoG act as an estuary, the surface layer flows out of the strait and the deep layer into the strait. Thus, hake (deeper) particles tend to remain in the Strait but many herring (shallow) particles are washed out of the Strait. In March, there is a current to the south (Fig. D4), which explains the southward advection of the northern hake particles at the beginning of the season (Fig. F5). In April and May, the current shifts. Tracks of 50 random particles (Fig. F11 a, b), released in the central, strait follow the two deep gyres (Fig. F11 c). The two baroclinic gyres are permanent features of SoG and have been observed by Stacey et al (1987). The gyres are present at 100, 200, and 290m depth and their speed is about 10cm/s (comparable to the tidal currents, Chang et al, 1976). These gyres cause quick East-West connection across the Strait. Although most of the herring northern particles stayed in the north and most of the southern ones stayed in the south, some particles were transported across and along the Strait.  47  This agrees with the genetics studies considering herring in SoG to belong to the same population structure (Beacham et al, 2001). Similarly, northern and central SoG are weakly connected for hake particles. 2.5  Conclusions ROMS and LTRANS models have been implemented for the Straits of Georgia and Juan  De Fuca. Herring and hake larval particles have been tracked for 3 consecutive springs (20072009). Particle concentrations have been calculated for different parts of the domain. Distributions for both types of particles are very sensitive to exact release date and location, as well as exact swimming depth. Diel vertical migration may be serving as a retention mechanism for herring larvae in the northern SoG. Herring base study results are in a reasonable agreement with field observations, showing agreement at 4 out of 6 snapshots. The comparison with hake was not so successful probably due to lack of initial data, as when and where hake spawn. Most of the herring particles released in the northern part of SoG stay in the north, and most of the remaining southern particles stay in the south. Although north and south parts of SoG are weakly connected, there is some transport of herring individuals along the strait and across the strait. Similarly, the majority of the hake particles, released in the north, stay in the north, and most of the ones released in central Strait stay in the central strait. Hake distributions do not show much interannual variability, because particles are affected by deep currents. Early in the spring, a current to the south advects some particles from northern and central SoG south. Two deep gyres in the central strait cause quick east-west transport across the strait, and make hake particles released in the central SoG gather towards its east side.  48  Because herring particles generally stay near the surface, wind is the main forcing which determines their distributions in the Strait of Georgia. If strong winds blow to the north during the hatching period (as it happened in 2007) herring larvae will be washed out of the Strait, leading to poor recruitment later. Alternatively, if the wind storms blow predominantly to the south (as in 2009), then most of the larvae will stay in the strait and very likely Comox larvae will dominate in the Strait. In 2008, strong winds to the north early in the season washed out Comox larvae, which hatched first. The winds shifted to the south later in the season allowing larvae from Nanaimo (dominant), Cowichan and Saanich to remain in SoG. Since the winds are the primary driver of herring larvae dispersal, there is a significant interannual variability and it is hard to identify congregation areas other than the Gulf and San Juan Islands.  49  Chapter 3: Conclusion Pacific herring and hake are a food source for people and play important roles in the Strait of Georgia ecosystem. Larvae of these species hatch during the spring - the season of change from winter to summer. Winter is associated with weak estuarine circulation in the Salish Sea, driven by low river discharge and strong winds blowing to the north causing downwelling over the continental shelf. Summer is associated with strong estuarine circulation driven by high river discharge, and winds to the south casing upwelling off the coast of British Columbia. Herring and hake larvae are dependent on ocean currents for transport into areas favorable for growth and survival. This study is one of the first Salish Sea studies to simulate transport of herring and hake larvae. In this project, an existing ROMS implementation for the Straits of Georgia and Juan De Fuca was coupled with LTRANS. ROMS simulated ocean circulation in the Salish Sea and LTRANS traced larval particles for 3 consecutive springs (2007-2009). Model results were used to map particle concentrations in different parts of the domain. Concentrations differ for both species because different, species-specific behavior and spawning grounds were incorporated into LTRANS. First, a sensitivity analysis was performed on particle release date and location. Herring PDF error due to releasing particles 2 gridcells away from the shore is 0.06, and the error due to releasing particles one day apart is 0.13. The corresponding hake errors are 0.17 and 0.2. Model results were very sensitive to the exact release date and location for both types of particles. This was important when reproducing real herring larvae hatching events. The herring particles had to be released as close to the spawning grounds as possible and the incubation time had to be calculated based on water temperature. Although PDFs are sensitive towards release date and location, important results are not affected. 50  Secondly, sensitivity tests for vertical swimming behavior were performed. Distributions for both types of particles are very sensitive to exact swimming depth due to vertical shear in the water currents. Different behavior changes distribution but there is no consistent pattern. Significantly less herring particles with DVM than surface drifters, released in the northern SoG, were washed out through Johnstone Strait. Behavior may be serving as a retention mechanism for herring larvae in the northern SoG. Since, plankton is abundant in the northern part of SoG this area is a good habitat for larvae to grow in. Next, simple interannual variability was explored in order to determine areas of aggregation or disaggregation. Although results showed large interannual variability in the distributions of herring particles, there are certain trends: most of the particles released in the northern part of SoG stayed in the north, and most of the southern particles stayed in the south. However, some particles were transported along SoG and across SoG, which is in agreement with genetics studies, qualifying SoG herring as the same stock. In all 3 years, herring particles from southern SoG seemed to gather around the Gulf and San Juan Islands. Opposite to herring particles, hake distributions do not show much interannual variability. Hake particles from the central SoG congregated in the central SoG and particles from the northern SoG stayed in the northern SoG. Since herring particles reside in the top 12 m of the water column, their dispersal is strongly influenced by the winds. Model simulations showed that if strong winds blow to the north during the hatching period (as it happened in 2007), herring larvae from the northern part of SoG will be washed out through the Johnstone Strait, and larvae hatched in Juan De Fuca will be pushed towards the Canadian coast of the strait, where the ebb tides are stronger, and the currents will carry the larvae to the open ocean. Alternatively, if the wind storms blow predominantly to the south (as in 2009), then most of the larvae from northern SoG will stay in 51  the Salish Sea, and larvae hatched in Juan De Fuca strait will be pushed to the US coast of the strait, where the flood tides are stronger, and the currents will advect the larvae towards Puget Sound. Because herring larvae distribution depends on the winds, and there is no simple mean path that the particles take, it is really hard to predict where the particles go. As hake particles swim deeper in the water column (50-200m), their distributions are shaped by deep currents. In most years, early in the