Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

New strategies for feeding salmonids : video monitoring and contrast enhancement of feed pellets Ang, Keng Pee 1999

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

Item Metadata

Download

Media
831-ubc_1999-463117.pdf [ 22.97MB ]
Metadata
JSON: 831-1.0089259.json
JSON-LD: 831-1.0089259-ld.json
RDF/XML (Pretty): 831-1.0089259-rdf.xml
RDF/JSON: 831-1.0089259-rdf.json
Turtle: 831-1.0089259-turtle.txt
N-Triples: 831-1.0089259-rdf-ntriples.txt
Original Record: 831-1.0089259-source.json
Full Text
831-1.0089259-fulltext.txt
Citation
831-1.0089259.ris

Full Text

NEW STRATEGIES FOR FEEDING SALMONBDS: VIDEO MONITORING AND CONTRAST ENHANCEMENT OF FEED PELLETS by KENG PEE ANG B.Sc. (Honours), University of Plymouth, Plymouth, England, U . K . , 1981 M . S c , University of Stirling, Stirling, Scotland, U . K . , 1989 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F D O C T O R O F P H I L O S O P H Y in T H E F A C U L T Y O F G R A D U A T E S T U D I E S (Department of Animal Science) We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y O F BRITISH C O L U M B I A August, 1999 © Keng Pee Ang, 1999 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, 1 agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Columbia Vancouver, Canada Date ttef*S€* (f, rfWf -DE-6 (2/88) Abstract The efficiency with which feed is converted to fish flesh is a prime problem for the aquaculture industry. In the practical sense, feed conversion ratio (FCR) refers to the amount of feed that is dispensed to the fish relative to the amount of weight gained. The present research dealt with the possibility of extending the feeding opportunity of caged fish from the current restricted area near the cage surface to the larger volume of the cage as a technique for reducing this ratio and increasing feeding rate. Procedures were developed to control both feeding rate and pellet loss after inspecting feeding behaviour using underwater-cameras. Tests indicated that BioFCR in camera-monitored cages were lower than in surface-fed cages. Spatial homogeneity in feeding (uniform feeding) was seldom observed in both camera-monitored and surface-fed cages. Pellet wastage was high i f fish were surface-fed or fed to a predetermined ration. One explanation for the lack of feeding uniformity was poor visibility of feed pellets towards the cage bottom. The effects of using a contrast-enhanced pellets and light intensity on feeding patterns were examined while feeding Atlantic salmon (Salmo salar L.) using the previously developed camera-monitored feeding method. The contrast-enhanced pellet tested (silver coated) was chosen after a series of laboratory-based fish detectability experiments had been conducted. The silver pellets were 41% more detectable than a conventional pellet to a human observer. Neither the use of conventional nor silver pellets permitted uniform feeding, as fish tended to feed in groups. Pellet discharge rate decreased as stocking density and light intensity increased, and was lower for silver pellets. Fish fed silver pellets preferred to feed near the cage bottom, i i and this preference increased the likelihood of pellet wastage. Feed discharge rate was, therefore, reduced in cages fed silver pellets to keep fish off the cage bottom. Overall, camera-monitored feeding is more effective than surface feeding, as FCR is improved and feeding rate can be optimised. Contrast-enhanced pellets offer fish a choice of where to eat. Light intensity affects spatial homogeneity of feeding, with the highest level of homogeneity under conditions of low light intensity and good water visibility. i i i T A B L E O F C O N T E N T S Abstract i i T A B L E OF CONTENTS iv List of Tables viii List of Figures x Acknowledgement xiii Chapter 1 1 G E N E R A L INTRODUCTION 1 1.1 Economic issues in the salmon farming industry 1 1.2 Scientific issues in the salmon farming industry 4 1.3 Overall research goals and hypotheses 6 1.4 Research objectives 6 1.5 Thesis organisation 7 Chapter 2 9 LITERATURE REVIEW 9 2.1 Fish behaviour 9 2.1.1 Swimming behaviour 10 2.1.1.1 Locomotion 12 2.1.1.2 Schooling behaviour 14 2.1.1.3 Swimming in artificial enclosures 16 2.1.1.4 Vertical/horizontal migration (in cages and in the wild) 17 2.1.1.5 Implications of stress of aquacultural practices on swimming behaviour 19 2.1.2 Feeding behaviour 20 2.1.2.1 Diel cycles of diurnalism, nocturnalism and crepuscularism 23 2.1.2.2 The feeding sequence in fish 24 2.1.3 Gastric evacuation rate in fish 30 2.2 The considerations for feeding fish 31 2.2.1 Types of feed 32 2.2.2 Feed dispensation 33 2.2.3 Feeding frequency 34 2.2.4 Food intake 35 2.3 Underwater vision 36 2.3.1 The underwater visual environment 37 2.3.2 Fish vision and implications of seeing underwater 38 Chapter 3 41 EFFECTIVENESS OF A SUB-SURFACE FISH A N D FEED MONITORING S Y S T E M 41 3.1 Summary 41 3.2 Introduction 41 3.3 Materials and methods 46 3.3.1 Development of camera feeder control system 47 3.3.1.1 Underwater viewing equipment 47 3.3.1.2 Mechanical feeders 48 3.3.1.3 Video feeding techniques 48 3.3.2 Feeding trials 50 iv 3.3.2.1 General field trial protocol 50 3.3.2.2 Experiment 1 52 3.3.2.3 Experiment 2 54 3.3.2.4 Experiment 3 54 3.3.2.5 Data collection and statistical analyses 55 3.4 Results and Discussion 59 3.4.1 Environmental record 60 3.4.1.1 Experiment 1 60 3.4.1.2 Experiment 2 60 3.4.1.3 Experiment 3 61 3.4.2 Experimental observations 61 3.4.2.1 Experiment 1 61 3.4.2.2 Experiment 2 62 3.4.2.3 Experiment3 63 3.4.3 BioFCR, growth, mortality, variation in daily ration amounts and feed dispensation 63 3.4.3.1 Experiment 1 64 3.4.3.2 Experiment 2 70 3.4.3.3 Experiment 3 74 3.5 Conclusion 80 Chapter 4 82 GROUP FISH FEEDING PATTERNS IN CAGES 82 4.1 Summary 82 4.2 Introduction 83 4.3 Materials and methods 84 4.3.1. Underwater viewing equipment 85 4.3.2. Feeding methods 87 4.3.3. Feeding behaviours 90 4.4 Results 93 4.4.1. Low visibility and light intensity 94 4.4.2. Continuous feeding 95 4.4.3. Batch feeding 101 4.4.4. Ration 104 4.5 Discussion and recommendations 104 4.6 Conclusion 109 Chapter 5 I l l L A B O R A T O R Y STUDIES ON THE DETECTABILITY OF PELLETS UNDER VARIOUS E N V I R O N M E N T A L CONDITIONS I l l 5.1 Summary I l l 5.2 Introduction 112 5.3 Materials and Methods 117 5.3.1 Making contrast-enhanced feed pellets 118 5.3.1.1 Contrast levels of coated pellets 122 5.3.1.2 Coat stability/integrity in moving water 124 5.3.1.3 Palatability/acceptability of pellets by fish 125 5.3.2 Laboratory experiments on pellet detectability by humans and fish 125 5.3.2.1 Detectability of pellets (human vision) 126 5.3.2.2 Detectability of pellets (fish vision, preliminary experiments) 129 5.3.2.3 Detectability of pellets (fish vision, final experiments) 135 5.4 Results 147 5.4.1 Detectability/visibility of pellets (human vision) 147 5.4.2 Detectability of pellets (fish vision) 151 5.5 Discussion 157 5.6 Conclusion 162 Chapter 6 163 EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES 163 6.1 Summary 163 6.2 Introduction 164 6.3 Materials and methods 167 6.3.1 Experimental site and design 167 6.3.2 Experimental pellets 171 6.3.3 Instrumentation 172 6.3.4 Data collection and statistical analyses 174 6.4 Results..... 180 6.4.1 Pellet characteristics and environmental readings 180 6.4.2 Observations on fish behaviour 182 6.4.3 Statistical analyses 187 6.4.4 Uniform feeding... 187 6.4.5 Feed ration (kg kg-fish"1) 191 6.4.6 Feed rates 192 6.5 Discussion 200 6.5.1 Trends 200 6.5.2 Feeding patterns 201 6.5.3 Feed ration amounts 202 6.5.4 Feeding rate 203 6.5.5 Surface, bottom and uniform feeding times 204 6.6 Conclusion and recommendation 205 Chapter 7 .' 206 O V E R A L L CONCLUSIONS A N D FUTURE CONSIDERATIONS 206 7.1 Conclusions on overall research 206 7.2 Incidental results 207 7.3 Practical applications 208 7.4 Patentable technology 209 7.5 Future considerations 210 REFERENCES 213 Appendix A - 1. Two-sample t tests (a = 0.10) for SGR, BioFCR and Mortality rates between CMCs and control cages, for Experiments 1, 2 and 3 (Chapter 3) 227 Appendix A - 2. F-test two-sample for variances (a = 0.05) for daily ration amounts in CMCs and control cages for Experiments 1, 2 and 3 (Chapter 3) 227 vi Appendix B- 1. Two-way analysis of variance (using S-Plus)for data from Bamfield Marine Station, with visible distance as outcome and light level (depth) and pellet type as the factors, for clear water condition. With 18 different types of pellets, there were 153 comparisons. Significance is indicated by "****». pj refers to fluorescent colours 228 Appendix B- 2. Two-way analysis of variance (using S-Plus) for data from Bamfield Marine Station, with visible distance as outcome and light level (depth) and pellet type as the factors, for turbid water condition. With 18 different types of pellets, there were 153 comparisons. Significance is indicated by "****". Fl refers to fluorescent colours 231 Appendix C. Colour prints V l l List of Tables Table 3.1. Experimental conditions and results (CMC and c indicate camera-monitored and control cages, respectively). Fish size, unless stated otherwise, was obtained by using a video-sizing technique 53 Table 3. 2. Mean ration amounts before and after Day 17 in Experiment 1. (CMC 66 Table 4. 1. Site characteristics of farms visited and the number of experimental feeding events monitored. Long-term trials were conducted at sites A , B and C 86 Table 4. 2. Feeding technique and feeding endpoint at different farming sites 88 Table 4. 3. Feeding formation, occurrence of pellet wastage, and relative numbers of competitive and aggressive behaviours during feeding events at the various farming sites. 96 Table 4. 4. Duration of feeding times (expressed as a percentage of total feeding time per meal) spent by fish at different locations (depths) when water visibility was greater than 3 m 99 Table 4. 5. Duration of times that fish in CMCs spent feeding at different levels (depths) in the water column, expressed as a percentage of total feeding time per meal or feeding event when water visibility was greater than 3 m for long-term sites (means ± SD) 100 Table 5.1. Reduction in light intensity due to light attenuation at different water visibility and depths in the sea 116 Table 5.2. Coloured and reflective materials used in coating. Check (•/) signs indicate colours tested 119 Table 5.3. Contrast levels (in descending order of magnitude) of 18 different pellet types 1 2 against a black background under high light (12.75+0.10 /anol photon s" m" ; diffused light from a 1000 W halogen lamp) 123 Table 5. 4. Contrast values of selected pellet types against different background colours under bright light (1000W quartz lamp) in air 124 Table 5.5. Pellet types chosen and their corresponding visual characteristics 143 Table 5.6 a. Maximum horizontal visible distance (human vision) of different pellet types under low ambient light (0.03±0.01 pmol photon s"1 m"2) and clear (9.35±0.07 NTU) seawater conditions at two depths 149 Table 5.6 b. Maximum horizontal visible distance (human vision) of different pellet types under low ambient light (0.03±0.01 //moi photon s"1 m"2) and turbid (21.8+0.5 NTU) seawater conditions at two depths 150 Table 5.7 a. Responses from detectability tests (fish vision) conducted under low light (0.007 //moi photon s"1 m*2) and clear water (3.6 ± 1.4 NTU) conditions 154 Table 5.7 b. Responses from detectability tests (fish vision) conducted under low 154 Table 6. 1. Initial fish numbers, mean fish weight and stocking density 170 Table 6. 2 Feeding fish interactively using fish feeding behaviours 176 Table 6. 3 Graphs plotted for highlighting outlier and trends with environmental conditions. 179 viii Table 6. 4. Environmental variables (means ± SD) 181 Table 6. 5 Summary of outcome of statistical analyses (two-tailed) conducted for various data sets 187 Table 6. 6. Duration of times that fish spent feeding at different levels (depths) in the water column, expressed as a percentage of total feeding time for morning (AM) and evening (PM) feeding events (means ± SD) 188 Table 6. 7. Mean and range of rations fed, in kg kg-fish"1 during morning (AM) and evening (PM) feeding events 191 Table 6. 8. Mean and range of feeding rates during morning (AM) and evening (PM) feeding events. Feed rate is the rate at which feed pellets are being discharged into the cage (or discharge rate) 193 i x List of Figures Figure 2. 1 Pathways of light entering a fish's eye 40 Figure 3. 1 Fish are crowded to the top few meters (X m) when fed using surface feeding activity (including feeding frenzy). The depth to which feed pellets can be viewed by the farmer from the surface is dictated by water clarity, and feed size and colour 42 Figure 3. 2 It was hypothesised that i f fish no longer crowd near the surface and redistribute themselves in the water column (X m + Y m) thus, extending the feeding volume (area utilised for feeding) then more fish will have the opportunity to eat and aggressive activities associated with feed competition will be reduced 44 Figure 3.3 a. Graph to show the percent differences between ration fed to control cages and camera-monitored cages (CMCs) in Experiment 1 67 Figure 3.3 b. Variation in daily amount of food fed to fish in camera-monitored cages (CMCs) in Experiment 1 68 Figure 3.3 c. Variation in daily amount of food fed to fish in control cages in Experiment 1 69 Figure 3.4 a. Graph to show that fish in camera-monitored cages (CMCs) achieved better BioFCRs in Experiment 2 because of poor food consumption by fish in the control cages, possibly as a result of feed wastage or that adequate ration was not provided. Rc is the average ration in the control cages, R c m c is the average ration in the CMCs, and the estimation of uncertainty was calculated using root-sum-square equation, which estimates uncertainty at the 99.9 % confidence interval (or 3 SD) (Doebelin, 1975) 71 Figure 3.4 b. Variation in the daily ration in camera-monitored cages (CMCs) in Experiment 2.72 Figure 3.4 c. Variation in the daily ration in control cages in Experiment 2. A l l cages were fed according to surface activity 73 Figure 3.5 a. Graph to show that fish in camera-monitored cages (CMCs) achieved better (lower) BioFCRs than control cages in Experiment 3 76 Figure 3.5 b. Variation in the daily ration consumed by fish in camera-monitored cages (CMCs) in Experiment 3 77 Figure 3.5 c. Variation in the daily ration consumed by fish in control cages in Experiment 3. A l l cages were fed according to surface activity 78 Figure 3.6 a Graph to show increasing feed discharge rate (kgfeed MTfish"1 min"1) as water visibility increased in Experiment 1 80 Figure 4. 1 (a) Starting position for a foraging fish (marked with a white dot), when executing an 'S'-shaped swimming pattern, (b) Mid-foraging position for the same fish, (c) End-foraging position for the same fish before repeating the sequence. Frames are 0.2 s apart. 92 Figure 4. 2 Ring-like structure characterised by organised foraging in continuously fed fish.. 97 Figure 4. 3 Spiral-like structure followed by disorganised foraging in fish fed in batches 102 Figure 4. 4 Feed pellets falling through the centre of a cage uneaten (pellet loss) 102 Figure 5.1. Apparatus designed of aluminium to measure pellet visibility (human vision) under different light and seawater conditions 128 Figure 5. 2 A n aquarium with horizontal grid (2.5 cm off the bottom) and the positions of the four quadrants for dropping pellets 131 Figure 5. 3 Plan view of wet laboratory showing experimental set-up and the position of the light source relative to the aquariums during detectability trials 133 Figure 5. 4 Cross-section of wet laboratory showing aquarium installation and the movable platform for adjusting distance of light source from the wall and to change light levels in the room 134 Figure 5. 5 Pellet suspending device 138 Figure 5. 6 Cross-section of aquarium showing the relative positions of pellets while suspended during detectability trials 140 Figure 5. 7 Three-way combinations for switching operator, suspending devices (DI and D2) and aquariums 144 Figure 5. 8 Different optical paths affecting detectability of silver and conventional brown pellets at different viewing angles and at different depths (and, therefore, light levels).. 161 Figure 6. 1 Layout of cages (measuring 1 2 m X 1 2 m X 2 0 m ) used in the experiment. C and T refer to control and treatment cages respectively 168 Figure 6. 2 Before feeding, fish swim near the sides of the cage in a ring-like structure leaving a hole in the centre 183 Figure 6. 3 When feeding started, fish began to fill the centre hole to feed as more fish swim upward to join the top fish 184 Figure 6. 4 A swarm is formed at the top at the beginning of a feeding event 184 Figure 6. 5 The fish swarm moved down and a discrete group of foraging fish with individual fish exhibiting the "S"-shaped foraging pattern became visible as the discharge rate was increased 185 Figure 6. 6 Fish feeding near the camera as they follow the pellets downward 185 Figure 6. 7 Fish respond to pellet depletion near cage bottom and new feed at the top by swimming back up to feed 186 Figure 6. 8 Swarm is formed again at the top as fish resume feeding 186 Figure 6. 9. Histograms showing similar frequencies of the occurrence of uniform feeding during the morning feeding events. Data were log-transformed (X ' = Log (X + 1)). L C , LT, HC and HT refers to low density control, low density treatment, high density control and high density treatment cages, respectively 189 Figure 6. 10. Histograms showing similar frequencies of the occurrence of uniform feeding during the evening feeding events. Data were log-transformed (X ' = Log (X + 1)). L C , LT, HC and HT refers to low density control, low density treatment, high density control and high density treatment cages, respectively 190 Figure 6. 11 Graph illustrating the reduction in total feeding time with increase in feeding rate (n=36; low density cages). Solid horizontal and vertical lines indicate the invariant feeding time and invariant feeding rate, respectively 194 Figure 6. 12 Graph illustrating the reduction in total feeding time with increase in feeding rate (n=36; high density cages). Solid horizontal and vertical lines indicate the invariant feeding time and invariant feeding rate, respectively 194 xi Figure 6. 13 Graph illustrating that feeding rate was not affected by water visibility (Secchi Disc Reading, m) in any experimental cage . 195 Figure 6. 14 Feeding rate (kg feed MT-fish"1 min"1) versus underwater light intensity levels (umol photon s"1 m"1) before feeding, at 3 m (a) and 8 m (b) for morning feeding events and similarly for evening feeding events at 3 m (c) and 8 m (d) for high density cages. HC and HT refer to high density control and treatment cages 197 Figure 6. 15 Feeding rate (kg feed MT-fish"1 min"1) versus underwater light intensity levels (umol photon s"1 m"1) during feeding, at 3 m (a) and 8 m (b) for morning feeding events and similarly for evening feeding events at 3 m (c) and 8 m (d) for high density cages. HC and HT refer to high density control and treatment cages 198 Figure 6. 16 Feeding rate (kg feed MT-fish"1 min"1) versus successive feeding days before feeding for low and high density cages (a and b respectively) and similarly during feeding (c and d respectively) 199 Figure 7.1 a Depth below which enhanced pellets should be used for greater detectability under different light levels in air and water visibility levels, assuming no shading effects due to fish 211 Figure 7.1b Depth below which enhanced pellets should be used for greater detectability under different light levels in air and water visibility levels, assuming shading effect due to fish 211 xn Acknowledgement I wish to thank my supervisors Dr. Royann J. Petrell and Dr. Beryl E. March for their invaluable guidance, consistent encouragement, and academic as well as psychological advice throughout the research period; Dr. Edward Auld and Dr. A l Castledine for their expert advice and industrial contribution respectively as well as for agreeing to be my committee members. My special thanks to Dr. Jim Thompson (for academic advice, his keen interest in my research and its potential industrial applications), Dr. Bob Miller (for assistance with fluid bed atomising technique), Dr. Allan Donald (for invaluable insights and advice with statistical analyses), to Mr. Meyede (Hi-To Fisheries) for supplying limitless amounts of fish scales and Mr. Sylvain Alie for help with the initial field trial arrangements and his continued interest in the research. I would also like to extend my appreciation to my colleagues for their help during field trials and experimentation: R. Gordon, A . Shieh, S. Paterson, C. Oikawa, J. Louie, T. Shelford, C. Canero, E. Wong, J. Miller, M . Armstead, C. Savage, N . Jackson and J. Pehlke. To Mr. Nick McKinnon and Dr. Hai San Zheng of B.C. Cancer Research Agency for their time and use of their expensive and sophisticated light measuring equipment and to the staff and technicians at Bamfield Marine Research Station. I am especially grateful to the many companies and funding agencies that were involved in this research; without whom this work would not have progressed beyond its conception. They include IAS Products Ltd, Agriculture Canada, Industry Science and Technology Canada, Natural Sciences and Engineering Research Council, National Research Council (IRAP program), British Columbia Ministry of Agriculture, Fisheries and Food, B.C. Packers Ltd. (BCP), Pacific National Group Ltd. (PNG), Pacific Aqua Salmon Farming Partners Ltd. (PASFP), Paradise Bay Seafarms (PBS) and Moore-Clark (Nutreco A/S). In particular, I thank all the site managers and farm site personnel from Sir Edward Island site (BC Packers Ltd.), Bawden Point site, Brinkley Point and Broughton Point sites (PASFP Ltd.), Saranac Island and Blue Heron sites (PNG Ltd.); and Orchard Bay, Conville Bay and Conville Point sites (PBS Ltd.) for their interests and participation in the field trial that was conducted at their respective sites. I acknowledge the kind contribution from Abbotsford Trout Hatchery of experimental fish used during the preliminary stages of the detectability trials. Most of all, I am forever indebted to the patience, support and silent sacrifices endured by my family throughout these last few years of my studies. xiii Chapter J. GENERAL INTRODUCTION Chapter 1 GENERAL INTRODUCTION The farming of salmonids in sea cages is a relatively recent phenomenon. Detailed accounts of historical developments in salmonid culture in the world are numerous including that of Laird (1996). The main species cultured is Atlantic salmon (Salmo salar) but considerable quantities of chinook (Oncorhynchus tshawytscha) and coho (O. kisutch) salmon are also produced. The major salmon producing countries include Norway, Chile, the U.K. , Canada and Japan. World production of farmed salmon increased from a few thousand tonnes 18 years ago to approximately 300,000 tonnes in 1990 (Shaw and Egan, 1996), over 450,000 tonnes in 1994 (Novotny and Pennell, 1996), approximately 560,000 tonnes in 1996 (Anon 1, 1996), 605,000 tonnes (Atlantic salmon) in 1997 and projected to be around 706,000 tonnes in 1998 (Anon 2, 1998). British Columbia (ranked fourth in the world as a salmon producer), has 103 salmon farming operations, concentrated in the coastal communities on Vancouver Island and produced 24,025 tonnes of salmon in 1996, worth more than $172 million (Anon 1, 1996). Presently, farmed salmon is British Columbia's largest agricultural export crop. 1.1 Economic issues in the salmon farming industry Aquaculture in general, and salmon farming in sea cages in particular, has been able to continue growing because of continued improvements in farming techniques, seed production, fish nutrition, disease control, and habitat access. The growth is also driven by the increases in market demand due partly to the decline in wild catches and partly to changes in people's lifestyles toward healthier diets. In salmon farming, stronger and better cage designs have 1 Chapter I. GENERAL INTRODUCTION allowed access to more remote and open waters while research and development of more nutritionally balanced diets has enabled fish to approach their genetic potential for growth under a given culture environment. Underlying all these issues is the pressure on aquaculture to be economically profitable and environmentally sustainable. In order to maintain profitability as production increases, the cost of production must decrease over time. Many aspects of fish farming have been researched, resulting in improvements to problems faced by farmers such as broodstock quality, egg hatchability, larval rearing and survival, smolt transfer (ponding) survival, various vaccines to combat disease outbreaks and feed wastage and accumulation underneath a cage, to name a few. The aquaculture industry has not been immune to public scrutiny for its accountability and responsibility towards the environment as it proceeds with the day-to-day processes. The public sensitivities for the environment partially resulted in a moratorium being imposed on further expansion of the aquaculture industry in British Columbia in 1996 that is still in effect today. Uneaten feed represents one of the biggest components of waste that is discharged to the environment. This waste is both expensive and detrimental to the environment. Feeding systems should be designed to minimise feed waste. In terms of feeding management, many aspects have not been fully researched. Advances in feed management could lead to substantial cost savings because feed costs consistently account for more than 50% of the production cost of a typical farm operation. Nutritionally superior feeds need to be delivered equally to every fish in a population, and to satiation at each feeding event. Effective feed management strategies should be developed to closely match fish appetite, which is highly variable. Effective feed management may also, ensure evenly sized fish, thereby eliminating the need for labour and time consuming grading process which is also stressful to the fish. 2 Chapter 1. GENERAL INTRODUCTION The efficiency of the farming or production process is measured in many ways including Specific Growth Rates (SGR) and Feed Conversion Ratio (FCR). In industry, FCR is expressed as Economic FCR (EcoFCR) or Biological FCR (BioFCR) depending on whether biomass gained by fish dying through the course of production is included or not, as shown in the following equations: SGR ( mjFinal Wt) - \n(Initial Wt) ^  Growth period, days xlOO (1.1) where (Final Wt) and (Initial Wt) are the final and initial average weight of fish respectively for the period of growth. SGR values are expressed in % body weight per day. „ „„_ Total dry rationused EcoFCR= - (1 2) (Final#xFinalWt)-(Initial#xInitialWt) ' BioFCR - Total dry rationused (Final #xFinal Wt) - (Initial#xInitialWt) + BiomassMorts ^ where Final # and Initial # are the final and initial fish numbers and Biomass Morts refers to the sum of weight gained by individual mortality (i.e. weight of individual mort minus initial average fish size) accumulated over the period of growth. The EcoFCR in British Columbia (B.C.) according to the Cooperative Assessment of Salmonid Health program (CASH) conducted through the B.C. Salmon Farmers Association in 1993 was approximately 1.5 for Atlantic salmon and 2.0 for chinook salmon over a production period ranging from 18 to 24 months for both species. Different farms have different operating practices and priorities thus resulting in different efficiency levels. FCRs 3 Chapter 1. GENERAL INTRODUCTION ranging from 1.0 to 1.9 for Atlantic salmon; and from 1.6 to 2.5 for chinook salmon are not abnormal. High FCR can be due to overfeeding or underfeeding. In an overfeeding situation, reducing feed wastage while feeding fish to satiation may greatly reduce FCR. Uneaten feed is also detrimental to the industry because it can accumulate underneath a cage to attract wild fish and/or deteriorate from microbial activities causing environmental concerns. 1.2 Scientific issues in the salmon farming industry The only source of food for fish kept in artificial enclosures such as cages comes in the form of formulated diets provided by the farmer. Feed delivery techniques and the understanding of fish feeding behaviour in sea cages become of paramount importance to ensure feeding efficiency. Much research has, therefore, been done to gain better understanding of the many biological processes associated with fish growth and survival under such environments. These include daily consumption rate and meal size, optimal feeding frequency, the ideal feed distribution pattern, as well as the optimal or preferred shapes, size and colour of feed pellets, as evidenced from the following literature review. There is, however, little known about the feeding 'space' occupied by fish during feeding and the feeding patterns within that space. The horizontal dimensions of length and width are fixed in a rigid holding structure like a sea cage, allowing a fixed number of fish to occupy the surface plane at which feed pellets enter the feeding area. Those fish that cannot enter the upper plane are forced to occupy even lower planes within the cage, essentially creating an underwater feeding volume. Under conventional feeding techniques, fish are fed using surface feeding activities such as splashing (aggressive feeding behaviour) to judge satiation or feeding endpoint. Farmers discharge feed at a rate according to how fish in the Chapter 1. GENERAL INTRODUCTION surface planes are observed to react to pellets in the water. Feeding is discontinued when the fish ignore food and/or when the total ration for the day has been fed. There have, however, been no studies to show whether fish at all planes in a cage can be satiated at the same time. Discontinuing feeding based on surface feeding behaviour may, therefore, deny food access to deeper fish. Forcing fish to feed near the surface may promote competition, and consequently, unequal access to food. Competition is an interaction between individuals in which one or more of the participants suffers a net loss of fitness and none show a net gain compared with values in the absence of the competitive interaction (Wooton, 1990). However, in the context of a cage environment, when fish are forced into the top few meters of the water column during feeding, it is the intraspecific competition that comes into play. Hassell (1976) recognised two forms of intraspecific competition, namely, scramble competition and contest competition. In scramble competition, the share of resources taken by an individual depends on the amount available and the population density; irrespective of the rank of the individual in the population. In contest competition, the resource is unequally divided because the share taken by high-ranking individuals (dominants) is independent of low-ranking individuals (subordinates). Territorial behaviour is a form of contest competition i f some fish that are capable being territorial, are prevented from doing so by other territorial fish (Wootton, 1990). Unequal access of fish to food can result in non-uniform feeding when fish are not feeding at all levels, and most of the feed pellets are consumed by fish in upper planes, leaving little for those in the lower planes. When food reaches the lower planes, it may not be as visible or detectable because of reduction in light intensity due to light attenuation underwater and high 5 Chapter I. GENERAL INTRODUCTION turbidity (Parsons et al., 1984). Shading by surface-feeding fish may further reduce light intensities at lower planes. An ideal feeding scenario occurs when food is made available not only uniformly over a 2-dimensional area but also within a 3-dimensional volume. 1.3 Overall research goals and hypotheses This research was, therefore, aimed at addressing the need in the salmon farming industry for improving F C R and fish growth. Scientifically, it aimed to establish the concept of extending the feeding volumes, to further our understanding of feeding behaviour within these feeding volumes and to decrease feeding time. Individual hypotheses will be discussed in subsequent relevant chapters. 1.4 Research objectives The objective of this research was to develop a feeding system that would incorporate underwater monitoring of fish feeding behaviour and pellets for controlling feeding rate and pellet wastage. The components of the study had the following objectives: (1) To test the performance of the above system against the traditional surface-activity-based feeding system using growth, mortality rates and FCR as comparative parameters, (2) To conduct exploratory research on salmon farms regarding the usage of the cage volume during feeding, (3) To determine if feed pellets can be made more detectable to salmonids under different light and water visibility levels and 6 Chapter I. GENERAL INTRODUCTION (4) To compare the effect of visually enhanced pellets against conventional pellets on feeding behaviours, feeding rate and ration in a commercial farm under typical farming and lighting conditions. 1.5 Thesis organisation . This thesis is organised as a series of papers (both published -Ang and Petrell, 1997 and Ang and Petrell, 1998, and unpublished) describing experiments and field trials conducted during the course of this research. Chapter 1 introduces the subject of fish farming as a young industry (when compared to the poultry or swine industries) that is still growing and lists the economic as well as scientific issues facing the industry. It lays out the overall research goals and hypotheses of my research as well as specific research objectives. The literature review is covered in Chapter 2 where the subjects of fish feeds, feeding techniques, feed wastage, swimming and feeding behaviours as well as fish underwater vision are reviewed separately. Chapter 3 describes the work done to develop and test the effectiveness of a sub-surface based fish feeding system utilising underwater cameras through three long-term field trials. Materials associated with Experiments 1 and 2 in Chapter 3 were published in 1997 as Ang and Petrell (1997). The underwater cameras used in this new feeding system to detect pellet wastage also facilitated the observation and scrutiny of the swimming and feeding behaviours of salmon in sea cages. The observational and experimental results on group feeding patterns in cages from the long term field trials, combined with results and discussion from shorter term trials at five more farm sites, are described in detail in Chapter 4 and published as Ang and Petrell (1998). 7 Chapter I. GENERAL INTRODUCTION It was discovered during the many feeding events conducted during periods of poor water visibility and low ambient lighting conditions that fish could not home in on individual pellets very easily, thus missing pellets while foraging vigorously during a feeding event. Laboratory studies were subsequently carried out to determine the characteristics of pellets that can aid in pellet visibility underwater under these adverse conditions. Chapter 5 describes the procedures that led to the production a pellet with enhanced visual properties designed to be more visible than a conventional brown pellet to both humans and fish, especially under very low lighting conditions. Chapter 6 describes the eventual reaction of fish fed the visually enhanced pellets as compared with fish fed conventional pellets. Chapter 7 summarises the overall research findings and conclusions and outlines future considerations toward a better understanding of fish behaviour in sea cages. 8 Chapter 2. LITERA TURE REVIEW Chapter 2 LITERATURE REVIEW This literature review is aimed at topics in fish swimming and feeding behaviours, in relation to the concept of feeding fish and the effects of feeding techniques on the farming process. These topics are followed by reviews of the underlying principles and physics of light and vision under water. Literature on individual aspects that are more pertinent to each stage of the research is then reviewed in greater detail in the introductions to each subsequent chapter in order to keep the chapters more complete and self-contained. 2.1 Fish behaviour Animal behaviour as a subject, has been well studied and documented in voluminous journals and periodicals (e.g. Animal Behaviour; Animal Monographs; Journal of Animal Behaviour; Animal Learning and Behaviour) as well as in books and manuals (e.g. Hafez, 1975; Reese and Lighter, 1978; Archer, 1979; Craig, 1981; Wood-Gush, 1983; Barnard, 1983; Huntingford, 1984; Monaghan, 1984; Ellis, 1985; Hart, 1985; Fraser and Broom, 1990; Monaghan and Wood-Gush, 1990; Lehner, 1996). The study of fish behaviour, however, has a shorter history (e.g. Huntingford, 1984; Pitcher, 1993). Pitcher (1993) noted that materials presented in the first edition of a book, entitled, "Behaviour of Teleost Fishes" (1986) indicated that fish behaviour is not just a simplified version of that seen in birds and mammals, but obeys the same ecological and evolutionary rules. 9 Chapter 2. LITERATURE REVIEW Fish have evolved by natural selection to live in many different environments including the occupation of a variety of habitats during their life cycles. Fish behaviour is, therefore, varied and present in all stages of the fish life. Detailed accounts of fish movements from the developing embryo up to different stages in the fish life cycle have been documented (e.g. Huntingford, 1993; Dill , 1977). Di l l (1977) noted that upon emergence (and even before emergence), young Atlantic salmon, Salmo salar, will dart and snap at solid objects moved in front of its head. Later, they begin to chase and nip at their companions at the same time as they start to feed while fish that are being chased, instinctively give escape responses in similar movements. Salmonid behaviour progresses to foraging and aggression, fear and (predator) avoidance, seaward migration (smoltification), homing, mating and spawning behaviour as the fish grow and mature, with a few exceptions. A good example of an exception is when some male parr in Atlantic salmon mature without migrating at all (Thorpe, 1989). These early maturing fish retain their tendency to swim into the current and remain on the spawning ground to compete both with other parr and with anadromous males to fertilise eggs (Huntingford, 1993). Other accounts of such individual differences and alternate behaviour in fish can be found in a review by Mogurran (1993). Similarly, the genetic basis of fish behaviour, also called behavioural genetics, can be found in a review attempting to link behaviour to genes by Danzmann et al. (1993). 2.1.1 Swimming behaviour In this review, swimming behaviour covers locomotion, schooling, swimming in artificial enclosures, vertical/horizontal migration patterns and the implications of stress of aquacultural practices on swimming behaviour. Swimming behaviour is manifested in fish from the most rudimentary lateral movements of the body, fins and tail by newly hatched fry. The young fish 10 Chapter 2. LITEM TURE REVIEW eventually pushes itself off the substrate, and swimming become mechanically and visually sensitive. Periodic sudden vertical movements made by the young salmonid, by flexing its tail and pushing against the substrate eventually bring the fish out of the gravel and into the stream. Light adaptation does not occur until after emergence of the young salmonid from gravel. The first feeding movements may start (even before emergence) from poorly co-ordinated biting (dart and snap) movements, directed indiscriminately at any small objects. These become more effectively directed only at potential food items and young fish begin to chase and nip at their companions (which, in turn, exhibit escape response) at the same time as they start to feed, with similar movements. Subsequent behaviour of young salmonids varies among species and progresses to foraging and aggression, fear and avoidance, seaward (horizontal) migration, homing, mating, and spawning. Most species of salmon, for example, have a migratory pattern that extends from gravel beds in rivers (as smolt) to the oceans and back (as matured fish). At almost every stage along the thousands of kilometres, they have to feed (for somatic and reproductive growth), excrete, seek shelter and avoid predators (among others) and eventually, reproduce to give rise to the next generation. The returning mature salmon do not feed during the final stretch of their journey upstream to the spawning ground. Hasler et al. (1978), suggest that the homing process that migratory salmon utilise to reach their home streams appears to have learned components because juveniles imprint on the odours of their home streams and are able to recognise these odours as they near their streams. 11 Chapter 2. LITEM TURE REVIEW 2.1.1.1 Locomotion According to Maier and Maier (1970), five fundamental types of locomotion exist, namely flagellar movement, changes in body shape (e.g. as in Paramecium), undulation, jet propulsion, and movement of extremities. Of these, undulation (or writhing, a pulsating, wave-like locomotor pattern) and movement of extremities are employed by many vertebrate fish while swimming through the water medium. In an aquatic medium, swimming becomes of paramount importance, being intimately interwoven with most other behaviours in the fish's life cycle. An excellent account of the physiology of fish locomotion (e.g. musculature, movement) can be found in Bone et. al. (1995) while aspects of optimisation of locomotion in fish (e.g. optimal body shapes, burst and coast swimming) has been covered by Weihs and Webb (1983). As such, only swimming patterns of fish, particularly salmonids in the wild as well as in cages will be discussed here. In the wild (ocean phase), coho salmon, Oncorhynchus kisutch, were tracked using depth-sensing ultrasonic transmitters, and observed to have average swimming depths (for four fish) ranging from 7.1 to 13.4 m with maximum depths reached of 53 to 74 m but spent most of the time (72.3 to 92.8%) in the upper 15 m of the water column (Ogura and Ishida, 1992). Other authors (Jakubsstovu, 1988; Holm et al., 1982; Westerberg, 1982) have also suggested that salmon generally swim within the upper 15 m with occasional dives to greater depths. In particular to Oncorhynchus tshawytscha, off the coast of Vancouver Island, most of the fish sampled during troll investigations were found below 48 m while the maturity of only maturing fish were located below 20 m (Healey, 1991). In particular to Atlantic salmon, the few fish that were tracked, moved up and down in the water column over 150 m and showed no depth preference. Jakupsstovu (1988). The general equation for optimal swimming speed was developed from theory and subsequently modified to include size effects by applying various empirical 12 Chapter 2. LITERA TURE REVIEW relationships based on trout and salmon as quoted by Weihs and Webb (1983) to take the form, V=0.5L0 4 3 , where V is expressed in m s"1 and L is the fork length in m. The average ground speeds of salmon ranged from 0.29 to 0.4 m per sec (m s"1). This agreed somewhat, with predicted speeds for migrating salmon of 0.5 m s"1 using the formula, V=0.5L043 (Bone et al, 1995). A 9 m basking shark, on the other hand, can reach 0.94 m s"1 whilst swimming at 8 m below the surface. The first and simplest behavioural parameter that a fish can control is its swimming speed, relative to the water mass. A common mode of locomotion among many migratory species is an alternation between burst and coast phases. In this mode of locomotion, a fish initially accelerates using the body and fins (the burst phase) followed by a comparable period of coasting with no propulsive motions referred to as the coast phase (Weihs and Webb, 1983). The burst/coast cycle is repeated when the coasting speed has slowed to the fish's initial value. This swimming behaviour has been shown to lead to large energy savings for sustained swimming over long distances (e.g. during migration). Swimming speed in fish is important also because the prey-predator relationship that determines whether a fish becomes prey or predator depends on their relative swimming speeds and the ability to execute a C-start (a prey's escape fast start response) or a strike (a predator's fast start) characterised by an S-start. Escape for a prey becomes possible when the prey enters a safe zone delimited by the predator's larger minimum turning radius, assuming that predation interactions ultimately occur between individuals of different sizes (Weihs and Webb, 1983). Pelagic fish such as salmon is able to control buoyancy, thus, enabling the fish to control lift and depth at which to reside in the water column. Control and rapid alteration of lift play an important role in escape once a prey detects the presence of a predator. In general, fish with swim bladders either move slowly, hover, or both and search for food only over a horizontal 13 Chapter 2. LITERA TURE REVIEW plane. Others (e.g. Salmonidae) alternate hovering and scanning the substrate for prey with short distance cruises (Gee, 1983). The minimum turning radius of a fish is expected to be determined when the centrifugal force equals the turning moment (Howland, 1974). Clearly, because all these manoeuvres (for potential prey or predator) require energy executed through muscles, and fish extremities, stressors which compromise the performance of body tissues and availability of energy, will undoubtedly influence the natural outcome of predator-prey interaction, which in turn, will determine the structure and composition offish populations and communities in an ecosystem. In other words, i f a fish's movement is slowed down by the effects of stress, it becomes prey more easily than unstressed individuals. Similarly, it will consistently fail to obtain food (which weakens it further or until it becomes prey to healthier individuals) because potential prey will escape with ease. 2.1.1.2 Schooling behaviour For schooling fish, many reasons have been suggested for the schooling behaviour (Cushing and Harden-Jones, 1968; Shaw, 1978) such as benefits in food search and escape (from predators), and hydrodynamics (Weihs and Webb, 1983). The word "school" refers to a group of fish swimming at about the same speed in roughly parallel orientation and maintaining constant nearest-neighbour distance, NND, usually in the order of one-half or one body length (Pitcher, 1993). It is suspected that NND is determined by an interplay of attraction mediated by visual stimuli and repulsion by lateral-line stimuli, thus maintaining the school structure and dynamics (Bone et al, 1995). Such synchronised motion in tight formations can be shown to reduce thrust (Weihs, 1975), and, therefore, energy required to move at any given sustained speed. Stress is expected to compromise the maintenance of such a structure because when enough individuals in a schooling 14 Chapter 2. LITEM TURE REVIEW formation are not able to hold their positions (by maintaining NND), the entire school can be jeopardised and the benefits lost. The school can either regroup (of reduced size, if sufficient numbers are healthy) and stressed individuals will be expected to fall behind and be excluded or left to join another level in the food chain hierarchy. Large schools are often oblate spheroid in shape, possibly to minimise the detection envelop underwater and so reduce predation. A "shoal", on the other hand, describes all social groups of fish, including schools as well as aggregations of fish with random orientation and varying NND (Bone et al., 1995). The major role for shoaling (including schooling) is protective, although it may play a part in food search (described later) because it is difficult for predators to select individual prey from a shoal (the confusion effect) and they usually settle on stragglers. It follows, therefore, that any stress factor that compromises the responsiveness of a fish's visual and lateral line sensory organs will also disrupt the structure and dynamics of the shoal, thus increasing the proportion of stragglers (or unhealthy individuals) that are lost from the shoal, much to the benefit of the predators. Pink salmon fry (Oncorhynchus gorbuscha) develop schooling behaviour to replace individual responses soon after emergence from gravel. Hoar (1958) found that pink salmon fry in shallow experimental troughs would not abandon the school when attacked (by crows), as did chum (O. Keta), coho (O. kisutch) and sockeye (O. nerka) frys and noted the obvious disadvantage of strong schooling behaviour in pink salmon in shallow streams (Heard, 1991). Schooling behaviour is exhibited in the marine phase of pink salmon life cycle but the anti-predator defense attribute of schooling during the juvenile stage is doubted (Heard, 1991). Some species (e.g. O. nerka) alternate between schooling and non-schooling periods (e.g. juvenile fish in Lake Washington; Burgner, 1991). Detailed coverage of schooling behaviour in Pacific salmon can be found in Groot and Margolis (1991). 15 Chapter 2. LITERATURE REVIEW 2.1.1.3 Swimming in artificial enclosures Domestication is a process of adaptation of organisms to an environment provided by man, a process which involves an evolutionary response through changes in gene frequencies between generations, and environmentally induced shifts in developmental processes and rates that occur within each generation (Price, 1984). Salmon in cages are faced with quite different sets of constraints than those experienced in the wild. The net cage forms a physical barrier that alters the "natural" swimming depths, speed and swimming patterns of fish. Indeed, the size of enclosures has been shown to significantly influence the swimming characteristic of planktivorous brook trout, Salvelinus fontinalis (Tang and Boisclair, 1993). Smaller enclosures restrict fish movements, forcing them to execute sharper and more frequent turns. As enclosure size increases, fish movements tend toward faster and more linear swimming patterns that may be more characteristic natural behaviours in the wild, including swimming patterns that resemble shoaling (with variable NND). The incessant swimming activity may be associated with an inherent migratory tendency related to optimum cruising speed (Weihs, 1973) or for foraging (Sutterlin et al., 1979). Although no stress parameters were measured for fish in the present study, it seemed likely that forcing fish to make sharper and more frequent turns would constitute unnatural behaviour resulting in additional stress and wastage of valuable energy. Other forms of unnatural or altered behaviour as a result of holding fish in net barriers include circular swimming orientation (clockwise or counter clockwise), swimming depths and speeds, all of which must be stressful to farmed fish albeit, to varying extents, but no supporting information is available. The maximum depth of the standard net cage usedToday (15-20 m) does coincide with the general swimming depths of salmonid in the wild (section 2.1.1.1) but does not allow fish to reach 16 Chapter 2. LITEM TURE REVIEW greater depths at will as in wild fish. A trade-off between predation risk and feeding rate has been suggested as the determining factor in the preferred depth for sokeye salmon, and the existence of anti-predation "windows" for feeding in near surface waters at dawn and dusk (Clark and Levy, 1988). A predator-prey relationship does not ordinarily exist in a monoculture net cage situation. However, individuals of a predatory species e.g. dogfish, can pass through the net mesh into a cage as juveniles and remain in the cage, either voluntarily, or are trapped in the cage as they grow, feeding on the abundant food supply and eventually predate on the farmed species (salmon). Bone et al. (1995) suggest that cannibalism can be severe in aquaculture if a sufficient size differential (where individual lengths differ by 50 %), is allowed to build up within a cohort. Domestication and genetic selection for fast growing, non-aggressive traits may reduce size variations and, therefore, cannibalism to varying extents. Dominance-subordinate relationships and territoriality in sea cages may, however, be more important than in the wild because of crowding, and the inability of subordinates to move away from the domain of the dominant individuals. 2.1.1.4 Vertical/horizontal migration (in cages and in the wild) Marked differences are found in the vertical distribution of different fish species in the natural environment, with many species making seasonal and diurnal vertical migrations (Neilson and Perry, 1990; Levy, 1990; Clark and Levy, 1988). Nevertheless, there seems to be no consistent diel pattern for all species of adult Pacific salmon, or individuals within a species (Ruggerone et al, 1990; Quinn et al, 1989). In contrast, Atlantic salmon in cages usually descended at dawn and ascended at dusk, swimming deeper in summer than winter (Ferno et al, 1995). Great variations in local densities were found, with up to 80% of the fish localised within a 1 m depth interval in 17 Chapter 2. LITERATURE REVIEW summer. Fish were, however, more evenly spread out in the cage in winter. Light apparently exerts a greater influence on fish in cages than in the wild, because the same study also found a negative correlation between light level and fish density at the surface in summer, and high light levels resulted in deeper fish distribution. The crowding of large numbers of fish at or near the bottom of the cage (as a result of avoidance of adverse stimuli such as light, predator, salinity or temperature at the surface) can be expected to be stressful to the fish as competition for both space and oxygen increases. As by-products of metabolism accumulate in the vicinity of individual fish, removal rates are reduced because of increased fish biomass, which acts as a barrier to water exchange. Fish distribution in cages is also influenced by feeding, with rapid upward swimming at the start of feeding and a gradual descent during the course of feeding (Ferno et al, 1995). Since feeding and migration are also major motivations of vertical movement in wild fish, Ogura and Ishida (1992) suggested that the variation of daily changes (in swimming depth) might have occurred due to the stage of maturity, food and physical environments and the distance from natal rivers. The vertical distribution of fish in net cages can also affect fish health because both the concentration of sea lice larvae and the risk of skin damage caused by U V light are high close to the surface (Johannessen, 1978; Bullock, 1988), as are the risks of attack from birds. Fish in net cages swim relatively rapidly in a ring or circle formation in the horizontal plane. Sutterlin et al, (1979) observed that swimming orientation (clockwise or counter clockwise) did not seem to be influenced by variations in the amplitude or direction of the water current. The path of a single fish might actually describe a series of spirals or ellipses with swimming radii and velocity gradually changing from revolution to revolution. This swimming characteristic also applied to individuals at different localised groups at the different swimming 18 Chapter 2. LITEM TURE REVIEW depths, with the exception of fish closest to the cage bottom where individuals tended to form unstructured groups, with apparently random swimming direction and speed. 2.1.1.5 Implications of stress of aquacultural practices on swimming behaviour In the wild, the benefits of catabolising energy reserves for use in overcoming or avoiding a stressful environment could include moving on to more favourable environments. However, when farmed fish are exposed to environmental or operational stressors in aquaculture, there is usually no means of avoidance or escape. The beneficial aspect of stress response (immediate survival) is, therefore, not apparent (Pickering, 1992). Where stressors are severe or prolonged, homeostasis may be irrecoverable. This fact may be manifested in certain behaviours of captive fish that would not normally be characteristic of the fish under "normal" (unstressed) conditions. Ferno (1989), for example, found a positive correlation between the number of lice and leaping activity of Atlantic salmon reared in cages, and explained that the leaping activity was triggered by the lice infestation (the stressor). During a leap, the fish accelerates to break surface and makes an uncontrolled landing. In the same study, the leaping activity was found to increase gradually until de-lousing, and a marked reduction occurred after de-lousing. In approximately 6% of the leaps, fish came into contact with the net wall, possibly damaging the epidermis. Such physical injuries can be expected to trigger additional stress responses in the already stressed individual. As well, the damaged epidermis will lead to an increase in susceptibility to more parasitic and other pathogenic infections because the fish's skin is one of its vital first lines of defence against disease entry. 19 Chapter 2. LITERATURE REVIEW Rolling activity, where the fish swims slowly upward and breaks the surface before descending again, was also considered to be associated with stress reactions in captive fish (Ferno, 1989). It was found to increase as leaping activity decreased after delousing. At the end of a roll, fish was observed to swim rapidly down and gas was released. The release of air from the swim bladder (especially during a delousing procedure) is interpreted as a stress reaction (Furevik et al, 1988), and a high rolling activity after delousing (itself a very stressful process), may be connected to a compensation of this gas loss (Ferno, 1989). Leaping and rolling activity levels can be simultaneously high at dawn and dusk in some farms that apparently have very low sea lice infestation (personal observation). It is possible that such activities may have a genetic basis, an inherent escape behaviour (when in confinement), a behavioural response to predator (seals, sea lion, dogfish at the bottom of the cage), or even a learned behaviour (from other individuals within the cage or from adjacent cages). 2.1.2 Feeding behaviour As one of the most diverse of living organisms, fishes occupy virtually every possible trophic role from herbivorous species to secondary carnivores (Wootton, 1990) and are represented in just about every conceivable aquatic environment (Pitcher, 1993). Over 22,000 species are teleost, to which nearly all of the fish of importance in commercial fisheries and aquaculture belong. This trophic diversity has meant great variability in both the size and range of food organisms consumed by fish. A classification of fishes from different ecosystems according to their food groups can be found in Hyatt (1979). The feeding behaviour of fish involves a hierarchical set of decisions. Once it has decided to feed, it is faced with a number of further choices including choosing a search path to minimise 20 Chapter 2. LITEM TURE REVIEW the energy loss per unit of search time (Hart, 1993). In congruence, the theory of optimal foraging is based on the evolutionary premise that individuals within a population that forage most efficiently and maximise their net rate of energy income will possess greater fitness and contribute more genes to future generations (Calow and Townsend, 1981). Here, fitness is taken to mean fish size and the ability to produce more eggs. For example, male (stickleback) fish will select and court larger females first. In terms of "cost and benefit", the theory proposes that natural selection favours a predator whose foraging strategy provides the highest possible rate of energy intake relative to energy expended in feeding (Krebs, 1978). Further, if the net or gross energy intake per unit time is to be maximised, then the most profitable prey will be those for which the cost of prey capture is minimal. One measure of this cost is the time taken to handle the prey (h) divided by the weight or energy content of the prey (r), or hlr. This measure of cost can be modified to include the time costs of searching and capture (Wootton, 1990). As the rate of encounter with the most profitable prey declines, the next most profitable prey should be included. Bluegill sunfish feeding on Daphnia showed that at low prey densities Daphnia of different sizes were consumed in the same proportions as they were encountered. At higher prey densities, the bluegill prey more selectively on the bigger Daphnia available. However, the bluegill did not completely stop feeding on the smaller Daphnia as the theory predicted (Werner and Hall, 1974) When fish apparently fail to choose a predicted optimal diet, it may be an indication that the variable assumed to be maximised has been incorrectly identified. That is, the fish is maximising something else. Fish may be lacking the information required to make an optimal choice. Another possible explanation is a lack of suitable genetic variation in the population so an optimal solution cannot evolve through natural selection (Wootton, 1990). Wootton (1990) also 21 Chapter 2. LITERA TURE REVIEW suggested that, while the optimal foraging theory assumes the forager has perfect knowledge of the profitabilities of different prey, it does not specify how such information is achieved. A forager's approach to an optimal diet may depend on a learning process or on relatively simple rules of thumb such as always select the largest prey visible (Krebs and Stephens, 1986). Nevertheless, according to Wootton (1990), there is evidence from both experimental and field studies that fishes do respond to differences in their rate of return as they forage in different regions of their environment. As suggested by an ideal free model originally developed by Fretwell and Lucas (1970) on birds, Milinski (1986) suggested that fish, in an environment in which food is distributed in patches of different densities, could distribute themselves so that no fish is able to increase its feeding rate by switching to another patch. The ideal free model assumes that the fish have perfect information about the profitabilities of each patch and that competition with conspecifics does not constrain the movement of the fish. For example, threespine sticklebacks presented with two patches of Daphnia of differing profitability distribute themselves between the patches in a ratio that closely approaches that predicted by the ideal free distribution, albeit taking a few minutes to reach the ratio (Milinski, 1979). However, a detailed analysis of feeding rates shows that the fish do not obtain equal rewards, but that some are competitively superior to others, so the distribution is not truly ideal and free (Milinski, 1984)b. In fact, when two sticklebacks are presented simultaneously with a large and a small Daphnia, over a series of presentations a good competitor takes more of the large daphnids (Milinksi, 1982). Wootton (1990) suggested that prey selection of the sticklebacks is not determined solely by mechanical constraints or by simple optimal foraging rules because the presence of a conspecific also exerts an effect, at least on the less capable competitor. This approach of optimisation as applied to animal behaviour, therefore, is subject to criticism since a solution to a 22 Chapter 2. LITERA TURE REVIEW problem is only optimal under a particular set of conditions: change the assumptions and the optimal solution will change (Hart, 1993). Nevertheless, the theory has contributed much to the understanding of foraging theory in fish and continues to be developed in the same area. (Tytler and Calow, 1985). 2.1.2.1 Diel cycles of diurnalism, nocturnalism and crepuscularism The rise and set of the sun each day imposes a strong influence on the feeding behaviour of fish, resulting in somewhat abstract feeding times including diurnal, crepuscular and nocturnal that has been used extensively in the literature. There is a considerable literature on diel and seasonal activity in relation to the feeding behaviour of fish. For example, Eriksson and Alanara (1990) noted that rainbow trout in Sweden, fed mainly during dawn and dusk during February and July, but fed chiefly by day during August and September. Later in the autumn months of October and November they fed mostly after dark and at night. Other authors (Adron et al, 1973 and Grove et al., 1978) also found that rainbow trout fed nocturnally during the winter months. Smith et ah, (1993) studied the diurnal and seasonal variation in behavioural indices of appetite in one-sea-winter Atlantic salmon in a sea cage in relation to environmental variables and fish swimming activity from October, 1990 (Autumn) to April, 1991 (Spring) and found marked seasonal variation in feeding behaviour. There were no marked morning and evening peaks of appetite. The former outcome was interpreted to indicate a reduction in appetite from autumn to winter and a rapid increase in appetite from late winter onwards. However, in contrast to the latter outcome, an earlier study carried out at the same site during June and July (summer) 1989 by Kadri et al. (1991), both appetite and swimming speed 23 Chapter 2. LITEM TURE REVIEW showed the same marked daily rhythm, being highest in the early morning and evening and lowest in early afternoon. 2.1.2.2 The feeding sequence in fish a. Appetite and satiation A prerequisite for achieving the goal of optimal feeding in fish is an understanding of some knowledge about how the appetite of fish varies with time and about how it affects the amount of feed that will be taken at a given instant. Knight (1985) defined appetite as the drive state that initiates arousal and appetitive behaviours leading to consummately behaviour, satiation and negative feedback (drive reduction). Other authors (Olsen and Balchen, 1992) consider appetite as the short term desire for food, often expressed as the amount of feed ingested per unit time (on a voluntary basis) and that, in the long term, appetite is controlled by a set point given by the central nervous system, depending on several factors including stomach fullness (related to the amount of feed previously eaten) and stomach evacuation rate. The same authors refer to long term appetite as hunger. Wootton (1990) defined hunger as the propensity to feed when given the opportunity, while appetite is the quantity of food consumed before the fish ceases to feed voluntarily. Under the latter definition, the motivational state hunger and the determination of appetite is generated by an interaction of two systems. The first is systemic demand, that is the demand for energy and nutrients generated by the metabolic rate, and the second is the rate at which the digestive system can process food. Satiation refers to the fish condition at cessation of voluntary feeding. Satiation time then, refers to the time from start of (active) feeding to voluntary cessation of feeding (Brett, 1971). Elliot (1975) and Grove et. al. (1978) suggest that the appetite of fish is strongly dependent on 24 Chapter 2. LITERATURE REVIEW stomach fill (or fullness), such that, after feeding, while the food is being digested and evacuated from the stomach into the intestines, the fish become increasingly motivated to approach and struggle for food. Knight (1985) also noted that, in general, appetite is inhibited by stomach distension but excited by gastric evacuation. However, according to Olsen and Balchen (1992), appetite involves metabolic, neurophysical and hormonal mechanisms and may be stimulated by metabolic factors such as levels of certain metabolites in the blood. Temperature affects the maximum rate of consumption through its effect on the rate of gastric evacuation and its effect on systemic demand (Wooton, 1990). At low temperatures fish may cease to feed, but as temperature increases, consumption rate increases up to a maximum. A further increase in temperature is marked by a rapid decrease in consumption. This decline in consumption at high temperatures is said to- be a systemic effect because the rate of gastric evacuation continues to increase (Elliot, 1972). Maximum appetite refers to the maximum amount of food consumed on a voluntary basis. Under fish farming conditions where food is readily available, the fish feeding rate for a particular feeding frequency will be determined by maximum appetite and the rate at which appetite returns after a meal (Dunbrack, 1988). Dunbrack also stated that because there is an inverse relationship between appetite and pre-feeding stomach fullness (Brett, 1971), a corresponding relationship between the rate constants of appetite return following a meal and gastric evacuation is implied. Statements of the above nature and results derived from experiments using (mostly) a limited number of juvenile fish feeding on natural foods over a limited period of time should be viewed with caution since pelleted feeds used in aquaculture would elicit different gastric evacuation characteristics (Dunbrack, 1988). Fish under culture conditions also co-exist in higher densities (Olsen and Balchen, 1992) and are therefore, further exposed to additional influencing 25 Chapter 2. LITERATURE REVIEW factors. When viewed in this context, information on aspects of fish feeding such as maximum appetite (and sustained feeding rate), the rate of appetite return and how they vary with body size and feeding frequency can be useful in designing optimal feeding schedules. Dunbrack (1988), working with juvenile coho salmon, referred to sustained feeding rate as the maximum feeding rate that can be sustained indefinitely under a particular feeding schedule and is calculated as the ratio of total food consumed over a time interval to the interval length. b. Arousal and food search Arousal and appetitive behaviour are direct results of intrinsic increases in appetite (Knight, 1985). De Ruiter (1967) classified search behaviour into locomotion, scanning via the sense organs and special search movements such as turning over leaves by blackbirds. Hyatt (1979) reviewed species-specific search patterns and concluded that there appeared to be a match between the kinds of prey exploited and specific search procedures by fish. In particular, lemon sharks, Negaprion brevirostris, which orientate to the strongest local current and swim rapidly upstream upon receiving chemical stimuli, are considered pursuit predators that live largely on a diet offish. Under culture conditions with fixed feeding schedules, Knight (1985) found that fish can learn to anticipate feeding times but considered arousal and aggregation near feeders to be due to the classical (Pavlovian) conditioning, just as sight or sound of farm workers near feeding areas could act as conditioned stimuli. Social facilitation was also considered important in high density fish culture, for example, arousal and excitation of one or a few fish leading to mass excitation of others (Knight, 1985). It follows from this consideration, therefore, that a whole population of fish could be stimulated to start feeding by the feeding action of a smaller group of fish. 26 Chapter 2. LITERATURE REVIEW The (Pavlovian) conditioning ability in fish could also help fish to learn to respond to and take a novel formulated diet (Knight, 1985) and would, theoretically, allow farmers to successfully adopt and change the optimal feeding regime and feed type from time to time to suit the prevailing farm conditions. c. Location and initial identification The location of food by fish (as in search) is commonly mediated by sensory systems that process visual, electrical, mechanical (turbulence and sound), or chemical stimuli (Hyatt, 1979). And since a complex set of optical, acoustic, tactile, chemical and electrical stimuli may be associated with each potential food item, Hyatt (1979) suggested that well-defined responses of fishes are usually a consequence of the summation of a complex of signals received through a combination of sensory pathways. Salmonids have always been considered to be visual feeders (Hoar, 1942) but they do not have to rely entirely on vision to feed (Eriksson and Alanara, 1990). Despite the generally poor quality of underwater images, fish depend a great deal on vision as a source of sensory information (Guthrie and Muntz, 1993). The same authors stated that because of the physical nature of light and its complex interactions with the environment, a variety of different properties of visible objects can be recognised, such as brightness, hue, texture and contour, as can more subtle differences of degree, such as patch size and pattern grain. Further, comparative properties such as colour contrast or brightness contrast can also be identified. Other sensory modes apart from visual detection used by fish to locate and identify food include electrodetection (orientation by means of electric stimuli especially in sharks and rays), mechanoreception (use of sound and turbulence as stimulus cues to locate food) and 27 Chapter 2. LITERA TURE REVIEW chemoreception (use of chemical stimuli) (Hyatt, 1979). Applications of the above information to the farming situation such as the manipulation of pellets to be more visible under water will be discussed later under "Types of feed". d. Capture In the wild, the relative vulnerability of a single food item to different fish species or of a variety of food items to a single species is often determined by match or mismatch of predator and prey characteristics during the approach, pursuit and attack phase (Hyatt, 1979). The probability of successful capture of prey, according to Hyatt (1979), is the integration of the fishes' swimming abilities with morphological characteristics. Capture success is also dependent on prey size and its avoidance behaviour (Knight, 1985). Hyatt (1979) conceded that, given the variety of body forms and the accompanying hydrodynamic properties, no general rules emerge with respect to body size and attack procedures by fishes, that is, different forms adopt different procedures. However, in general, fishes capable of high burst speeds regularly capture highly mobile prey such as other fishes while those that possess much lower burst speeds feed mainly on less mobile prey. At the opposite end of the spectrum of search and capture behaviour is the "sit and wait" procedure, which often involves stealth. In macrophagous carnivores, capture involves orientation of all or part of the body relative to the food item and then possibly a forward striking movement. Most teleost take small food particles into the mouth whole by creating an inertial suction, that is, a sudden negative pressure is created by rapid expansion of the orobranchial cavity (Hyatt, 1979). By deduction, farmed fish 28 Chapter 2. LITERATURE REVIEW would take feed pellets in a similar manner making pellet size an important consideration when manufacturing formulated feed. e. Orobranchial manipulation and testing Once in the mouth, food items may have to be manipulated prior to swallowing. Just as fishes cannot successfully pursue and capture every prey detected, they will not successfully handle and ingest every prey captured (Hyatt, 1979). The characteristics of the prey (morphological e.g. possession of spines and behavioural e.g. escape adaptations) are just as important as the size, shape, position and mechanics of jaws, mouth, branchial arches and dentition of the predatory fish in determining patterns of food acquisition. Hyatt (1979) provides a detailed account of morphology, handling and ingestion of food for further reference. Finally, palatability of foods is assessed by fishes during the handling phase of food exploitation (Hyatt, 1979) before swallowing. Sutterlin and Sutterlin (1970) suggested that, for many fishes, chemoreception is utilised again in the food handling stage to determine whether a particular food item is eventually ingested. f. Swallowing or rejection Swallowing represents the final stage in the acquisition of food in fish. It implies that the prey or food has passed all tests that it has been subjected to by the predatory fish such as size, shape, texture, palatability and energy content. Rejection of unsuitable food or prey is often forceful in fish. Stradmeyer (1990) in observing the response of fish to pellets noticed that fish actively spit out pellets during rejection of food. Information is lacking on the rejection of immobilised, injured or unsuitable prey in the wild, possibly because fish will have already judged the suitability of a potential prey through its 29 Chapter 2. LITERA TURE REVIEW system of sensory modes before attacking, as described earlier and rejection at the handling stage is therefore, rarely observed. 2.1.3 Gastric evacuation rate in fish Gastric evacuation rates in fish directly affect the number of meals that can be fed to fish in sea cages on a daily basis. In general, as temperature increases, gut evacuation rate increases allowing multiple feeding in the summer months and only one feeding per day during the colder fall, winter and spring months. The gastric evacuation rate is probably controlled or influenced by both external (including temperature, feed type and availability) and internal factors or mechanisms. In fact, all models (whether linear, exponential, power exponential, square root, logistic, logistic restricted, Gompertz or Gompertz-restricted) developed to describe gut evacuation rate in fish contain the temperature term. Following a meal, a number of different gastrointestinal peptide hormones may be released into circulation and both the amount and time of release may be affected by meal composition and rate of gastric evacuation which may, in turn, be influenced by individual hormones or interactions between hormones (Jobling, 1986). The evacuation of food from the stomach is not thought to be a continuous smooth process but may occur in a pulse-like (step-wise) fashion (Jobling, 1986). Jobling (1986) further hypothesised that the pattern of emptying is influenced by feedback signals from receptors located in the upper intestinal tract and by factors affecting the rate of physical/chemical breakdown of the food particles, principally temperature. Gastric emptying patterns are determined by the interaction of a number of factors including propulsive forces generated by the gastric musculature, inhibitory feedback, central nervous input and the anatomical structure of the stomach and pyloric region. Smith (1989) 30 Chapter 2. LITEM TURE REVIEW suggested that the factors most relevant to gastric evacuation studies are fish species, fish size, temperature, food quality, meal size and feeding history of the experimental fish. Kapoor et al. (1975) demonstrated that in general, there is a direct relationship between secretion of gastric juice and temperature such that, for a 10 °C rise in temperature, there is a 3-4 fold rise in rate of digestion up to an optimum value. Storebakken and Austreng, (1987) suggested that high evacuation rates allow more food to be consumed in total thus giving more weight gain. 2.2 The considerations for feeding fish The most important aspect of fish behaviour related to aquaculture is the feeding behaviour and possible hierarchical systems among the individuals. While most research is focused on feeding experiments in various types of tanks and is, therefore, fairly representative of domesticated fish, little attention has been paid to hierarchical indicators such as spatial distribution and aggression or avoidance. The concept of feeding fish is the same as that of feeding other food animals, that is, to nourish the animals to the desired level or form of productivity as profitably as possibly. The differences arise from an aqueous environment that dictates dimensions beyond those considered in feeding land animals. The nutrient requirement and feeding practices for fish are also unique. The necessity to feed fish efficiently has been a natural accompaniment to intensification of culture for increased productivity in fish farming. Starting with low organic or inorganic fertilisation of ponds, the feeding of farmed fish has progressed through the use of supplemental feed (usually in crude or concentrated feed material) to provision of all nutrients required by the fish in the form of prepared or optimally formulated feed. As aquaculture has become more 31 Chapter 2. LITERATURE REVIEW dependent on prepared feeds, the need for nutritionally complete feeds also becomes critical especially in highly modified environments such as net pens, floating cages, raceways and tanks. 2.2.1 Types of feed It is rarely possible to feed fish in intensive farming systems on a natural diet for reasons including difficulty in dispensation, maintenance of quality (shelf-life) and obtaining sufficient quantity on a constant basis. To date, the fish farming industry has spent enormous amount of time and money on developing the optimum feed, in trying to improve its nutritional value, movement through water, shape, size, colour, texture, smell and palatability. The shape, texture and colour of pellets has often been determined by the economics of manufacture rather than by any preference of the fish (Metcalfe, 1990). The easiest shape of pellet to produce is a roughly spherical "crumb" but salmon were found to be far less likely to approach and attack crumb-shaped compared with rod-shaped pellets composed of the same material. Stradmeyer et al. (1988) attributed this to the fact that most of salmon's natural food is elongate and a round object is more likely to be something inedible such as dirt. The types of feed available to the fish farming industry include pelleted or extruded pellets, meals and crumbles, larval feeds (including micro-encapsulation), moist diets and flaked diets. A dense pellet that sinks rapidly in the water is produced by steam pelleting through compression. The extrusion process, on the other hand, produces a low-density pellet that floats in the water. Extrusion is a process by which feed material is moistened, pre-cooked, expanded, extruded and then dried. Other feed types will not be discussed here because they are rarely used in salmonid culture. In general, an optimal feed pellet should be nutritionally adequate, palatable, attractive to 32 Chapter 2. LITERATURE REVIEW the fish (by adding attractants, smell) and be shaped as to elicit the greatest feeding response from the feeding fish. 2.2.2 Feed dispensation The method of feed dispensation can influence the population structure of fish under culture in at least two ways: namely, the size distribution within the population and overall growth of the population. This coincides with the two goals of a fish farmer as described by Noakes and Grant (1990), that is, to maximise the efficient production of fish and to divide this production into parcels of relatively uniform size. Variations in fish size are often attributed to agonistic behaviour and territoriality in fish. Dominance hierarchy is especially pronounced in a slow continuous distribution of food which allows dominant individuals to monopolise a large share of resources through defence of space (Thorpe et al. 1990). This same monopolistic characteristic then leads to uneven sized "parcels" of fish production. Noakes and Grant (1990) suggested that food should NOT be presented from a few, concentrated and predictable points. They added that such a system would promote resource defence or dominance hierarchies with the largest or most aggressive fish occupying sites below the feeding point. Even if no aggression occurred, the high density of fish attracted to the concentrated food sources might promote interference competition (Gillis and Kramer, 1987) and inefficient foraging. Food should be delivered to fish in a manner as to render it indefensible (Dill et. al. 1981; Pucket and Dill , 1985) and therefore, unprofitable for any individual fish to defend. This can be achieved by scattering food evenly over a wide area of the surface of the water. Fish would then respond to an even distribution of food by distributing themselves uniformly over the tank (or 33 Chapter 2. LITERA TURE REVIEW cage). This ininimises aggression and maximises foraging efficiency (Noakes and Grant, 1990). The initiation of scramble competition (as a result of even distribution of feed) will result in more uniform growth rates in a population (Noakes and Grant, 1990). 2.2.3 Feeding frequency Some authors (e.g. Noakes and Grant, 1990) suggest that there may be a minimum number of feedings per day, based on limitations set by fioatability of food or the oxygen demand created by adding food to the system (e.g. tank). In a farming situation, the number of meals per day is greatly influenced by the water temperature and labour availability. Most farms today feed fish during the early daylight hours with the total ration (% body weight per day) divided over several feeding batches in a rotational pattern from cage to cage until the entire ration has been fed. Despite knowledge that there are pronounced diel and seasonal differences in feeding behaviour of salmonids, and that feeding patterns are timed by the light/dark cycle, farmers continue to schedule feeding time and frequency to suit their schedule and not the fishes'. Feeding method and schedule are highly critical aspects of feed management in the farming process and should be given priority over all other activities. Information from the literature also tends to suggest that fish feeding behaviour under farming conditions is still influenced by the natural feeding patterns of fish in the wild (e.g. Eriksson and Alanara, 1990) because farmed fish today are only several generations removed from wild stocks. 34 Chapter 2. LITERATURE REVIEW 2.2.4 Food intake Fish, like all animals, depend on an income of energy and nutrients acquired by their feeding activities to survive, grow and reproduce. Balon (1986) lists four types of nutrient acquisition in the ontogeny of fishes -endogenous (yolk), absorptive (from the environment via body surface), exogenous (orally ingested and intestine digested) and mixed (a combination of two or all three). The food taken in by fish depends on several factors including visual conditions, particle size and distribution, and confusion effects (Olsen and Balchen, 1992). Visual conditions include light intensity, water turbidity and particle colour/background contrast. Milinski (1984a) introduced the idea of a "Confusion cost" incurred by farmed fish during feeding when describing the distraction produced by the presence of many alternative food items. For example, Jakobsen et al. (1987) found that farmed salmon grew faster when fed on two colours of pellet simultaneously than when fed on either colour alone. It is thought that the confusion is reduced when the diet is more varied (Metcalfe, 1990). Stradmeyer et al. (1988) found evidence that commercial pellets commonly used by farmers are considered to be harder in texture than those that produce an optimal response from the fish. In the wild, the rate of food intake is determined by prey type and abundance, presence/absence of conspecifics and predators; hunger and appetite levels. On the other hand, the chances that an individual will be attacked depend on the density of the prey, the density of the predators and often on the size of the predator. The relationship between the number of prey attacked per predator and the density of the prey is represented by an asymptotic functional response curve (Wootton, 1990). As the density of prey increases, the number attacked increases. The asymptote may be produced firstly by the fish becoming satiated and secondly, by the time available for foraging. Experimental studies tend to suggest that when food is distributed in 35 Chapter 2. LITERA TURE REVIEW patches, a fish can increase its feeding rate by joining a shoal (Pitcher, 1986). In contrast, conspecifics in salmonid mono-culture show more aggressive behaviour than in duo-culture of, for example, salmon and charr (Abbott and Dill , 1985) The presence of a predator has been shown to reduce the food intake of a fish because the fish is less likely to forage (Wootton, 1990) and actively balances the risk of predation with energy intake. Koebele (1985) found that even the sight of a dominant was enough to reduce the food intake of subordinate cichlids (Tilipia zilli), despite being given excess food. The implications of this to feeding salmonid in cages when seals are ever present should be considered alongside other stress factors (e.g. fear) in fishes. In discussing hunger and appetite, it is apparent that under culture conditions of abundant food supply, restrictions on foraging time become unimportant and food intake is then determined by the maximum appetite (refer to previous sections on appetite) of the fish and the rate at which appetite returns or the gastric evacuation rate. Finding the optimal ration size to satiate fish under commercial farming conditions is not simple because of the unpredictability of fish appetite. 2.3 Underwater vision Vision is vitally important in a lot of animals on this earth relative to the environments they inhabit. However, light levels underwater are several magnitude lower than those above water because of the paths that light rays have to pass through before entering the water surface and also because of the medium of transmission once underwater. For example, light from the sun, moon and stars is first filtered by the atmosphere before entering the water column (Leow and McFarland, 1990). From the water surface, light can be further affected by suspended solids 36 Chapter 2. LITERATURE REVIEW present in the water (turbidity), and salinity and temperature profiles relative to depth (through salinoclines and thermocline with differing relative densities). 