spring, a current to the south advected some hake particles from northern and central SoG south. A pair of deep gyres in the central strait cause quick east-west transport across the strait, and make hake particles released in the central SoG gather towards its east side. Herring base study for 2009 was used for comparison with field observations. Spawning locations and dates were taken from observations, and incubation times were calculated based on water temperature. Model simulations for herring particles were in reasonable agreement with field observations, showing agreement at 4 out of 6 days of fieldwork. Comparison between modeled distributions and observed hake abundance was not so successful probably due to lack of initial data, as when and where hake spawn. Simulations from herring base study showed that, in 2007, most larvae in SoG were washed out, which was in agreement with herring juvenile surveys performed later that year, reporting the lowest juvenile catch rate. In 2008, strong winds to the north early in the season washed out Comox larvae, which hatched first. The winds shifted to the south later in the season allowing larvae from Nanaimo (dominant), Cowichan and Saanich to remain in SoG. In 2009, predominant winds were to the south and most of the larvae were retained in the Salish Sea. Comox larvae were dominant in almost all parts of the strait. There is not one single spawning area that is important for herring larvae in SoG, larvae from different spawning grounds are important in different years. 52  This study brings some useful insight into herring and hake larval dispersal, fishes of significant importance in the Salish Sea, improving our understanding of larval dispersal in SoG and the drivers for it. Comprehensive understanding of factors influencing larval transport will inform conservation decisions and may ultimately provide the scientific basis for adaptive management strategy. The full inputs to this strategy remain to be determined, but its implementation will help to better protect important fisheries in British Columbia. Model results from this study showed that the current combination of ROMS and LTRANS works reasonably well for herring larvae in the Salish Sea. There are a few options for future work using the two models, such as: run the current configuration for more years to better determine patterns of larval distributions, or develop behaviors for other larval species and study their distribution in SoG, or incorporate feeding and mortality in LTRANS. However, my preference would be to use ROMS, which is already developed for the whole British Columbia coast (Masson and Fine, 2012), and study larval dispersal in British Columbia coastal waters, outside the Salish Sea. This will give us a more complete picture of herring larval transport in BC waters.  53  Bibliography Alderdice D.F. and F. P. J. Velsen. 1971. Some effects on salinity and temperature on early development of Pacific herring (Clupea pallasi). J. Fish. Res. Bd. Canada 28: 1545-1562. Bailey, K.M. 1982. The early life history of the Pacific Hake, Merluccius Productus. Fish. Bul: 80(3), pp 589-598 Bartsch, J. 1993. Application of a circulation and transport model system to the dispersal of herring larvae in the North Sea, Cont Shelf Res, Vol.13, No.12, pp. 1335-1361 Batty, R.S. 1984. Development of swimming movements and musculature of larval herring (Clupea harengus). J. exp. Biol: 110, pp 217-229 Baumann, H., P. Pepin, F.J.M. Davidson, F. Mowbray, D. Schnack, and J.F. Dower, 2003. Reconstruction of environmental histories to investigate patterns of larval radiated shanny (Ulvaria subbifurcata) growth and selective survival in a large bay of Newfoundland. – ICES Journal of Marine Science, 60: 243–258. Beacham, T., J. Schweigert, C. MacConnachie, K.D. Le, K. Labree and K.M.Miller, 2001. Population structure of herring (Clupea pallasi) in British Columbia: an analysis using microsatellite loci. Canadian Science Advisory Secretariat, Research Document 2001/128. Bissett, W.P., J.J. Walsh, D.A. Dieterle, K.L. Carder, 1999a: Carbon cycling in the upper waters of the Sargasso Sea: I. Numerical simulation of differential carbon and nitrogen fluxes, Deep-Sea Res., 46, 205-269 Blaxter, J.H.S., 1962. The feeding of herring larvae and their ecology in relation to feeding. CalCoFI Rep., vol. 10, 1962 Budgell, W.P., 2005: Numerical simulation of ice-ocean variability in the Barents Sea region, Ocean Dynamics Chang, P., S. Pond, and S. Tabata, 1976. Subsurface currents in the Strait of Georgia, west of Sturgeon Bank. J. Fish. Res. Board Can. 33, 2218–2241 Cleary, J.S., and J.F. Schweigert, 2011. Stock Assessment and Management Advice for the British Columbia Herring Stocks: 2010 Assessment and 2011 Forecasts., Canadian Science Advisory Secretariat, Research Document - 2011/115 Cowen, R.K., and S. Sponaugle 2009. Larval dispersal and marine population connectivity, Annu Rev. Mar. Sci., 1, pp. 443–466 Crawford, W. R. and G.S. Jamieson, 1995. Modeling advection of Dungeness crab (Cancer magister) in Dixon Entrance and northern Hecate Strait, British Columbia. Alaska Sea Grant Report, 96-02, pp 489-506  54  de Young, B., Anderson, J., Greatbatch, R.J. 1994 Advection-diffusion modeling of larval capelin (mallotus-villosus) dispersion in Conception Bay, Newfoundland, Can. J. Fish. Aquat. Sci. 51 (6), pp 1297-1307 Dickey-Collas, M., Bolle, L.J., van Beek, J.K.L., Erftemeijer, P.L.A., 2009. Variability in transport of fish eggs and larvae. II. Effects of hydrodynamics on the transport of Downs herring larvae. Marine Ecology Progress Series 390, 183–194. DFO - Herring spawn tables: http://www.pac.dfo-mpo.gc.ca/science/species-especes/pelagicpelagique/herring-hareng/herspawn/tabsfram-eng.htm DFO - Herring Database: Fisheries and Oceans Canada. Pacific Region. Institute of Ocean Sciences Data Archive. Ocean Sciences Division. [Internet]. Available from http://www.pac.dfo-mpo.gc.ca/science/species-especes/pelagic-pelagique/herringhareng/herspawn/tabsfram-eng.htm DFO - Lighthouse data: Fisheries and Oceans Canada. Pacific Region. Institute of Ocean Sciences Data Archive. Ocean Sciences Division. [Internet]. Cited 2011-03-14. Available from: www.pac.dfo-mpo.gc.ca/science/oceans/data-donnees/lighthouses-phares/indexeng.htm DFO. 2011. Stock Assessment Report on Pacific Herring in British Columbia in 2011. DFO Can. Sci. Advis. Sec. Sci. Advis. Rep. 2011/061. Available from http://www.dfo-mpo.gc.ca/csassccs/Publications/SAR-AS/2011/2011_061-eng.pdf Di Lorenzo, E., M. G. G. Foreman, and W. R. Crawford (2005) Modeling the generation of Haida Eddies, Deep Sea Res., Part II, 52, 853–873. Fox, C.J., and J.N. Aldridge, 2000. Hydrographic circulation and the dispersal of yolk-sac herring (Clupea harengus) larvae in the Blackwater Estuary, J. Mar. Biol. Ass. U.K., 80, 921-928 Fennel, K., J. Wilkin, J. Levin, J. Moisan, J. O'Reilly, and D. Haidvogel, 2006: Nitrogen cycling in the Middle Atlantic Bight: Results from a three-dimensional model and implications for the North Atlantic nitrogen budget, Global Biogeochem. Cycles, 20, GB3007, doi:10.1029/2005GB002456 Ferreira, A. S. de A., E. K. Stenevik, K. W. Vollset, R. Korneliussen, A. Folkvord 2012. Vertical migration of Norwegian spring-spawning herring larvae in relation to predator and prey distribution Mar Biol Res, 8:7 Foreman, M. G. G., W. Callendar, A. MacFadyen, B. M. Hickey, R. E. Thomson, and E. Di Lorenzo (2008), Modeling the generation of the Juan de Fuca Eddy, J. Geophys. Res., 113, C03006, doi:10.1029/2006JC004082. Fox, C.J., P. McCloghrie, E.F. Young, R.D.M. Nash, 2006. The importance of individual behaviour for successful settlement of juvenile plaice (Pleuronectes platessa L.): a modelling and field study in the eastern Irish Sea. Fish. Oceanogr., 15, pp. 301–313 55  Gallego, A., E.W. North, P. Petitgas, 2007. Introduction: status and future of modelling physical–biological interactions during the early life of fishes. Mar. Ecol. Prog. Ser., 347, pp. 122–126 Guan, L. Dower, J. F., S. M. McKinnell, P. Pepin, and E. Pakhomov (In prep). A comparison of spring ichthyoplankton assemblages in the Strait of Georgia (British Columbia, Canada): the early-1980s and the late 2000s. Haegele, C.W., and J.F. Schweigert. 1985b. Distribution and characteristics of herring spawning grounds and description of spawning behavior. Can. J. Fish. Aquat. Sci. 42 (Suppl.1):39-55. Hake Assessment 2012. Status of the Pacific hake (Whiting) stock in U.S. and Canadian Waters in 2012. Report of International Joint Technical Committee for Pacific hake. Available from: http://www.pcouncil.org/wp-content/uploads/Hake_2012_Assessment.pdf Harrison, P., J., J. D. Fulton, F. J. R. Taylor, and T. R. Parsons. 1983. Review of the biological oceanography of the Strait of Georgia: pelagic environment. Can. J. Fish. Aquat. Sci. 40: 1064 - 1094 Hay, D.E. 1985. Reproductive biology of Pacific herring (Clupea harengus pallasi). Can. J. Fish. Aquat. Sci. 42 (Suppl. 1):111-126. Hay, D.E. 1986. Effects of delayed spawning on viability of eggs and larvae of Pacific herring. Trans. Am. Fish. Soc. 115:155-161. Hay, D.E., M.C. Healy, L.J. Richards, J.B. Marliave, 1989. Distribution, abundance, and habitat of prey fishes in the Strait of Georgia. In: Vermeer, K. and R.W. Butler (Eds.) The ecology and status of marine and shoreline birds in the Strait of Georgia, British Columbia. Ottawa: Canadian Wildlife Service Special Publication. pp 37-49 Hay, D. E. 1990. Tidal influence on spawning time of Pacific herring (Clupea harengus pallasi). Can. J. Fish. Aquat. Sci. 47: 239C2401. Hay, D.E. and McCarter, P.B. (1997) Larval distribution, abundance, and stock structure of British Columbia herring. Journal of Fish Biology: 51 (Supplement A) pp 155-175. He, R., D. J. McGillicuddy Jr., B. A. Keafer, and D. Anderson (2008). Historic 2005 toxic bloom of Alexandrium fundyense in the western Gulf of Maine: 2. Coupled biophysical numerical modeling, J. Geophys. Res., 113, C07040 Hedstrom, K., 2011, Hong Kong ROMS Workshop. (https://www.myroms.org/wiki/index.php/HK_2011) Lecture 14: MPI parallel programming in ROMS. Helbig, J.A., and P. Pepin, 2002. The effects of short space and time scale current variability on the predictability of passive ichthyoplankton distributions: an analysis based on HF radar observations. Fish Oceanogr 11:175-188 56  Hourston, A. S. and C. W. Haegele. 1980. Herring on Canada’s Pacific Coast. Can. Spec. Publ. Fish. Aquat. Sci. 48:23 p. Iwamoto, E., M.J. Ford, R.G. Gustafson. 2004. Genetic population structure of Pacific hake, Merluccius productus, in the Pacific Northwest. Environmental Biology of Fishes 69: 187– 199 Incze, L., H. Xue, N. Wolff, D. Xu, C. Wilson, R. Steneck, R. Wahle, P. Lawton, N. Pettigrew, Y. Chen, 2010. Connectivity of lobster (Homarus americanus) populations in the coastal Gulf of Maine: part II. Coupled biophysical dynamics, Fish. Oceanogr., 19(1), pp 1-20 King, J.R. and G.A. McFarlane. 2006. Shift in size-at-age of the Strait of Georgia population of Pacific hake (Merluccius productus), CalCOFI Rep., 47, pp. 111–118 LeBlond, P. H. 1983. The Strait of Georgia: functional anatomy of a coastal sea. Can. J. Fish. Aquat. Sci. 40: 1033–1063. Mackas, D. L., G. C. Louttit, and M. J, Austin. 1980. Spatial distribution of zooplankton and phytoplankton in British Columbian coastal waters. Can. J. Fish. Ayuat. SCB. 37:14761487. Marinone, S. G. and J. Fyfe. 1992. Residual currents in the central Strait of Georgia, B.C. Atmos.-Ocean, 30, 94–119. Marinone, S. G., S. Pond and J. Fyfe. 1996. A three-dimensional model of tide and wind-induced residual currents. Estuar. Coast. Shelf Sci., 43, 157–182. Mason, J.C., O.D. Kennedy and A.C. Phillips (1982a & b) The Strait of Georgia ichthyoplankton survey, 1981. Ichthyoplankton cruises 1-8. Can. Data Rep. Fish. Aquat. Sci. 360: 245 p and 364: 245 p. Masson, D. and Cummins, P.F. 2004. Observations and modeling of seasonal variability in the Straits of Georgia and Juan de Fuca. J. Mar. Res., 62, 491–516 Masson, D. and Fine, I. 2012. Modeling seasonal to interannual ocean variability of coastal British Columbia. J. Geophys. Res., 117, C10019, 14 pp. Masson D., and Pena, A., 2009. Chlorophyll distribution in a temperate estuary: The Strait of Georgia and Juan de Fuca Strait. Estuarine, Coastal and Shelf Science 82: 19–28 McFarlane, G.A. and R.J. Beamish. 1985. Biology and fishery of Pacific whiting, Merluccius productus, in the Strait of Georgia. Mar. Fish. Rev. 47: 23–34 Megrey, B. A., and S. Hinckley, 2001. Effect of turbulence on feeding of larval fishes: a sensitivity analysis using an individual-based model, ICES Journal of Marine Science, 58: 1015–1029  57  Miller, T.J., L.B. Crowder, J.A. Rice, and E. A. Marschall. 1988. Larval size and recruitment mechanisms in fishes: Toward a conceptual framework. Can. J. Fish. Aquat. Sci. 45: 16571670 Miller, T.J., 2007. Contribution of individual-based coupled physical–biological models to understanding recruitment in marine fish populations. Marine Ecology Progress Series 347, 127–138. North, E. W., Z. Schlag, R. R. Hood, M. Li, L. Zhong, T. Gross, and V. S. Kennedy. 2008. Vertical swimming behavior influences the dispersal of simulated oyster larvae in a coupled particle-tracking and hydrodynamic model of Chesapeake Bay. Marine Ecology Progress Series 359: 99-115 North, E. W., Gallego, A., Petitgas, P. (Eds). 2009. Manual of recommended practices for modelling physical–biological interactions during fish early life. ICES Cooperative Research Report No. 295. 111pp. North, E. W., E. E. Adams, S. Schlag, C. R. Sherwood, R. He, S. Socolofsky. 2011. Simulating oil droplet dispersal from the Deepwater Horizon spill with a Lagrangian approach. AGU Book Series: Monitoring and Modeling the Deepwater Horizon Oil Spill: A Record Breaking Enterprise Olesiuk, P.F., M.A. Bigg, G.M. Ellis, S.J. Crockford, and R.J. Wigen. 1990. An assessment of the feeding habits of harbour seals (Phoca vitu7ina) in the Strait of Georgia, British Columbia, based on scat analysis. Can. Tech. Rep. Fish. Aquat. Sci. 1730: 135 p. Pawlowicz, R, O. Riche, and M. Halverson, 2007. The circulation and residence time of the strait of Georgia using a simple mixing - box approach, Atmos-Ocean, 45:4, 173-193 Peck, M.A., and M. Hufnagl 2012. Can IBMs tell us why most larvae die in the sea? Model sensitivities and scenarios reveal research needs. J. Mar. Sys. 93, 77–93 Pineda, J., J.A. Hare, and S. Sponaugle. 2007. Transport and dispersal in the coastal ocean and consequences for population connectivity, Oceanography, 20 (3), pp. 22–39 Powell T. P., C. V. W. Lewis, E. N. Curchitser, D. B. Haidvogel, A. J. Hermann, E. L. Dobbins (2006), Results from a three-dimensional, nested biological-physical model of the California Current System and comparisons with statistics from satellite imagery, J. Geophys. Res., 111, C07018, doi:10.1029/2004JC002506. Robinson, C.L.K., Morrison, J., Foreman, M.G.G. 2005. Oceanographic connectivity among marine protected areas on the north coast of British Columbia, Canada, Can. J. Fish. Aquat. Sci. 62(6) 1350-1362. Schlag, Z. R., E. W. North, and K. A. Smith. 2008. Larval TRANSport Lagrangian model (LTRANS) User’s Guide. Technical Report of the University of Maryland Center for Environmental Science Horn Point Laboratory. Cambridge, MD. 147 p. 58  Schweigert, J. F., Boldt, J. L., Flostrand, L., and Cleary, J. S. 2010. A review of factors limiting recovery of Pacific herring stocks in Canada. – ICES Journal of Marine Science, 67: 000– 000. Sentchev, A. and K. Korotenko, 2007. Modelling distribution of flounder larvae in the eastern English Channel: sensitivity to physical forcing and biological behaviour, Mar. Ecol.-Prog. Ser., 347 (2007), pp. 233–245 Shchepetkin, A. F., and J. C. McWilliams, 1998. Quasi-monotone advection schemes based on explicit locally adaptive dissipation, Monthly Weather Rev., 126, 1541-1580. Shchepetkin, A. F., and J. C. McWilliams. 2005. The Regional Ocean Modeling System: A splitexplicit, free-surface, topography following coordinates ocean model, Ocean Modelling, 9, 347-404. Shchepetkin, A. F., and J. C. McWilliams. 2008. Computational kernel algorithms for fine-scale, multi-process, long-time oceanic simulations. In: Handbook of Numerical Analysis: Computational Methods for the Ocean and the Atmosphere, eds. R. Temam & J. Tribbia, Elsevier Science, ISBN-10: 0444518932, ISBN-13: 978-0444518934. She, J., and J. M. Klinck, 2000. Flow near submarine canyons driven by constant winds, J. Geophys. Res., 105(C12), 28,671–28,694, doi:10.1029/2000JC900126. Song, Y. and D. B. Haidvogel, 1994. A semi-implicit ocean circulation model using a generalized topography-following coordinate system, J. Comp. Phys., 115 (1), 228-244. Stacey, M. W., S. Pond, P. H. LeBlond, H. J. Freeland and D. M. Farmer. 