2.3.1 The underwater visual environment The optical characteristics of water affect illumination intensity, spectral quality, and directional distribution (Fernald, 1993). The maximum rate at which photons arrive at the sea surface from an overhead sun in a sky of average cloudiness is about 4000 //moi photon s"1 m"2 (Tett, 1990). These photons are reflected, refracted, and to some extent, polarised at the water surface before becoming available to fish underwater. These physical transformations are themselves critically dependent on the surface of the water because typically more light enters through a still surface than a rough one (e.g. with wave action). Such transformations at the water surface, therefore, generally mean that the underwater visual environment faced by fish depends on surface conditions. Specifically, wave actions at the surface influence both temporal and spatial fluctuations in the intensity of underwater light levels to depths of up to 70 m with fluctuations differing at different depths according to the wavelength of the dominant waves (Leow and McFarland, 1990). The visual environment also changes throughout the day, being different at noon with the sun overhead, and at dawn and dusk, when sunlight arrives at a lower angle through a longer atmospheric path, resulting in much lower intensities, shifted towards the shorter wavelengths (Muntz and McFarland, 1977) Image formation underwater is greatly influenced by scatter (redirection of photons due to reflection, refraction and diffraction). Scatter decreases the penetration of light through water and diminishes the directionality of light, eliminating shadows and hence directionality 37 Chapter 2. LITERATURE REVIEW cues. Underwater light coming from an image is also scattered thus reducing brightness and, therefore, reducing available information about the image (Lythgoe, 1972). 2.3.2 Fish vision and implications of seeing underwater Comprehensive reviews on fish vision have been published by several authors including A l i (1975), Lythgoe (1979), Nicol (1989), and Douglas and Djamgoz (1990). The optics of animals living underwater differs from those of animals living in air in one fundamental aspect: underwater there is no air-cornea interface which provides significant optical power to land-dwelling vertebrates (Fernald, 1993). For example, in humans, the air-cornea interface provides about 43 diopters of refractive power while the lens provides 13 diopters (Westheimer, 1968). Eyes underwater, therefore, have no optical benefit from the cornea because its refractive index is nearly equal to that of water (Fernald, 1993) and the lens have to provide all the dioptric requirements. Fish lenses are spherically shaped (to compensate for eye-size constraint) and have very high refractive strength or dioptric power. The amount and spectral quality of light entering a fish's eye will differ according to the line of sight involved (Levine et. al., 1980). Views overhead will involve the shortest optical path lengths and, therefore, luminance attenuation and spectral shift will be least (Figure 2.1; pp. 36, Arrow 'a'). In fact, Snell's window dictates that objects overhead above the water surface, within a solid angle of 97° can be viewed. Horizontal lines of sight are subject to lower light levels and greater spectral shifts due to the longer optical paths involved (Guthrie, 1980). The apparent contrast of objects viewed horizontally will be reduced by veiling light scattered from particles between the observer's eye and the object (Arrow 'e'). The effect of the veiling light becomes greater as the object moves farther away from the eye. The spectral 38 Chapter 2. LITERA TURE REVIEW reflectance of the object is increasingly shifted toward that of the dominant space light and the object brightness approximates that of background brightness. The downward line of sight from the eye involves the longest optical pathway, with the greatest effects on both object brightness and its spectral properties (Guthrie, 1980). Please see Figure 2.1 below for the many possible pathways of light entering a fish's eye. 39 Chapter 2. LITERATURE REVIEW Organic • Snell's window Figure 2. 1 Pathways of light entering a fish's eye. Arrows indicate from the left: a-direct overhead path (high intensity, daylight spectrum); b-long path scattered from bottom (low intensity, modified spectrum); c-light scattered from suspended particles (veiling light); d-light scattered from submerged object; e-path as in d, but light largely absorbed by suspended particles and therefore, not reaching the observer; f-as in c, but longer light path (greater spectral shift, less brightness); g-as in b, but shorter light path reduces the spectral shift produced by differential absorption and reflectance; h-as in g and b, but total reflectance from water surface has occurred, and spectral shift is greater due to long light path through turbid epilimnion. Adapted from Guthrie (1986) 40 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Chapter 3 E F F E C T I V E N E S S OF A S U B - S U R F A C E FISH A N D F E E D M O N I T O R I N G S Y S T E M 3.1 Summary Traditionally, feeding rates and amounts of feed provided to fish are based on a subjective interpretation of surface feeding behaviour, as feeding participation rate was relatively unknown and the fraction of the cage utilised for feeding (feeding volume) was small. Furthermore, pellet wastage was not monitored. The following study demonstrated that feeding efficiency was improved by both extending the feeding volume with the use of underwater observations of fish behaviour and controlling pellet loss. Feeding trials were conducted on three salmon farms. 3.2 Introduction Feeding fish in a three-dimensional sea cage, where the cage bottom and often the fish are not visible from the surface, is a complex operation. In an aquatic medium, feed pellets offered to fish quickly disappear from sight after hitting the water surface making it very difficult to judge feeding rates and cessation. The traditional use of surface feeding activity (e.g. feeding frenzy or fish splashing, mouthing, etc.) to determine the feeding rate and amount forces fish to crowd into the top few meters of the water column to compete for pellets causing scale loss and stress. This method of using surface feeding behaviour to feed fish forces fish to utilise a small upper fraction of the cage because the farmer will stop or reduce the feed discharge rate if fish are not visible from above and/or surface visible fish reject pellets (Figure 3.1). 41 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Feeding volume Walkway Portion of feed pellets as viewed by the farmer from the surface Float Fish nearest the surface has the opportunity to feed first Fish here may have to wait their turn Figure 3. 1 Fish are crowded to the top few meters (X m) when fed using surface feeding activity (including feeding frenzy). The depth to which feed pellets can be viewed by the farmer from the surface is dictated by water clarity, and feed size and colour. 42 Chapter 3. EFFECTIVENESS OF A SUB-SURFA CE FISH AND FEED MONITORING SYSTEM The fraction of the cage utilised for feeding can be described as the feeding volume by considering the following. In a sea cage, horizontal dimensions of length and width are fixed, allowing a fixed number of feeding fish to occupy the surface plane at which feed pellets enter the feeding area. Those fish that do not make it to the upper plane are forced to occupy a second lower plane, and lower still for other fish, essentially creating the underwater feeding volume (see Figure3.1). As so little is known about the actual feeding volume, it may be insufficient to only use surface observations to control feeding. The hypothesis in this research was, therefore, feeding fish using underwater observations of fish and pellet loss to extend the feeding volume may affect FCR, growth and mortality (Figure 3.2). Using an underwater camera to observe fish behaviour and pellet loss extends the feeding volume because observations of fish and feeding behaviour are not restricted to the surface. 43 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Figure 3. 2 It was hypothesised that if fish no longer crowd near the surface and redistribute themselves in the water column (X m + Y m) thus, extending the feeding volume (area utilised for feeding) then more fish will have the opportunity to eat and aggressive activities associated with feed competition will be reduced. The underwater camera is used to observe fish behaviour and pellet loss, which eliminates the need to use surface feeding behaviour to feed fish. 44 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM On commercial farms, performance of the farming process is often measured using feed conversion ratios (FCRs) and growth rates. Wastage directly affects FCRs and consequently, the economics of fish farming. Although FCRs for dry feed have been reported to be close to 1 or less for small fish and, under experimental conditions, for larger fish (Austreng et al., 1987), the FCR in British Columbia (in 1993) according to the Cooperative Assessment of Salmonid Health program conducted through the B.C. Salmon Farmers' Association is approximately 1.5 for Atlantic salmon and 2.0 for chinook salmon over the entire production period. The actual FCR may be worse (higher) because wastage and underfeeding were not considered. Reports in the literature indicate that 15-40% of the dry feed offered may leave a cage uneaten (Gowen et al., 1985; Thorpe et al., 1990; Seymour and Bergheim, 1991). The highest levels of wastage have been associated with automatic feeders when the broadcast area was too small or/and dispensation rate was too high. Uneaten feed is a problem to farming because it accumulates underneath a cage where it is subject to microbial spoilage and attracts wild fish. For farming conditions, special sensors have been used to detect feed pellets (Juell, 1991; Blyth et al, 1993; Skjervold, 1993; Juell et al, 1993; Foster et al, 1995). An infra-red sensor has, for instance, been used to control an automatic feeding system in Tasmania (The AquaSmart from the Tasmania Technopark Centre: Blyth et al., 1993). Airlift pumps can be installed in the bottom of cages in order to return uneaten feed to the surface (Feedback Feed Control System; Talbot, 1995). An inexpensive, underwater camera connected to a surface viewing monitor has been used in the Bio-Resource Engineering laboratory to observe the movement of pellets within cages (Foster et al, 1995). The system utilises unique image characteristics of pellets (sinking). The sinking pellets appear snowy white against a black 45 Chapter 3. EFFECTIVENESS OF A SUB-SURFA CE FISH AND FEED MONITORING SYSTEM background if the camera faces downward and black against a white background if the camera faces toward the water surface. In the present study, a camera-based feed control system was developed and tested for the feeding of fish to satiation near the cage bottom without pellet loss. A vision-based system was examined because (1) it would permit direct visual detection of fish and pellets without any need for system calibration, (2) feed dispensation rate could be varied according to both fish activity and pellet loss, and (3) the position of the sensor (video camera) within a cage need not be fixed but can vary depending on the depth of fish schooling and on environmental conditions. The specific objectives of this study were: 1) To develop a feeding system, which incorporates underwater monitoring of fish feeding behaviour and pellets, for controlling feeding rate and pellet wastage, 2) To test performance of the above system against the traditional surface-activity-based feeding system using FCR, growth and mortality rates as comparative parameters. In addition, while testing the new feeding system, feed dispensation rate, surface and underwater observations of feeding activities were recorded. 3.3 Materials and methods Three long-term (2-3 months duration) feeding trials were conducted at commercial salmon farms in British Columbia. Two of these trials were conducted with Atlantic salmon, Salmo salar, and one with chinook salmon, Oncorhynchus tshawytscha. Three sites permitted system testing under two different stocking densities and on two species. Under the present 46 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM feeding system, salmon were fed using mechanical feeders while viewing with a monitor connected to an underwater video camera placed below 1 -2 m below the feeding aggregation. Subsurface feeding activities and pellet loss were used to judge satiation and feed dispensation or discharge rate. As controls, fish in similar cages were fed using only the visually perceived decline in surface activities as an indicator of satiation. Details of experimental design and set-up follow. 3.3.1 Development of camera feeder control system 3.3.1.1 Underwater viewing equipment Surveillance video cameras (Panasonic WV-BD400 and Cohu 4915-2000/0000) each with a Cosmicar (Pentax Canada Inc., Vancouver, B.C.) 4.8 mm lens were used. The cameras were environmentally sealed by custom built underwater housings (International Hardsuits, North Vancouver, B.C.). The resulting viewing volume was a right rectangular pyramid of height, as measured perpendicular to the base of the pyramid, estimated to be the visibility (m) within the water column as defined by the Secchi disc reading. No artificial lights were used. The umbilical cords from the video cameras were connected to a Panasonic WJ-FS10 digital frame switcher (DFS). A Panasonic AG-1960 S-VHS recorder was used to record special events, and a Panasonic TR-930 CB video monitor was connected to the DFS to allow for direct observation of images captured by the video cameras. The vertical placement of the cameras in the cages was 1 to 2 m below the feeding zone (i.e. 5-12 m deep depending on fish vertical distribution during feeding). The lens of the cameras faced toward the water surface. In that position, the feed pellets and fish were viewed as distinct black objects against a light grey background. 47 Chapter 3. EFFECTIVENESS OF A SUBSURFACE FISH AND FEED MONITORING SYSTEM 3.3.1.2 Mechanical feeders Mechanical broadcast feeders/blowers (Aero-Spreader 200, Feed Companion, IAS Products Ltd., Vancouver, B.C.) were used to ensure an even and uniform broadcast of feed pellets in all test cages. This was accomplished with the feed control system and a flexible, movable spout. The discharge rate and throwing distance provided by the mechanical feeder were adjustable. Its hopper capacity was 50 kg (i.e. the size of two typical feed bags). 3.3.1.3 Video feeding techniques Prior to conducting the feeding experiments, procedures to feed fish quickly without pellet loss were developed by studying underwater feeding video footage. The feeding endpoint was defined as the time at which fish are satiated. This endpoint is judged by the fact that fish at the bottom of the feeding school had eaten and/or when pellet loss could not be controlled by adjusting the feeding rate. The following procedures were employed throughout the field trials: (1) The feed dispensation rate was as high as possible without generating pellet loss. (2) Pellets could escape through the cage side depending on the current velocity. If the water current became a problem in this connection, feed dispensation was discontinued and resumed when the current velocity sufficiently decreased to permit the pellets to sink into the fish-feeding zone. (3) Feed was scattered over the entire cage surface because growth rate was found to increase with the size of the feed scatter area (Thomassen and Lekang, 1993). When fish were observed to feed at the surface and the net panels were not billowed or deflected, the feed scattering area covered most of the cage surface. Discharge rate (kg min"1) was observed to increase with increasing scatter area and spout movement. 48 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM (4) The discharge spout of the feeder had to be continually moved. Keeping it still, so that pellets fell in the same place, resulted in wasted feed and/or feeding frenzies (i.e. intense level of splashing on the surface). Unless otherwise stated in Procedure 7, pellets were, therefore, scattered by continually moving the discharge spout of the feeder from side to side and in a circular motion. This is in agreement with Kadri et al. (1995), who surmised, after studying the feeding interactions of 19 fish, that food should be presented so that its presence is unpredictable in both time and space. (5) If fish were not eating at the surface and/or the net panels were billowed or deflected, broadcast area and feeding rate were reduced. The broadcast area was reduced by restricting it to an area above the camera. This action minimised the loss of pellets through the sides of a cage. (6) The motors of the mechanical feeders were not turned off during a feeding event, because fish became accustomed to the noise. Without the noise, they would cease foraging even if pellets were still visible in the water. (7) The feeding method was varied (interactively) throughout each feeding event in order to maximise feed intake by fish. One of the following four methods were used, (a) Feed was offered in small pulses lasting two sweeps around the cage, (b) Feed was offered in longer pulses of four to six sweeps around the cage, (c) Fish were continuously fed in a sweeping motion around the cage for up to 3 min. (d) Fish were continuously fed in a sweeping motion for longer than 3 min. Pulses and short continuous feeding were used to check for feed consumption at the beginning of a feeding period and between bags of feed, and to help determine the endpoint for the feeding event. 49 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM (8) Feeding was terminated if pellet loss could not be prevented through any of the above mentioned techniques. 3.3.2 Feeding trials Three sets of experiments were conducted on commercial sea cage farming sites to test the potential benefits of camera-monitored feeding over conventional feeding using surface activity to judge satiation and feed discharge rates. The performances of these two methods were compared in terms of feed conversion, growth and mortality. Feed dispensation rate is expressed in 'kg feed MT-fish"1 min"1' (MT refers to metric tonne) throughout this thesis instead of the S.I. unit of 'kg feed kg-fish"1 s"1 in order to avoid expressing very small values with up to six decimal places (e.g. 1.3E-6). 3.3.2.1 General field trial protocol A l l camera-monitored cages (CMCs) were fed by researchers, while the control cages were fed by experienced farm workers. Neither group had access to the feeding records of the other group, thereby ensuring independent responses to fish feeding behaviour. Both groups worked the regular farm shifts of 8 to 10 d on and 4 to 6 d off, so that individuals tending the machinery varied over the course of the experiment. The variation in personnel made it very unlikely that any one person could influence the results of the study. On the farms visited, cages were normally hand-fed. For the duration of the feeding trials, all cages were fed with mechanical feeders (see section 3.3.1.2). A l l farm workers were instructed to treat the feeder as a "mechanical arm" and to follow video feeding techniques 2 and 4 above, and to scatter pellets widely over the cage. The mechanical feeders were used and 50 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM instructions given to ensure that the different effects we observed were caused by different interpretations of feeding activity and not by broadcast method or current speed. Feed dispensation in control cages was based on standard hand feeding practices. Feeding was discontinued when pellets were not consumed before they faded into the water column. The discharge rate was judged relative to the level of surface feeding activity. Mouthing (seeing open mouths) and surface splashing were the most commonly used indicators. Dispensation rate increased or decreased depending on the number of those types of occurrences. A farm worker usually "fed to the bag", i.e. a bag (25 kg) was emptied even if fish feeding activity was slow, or conversely not started if it appeared that the fish would not consume another complete bag. Under those conditions, at least one bag was always fed in any given feeding session. Experiments commenced 1 wk after the cameras were introduced into the CMCs. Prior to the experimental period, the fish population in cages had been divided to reduce stocking density. During the splitting operation at site B, fish were individually counted and maturing fish (grilse) were removed. Fish at sites A and C were stocked into experimental cages according to stocking densities as planned by the farm management. The numbers of fish in each cage are shown in Table 3.1. The time period for each experiment was selected to ensure a theoretical growth of approximately 25% using an exponential growth equation as follows: Wt2 = WtxeGl'100 (3.1) where Wt] and W t 2 are starting and ending weights of experimental fish, G is the specific growth rate in % body weight per day (obtained from feed table produced by feed manufacturers and t is the experimental period in days. 51 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM The existing feeding schedule on each farm was followed so that the experiments would be as non-disruptive as possible to existing routines and fish. Farm sites and cages were selected on the basis of having available six cages of similar fish strain, size and stocking density. Care was taken to select cages with similar water flow. Trials were conducted in triplicate with three camera-monitored cages and three control cages. A l l cages were randomly assigned. Details particular to each experiment follow. 3.3.2.2 Experiment 1 The experiment commenced on March 8, 1994 and lasted 56 d. The farm site consisted of 24 cages positioned in a two-by-two array. Cages measured 15 m x 15 m x 20 m deep. Experimental fish were Atlantic salmon of the Mowi strain, averaging 1.1 kg individual weight 3 3 with an initial stocking density of approximately 1.5 kg m" and 1.4 fish m" (Table 3.1). Experimental cages had similar water flow conditions because they were not positioned at system corners or adjacent to farm structures. They were located on both ends of the farm with half of the cages at either end. A l l of the fish on the farm were fed once a day on alternate days with half of the system (12 cages) being fed each day in the morning from 8:00 A . M . onward. They had been on that restricted feeding regime as set by the farm management since November 1993. Dry pelleted (44% protein, 18% fat) feed, 5-7 mm in diameter was offered. 52 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Table 3.1. Experimental conditions and results (CMC and c indicate camera-monitored and control cages, respectively). Fish size, unless stated otherwise, was obtained by using a video-sizing technique. Cage Initial Mort, Initial size, Final size, SGR % BW Total numbers % kg±SD kg±SD d"1 ration, kg BioFCR Experiment 1 1.1 CMC (Atlantic salmon) 6161 0.03 1.04±0.20 1.48+0.24 0.63 1910 0.71 1.2 CMC 5845 0.03 1.16+0.21 1.73+0.30 0.71 2039 0.61 1.2 CMC 6111 0.07 1.12+0.18 1.52+0.25 0.55 1796 0.74 Average 6039 0.04+0.02* 1.11+0.20 1.58+0.26 0.63+0.08 1915 0.69+0.07* 1.4 c 6727 0.13 1.05+0.19 1.47+0.25 0.61 2200 0.77 1.5 c 6035 0.10 1.22+0.25 1.72+0.26 0.61 2300 0.76 1.6 c 6439 0.28 1.10+0.15 1.52+0.30 0.58 2300 0.86 Average 6400 0.17±0.10* 1.12±0.20 1.57±0.27 0.60±0.02 2267 0.8010.06* Experiment 2 2.1 CMC (Atlantic salmon) 4687 0.38 3.88+0.64 4.71+0.95 0.47 5255 1.40 2.2 CMC 3873 0.41 3.90+0.85 5.0+1.00 0.61 3986 0.94 2.3 CMC 4974 0.26 3.34+1.00a 4.28+1.20 0.60 6079 1.30 Average 4511 0.350.08* 3.71±0.84 4.66±1.10 0.56±0.08* 5107 1.21+0.24* 2.4 c 5094 0.49 3.90+0.95 4.61+1.00 0.41 5425 1.50 2.4 c 3873 0.75 3.66+0.80 4.27+1.00 0.38 4125 1.80 2.6 c 5059 0.53 3.51±1.00 a 4.26+1.20 0.47 5500 1.50 Average 4675 0.59±0.14* 3.69±0.92 4.38±1.10 0.4210.05* 5017 1.6010.17* Experiment 3 3.1 CMC (chinook salmon) 7921 7.6 1.80+0.38 2.56±0.57b 0.56 7889 1.38 3.2 CMC 11892 5.7 1.87+0.41 2.44±0.53 b 0.42 8182 1.25 3.3 CMC 12717 6.1 1.67+0.35 2.35±0.54 b 0.54 9580 1.16 Average 10843 6.5+1.0 1.78±0.38 2.45±0.55 0.5110.07* 8550 1.26+0.11* 3.4 c 9843 6.5 1.91+0.42 2.38±0.44 b 0.35 8438 1.90 3.5 c 12303 6.4 2.06+0.43 2.31±0.51 b 0.18 9050 2.97 3.6 c 13411 6.7 1.88+0.42 2.28±0.57 b 0.31 9563 1.62 Average 11852 6.5±0.2 1.95±0.42 2.32±0.51 0.2810.09* 9017 2.1610.71* * Values are statistically significant (a=0.10, two-tailed) between C M C and Control cages. a Due to technical difficulties associated with the cage grading operations, video-sampling data could not be used. Average size of 100 fish in these two cages was obtained by weighing approximately 25 fish per seine for four seines during a cage splitting operation. b Due to technical difficulties associated with the camera equipment, video-sampling was not possible. Average size was, then, obtained by weighing 100 fish that were randomly selected from a physical sample representing approximately 10% of a cage population. SD: standard deviation; BW: body weight; SGR: specific growth rate; BioFCR: biological feed conversion ratio. 53 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM 3.3.2.3 Experiment 2 The feeding trial commenced on July 14, 1994 and lasted 41 d. The farm site consisted of 32 cages positioned in a two by two array as shown in Appendix C (first colour print). Cages measured 1 5 m x l 5 m x l 2 m deep. Experimental fish were Atlantic salmon of the McConnell strain, averaging 3.7 kg individual weight with an initial stocking density of approximately 6.2 kg m"3 and 1.7 fish m"3 (Table 3.1). A l l experimental cages were open to the sea on one side. One of the cages (Cage 2.1) was located adjacent to a floating building. Four of the cages were located at one end of the farm, while two others (1 control and 1 CMC) were located adjacent to each other in the middle of the farm. As set by the management, all the fish on this farm site had been fed twice a day, in the morning starting from 8:00 A . M . onward and again starting past 6:00 p.m. since late Spring. This feeding schedule was followed during the trial. Dry pelleted feed (46% crude protein and 23% fat), 9 mm in diameter was offered. 3.3.2.4 Experiment 3 The feeding trial commenced on September 8, 1994 and lasted for 63 d. The farm site has an assembly of twelve 1 5 m x l 5 m x l 6 m deep cages and four units of 30 m x 48 m x 30 m deep cages on the same system with the latter groups of cages sandwiched between two groups of 15 m x 15 m cages. Experimental fish were chinook salmon, Oncorhynchus tshawytscha averaging 1.9 kg individual weight with an initial stocking density of approximately 5.3 kg m"3 and 3.2 fish m" (Table 3.1). A l l six cages selected for the study were positioned in a straight row with four cages having one side facing the open ocean. One experimental and one control cage (end cages) had two sides facing the open ocean. 54 Chapter 3. EFFECT! VENESS OF A SUB-SURF A CE FISH AND FEED MONITORING SYSTEM As set by the farm management, all the fish at the farm site had been fed once a day in the morning starting from first light (usually 7:00 A.M.) and finishing around 11:00 a.m. to 1:00 P.M. prior to the experimental period, therefore, this feeding schedule was maintained. Dry pelleted feed (46% protein, 23% fat), 9 mm in diameter was offered. 3.3.2.5 Data collection and statistical analyses Environmental record. Water temperature, dissolved oxygen (YSI Model 57) and salinity (YSI Model 33) were measured daily at the surface, 5 and 7 m depths. Light intensity at the water surface and at different depths (LI-COR, LI-185B, LI-190 SA quantum sensor), climatic conditions and water transparency (Secchi Disc Reading) were also recorded daily at 2:00 P.M. Light intensities in air were recorded at the beginning and ending of a feeding event in Experiments 2 and 3, and only at 2:00 P.M. in Experiment 1. Feeding response in camera-monitored cages. At the start of a daily feeding, during feeding and at feeding endpoints, fish behaviour was recorded. The types of feeding behaviour were classified as mouthing, surface splashing, feed intake and foraging behaviour indicated by "S" shaped foraging pattern. Ubiquitous, few or nil were the adjectives to describe the number of such occurrences. The "S" shaped foraging pattern is similar to a very slowly manoeuvred "S" start, chasing pattern (Weihs and Webb, 1983). 55 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Growth and FCR calculations. The time to discharge the contents of a feed bag (25 kg) was measured, and any food remaining in a bag at the end of a daily feeding session was weighed. Feeding time excluded the time to inspect for pellet loss after a short pulse or short continuous feed. Standard deviations (SD) for all means (throughout this thesis) were calculated using the non-biased or "n-1" method with the following equation: For the purpose of this experiment, every fish mortality was weighed and the amount of weight gained by each mortality was obtained by subtracting the initial average fish size from the weight of the mort. The Biological Feed Conversion Ratio (BioFCR) per cage was calculated using equation 1.3 from Chapter las follows: where Final # and Initial # are the final and initial fish numbers; Final Wt and Initial Wt are the final and initial average fish weights (kg), and Biomass Morts refers to the sum of weight gained by individual mortality (i.e. weight of individual mort minus initial average fish size) accumulated over the period of growth. (3.2) BioFCR = Total dry ration used [1.3] (Final ttxFinalWt)- (Initial#xInitiaWt) + BiomassMorts 56 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Growth rate of experimental fish was calculated as Specific growth rate (% BW d"1) using Equation 1.1 from Chapter 1. Whenever possible, average fish size was measured using a computerised, video camera-sizing method developed in the Bio-Resource Engineering laboratory (University of British Columbia) (Naiburg et. al., 1993; Petrell et. al., 1996). When the video method was tested in sea cages with known individual salmonid weights, no significant difference was evident between video measured and actual average size provided the cameras were positioned in preferred fish swimming locations (Petrell et. al., 1996). The video method was chosen over the conventional seining and dip netting method because the sampling error (bias) in the conventional method was reported by participating farmers and Klontz (1993) to be 15%. Unless otherwise stated in Table 3.1, video fish weight estimates were averaged per cage at the beginning and end of the experimental period, and were used to calculate fish growth. To obtain the two readings, cameras were lowered to the lowest depth with acceptable background lighting. If sufficient numbers of fish were found at this filming depth, then one or two additional positions/depths were filmed. Cameras usually faced the net, because in that way better background lighting could be achieved. At least 30 to 40 min of footage was recorded at each camera position. Due to problems beyond our control, video sampling could not be conducted in two cages (camera-monitored cage 2.3 and control cage 2.6, first sampling) and all cages during final samplings in Experiment 3. In those cases, cages were sampled as noted in Table 3.1. Care was provided to ensure that the same and most accurate sizing method was applied to both experimental treatments. Sampling method can impact growth rate because Specific growth rates are calculated based on the initial and final weights of experimental fish according to Equation 1.1. 57 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Fork length, mass, girth and height of at least 50 fish per cage were obtained at the beginning and end of the experiments, and were used to check the calibration of the video sizing system. The minimum number, n, of video fish images sampled per video sampling position was calculated using a standard statistical method: „ . ( " ' > " , • * ' ) (3.3) where a 2 = the population variance, kg, D = half-width of the desired confidence interval (maximum acceptable difference between the sample mean mass and the population mean mass, i.e. 5% of the sample mean, kg) and 1.96 is the critical value of the cumulative normal variable at the 95% confidence interval. The sample number varied from 80 to 100 fish. Statistical analyses. The effect of eliminating or reducing feed wastage and extending the feeding volume on growth rate, FCR and mortality of experimental fish was difficult to anticipate because these variables are affected by many factors such as water temperature, dissolved oxygen profile, predators, disease, smolt quality and genetics. At the conception stage of this study, some farmers believed that some degree of feed wastage was necessary for maximum fish growth. The hypothesis in the two-sample t test selected for testing significant difference in growth between treatment and control cages was, therefore, expressed as a two-tailed hypothesis, i.e. Ho: \i\ = | i 2 and HA: \ii * u 2 . Similarly, for BioFCR, since by definition, BioFCR is a function of ration amounts fed over weight gained. The effect of controlling feed discharge using underwater cameras on mortality was also difficult to predict (two-tailed). The level of significance (a) for all tests was 0.10. This is because of the small size of the experiment (n=3). According to Steel and Torrie (1960), "The levels of 5 % and 1 % are arbitrary but 58 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM appeared to have been adequate in the field of agriculture where they were first used. In the case of small-sized experiments, it is possible that the null hypothesis will not likely be rejected if these levels are required, unless a large difference exists. This suggests the choice of another level of significance, perhaps 10 % for small experiments". Levels used other than 5 % and 1 % have to be clearly stated. The two-sample t test assumes, on a theoretical basis, that both samples come at random from normal populations with equal variances. These assumptions are not always true, for example, in biological studies. However, numerous studies have shown that the t test is robust enough to stand considerable departures from its theoretical assumptions, especially i f the sample sizes are equal or nearly equal, and especially when two-tailed hypotheses are considered (Zar, 1999). A l l data were analysed using statistical modules included in MS Excel (Microsoft Corp., 1997). Mean ration amounts fed before and after Day 17 in Experiment 1 were also analysed using the two-sample t test with the two-tailed hypotheses expressed as, Ho: \i\ = \ij and H A : \i\ * \i2- Variability in daily ration amounts between CMCs and control cages were tested using the F test Two-sample for variances in MS Excel (Microsoft Corp., 1997) with a = 0.05) and hypotheses expressed as, Ho: C M 2 = c>22 and 2 2 H A : O"I ^ 0 2 . The F test assumes that samples are taken at random from normal populations. Detailed summaries of selected analyses are attached as appendices in this thesis. 3.4 Results and Discussion Chinook and Atlantic salmon (>1.75 kg in size) in the CMCs achieved higher growth rates (0.56 % d"1 and 0.51 % d"1) when compared with fish in the control cages (0.42 % d"1 and 0.28 % d"1). Biological feed conversion ratios were also better or lower for fish CMCs (1.21 and 1.26) than for fish in the control cages (1.60 and 2.16). Smaller Atlantic salmon (<1.75 kg) 59 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM achieved better BioFCRs (0.69) in CMCs than in control cages (0.80) but growth in those cages was not significantly greater than in control cages. This was attributed to the feeding frequency (alternate day feeding) and poor pellet detection by the salmon throughout most of the experimental period. In CMCs, feed dispensation rate (kg feed MT-fish"1 min"1) varied depending on lighting conditions and broadcast method (and appetite), and ration level changed constantly. The camera monitoring system was successful, because it was able to extend the feeding volume thereby increasing the opportunity for fish to consume pellets throughout the water column, as well as control for pellet loss. 3.4.1 Environmental record 3.4.1.1 Experiment 1 Salinity varied over the course of the experiment from 14 to 32 ppt on the surface, and remained at approximately 30 ppt at a depth of 7 m. Low surface salinity values were due to runoff from spring rains and snow melts from nearby mountains. Visibility limit (water transparency) varied from 1.5 to 9 m. Low values (1.5 - 3 m during 75% of the experimental period) were due to either algae and/or silt. Light levels in air ranged from 100 to 900 //moi photon s"1 m"2. Dissolved oxygen (mg l"1) ranged from 8.5 to 12 at the surface, and was constantly 8 mg l" 1 at a depth of 7 m. Water temperature ranged from 7 to 12 °C on the surface, and between 7 to 8 °C at a depth of 7 m. 3.4.1.2 Experiment 2 Salinity varied over the course of the experiment from 21 to 25.5 ppt on the surface, and from 23 to 27.5 ppt at 7 m. Low surface salinity values were due to heavy rains and runoff from nearby 60 Chapter 3. EFFECTIVENESS OF A SUB-SURFA CE FISH AND FEED MONITORING SYSTEM mountains. Visibility limit (water transparency) ranged from 3.5 to 6 m. Low values were due to 1 9 the presence of algae. Light levels in air varied from 9 to 150 //moi photon s" m" at the beginning and from 100 to 900 //moi photon s" m" at the end of the morning feed; and from 25 to 900 //moi photon s"1 m"2 at the beginning and from 10 to 720 //moi photon s"1 m"2 at the end of the evening feed. Dissolved oxygen (mg l' 1) ranged from 7.4 to 10.4 at the surface, and from 6.4 to 9.5 at 7 m. Water temperature ranged from 15 to 18 °C on the surface, and between 12 to 14.5 °C at 7 m. 3.4.1.3 Experiment 3 Salinity varied over the course of the experiment from 21 to 25.5 ppt on the surface, and from 23 to 26.5 ppt at 7 m. Visibility limit (water transparency) ranged from 4.5 to 8 m. Light levels in air varied from 39 to 114 //moi photon s"1 m"2 at the beginning and from 20 to 1080 /miol photon 1 0 1 s" m" at the end of the feeding event. Dissolved oxygen (mg 1") ranged from 6.4 to 10.0 at the surface, and from 6.7 to 9.2 at 7 m. Water temperature ranged from 9.5 to 15.5 °C on the surface and between 7 to 15.0 °C at 7 m. 3.4.2 Experimental observations 3.4.2.1 Experiment 1 Fish would generally migrate toward the surface to feed, and this action would leave large open areas near the cage bottom with few fish. Generally at the end of a feeding event, fish were seen to be consuming pellets and lightly splashing on the surface even as pellets were being wasted. When the Secchi Disc reading went below 3 m, fish often appeared unable to capture the pellets, even at the lowest possible feeding rate. Fish feeding under poor visibility exhibited "S" shaped 61 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM foraging swimming patterns from the surface to camera depth. When two or more fish converged on a pellet, they often missed it completely, or frequently pellets bounced off their bodies. During periods of higher levels of visibility, "S" shaped foraging patterns were usually not evident at the surface. 3.4.2.2 Experiment 2 Crowding of fish, because they were larger, was greater in Experiment 2 than in Experiment 1. The camera in Experiment 2 often had to be placed near the cage bottom in order to be underneath the fish feeding school. Usually surface feeding activities of mouthing, splashing and surface feed intake were ubiquitous at the beginning of a feeding event but subsequently, only fish at the bottom of the school were still foraging and consuming pellets. There was no agreement between the levels of surface and bottom feeding activities 94% of the time at the feeding endpoint during the morning feed and 92% of the time during the evening feed. Approximately 82% and 74% of the time during the morning and evening feed respectively, no fish were evident on the surface when pellet loss was detected (i.e. the farmers, basing cessation of feeding on surface activity would, therefore, have stopped feeding before the researchers). Fish viewed immediately above the camera did not begin to feed actively until surface fish had discontinued eating, and usually did so by exhibiting "S" shaped foraging patterns and showing successful feed capture. No signs of competitive or aggressive behaviours were evident at the camera level. 62 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM 3.4.2.3 Experiment 3 Fish would generally migrate upward to the surface to feed as in Experiments 1 and 2. At the start of feeding, vigorous surface activity usually occurred. On a few days, fish did not eat. The camera was typically at 8 m of depth. Sub-populations of fish, comprising 30 to 40 individuals, would swim gently up and down past the camera level in gradual waves, with each group sometimes stretched out in a wide spiral over several meters, as if following a leader throughout the feeding event. Surface feeding activities (some feed intake, no mouthing or splashing) continued to be obvious but weak when bottom feeding activity indicated a lack of feed consumption and pellet loss was observed. On average, this occurred 80% of the time during the 56 d of experimentation. During periods of good visibility (>7 m) fish were occasionally observed to follow pellets down to the camera level. 3.4.3 BioFCR, growth, mortality, variation in daily ration amounts and feed dispensation The average amount of food pellet dispensed per fish was not significantly different (P < 0.10) among the six test cages in all three experiments (data not shown). No significant change in size distribution in the cages occurred from the beginning to the end of the experiments (P < 0.10). The duration of the trial period was probably too short for measurable changes to occur in size distribution. Mortality was less than 1 % in terms of initial numbers for Experiments 1 and 2 and was 6.5% for Experiment 3 (Table 3.1). Predation loss contributed greatly to the losses in Experiment 3 but to a much lesser extent in Experiments 1 and 2. Individual weights of mortalities resulting from predation could not be measured (because of carcass mutilation) and weight gained by these morts were consequently estimated from average weights of other morts retrieved during the 63 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM dive. The mortality rates for CMCs were significantly lower than in the control cages during Experiment 1 (t = -2.212, df=4,p = 0.0914) and Experiment 2(t = -2.583, df=4,p = 0.0611), both of which were conducted at two different stages of the growth cycle. Our practice of inducing fish to feed lower in the water column to avoid surface 'feeding frenzies' probably reduced the number of stressful events and, therefore, the mortality rate in the CMCs. There were no significant differences (t = -0.114, df = 4,p = 0.9148) in mortality rates for fish in CMCs and control cages in Experiment 3. (See Appendix A - l and A-2 for details). 3.4.3.1 Experiment 1. The BioFCR in Experiment 1 for the CMCs were significantly better (i.e. lower) than the control cages by 13% (t = -2.176, df = 4, P O.0476) (Table 3.1). Growth rates, on the other hand, were not significantly different from those of the control cages. A l l the six cages had BioFCR values that were below one, which may be typical of fish that had been on an alternate day feeding regime for several months. Several investigators (e.g. Austreng et al, 1987; Storebakken and Austreng, 1987a, 1987b) reported FCR values for salmonids as low as 0.9 and 1.1 for fish ranging in size between fry and 5 kg at temperatures between 2 and 16 °C. Growth rates of 0.55 to 0.71 % d"1 in the present study were slightly lower (being on alternate day feeding regime) but within the range for growth rate of Atlantic salmon (600-2000 g) in sea cages provided by Austreng et al. (1987). Mass gain appeared to occur mostly along the length of the fish, because they appeared to be thin and long. Calculations of average growth based on increases in fork length (0.21% d"1), height (0.14% d"1) and girth (0.05% d"1) over the experimental period confirmed this conjecture. 