1987. An analysis of the low-frequency current fluctuations in the Strait of Georgia, from June 1984 until January 1985. J. Phys. Oceanogr., 17, 326–342. Stacey, M. W., S. Pond, P.H. LeBlond, 1991. Flow dynamics in the Strait of Georgia, British Columbia, Atmos-Ocean, 29 (1), pp 1-13 Stevenson, J.C. 1962. Distribution and survival of herring larvae (Clupea pallasi Valenciennes) in British Columbia waters. J. Fish. Res. Bd. Canada 19:735-810. Taylor, F.C.H. 1964. Life history and present status of British Columbia herring stocks.0020Bulletin of the Fisheries Research Board of Canada 143: 81p. Therriault, T.W., D.E. Hay, J.F. Schweigert, 2009. Biologic overview and trends in pelagic forage fish abundance in the Salish Sea (Strait of Georgia, British Columbia). Marine Ornithology 37: 3–8. Thompson, M., and J. Schweigert, 2009. Strait of Georgia Juvenile Herring Survey, September 2007, Can. Manuscr. Rep. Fish. Aquat. Sci. 2871 Thompson, M., C. Fort, and J. Schweigert, 2009. Strait of Georgia Juvenile Herring Survey, September and October 2008, Can. Manuscr. Rep. Fish. Aquat. Sci. 2908 59  Thompson, M., C. Fort, and J. Schweigert, 2010. Strait of Georgia Juvenile Herring Survey, September and October 2009, Can. Manuscr. Rep. Fish. Aquat. Sci. 2921 Thomson, R.E. 1981. Oceanography of the British Columbia coast. Can. Spec. Publ. Fish. Aquat. Sci. 56: 291 p. Tully, J.P. and A.J. Dodimead. 1957. Properties of the water in the Strait of Georgia and influencing factors. J. Fish. Res. Board Can.14: 241-319. Vikebø, F.B, Å. Husebø, A. Slotte, E.K. Stenevik, and Vidar S. Lien, 2010. Effect of hatching date, vertical distribution, and interannual variation in physical forcing on northward displacement and temperature conditions of Norwegian spring-spawning herring larvae, ICES J. Mar. Sci. 67(9): 1948-1956 Vikebø FB, B. Ådlandsvik, J. Albretsen, S. Sundby, E.K. Stenevik, G. Huse, E. Svendsen, T. Kristiansen, E. Eriksen, 2011. Real-Time Ichthyoplankton Drift in Northeast Arctic Cod and Norwegian Spring-Spawning Herring. PLoS ONE 6(11): e27367 von Westerhagen, H., and H. Rosenthal, 1979. Laboratory and in-situ studies on larval development and swimming performance of Pacific herring, Clupea harengus pallasi. Helgo. Wiss. Meeresunters. 32539-549. Waldichuk, M. 1957. Physical oceanography of the Strait of Georgia, British Columbia. J. Fish. Res. Board Can. 14(3): 321–486. Ware, D.M., and R. E. Thomson, 2005. Bottom-Up Ecosystem Trophic Dynamics Determine Fish Production in the Northeast Pacific. Science, 308 (5726), 1280-1284. Warner, J. C., C. R. Sherwood, R. P. Signell, C. K. Harris, and H. G. Arango, 2008. Development of a three-dimensional, regional, coupled wave, current, and sedimenttransport model, Computers & Geosciences, 34, 1284-1306 Werner, F.E., Cowen, R.K., Paris, C.B., 2007. Coupled biological and physical models present capabilities and necessary developments for future studies of population connectivity. Oceanography 20 (3), 54–69 Yao, T., S. Pond, and L. A. Mysak. 1985. Profiles of low-frequency subsurface current fluctuations in the Strait of Georgia during 1981 and 1982, J. Geophys. Res., 90(C4), 71897198 Yin, K., R.H. Goldblatt, P.J. Harrison, M. A. St. John, P.J. Clifford, R.J. Beamish, 1997. Importance of wind and river discharge in influencing nutrient dynamics and phytoplankton production in summer in the central Strait of Georgia, Mar. Ecol. Prog. Ser., 161: 173-183  60  Appendix A Regional Ocean Modeling System (ROMS) A.1  Brief general description of the model ROMS solves the incompressible, hydrostatic, primitive equations with a free sea surface,  horizontal curvilinear coordinates, and a generalized terrain-following vertical coordinate which can be constructed to increase resolution near the surface or near the bottom (Song and Haidvogel, 1994). The predictive variables are surface elevation, barotropic (fast) and baroclinic (slow) horizontal velocity components, and water properties such as temperature and salinity. ROMS incorporates several coupled models for biogeochemical (Powell et al., 2006; Fennel et al., 2006), bio-optical (Bissett et al., 1999a), sediment (Warner et al., 2008), and sea ice (Budgell, 2005) applications, which were not used here. The primitive equations for momentum are approximated using a mode-splitting method which allows the separation of the barotropic and baroclinic components in the model. The mode-splitting is done using a time-stepping scheme, which is constrained to sustain volume conservation and to maintain the constancy of the tracers (such as temperature and salinity) (Shchepetkin and McWilliams, 2005). The time-stepping is a leapfrog/Adams–Moulton, predictor–corrector scheme, which is third-order accurate in time and very robust. This scheme allows larger time steps, by a factor of about four, which compensates for the increased cost of the predictor-corrector algorithm. Also, the model includes a variety of advection schemes: second- and forth-order centered differences; and third-order, upstream biased. We used the model default scheme - third-order, upstream biased (Shchepetkin and McWilliams, 1998). ROMS is a modular code written in F90/F95 and uses C-preprocessing to activate the various physical and numerical options. The code can be run on either serial or parallel computers.  61  A.2  Vertical sigma levels in SoG Sigma-coordinate system is a stretched vertical coordinate system which essentially  "flattens out" the bottom (Fig A1b). Sigma levels are the vertical layers in a sigma-coordinate system. In our case, in order to enhance resolution at the surface, the sigma-levels at the surface have smaller thickness than the bottom ones. Also, there are more sigma-levels at the surface than the bottom.  Fig. A1. Vertical profile with sigma levels. a) Transect in SoG, for which the depth profile and the sigma vertical levels are shown in b).  A.3  ROMS configuration specifically for this study ROMS for SoG was configured by Diane Masson (a similar model implementation for  the entire BC coast is described in detail in Masson and Fine, 2012). However there are a few differences specific for this study: (a) temporal range of the output files and (b) frequency of the stored output data, (c) initial conditions, (d) the processing option - PERFECT_RESTART was enabled, and (e) the model was run on a different server. Details follow: 62  (a) Prior to this study, D. Masson completed continuous 3 year run (2007-2009). However, in this project, only ROMS results for three months (March through May) in each of the three years was required. Since larvae hatch in the spring, they are heavily dependent on the ocean currents during that time of the year. (b) D. Masson stored hydrographic data once a day, while for this study it was necessary to record current velocities more often, in order to resolve the tidal effect. It was chosen to store ROMS output data every 30min (section 2.2.5). (c) In order to initialize ROMS runs for the spring of each year, history files, generated by D. Masson, were used as initial conditions files. However, those files do not contain all necessary parameters for the model to perform the “perfect restart”, i.e. to continue from where it had stopped and produce the same results as if the model had run continuously. Instead, ROMS interpolates the missing information and a comparison between velocities in the interrupted case versus continuous run shows that the velocities in the two cases differ in magnitude and direction. To work around that problem, a history file for late January in each year, generated by D. Masson, was used as initial conditions file for each year. The model was run for 30 days with the PERFECT_RESTART pre-processing option enabled. This period allows the model sufficient time to adjust from non-perfect restart and produce results much closer to D. Masson’s. ROMS results generated by D. Masson and files produced for this study were compared as follows: time series of surface current velocities and surface height at four points in SoG (Fig. A2) were constructed (Fig. A3). Only the plots for the year of 2007 are shown but similar assessments were made for the other two years. According to Fig. A4 more than 80% of grid cell velocities have a difference less than 0.17m/s. The maximum surface height difference is less than 25cm (Fig. A6). Another reason, for small differences occurring in the computed values, is that ROMS was run on a different server than the original model, configured by D. Masson. 