64 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM It is probable that poor visibility for most (75%) of the experimental period in Experiment 1 contributed to the difference in BioFCR. Failure of the fish to detect pellets, even at the lowest possible feed discharge rate in the CMCs during times of heavy siltation, prevented researchers from feeding fish completely to satiation. CMCs achieved better BioFCRs because the control cages were underfed as compared to the CMCs when the visibility (mean Secchi Disc Reading of 7.2 ± 1.2 m) allowed effective food detection during the first 25% of the experimental period and had wastage with poor visibility (mean Secchi Disc Reading of 3.0 ± 0.9 m) as compared to the CMCs during the remaining 75% of the experimental period. The idea that visibility affected daily ration was tested after graphing the percent difference in average daily ration between CMCs and control cages (Figure 3.3 a). In response to the marked reduction in visibility from the first 25% to the remaining 75% of the experimental period, the corresponding reduction in daily ration for the CMCs (38.2 ± 1.8%) were significantly greater (t = 11.2558, df = 4, p = 0.0004) than those for the control cages (mean 8.6 ± 9.3%) (Table 3.2). Figure 3.3 b and Figure 3.3 c show that the ration variability between cages and daily variability was much higher in the CMCs (F = 3.719, vj = V 2 = 92,p = 0.0000) than in the control cages in Experiment 1. Fish in CMCs were, therefore, being more precisely fed by controlling pellet dispensation rate through prevention of pellet loss and promotion of feeding lower in the water column. Light intensity and turbidity in water inevitably combine as factors to affect the detectability of food pellets by fish. Turbidity affects foraging success of particulate feeding juvenile and adult fishes by reducing reaction distance (Vinyard and O'Brien, 1976; Crowl, 1989). With formulated pellet food particles, the visual characteristics of size, shape, colour and movement are as important as gustatory characteristics in improving pellet capture and ingestion 65 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM by salmonids including Atlantic salmon parr (Stradmeyer et al, 1988; Stradmeyer, 1989). In the present experiments, poor pellet detectability corresponded to "S" shaped foraging patterns without obvious signs of competitive behaviours. Table 3. 2. Mean ration amounts before and after Day 17 in Experiment 1. (CMC and c indicate camera-monitored and control cages, respectively). Cage Mean ration amount before Day 17, k g ± S D (Mean visibility, 7.2 m) Mean ration amount after Day 17, kg± SD (Mean visibility, 3.0 m) Reduction, % ± SD Experiment 1 (Atlantic salmon) 1.1 CMC 110.4 66.3 39.9 1.2 CMC 105.1 64.8 38.3 1.2 CMC 98.6 62.8 36.3 Average 104.7 ± 5.9 a 64.6 ± 1.8 b 38.2 ± 1.8 1.4 c 77.8 79.5 -2.2 1.5 c 97.2 83.0 14.6 1.6 c 91.7 79.5 13.3 Average 88.8 ± 10.0 a 80.7 ± 2.0 a 8.6 ± 9.3 Values with different letters are significantly different (cc=0.05) for mean ration amounts before and after Day 17. SD: standard deviation. 66 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM (Rc-Rcmc)/Rc x 1 0 0 + / - 3 SD Figure 3.3 a. Graph to show the percent differences between ration fed to control cages and camera-monitored cages (CMCs) in Experiment 1. CMCs achieved better (lower) BioFCRs than the control cages because the control cages were underfed (negative y values) as compared to the CMCs when good visibility before Day 17 allowed effective food detection and had feed wastage (positive y values) with poor visibility as compared with the CMCs after Day 17. Rc is the average ration in the control cages, R c m c is the average ration in the CMCs, and the estimation of uncertainty was calculated using root-sum-square equation, which estimates uncertainty at the 99.9 % confidence interval (or 3 SD) (Doebelin, 1975). 67 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM 0.020 0.018 JS 0.016 c M 0.014 _c " M - i 0.012 E O 0.010 rati 0.008 & 0.006 cs Q 0.004 0.002 0.000 10 Day 17 5 20 25 30 35 40 45 50 55 Day • • Cage 1.1 Cage 1.2 Cage 1.3 60 Figure 3.3 b. Variation in daily amount of food fed to fish in camera-monitored cages (CMCs) in Experiment 1. The variation is probably in response to variation in appetite levels of fish. In response to the marked reduction in visibility starting on Day 17 to the remainder of the experimental period, the corresponding reduction in mean daily ration in the CMCs was 38.2±2.5%. 68 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Figure 3.3 c. Variation in daily amount of food fed to fish in control cages in Experiment 1. Ration in control cages, which were fed based on perceived levels of surface activity, did not decline due to poor water clarity starting on Day 17 because fish continued to exhibit signs of feeding activity at the surface even though they could not easily detect pellets. 69 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM 3.4.3.2 Experiment 2. In Experiment 2, the mean BioFCR in the CMCs was 1.2, and the mean BioFCR in the control cages was 1.6. Values for BioFCR for fish in the CMCs were significantly lower (by 24 %) than for fish in the control cages (t = -2.2508, df = 4, p = 0.0876). Growth rates (% body weight d"1) for CMCs were significantly higher (t = 2.7135, df = 4, p = 0.0533) than those of the control cages (Table 3.1). The mean growth rate (0.56 % d"1) in the CMCs agreed with figures for fish over 2000 g size, reported by Austreng et al. (1987) for temperatures around 12 and 14 °C, while the mean growth rate (0.42 % d"1) of the control cages was lower. The latter would indicate poor food consumption, possibly as a result of feed wastage or that adequate ration was not provided (Figure 3.4a). Figure 3.4 b and Figure 3.4 c show that the ration variability between cages and daily variability was much higher in the CMCs (F = 1.4784, V [ = 120, v 2 = 122, p = 0.0162) than in the control cages in Experiment 2. Fish in CMCs were, therefore, being more precisely fed by controlling pellet dispensation rate through prevention of pellet loss and promotion of feeding lower in the water column. It can be ascertained that the higher growth rates achieved by fish in CMCs were due to the large feeding volume in the CMCs as compared to control cages. The choice location for fish in a sea cage in terms of food density and detectability is the upper few meters of a cage, which should imply that most fish would eventually eat there. If stocking density was high enough to prevent the immediate usage of the upper few meters of the cage by all fish, then the most dominant fish would probably feed first, unless they had chosen to pursue other activities such as predator or light avoidance (Ferno et al., 1995). Although it is has been assumed that high density prevents defense of territory, Ferno and Holm (1986) found that at approximately 10 kg 70 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM m"3 a few dominant Atlantic salmon parr could defend certain areas of a tank and thus, possibly, gaining preferential access to any food that arrived in these areas. 40 "(Rc-Rcmc)/Rc x 1 0 0 + / - 3 SD Figure 3.4 a. Graph to show that fish in camera-monitored cages (CMCs) achieved better BioFCRs in Experiment 2 because of poor food consumption by fish in the control cages, possibly as a result of feed wastage or that adequate ration was not provided. R c is the average ration in the control cages, R c m c is the average ration in the CMCs, and the estimation of uncertainty was calculated using root-sum-square equation, which estimates uncertainty at the 99.9 % confidence interval (or 3 SD) (Doebelin, 1975). 71 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM Cage 2.1 Cage 2.2 Cage 2.3 Figure 3.4 b. Variation in the daily ration in camera-monitored cages (CMCs) in Experiment 2. The ration variability between cages and daily variability was much higher in the CMCs than in the control cages (Figure 3.4c). CMCs were, therefore, being more precisely fed by controlling pellet dispensation rate through prevention of pellet loss and promotion of bottom feeding. 72 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM 0.0121 Cage 2.4 Cage 2.5 Cage 2.6 Figure 3.4 c. Variation in the daily ration in control cages in Experiment 2. All cages were fed according to surface activity. 73 Chapter 3. EFFECTIVENESS OF A SUB-SURFA CE FISH AND FEED MONITORING SYSTEM In a sea cage, dominance relationships could develop when stocking density is low, and once established persist until the cage is harvested. It has been suggested in a review by Metcalfe (1990) that while territoriality is a strategy that may be turned on or off, dominance relationships are harder to eliminate. A feeding system that favours dominant fish would, therefore, promote poorer growth by subordinates because they must remain in or move to less profitable feeding locations. To overcome problems associated with dominant fish, it has been stated that food should be dispensed to fish such that it is indefensible (Dill et al., 1981; Puckett and Dill , 1985). Feeding fish so that the food enters the water in one plane (i.e. the surface plane) makes it difficult to ensure this condition of indefensibility. The problem was compounded in the control cages where satiation was judged based on surface activity. Premature cessation of feeding thus resulted in limited accessibility of food to fish lower in the water column because video observations indicated that fish lower in the water column fed actively after fish at the surface showed no further interest in feeding or had disappeared from the surface. 3.4.3.3 Experiment 3. In Experiment 3, the mean BioFCR in the CMCs was 1.26, while that of the control cages was 2.16. The BioFCR values for fish in the CMCs were significantly improved (by 42 %) over fish in the control cages (r = -2.162, df = 4, p = 0.0967). The mean growth rate for CMCs (0.51% body weight d"1) was significantly higher (t = 3.4601, df=4,p = 0.0258) than that of the control cages (0.28% body weight d"1) by 82% (Table 3.1). The growth advantage gained by fish in the CMCs may be as a result of underfeeding in control cages as compared to the CMCs (Figure 3.5 a). 74 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM The feeding behaviour of the chinook salmon in Experiment 3 differed from that of Atlantic salmon used in Experiment 2 in the CMCs. Instead of disappearing completely from the surface near the end of a feeding event like the Atlantic salmon usually did, at least a few fish were often observed to feed near the surface at the end of a feeding event. This persistence of surface feeding activity alone could not provide accurate cues for judging overall satiation. Figure 3.5 b and Figure 3.5 c show that the ration variability between cages and daily variability was much higher in the CMCs (F = 1.7362, vi = V 2 = 59, p = 0.0180) than in the control cages in Experiment 3 (similar to Experiments 1 and 2). Fish in CMCs were, therefore, being more precisely fed by controlling pellet dispensation rate through prevention of pellet loss and promotion of feeding lower in the water column. Unequal FCR values between test and control groups of fish in other feeding experiments have also been explained by food wastage and/or sub-optimal amounts of ration in control groups (Juell et al, 1993). Alanara (1992) found that growth and FCR were affected by restricting access time to demand feeders. Maximum daily satiation time varied in the experiments from 25 to 120 min in Experiment 1, 10 to 70 min in Experiment 2, and 10 to 51 min in Experiment 3. Satiation time (approximately 43 min) for small sockeye salmon (Brett, 1971) was similar to the average time for Atlantic salmon in Experiment 2 at similar water temperatures. Extremely short satiation times could probably be attributed to the presence of predators. These values for satiation times were similar to those noted in a study using the Aquasmart adaptive feeding system equipped with a pellet sensor (Blyth et ah, 1993). 75 Chapter 3. EFFECTIVENESS OF A SUB-SURFA CE FISH AND FEED MONITORING SYSTEM (Rc-Rcmc)/Rc x 1 0 0 + / - 3 SD Figure 3.5 a. Graph to show that fish in camera-monitored cages (CMCs) achieved better (lower) BioFCRs than control cages in Experiment 3. The growth advantage gained in the CMCs was due to control cages being underfed on average as compared to the CMCs. Rc is the average ration in the control cages, R c m c is the average ration in the CMCs, and the estimation of uncertainty was calculated using root-sum-square equation, which estimates uncertainty at the 99.9 % confidence interval (or 3 SD) (Doebelin, 1975). 76 Chapter 3. EFFECTIVENESS OF A SUB-SURFA CE FISH AND-FEED MONITORING SYSTEM 0.018 0.016 Cage 3.1 Cage 3.2 Cage 3.3 Figure 3.5 b. Variation in the daily ration consumed by fish in camera-monitored cages (CMCs) in Experiment 3. As in Experiments 1 and 2, the ration variability between cages and daily variability was much higher the CMCs than in the control cages (Figure 3.5c) in Experiment 3. 77 Chapter 3. EFFECTIVENESS OF A SUB-SURFA CE FISH AND FEED MONITORING SYSTEM 0.014 i ja 0.012- /': Cage 3.4 Cage 3.5 Cage 3.6 Figure 3.5 c. Variation in the daily ration consumed by fish in control cages in Experiment 3. All cag< were fed according to surface activity. 7 8 Chapter 3. EFFECTIVENESS OF A SUB-SURF A CE FISH AND FEED MONITORING SYSTEM Feed discharge rates in CMCs averaged approximately 0.20 ± 66% kg feed MT-fish"1 min"1 in Experiments 2 and 3, and 0.12 ± 35%) kg feed MT-fish"1 min"1 in Experiment 1. It varied with Secchi Disk Readings within the range of 1.5 to 9 m in Experiment 1 (Figure 3.6 a). With true ration required by fish varying daily and not known a priori, a fixed discharge rate in terms of kg feed min"1 and satiation time could be wasteful. We have found that this is because the discharge rate should be varied according to appetite and fish activity near the cage bottom. Discharge rate should, however, be as fast as possible without generating pellet loss throughout the feeding event. Other feed dispensation studies where food intake has been measured have not found discharge rate to affect growth (e.g. Juell et al. 1994). 79 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM •a S K 0.25 0.20 0.15 0.10 0.05 Y=0.01x+0.07 R 2=0.45 Outlier points u.uu 1 2 3 4 5 6 7 Water visibility (Secchi Disc Reading, m) Figure 3.6 a Graph to show increasing feed discharge rate (kg feed MT-fish"1 min"1) as water visibility increased in Experiment 1. When the Secchi Disc Reading fell below three, fish could not always capture pellets. At 9 m of visibility, fish in two cages (i.e. the two outlier points) showed little appetite which could be attributed to a predator /light avoidance behaviour. 3.5 Conclusion Subsurface-feeding activities and pellet loss constitute better indicators of satiety in salmon reared in sea cages than do surface feeding activities. Feeding to the point that feed wastage occurs is not a necessary condition for achieving maximum growth. For Atlantic salmon, average BioFCR values for camera-monitored and control cages were 0.95 and 1.20, 80 Chapter 3. EFFECTIVENESS OF A SUB-SURFACE FISH AND FEED MONITORING SYSTEM respectively. This represents a 21% improvement in BioFCR for the CMCs. The corresponding improvement in BioFCR for chinook salmon in CMCs is 42% over fish in control cages. These figures translate to 37% improvement in BioFCR for both species over their respective overall industry averages of 1.5 for Atlantic and 2.0 for chinook salmon (British Columbia Salmon Farming Association). When not on a starvation feeding regime (Experiments 2 and 3), fish in CMCs had improvements of 34% and 82% higher growth rates over fish in control cages for Atlantic and chinook salmon respectively. While there was no difference in the mortality rates for camera-monitored and control cages for chinook salmon, mortality rates in the CMCs were 56-76 % lower than the rate in the control cages for Atlantic salmon. Under practical conditions where the complexity of allocating food when the appetite level of fish, water quality, fish distribution and lighting conditions are uncontrollable, the use of an underwater viewing device to judge feeding activity is of critical importance. A camera-monitored system permits direct visual detection of fish and pellets without any need for system calibration, immediate and direct input as to why pellets are being wasted and easy camera placement within a cage once the vertical distribution of fish is established. In this study, the underwater camera was shown to effectively extend the feeding volume because fish no longer congregate on the surface to feed and instead re-distribute themselves vertically as feed pellets become available even at greater depths (see Figure 3.2; 40). The future vision is a feeding system operated by a fish farmer who sits in front of a monitor and electronically controls mechanical feeders in many cages. 81 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CA GES Chapter 4 GROUP FISH FEEDING PATTERNS IN CAGES 4.1 Summary Prior to these experiments little was known about how, where and when fish eat during a feeding event or i f large numbers of fish feed simultaneously. To investigate these questions, observations of subsurface and surface feeding activities were recorded on salmon farms that utilised a variety of feeding techniques. Through this research, several distinct feeding patterns were discovered. When the visibility limit (water transparency) was less than 3 m, fish apparently could not readily detect pellets. Under higher visibility and i f fish were being continuously fed, they formed a ring-like structure, characterised by organised foraging. When fish were being fed in batches and while feed was still in the water, they initially formed a spiral-like structure, and then started to forage in a disorganised style. The foraging pattern by individual fish was an "S"-shaped searching manoeuvre, and the amplitude of a swerve was greater in batch feeding than in continuous feeding. Feed wastage varied with feed delivery rate and the method used to judge feeding endpoint. Wastage and underfeeding were associated with ration feeding. With the exception of one farming site, simultaneous feeding at the surface and near the cage bottom was uncommon. One of the hypotheses, formulated from the data, is that uniform feeding may be achieved if feeding rate is optimal and pellets are detectable throughout the water column. 82 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CAGES 4.2 Introduction Ideally, individual fish in a cage should have a choice of where to eat and for how long, and whether to eat or to pursue other activities such as predator avoidance (Greig-Smith, 1990; Metcalfe, 1994). In reality, feeding systems have largely been designed for material and labour efficiencies, and have ignored the animal. Better systems could be designed if more information on how to feed whole populations of fish in commercial-size cages was available. Competitive behaviours could be reduced if all fish could feed similarly throughout the cage (i.e. uniformly) because they would have ample access to food (Jobling et al., 1995). Moreover, the overall feeding time for an entire population would be lower than the time needed to feed individual or groups of individual fish. Increasingly, the behaviour of caged fish is being studied and incorporated into the design of new feeding systems as new sensing devices are being developed (e.g.: Skjervold, 1993; Juell et al., 1993; Juell, 1991) although, the test cages are often smaller than those commonly used in the industry. Juell and Westerberg (1993) developed a technique to track an individual salmon in a 7 m deep cage. The fish tracked during a feeding event did not always move into a zone where feed was available. In that experiment fish were being fed to a set ration with an automatic feeder, which spread batches of feed every 3 min. In experiments conducted in the first part of the present study (Chapter 3), underwater video cameras were effectively used to control continuous feeding of fish in 20 m deep cages on salmon farms. The density of fish in various sections of a 7 m deep cage was monitored during continuous feeding in another investigation using echo integration (Juell et al., 1994a). 83 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES The specific objectives of the portion of the study described in this chapter were: (1) To describe overall group feeding patterns in terms of general fish feeding behaviours that occurred while using different feeding systems under field (fish farming) conditions (2) To measure times spent by fish feeding at different levels including uniform feeding times relative to total feeding times Fish behaviours and pellet wastage were recorded prior to, during feeding, and at feeding endpoints with the use of underwater cameras. Surface feeding behaviours were compared to subsurface behaviours. Observations were used for formulating hypotheses for future research work and suggesting ways to feed fish more effectively. 4.3 Materials and methods Descriptive research on feeding was conducted on different farming sites rearing either Atlantic, Salmo salar L. or chinook, Oncorhynchus tshawytscha salmon of different sizes and stocking densities under different feeding conditions and seasons (Table 4.1). For details on uses of and differences between description and experimental research see Lehner (1996). A l l fish were tested under the regular feeding regime of each farm site. Data collected while testing a new feeding system at three sites A , B and C were analysed and reported in Chapter 3. The testing period at each site ranged from 2 to 3 months. Data on growth and feed conversion from sites A and B (Atlantic salmon) and C (chinook salmon) are described in Chapter 3. An in-depth description of the fish behaviour recorded during these three trials will be reported in this chapter. As these trials report the first time that subsurface and surface feeding activities have been examined, additional field observations were recorded (as in the longer-term 84 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES experiments) on a broader range of sites and feeding conditions to search for common behavioural patterns. Existing feeding systems at sites D through H (Table 4.1) were, therefore, studied. The experimental time frame (3 to 4 d) was shorter than the time frame in the previously mentioned longer-term trials, since growth and feed conversion were not experimental variables. Water temperature and dissolved oxygen (YSI Model 57) were measured daily at the surface and at 5 m depth. Water transparency (limit of water visibility; Secchi Disc Reading) was recorded daily at 2:00 p.m. (see Table 4.1, for environmental characteristics of sites). A Secchi Disc Reading is the water depth at which a black and white disc is just visible to the human eye. 4.3.1. Underwater viewing equipment The same viewing set-up used during the long-term trials was used for the short-term trials (see Section 3.3.1.1 for details). In all trials, a camera was placed in the cage centre facing the surface. It was suspended by a rope to a depth below the swimming aggregation, and was used to check for pellet loss while providing a more complete picture of fish distribution and behaviour in the water column. The average camera depth was approximately 8 m but ranged from 5 to 12 m depending on the vertical swimming distribution of fish and cage depth. Feeding events were videotaped using the underwater camera once a week during the long term feeding trials, while all short-term trials were taped. Tapes were used to track fish movement, and to describe the structure of the visible feeding aggregation in terms of feeding and swimming patterns. 85 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Table 4. 1. Site characteristics of farms visited and the number of experimental feeding events monitored. Long-term trials were conducted at sites A, B and C. Site # exp. # feeding Fish Cage size, Stocking Water Dissolved Visibility, Season3 cages events size, kg m density, kgm"3 temp., °C oxygen mg r1 m A 1 3 84 1.2 15x15x20 1.5 7.0-12.5 7.1-11.7 1.2-9.0 W/Spr B 1 3 246 3.7 15x15x12 6.2 14.0-18.0 6.6-10.4 3.5-5.0 Sum C 2 3 180 1.9 15x15x16 5.3 9.5-15.2 6.8-11.2 4.5-8.0 Sum/F D 2 2 3 2.9 15x15x16 5.2 7.9-9.0 8.9-9.0 5.5-8.0 W E 1 5 10 1.0-1.8 15x15x12 1.5-3.6 9.0-10.0 8.0-8.9 8.0-13.0 Spr F 1 2 2 9.1 15x15x12 11.9 9.0-10.0 8.0-9.0 7.0-8.0 Spr G 2 1 3 2.1 13x13x20 3.1 6.8-7.2 9.3-9.9 13.0 Sum H 1 3 4 2.5-2.9 15x15x20 4.4-7.8 n/a n/a 2.5-3.5 Sum n/a: Not available. xSalmo salar 2Oncorhynchus tshawytscha 3 W: winter: Spr: spring; Sum: summer; F: fall/autumn. 86 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES 4.3.2. Feeding methods A broad spectrum of feeding techniques commonly used in the industry was examined (Table 4.2), and all fish were well accustomed to their particular feeding system. Common to all systems was the size and location of the pellet scattering area. Pellets were discharged toward the cage centre and covered over a third of the cage surface. Discharge techniques included combinations of either mechanical or manual feed deliveries, and batch or continuous broadcast systems. Techniques were further classified according to meal frequency and discharge rate or output. Feeding endpoints were based on either a predetermined ration, a modified ration, surface activity, pellet loss and subsurface activity (camera system) or just pellet loss (e.g. uplift/feedback system). Specific terminology is described below: Batch feeding refers to pellets being scattered either manually or mechanically in a burst-like fashion (sites D-H). On farms that practised batch feeding, there would be several feeding bursts in quick succession, then either a short or long pause or interval, before the next series of shortly spaced bursts. In continuous feeding (sites A , B and C) there were generally no bursts or pauses. At Site C (Table 4.1), however, there was a pause after 75 kg of feed were fed into a cage in order to use the mechanical feeder to feed two other cages. To check for a feeding response, batch feeding was briefly employed on sites A and C at the beginning of a feeding event and when a bag of feed (25 kg) had nearly been discharged. Continuous broadcast feeding was achieved by using a mechanical, feed blower with a movable spout. The spout was manually moved from side to side and in a circular motion in order to continuously scatter feed evenly and randomly until the feeding endpoint was reached. 87 Chapter 4. GRO VP FISH FEEDING PA TTERNS IN CA GES Table 4. 2. Feeding technique and feeding endpoint at different farming sites. Site Discharge technique Output rate Feeding interval Meal frequency Feeding endpoint method A Continuous, mechanical blower Variable to 0.00025 kg/kg fish/min None Every other day Video camera method B Continuous mechanical blower Variable to 0.00025 kg/kg fish/min None Twice a day Video camera Method C Continuous mechanical blower Variable to 0.00025 kg/kg fish/min Paused after 75 kg Once a day Video camera method D Batch (large portions) manual hand feeding 5 kg/scoop, 2-3 scoops in succession 2-5 min Once a day Ration E Batch (small portions) rotary 0.25-0.5 kg/burst lasting 2-3 s 5-7 min 3 times a day Surface activities Batch (small portions) manual hand feeding 0.25-0.5 kg/scoop, 2-4 scoops every 1-2 min for 10 min Variable 3 times a day Surface activities F Batch (small portions) rotary 0.25-0.75 kg/burst lasting 2 s 8 min 3 times a day Surface activities Batch (small portions) manual hand feeding 1.0 kg/scoop, 2 scoops every 2 min for 10 min Variable 3 times a day Surface activities G Batch (medium portions) manual hand feeding 2.5 kg/scoop, 2-10 scoops in succession 10-15 min Once a day Feedback feed control system H a Batch (variable portion) programmable blower One 5-9 kg burst every 7-13 min 45 min to 1 hr Twice a day Modified ration (with surface activity when topping up) H b Nearly continuous batch (variable portions) programmable blower 1 -9 kg/burst every 9-68 s for 3-9 min 1-2 hr Twice a day Modified ration (with surface activity when topping up) a: Stocking density of 4.4 kg m" b: Stocking density of 7.8 kg m' 88 Chapter 4. GRO UP FISH FEEDING PA TTERNS IN CA GES At Site H, one experimental cage was fed nearly continuously as 1-9 kg of feed was discharged every 9-68 s for 3-9 min between non-feeding intervals of 1-2 h (to feed other cages). Output defines the amount of feed (kg) in a single burst as in the amount of feed in a standard scoop used in hand feeding, or pulse of a rotary or blower feeder. The feeding endpoint as judged using surface activity occurred when pellets were not consumed before they appeared, from the surface, to be fading into the water column (sites E and F). The output rate (kg per given time) in experiments using surface activity and ration (Site D) for determining the feeding endpoint was judged relative to the level of surface feeding activity. Mouthing (i.e. seeing open mouths), surface splashing and visual evidence of feed consumption were the commonly used indicators of surface feeding activities. Output rate was increased or decreased depending on the number of these indicators. Output of the mechanical feeders at sites A, B, C, E , F and H was adjustable. In a modified ration method (site H), a predetermined ration was discharged using a programmable automatic feed blower ( A K V A ) in feeding batches followed by "topping up" with batches of extra feed (2-5 kg burst every 23-43 s) until no apparent surface feeding activity. Two types of pellet detection systems were examined, namely the Feedback Feed Control System (see Talbot, 1995) (Site G) and the video camera method (see Section 3.3.1.1, for camera positioning) (sites A , B and C). In the Feedback Feed Control System, feed at the bottom of the cage was returned to the surface via an airlift pump. In both types of pellet detection systems, the feeding endpoint was when pellet loss could not be prevented after adjusting the output rate. Output rate in the video camera method was also adjusted to meet the level of foraging activity at the bottom of the feeding aggregation. 89 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CA GES 4.3.3. Feeding behaviours Feeding behaviours near the surface and at camera level (subsurface) were recorded. Swimming formations were described. A regular swimming formation was described as fish swimming around a cage forming circular rings similar to a closed-figure school formation (Sutterlin et. al., 1979; Breder, 1965). In a "regular" swimming formation, rotational speeds of the fish that form the rings appear similar. Modes outside the "regular" swimming formation were described as they occurred. Surface feeding activities included mouthing, splashing, foraging and feed capture. Bottom feeding activities included foraging and feed capture. Foraging activity by individual fish was described as a fish breaking away from a regular or directed swimming mode to exhibit searching action with its head swerving sharply to the right or left (Figure 4.1) in a classical "S" shape style (Weihs and Webb, 1983). The "S" shaped foraging pattern was also described as being executed either quickly as in a chase or very slowly as in a search. Avoidance (a fish approaching a food pellet turns away to avoid a competitor approaching the same pellet) or beaten (fish attempting to catch a sinking pellet is beaten to it by a competitor without any sign of avoidance) behaviours were recorded as feeding competition (Kadri et. al., 1996). Aggressive behaviours were recorded when nipping or chasing was observed. Activities were subjectively quantified as either ubiquitous (apparently involving the majority of the visually perceivable fish), numerous (many but not all fish) and little or no perceivable activity. Pellet loss was recorded when pellets were observed to fall through the camera view area. Loss was recorded as numerous when more than ten pellets were evident over several frames. 90 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CA GES The amount of time spent by fish feeding near the surface, at the camera level, and feeding at both of those levels; and the time elapsed before fish started to feed near the camera level were extracted from data sheets. For the long-term trials, a sample containing 21 consecutive feeding events were analysed because the variance generally remained large relative to the average even as sample size increased. Uniform feeding time was defined as the amount of time during a feeding event that fish were found feeding both near the surface and camera level. These various feeding times were expressed as a percentage of the total feeding time for each feeding event to standardise units for comparison. 91 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CA GES Rope suspending the camera Figure 4. 1 (a) Starting position for a foraging fish (marked with a white dot), when executing an 'S ' -shaped swimming pattern, (b) Mid-foraging position for the same fish, (c) End-foraging position for the same fish before repeating the sequence. Frames are 0.2 s apart. 92 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Figure 4.1 (continued). 4.4 Results No differences between chinook and Atlantic salmon in terms of feeding behaviours reported herein were observed. Overall, in batch feeding systems, different levels of pellet wastage were observed during a feeding event and at the feeding endpoint. Independent of the feeding system, fish appeared to have difficulty locating pellets when the limit to visibility was less than 3 m. Under higher visibility, two distinct group-swimming formations emerged, one was associated with continuously fed fish and the other with batch fed fish. No feeding system was consistently able to provide for uniform feeding. Generally, the levels of mouthing, splashing, foraging and feed capture were ubiquitous to numerous on the surface at the start of a feeding event, while below at the camera level, fish 93 Chapter 4. . GROUP FISH FEEDING PATTERNS IN CAGES would be seen to be migrating upward in the water column. Exceptions to this initial feeding response were found at chinook salmon sites C (video camera method: ten out of 180 feeding events), D (fed to ration: one out of three feeding events) and G (Feedback Fed Control System: three out of three feeding events). Apart from one feeding event at Site D, when no initial surface feeding response was evident, fish subsequently, did not eat elsewhere as well. The presence of predators (seals) was offered by farmers as an explanation for the apparent lack of appetite. Fish lacking an initial feeding response formed an inverted bell-formation with a large hole in the middle, few fish circulating along the perimeter and several to few fish at the cage bottom. The Feedback Feed Control System at Site G malfunctioned and observational results from this site were only mentioned here to describe a fish aggregation when fish had no feeding response. At Site C (feeding to pellet loss with the video camera method), pellet loss was low even when fish did not eat because output was low and feeding was discontinued when it became apparent that the fish were not foraging or eating. When there was an initial feeding response somewhere in the cage, different swimming formations and levels of pellet wastage were observed to be associated with feeding methods, and stocking and visibility conditions as described in the following sections. 4.4.1. Low visibility and light intensity Visibility (< 3 m) and light levels near twilight were observed to affect foraging success. During 52% of the feeding events at Site A and 50% of the feeding events at Site H, the water visibility was less than 3 m. As fish were being fed under poor water visibility and varying levels of light intensity, mouthing and splashing activities were ubiquitous to numerous while pellet loss was unavoidable (Site A , continuous feeding, video camera method) or numerous to 94 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CAGES substantial (Site H , 5-9 kg per 6-8 s burst feeding every 7-13 min., modified-ration method). Pellet loss was substantial, although fish were actively foraging while being fed near twilight (sunset) during one feeding event at Site H (1-9 kg/burst every 9-68 s for 3-9 min with 1-2 h non-feeding interval, modified-ration method). Fish feeding under apparent conditions of poor pellet detectability exhibited 'S'-shaped foraging manoeuvres both near the surface and down to 5-7 m under the water (Figure 4.1). The head swerves in the manoeuvres were widely executed. Competition was not evident. Fish were sometimes observed to converge onto a pellet but completely miss it. 4.4.2. Continuous feeding Continuous and nearly continuous feed delivery produced a regular swimming formation (sites A, B, C and H) (Table 4.3, Figure 4.2). The regular swimming formation was obvious near the surface at the beginning of feeding in all cages fed continuously or nearly continuously. The entire formation was also visible from below water at sites A and C, while only occasionally visible from below the water in the higher stocking density cages at sites B and H. The 'S ' -shaped foraging pattern was gently manoeuvred as fish swam with tails constantly beating and without coasting. The swerve in the head was barely evident. Pellets were consumed in an orderly fashioned as the fish circulated around the cage. Competitive or aggressive behaviours were not observed (Table 4.3). 95 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CA GES Table 4. 3. Feeding formation, occurrence of pellet wastage, and relative numbers of competitive and aggressive behaviours during feeding events at the various farming sites. Site Discharge technique Feeding formation Pellet wastage Competitive behaviours Aggressive behaviours A' Continuous, mechanical blower Regular swimming formation breaking down to no apparent structure 1-2 (sporadic) Absent, except at the beginning Absent A 2 Continuous, mechanical blower Disorganised pattern, no apparent structure 1-2 (sporadic) Not perceivable Absent B 3 Continuous mechanical blower Regular swimming formation breaking down to no apparent structure 1-2 (sporadic) Absent, except at the beginning Absent C Continuous mechanical blower Regular swimming formation breaking down to no apparent structure 1-2 (sporadic) Absent, except at the beginning Absent D Batch (large portions) manual hand feeding Spiral vortex breaking down to no apparent structure Substantial after the third burst of feed, nil at endpoint Ubiquitous to numerous on the surface Absent E Batch (small portions) rotary mechanical spreader Spiral vortex breaking down to no apparent structure Absent except at the feeding endpoint Ubiquitous to numerous on the surface Numerous on the surface Batch (small portions) manual hand feeding Spiral vortex breaking down to no apparent structure Absent except at the feeding endpoint Ubiquitous to numerous on the surface Numerous on the surface F 3 Batch (small portions) rotary mechanical spreader Not recorded Numerous during the last half hour Not measured Not measured F 3 Batch (small portions) manual hand feeding Not recorded Numerous during the last half hour Not measured Not measured H l . a Batch (variable portion) programmable blower Either spiral vortex on the surface or following pellets downward Substantial throughout Ubiquitous to numerous oh the surface Absent H2,b Nearly continuous batch (variable portions) programmable blower Regular swimming formation breaking down to no apparent structure Numerous for the last 5 min during the twilight feeding Absent, except at the beginning Absent ' High visibility conditions; 2 Low visibility conditions 3 High stocking density did not permit consistent subsurface viewing of the swimming aggregation. a Stocking density of 4.4 kg m"3 b Stocking density of 7.8 kg m"3 96 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Figure 4. 2 Ring-like structure characterised by organised foraging in continuously fed fish. During feeding, the feeding formation would break down as fewer and fewer fish participated in it. Foraging and non-foraging fish were easily distinguishable by the way they swam. Non-foraging fish coasted after thrusting with their tails, while actively foraging fish did not appear to coast. At the low stocking density Atlantic salmon Site A , sporadic and few incidents of both foraging behaviour and pellet intake occurred on the surface and at the camera level at the feeding endpoint (Table 4.4). At the higher stocking density Atlantic salmon Site B, the level of surface feeding activities seldom corresponded with subsurface feeding activities. On average, in 7% of the feeding events, fish near the surface and at the camera level were not feeding at the feeding endpoint, and in 78% of the feeding events, fish at the camera level were the last to finish eating (see Chapter 3 for details). 97 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Surface feeding activities often declined to non-perceivable levels and no fish were visible from the surface while subsurface activities would continue to indicate numerous incidents of feed capture and the fish displayed slowly executed "S"-shaped foraging activity. On average, in 80% of the feeding events at the chinook salmon Site C, levels of surface feeding activities such as splashing, mouthing and pellet capture continued to be obvious when pellet loss became unavoidable. During 22 out of 180 feeding events there was agreement between the level of subsurface and surface feeding activities at the feeding endpoint. At Site H (programmable, nearly continuous modified ration), agreement between the level of surface and subsurface feeding activities at feeding endpoints occurred in two out of two feeding events for fish stocked at 7.8 kg m"3 (Table 4.4). The level of foraging activity near the camera level at Site A was always similar to the level at the surface (100% uniform feeding). At Site B, fish spent variable amounts of the total feeding time near the surface (57+27% to 76±35% morning feed; 22+25% to 69±39% evening feed) and the camera level (24+19% to 40+21% morning feed; 28±25% to 48±19% evening feed). The percentage of time spent both on the surface and near the camera (uniform feeding time) by fish was low (2±6% to 14±19% morning; 1±4% to 23±25% evening feed). In both morning and evening feeding events, fish only started to feed near the camera level after at least 51% of the feeding event had elapsed (Table 4.5). In contrast, fish at Site C spent more time (> 95%>) feeding both near the surface and camera levels resulting in a larger average uniform feeding time but they took comparable time to start feeding near the camera level. At Site H, an average uniform feeding time of 65% was achieved over 2 d by feeding nearly continuously (Table 4.4). 98 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Table 4. 4. Duration o f feeding times (expressed as a percentage of total feeding time per meal) spent by fish at different locations (depths) when water visibility was greater than 3 m. Site Discharge technique Feeding endpoint method # of feeding endpoints when subsurface and surface feeding activities were similar Time elapsed before feeding at camera level, mean±SD, % Uniform feeding time, mean±SD, % A Continuous mechanical blower Video camera method 84 out of 84 0 100 B Continuous mechanical blower Video camera method 14 out of 246 63±19 10±15 C Continuous mechanical blower Video camera method 22 out 180 53±29 43±29 D Batch (large portions) manual hand feeding Ration method 1 out of 2 0 50±35 E Batch (small portions), rotary and manual hand feeding Surface activities method 10 out of 10 100 0 F Batch (small portions) rotary and manual hand feeding Surface activities method 1 out of 2 Not measured Not measured G Batch (medium portions) manual hand feeding Pellet loss method, device malfunction n/a n/a n/a H 1 Nearly continuous batch (variable portions) programmable blower Modified ration method 2 out of 2 35±0 65±0 1 For stocking density o f 7.5 kg m' SD: Standard deviation 99 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Table 4. 5. Duration of times that fish in CMCs spent feeding at different levels (depths) in the water column, expressed as a percentage of total feeding time per meal or feeding event when water visibility was greater than 3 m for long-term sites (means + SD). Cage# Surface 1 feeding, % Bottom feeding, % Uniform feeding, % Time to start bottom feed, % Site A 1-3 SiteB Morning feed: 1 2 3 Evening feed: 1 2 3 100 57 ± 2 7 67 ± 3 5 66 ± 16 22 ± 2 5 69 ± 3 9 59 ± 3 5 100 24 ± 19 40 ±21 32 ± 17 46 ± 15 28 ± 2 5 48 ± 19 100 2 ± 6 14 ± 19 7 ± 13 1 ± 4 13 ± 2 6 23 ± 2 5 75 ± 19 59 ±21 68 ± 17 53 ± 15 71 ± 2 5 51 ± 19 SiteC 1 2 3 95 ± 11 96 ± 16 96 ± 12 45 + 31 53 ± 2 9 40 ± 2 6 42 ± 3 0 49 ± 3 0 38 ± 2 7 54 ±31 46 ± 2 9 59 ± 2 6 Surface feeding refers to the duration of time that fish spend feeding near the surface (swarm at top). 2 Bottom feeding refers to the duration of time that fish spend feeding near the cage bottom (or camera level). 3 Uniform feeding refers to the duration of time when fish could be observed to feed both near the surface (from data recording sheets) and near the cage bottom (from camera observation). 4 Time to start bottom feed refers to the amount of time that elapsed from the start of feeding before fish begin to move down toward the camera level. Data were obtained by playing back tapes of recorded feeding events and matching fish movements (migration) to the corresponding times (clock) displayed on the monitor screen. SD: standard deviation. 100 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Pellet loss in all the cages that were fed continuously with the video camera method could be controlled until the feeding endpoint, except, as previously reported (see Section 4.4.1), pellet loss occurred near the end of a feeding event at twilight (sunset) during one out of the two feeding events at Site H. 4.4.3. Batch feeding In batch feeding, fish would spiral upward to feed, and, thereafter, feed in a disorganised and competitive manner (Fig. 4.3). At sites A , C, D, E and H, the spiral vortex could be seen from underwater (i.e. camera view). Competitive behaviours on the surface decreased from ubiquitous at the beginning of a feeding event to numerous toward the end of a feeding event. While pellets were visible, 'S'-shaped foraging patterns on the surface (Figure 4.1) were ubiquitous at the beginning of a feeding event, and numerous throughout the feeding event. The amplitude of a swerve in the 'S'-shaped pattern in batch feeding was greater than in continuous feeding. Prolonged surface feeding frenzies, comprising of splashing, mouthing and competitive behaviours, were evident at the beginning of a feeding event. While feeding 2.5 kg fish in 5-9 kg bursts every 7-13 min at Site H, fish stayed on the surface when the feeder was active, then followed pellets downward into the water column. At all sites, it was not uncommon for fish subjected to batch feeding to change from a regular swimming formation when pellets were unavailable to a spiral vortex when pellets were available. Structural changes in swimming formations depended on the size of the broadcast area and output rate. 101 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES Pellets were lost during the last half hour of the feeding events at Site F, when numerous pellets were evident after these large salmon (9 kg) were fed five bursts (750 g burst" ') of a rotary feeder or five scoops (1 kg scoop"1) by hand (Table 4.3). While hand feeding to ration at Site D, substantial pellet loss occurred when three scoops (5 kg scoop"1) were fed in quick succession. In general, pellets lost at Site D sank from the centre of the cage (Fig. 4.4). No pellet loss was evident until the feeding endpoint at Site E while feeding very small scoops or bursts of feed. Similar to the dissolution of a regular swimming aggregation, the structure of the spiral vortex broke down as fewer and fewer fish remained within the structure during the feeding event. For example, when Atlantic salmon (1.8 kg average weight, site E) were batch fed very small portions (either mechanically or manually) based on the level of surface activities over the daylight period, the feeding event started with an obvious spiral vortex. Soon after, fish started to forage very slowly and in a circular fashion around the cage. Numerous incidents of both types of competitive feeding behaviours and aggressive behaviour on the surface occurred while fish competed for the few pellets offered (Table 4.3). Such foraging behaviour continued throughout the feeding event. No near camera level feeding occurred, as pellets did not reach that location. Pellet loss at the cage bottom was evident only at satiation as perceived using surface activities. At Site D (chinook salmon), fish at the camera level immediately started to forage when fed using 5 kg scoops in two feeding events. Uniform feeding time averaged 50% of the total feeding time. During one of the two feeding events, most of the feeding occurred near the cage bottom, while during the other event, uniform feeding occurred throughout the event (Table 103 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CA GES 4.4). At the end of one of the two feeding events, feeding was not perceivable on the surface, while at the camera level, fish were still foraging. 4.4.4. Ration At site D, fish did not eat during one of the three feeding events, which resulted in heavy pellet wastage. When fish ate at that site, subsurface feeding activities indicated that fish were not satiated at the feeding endpoint (Table 4.4). During the final day of video-recording, the farm employee continued to feed after the ration was allocated to determine when foraging at the camera level would have ceased and pellet loss would have become unavoidable. As a result, 25% more feed was allocated to the fish during that feeding event. Modified ration method gave mixed results at site H. Pellet wastage occurred throughout two events, in the last 5 min in another event and was absent in one out of the four feeding events. 4.5 Discussion and recommendations Observations from this study indicated that continuous and nearly continuous feeding produced an organised and directional swimming formation while feeding small amounts in bursts in a cage produced a spiral vortex and disorganised feeding. The latter case appeared to reduce the feeding opportunity of fish because fish were forced to compete and forage heavily for the limited supply of feed. Marked changes in swimming direction and speed due to foraging and competition were evident. Long-term usage of this form of feeding whether by hand or machine would, therefore, appear to be an energy inefficient method to feed fish. Boisclair and Tang (1993) found from empirically derived equations that routine swimming, characterised by 104 Chapter 4. GROUP FISH FEEDING PA TTERNS IN CA GES marked changes in speed and direction, were 6.4 to 14 times more energy expensive than swimming patterns executed at constant speed. Continuously broadcasting feed produced an organised swimming formation characterised by most fish feeding in an apparently non-competitive fashion while circulating around the cage. This is similar to Kokanee salmon (Oncorhynchus nerka), which are mainly schooling fish and exhibit little territorial behaviour (Hutchinson and Iwata, 1997) when presented with food source that is abundant yet patchy. In the wild, salmon exhibit three types of foraging behaviours or strategies namely, territorial strategy, floater (drifter) strategy and schooling strategy depending on the nature of food sources. When forced to wait for food to drift by, fish become drift feeders and are very territorial, taking food and quickly returning to its territory. Continuous feeding may be more energy efficient for fish than batch feeding. Competitive behaviours may be reduced by other mechanisms as well. For example, Kadri et al. (1996) surmised after studying the feeding interactions of 19 fish, that food should be presented so that it is unpredictable in both time and space to reduce the likelihood of competition to achieve maximum food intake. Pellet loss at the batch-fed Site E was insignificant when feeding (using surface activities) very small amounts of feed over a long time period, although, aggressive and competitive behaviours were common on the surface. Kadri et al. (1991) monitored the daily feeding rhythms of Atlantic salmon and implied that a feeding regime that dispenses food at a low rate throughout the day may not provide enough food when fish were hungry, yet cause substantial wastage the rest of the time. In this study, wastage was controlled by the workers at Site E who carefully dispensed feed in such a way as to guarantee its consumption at the cage surface, thus forcing the fish to rise to the surface to eat. At stocking densities higher than 105 Chapter 4. GRO UP FISH FEEDING PA TTERNS IN CA GES those tested at Site E, fish may not all be able to migrate upward to feed and workers may not be able to see effectively from the surface i f fish are consuming all the pellets. The latter case could have been the cause of the pellet wastage exhibited during the latter stages of the feeding event at Site F, where fish were stocked at three times the level of Site E yet fed in a similar way. In some batch feeding events, pellets sank down to the cage bottom from the centre of the cage. Explanations for this wastage could be related to excessive output rate, fish foraging behaviour and the point over which pellets were distributed (i.e. centre of cage). It is conceivable that foraging fish swimming in the 'S ' - shaped pattern (requiring a large turning radius) will miss capturing the same pellet to avoid a collision. Smith et al. (1993) observed that in a net cage, pellets appeared more likely to be missed when two or more fish attacked the same pellet. Fish feeding pattern can be used to assess fish appetite level, since feeding formations evident at the beginning of a feeding event were always observed to deteriorate. For example, if a brief but intensive burst of feed within a small area does not promote an obvious spiral vortex, then fish can be assumed to be nearly satiated. Short-term usage of intensive batch feeding for this purpose should neither cause appreciable harm to fish nor pellet loss. Feeding patterns can also be used to judge the apparent level of feeding 'readiness', because feeding inactivity at the beginning of a feeding event was a good indicator that fish would subsequently not eat. Other researchers mentioned that at the beginning of a feeding event, the density of fish near the surface increased (Juell et al., 1994 a and b; Bjordal et al., 1993). Feeding fish without an initial response of mouthing, pellet consumption or formation of feeding pattern could increase the incidence of pellet loss through the water column. 106 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES No feeding method consistently fed fish uniformly throughout the cage. Only at Site A , where stocking density was the lowest of all the sites (-1.5 kg m") was uniform feeding apparent, with subsurface and surface foraging activities obvious from the beginning to the end of feeding events for over 2 months. The explanation for uniformity here and not elsewhere, could be related to the low stocking density or to the fact that the fish were hungrier due to their alternate day feeding regime. At most other sites, subsurface foraging and pellet consumption became obvious only after the feeding event was well on its way, and once established, generally continued until the end of the feeding event. The lack of uniform feeding could be due to a problem with pellet detectability near the bottom of the cage, competition, preference to eat elsewhere or pursue other activity, and/or insufficient feed output rate. At Site B, high surface water temperatures (13-18°C) might have prevented fish from spending more time feeding near the surface (57-66%) as compared to site C where surface feeding times were greater than 95% when surface water temperatures were lower (Tables 4.1 and Table 4.5). In this study, pellet detectability was influenced by several factors: stocking density or distribution of fish within a cage (i.e. too dense near the surface and light could be sufficiently blocked), air light intensity as well as water clarity. Another study confirmed pellet wastage is uncontrollable while feeding fish at twilight (Blyth et al., 1993). Light intensity and turbidity in water inevitably combine as factors to affect the detectability of food pellets by fish. Turbidity affects foraging success of particulate feeding juvenile and adult fish by reducing reaction distance (Crowl, 1989; Vinyard and O'Brien, 1976). With formulated pellet food particles, the visual characteristics of size, shape, colour and movement are important (as are organoleptic ones) in improving pellet capture and ingestion by Atlantic salmon parr, (Stradmeyer, 1989; Stradmeyer et al., 1988). In the present 107 Chapter 4. GRO UP FISH FEEDING PA TTERNS IN CA GES experiments, when visibility fell below 3 m, fish could not easily capture, although vigorous splashing and mouthing occurred on the surface. The fish feeder's response (based on the number of incidents of splashing and mouthing activities) would have been to feed more quickly to satiate the fish, resulting in greater pellet loss. Feeding more slowly, however, would have increased the capture success rate. It may be possible to adjust feeding rates and delivery method to promote uniform feeding. With a pellet detector, uniform feeding may be achieved without pellet loss. Juell et al. (1994a) found no effect of feeding intensity on either feed conversion or growth while continuously feeding fish to pellet loss. The present study indicates that success of foraging near the bottom of the cage is just as important as pellet loss for judging satiation and determining feeding intensity or output rate. Feed output rates should match the foraging success rate and not be fixed. Surface light blockage due to high fish density may be prevented by feeding at the highest feeding rates possible for efficient foraging, because low output/feeding intensity rates (1.9 pellets fish"1 min"1) were found to promote higher fish densities near the surface than higher rates (Juell et al., 1994b). To avoid pellet loss and high energy swimming patterns, fish can be fed continuously by scattering the pellets widely and unpredictably both in time and space. This criterion may be difficult to meet for cages larger than 15 m by 15 m in area, since many types of feeders have a limited broadcast spread. It is conceivable that feeder spread patterns could be adopted to simulate swimming vortices so that fish would swim in an organised pattern while feeding. Feeding to a predetermined ration or to satiation as judged by surface feeding activities cannot be recommended. Fish achieved less growth when fed to a predetermined daily ration as compared to being fed to obvious pellet loss (Juell et al., 1993). Inaccuracy with the ration 108 Chapter 4. GROUP FISH FEEDING PATTERNS IN CAGES method of feeding has been indirectly attributed to poor predictive models, inaccurate estimation of fish numbers and variable appetite, and the cause of over or underfeeding (Kimura et al., 1993). This study for the first time shows direct evidence of underfeeding and pellet wastage in the ration method. Feeding fish based on surface feeding activities did not guarantee that bottom foraging fish had an opportunity to feed to satiation nor was pellet wastage prevented in all cases. Feed discharge rate should match actual feed capture rates and pellet detection ability. To achieve this, output of feeders should be adjustable and capable of exceeding consumption rate. Under these operating conditions, an operator will be able to adjust the feeding rate throughout a feeding session to provide feed without wastage. To promote uniform feeding, more research must be conducted to determine the factors affecting its occurrence in the industry. Factors that have not been considered in this study but could be considered in future research include cage size, fish strain, temperature effects and hatchery rearing conditions as well as prior conditioning or entrainment effects. 4.6 Conclusion Subsurface and surface feeding activities seldom coincided. Pellet wastage was associated with the apparent inability of fish to either detect or to successfully capture pellets, or was due to too high output rate. Pellet wastage and underfeeding were associated with the predetermined ration method of feeding. 109 Chapter 4. GRO UP FISH FEEDING PA TTERNS IN CA GES Two distinct types of feeding formations were observed. One was associated with continuous feeding, which consisted of fish circulating around the cage in an organised way. The other form of feeding formation was the spiral vortex, which was associated with batch feeding. Both types of structures tended to break down over the feeding event, and this breakdown may be exploited toward finding another way to judge satiation. The foraging pattern of individual fish was an 'S'-shaped searching manoeuvre. The amplitude of a swerve was greater in batch feeding than in continuous feeding. The difference in foraging patterns between continuous and batch feeding may lead to a difference in energy requirements by fish. More research is needed to further our understanding of the impact of non-uniform feeding and the effect of feeding patterns on long-term growth rate and feeding efficiency. Uniform feeding may be achieved provided fish can detect pellets throughout the water column and are fed at a rate sufficient for pellets to reach the fish near the bottom of the cage. 110 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Chapter 5 L A B O R A T O R Y STUDIES O N T H E D E T E C T A B I L I T Y OF P E L L E T S U N D E R V A R I O U S E N V I R O N M E N T A L CONDITIONS 5.1 Summary This research was conducted to find a pellet type that was more detectable than the conventional brown pellet under conditions of low light and/or high turbidity. The contrast of 18 differently coated pellets varied with different coloured backgrounds. The detectability by fish of the same 18 differently coated pellets were compared with conventional pellets by offering them to fish (rainbow trout, Oncorhynchus mykiss) in 160 L aquariums at a low ambient air light level of 0.007//mol photon s"1 m"2. Five of the 18 coated pellet types were selected for subsequent detectability studies under similar conditions of low light (0.007//mol 1 2 photon s" m" ) and two turbidity levels. In a 160,000 L seawater tank, the visibility of coated pellets to the human eye were measured at two depths, one ambient air light level (0.03±0.01 //moi photon s"1 m"2) and two turbidity levels (9.35±0.07 N T U and 21.8±0.05 NTU). Dark-coloured pellets were more detectable than bright coloured pellets under bright ambient conditions typical of near surface feeding and highly turbid water. Reflective (silver) and white pellets were more detectable than darker coloured pellets for darker lighting conditions (< 1 2 0.007 //moi photon s" m") typical of low feeding depths, highly stocked cages and cloudy days or shaded environments. Mixing bright and dark-coloured pellets may be beneficial under environmental conditions that have variable lighting and turbidity conditions. Unlike reflective 111 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS or silvery pellets, coloured (chromatic) pellets may not be suitable for highly variable environments because the latter undergo colour shifts with changing depths (lighting) and turbidity conditions. 5.2 Introduction Observations from the descriptive research discussed in Chapter 4 suggest that pellet detectability is influenced by several factors including stocking density or distribution of fish within a cage, air light intensity as well as water clarity. The influence of light is confirmed in a study by Blyth et al. (1993) where pellet wastage was shown to be uncontrollable while feeding fish at twilight. Light intensity and turbidity in water inevitably combine as factors to affect the detectability of food pellets by fish. Turbidity affects foraging success of particulate feeding juvenile and adult fishes by reducing reaction distance (Vinyard and O'Brien, 1976; Crowl, 1989). Brightness contrast is most often the determining factor in underwater visibility of objects (Lythgoe, 1975). Contrast detection is an especially acute problem for fish because contrast is severely degraded underwater (Lythgoe, 1979). An object must usually reflect a perceivably greater or lesser amount of light than the background for it to be visible underwater (Douglas and Hawryshyn, 1990). The same authors specify that the difference between an object and the background radiance is further decreased as it moves away from the fish because both light absorption by water and light scattering from suspended material can attenuate the image forming light. 112 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS An animal's sensitivity to contrast is measured as the minimum difference in radiance between an object (Ro) and its background (R b) that the eye can detect as a function of the background radiance (Rb) as represented by the following equation: c = ( ^ z A ) ( 5 > 1 ) where C is the normalised contrast level. This ratio, (R<, - R b ) / Rb, is termed the Weber fraction or contrast threshold and can range from 0.003 to 0.05 depending on the species of teleost and the behavioural method of determination (Douglas and Hawryshyn, 1990). Feeding performance of fish is adversely affected by poor visibility in water. In the case of larval bluegill, Lepomis macrochinus, feeding on crustacean zooplankton, turbidity from suspended sediments reduced consumption of zooplankton when light intensity in part of the enclosure fell below a threshold estimated at <450 lux (Miner and Stein, 1993). Under experimental conditions where illumination is artificially reduced, Blaxter (1970) found that the threshold for feeding averaged about 10"1 lux. The foraging rate of larval herring, Clupea harengus harengus, also declined with increasing turbidity because of reduced reaction distances (Johnston and Wildish, 1982). With adult fish, Barrett et al. (1992) showed that foraging by adult rainbow trout, Oncorhynchus mykiss, decreased with increasing turbidity. Commercially available salmonid feed pellets are cylindrical in shape and are sized based on its diameter (e.g. a 3.5 mm pellet has a diameter of 3.5 mm and a length of approximately 1.5 times the diameter). Salmonid pellets have been developed from studies on some of the physical 113 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS characteristics of salmonid prey such as particle size (Wankowski, 1979; Wankowski and Thorpe, 1979; Irvine and Northcote, 1983) and movement (Rimmer and Power, 1978; Irvine and Northcote, 1983) and on test pellets such as taste and smell (Sutterlin and Sutterlin, 1970; Adron and Mackie, 1978; Stradmeyer, 1992; Oikawa and March, 1997), colour (Ginetz and Larkin, 1973; Clarke and Sutterlin, 1985; Shelton, 1987; Browman and Marcotte, 1987; Stradmeyer and Thorpe, 1987), texture (Lemm and Hendrix, 1981; Lemm, 1983) and pellet dimensions (Stradmeyer et al, 1988; Smith et ah, 1995). For fish predators hunting by sight, the important visual characteristics of the prey are size, contrast with the background and movement. (Wootten, 1990). Ginetz and Larkin (1973) state that for fish perception, contrast may be more important than the colour of the feed. Pellet detectability by both human and fish vision is affected by contrast, which in turn is affected by many factors that cannot be controlled by the fish farmer. Knowledge on detectability of food items under different environmental and stocking conditions is limited. Little research has been specifically focused on making salmonid feed pellet more detectable by increasing contrast of the pellet surface. Contrast can be increased by increasing object radiance over background radiance, and in pellets, this can be achieved by making them bright white or more reflective regardless of orientation of the pellets to the incident light. The visual properties of pellets in sea cages had been studied using cameras in the imaging laboratory of the Department of Bio-Resource Engineering at the University of British Columbia (Foster et al, 1995). Commercial fish feed pellets are normally brown in colour (derived from natural colour of ingredients) and appear black from below and side, and white when viewed from above. The reflectivity of commercial pellets was measured to be approximately 6% (Foster et al, 1995). In fish farms where fish are fed using surface activity and the fate of pellets in the water column to 114 Chapter 5. LABORA TORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS judge feeding rate and feeding endpoint, pellet visibility from the surface (vertical plane) is vital. Pellets that are more visible allow the farmer more time during which to observe the fate of pellets (i.e. eaten or ignored) before the pellets sink into the water column. When viewed from the side (horizontal plane), the contrast of brown pellets in water is good near the surface where ambient light level is high (bright background) but very poor at 8 m or at the bottom of the feeding school because of greatly reduced light levels (dark background). The deteriorating contrast can be explained by the reduction of light with depth in the water column. This light reduction is expressed in terms of the vertical extinction coefficient k, also called attenuation coefficient as shown in the following equation (Parsons et ai, 1984): Id=I0e-kd (5.2) where I0 is the incoming light intensity, Id the light intensity travelling a distance (depth) of d. The coefficient k can be estimated from Secchi disc reading (D s, in meter) using an empirical relation as follows: 1.7 k-— (5.3) A where k is the estimated extinction coefficient and 1.7 is a constant which, according to Idso and Gilbert (1974), gives fairly close extinction coefficients when compared to optically measured coefficients in a wide range of water visibility (turbid to clear ocean water) covering D s between 1.9 and 35 m (Parsons et. al., 1984). 115 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5.1 below shows the reduction (%) in light intensity due to light attenuation at different visibility levels (from Secchi disc readings, D s) at 3, 5 and 8 m depths using Equations 5.2 and 5.3. Table 5.1. Reduction in light intensity due to light attenuation at different water visibility and depths in the sea. Water visibility (Ds), m Percent reduction At 3 m depth in light intensity At 5 m depth At 8 m depth 3.0 82% 94% 99% 5.5 60% 79% 92 % 7.0 52% 71 % 86% 12.0 35% 51 % 68% D s: Secchi disc reading The reduction (in light intensity) may actually be greater when fish are crowded at the surface, blocking out more light. Increasing the reflective index of a pellet should theoretically, improve its detectability underwater. The angle of light incident upon the object is very important. For example, a sudden bright flash from the reflective parts of a predator's body may act as an initial warning of a predator's presence before its general outline becomes perceivable (Loew and McFarland, 1990). Fish are silvery so they visually blend into the water when they are vertically orientated (upright) to reduce their own visibility to predators (Muntz, 1990) on a horizontal plane because the silvery surface of the prey reflects light of equal intensity to the surrounding 'space' light. However, when the silvery surface is not upright, the intensity of reflected light from the prey is greatly reduced causing unequal light intensities between reflected and 'space' light, thereby, increasing contrast and detectability of the prey. In this connection, Ward (1919) suggested that predators preferentially select unfit silvery fish when they cannot maintain a 116 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS vertical position and therefore, become more conspicuous. In fact, when viewed from above and from the side, a silver surface reflects all light it receives thus making it very bright when not orientated vertically. A further suggestion that pellet detectability might be improved by increasing shininess or brightness is the observation of Dendrinos et al. (1984) that feeding efficiency of larval sole, Solea solea was improved when the artemia provided were brightly stained. In the present study, it was hypothesised that contrast-enhanced (brighter) pellets were more detectable than conventional brown pellets under a dark (low light intensity) environment for both humans and fish. The specific objective of this stage of the research was to test the effect of contrast levels of pellets on pellet detectability by fish and humans under different environmental conditions. 5.3 Materials and Methods The visual characteristics of pellets used in all tests in this stage of the research were achieved by applying different coating materials to conventional brown pellets. Different coating methods were also tried and the most effective method was then used for subsequent coatings. The coated pellets were subjected to the following tests to determine the water stability of the coatings and the extent to which contrasts was enhanced by the coatings: a. Measurement of contrast levels of pellets (in air) against different coloured backgrounds using the spot meter and luminance analysis software. b. Coat stability tests. c. Pellet palatability d. Pellet detectability by humans in seawater 117 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS e. Pellet detectability in freshwater by fish (Parts 1 and 2) Details on procedures employed at each of these tests follow: 5.3.1 Making contrast-enhanced feed pellets Initially, both non-fluorescent and fluorescent colours, and reflective materials were systematically considered for testing (Table 5.2). These included materials in various forms such as powder (e.g. cosmetic products), paste (natural guanine), dye (food grades) and paints (liquid and spray paints). Non-fluorescent colours (paints and dyes) coated on pellets included white, grey, red, orange, yellow, green, blue, violet, brown and eggplant while fluorescent colours (paints only) used included red, orange, yellow, green and blue. Reflective materials included natural guanine, titanium-dioxide-coated mica (TCM, MagnaPearl Corp.) and cosmetic products (e.g. nail polish, aluminium and gold-coloured dust, and sparkled gels). A wide range of colours (including colours described in the literature as having high contrast against specific background conditions) were initially tested in order to provide as wide a range as possible in terms of realising any potential benefits that each colour might have. 118 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5. 2. Coloured and reflective materials used in coating. Check (•/) signs indicate colours tested. White Grey Red Orange Yellow Green Blue Brown Eggplant Reflective3 (violet) Paints Liquid • • • • • • • • Spray • • • • • • • • Fluorescent • • • • • Powder 1 • • • • Dyes • • • • • • • Paste2 / 1 Various cosmetic products including nail polish, aluminium, titanium-dioxide coated mica (TCM) and gold coloured dust. 2 Natural guanine. 3 Reflective attributes of coated pellets refer to a silvery or sparkly appearance. Natural guanine was extracted from salmon scales that had been removed from fish at a processing plant. The scales were initially rinsed with water to get rid of skin material and mucus before placing them in 1 L conical flasks in a mixture of 1:2 parts scales to water and mechanically stirred with a magnetic stirrer for approximately 18 h. The solution was filtered using a kitchen sieve and the supernatant then centrifuged at a laboratory in the Department of Animal Science, to concentrate the guanine. Natural guanine concentrates processed from this method were mixed with distilled water in a ratio of 0.25:1 water to guanine to form a paste and coated on pellets by rolling and shaking in a small 250 ml glass beaker for approximately 10 s or until the pellet is completely coated with guanine. Natural guanine was apparently easily extracted from salmon scales and produced long lasting, intact coats on pellets. Guanine is available commercially but all products are packaged in non-edible (and classified under 119 Chapters. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS "incidental ingestion" e.g. as an ingredient in lipsticks) medium to preserve its structural integrity and, therefore, not suitable as fish feed. It is also very expensive. Other coatings were applied using several methods including rolling and shaking pellets in a small container, spraying directly on pellets, open- pan coating and fluid bed atomisation. The first two methods were performed at a laboratory in the Department of Chemical and Bio-Resource Engineering while the last two methods were tested at a laboratory in the Department of Pharmaceutical Sciences. Spraying pellets with coloured paint sprays was straightforward but produced coats of varying quality because of micro-bubbles formed as air in the miniature cracks of a pellet is pushed out and replaced by the paint. Most powdered and paste materials were easily coated on pellets by shaking and rolling (i.e. "stirring") them in a sprinkling of the powder or a small amount of the paste or paint in a 250 ml beaker. Pellets were coated in batches of 100 g (with powder and dyes) and 50 g (with paints and paste). The open pan coating method utilised a 5 L glass bowl attached to a rotating shaft at a 45° angle. Pellets were placed in the bowl and coating material added in small quantities as pellets were "tossed and turned" by the rotating action. A hair dryer was used to dry pellets between coatings to prevent pellets from sticking together. This manual pan coating method produced good intact coats but proved laborious on a small scale and was abandoned. Large commercially available fully automated (drying and coating) pans might be usable for producing large quantities of pellets. The most complex method tried was the fluid bed atomiser commonly used in the pharmaceutical industry for preparing tablets. The fluid bed atomiser method produced the most intact coats (double coats) but the process was slow (250 g of 3.5 mm pellets per 20-30 min), expensive (requiring air compressor, peristaltic pump, nitrogen gas, heating to 40-60°C) and the powder must be fine enough to pass through the 120 Chapter 5. LABORA TORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS atomising nozzle. It was only attempted for three or four types of powders (e.g. cosmetic aluminium, yellow dye and titanium white) as an exercise to assess the merits of available procedures. A l l pellets coated were 6.5 mm in diameter, appropriate for experimental fish in the laboratory. The application rate of each coating material (regardless of coating method) was determined by trial and error until an effective intact coat was obtained. The application rate (AR, %) was calculated as: AR =^-xl00 (5.4) where A c is the amount of coating material used (g) and A p is the amount of pellets coated (g) Application rates ranged from 4.5-5% for paints and paste, approximately 0.5% for silvery powder (MTD) and 2-2.5% for powder used in the atomising method. Coating pellets by rolling and shaking coating materials in a small container was the most straightforward and easiest method to use. Paints and paste produced the most intact coats by shaking. Most powders and dyes coated well but application rates were difficult to standardise because too much dye caused smearing while too little produced incomplete coats and optimal rates varied greatly with different coloured dyes. The contrast of each type of coated pellet was measured and its coat tested for stability or integrity in moving (stirred) water according to the following methods. Subsequently, the pellets' palatability or acceptability to fish was also examined. 121 Chapters. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS 5.3.1.1 Contrast levels of coated pellets Six pellets at a time were placed close to each other on different coloured paper as background. Lighting was supplied by a 1000 W constant spectrum flood/spot lamp (IANIRO Lighting Equipment Ltd.) commonly used for photo shoots, positioned 1.8 m directly above the pellet formation, with its beam pointed towards the ceiling to diffuse the light rays. The luminance level of each of 18 pellet types (Appendix 1) was measured using a spot meter (Minolta 1° Spotmeter F; Lens and Shutter, B.C) and its contrast level was calculated using the Weber fraction (Equation 5.1). The spot meter was used to obtain E V integers (electron volt) of a small circular area equivalent to 1° angle subtended at the viewing piece. The EV integers were then converted to luminance readings (cd m"2) using conversion tables provided by the manufacturer. Contrast levels of each pellet against the background were calculated as dimensionless values as in Table 5.3. 122 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5 .3 . Contrast levels (in descending order of magnitude) of 18 different pellet types against a black background under high light (12.75+0.10 //moi photon s"1 m" 2; diffused light from a 1000 W halogen lamp). Pellet type Contrast values Posi t ive / 1 negative Fluorescent yel low 3.46 + White 3.15 + Fluorescent orange 2.63 + Fluorescent red 2.14 + Yel low 1.73 + Silver 1.39 + Orange 0.57 + Blue 0.40 -Red 0.37 + Natural guanine 0.37 + Uncoated 0.26 -Sparkles 0.19 + Eggplant 0.16 -Green 0.16 -Fluorescent blue 0.16 -Grey 0.04 + Fluorescent green 0.03 -Brown 0.03 -1 Positive/negative values indicate pellets of higher/lower brightness than background, respectively. Photographs (Kodak Gold Plus 100) of the different coated pellets were taken in a darkroom/studio, using a standard 35 mm camera under the same source from a 1000 W constant spectrum flood/spot lamp (IANIRO Lighting Equipment Ltd.) commonly used for photo shoots. Both the spot meter readings and the photographs were taken with pellets placed on a non-reflective white, ocean-blue and dull black coloured paper as the background. These background colours were chosen to resemble different possible field conditions for pellets. Colour prints were scanned and analysed for contrast of selected pellets using Photo-Deluxe (Microsoft Corp.) (Table 5.4). 123 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5. 4. Contrast values of selected pellet types against different background colours under bright light (1000W quartz lamp) in air. Pellet type Pellet luminance level1 Pellet contrast against different background colours 2: Blue (118.45') Black (33.22') White (191.22') Silver 121.50 0.03 2.66 0.37 Brown 12.67 0.89 0.62 0.93 Guanine 122.94 0.04 2.70 0.36 White 200.41 0.69 5.03 0.04 1 Luminance levels range from 0 (darkest) to 255 (brightest) on a grey scale. 2 Contrast values calculated using Equation 5.1. 5.3.1.2 Coat stability/integrity in moving water Two pellets from each type of coating material and coating method from Table 5.2 were tested for stability of their coats in water by placing a single pellet from each type of coating in a beaker (2 L) of tap water which had been pre-stirred with a glass rod. The stirred water simulates the sinking motion of a pellet in the cage. The time taken for the pellet to lose approximately half of its coating material (i.e. revealing approximately 50% of pellet brown colour) was recorded. Pellets with coatings that did not last longer than 2 min were eliminated from subsequent tests. The time frame of 2 min was based on the length of time that a pellet of similar dimension would take to sink past the feeding volume in a sea cage. In stirred water, coats from paints and paste stayed intact for longer than 2 min. A few cosmetic materials (e.g. aluminium powder) lost 50% of coating materials between 52-66 s while pellets coated with dyes started to form a plume between 1-2 min. In terms of coating method, most coats produced from the shaking and rolling method lasted longer than 2 min while those coated using the fluid bed atomiser method lasted just around 2 min. Sprayed pellets had no difficulty keeping coats intact. Based on this test and previous considerations 124 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS (e.g. ease of use, costs, and coating speed), the shake and roll method was used to coat all pellet types (except white, which was coated using the atomising method) for subsequent tests. 5.3.1.3 Palatability/acceptability of pellets by fish Acceptability tests were conducted with large (average weight 800 g) rainbow trout kept in eight fiberglass tanks (150L) in groups of 15-20 fish per tank (Appendix C). Five of each type of pellets were slowly fed to fish, one at a time, taking care to determine the fate of a pellet fed before introducing the next one. Fish were not fed for 24 h before commencing the tests. Pellets were fed in rotation from tank 1 to 8, feeding only one pellet type to each tank per time. Level of feeding response was recorded as poor, spitting out pellets, and readily eaten. Coating materials designed to alter the appearance of coated pellets could theoretically, also alter the palatability of coated pellets by blocking or masking pellet flavour. In the palatability or acceptability test, most coated pellets were readily consumed except for pellets coated with nail polish glitter (5 out of 5 pellets were spat out by fish). It was, therefore apparent, that the appearance (and taste) of pellets did not adversely affect consumption of pellets by fish under the conditions tested. The pellets coated with nail polish glitter were probably rejected due to both its taste and texture because the material hardened (or set) to form a solid crust that was extremely rough to the feel. 5.3.2 Laboratory experiments on pellet detectability by humans and fish Experiments to test the effect of pellet contrast level on underwater detectability by humans and by fish were conducted at the Bamfield Marine Research Station (Vancouver Island) and at UBC's Department of Animal Science (South Campus, Vancouver) respectively. Visibility of 125 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS pellets from a horizontal view in seawater was conducted at Bamfield Marine Research Station. A set of preliminary experiments was conducted prior to the main experiment on detectability of pellets by fish, in order to establish basic experimental conditions such as fish numbers and light threshold levels. 5.3.2.1 Detectability of pellets (human vision) (a) Coated pellets Large (16 mm) commercially available brown pellets (for feeding yellowfin tuna in Japan) were coated with the previously established range of colours (Table 5.2). The 16 mm pellets were used for this test so that their images on the viewing monitor could be easily detected by the human eye. Materials in cosmetic powder form and dyes were not used because the entire colour range coated on the 6.5 mm pellets was available in paint (liquid, spray and fluorescent) forms which were easy to apply. Reflective coating was still applied in powder form but the sparkly coating from nail polish was excluded because it was one of the few pellet types rejected by fish in the acceptability test. (b) Equipment and experimental set-up The testing apparatus was designed and constructed of aluminium (Figure 5.1), and consisted of an underwater black and white CCTV video camera (Panasonic WV-BP310, B M S Communications Services Ltd., Burnaby) mounted on the horizontal arm. The vertical arm allowed tests to be conducted at different depths by lifting or lowering the apparatus. A l l tests were conducted at Bamfield Marine Research Station in an indoor large 160,000 L seawater pool (2.5 m deep by 10 m in diameter). Pellet images were viewed from a black and white 126 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS video monitor (Panasonic TR-930 CB). A single pellet was suspended in front of the camera via a transparent monofilament attached to a telescopic rod (Figure 5.1). The pellet was gently jigged and moved (by extending the rod) slowly away from the camera while keeping it within the camera view area, until its image (maintained at the centre of the monitor) became barely visible. The distance of the pellet from the camera was then measured with a tape measure attached to the rod and recorded as the maximum visible distance for that particular pellet type. Pellets were moved or jigged to imitate a sinking pellet in a sea cage. Movement also allows an object to be visible at a longer distance. The completely enclosed pool allowed these detectability tests to be conducted under low light environment. Diffused artificial light from a series of fluorescent tubes (5000 K S40 Designer 50, 94 cm, natural light spectrum, Sylvania Lighting Products Ltd., B.C) were adjustable by varying the number of lighted tubes and shading the tubes with strips of card board box. Light levels were adjusted by trial and error until both the most and least detectable pellets were visible within measurable distance on the telescopic rod. Detectability tests were conducted under low air light (0.03±0.01 //moi photon s"1 m"2) and clear water (9.35±0.07 NTU) for two depths (0.4 and 1.5 m) and repeated under low light and turbid (21.8+0.5 NTU) water conditions. Turbidity was increased by adding silt (i.e. Sperser granular Volclay bentonite of particle size between 210 and 840 microns; American Colloid Company, OCL, B.C.). Light levels in air were measured using a terrestrial light sensor (LI-COR Quantum Sensor, Model LI-190S A) while underwater light levels were measured using an underwater spherical light sensor (LI-COR Spherical Quantum Sensor, Model LI-193SA). A l l sensors were connected to a datalogger (LI-COR, Model LI-1000) for instantaneous readings. 