63  Fig.A2. Four points for which time series of surface current velocities between D. Masson’s configuration and this study were constructed.  Fig.A3. Comparison of the magnitudes of surface current velocities between D. Masson (dots) and this study (line).  64  Fig.A4. Differences in surface currents velocities between ROMS results done for this study and D. Masson. Ten bins were organized as: bin1 has the count of all velocities having magnitude difference between 0 and d/10, …, bin 10 has the count of all velocities with magnitude difference between 9d/10 and d, where d is the largest difference observed.  Fig.A5 Surface height comparison. Dots – D. Masson, line – ROMS output generated for this study  65  Fig.A6. Differences in surface height for the whole domain between D. Masson and this study. Color bar represents difference in meters.  A.4  ROMS rho (density) grid The model variables are computed on a staggered Arakawa C grid, where state variables  (such as rho) and velocities are defined on staggered points (Fig. A7). Density is evaluated between points where the currents are calculated.  66  Fig. A7. ROMS staggered horizontal grid (or rho-grid). Figure is taken from https://www.myroms.org/wiki/index.php/Numerical_Solution_Technique  A.5  Frequency of ROMS output Fig. A8 and Fig. A9 show comparison between probability density functions (PDFs)  calculated for different sets of ROMS output files (data stored every 15min, 30min, 60min, and 120min) and for different groups of particles (1000, 2000, and 4000 particles). PDFs will converge as ROMS output becomes more and more frequent. The PDFs for the 120min and 60min output are quite different than the rest of the PDFs. This suggests that storing data every 60min or 120min is not suitable.  67  Fig. A8 PDFs calculated for different frequency of ROMS output and different number of particles.  68  Fig. A9. PDFs calculated for 15min, 30min, and 1h ROMS output and 1000, 2000, and 4000 particles.  Appendix B Lagrangian TRANSport model (LTRANS) B.1  Interpolation of some water properties LTRANS reads hydrographic data from ROMS output and interpolates water properties  at the particle location. Current velocities are interpolated from the original ROMS grid points: u grid points are used to calculate u-velocity, v grid points are used for v-velocity, and rho (density) grid points are used for sea surface height, w-velocity, salinity, and diffusivity calculations (Schlag et al. 2008) (Fig. B1). Because the current velocities are interpolated,  69  LTRANS does not resolve velocities near the coast well. (Explanation about ROMS rho grid is included in Appendix A, section 4).  Fig. B1 Boundaries and horizontal velocities interpolated from ROMS predictions. Figure adapted from Schlag et al.(2008)  B.2  Flow diagram of LTRANS LTRANS has an external and an internal time step. The external time step is the time step  of ROMS output (e.g. 30min). The internal time step is the time interval during which particle movement is calculated (e.g. 2 min). The internal time step is non-adaptive and smaller than the external time step thus preventing particles from moving in large jumps that could cause inconsistencies between predictions of the hydrodynamic model and the particle tracking model. (Fig. B2). Fig. B2 shows a schematic of LTRANS algorithm. After LTRANS initializes and reads the first ROMS output file, particle position is renewed based on movement due to advection, turbulence and behavior. Particle location is updated in LTRANS time step cycle as many times as it is necessary until it is time to read the next ROMS output file (e.g. 30min / 2 min = 15 times). The process of loading hydrographic data and updating the particle location based on 70  advection, turbulence and behavior is re-iterated until ROMS data for the whole particle tracking period is read.  Fig.B2. Flow diagram of LTRANS. Figure adapted from Schlag et al. 2008.  B.3  Determining LTRANS internal time step by comparing numerical and analytical  solutions In order to choose an appropriate internal time step, numerical solution of diffusion equation was compared to the analytical solution in the following way: batches with different number of surface drifters (1000, 2000, 4000, 6000, 8000 and 10000 particles in each batch) were released in the middle of SoG and they were let float for 28 days. Water velocities were set to zero, so the particles changed position only due to turbulent diffusion. LTRANS was run with five different time steps (60s, 120s, 300s, 600s, and 1200s) for each of the batches. The longest distance d that the particle floated away from the release location was divided by 10 and ten  71  concentric circles were drawn around the release location. (Fig.B3) The smallest circle has radius d/10, the next circle: 2d/10, etc…, the biggest circle is of radius d.  Fig. B3. Particle locations after floating for 28 days. Particles were released at the center of the 10 concentric circles. The radius of the biggest circle is equal to the longest distance that a particle traveled. The smallest circle’s radius is 1/10 of the biggest circle radius.  Particle concentration for each band of the circles was calculated as the number of particles in each band divided by the total number of particles. Concentrations determined numerically for batches of 6000, 8000 and 10000 particles are shown on Fig. B4, B5, and B6, where different color denominates different LTRANS internal time step.  Particle concentrations determined numerically should converge to the analytical solution of the diffusion equation:  The analytical solution for the 2D case is:  72  Where C is the concentration; Dx, Dy, Dz are the diffusivity constants in x, y, z direction; M/L is the number of particles; (xo, yo) are the coordinates of the release location: (x, y) are the particle coordinates after time t. Concentrations determined analytically for batches of 6000, 8000 and 10000 particles are plotted on Fig. B4, B5, and B6 with red line.  The following figures (Fig. B4, B5, and B6) and tables (Table B1, B2, B3) show that time steps of 300s or less are converging to the analytical solution.  73  Fig.B4. Comparison between the numerical and analytical solutions for particle concentrations. LTRANS was run with different internal time steps for a batch of 6000 particles. The bottom plots enlarge particle concentration in band 1, 2 and 3.  Fig.B5. Comparison between the numerical and analytical solutions for particle concentrations for a batch of 8000 particles.  74  Fig.B6. Comparison between the numerical and analytical solutions for particle concentrations for a batch of 10000 particles.  75  Table B1: Concentration differences between the analytical and numerical solutions for a batch of 6000 particles. Band 1 is the most inner circle; Band 10 is the most outer band. IDT (s)  Band 1  Band 2  Band 3  Band 4  Band 5  Band 6  Band 7  Band 8  Band 9  Band 10  1200  4.71E-06  5.41E-05  9.00E-05  9.42E-05  6.98E-05  4.23E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  600  1.36E-05  4.90E-05  9.31E-05  9.29E-05  6.98E-05  4.23E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  300  5.19E-06  4.81E-05  9.25E-05  9.45E-05  7.00E-05  4.24E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  120  1.77E-05  5.20E-05  9.01E-05  9.42E-05  6.92E-05  4.22E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  60  7.62E-06  5.20E-05  9.38E-05  9.46E-05  6.98E-05  4.21E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  Table B2: Differences in concentrations between the analytical and numerical solutions in each band of circles for a batch of 8000 particles. IDT (s)  Band 1  Band 2  Band 3  Band 4  Band 5  Band 6  Band 7  Band 8  Band 9  Band 10  1200  4.71E-06  5.41E-05  9.00E-05  9.42E-05  6.98E-05  4.23E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  600  1.36E-05  4.90E-05  9.31E-05  9.29E-05  6.98E-05  4.23E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  300  5.19E-06  4.81E-05  9.25E-05  9.45E-05  7.00E-05  4.24E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  120  1.77E-05  5.20E-05  9.01E-05  9.42E-05  6.92E-05  4.22E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  60  7.62E-06  5.20E-05  9.38E-05  9.46E-05  6.98E-05  4.21E-05  2.24E-05  1.06E-05  4.56E-06  1.77E-06  Table B3: Concentration differences between the analytical and numerical solutions for a batch of 10000 particles. IDT (s)  Band 1  Band 2  Band 3  Band 4  Band 5  Band 6  Band 7  Band 8  Band 9  Band 10  1200  9.33E-06  6.52E-05  1.12E-05  1.18E-05  8.72E-05  5.29E-05  2.80E-05  1.33E-05  5.70E-06  2.21E-06  600  1.40E-05  6.12E-05  1.17E-05  1.16E-05  8.72E-05  5.29E-05  2.81E-05  1.33E-05  5.70E-06  2.21E-06  300  3.33E-06  6.25E-05  1.15E-05  1.18E-05  8.75E-05  5.30E-05  2.80E-05  1.33E-05  5.70E-06  2.21E-06  120  1.34E-05  6.56E-05  1.14E-05  1.18E-05  8.63E-05  5.28E-05  2.80E-05  1.33E-05  5.70E-06  2.21E-06  60  2.35E-06  6.39E-05  1.16E-05  1.18E-05  8.74E-05  5.26E-05  2.80E-05  1.33E-05  5.70E-06  2.21E-06  76  B.4  LTRANS internal time step - particles with DVM In order to determine the appropriate time step, test runs with three different time steps  (60s, 120s and 300s) were performed (Fig. B7 and Fig. 5). In all three tests, particles float for the first six days and then they start performing DVM. During that time strong winds blow and induce mixing of the water column.  