127 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS To ceiling Metal sleeve withclamp Rope and pulley for supprting rod Aluminum angle bars Figure 5.1. Apparatus designed of aluminium to measure pellet visibility (human vision) under different light and seawater conditions. 128 Chapter 5. LABORA TOR Y STUDIES ON DETECTABILITY OF PELLETS UNDER VARIO US ENVIRONMENTAL CONDITIONS 5.3.2.2 Detectability of pellets (fish vision, preliminary experiments) The purpose of these preliminary detectability experiments were to: (1) determine the most practical number of fish to put into one tank when conducting the experiments, (2) develop a procedure and design/build the equipment for studying detectability of pellets by fish, (3) determine the light threshold level that will allow fish to distinguish between two pellets offered effectively (under laboratory condition), (4) conduct small-scale (6 replicates) detectability tests for pellet types of all available colours. (a) Experimental fish, tanks and lights Fish for these preliminary experiments were rainbow trout, Oncorhynchus mykiss, of average weight 213 + 49 g, from Abbotsford Trout Hatchery (B.C.) and distributed at random to twelve 150 L oval fiberglass holding tubs. Fish were then allowed to acclimatise under laboratory conditions of constant photoperiod (12L:12D) and ambient water temperature of 11-13°C and fed maintenance ration for 1 month. Fish scheduled for experimentation were not fed on the day of transfer to 160 L glass aquariums (test tanks) and left overnight to allow them time to recover from the transfer. Six identical aquariums (test tank), each measuring 88 cm by 44 cm by 50 cm high, were completely lined with black plastic lining on the inside (on all four vertical sides and the bottom). The plastic sheet was held in place by duct tape and weighted with a metal frame (86 cm by 42 cm) constructed of metal rod (1 cm diameter). In this setup, light would mostly enter the aquarium through the water surface, similar to situations in a sea cage. 129 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Fish must be calm and "comfortable" with their surroundings before they can be expected to perform well in a behaviour-dependent experiment such as this. To determine the most practical number of fish to put in each tank, one, two, three or four fish were placed in an aquarium and left overnight. Their behaviours were observed for 2-3 h after transfer and again the next morning, and recorded as calm, slightly agitated or skittish. For the purpose of this study, fish were considered calm when they maintained their positions for long periods of time, ventilated at a rate similar to that when they were undisturbed in the holding tank with up to 15 other fish, and moved in a gentle slow motion around the tank. Slightly agitated fish swam around the tank more frequently than calm fish but not at high speeds, and had ventilation rates similar to calm fish. Skittish fish were nervous, swimming rapidly from one end of the aquarium to the other and forcefully knocking their heads on the end wall with a thump after each swimming burst. Results from these tests showed that fish were calmest when placed three or four to a tank, remaining calm overnight and into the next day. Single fish per tank were skittish and did not calm down until transferred back to its holding tank the next day. When two fish were kept in each tank, both fish appeared slightly agitated, swimming frequently around the tank. A l l subsequent experiments were conducted with three instead of four fish per tank because the former produced the desired result (calm fish) with the minimum number of fish required. (b) Pellet detection procedure: The dropping method To study fish ability to detect pellets of different contrast, a pellet-dropping method was initially developed. In this method, a pair of pellets (one coated and one uncoated brown pellet) was simultaneously offered to the three fish in an aquarium by dropping the two pellets 130 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS close to the water surface until 20 pellets of each type have been offered. Pellets were dropped close to the water surface in order to reduce splashing which could alert the fish to the pellets' locations. To randomise location of drops, the tank surface was roughly divided into four sections of equal size and assigning them as Quadrants 1-4 (Figure 5.2). The numbers 1-4 (to coincide with quadrant 1-4) from a random table (Snedecor and Cochran, 1967) were used to determine the location of a drop. A horizontal grid (1.5 cm mesh size; 86 cm by 42 cm) placed 2.5 cm off the tank bottom prevented fish from picking up a pellet once it had fallen past the fish thus ensuring that pellets were consumed as a function of their detectability while in the water column. After each test, fish were quickly transferred back to the holding tank. Glass support structure Black horizontal grid (1.5 cm mesh) Figure 5.2 An aquarium with horizontal grid (2.5 cm off the bottom) and the positions of the four quadrants for dropping pellets. 131 Chapters. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS (c) Light threshold for fish under laboratory conditions Once the most practical number of fish to use per tank had been ascertained, and the detection procedure (i.e. dropping method) had been developed, a threshold light level at which fish were just able to distinguish between two pellets of unequal contrast or detectability had to be established. Light from fluorescent tubes (5000 K S40 Designer 50, 94 cm, natural light spectrum, Sylvania Lighting Products Ltd., B.C) was diffused by placing tubes end to end at the aquarium level, on the side opposite to the experimental aquariums in the laboratory (Figure 5.3). The 5000 K S40 Designer 50 fluorescent tubes were selected because the light produced simulated natural light at high noon on a bright sunny day (Sylvania Lighting Products Ltd., B.C). The tank side of the tubes was also shaded with a 23 cm high strip of cardboard so that no direct light rays could reach the experimental tanks. Diffused light was necessary to reduce shadow formations. The level of diffused light was adjusted by increasing or reducing the distance between the light tubes and the wall. Pellets coated white in colour were offered to fish in a dropping manner under a light 1 2 level of 0.27 //moi photon s" m" (light level when fluorescent tubes were shaded but positioned ~ 30 cm (farthest) from the wall (Figure 5.4). To find a threshold light level under laboratory conditions, the light level was adjusted from when fish were able to detect both pellets offered equally well (0.27-0.009 //moi photon s"1 m"2) to a level when fish were not able to detect any pellet at all (outside measurable range of sensor). 132 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Water and air supply (inflow) from ceiling Cardboard shades 23 cm high Black lustre netting taped to wall Fluorescent light tubes placed on a platform 1.0 m above the floor, 20 cm from the wall Door Restraining netlon screen Restraining netlon screen Glass aquariums, 8 x 44 x 50 cm high Fiberglass holding tanks, 150 L Station 2: Pellet suspending device, installed Operator #2 To heat exchanger Station 1: Pellet suspending ' device, installed y Operator #1 Figure 5. 3 Plan view of wet laboratory showing experimental set-up and the position of the light source relative to the aquariums during detectability trials. 133 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Black lustre netting taped to wall Cardboard shades Fluorescent light tubes Movable platform for adjusting distance of light from the wall Water and air supply (inflow) fromceiline 0 Fibreglass holding tank Aquariums U-tube drainage system Concrete support blocks Plywood sheet Figure 5. 4 Cross-section of wet laboratory showing aquarium installation and the movable platform for adjusting distance of light source from the wall and to change light levels in the room. Detectability tests were conducted using the dropping method, under a range of ambient light levels using the pellet type which had the highest contrast level (white), to determine the threshold light level. The threshold light level that resulted in a significant difference between a high contrast pellet and the uncoated pellet (under laboratory conditions) was 0.002 //moi photon s"1 m"2. However, when testing pellet types with lower contrast, the most practical light level was 0.007 //moi photon s"1 m"2 (LI-COR terrestrial sensor, LI-190S A). Subsequently, pellets coated white, grey, red, orange, yellow, green, blue, brown, eggplant, fluorescent colours (red, orange, yellow, green and blue), natural guanine and silver were all tested under the latter threshold light level. 134 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS (d) Replicate and statistical analyses A set of detectability tests consisted of offering 20 pellets of a single coated pellet type and 20 conventional pellets to three fish in one aquarium. Pellets were offered, two at a time, in pairs of one coated and one conventional pellet. Uneaten pellets were quantified and data were statistically analysed (Student's t-test of means) to test the effect of contrast levels on consumption and therefore, detection level. Tests on each type of coated pellet were conducted in six replicates using six different aquariums. Only pellets coated white, silver and natural guanine were significantly more detectable (a=0.05) than conventional uncoated pellets. None of the colours including the fluorescent colours significantly improved the detectability of pellets by fish (data not shown because this was a preliminary short-term experiment). 5.3.2.3 Detectability of pellets (fish vision, final experiments) The objective of conducting an extended set (16 replicates per pellet type) of experiments on detectability of pellets by fish was to allow pellet types of different detectability levels under different background illumination to be compared. This would in turn, enable the selection of one pellet type that could be considered to be most suitable for increasing pellet detectability under a variety of different light and water conditions. (a) Experimental fish and tanks Approximately 150 rainbow trout (153 ± 25 g initial average weight from Sun Valley Trout farm Ltd., Mission, British Columbia) were distributed at random to twelve 150 L oval fiberglass holding tubs for the final detectability experiments. The fish were allowed to acclimatise under laboratory conditions of constant photoperiod (12L:12D) and ambient water 135 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS temperature of 6-8°C and were fed a maintenance ration for 2 months. Using a heat exchanger, the water temperature was gradually raised to 13°C over a period of 18 d at VA - Vi°C d"1. Fish were acclimatised for a further 2 wk at this temperature before starting the experiments. Fish scheduled for experimentation were not fed on the day of transfer to 160 L glass aquariums as in previous trials, and were left overnight to allow them time to recover from the transfer. Another six identical units of aquariums with similar dimensions were added to the six existing ones used for the preliminary trials. These tanks were similarly lined with black plastic lining on the inside (on all four vertical sides and the bottom), so that all 12 tanks were identical in all respects. A new batch of rainbow trout had to be conditioned for this part of the experiment because the batch used for the preliminary trials had grown too big (>800 g) by the time this set of experiments was ready to commence. (b) Pellet detecting procedure: The suspending method In the preliminary pellet dropping experiments, fish might or might not have had an equal initial opportunity to detect and attack either of the pair of pellets. The effective reaction distance in that method depended on where the fish were in the water column when the pellets were dropped. Fish could also theoretically learn to expect pellets both temporally and spatially. These potential biases were reduced in the said experiment by randomly dropping the pair of pellets at unpredictable locations from the water surface in the tank and randomly changing the positions (left or right) of coated and conventional pellets. However, in order to ensure that all experimental fish in a tank had an equal opportunity to detect and attack either pellet, fish in this final detectability experiment were confined to one end of the tank at the start of each test. The confining screen was pulled out of the tank, once the pair of pellets to be 136 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS tested has been suspended at the opposite end of the tank, much like the horseracing "start" procedure. Both pellets were suspended equidistant from the screen. This method is referred to as the suspending method. (c) Pellet suspending device and screen A pellet-suspending device (Figure 5.5) was designed and constructed by soldering two metal rods of equal length (diameter 2.38 mm; length 38 cm) to form a T-configuration. The T-configuration consist of the top of the T (designated as "rod A") and the body of the T (designated "rod B"). Rod B was inserted into a metal sleeve (inside diameter 3.08 mm; outside diameter 6.33 mm; length 33 cm), such that rod B was able to swing freely inside the sleeve. The friction between rod B and the inside of the metal sleeve was adjusted using varying lengths of Scotch tape, so that rod B would only rotate when there was a definite pull from either end of the suspending beam (see Figure5.5). A pair of pellets (always a control pellet and a pellet with the coating being tested) was suspended on either end of rod A (which became the suspending beam) via transparent mono-filaments. Each pellet was held by a U -shaped flexible hook (and spring loop) formed from black stove pipe wire (20 ga) which gripped the pellet well enough during the experiment while allowing easy replacement of pellets at the end of each trial. Pellets were gripped on the ends so that the sides of pellet would be fully exposed to the fish. 137 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS I Spring loop Black U-shaped hook Q4r^-" Feed pellet Figure 5. 5 Pellet suspending device 138 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS A rigid screen with dimensions to match the inner measurements of a cross-section of the aquarium, was constructed from a wooden frame and black plastic Netlon netting with 8 mm by 8mm mesh size (Netlon Z-30, Tanaka Sanjiro Co. Ltd). The screen restrained fish to one end of the aquarium while a pair of pellets was suspended 24.5 cm off the bottom and 12.5 cm away from either side of the aquarium. Under this configuration, the pellets in each pair were 18.0 cm.apart (Figure 5.6). The dimensions for placement of the pellets were based on fish size. Specifically, fish were able to swim freely below, above (but not directly above each pellet because of the mono-filament), in between or to the sides of the pellets without having to touch or bump into the pellet accidentally, i f they could not detect the pellet. The screen was lifted in one quick movement to allow fish equal starting distance from which to detect either pellet. Restraining fish with a screen ensured that the primary factor determining which pellet would get attacked was the reactive distance associated with that particular pellet type. 139 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Rod A 5.5 cm Water level 50 cm ^12.5 cm^ 20 cm 61 cm 18 cm Black plastic ends L2.5 cm 24.5 cm Mono-filament Pellet Tank wall with black plastic lining Solid metal frame (86 cm by 42 cm) to hold plastic sheet in place 44 cm Figure 5. 6 Cross-section of aquarium showing the relative positions of pellets while suspended during detectability trials. 140 Chapter 5. LABORA TOR Y STUDIES ON DETECTABILITY OF PELLETS UNDER VARIO US ENVIRONMENT A L CONDITIONS (d) Testing sequence and fish response A sequence of steps from start to end of a trial was established and strictly followed for all trials as follows: a. Wipe fingers (forefinger and thumb) with damp paper towel before picking up any pellet, b. Select position of pellet type and install pellet by inserting lengthwise between the U-shaped hook, always installing conventional pellet before coated pellets to avoid contamination of conventional pellet with coating material, c. Wipe fingers after handling each pellet, d. Confine fish to one end of the aquarium and secure screen in position, e. Install pellet-suspending device on aquarium such that the pair of pellets is suspended in the water, f. Remove screen and start timer as soon as screen is out of the water, g. Observe beam and wait for tilt (response), h. Stop timer and lift device immediately upon getting a response (attacked pellet) or after 2 min (if no response), i . Record the response and the response time (when required), j . Remove used pellets, wipe fingers and repeat steps a to i . A response depended on a fish detecting a pellet successfully, aligning itself for a strike and attacking the pellet by "inhaling" it into its mouth. The attack produced a distinct pull or jerk on the mono-filament resulting in an abrupt tilt of the T-top to the side of the pull, constituting a response. A false response was observed to result from eddies created by a tail flip, the tail flip itself or when a fish accidental brush up against a pellet. An R or a C was registered as a response if the conventional or coated pellet, respectively was attacked. Pellets were removed from the tank after 2 min (from removal of screen), and a no response (NR) was registered i f fish did not attack either pellet. The time frame of 2 min was selected for the following reasons: 1. In preliminary trials, most responses occurred within 2 min 141 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS 2. To ensure that the time difference in experimentation between fish in the first set of trials and the last set (3 - 3lA h) is not exceedingly large. 3. In the field, pellets rarely stay within a fish feeding aggregation for more than 2 min. Threshold light intensity level in air, used throughout the clear and silted water sets of trials was 0.007±0.002 /jmol photon s"1 m"2 (reading from terrestrial sensor, LI-190SA) or 1 2 0.014±0.002 //moi photon s" m" (if taken from spherical sensor, LI-193SA). The light levels in water (clear or silted) would be much lower than levels in air but the LI-COR Quantum light sensors used were not sensitive enough at these intensities to register any meaningful readings. The time taken for a fish to respond (pellet attack) after being released was also recorded for a segment of the experiment to obtain a sample of response times and to determine whether response time varied between pellet types. (e) Pellet selection In this final set of detectability tests on fish, five types of pellets were selected based on the relative merits of each pellet type under its respective optimum performing conditions of light and water conditions. Pellets coated with natural guanine were chosen to see how well synthetic materials performed against a natural product (natural guanine). The selected pellet types were randomly assigned code names "PI" to "P5" to eliminate operator bias (Table 5.5). 142 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5 . 5 . Pellet types chosen and their corresponding visual characteristics. Pellet type Visual characteristics Coating material Basis for selection PI White Synthetic Highest contrast in air. 1 Very visible underwater at low light, clear water.2 Most detectable by fish.3 P2 Green Synthetic Low contrast in air. 1 Neutral to less detectable than conventional pellet at low light, clear and silted water. 2 As detectable as conventional pellet.3 P3 Red Synthetic Low contrast in air. 1 Very visible at low light and turbid water but neutral to worse than conventional pellet at low light, clear water.2 As detectable as conventional pellet.3 P4 Reflective/ Silver Synthetic High contrast in air. 1 More visible than conventional pellet at low light, clear water but neutral at low light silted water.2 More detectable than conventional pellet.3 P5 Guanine Natural High contrast in air. 1 More visible than conventional pellet at low light, clear water but neutral to worse than conventional pellet at low light, silted water. 2 More detectable than conventional pellet.3 1 Contrast levels of pellets against a black background. See Table 5.4, for details of contrast levels against different background colours. 2 From results of tests on visible distance with human vision and underwater camera. 3 From results of preliminary detectability tests using rainbow trout. (f) Replicates (clear and turbid water) and statistical analyses Replicates. The aquarium and pellet type to be tested each day was assigned in a randomised block design. Fish from the fiberglass holding tubs were used following a random pattern. This gave experimental fish similar resting periods before being used again. Two sets of six aquariums were used for the detectability tests with the sixth aquarium as backup in case fish in any one of the five aquariums were inactive or conversely, hyperactive (chasing and fin nipping). With 12 aquariums, it was possible to conduct 2 sets of identical experiments concurrently. This set-up 143 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS required two operators and 2 pellet suspending devices, all of which were switched during the course of the 8 d experiment to reduce bias from either operator effect or device effect. A n ideal switching program would be to follow a three-way combination set-up as shown in Figure 5.7 below and randomly selecting which 2 d combination to use first. However, because this set-up was not put in place prior to the experiments, switches were conducted every 2-3 d arbitrarily between operator, suspending devices and the 2 sets of aquariums. D l D2 Aquarium 7 Aquarium 1-6 Operator A Operator B Figure 5. 7 Three-way combinations for switching operator, suspending devices (Dl and D2) and aquariums. The experiment was conducted under low light and with clear water for 16 replicates and repeated under low light and with turbid water for additional 16 replicates. The pellet type to be tested in a particular aquarium was assigned at random each day. Small pieces of paper were written with codes for the different pellet type (PI, P2, P3, P4 and P5) and shaken in a small box before picking them up one at a time. The sequence with which the coded papers were picked up was then used as the sequence to assign pellet type to the aquariums. Aquarium numbers 6 and 12 were always designated backup aquariums. 144 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Light and turbidity levels established in the preliminary trials were used. Measured quantities of Sperser granular Volclay bentonite (American Colloid Company, OCL, B.C.) of particle size between 210 and 840 microns (seived to select smaller particle ~ 210 microns) and water were mixed to form a slurry which was added to the aquarium before conducting each set of tests in the silting experiment. Turbidity readings were taken before and after each test. The placement of pellets on the left or right side of the aquarium was determined using a random table (Snedecor and Cochran, 1967; odd numbers = left; even numbers = right) while ensuring that there was an equal number of lefts and rights in each set of trials. Statistical analyses. A two-way analysis of variance was conducted on data from Bamfield Marine Station (human vision) where two observations were made on each of 18 different types of pellets at two depths and two water clarity conditions. The two-way analysis of variance was set up with distance as the outcome and depth and pellet type as the factors and a was set at 0.05. The Tukey test (95 % confidence intervals for specified linear combinations) was used to compare each pair of pellet types. In the final detectability experiment with Rainbow trout in the laboratory, the goodness-of-fit test was selected for testing significance between coated and non-coated pellets because data were counts of successes. Contingency table analysis was selected for testing significance between the different coated pellet types for the same reason, i.e. because one is comparing "success rates" when the data are counts of successes (S-Plus v3.3; MathSoft, Inc. and Stata Statistical Software v4.0; StataCorp). An alternative approach, the analysis of variance with percentages or proportions of successes combined with an arcsine data transformation, is valid 145 Chapter 5. LABORA TORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS when there are many observations, but the chi-square test used in the contingency table approach is valid for smaller numbers of observations. The question of dependence arises when multiple observations are made on the same fish, as was the case in these experiments. The dependence is intricate but a conservative adjustment can be made to the chi-square test by assuming the worst possible case. This case is the one in which the dependence is complete, that is, each fish acts in exactly the same way each time. A total of 150 fish were used in the 96 separate sets of observations (including fish in backup aquariums). Thus, each fish was observed at the most, two times. One validated correction to the chi-square statistic (Brier, 1980) is to divide it by 1 + (n-l)r, where n is the number of observations and r is the intraclass correlation coefficient. If we have complete dependence, then r = 1. Taking n = 2 (because each was observed at most two times), yields a correction factor of 1 + 1 = 2. In other words, one can correct the chi-square statistic for intra-fish dependence by dividing it by 2. This correction combined with the Bonferroni correction (which in itself is conservative), leads to extremely conservative tests of significance. It is possible, for example, that i f we were able to unravel the complicated dependence structure in these experiments, the correction factor would be closer to 1 than it is to 2. Thus, any of the following tests that are significant at the p = 0.005 level can be considered extremely significant. It is likely, in fact, that applying a significance level of p = 0.01 to each test would result in an experiment-wide Type I error rate of 0.05, the conventional level. The 'no response' (NR) variable was treated as a nuisance variable (by original design) and was ignored in all statistical analyses. 146 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS 5.4 Results Under low turbidity (3.6±1.4 NTU) water conditions, pellets coated white, silver and guanine were significantly more detectable than uncoated pellets. Detectability of the green and red coloured pellets was not significantly different from the detectability of the uncoated pellet. Under high turbidity (20.1+1.5 NTU) conditions, pellets coated white, silver and guanine were also significantly more detectable by fish than uncoated pellets. The visible ranges for white, silver and the fluorescent coatings were broader than the range for uncoated pellets under low turbidity (9.35±0.07 NTU) seawater conditions. On the' other hand, under higher turbidity (21.8+0.5 NTU), no coated pellet type differed appreciably from conventional pellets in visible distance although visible distances for darker pellets (e.g. red, eggplant and blue) as well as white were slightly greater. Most fluorescent pellets were less visible than conventional pellets at high turbidity. The effect of silt was to brighten the environment due to a phenomenon described in literature as veiling brightness. 5.4.1 Detectability/visibility of pellets (human vision) Results of visibility tests in seawater conducted at BMS are summarised in Tables 5.6 a and 5.6 b. Pellet types were ordered according to the maximum distance that their images remained visible on a monitor connected to the viewing underwater camera. Visible distances for different pellet types could not be statistically analysed for significant differences because only two observations (n=2) were taken for each type of pellet. However, at a depth of 0.4 m, three groups of pellets with markedly different means were evident based on magnitude of difference (> 0.74 m) between individual pellets at low light (0.03±0.01 //moi photon s"1 m"2) and clear 147 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS (9.35±0.07 NTU) seawater conditions (Table 5.6 a). Group A (mean visible distance of 4.26 m) contained pellets coated white, silver and all the fluorescent colours studied. Group B (mean 2.29 m) was comprised of natural guanine and yellow while Group C (mean 1.31 m) contained sparkly, grey, orange, conventional uncoated pellet, blue, red, green and brown. At 1.5 m, a similarly approximate grouping was also evident (individual difference > 0.26 m) but absolute distances of each group were reduced by up to 40%. Group A ' (mean 2.46 m) contained the pellets visible at 0.4 m plus yellow and natural guanine. Group B ' comprised of a single pellet type i.e. orange (1.80 m) and could be part of either group A ' or C ' . Group C' (mean 1.24 m) comprised of most of those visible at 0.4 m minus orange and green but included eggplant (violet). Under low light (0.03+0.01 //moi photon s"1 m"2) and turbid (21.8+0.5 NTU) seawater conditions, white pellets remained more visible (~ 6%) than conventional brown pellets. No groups of pellets with distinguishable means were evident at either depth studied. In fact, pellets belonging to Groups A and B of the previous (clear seawater) tests (e.g. silver, natural guanine and all the fluorescent coloured pellets) did not differ much from conventional brown pellets in visible distance (Table 5.6 b). When ordered, dark coloured pellets (e.g. red and eggplant) appeared at the top of the list with fluorescent colours at the bottom of the list. The results from the two-way analysis of variance indicated that both type and depth influence the visible distance of pellets (very small p values). For clear water conditions, pellet types had F(17,53) = 16.52, p = 0.000 and for depth, F(l,53) = 32.73, p = 0.000 For turbid water condition, dot significant but pellet type was significant, F(17,53) = 16.52, p = 0.000 and 148 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS for depth, F(17,53) = 10.31, p = 0.000. See Appendix 2 for differences between coated pellet types. Table 5.6 a. Maximum horizontal visible distance (human vision) of different pellet types under low ambient light (0.03±0.01 /anol photon s"1 m"2) and clear (9.35±0.07 NTU) seawater conditions at two depths. Depth: 0.4 m Depth: 1.5 m Underwater light level: 0.017±0.007 /moi Underwater light level: 0.012±0.007 //moi photon photon s"1 m"2 s"1 m"2 Pellet type Maximum visible Pellet type Maximum visible distance, distance, m m White 4.60 ~^ White 2.74 Fluorescent yellow 4.58 Fluorescent yellow 2.72 Fluorescent red 4.39 Fluorescent red 2.69 Fluorescent orange 4.29 )>- Group A Fluorescent orange 2.68 Group A' Fluorescent blue 4.25 Mean=4.26 Yellow 2.45 Silver 3.98 Silver 2.39 Mean=2.46 Fluorescent green 3.75 _^ Fluorescent green 2.28 Natural guanine 2.38 ^_ Group B Fluorescent blue 2.16 Yellow 2.20 _ Mean=2.29 Natural guanine 2.06 J Sparkly/speckled 1.46 Orange 1.80 Group B' Grey 1.44 Sparkly/speckled 1.41 Orange 1.43 Blue 1.34 UNCOATED 1.39 V- Group C Eggplant 1.24 Blue 1.23 Grey 1.23 Group C f Mean=1.24 Red 1.21 Mean=1.31 UNCOATED 1.21 Green 1.19 Red 1.15 Brown 1.16 Brown 1.12 Eggplant 1.12 Green 1.08 U N C O A T E D refers to conventional pellet that has not been coated. Groups A, B and C are pellet categories with markedly different means (individual differences >0.74 m). Groups A' , B' and C are pellet categories with markedly different means (individual differences >0.26 m). 149 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5.6 b. Maximum horizontal visible distance (human vision) of different pellet types under low ambient light (0.03±0.01 /moi photon s'1 m"2) and turbid (21.8±0.5 NTU) seawater conditions at two depths. Depth: 0.4 m Depth: 1.5 m Underwater light level: 0.012±0.007 //moi photon s"1 m"2 Underwater light level: < 0.012 //moi photon s'1 m"2 Pellet type Maximum visible Pellet type Maximum visible distance, m distance, m Red 1.53 Eggplant 1.55 Eggplant 1.47 Red 1.51 White 1.43 Blue 1.48 Blue 1.43 Sparkly/speckled 1.47 Grey 1.42 Green 1.45 Green 1.41 White 1.45 Brown 1.40 Grey 1.41 Fluorescent yellow 1.34 Brown 1.39 UNCOATED 1.34 UNCOATED 1.36 Sparkly/speckled 1.34 Fluorescent yellow 1.36 Orange 1.32 Silver 1.29 Silver 1.31 Natural guanine 1.27 Natural guanine 1.31 Fluorescent blue 1.26 Yellow 1.31 Orange 1.25 Fluorescent blue 1.23 Fluorescent green 1.24 Fluorescent red 1.15 Yellow 1.20 Fluorescent orange 1.15 Fluorescent red 1.19 Fluorescent green 1.14 Fluorescent orange 1.16 U N C O A T E D refers to conventional pellet that has not been coated. 150 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS 5.4.2 Detectability of pellets (fish vision) Approximately 400 pairs of pellets were offered to fish during preliminary pellet selection and light intensity adjustment stages conducted in the summer of 1996. The threshold level in clear (3.6±1.4 NTU) water at which fish (rainbow trout, Oncorhynchus mykiss), were able to detect a difference between conventional brown (low contrast) and white (high contrast) pellet was 0.002 //moi photon s"1 m"2. However, fish were able to distinguish between conventional brown pellets and other types of coated pellets (having contrast that is lower than that of a 1 2 white pellet) at a light level of 0.007//mol photon s" m" . Results from the final experiment on detectability of pellets by rainbow trout, revealed that under similar low light (i.e. 0.007/anol photon s"1 m"2) and clear water conditions, pellets coated white, silver and guanine were significantly more detectable than conventional brown pellets (x goodness of fit, oc=0.05). Green and red coloured pellets were not significantly different from conventional brown pellets (Table 5.7 a). Under low light and turbid (20.1±1.5 NTU) water conditions, pellets coated white, silver and guanine were also significantly more detectable than conventional brown pellets although to a lesser extent (x goodness of fit, a=0.05) (Table 5.7 b). The detection of green and red coloured pellets was not significantly different from conventional brown pellets. Under low light and clear water conditions, pellets coated white, silver and guanine were significantly different from pellets coated green and red but not from each other (multiple comparison, x contingency table, ct=0.05, 1 df). Responses from pellets coated green were not significantly different from those coated red (Table 5.7 a) both under low light clear water and low light turbid water conditions. Under low light and clear water, there was no change in the pattern of significance when data were analysed using multiple comparison x contingency 151 Chapters. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS table with Bonferroni correction (a=0.005). Under low light and turbid water conditions, the Bonferroni correction analyses (a=0.005) changed the pattern of significance so that the visibility of pellets coated white was not significantly different from that of pellets coated green, silver and guanine; pellets coated green were not significantly different from those coated red and silver; and pellets coated silver were not significantly different from those coated with guanine. The only pellets that were significantly different were those coated white from green, red from silver and red from guanine. While required as part of a proper statistical procedure, the Bonferroni correction resulted in an extremely stringent condition (a=0.005) for statistical significance. A true null hypothesis occasionally will be rejected, which means that one has committed an error in drawing a conclusion about the sampled population and this error will be committed with a frequency of a. This rejection of a null hypothesis when it is in fact true is known as a Type I error. Conversely, if Ho is in fact false, a statistical test will sometimes not detect this fact, and one shall thus reach an erroneous conclusion by not rejecting Ho. The probability of not rejecting the null hypothesis when it is in fact false is represented by p. This error is known as a Type II error. For a given sample size, n, the value of a is inversely related to the value of (3. This means that lower probabilities of committing a Type I error are associated with higher probabilities of committing a Type II error. Thus, by experience, and hence by convention, an a of 0.05 is considered to be a "small enough" chance of committing a Type I error, while not being so small as to result in "too large a chance" of committing a Type II error (Zar, 1999). Both types of error may be reduced by increasing the sample size. When n is small, as can be the case in some field trials with limited experimental units, the power of a 152 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS statistical test is much reduced since, for a given a, larger sample size will result in statistical testing with greater power (1-P). Industrial decision making is influenced by the need to obtain outcomes that favour a new product or a service and so the significance levels tend to be set higher than 0.05 (e.g. 0.10) for rejection of Ho. When combined with small sample sizes (e.g. to reduce cost), industrial decision making may be more prone to commit Type I errors. Scientific research, on the other hand, focuses on contributing to knowledge and the pressure is on reducing both Type I and Type II errors. Thus, sample sizes are larger and significance levels are set at typically 0.05. However, in some experiments where a 5 % chance of an incorrect rejection of Ho is unacceptably high, the level of significance is sometimes set at 1 %. The changes to the pattern of significance between coated and conventional pellet or between the different pellet types after applying this Brier (1980) correction to account for data dependence was minimal and affected only data from the experiment conducted under turbid water condition. That is, there was no longer any statistical difference between coated pellet type P4 and non-coated pellet (%2 = 9.93; p = 0.0016; critical %2 value is 10.60) (Table 5.7b). Significance differences between the coated pellet types also affected data only from the turbid water conditions as shown in Table 5.8a and Table 5.8b, respectively. That is, there are no longer any significant differences for coated pellet types P2 vs P4 and P3 vs P5 under turbid water condition (Table 5.8b). 153 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5.7 a. Responses from detectability tests (fish vision) conducted under low light (0.007 //moi photon s"1 m"2) and clear water (3.6 ± 1 . 4 N T U ) conditions. Coated Pellet type Conventional Responses: Coated NR Brier-corrected %2 value (p value) PI 15 102 43 32.35 (0.000) P2 65 53 42 0.61 (0.435) P3 46 54 60 0.32 (0.572) P4 23 94 43 21.54 (0.000) P5 26 94 40 19.27 (0.000) N R : N o response Underscored coated responses are significantly different from conventional pellets (Goodness of fit %2, cc=0.05). Critical value o f Brier-corrected x 2 value is 10.60 Table 5.7 b. Responses from detectability tests (fish vision) conducted under low light (0.007 //moi photon s"1 m"2) and turbid water (20.1 ± 1 . 5 N T U ) conditions. Coated Pellet type Conventional Responses: Coated NR Brier-corrected x 2 value (p value) PI 37 92 31 11.72 (0.001) P2 58 76 26 1.21 (0.272) P3 52 54 54 0.02 (0.891) P4 40 91 29 9.93 (0.002) P5 30 80 50 11.36 (0.001) N R : N o response Underscored coated responses are significantly different from conventional pellets (Goodness of fit %2, ot=0.05). Critical value o f Brier-corrected %2 value is 10.60 154 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5.8 a. Multiple comparison of different coated pellet types under low light (0.007 //moi photon s"1 m'2) and clear water (3.6 + 1.4 NTU) conditions. Pellet types Bonferroni-corrected X 2 value df Brier-corrected %2 value P value Outcome PI vs P2 46.75 2 23.38 0.000 S PI vs P3 33.33 2 16.67 0.000 S PI vs P4 2.01 2 1.00 0.607 NS PI vs P5 3.27 2 1.64 0.440 NS P2 vs P3 6.44 2 3.22 0.200 NS P2 vs P4 31.49 2 15.75 0.000 S P2 vs P5 28.20 2 14.10 0.001 S P3 vs P4 21.28 2 10.64 0.005 S P3 vs P5 20.37 2 10.19 0.006 S P4 vs P5 0.29 2 0.15 0.928 NS df: degrees of freedom. S: significant NS: Not significant Critical value of Brier-corrected %2 value is 7.88 155 Chapters. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS Table 5.8 b. Multiple comparison of different coated pellet types under low light (0.007 //moi photon s"1 m"2) and turbid water (20.1 ± 1.5 NTU) conditions. Pellet types Bonferroni-corrected X 2 value df Brier-corrected X 2 value P value Outcome PI vs P2 6.61 2 3.31 0.190 NS PI vs P3 18.64 2 9.32 0.009 S PI vs P4 0.19 2 0.09 0.212 NS PI vs P5 6.03 2 3.01 0.222 NS P2 vs P3 16.60 2 8.29 0.031 S P2 vs P4 13.85 2 6.93 0.300 NS P2 vs P5 4.82 2 2.41 0.016 NS P3 vs P4 18.54 2 9.27 0.010 S P3 vs P5 11.10 2 5.55 0.062 NS P4 vs P5 7.72 2 3.86 0.145 NS df: degrees of freedom. S: significant NS: Not significant Critical value of corrected %2 value is 7.88 156 Chapter 5. LABORA TORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS 5.5 Discussion Results from measurements of pellet contrast against different background colours suggest that a pellet can have variable contrast values and, therefore, detectability will change with changes in the characteristics of its background environment in terms of lighting and colour conditions. Under bright ambient light in air, bright objects (e.g. white, silver) against dark background (e.g. black) produces the highest contrast. Similarly for the reverse condition of dark objects (e.g. brown, red, blue) against bright background (e.g. white). This phenomenon also holds true for objects at low light intensities (0.03 //moi photon s"1 m"2) in aquatic environments where brighter objects (white, silver and fluorescent colours) can be seen at farther distances than darker ones against a dark background. The reverse condition of darker objects being more visible under a brighter surrounding environment is seen under silty water condition. The addition of silt (during the second part of the experiments at Bamfield Marine Research Station) to simulate turbid field conditions, "brightened" up the background environment. The brightening effect of silt is due to light scattering, i.e. light repeatedly reflected and refracted off each of the particles. This multiple scattering of light creates "veiling brightness" and reduces contrast (Lythgoe, 1988). The veiling brightness effect reduces the apparent contrast of objects (Gazey, 1970) for a given light intensity in air. This is evident from the experiments performed at BMS where, under a constant light level in air of 1 2 0.03 //moi photon s" m" , the visible distance of a bright pellet (e.g. white) is reduced by up to 69% (from 4.60 m to 1.43 m at 0.4 m depth) with silting. The visible distance of a dark pellet (e.g. red) did not improve by a corresponding amount (i.e. only 26%) due to the silting effect. The reduction in visible distance of bright pellets (including the fluorescent colours) due to 157 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS silting in the said experiments could also be influenced by the reduction in actual light available for viewing pellets (O.O 12 //moi photon s"1 m"2). In preliminary experiments on pellet detectability by fish (dropping method), fluorescent colours were not more detectable than conventional brown pellets under clear or turbid water conditions at low light intensity (0.007 //moi photon s"1 m"2). This was probably because the ambient light level (especially of ultraviolet) in the tank was below which fluorescence could be effective. Fluorescent colours were eliminated from further tests with fish because they were not readily available, often toxic, and expensive. Under laboratory conditions of small darkened holding facilities (e.g. aquariums), the light threshold at which rainbow trout can selectively detect a high contrast object (white pellet) over a similar-sized object (conventional brown pellet) of lower contrast level was 0.002 //moi photon s"1 m"2. That level was similar to the "threshold" reported in literature of 0.002 //moi 1 2 photon s" m" for fish in the wild, during light/dark adaptation (equivalent to late dusk or early dawn; Blaxter, 1970). However, the actual light level used for subsequent tests on detectability was higher (0.007 //moi photon s" m") probably because more light was required for fish to successfully detect pellets with lower contrast. In open waters, more light is able to penetrate the water thus providing fish a brighter environment at an equivalent surface light level of 1 2 0.002 //moi photon s" m" as reported by Blaxter (1970). Contrary to visibility tests at BMS, bright pellets (i.e. white, silver and guanine) remained significantly more detectable to fish in aquariums during the silting experiments under low light and turbid water conditions (final detectability tests). This was probably a result of the viewing angle at which fish in the aquarium were able to see pellets. In the tank, 158 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS the viewing angle was dependent on whether the fish was swimming above or below the suspended pellets (Figure 5.7). Unlike dark coloured pellets (i.e. uncoated, green and red), the brightness attributes of bright pellets (white, silver and guanine) would continue to manifest themselves and be effective (by reflecting downwelling light) to a fish looking down on a pellet thus reducing the veiling effect. Dark coloured pellets actually simulate the camouflaging attributes of the darker dorsal surfaces of fish when viewed from the water surface. The reverse would be true for a fish looking up toward the pellets, i.e. dark coloured pellets would appear more detectable (against the background surface light) than bright ones, while the bright pellets would simulate the camouflaging attributes of the silver ventral side or belly of a fish. 1 9 Under low light (<0.03 //moi photon s" m") bright pellets, especially white, were considerably more detectable than conventional brown pellets under most water conditions. Unlike other colours (fluorescent and non-fluorescent), white colour and reflective attributes such as silver reflect light at all available wavelengths and therefore, do not undergo colour shifts with changes in wavelength composition of ambient light. Fluorescent colours often appear different in air (with higher proportion of ultraviolet radiation) and water. Above water fluorescent colours appear brighter to the human eye than non-fluorescent colours. In clear seawater, non-fluorescent colours undergo colour shifts when viewed at depths (~7 m) with a notable shift for red and orange (Johnson, 1984). Longer wavelength reds shift to brown and then black with greater depths (>18 m). In terms of detectability of feed pellets within the feeding volume of a cage, the combined results of experiments on detectability of pellets by fish and humans, suggest that bright (e.g. white and silver) pellets are most detectable under low light and therefore, dark water conditions, typical of depths at the lower layers of the feeding volume. Silver pellets will 159 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS sparkle to permit better viewing by humans. In the upper layers of the feeding volume where surface light intensities are usually higher, dark pellets (e.g. conventional brown pellets) will be more detectable than brighter pellets when viewed horizontally. In Figure 5.8, arrows a and b represent optical paths from the silver and conventional (brown) pellet respectively, when viewed upward. Both pellets are detected equally well because solid objects form dark silhouettes against a bright background (the surface light). Arrows c and d represent optical paths from the two pellets when viewed horizontally. Contrast values for silver and brown pellets differ greatly with the silver pellets being more detectable because more light from the upper light source is reflected off the silver pellet to the observer. Brown pellets reflect even less light as light intensities decline with increasing depth and eventually approximate the space light resulting in poor contrast as represented by arrow f. Silver pellets retain their brightness (represented by arrow g) as the background darkens thus improving detectability relative to the brown pellet. The viewing angle looking downward also contributes further to the increased detectability of silver pellets while reducing the detectability of brown pellets. When pellets in water were viewed from above water, brighter pellets (white and silver) were visible to greater depths than conventional brown pellets. Increased visibility allows an observer more time to judge the fate of pellets (i.e. either captured or ignored by fish) before they sink from view. The white pellets used in the experiments were difficult to make and required materials to be coated at high application rates. Reflective pellets such as silver, on the other hand, were easily produced in large quantities at the lowest application rate and the coating materials were available relatively cheaply. Consequently, reflective silver pellets were 160 Chapter 5. LABORATORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS recommended for use in the next stage of the research (i.e. to extend feeding volume using contrast-enhanced pellets, Chapter 6). L I G H T S O U R C E Humic particles Figure 5. 8 Different optical paths affecting detectability of silver and conventional brown pellets at different viewing angles and at different depths (and, therefore, light levels). 161 Chapter 5. LABORA TORY STUDIES ON DETECTABILITY OF PELLETS UNDER VARIOUS ENVIRONMENTAL CONDITIONS 5.6 Conclusion Pellet contrasts can change with changes in the characteristics (lighting and colour conditions) of its background environment. The contrast of conventional pellets was successfully enhanced by coating them with titanium-dioxide-coated mica (TCM) and other bright materials. At low light intensities, bright pellets (e.g. white, silver and fluorescent colours) have high contrast against a dark background (e.g. black) and can be viewed at considerably farther distances (human vision) or be significantly more detectable to fish than conventional brown pellets under clear or turbid water conditions. Reflective pellets such as those coated with T C M can be produced in huge quantities with relative ease and low costs and may be a good choice for increasing pellet contrast of conventional commercially available pellets. 162 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES Chapter 6 E F F E C T OF C O N T R A S T - E N H A N C E D P E L L E T S O N S A L M O N I D F E E D I N G P A T T E R N S IN S E A C A G E S 6.1 Summary Salmon feeding activities were non-uniform, varying both spatially and temporally during a feeding event. Theoretically, the rate of feeding per meal should increase if uniform feeding was achieved. A n experiment was performed over a period of approximately 1.5 months to test whether uniform feeding could be achieved if fish were provided the means to detect and capture pellets throughout the cage so that the proportion of the volume of the cage utilised for feeding (feeding volume) was increased. Fish in separate cages on a commercial fish farm were fed either silver (contrast-enhanced) pellets or conventional dark brown pellets while using underwater cameras to monitor fish and control pellet loss near the cage bottom. The silver pellet could be detected by a human observer at noon on a sunny day (Secchi disc reading of 9 m) up to 41% deeper than a conventional brown pellet. Feeding rates per meal decreased with underwater light intensity and as stocking density increased. The feeding rate for silver pellets was lower than the rate for conventional brown pellets. The fraction of the total feeding time spent at different levels in the water column (surface, near the cage bottom and throughout the cage), among all cages was not significantly different. Salmon offered silver pellets preferred feeding near the cage bottom (~ 10 m), thereby increasing the feeding volume. As a consequence of this preference, feeding rate was reduced in the silver pellet cages to prevent pellet wastage. It was surmised that the size of the feeding volume affects feeding rate but does not promote uniform feeding. Vertically increasing the feeding 163 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES volume by increasing pellet detectability or stocking density increased the amount of time required for pellets to sink through the feeding volume, thus, increasing the feeding time instead of decreasing it as was initially expected. Apparently, the feeding patterns exhibited by the fish did not permit an increase in feeding rate to compensate for a larger feeding volume. Uniform feeding will not be achieved unless the observed feeding patterns, which could have formed as a result of environmental factors, prior conditioning and genetics, can be modified. Future work could be focused on the early detection and understanding of the formation of these patterns. It is recommended that contrast-enhanced pellets be used because they offer fish a choice of where to eat and provide farmers with a better way of judging feeding rates from the surface. 6.2 Introduction Feeding is affected by many factors including temperature, dissolved oxygen, visibility, appetite, feed availability, feed size and palatability. The ideal feeding method has been envisioned as being one in which competition and aggressive behaviours are minimised while foraging efficiency is maximised (Noakes and Grant, 1990). A feed discharge method that broadcasts feed over a wide area of the surface feeding plane can make food indefensible and reduce competition. Food should be delivered in a manner as to render it indefensible (Dill et ai, 1981; Puckett and Dill , 1985). The delivery of food so as to be indefensible minimises aggression and maximises foraging efficiency (Noakes and Grant, 1990). One foraging model developed using birds suggests that in response to a patchy distribution of food, birds would distribute themselves so that no bird is able to increase its feeding rate by switching to another patch. This phenomenon 164 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES described as an ideal free distribution by Fretwell and Lucas (1970) might be applied to fish in a cage i f fish have unconstrained movements. Conversely, interference competition is promoted by concentrated food sources (Gillis and Kramer, 1987) and inefficient foraging. Dominance hierarchy is especially pronounced with slow continuous distribution of food, allowing dominant individuals to monopolise a large share of resource through defence of space (Thorpe et al., 1990). In a sea cage environment, as in smaller enclosures such as tanks and raceways, indefensible feeding should be achievable when feed is broadcast uniformly over the surface of the enclosure and at a rate high enough to ensure that feed can reach fish at the bottom of the feeding school. Swimming patterns of fish in a sea cage environment can change with different conditions such as stocking density. Juell and Westerberg (1993) observed that an Atlantic salmon with a telemetric tag attached, would change in swimming behaviour from being motionless or slowly drifting along the net wall to swimming consistently in a counter-clockwise direction when the number of fish in the sea cage increased from 30 to 530. When fed with an automatic feeder, which dispensed food in batches over an area 3 m in diameter, every 3 min, one tagged fish was observed to not feed at all in five out of 13 feeding periods. This observation may imply that individuals in sea cages do not feed every time food is offered under the prescribed feeding regime, but this may not always be the case when other feeding methods are used. However, the same authors (Juell and Westerberg, 1993) did conclude, from the mean swimming depths of tagged fish, that certain fish clearly belonged to the deep swimming group of the population. This could mean that fish in a sea cage are comprised of populations that are deep swimming, shallow swimming and perhaps, also, of a whole range of different swimming depths which may or may 165 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES not mix during feeding. Jacobsen et al. (1987) suggested that easily perceived food items should result in higher feeding efficiency, consumption and growth rate of Atlantic salmon. Salmonids are primarily visual feeders (Hoar, 1942; Eriksson, 1973; Higgins and Talbot, 1985; Thorpe et ah, 1988). Visual characteristics of the food that can make it more detectable, such as contrast, colour and size, as well as feed discharge/broadcast method and rate are, therefore, important factors influencing food availability, especially at greater depths where light intensities are low. Lythgoe (1988) stressed that vision underwater is poorer than in air so that the initial detection of objects, even at fairly close range, often requires high contrast sensitivity (Loew and McFarland, 1990). With sufficient light, the most important visual problem of a feeding fish is to detect objects at the greatest possible distance (Douglas and Hawryshyn, 1990) thereby, increasing the reactive distance. Under this condition fish will have more time to proceed through the behavioural components (Stradmeyer, 1989) of the feeding response such as orientation and approach, to increase success rate of capture. Muntz (1990) stated that the detection and recognition of objects under water depends on the rectilinear propagation of light between the stimulus and the background as seen from the observer's viewpoint. An object becomes visible i f the observer is able to detect a difference (contrast of pattern, movement, texture, colour or brightness) between it and the environment (background) for a given angle of view. Salmon feeding activities in a cage are mostly non-uniform, varying both spatially and temporally during a feeding event (Chapter 4; Juell and Westerberg, 1993). It was hypothesised in Chapters 3 and 4 that one of the reasons for this lack of uniform feeding was that fish could not see pellets adequately at lower depths. This lack of pellet detectability might force fish to feed in small groups even under conditions of food indefensibility. Theoretically, 166 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES the rate of feeding per meal should increase i f uniform feeding is achieved. Maximum foraging efficiency and uniform feeding might be accomplished in aquaculture by ensuring accessibility and detectability of pellets to the entire feeding population within one feeding event until all feeding fish are satiated without pellet loss. Unlike conditions in the wild, food wastage is an issue in aquaculture as it affects the environment and farm economics. In this stage of the research, it was hypothesised that contrast-enhanced pellets would permit uniform feeding, and if so, a faster feed dispensation rate. The specific objective was to study the effects of contrast-enhanced pellets on fish feeding patterns (feeding location and feeding rate) in a sea cage. As this is a field study where environmental conditions and some husbandry conditions (ration) were not controllable, they were recorded and related to variation in feeding patterns. 6.3 Materials and methods 6.3.1 Experimental site and design The trial commenced on July 7, 1997 on a commercial fish farm off the east coast of Vancouver Island, B.C. After waiting for approximately 4 mos beyond the anticipated start date, four cages were made available for the experiment instead of the agreed upon six cages. Cages used measured 12 m by 12 m by 20 m deep during slack tide. Fish in the cages had just been counted during a sorting exercise by management. Experimental cages had similar flow and lighting conditions. They were not near any building and all cages had one side exposed to the open sea (Figure 6.1). A l l cages on the site had been installed by the management with tarpaulin skirting from the surface to 12 m depth as a protective measure against prevailing algal blooms from the beginning of summer. Water was pumped up from 18 m depth during 167 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES Figure 6. 1 Layout of cages (measuring 12 m X 12 m X 20 m) used in the experiment. C and T refer to control and treatment cages respectively. algal blooms and when the dissolved oxygen level was low (< 6 ppm). During the same period, transfer cages were also temporarily positioned and moved from around experimental cages to accommodate sorting/grading of remaining cages on the site. Fish at the selected site were Atlantic salmon {Salmo salar L.), of the McConnell strain. Prior to the experimental period, and as set by the management, fish had been fed twice a day starting in the morning at 9:00 A . M . onward and again starting past 5:00 P .M. using a computerised central feeding system (Akva Feeding Systems Inc.). The Akva feeding system dispensed small amounts of feed into a cage at short intervals through plastic tubes in a pre-168 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES programmed sequence during each feeding session. For this experiment fish were conditioned to mechanical feeders (IAS Products Ltd., Series 200 Feed Companion), and to the video feeding techniques of determining feeding endpoints as established in Ang and Petrell (1997), as soon as each cage became available from July 7, 1997 onward. The introductory period varied from 1 wk to 1 mo per cage. One week was shown to be sufficient in our previous trials (Chapter 3). Actual experimental period was between July 26 t h and August 23 r d, 1997. Prior to the experiment, fish were fed silver pellets over a period of at least 3 d. During this 3 d period, silver pellets were mixed with conventional pellets and the percentage of silver pellets in the daily ration was gradually increased with subsequent feeding events. After 3 d, no rejection of food (spitting) was observed. Average fish weights (Table 6.1) were obtained by researchers using underwater stereo cameras and imaging software (FICASS, Petrell et ah, 1997). In the FICASS biomassing technique, stereo cameras were lowered into the water and positioned where fish appeared to be densest. Then approximately 25 min of footage was recorded and used to size 100 fish with imaging software. Only the initial weights of fish were used because of the short-term nature of this experiment. Two cages were stocked at a lower density (L cages) than two other cages (H cages). The cages holding similar densities were adjacent to each other. The control pellets and the contrast-enhanced pellets were fed to fish in both the L and H cages as indicated in Table 6.1. The respective pellet types (contrast-enhanced or conventional) were assigned to experimental cages at random by coin tossing. 169 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES Table 6. 1. Initial fish numbers, mean fish weight and stocking density. Cage Initial fish numbers Initial fish weight, mean ± SD, kg Stocking density, kg m"3 Fish mortality over period, % LC 7912 2.28 1 0.44 6.2 0.18 LT 8724 2.42 1 0.43 7.4 0.09 HC 11012 2.73 ±0.52 10.4 0.45 HT 11478 2.7810.41 11.1 0.03 L C , L T , HC and HT refer to low density control and treatment, and high density control and treatment cages respectively. SD is the standard deviation (n=100) Experimental cages were labelled L C , LT, HC and HT where the letters C and T refer to control and treatment cages, respectively. Fish in control cages were fed conventional pellets while those in treatment cages were fed the contrast-enhanced (silver) pellet. Each feeding event during this experiment was treated as one data point for the purpose of statistical analyses. Fish were not fed during strong current flows (> approximately 15 cm s"1), when net panels were billowed over the camera, high algal counts and low dissolved oxygen (< 6 mg l"1). Each of the four experimental cages had an underwater camera (Fisheye Inc.) installed at 8 m depth facing towards the water surface prior to the experimental period as they had been routinely used by management prior to the experimental period to monitor pellet wastage. Two groups of three researchers per group worked in alternating shifts of 8 d on and 5 d off for the duration of the experiment. To ensure a blind experiment, the feeding station was shielded from the video monitor and recording station (monitoring station) by placing all equipment associated with the latter in a walled shed, which also protected the electronics from the 170 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES weather. The appropriate camera in a feeding cage was connected to the battery and video monitor by the person operating the feeder, prior to each feeding event and without the knowledge of the person manning the viewing station. Feeding events were selected for recording on tape at least once a day for future reference. A l l times associated with the start and stop of feeding, fish vertical migration pattern (including swarming activities) and pellet loss were recorded in accordance with previously established procedures for determining end-points of feeding events (Chapter 3; see also section on data collection in this chapter). 6.3.2 Experimental pellets Both the conventional and silver pellets were prepared at Moore-Clark's manufacturing plant in Vancouver. In order to ensure that conventional and silver feed offered to fish differed only in their visual characteristics (and not nutritionally), feed pellets were prepared in batches such that approximately 50% of each batch would be coated to become the silver feed with the balance becoming conventional feed. The silver pellets used in this experiment were of the same type of contrast-enhanced (silver) pellets described in Chapter 4 but detailed formulation and processing techniques have not been included because the technology is currently patent pending. The feeding regime established by the farmers was followed by feeding fish in the L cages with mixed pellets, thus requiring the production of two sizes of pellets of 6.5 mm and 8.5 mm diameter for both conventional and silver feeds. Fish in L cages were fed one size of pellets (8.5 mm). Silver pellets appeared oilier than conventional pellets after packing and storage despite being manufactured with a similar formulation and batch probably because of the addition of the reflective coating and/or the coating method that was utilised at the mill. 171 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES The effect of the silver coating on pellet sinking speed was tested by dropping conventional and silver pellets (8.5 mm), one at a time, into a large plastic test-tube (2.0 m by 0.16 m diameter) designed for testing the sinking characteristic of pellets (Moore-Clark Vancouver plant). The sinking speed was determined by obtaining the time taken for a pellet to sink 0.43 m after it has reached terminal velocity. The tests were repeated with a new pellet each time until 20 sets of readings each for conventional and silver pellets had been obtained. A two-sample t-test (a=0.05) was used to test for statistical significance between the two sets of readings taken. The depths to which conventional and silver pellets were visible in the water from above water were tested at noon on a sunny day, using 15 mm pellets by slowly lowering a single pellet suspended from a thread. Water transparency was measured using the Secchi disc. The depths at which the pellets faded from view were recorded. The tests were repeated 20 times with a new pellet each time for conventional and silver pellets. Statistical significance between the two sets of readings were tested using the two-sample t-test (a=0.05). 6.3.3 Instrumentation The pellet viewing and recording station consisted of a black and white video monitor, a video cassette recorder (VCR), a digital frame switcher (DFS), a 12 volt deep cycle DC battery, a datalogger, a pair of citizen band transceivers (Radio Shack walkie-talkie, Model TRC-225), air horn and recording sheets. Instructions to start and stop feeding, as well as information on fish behaviour and weather conditions outside the shed were relayed through the walkie-talkie. The air horn was used to alert the feeding station of pellet loss if the walkie-talkie failed. Signal 172 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES and power cables from the four cameras were connected to the DFS in the viewing shed via a series of low loss coaxial/power extension cables laid along the sides of the central walkway. Environmental variables such as light intensity levels in air and at various water depth, dissolved oxygen, salinity, turbidity, temperature and water transparency (Secchi disc) during feeding events were also recorded. Environmental data especially light data were needed because it might be darker than expected at lower levels i f all fish were feeding at the same time, and it is standard practice to record the conditions under which experiments take place. Sensors for recording environmental variables included those for measuring light intensities and for water quality parameters. Light intensities were measured using a terrestrial light sensor (LI-COR Quantum Sensor, Model LI-190SA) placed atop the roof of the viewing shed, and four underwater light sensor (LI-COR Spherical Quantum Sensor, Model LI-193SA) two each in two different cages positioned at approximately 3 m and 8 m depths. Water quality parameters were measured using a turbidity meter (D and A Instrument, Model OBS-1), one current meter (Ocean Inc., Model S4) positioned permanently at the North end of the floating system, and three multi-parameter water quality sensors (one H Y D R O L A B MiniSonde and two HYDRO L A B Datasonde 4 multiprobes). The turbidity meter gave spurious readings during the experiment so information on water visibility relied totally on Secchi disc readings. The MiniSonde multi-probe was suspended at 8 m depth and moved from cage to cage depending on which cage was being fed. The Datasonde 4 multiprobes were positioned, one at the North end and the other one at the middle of the farm. The multiprobes in each of the H Y D R O L A B s were designed to record (and store in memory) up to 15 water quality variables continuously but were programmed to measure only water temperature, dissolved oxygen, salinity and turbidity in this study. A salinometer (YSI Model 33) and one dissolved oxygen meter (YSI 173 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES Model 57) with their sensors positioned at 1 m depth were used to measure salinity and dissolved oxygen levels in water immediately outside the viewing shed. A l l light sensors and the turbidity meter were connected to a datalogger (LI-COR, Model LI-1000) programmed to start registering data 10 min before a feeding event. A set of readings was taken every minute during the feeding event and readings were discontinued 10 min after a feeding event has ended. 6.3.4 Data collection and statistical analyses Data collection Dissolved oxygen varied at the site, and as a corrective measure for when it was low, water was pumped from below each cage by the management in order to increase the level of dissolved oxygen. Water transparency (Secchi disc reading) was measured daily at 2:00 P .M. The times to start and stop feed discharge were recorded on recording sheets containing columns designated for time and the various activities associated with feeding. These included the fish swimming pattern, fish surface feeding activity and the feeding technique employed during feed discharge (e.g. high or low continuous discharge). Researchers feeding fish at the cage side were encouraged to describe fish behaviour at the surface and to draw pictures when necessary while those monitoring the feeding from inside the shed were encouraged to do the same when observing behaviours on the video monitor. This facilitated the recording of unusual or unexpected visual feeding or swimming patterns. Surface activities recorded were in addition to the ones already described in Section 4.3.3 of Chapter 4 (and published in Ang and Petrell, 1997) and included the following: 1. fish eating/mouthing; 2. fish at the surface but not eating; 3. splashing/frenzy; 4. swimming in an organised ring-like structure (clockwise 174 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES or counter-clockwise) or in a disorganised fashion (spiral-like structure). Sub-surface feeding activities (also in addition to section 4.3.3 of Chapter 4) included the following: 1. presence /absence of a feeding swarm (dark mass of feeding fish); 2. fish swimming in an organised or a disorganised fashion; 3. fish swimming upward/downward; 4. fish swimming near the camera and eating/not eating; 5. swarm moving down or swarm disintegrating with fish moving back up to the upper levels and; 6. presence/absence of a discrete group of actively foraging fish which appeared agitated and swimming in vigorous "S"-shaped patterns apparently capturing pellets as they follow the sinking pellets downward. These observations were used to adjust feeding rates and feeding patterns, thereby feeding fish interactively without pellet loss as shown in Table 6.2. Duration of times that fish spent feeding at different levels (depths) in the water column was obtained from data recording sheets. Surface feeding refers to the duration of time that fish spend feeding near the surface. Bottom feeding refers to the duration of time that fish spend feeding near the camera depth. Uniform feeding refers to the duration of time when fish are observed to feed both near the surface (from data recording sheets) and near the cage bottom (camera level). The time to start bottom feed refers to the amount of time that elapsed from the start of feeding before fish begin to move down to feed near the camera level. Frequency of vertical migration refers to the number of times that fish move up and down during the course of a feeding event or meal. This frequency is obtained by counting how often fish (swarm) alternated between feeding near the surface and near the camera depth throughout the feeding event. 175 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES Table 6. 2 Feeding fish interactively using fish feeding behaviours. Fish behaviours observed Interpretations and actions required Swarm at top in an organised fashion Swarm at top in a disorganised spiral-like fashion (frenzy) Swarm moving down but not reaching the camera level Swarm moving down with discrete group of "S"-shaped foraging fish visible on the monitor Fish eating near the camera Discrete group of foraging fish disintegrate with fish moving back up 1. Fish feeding near the surface with pellets spread to sufficiently cover cage surface. 2. Discharge rate is not fast enough to satisfy consumption rate by fish i.e. discharge rate < onsumption rate. Action: Increase discharge rate of pellets. 1. Fish feeding near the surface with pellets discharged in a small area not sufficiently cover cage surface (point source of food). 2. Discharge rate < consumption rate. Action: Increase spread of pellets and increase discharge rate of pellets. 1. Fish feeding deeper in the water column. 2. Discharge rate approximately matched by pellet consumption rate. Action: Maintain the current discharge rate. 1. Fish feeding deeper in the water column. 2. Discharge rate > consumption rate. Action: Reduce discharge rate or stop discharge temporarily. Fish feeding near the bottom of the cage. Action: Discontinue discharge immediately if not already discontinued because pellet loss will follow shortly if discharge is not discontinued. Discharge rate < consumption rate. Action: Resume feeding at a rate less than the rate just prior to swarm moving down. The times during which pellets were discharged were obtained by adding the time intervals between each start and stop times during a feeding event and the total figure is referred to as the total feeding time. The keywords/phrases from Table 6.2 above were also used to determine duration of surface feeding, bottom feeding for fish and subsequently the duration of uniform feeding, when fish were observed to feed near the surface and at depths simultaneously. The times it took for the fish (swarm) to start moving down while feeding (each day) were similarly obtained from the recorded data. 176 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES Statistical analyses. Data from fifteen experimental days and four cages were available. This is important, as fish feeding behaviour vary daily. In order for the enhanced pellet to be accepted by industry, it would have to be effective under a wide range of environmental and husbandry conditions. The experiment was, somewhat, constrained by the unexpected experimental condition given by the farm management (i.e. four cages and two stocking densities). As the effect of stocking density on my experimental variables was not known, experimental variable from the two lower stocking density cages were tested against data from the two higher stocking density cages (one-factor nested A N O V A , factor: cage density, cages are nested). It was determined that stocking density had an effect during the morning feeding events (see results), therefore, the two density levels were treated as two separate experiments in subsequent statistical tests. As well, the time of day was, treated as a factor in the statistical test because the effect of time of day (morning and evening feeding) on the experimental variable was not known. The effect of pellet type on the fraction of time spent by fish feeding uniformly was initially analysed by inspecting the histograms. Data were highly skewed toward the left because there were many zero values. As recommended by Sachs (1984), the data were, therefore, transformed using the Log-transformation (X' = Log (X+l), where X is a data point). After transformation, the data sets were still not normal (Kolmogorov-Smirnov test). No further statistical tests were done, as it was obvious that there were no effect of pellet type on occurrence of uniform feeding (see results). To look for an explanation for the general lack of uniform feeding throughout the experimental period, the following tests were conducted. One, the effect of pellet type on the 177 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES fraction of time spent feeding near the cage bottom was tested using a two-sample t test (two-tailed) on the transformed data (transformed by log-normal as the data sets were not normal). This fraction does not imply that i f fish are not feeding near the bottom, they are feeding on the surface, therefore, it is not considered of binomial distribution. The Log-transformation was used as described above to normalise the data set as required by parametric tests. If the data set did not meet the conditions of normality and equal variance, then the Mann-Whitney Rank Sum test (non-parametric) was used. Additionally, ration amounts (kg kg-fish"1) and feeding rate (kg MT-fish"1 min"1) were tested using single-factor nested A N O V A (pellet type as a factor) and the two-sample t tests, respectively. Data for feeding rate was also transformed. As in Chapter 3, a significance level of a = 0.10 was used for the same reason as stated. Graphs were plotted using MS Excel (Microsoft Corp., 1997) to look for obvious outlier and any trend that may indicate a confounding relationship between the parameter of interest and environmental variables such as light intensity (Table 6.3). A data point was discarded as an outlier i f it was outside the region "mean + 4 SD", where the mean and standard deviation (SD) were computed without the value being suspected of being an outlier (Sachs, 1984). Parameters that did not have an obvious trend as described above were analysed for statistical significance using M S Excel (Microsoft Corp.) and SigmaStat(Jandel Scientific). 178 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES Table 6. 3 Graphs plotted for highlighting outlier and trends with environmental conditions. Graph plotted Graph type Outcome Ration on consecutive days for morning and evening feeding events (low-density and high-density cages) Feeding times on consecutive days for morning and evening feeding events (low-density and high-density cages) Feeding rate on consecutive days for morning and evening feeding events (low-density and high-density cages) Ration per meal and daily ration versus Secchi Disc Reading (water transparency) for morning and evening feeding events (low-density and high-density cages) Ration per meal versus feed rate for low-density and high-density cages Total feeding time versus Secchi Disc Reading (water transparency) for low-density and high density cages Total feeding time versus feed rate for low-density and high-density cages Feed rate versus Secchi Disc Reading (water transparency) for morning and evening feeding events (low-density and high-density cages) Feeding rate versus light intensities for morning and evening feeding events (low-density and high-density cages) Bar graph (graphs not shown) Bar graph (graphs not shown) Bar graph (graphs not shown) Scatter plot Scatter plot Scatter plot Scatter plot Scatter plot Scatter plot Two qualified outliers found: one feeding time for HC cage (evening feeding) and one feeding time for HT cage (evening feeding) No outlier No outlier No apparent trend No apparent trend No apparent trend Yes, trends apparent No apparent trend Apparent trend found in: morning events before feeding for high-density control cage (HC) with light at 3 m but not for treatment cage (HT) and; morning events during feeding for both and high-density control (HC) and treatment (HT) cages with light at 8 m. No apparent trend for evening feeding events 179 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES 6 A Results Uniform feeding was not achieved by using contrast-enhanced pellets. Basically, the fish tended to avoid light and to feed in groups. Details follow. 6.4.1 Pellet characteristics and environmental readings The mean sinking speeds (± SD) for conventional and silver pellets were 0.128 + 0.006 and 0.127 ± 0.006 m s"1 respectively. There was no significant difference in sinking speed between conventional and silver pellets. At noon, on a sunny day (light level not measured, Secchi disc reading 9 m), the mean visible depths (± SD) before pellets faded from view were significantly different (two sample t, p < 0.05) for conventional and silver pellets and were 2.18 ± 0.08 and 3.08 ± 0.08 m, respectively. This meant that silver pellets could be seen 41% deeper in the water column than conventional pellets when viewed from above water. Water temperatures varied from 10 °C to 14 °C during the experiment with an average of 12 °C. Dissolved oxygen fluctuated between 5.5 mg l " 1 and 12.0 mg l " 1 . Salinity ranged from 26 ppt to 32.7 ppt Light intensity in air varied from 194.5 to 1,662.5 //moi photon s"1 m"2 during the morning feeding 1 2 and from 30.7 to 891.0 //moi photon s" m" during the evening feeding. Light intensities underwater (at approximately 3 m) ranged from 13.4 to 1252.0 //moi photon s"1 m"2 in the morning feed and from 1.4 to 320.1 //moi photon s"1 m"2 during the evening feeding. Light intensities underwater were much lower at approximately 8 m and ranged from 0.9 to 589.8 1 9 1 9 //moi photon s" m" during the morning feeding and from 0.02 to 33.2 //moi photon s m during the evening feeding. Averages of readings for these same environmental variables are included in Table 6.4). Light intensity levels in air were measured continuously throughout 180 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES each feeding event. Water transparency (Secchi disc reading, m) or visibility measured at 2:00 P .M. each day, ranged from 2.0 m to 9.0 m but was generally around 3.5 m. Low readings were probably due to algae and not silt since rainfall was negligible both prior to and during the experimental period. The latter also coincided with historical periods of algal blooms in the area (enhanced by the long daylight hours and high water temperatures). Daily average current speeds during the experimental period were between 0.07 and 0.10 m s"1. The maximum and minimum current speeds were 20 cm s"1 and 0 cm s"1. Table 6. 4. Environmental variables (means ± SD). Water 1 temperature, °C Dissolved 1 oxygen, mg/1 Salinity, 1 ppt Light in air, 2 //moi photon s-1 m"2 Light at 3 m, 2 //moi photon s" Light at 8 m, 2 //moi photon s' AM FEEDING: LC n=21 12.2 ± 1.6 n=21 7.4 ± 1.7 n=21 28.3 ± 1.7 n=9 to 15 1310 ±160 n=9 to 15 2 3 0 ± 2 1 0 n=9 to 15 200 ±260 LT 12.3 ± 1.6 8.1 ±2 .2 29.5 ±5 .4 1230±160 250 ±130 20 ± 14 HC 12.6+1.6 7.8 ± 1.8 31.1 ±3 .0 940 ±470 350 ±410 23 ± 3 3 HT 12.1 ± 1.7 7.8 ± 1.7 23.3 ± 13.2 980 ±470 150±130 5.0 ±5.0 PM FEEDING: LC n=21 13.2 + 2.2 «=27 7.7 ±2.3 n=21 29.6 ±3.6 n=3 to 21 80 ±9 .0 n=3 to 21 Sensor down N=3 to 21 27 ±5 .0 LT 12.7 ±2 .3 7.9 ± 1.7 31.1 ±3 .7 58 ±9 .0 14 ±3 .0 2.0 ± 1.0 HC 13.6 ±2 .9 7.9 ± 1.5 32.7 ±3 .0 300 ±300 100±120 11 ±9 .0 HT 11.7 ±1.1 6.7 ± 1.3 30.3 ±3.3 230 ±260 32 ± 3 6 1.0 ±2.0 1 Measured at 3 m depths 2 Light intensity levels averaged all data points (at beginning, during and end of feeding) over experimental period. 'Sensor down' means equipment malfunction. LC, LT, HC and HT refers to low density control cage, low density treatment cage, high density control cage and high density treatment cage respectively. Daily light levels are available from Canero (1999). 181 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES 6.4.2 Observations on fish behaviour Fish behaviour changed a great deal during feeding but there was no change in overall patterns over time (successive feeding events) during the experimental period. Before feeding, fish tended to swim near the sides of the cage in a ring-like structure leaving a hole in the centre thus giving a cylindrical shaped formation (Figure 6.2). As pellets entered the water at the beginning of feeding, some fish reacted by filling up the centre hole from the sides while more fish moved up from below the camera level (Figure 6.3, and Figure 6.4). Many fish remained near the cage bottom. A disorganised feeding pattern on the surface could be seen forming in what was described in the previous section as a swarm. The swimming pattern that resulted was dependent on the spread of the pellets (2-dimensional aspect) while the feeding depths (fish moving down/up) were dependent on the discharge rate (3-dimensional aspect). As discharge rate increased, the feeding fish responded by moving down (swarm down) where a discrete group of foraging fish formed with individual fish exhibiting the "S"-shaped foraging pattern (Figure 6.5). When discharge rate was not reduced or temporarily stopped in time, the foraging fish eventually reached the camera level where some fish could be seen to capture pellets near the camera (Figure 6.6). Pellet loss was inevitable because pellets became inaccessible as they left the cage bottom. Fish then moved back up to the surface when pellets were depleted (either consumed or lost) near the cage bottom (Figure 6.7). This upward movement was also in response to pellet availability at the surface again as feed discharge was resumed at a lower rate than before. Fish quickly formed a swarm at the top again as they resumed feeding (Figure 6.8). This action and reaction were repeated until the endpoint was reached. Throughout, it was clear that there were feeding and non-feeding groups of fish. The endpoint was reached when pellet loss was inevitable even when feed was discharged at a very 182 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES low rate. This rate was arbitrarily fixed as the rate at which it was deemed to be not worthwhile to continue feeding and was approximately less than 0.5 kg per 15 sec for this experiment or approximately 1/10th of the peak discharge rate. Figure 6. 2 Before feeding, fish swim near the sides of the cage in a ring-like structure leaving a hole in the centre. 183 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES Figure 6. 3 When feeding started, fish began to fill the centre hole to feed as more fish swim upward to join the top fish. Figure 6. 4 A swarm is formed at the top at the beginning of a feeding event. 184 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES Figure 6. 5 The fish swarm moved down and a discrete group o f foraging fish with individual fish exhibiting the "S"-shaped foraging pattern became visible as the discharge rate was increased. Figure 6. 6 Fish feeding near the camera as they follow the pellets downward. 185 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES 6.4.3 Statistical analyses Results of statistical analyses are summarised in Table 6.5 below: Table 6. 