Fig.B7 Particle vertical position for LTRANS runs with internal time step 60s (a) and 300s internal time step (b).  B.5  Sufficient number of particles released in one location In order to establish the sufficient number of particles, the same algorithm was used  again: a few subsequent LTRANS runs were made with the same set of ROMS output files (data stored every 30min) and each time the number of particles is increased. PDFs for each of the runs were calculated and Fig.B8 displays that the PDFs converge for batches with higher number of particles (6000 and above). Therefore it is appropriate to release 6000 particles in one batch.  77  Fig. B8. PDFs calculated for different number of particles.  B.6  Changes in LTRANS (version 2) relevant to this study There are a few things changed in LTRANS, version 2, in order to fix bugs in the code or  to better adapt parts of the code for this study. (a) Parameter_module.f90, line 16: include clause should contain the correct path to LTRANS.h (b) The following parameters were not defined in parameter_ module.f90: readDens, constDens.  The  code  which  uses  readDens  and  constDens  was  disabled  in  hydrodynamic_module.f90. (Lines: 666, 840, 1281, 1305-1307) (c) LTRANS returned the error “Jumped over a u element” when a particle started floating in one grid cell channel. The model could not recognize the "wet" 1 grid cell channels. (Fig. B9). The model used to return a similar error (“Jumped over a v element”) when a particle was crossing east-west one grid cell channel. To work around that problem, the parts of the code, 78  which checked if a particle is in a u or v grid cell with at least one “wet” corner, were blocked. Lines 854 through 857 are commented in LTRANS.f90 file. Only the check if a particle is in a “wet” rho grid cell was left.  Fig. B9. Particle crossing into one grid cell channel. Wet rho grid cells represented with green color, white color denotes land; blue dot - particle’s previous location; red dot particle’s current location; magenta and black dots - u nodes.  (d) Particles move up and down due to behavior after they have been advected. Lines 1120-1122 in LTRANS.f90: IF (Behavior.NE.0) CALL behave(Xpar,Ypar,Zpar,Pwc_zb,Pwc_zc,Pwc_zf, P_zb,P_zc,P_zf,P_zetac,par(n,pAge),P_depth,P_U,P_V,P_angle,  &  &  n,it,ex,ix,ix(3)/DBLE(86400),p,bott,XBehav,YBehav,ZBehav)  were moved down further, right before: if(Behavior == 7)then…. (e) Behavior type 102 added – particles float at certain predefined depth zp plus or minus tolerance depth ztol. For example, particles with such behavior and zp = 10m, ztol = 2m will float between 8 and 12m depth.  79  Appendix C Lighthouse data C.1  Lighthouse positions and herring particle release spots  Fig.C1. Map of particle release locations (red dots) and lighthouses (blue dots).  Temperature data, for calculating incubation time for particles released nearby Comox, was taken from the lighthouse on Chrome Is; for particles released close to Nanaimo – from Entrance Is.; and particles in Cowichan area – from Active Pass. C.2  Comparison between ROMS output and lighthouse temperature data ROMS temperature series from the surface layer (Mar, April 2009) at the locations of  Chrome Is, Entrance Is, and Active pass lighthouses were plotted on Fig C2 and compared to the lighthouse measurement. The thickness of ROMS surface layer at Chrome Is. is 0.29m, at Entrance Is. is 0.37m, and Active Pass is 0.54m. On average, the difference between lighthouse data and model output is about 1oC. That leads to maximum 3 days difference in incubation time.  80  Fig.C2 ROMS versus lighthouse temperature. ROMS temperature is plotted with line and lighthouse measurements with dots.  Appendix D Vertical profiles of horizontal velocities D.1  Surface currents Depth profiles of the along-shore velocities (v velocities) near Comox (Fig. D1) and  Nanaimo (Fig. D2) show that there is very little or no shear. On the other side, depth profiles of the horizontal velocities at a point in the middle of SoG (Fig. D3) show vertical shear and stronger currents than near Nanaimo and Comox.  81  Figure D1. Along-shore velocities near Comox. Currents were de-tided by applying a low pass filter.  Figure D2. Along-strait velocities near Nanaimo. Currents were de-tided by applying a low pass filter. Positive means north-west (along the strait).  Figure D3. Along-strait velocities in central SoG. Currents were de-tided by applying a low pass filter. Positive means north-west (along the strait).  82  D.2  Deep currents Depth profiles of the along-shore velocities (v velocities) near Montgomery Bank (Fig.  D4) show that there is shear. In comparison to the surface in the previous section, deep currents are not as strong. Deep currents are ~0.2m/s and surface currents reach ~1m/s.  Figure D4. Along-strait velocities near Montgomery Bank. Currents were de-tided by applying a low pass filter. Positive means north-west (along the strait).  Appendix E Herring particles E.1 Base Study E.1.1  Herring particle distributions in 2007 During week 1 and 2 after the first release day, the particles were concentrated mostly in  the north-west part of SoG and with some of them around the Gulf Is (Fig. E1, plot A). By the end of week 3, the majority of the particles moved north of Texada Island, and 28% of them were washed out through Johnstone Strait. (Fig. E1, plot B) Up to that week, particles only from Comox and Nanaimo were present in the Strait. During week 4, 59% of the particles were washed out and new particles were released in Saanich (Fig.E1, plot C). Most of the particles were concentrated north-east of Texada Island. The relative contribution of Saanich particles was greatest around Gulf and San Juan Islands. Throughout weeks 5 and 6 (Fig.E1, plot D), up to 74% of the particles were washed out from SoG, and all of them originated in Comox. Particles 83  remaining in the Strait were uniformly distributed. Comox particles outnumbered Nanaimo particles in the northern and central strait. In the southern SoG and Juan De Fuca approximately 1/3 of particles came from each of the statistical areas (Comox, Nanaimo and Saanich). In 2007 no reported spawning occurred in Cowichan.  Figure E1. Base study distribution maps for 2007. Jet colors show PDFs, particle distributions on the date in the title. Black, white and grey colors indicate statistical areas in which particles were released. Each PDF area is designated a pie, specifying the relative number of particles originating from different statistical areas. The bottom left pie shows the relative amount of particles released in different statistical areas for the whole Salish Sea. The ratio in the title indicates the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca. 84  E.1.2  Herring particle distributions in 2008 For the initial two weeks, only larvae particles from Comox were present in the Strait and  they dispersed in the northern SoG (Fig. E2, plot A). 18% of the particles were washed out through the Johnstone Strait. During week 3, larvae started hatching in Nanaimo and even more particles (60%) were washed out of the system (Fig. E2, plot B). The bulk of the remaining particles concentrated in the northern part of SoG. Most of the particles in the Central and South SoG originated from Nanaimo (SA17). During week 4 and 5, 77% of all the particles were washed out (Fig. E2, plot C). The leftover particles gathered in the NE part of SoG or around the Gulf Is. Most of the particles in the central and southern SoG and JDF had started in Nanaimo (SA 17). Cowichan (SA18) particles were released, and their relative contribution is noticeable in the areas including Gulf and San Juan Is. During week 6, 7, and 8, new particles were released in Nanaimo, Cowichan and Saanich (Fig. E2, plot D). The particles were most concentrated around the Gulf Is. Up to week 7, Comox particles dominated in the Northern Strait but then they were outnumbered by Nanaimo particles. In the central Strait, Nanaimo particles dominated Comox particles. In the Southern SoG and JDF, particles originated from Nanaimo, Cowichan and Saanich and Nanaimo particles had the greatest relative contribution. After wk6, 99% of the washed out particles came from Comox and 1% from Nanaimo. After wk8, 91% of the washed out particles came from Comox and a little bit less than 9% came from Nanaimo, less than 1% came from Cowichan.  85  Figure E2. Base study distribution maps for 2008. Jet colors show PDFs, particle distributions on the date in the title. Black, white and grey colors indicate statistical areas in which particles were released. Each PDF area is designated a pie, specifying the relative number of particles originating from different statistical areas. The bottom left pie shows the relative amount of particles released in different statistical areas for the whole Salish Sea. The ratio in the title indicates the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca.  E.1.3  Herring particle distributions in 2009 During week 1 and 2, particles were released in Comox and Nanaimo (Fig.E3, plot A).  The majority of the particles gathered on west side of SoG. Interestingly, all particles in Puget Sound came from Comox, and all particles in JDF came from Nanaimo. Nanaimo particles had greatest relative contribution around Gulf Is. Until week 3, no particles were washed out. During 86  week 4, new particles were released in Nanaimo and Cowichan. 2% of the particles were washed out (Fig.E3, plot B). In week 5, particles from Saanich were released. Cowichan and Saanich particles had the greatest relative contribution around the Gulf Is (Fig.E3, plot C). At the end of week 6, only 12% of the particles were washed out of the system (Fig.E3, plot A), and 99% of the washed out particles came from Comox, and 1% from Nanaimo. From week 2 to 6, the particles were the most concentrated in two areas: between Texada and Vancouver Is, and around the Gulf Is. Comox particles dominated in all PDF areas, except around the Gulf Is.  87  Figure E3. Base study distribution maps for 2009. Jet colors show PDFs, particle distributions on the date in the title. Black, white and grey colors indicate statistical areas in which particles were released. Each PDF area is designated a pie, specifying the relative number of particles originating from different statistical areas. The bottom left pie shows the relative amount of particles released in different statistical areas for the whole Salish Sea. The ratio in the title indicates the percentage of particles washed out through Johnstone Strait versus the mouth of Juan De Fuca.  E.2 E.2.1  Sensitivity studies Sensitivity to start position (2 grid cells apart study) Batches of 6000 particles were released in pairs two grid cells apart in different parts of  SoG (Fig.E4).  Fig.E4 Batches of particles released in pairs in different locations in SoG and Juan De Fuca Strait. Batches in each pair are 2 grid cells apart.  After tracking the particles for 4 weeks, the concentrations (PDFs) for each batch and each PDF area were calculated. ∆C is the difference in concentrations between the batches of particles in one pair for one PDF area. The “2 grid cell” error is 0.06 (Fig E5).  88  Fig.E5 Estimate error for herring PDFs in “Two grid cells apart “study. <C> denotes mean particle concentration in one PDF area, and ∆C is the difference in concentrations between the batches of particles in one pair (released two grid cells away) for one PDF area.  E.2.2  Sensitivity to start date (One day apart study) Batches of particles were released in pairs in the same location but just one day apart.  Four pairs were released in 3 different locations: north and south of Comox and close to Comox. The maximum error from releasing the particles one day apart is 0.13 (Fig E6).  Fig.E6 Estimate error for herring PDFs due to releasing particles one day apart. <C> denotes mean particle concentration in one PDF area, and ∆C is the difference in concentrations between the batches of particles in a pair (released one day apart) for one PDF area.  89  E.2.3  Sensitivity to vertical swimming behavior In all three years, fewer particles with DVM from Comox, Nanaimo and Westview got  washed out than surface drifters from the same areas (Fig E7).The opposite is true for particles from Cowichan, Saanich, and JDF: less surface drifters got washed out than particles with DVM.  Fig E7. Percentage of surface drifters and particles with DVM washed out through Johnstone Strait and/or mouth of Juan De Fuca Strait  E.2.4  Interannual variability for the sensitivity study  (Please see videos) The goals of this section are to describe the wind impact on herring dispersal for particles released in different spots in the Salish Sea, as well as to portray the interannual variability in particle distributions. 2007 Fraser particles, released on Mar 19th, 2007, were pushed towards the shore in Boundary Bay, by 90  winds to the north for the first 10 days of their tracking period. A shift in the winds to the south, on Mar 29th, allowed them to spread out in southern SoG and San Juan Islands. On the same day, Comox particles were also released. Between April 1st and 8th, strong winds to the south forced Fraser, Comox, and just released Nanaimo particles to the south. Nanaimo particles dispersed around the Gulf Is. Most of the Fraser particles floated around San Juan Is and some of them were driven into Juan De Fuca strait. On April 8th, winds shifted to the north, moving Comox, and recently released Howe Sound and Pender particles to the north. During the next two big wind events (around April 13th and 16th) strong winds to the north washed out almost half of Comox and some Westview particles north, into Johnstone Strait. Some of the Nanaimo particles were also moved north but majority stayed around the Gulf Is. Cowichan and Saanich particles were released on April 15th and Juan De Fuca particles on April 16th. Shortly after that, the winds in the southern SoG and JDF blew to the south. The majority of Cowichan and Saanich particles were retained around San Juan Islands. JDF particles were moved towards the US side of JDF strait where the flood tides are stronger (Thomson, 1981), and the currents transported them towards San Juan Is and Puget Sound. Around April 24-27, strong winds to the north washed out most of Comox and Westview particles and almost half of Howe Sound ones. At the end of their tracking periods, most of the Fraser particles were still in Boundary Bay, the majority of Comox, Westview and Howe Sound particles were washed out, Pender particles did not disperse very much, most of Nanaimo and Saanich particles could be found around the Gulf Islands; and Cowichan and JDF particles ended up in Puget Sound.  2008 Until March 27th, 2008, strong winds to the north pushed Fraser particles (released on Mar 18th) into central and northern SOG and Comox particles (released on Mar 21st) into northern SoG. 91  After a week of weak and shifty winds, predominantly to the south, another storm blew to the north between April 4th and 7th. It forced the newly released Nanaimo particles to move quickly to the northern part of SoG. During that storm, 79% of Comox and 95% of Westview particles were washed out of the system. Fraser particles were dispersed everywhere in the straits of Georgia and JDF. The recently released Pender and Howe Sound particles stayed close to their start locations. Winds shifted to the south, and became strong between April 14th and 18th. During that time Cowichan and JDF particles were released. On April 19th, the JDF particles left their release location, a sheltered cove, and started floating in JDF strait. The winds shifted to the north and kept the particles close to the northern shore of JDF strait until April 23rd, where the ebb tides are stronger (Thomson, 1981), and the water currents carried them towards the open ocean. Then the winds changed to the south pushed Cowichan particles into San Juan Is and eastern JDF strait and kept Saanich particles around San Juan Is. At the end of their tracking periods, the majority of Comox, Westview and JDF particles were washed out of the Salish Sea, most Nanaimo particles were advected to the Northern SoG, Fraser particles were scattered in central and South SoG and around San Juan Is. Howe Sound and Pender particles did not disperse too much.  2009 Shifting winds at the end of March spread Fraser particles around the Gulf and San Juan Islands. On April 2nd Comox and Westview particles were released, and during the next few days strong winds to the southwest pushed the particles towards the Vancouver Is coast. Between April 9th and 13th strong winds to northeast washed out about 40% of Westview particles. During that period, particles from Howe Sound, Pender and Nanaimo were released, and were forced to move into the northern part of SoG. Between April 23rd and 30th, strong winds to southwest 92  drove Comox, Nanaimo, Westview , Pender, and some Howe Sound particles towards Vancouver Is and the east side of Texada Is. Most of the particles were retained in the system (except the Westview ones with over 70% of them washed out). In the meantime Saanich, JDF, and Cowichan particles started. On April 22nd, JDF particles were blown to the southern part of JDF strait, where the currents advected them towards Puget Sound and San Juan Is. At the end of their tracking periods, most of the Comox, Nanaimo and Pender particles were advected to the northern part of SoG, most of Westview particles were washed out, the majority of Fraser particles were retained around San Juan Is, particles from Cowichan and Saanich stayed around the Gulf and San Juan Is, JDF particles ended up in Puget Sound, Pender particles did not disperse much.  93  Fig. E8 Concentration maps for northern surface drifters. Color shows PDFs. Plot title denotes the release location and tracking end date. Particles were tracked for 28 days. Ratio in the top right corner indicates percentage of particles washed out through Johnstone Strait (north) versus mouth of Juan De Fuca (south).  94  Fig E9. Concentration maps for surface drifters released in central SoG. Color shows PDFs. Plot title denotes the release location and tracking end date. Particles were tracked for 28 days. Ratio in the top right corner indicates percentage of particles washed out through Johnstone Strait (north) versus mouth of Juan De Fuca (south).  95  Fig E10. Concentration maps for southern surface drifters. Color shows PDFs. Plot title denotes the release location and tracking end date. Particles were tracked for 28 days. Ratio in the top right corner indicates percentage of particles washed out through Johnstone Strait (north) versus mouth of Juan De Fuca (south).  E.2.5  Particle Average paths Particle average paths (Fig E11 – E16) confirm some of the observations from above: - In all years, the average paths of the northern particles (Comox, Westview, and Pender  Harbor) mostly stay in the northern part of SoG, regardless of their behavior (Fig. E11 and E12)  96  - In all years, the average paths of the southern particles (Cowichan, Saanich and JDF) stay in the southern part of SoG, or are washed out through Juan De Fuca Strait, regardless of their behavior. (Fig. E13 and E14) - No clear trend for particles from Nanaimo, Howe Sound and Fraser (Fig. E15 and E16) - The standard deviation circles are tighter (smaller radius, more concentric) for particles with DVM than the surface drifters in southern SoG (Fig. E1 and E12). Also, Comox DVM particles have tighter, more concentric circles than surface drifters in 2009 (Fig. E13 and E14). The same is true for Howe Sound particles in 2008 too (Fig. E15 and E16).  97  Fig E11. Average paths for Northern surface drifters. Particles were released in the location denoted on each plot. Particles were released 28 days earlier than the date shown on each plot.  Fig E12. Average paths for Northern particles performing DVM. Particles were released in the location denoted on each plot. Particles were released 28 days earlier than the date shown on each plot.  98  Fig E13. Average paths for Southern surface drifters. Particles were released in the location denoted on each plot. Particles were released 28 days earlier than the date shown on each plot.  99  Fig E14 Average paths for Southern particles performing DVM. Particles were released in the location denoted on each plot. Particles were released 28 days earlier than the date shown on each plot.  100  Fig E15. Average paths for Central surface drifters. Particles were released in the location denoted on each plot. Particles were released 28 days earlier than the date shown on each plot.  101  Fig E16. Average paths for Central particles performing DVM. Particles were released in the location denoted on each plot. Particles were released 28 days earlier than the date shown on each plot.  102  Appendix F Hake larvae F.1  Vertical behavior  Fig. F1 Vertical distribution of (a) larvae off the coast of California and (b) eggs and larvae in Puget Sound. Figure is taken from Bailey (1982)  F.2  Swimming speeds for some larval lengths The speeds for some larval lengths as estimated by Miller et al (1988) are summarized in  table F1. Table F1. Larval speeds calculated for different larval size. Larval Length 3mm (newly hatched hake) 9mm (~ 4 weeks old hake larva or a newly hatched herring)  Ave speed (cm/s) 0.25 0.8  10.5mm herring larva  0.96  13.5mm herring larva  1.26  22mm (4-5 weeks old herring)  2.1  103  F.3  Sensitivity to start position and date  Fig. F2. Estimate error for (a) two grid cells apart study and (b) one day apart study.  F.4  Sensitivity to vertical swimming behavior  Fig. F3 Concentration maps for particles released in 2009 near Halibut Bank. Particles have 3 types of behavior – hake larvae like, floaters at 75m, floaters at 200m. Particle release date is shown. Particles were tracked for 28 days.  104  Fig. F4 Concentration maps for particles released in 2009 near Montgomery Bank. Particles have 3 types of behavior – hake larvae like, floaters at 75m, floaters at 200m. Particle release date is shown. Particles were tracked for 28 days.  105  F.5  Interannual variability  Fig. F5 Concentration maps for hake larval particles released near Montgomery Bank (red dot). Particle release date is shown. Particles were tracked for 28 days.  106  Fig. F6 Concentration maps for hake larval particles released near Texada Is (red dot). Particle release date is shown. Particles were tracked for 28 days.  107  Fig. F7 Concentration maps for hake larval particles released near Halibut Bank (red dot). Particle release date is shown. Particles were tracked for 28 days.  108  Fig. F8 Concentration maps for hake larval particles released near Nanaimo (red dot). Particle release date is shown. Particles were tracked for 28 days.  109  F.6  Particle average paths  Fig. F9 Average paths for particles released in 2009. Circles represent the standard deviation or the dispersion from the average position after week 1, 2, 3, and 4. The circles are plotted 1wk, 2wks, 3wks, and 4wks after release. Particle release date is shown. Particles were tracked for 28 days.  110  F.7  Water currents at different depths in the Northern SoG  Fig. F10 Water parcels paths at different depths (75m and 200m) as if the water parcels were observed from a fixed point near Montgomery Bank. Paths were plotted for 28 days, starting on the 1st of each month.  111  F.8  Currents in the Central SoG  Fig. F11 Particle tracks and deep circulation. Random 50 tracks of particles, released near Halibut Bank (a) and Nanaimo (b). The particle tracks follow deep gyres plotted on (c). For (c) mean currents at 100m are averaged over 3 years. Figure is courtesy of D. Masson.  112  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0073588/manifest

Comment

Related Items