5 Summary of outcome of statistical analyses (two-tailed) conducted for various data sets. Parameters analysed Transform Normality Equal variance Significance test test Tests on ration (kg kg-fish'1) A M feeding: High density versus low density cages PM feeding: High density versus low density cages AM: Control versus treatment PM: Control versus treatment AM versus PM Tests on feeding rate (kg MT-fish'1 min1) A M feeding: High density versus low density cages PM feeding: High density versus low density cages AM: Control versus treatment PM: Control versus treatment AM versus PM Tests on Percent bottom time All factors had no significant difference No No No No No Passed Passed Passed Passed Passed Passed Passed Passed Passed Passed Yes (0.05) Yes (0.05) No (0.10) No (0.10) Yes (0.05) Yes Yes Yes Yes No Passed Passed Passed Passed Passed Passed Passed Passed Passed Passed Yes (0.05) No (0.05) Yes (0.10) Yes (0.10) Yes (0.05) 6.4.4 Uniform feeding Table 6.6 shows the mean times of surface, bottom and uniform feeding, expressed as percentages of total feeding time. Log-transformed data on uniform feeding were also not normalised. Histograms of uniform feeding (transformed data) are shown in Figure 6.9 (morning feeding events) and Figure 6.10 (evening feeding events). As mentioned previously, all histograms were skewed to the left due to many zero values. The silver pellets did not affect this pattern as the main activity of the feeding fish was to move up and down in groups so that 187 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES they were either near the surface or near the cage bottom. Further, there were no significant differences between pellet types for the fraction of time spent feeding near the cage bottom. Table 6. 6. Duration of times that fish spent feeding at different levels (depths) in the water column, expressed as a percentage of total feeding time for morning (AM) and evening (PM) feeding events (means ± SD). Cage Surface 1 feeding, % Bottom 2 feeding, % Uniform 3 feeding, % Time to start4 bottom feed, % Frequency of5 vertical migration AM FEEDING: LC (n=18) 52 ± 35 58 ± 2 6 34 ± 3 9 24 ± 2 6 1.7 ± 1.0 LT (n=18) 57 ± 3 3 63 ± 2 2 36 ± 3 3 22 ± 15 1.7±0.8 HC (n=19) 50 ± 3 3 68 ± 3 0 32 ± 3 2 21 ± 3 0 1.5 ± 0.8 HT (n=19) 52 ± 3 3 68 ± 2 7 33+33 29 ± 3 2 1.5 ±0.7 PM FEEDING: LC (n=18) 39 ± 4 0 69 ± 2 8 30 ± 3 7 29 ± 4 3 1.2 ±0 .4 LT (n=18) 47 ± 3 9 64 ± 2 5 30 ± 3 6 35 ± 3 2 1.5 ±0 .6 HC(n=16) 41 ± 3 8 60 ± 3 0 23 ± 2 8 22 ± 2 2 1.5 ± 1.0 HT (n=17) 42 ± 4 0 75 ± 2 5 26 ± 3 6 22 ± 3 6 1.2 ±0 .6 1 Surface feeding refers to the duration of time that fish spend feeding near the surface (swarm at top). 2 Bottom feeding refers to the duration of time that fish spend feeding near the cage bottom (or camera level). 3 Uniform feeding refers to the duration of time when fish could be observed to feed both near the surface (from data recording sheets) and near the cage bottom (from camera observation). 4 Time to start bottom feed refers to the amount of time that elapsed from the start of feeding before fish begin to move down toward the camera level. 5 Frequency of vertical migration refers to the number of times that fish move up and down during the course of a feeding event or meal. Data were obtained by playing back tapes of recorded feeding events and matching fish movements (migration) to the corresponding times (clock) displayed on the monitor screen. L C , LT, HC and HT refer to low density control cage, low density treatment cage, high density control cage and high density treatment cage respectively. 188 % — c fe I | ! a, I I to ! o o o o o r - ^ v o i y ^ T t m r s — o H u u Mor 0.35 1 ro j •o fN © 1 fN B o 5 0.15 o 1— o o 1 1 1 1 1 1 1 1 1 1  1 1 1 1 1 o o r ^ ^ o i n ^ - m f N " — i c \.>ii.)nl),>.i,| 3 sa m <N —< o U U H X OiJ rt U u X gp o •a M <u .fl ?| $ 2 c-i—i a 1 a | l | | -5 8 OX) .5 I? <D fl fl +^ fl V> o <£» fl H H PH fl - f l o fl | K > •4-* Ti I—I ° J o U OH 55 •• v) fl *—1 r>-l - L U "3 rt rS u 'g ( s o | . f l O r t fl 11 >, o ^ fl - f l T3 « s G 00 .2 -fl ON 05 • Si p rt V3 ON 00 Q -fl ?0 te to fe H CU CO rt U H h I H- — I • S cc m rs| < H cu Ml rt O u u co rt u £ -fl cu fl S cu rt , < D CD ^ fl .fl £ fl 'c£ cu fl > cu CD T3 I I fl^ •e § oo o .S >, T3 CU on C cu O CO <+H cu H o cu o c cu o CU o Jfl u ^ -O f_ CU I—I fl . cu ^ * + 1 >< .3 oo on O I I  -fl w on ^ on ( D § < H . O ( 3 0 < + H o a on rt cu > ' f l o cu fl. on CU on' CU CO rt cu -*-» C cu (U on C cu -a .fl 00 T3 fl rt sd -v V O cu <D >&H 3 rt bp 15 Q fl O CU on C cu 13 O ON Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES 6.4.5 Feed ration (kg kg-fish"1) The feed ration (kg kg-fish"1) was statistically significant for some of the factors but can not be considered practically significant because the difference would represent in the worst case scenario only 19 g of food per fish per meal (Table 6.7). Rration per meal and daily rations were not affected by water visibility (Secchi Disc Reading, m) for either of the low density cages (LC and LT) nor the high density cages (HC and HT) (Data not shown). Ration per meal was also not influenced by feed rate. Table 6. 7. Mean and range of rations fed, in kg kg-fish"1 during morning (AM) and evening (PM) feeding events. Cage Ration, k g kg-f ish" 1 Means ± S D Ration, k g kg-f ish" 1 M i n i m u m Ration, k g kg-f ish" 1 M a x i m u m AM FEEDING: L C (n=15) 0.0058 ± 0 . 0 0 1 8 " 0.0027 0.0094 L T (n=15) 0.0051 ± 0 . 0 0 1 5 b 0.0028 0.0080 H C (n=15) 0.0067 ± 0 . 0 0 1 5 0.0041 0.0093 HT(n=15) 0.0066 ± 0 . 0 0 2 1 0.0034 0.0110 PM FEEDING: L C (n=15) 0.0050 ± 0 . 0 0 1 4 " 0.0023 0.0074 L T (n=15) 0.0040 ± 0 . 0 0 1 4 b 0.0016 0.0070 H C (n=15) 0.0058 ± 0 . 0 0 1 5 " 0.0026 0.0080 H T (n=15) 0.0055 ± 0 . 0 0 1 4 b 0.0029 0.0081 a ' b Means with different letters are statistically different (p < 0.10). LC, LT, HC and HT refer to low density control cage, low density treatment cage, high density control cage and high density treatment cage, respectively. 191 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES 6.4.6 Feed rates The mean feeding rates were significantly higher in control than treatment cages for both stocking densities and for both morning and evening events (Table 6.8). However, the mean feeding rate of the high density control cage was higher than the mean feeding rate for the treatment cage in the evening feed at the a=0.1 level. Mean feeding rate for the low density control cage was statistically different from treatment cages for both morning and evening events. The unit of kg feed MT-fish"1 min"1 was used also because it is a common unit employed in industrial field trials. Feed rate was expressed as kg min"1 to show the actual rate of discharge of feed into the cages that fish were able to consume during each feeding event (Table 6.8). Figure 6.11 and Figure 6.12 shows the power (negative) relationship between total feeding time (min) and feeding rate (kg feed MT-fish"1 min"1) for low density and high density cages respectively. The vertical solid lines in both graphs indicate a feed rate that is invariant to feeding times. Under this scenario, the total feeding time increases as the number of fish participating in feeding increased. The horizontal solid lines on the other hand indicate that the total feeding time will remain the same while feeding rate increases in response to an increase in the number of fish participating in feeding. The goal of uniform feeding is expected to be achievable with feeding rates to the right hand end of this solid horizontal line. Feeding rate was not affected by water visibility (Figure 6.13). 192 I ^  t3 0 S H 2 5- - ro SL ffi 3 ' § I a. & X a. ft) - i 2 3 " r l f-t- _ O <T> o a> a. P p « Op' 3 O Si 3 O o S o ^ p el-'s s , , CD q CD * a a. / - - v g ^ w A <3 P ro 3 o p OP ft> rjq 3" D-ft) 3 o o 3 o p OP rt p 3 CL OP 3* a. fo 3 ft) 91 ft> a o p OP H o to 4i. X O o 4^ b o H to to 4^ to r H r 5 X H o '5s II II o o o p OS 4^ to to U J to o to p U J o b o p b o OS -O OS t o Os r o to 4^ ON to >— 4=> - J p o o © O O O O ^-J OS b o 4*. Ln Ln b o 1+ 1+ 1+ 1+ 1+ 1+ 1+ 1+ o o o o O O O o to U J 4^  to io to O o- o- o a" CT" o 4^ bo U J to Os o to 4*. to ui O U J Ln bs ON bo Os Ui l/i 1+ 1+ 1+ 1+ 1+ 1+ 1+ 1+ 4^ Ln 4*. Ui 4 -^ 4*. O o 4^ U J 4^ to 4^ os 4^ to U J O p 00 ro -fl 2 H a. q w S-1+ ° s; S> i. CO CL 2 5' 3 e 3 TI ^- co ft* ro H a. 3i 3 J~ ^ 3 OQ 5' CO1 i . ro ro ft* ro £ ^ °-3i in 3" (JQ — ro CL ?3 ro o 1+ cro o o 3 a s. 3 ?0 P I g 3 ?0 2 5-B ? | 0,3 I g. 3 H ft p r i - ft P ^ ft 0 0 cr 1- 2 3 ft OP P Cu 3 3" O. p 1 - 1 p OP 3 8 OP CL ft) 5' Q OP ft < ft) 3 T I ft ft CL & ft 3" ft ft O 3" ft ft CL 5 O s I Co 1 I Co I I 1 ffl Co o Co Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES 0 -I , , , , 1 , , 1 1 1 1 1 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.30 Feeding rate, kgfeed/MTfish/min s LC A LT Figure 6. 11 Graph illustrating the reduction in total feeding time with increase in feeding rate (n=36; low density cages). Solid horizontal and vertical lines indicate the invariant feeding time and invariant feeding rate, respectively. LC and LT refer to the low density control cage and the low density treatment cage respectively. 0 HC A HT Figure 6. 12 Graph illustrating the reduction in total feeding time with increase in feeding rate (n=36; high density cages). Solid horizontal and vertical lines indicate the invariant feeding time and invariant feeding rate, respectively. HC and HT refer to the high density control cage and the high density treatment cage respectively. 194 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES | CO 7 3 O d> ra u-7 3 CU CU LL 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 I • I ii is s 3 4 5 6 7 S e c c h i Disc R e a d i n g 10 • LC LT HC HT Figure 6. 13 Graph illustrating that feeding rate was not affected by water visibility (Secchi Disc Reading, m) in any experimental cage. 195 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES Figure 6.14 shows the relationships between feeding rate (kg feed MT-fish"1 min"1) and underwater light intensities (umol photon s"1 m"1) before feeding, at approximately 3 m (Figure 6.14 a) and 8 m (Figure 6.14 b) for the morning feeding events, and at approximately 3 m (Figure 6.14 c) and 8 m (Figure 6.14 d) for the evening feeding events (high density cages). Figure 6.15 shows similar relationships between feeding rate (kg feed MT-fish"1 min"1) and underwater light intensity levels (umol photon s"1 m"1) during feeding, at approximately 3 m (Figure 6.15a) and 8 m (Figure 6.15b) for the morning feeding events and at 3 m (Figure 6.15c) and 8 m (Figure 6.15d) for the evening feeding events (high density cages). The power (negative) relationship is very strong (R > 0.95) in the morning feeding events but non-existent in the evening feeding events both before (Figure 6.14) and during feeding (Figure 6.15). The trend lines shown in these two figures when they prevail should not be treated as extrapolative or absolute because of the approximate depths of the light sensors. No graph was plotted for low density cages because of insufficient data points as a result of the malfunctioning of one of the sensors. Figure 6.16 shows graphs of feed rate (kg feed MT-fish"1 min"1) for successive feeding days for low and high density cages (a and b respectively) and similarly during feeding (c and d respectively). This indicates that experimental fish did not change in their overall behavioural patterns and probably did not learn or unlearn previously acquired feeding patterns e.g. the lack of organised ring-like feeding structures associated with or expected from a continuous broadcast method of feeding. 196 CD> > O 3 OQ' cr o> _ 3 ^ CD c/3 CJ CD S" s & =^ a 3 O g OQ 2 3 - i S &w Cf „ CD a. o CD cn 2 3 " o p CTQ on — D. cr 3 3. 3» 3 CD < CD 3 5' CTQ c 3 D-CD CD P 5' Z CD 3" < rt CD 3 3 cn a S. §. 3 5" ^ % o 2. s —' cn P ^ 3 *P °- I oo g 3 ^* ^ !=" D. O <-f -< cn 3" cfp' 3 3" a. ^ ^ CD ,£ CTQ CD CD rv cn fcrt • 3 a 3 0 o a P 3 H p Feed rate Feed rate o o o p o o o o o o o o o o o o o o o o C O 3 • d 3 3 3 co o o o o o p p o p o o o o o o o ^ - » M M C o u i i j i c n b i b ) o c n o c n o c n o u i o c n o • IE O c o 3 • d CQ 3" CD 3 0) o CO o o 4^ o c n o o o 8 i Feed rate Feed rate o Ko o o o o in b) N o o o p p o p o o o o o o o • I o 00 3 o X d 2, oi CD 3 CO O • I O co 3 • co CD 3 (A ro o CO o 4v. o OI o o o o S I Co 1 1 t-< Co I f 1! o Co VO oo VO VO ct> ^ rt h-h O w cv -a 3? rt rt (1 O o P fl a . < o 3 2-3 p ^ w — s- °- s g 2 a . s. 8 js. «3 5T 3 2 — sT S 3 2 . g o 5 OP "I CTQ -P ST 5 rt i - S- tc O rrt a . a . « ft OQ c O N rt rt s* 5' OQ rt OQ rt rt a . , , 3 i i*> g P o sr o - 3. a . 3 , rt rt £. P 3^ „ 33 r5 rt rt OQ o 3 rt <-f rt c - < ' 3 rt <5 v o p OQ P OQ' CT 3" rt „ 3T O - p rt < S 5" o o 3 T3 rt o a . p c OQ -« rt 3j -p OQ °- ST 3 " ft ^ rt ^ 3-^ rt 2 . 3 ^ "3 rt 2 S. 3 o 3 a . rt • o P p I— OQ H rt H •3 O * i 2. ffi <: rt >-( c c/i C o o rt < rt rt rt a . OQ P a* rt rt rt £* 3' OQ P 3 Q. 3 " OQ' 3 " a . rt 3 o p OQ rt P 3 a . Feeding rate o o Feeding rate o o 8 8 £ 8 § S r o • • • • • * • » • • • • • • • • • « • • » • « • • • • • « • Feeding rate Feeding rate o • * e X a x a • « • « • • • • • « • • • • 5 1 o o s s Co 1 I t -Co 1 s Co fe Co Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES 6.5 Discussion Uniform feeding was not achieved using contrast-enhanced pellets because fish preferred to feed near the cage bottom and feeding rate was directly affected by light intensity. These findings are contrary to prior studies. As well, the description on fish movements that were recorded here add to the literature on the usage of a cage by fish. The size of the feeding volume affects the feeding rate, but did not promote uniform feeding. 6.5.1 Trends There was no apparent association between amount of ration fed or feeding time and water visibility, but feeding rate was influenced by underwater light intensity. As well, there was a clear association between total feeding time and feeding rate in both the low and high density cages (Figures 6.11 and 6.12, respectively). Uniform feeding is theoretically achievable and lies to the right-hand-end of the solid horizontal line shown in Figures 6.11 and 6.12. In Chapter 3, during an earlier trial there was an association between feeding rate and water visibility (Secchi Disc Reading). The difference between the previous and the current experiment is water quality. In the previous experiment, water became silty under low light intensities while in this experiment, low water visibility was associated with algal laden water (1-2 m) under high light intensities. The power law relationship of feeding rate and underwater light intensity is indicative of the fish preference to feed lower in the water column in order to avoid light, which is similar to observations on light avoidance by Ferno (1995). The lack of trends (feeding rate and underwater light intensity) during the evening feeding events probably resulted from the 200 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES confounding influence of the hills close to the experimental site, which tended to shade the cages during feeding. Ferno (1995) discussed light avoidance in Atlantic salmon but did not discuss the use of graphs that display the relationship between feeding rate and underwater light intensities as provided in this Chapter. Contrary to earlier work (e.g. Blaxter, 1970) which reported feeding rate to progressively fall during dusk or at artificially reduced illumination, this study found feed rate to increase with lower light intensities. One possible explanation for this difference could be that light levels in Blaxter's experiments were readings in air and not underwater readings which are the case in this study and therefore, underwater light intensities at dusk would be much lower than those recorded in this experiment. Wild salmon are found in many strata with no clear understanding of preferences as yet (Groot and Margolis, 1991). After comparing the associations between light and feeding rate, and feeding time and feeding rate, uniform feeding apparently is associated only with low light conditions while the lowest feeding rates were associated with high light intensity. 6.5.2 Feeding patterns In between feeding events, fish at this experimental site swam in a clockwise or anti-clockwise direction (direction was cage specific) when swimming near the surface but seldom fed in this same organised ring-like structures. This lack of feeding pattern may be a result of fish having been conditioned to the feed discharge pattern typical of the Akva feeding system, which dispenses feed to each cage in small quantities, spread over a fixed area of the water surface at pre-programmed intervals. Continuously feeding the experimental fish for over 2 months using mechanical feed blowers (IAS Products Ltd.) and spreading pellets over a wide area of the 201 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CAGES water surface as established in previous feeding trials in this study did not increase the incidence of organised swimming pattern during feeding. An organised structure during feeding permits more fish into a given volume when swimming in an orderly "school-like" formation, allowing more pellets to be discharged per unit time without pellet loss (as opposed to a spiral-like feeding pattern (see Chapter 3). It also eliminates the need for fish to forage or scramble, as in a feeding frenzy, in order to obtain food. 6.5.3 Feed ration amounts Feed tables (Moore-Clark Nutreco) designed using predictive growth models for known species are sometimes used as a guideline for farmers to determine the amount of daily ration to be fed based on water temperature and fish biomass. In this study, the ration for the experimental fish at water temperatures around 12 °C, according to feed tables from Moore-Clark should be about 0.95 % and 0.88 % body weight for L and H cages respectively. These figures indicate that fish fed neither conventional nor silver pellets were fed below recommended ration levels. Consequently, it can be assumed that fish fed silver pellets were not compromised in terms of food consumption during the experiment. Average amounts of daily ration fed in all experimental cages during the experiment were close to or higher than recommended ration from feed tables supplied by salmonid feed manufacturers or levels practised by fish farms in the industry today. It is highly probable that the experimental ration levels reached during this feed trial are attributable to the use of the camera feeding method established in earlier feed trials (Chapter 3). It also ensured that fish fed either conventional or silver pellets were not compromised in terms of food consumption and/or growth. There was significant difference between morning and evening feeding rations for both L and H cages (Table 6.4). Fish ate 202 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES more in the morning than in the evening for all experimental cages, perhaps as a direct result of entrained feeding pattern established by the feeding regime prior to the experiment and/or light intensity. Rations between control and treatment cages, however, were not statistically significant. 6.5.4 Feeding rate In both L and H cages, fish fed faster overall on regular brown pellets than on treatment pellets. This was contrary to the hypothesis that fish would feed faster i f pellets offered were more visible or detectable. Silver pellets were indeed more easily detected by fish than conventional pellets (confirmed in laboratory trials, previous section). Silver pellets also remain more visible than conventional pellets deeper in the water column where light intensities are greatly reduced. Fish in the LT and HT cages have a choice of where to feed when offered silver pellets that were more visible than conventional pellets, and essentially tend to feed deeper in the water. This meant that pellets took longer to reach the fish, thus increasing the total feeding time and subsequently feeding rate. The preference of fish fed silver pellets to feed deeper in the water column resulted in them feeding nearer to the camera. The close proximity of the feeding fish to the camera combined with the feeding method of reducing feed discharge rate when fish are down by the camera may have further reduced the feeding rates in the treatment cages. Hence, vertically increasing the feeding volume by increasing pellet detectability or stocking density increased the amount of time required for pellets to sink through the feeding volume, thus, increasing the feeding time instead of decreasing it as was initially expected. Apparently, the feeding patterns exhibited by the fish did not permit an increase in feeding rate to compensate for a larger feeding volume. 203 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PATTERNS IN SEA CAGES 6.5.5 Surface, bottom and uniform feeding times There were no statistically significant differences between control and treatment cages for time spent by fish feeding near the surface (surface feeding), deeper in the water column (bottom feeding) and simultaneous feeding (uniform feeding). This lack of significant differences suggests that the silver pellets did not encourage fish to feed more uniformly and/or simultaneously. Other factors probably have a greater influence on feeding patterns and these could include behavioural entrainment and feeding method. In this experimental scenario, fish were offered a choice of where to feed when silver pellets were offered and they chose to feed near the cage bottom. However, they were prevented from doing so as previously stated, by the researchers using the prescribed feeding technique that was used, which terminated feed discharge when fish were feeding near the camera level. Fish may have been instrumentally conditioned to feed in groups, thus preventing uniform feeding. Entrainment influence is associated with dominance-subordinate relationships within a cage population where dominant fish continuously maintain their feeding depth in relation to subordinate fish such that no fish is able to move past the dominant fish to feed at the top when the dominant fish move down to feed. The fish farming industry is not considering these feeding patterns while feeding fish for various reasons. 1. They (the farmer) feed fish only at the surface even though fish prefer to feed deep in the water column and 2. The farmer encourages fish to compete near the surface thereby creating group feeding activities. It was surmised that the size of the feeding volume affects feeding rate but does not promote uniform feeding. Uniform feeding will not be achieved unless the observed feeding 204 Chapter 6. EFFECT OF CONTRAST-ENHANCED PELLETS ON SALMONID FEEDING PA TTERNS IN SEA CA GES patterns, which could have formed as a result of environmental factors, prior conditioning and genetics, can be modified. 6.6 Conclusion and recommendation The hypothesis that contrast-enhanced pellets would permit uniform feeding was not met in this experiment. The specific objective of studying the effects of contrast-enhanced pellets on fish feeding patterns and extension of the feeding volume in a sea cage was achieved as valuable information on the feeding patterns of fish in a sea cage was obtained. One of the observations noted in this part of the research was that fish preferred to feed lower in the water column when given a choice of where to feed. Other possible interpretations for the longer time required to feed silver pellets were that 1. as fish move down to feed, pellets take longer to reach the feed because no fish moved up to repopulate the upper feeding zone and to intercept pellets in that zone. As a result researchers had to continuously reduce or temporarily stop feeding to prevent pellet loss, 2. The possibility that fish needed time to get used to a novel diet (silver pellets) from the conventional brown colour of the food they were previously fed may explain some of the reduction in rate of feeding in treatment cages. This experiment was not, however, designed to study the latter possibility. Studies on roles of chemical stimuli on feeding in fish suggest that feeding history (Mearns, 1985; McBride et. al., 1962) and learning may influence the food type most acceptable to fish. Feeding fish the silver pellet much earlier in their life stage (e.g. to smolts at seawater transfer or even during the hatchery stage) may reduce both the effect of dominance-subordinate hierarchy and the novelty of a new pellet type. 205 Chapter 7. OVERALL CONCLUSIONS AND FUTURE CONSIDERATIONS Chapter 7 OVERALL CONCLUSIONS AND FUTURE CONSIDERATIONS 7.1 Conclusions on overall research The research aim of addressing the need for improving FCR and fish growth in the salmon farming industry was achieved in the study as reported in Chapter 3. FCR (biological), growth and mortality rates were greatly improved through the use of a newly developed feeding method that utilised underwater cameras to view feeding behaviour and pellet loss when monitoring feeding rather than utilising surface feeding activities. The concept of extending the feeding volume was established with visualisation and interpretation of research results (Chapters 3 and 6). Detailed feeding patterns associated with different feeding methods were documented, and the process of feeding fish was refined with the use of feeding behaviours as cues to interactively feed fish. Specifically, two distinct types of feeding formations were observed. One was associated with continuous feeding, and it consisted of fish circulating around the cage in an organised way. The other form of feeding formation was the spiral vortex, which was associated with batch feeding. Both types of structures tended to break down over the feeding event. The foraging pattern by individual fish was an 'S'-shaped searching manoeuvre. The amplitude of a swerve was greater in batch feeding than in continuous feeding. It also became evident that feeding homogeneity throughout the cage was uncommon in the fish farming industry of British Columbia. One of my hypotheses for the lack of feeding homogeneity was that fish might not be able to detect pellets throughout the cage. This hypothesis was tested by examining the effect 206 Chapter 7. O VERALL CONCL USIONS AND FUTURE CONSIDERA TIONS of contrast enhanced pellet on feeding behaviour. Under low light and high turbidity levels in a laboratory setting, a pellet coating was identified, which was more detectable than conventional pellets. In the course of these laboratory studies, a reliable method for using fish to choose between pellets was developed using a balanced beam that can rotate freely within a metal sleeve. This method may be used in future investigations to identify other feed preferences. Under field conditions, contrast-enhanced pellets failed to achieve homogeneous feeding throughout the cage, rather fish persisted to fed in groups during different time intervals. The use of visually enhanced pellets to feed fish has revealed that given a choice of where to feed in the water column, fish preferred to feed lower in the water column than near the surface. 7.2 Incidental results Some observations led to new findings on fish behaviour. These findings are classified as incidental. The association between feeding rate and underwater light intensity levels (Chapter 6) is one of the major incidental findings. Apparently, fish avoid bright light, as feeding decreases with increase in light intensity. This association with feeding rate and light could have far-reaching ramification on sea cage aquaculture, as cages currently tend to be shallow. The use of underwater video monitoring of fish feeding has also demonstrated that fish behave differently when fed continuously as opposed to batch feeding, except when fish have no prior exposure to continuous feeding or intense feeding to provide indefensible food source. Further work should be conducted to compare FCR and growth under continuous and batch systems of feeding in order to evaluate the respective feeding efficiency. As shown through direct video observations, feeding based on pre-determined amounts of ration should be avoided because this practice can lead to over- or under-feeding. 207 Chapter 7. OVERALL CONCLUSIONS AND FUTURE CONSIDERATIONS 7.3 Practical applications Like most applied research, results from the studies conducted during this research have both scientific and practical applications. The method of feeding fish using underwater cameras to monitor pellet loss as well as fish feeding behaviour has been extensively adopted by the international fish farming community and has proven itself to be effective at reducing the FCR while increasing growth rate when used correctly. However, using pellet loss to determine how fast to feed fish and to determine when to stop feeding can still be expensive because it involves sacrificing/wasting a number of pellets before taking each remedial action (i.e. slow down or stop feed discharge). Since feeding behaviour can change in response to differing appetite levels and differing discharge rates, the entire feeding process can be conducted by monitoring fish behaviour. Specifically, farmers can be trained to feed fish by increasing the rate of discharge to a point when the population of feeding fish begins to move down toward the camera. Individual fish then become visible on the monitor, as they follow the pellets down, swimming in quick motions in the "S"-shaped pattern while foraging for pellets. The farmer can categorise this activity as "quick swimming" and learn to recognise and focus on this feeding pattern to decide what action to take. If the pellet discharge rate is reduced or temporarily stopped when this group of foraging fish first start to move down, the population of fish will eventually cease the foraging motions and begin to move back up because most of the pellets will have been consumed. Pellet discharge is resumed at a slower rate and the process of continuously reacting to fish movement up and down the water column can allow the fish to feed between the surface and the camera thus minimising pellet wastage. 208 Chapter 7. OVERALL CONCLUSIONS AND FUTURE CONSIDERATIONS The feeding process can, therefore, be simplified to one basic objective of matching pellet discharge rate to the consumption rate of fish. When considered over the whole production cycle, the reduction in amount of feed wasted at each feeding event and consequently the amount of money saved as well as the further improvement in FCR can be significant. Overall organic discharge to the environment is also reduced. Camera use has also increased first hand knowledge and understanding of the animals that fish farmers have taken under their care. The ability of the new feed monitoring system to detect pellet waste when it happens has taken much of the guess work out of the process of feeding fish to appetite (or to satiation). 7.4 Patentable technology A patent has been applied for by The University of British Columbia internationally for the technology that was developed during the later stages of this research. The technology to produce and use contrast-enhanced pellets has been licensed to a multinational company giving it sole rights to produce and sell contrast-enhanced pellet to the fish farming industry world-wide. Applications to include T C M in animal feed have also been applied for simultaneously in selected countries and to date, Chile has approved the use of titanium-dioxide-coated mica (TCM) as a feed additive in feed for food animal. Applications are pending in Canada and the European Union countries. The use of contrast-enhanced pellets under commercial farming situations may require prior knowledge of underwater turbidity and light conditions. Underwater light levels in cages with relatively high stocking densities are greatly reduced at depths as compared to a system without fish. For example, in cages stocked with fish of 2.8 kg or larger at 10.4 kg m"3 or 209 Chapter 7. OVERALL CONCLUSIONS AND FUTURE CONSIDERATIONS denser, only a small fraction (~0.07-0.37%) of the surface light is left (See Chapter 6). The relative depths below which enhanced pellets could be used more effectively than conventional pellets under different water visibility condition and light intensity levels in air are shown in Figure 7.1, assuming no shading by fish (Figure 7.1a) and with shading or overcrowding effects (Figure 7.1b). 7.5 Future considerations Results reported in Chapter 3 showed that BioFCR and growth were improved with the use of underwater cameras to control pellet loss while feeding fish. However, it could not be claimed that the improvements in BioFCR and growth were directly attributable to the amount of pellets saved from wastage when using underwater cameras to feed fish because cameras were not used in the control cages. Future experiments could involve putting cameras into the control cages. Additional cameras would necessitate a third group of researchers, who would have no contact with those feeding camera-monitored cages or control cages, or alternatively, images from cameras in the control cages could be taped and not viewed until the end of the experiments. This study has revealed that more research need to be conducted to determine the factors affecting the occurrence of uniform feeding of fish in the industry. Those factors include cage size, fish strain (genetics), temperature and hatchery rearing conditions as well as prior conditioning or entrainment effects. Further, the two distinct types of feeding formations associated with continuous and batch feeding may be exploited toward finding another way to judge satiation since both types of structures tended to break down over the feeding event. The difference in foraging patterns between continuous and batch feeding may lead to a difference 210 Chapter 7. OVERALL CONCLUSIONS AND FUTURE CONSIDERATIONS 0 i 1 1 1 1 i 0 2 4 6 8 10 Visibility, m A t 2 5 u m o l / s / m 2 1 0 0 u m o l / s / m 2 1 6 0 0 u m o l / s / m 2 6 4 0 0 u m o l / s / m 2 Figure 7.1 a Depth below which enhanced pellets should be used for greater detectability under different light levels in air and water visibility levels, assuming no shading effects due to fish. 0 H 1 1 1 1 1 0 2 4 6 8 10 Visibility, m A t 2 5 u m o l / s / m 2 1 0 0 u m o l / s / m 2 1 6 0 0 u m o l / s / m 2 6 4 0 0 u m o l / s / m 2 Figure 7.1 b Depth below which enhanced pellets should be used for greater detectability under different light levels in air and water visibility levels, assuming shading effect due to fish. 211 Chapter 7. OVERALL CONCLUSIONS AND FUTURE CONSIDERATIONS in energy requirements by fish, due, in part to the difference in amplitude as well as frequency of 'S'-shaped swerves exhibited in each type of pattern. As such, it may be beneficial to study the bioenergetics of fish in sea cages as an additional approach to improving efficiency of production. As stated in Section 6.6, there were a few possible reasons for why contrast-enhanced pellets did not permit uniform feeding and a faster feed discharge rate. These include: 1. fish may have been pre-conditioned to feed in groups, 2. bright lighting restricts feeding on the surface, and 3. fish may simply not be able to cooperate to permit uniform feeding. It may be beneficial to conduct experiments on contrast-enhanced pellets in the hatchery for two reasons, namely, 1. To examine the potential benefits of using silver feed to feed pre-smolts and, 2. To eliminate the effect of novelty of diet when experiments are conducted at the seawater (on-growing) stage or when silver feed is routinely used in the seawater stage to harvest. Finally, it should be stated that the impact (s) of uniform feeding or the lack of uniform feeding on the industry has not been fully understood but it represents a worthwhile area to spend the energy and time because of its potential influence on feeding efficiency. 212 APPENDICES R E F E R E N C E S Abbott, J .C, and Dill , L . M . , 1985. Patterns of aggressive attacks in juvenile steelhead trout (Salmo gairdneri). Can. J. Fish. Aquat. Sci. 42:1702-1706. Adron, J.W., Grant, P.T., and Cowey, C.B., 1973. A system for the quantitative study of the learning capacity of rainbow trout and its application to the study of food preferences and behaviour. J. Fish Biol. 5:625-636. Adron, J.A., and Mackie A . M . , 1978. Studies on the chemical nature of feeding stimulants for rainbow trout (Salmo gairdneri Richardson). J. Fish Biol. 13:303-310. Alanara, A. , 1992. The effect of time-restricted demand feeding on feeding activity, growth and feed conversion in rainbow trout (Oncorhynchus mykiss). Aquacult. 108:357-368. A l i , M.A. , 1975. Vision in Fishes: New Approaches in Research (lectures). N A T O Advanced Study Institute on New Approaches to the Study of Vision in Fishes. N R C Canada. Plenum Press, New York. 836 pp. Ang, K.P. and Petrell, R.J., 1997. Control of feed dispensation in seacages using underwater video monitoring: effects on growth and food conversion. Aquacult. Eng. 16:45-62. Ang, K.P. and Petrell, R.J., 1998. Pellet wastage, and subsurface and surface feeding behaviours associated with different feeding systems in sea cage farming of salmonids. Aquacult. Eng. 18:95-115. Anon 1, 1996. BCSFA Backgrounder, BC Salmon Farmers Association. Anon 2, 1998. Seafarm Business Review, January 1998 (No.3 and 4, Volume 2). Archer, J., 1979. Animals under stress. Studies in Biology no. 108. Edward Arnold (Publishers) Limited, London. 59 pp. Austreng, E., Storebakken, T., and Asgard, T., 1987. Growth rate estimates for cultured Atlantic salmon and rainbow trout. Aquacult. 60:157-160. Balon, E.K., 1986. Types of feeding in the ontogeny of fishes and the life-history model. In: Simenstad, C.A., and Caillet, G .M. (Eds.), Contemporary studies on fish feeding: the proceedings of GUTSHOP '84, Dr W. Junk Publishers, Dordrecht, The Netherlands. 334 pp. Barnard, C.J., 1983. Animal Behaviour: ecology and evolution. Croom Helm, London. 339 pp. 213 APPENDICES Barrett, J .C, Grossman, G.D., and Rosenfeld, J., 1992. Turbidity-induced changes in reactive distance of rainbow trout. Trans. Am. Fish Soc. 121: 437-443. Bjordal, A , Juell, J.E., Lindem, T., and Ferno, A. , 1993. Hydroacoustic monitoring and feeding . control in cage rearing of Atlantic salmon (Salmo salar L.). In: Reinertsen, H., Dahle, L .A. , J0rgensen, L. and Tvinnereim, K . (Eds.), Fish Farming Technology. Balkema, Rotterdam. 482 pp. Blaxter, J.H.S., 1970. 2. Light 2.3. Animals, 2.32 Fishes. Kinne, O. (Ed.), Marine Ecology, Vol.1, Part 1. Wiley, London, pp. 213-320. Blyth, P. J., Purser, G. J., and Russell, J. F., 1993. Detection of feeding rhythms in seacaged Atlantic salmon using new feeder technology. In: Reinertsen, H. , Dahle, L .A. , J0rgensen, L . and Tvinnereim, K . (Eds.), Fish Farming Technology. Balkema, Rotterdam. 482 pp. Boisclair, D., and Tang, M . , 1993. Empirical analysis of the influence of swimming pattern on the net energetic cost of swimming in fishers. J. Fish Biol. 42:169-183. Bone, Q., Marshall, N.B. , and Blaxter, J.H.S., 1995. Behaviour. Biology of Fishes. 2nd. ed. Blackie Academic and Professional, New York. 332 pp. Breder, C. M . , 1965. Vortices and fish schools. Zoologica 50:97-114. Brett, J.R., 1971. Satiation time, appetite and maximum food intake of sockeye salmon, Oncorhynchus nerka. J. Fish. Res. Bd. Can. 28:409-415. Brier, S. S., 1980. Analysis of contingency tables under cluster sampling. Biometrika 67:591-596. Browman, H.I., and Marcotte, B .M. , 1987. The effect of zooplankton abundance on feeding behavior and prey size selection in Atlantic salmon, Salmo salar, alevins. Holarctic Ecol. 10:163-170 Bullock, A . M . , 1988. Solar ultraviolet radiation: a potential environmental hazard in the cultivation of farmed fish. In: Muir, J.F., and Roberts, R.J. (Eds.), Recent Advances in Aquaculture, Vol. 3. Croom Helm, London, pp. 139-224. Burgner, R.L., 1991. Life history of sockeye salmon (Oncorhynchus nerka), In: Groot, C , and Margolis, L. (Eds.) Pacific salmon life histories. UBC Press, Vancouver, 564 pp. Calow, P., and Townsend, C.R., 1981. Ecology, energetics and evolution. In: Townsend, C.R., and P. Calow, P. (Eds.). Physiological Ecology: An Evolutionary Approach to Resource Use, Blackwell, Oxford. 393 pp. 214 APPENDICES Canero, C , 1999. An examination of the changes in % light transmittance during fish farming in seacages. Undergraduate thesis available from the Department of Chemical and Bio-Resource Engineering, University of British Columbia, Vancouver. 200 pp. Clark, C.W., and Levy, D.A., 1988. Diel vertical migrations by juvenile sockeye salmon and the anti-predation window. Am. Nat. 131:271-290. Clarke, L.A. , and Sutterlin, A . M . , 1985. Associative learning, short term memory, and colour preference during first feeding by juvenile Atlantic salmon. Can. J. Zool. 63:9-14. Craig, J.V., 1981. Domestic Animal Behavior: Causes and implications for animal care and management. Prentice-Hall, Inc., N.J. 364 pp. Crowl, T.A., 1989. Turbidity effects on largemouth bass-crayfish interactions: a predation component model. Hydrobiologia 183:133-140. Cushing, D.H., and Harden-Jones, F.R., 1968. Why do fish school? Nature 218:918-920. Danzmann, R.G., Fergusion, M . M . , and Noakes, D.L.G., 1993. The genetic basis of fish behaviour. In: Pitcher, T.J. (Ed.), Behaviour of Teleost Fishes, 2nd. ed. Chapman and Hall, London. 715 pp. Dendrinos, P., Dewan, S., and Thorpe, J.P., 1984. Improvement in the feeding efficiency of larval, postlarval and juvenile Dover sole (Solea solea L.) by the use staining to improve the visibility of Artemia used as food. Aquacult. 38:137-144. De Ruiter, L., 1967. The feeding behaviour of vertebrates in the natural environment. Am. Physiol. Soc. 1(6):97-116. Dill , P.A., 1977. Development of behaviour in alevins of Atlantic salmon, Salmo salar, and rainbow trout, S. gairdneri. Anim. Behav. 25:116-21. Dil l , L . M . , Ydenberg, R.C., and Fraser, A.H.G. , 1981. Food abundance and territory size in juvenile coho salmon (Oncorhynchus kisutch). Can. J. Zool. 59:1801-1809. Doebelin, E.O., 1975. Measurement system: application and design. McGraw Hill . 772 pp. Douglas, R.H., and Djamgoz, M.B.A. (Eds.), 1990. The Visual System of Fish. Chapman and Hall, London. 526 pp. Douglas, R.H. and Hawryshyn, C.W., 1990. Behavioral studies of fish vision: an analysis of visual capabilities, in Douglas, R.H., and Djamgoz, M.B.A. (Eds.), The Visual System of Fish, Chapman and Hall, London. 526 pp. 215 APPENDICES Dunbrack, R.L., 1988. Feeding of juvenile coho salmon (Oncorhynchus kisutch): maximum appetite, sustained feeding rate, appetite return, and body size. Can. J. Fish. Aquat. Sci. 45:1191-1196. Elliot, J.M., 1972. Rates of gastric evacuation in brown trout, Salmo trutta L. Fresh. Biol. 2:1-18. Elliot, J.M., 1975. Number of meals in a day, maximum weight of food consumed in a day and maximum rate of feeding for brown trout, Salmo trutta. Fresh. Biol. 5:287-303. Ellis, D.V., 1985. Animal Behavior and its Applications. Lewis Publishers, Inc., MI. 329 pp. Eriksson, L.O., 1973. Spring inversion of the dial rhythm of locomotor activity in young seagoing brown trout, Salmo trutta L., and Atlantic salmon, Salmo salar L. Aquilo Ser. Zool. 14:68-79. Eriksson, L-O. and Alanara, A. , 1990. Timing of feeding behaviour in salmonids. In: Thorpe, J.E., and Huntingford, F.A. (Eds.). The importance of feeding behaviour for the efficient culture of salmonid fishes. The World Aquaculture Society, Baton Rouge. 60 pp. Fernald, R.D., 1993. Vision. In: Evans, D.H. (Ed.), The Physiology of Fishes, CRC Press, Boca Raton. 592 pp. Ferno, A. , 1989. Behavioural studies of salmon reared in marine net pens. Ethology in aquaculture: Applications of behavioural studies in aquaculture. Proceedings from minisymposium, Trondheim, Norway, pp. 42-47. Ferno, A. , and Holm, M . , 1986. Aggression and growth of Atlantic salmon parr. I. Different stocking densities and size groups. FiskDir. Ser. HavUnders. 18:113-122. Ferno, A. , Huse, I, Juell, J-E., and Bjordal, A. , 1995. Vertical distribution of Atlantic salmon (Salmo salar L.) in net pens: trade-off between surface light avoidance and food attraction. Aquacult. 132:285-296. Foster, M . , Petrell, R.J., Tito, M . , and Ward, R., 1995. Detection and counting of uneaten food pellets in a net pen. Aquacul. Eng. 14 (3):251-270. Fraser, A.F., and Broom, D.M., 1990. Farm animal behaviour and welfare. 3rd ed. Bailliere Tindall, London. 437 pp. Fretwell, S.D., and Lucas, H.L., 1969. On territorial behavior and other factors influencing habitat distributions in birds. Acta Biotheoretica 19 : 16-36. Furevik, D., Huse, I., Bjordal, A. , and Ferno, A. , 1988. Surface activity of Atlantic salmon (Salmo salar) in net pens. Coun. Meet. int. Coun. Explor. Sea (F:19). 216 APPENDICES Gee, J.H., 1983. Ecologic implications of buoyancy control in fish. In: Webb, P. W. and Weihs, D., (Eds), Fish Biomechanics. Praeger, New York. 398 pp. Gillis, D.M. , and Kramer, D.L., 1987. Ideal interference distributions: population density and patch use by zebrafish. Anim. Beh. 35:1875-1882. Ginetz, R.M. , and Larkin, P.A., 1973. Choice of colors of food items by rainbow trout (Salmo gairdneri). J. Fish. Res. Bd. Can. 30:229-234. Gowen, R.J., Bradbury, N.B. , and Brown, J.R., 1985. The ecological impact of salmonid farming in Scottish coastal waters: a preliminary appraisal. ICES, C M . 1985, F:35, Ref. E+C, Sess. W, 13 pp. Greig-Smith, P., 1990. Foraging behaviour. In: Monaghan, P. and Wood-Gush, D. (Eds.), Managing the Behaviour of Animals. Chapman and Hall. London. 256 pp. Groot, C. and Margolis, L. , 1991. Pacific Salmon Life Histories. U B C Press. Vancouver. 564 pp. Grove, D.J., Loizides, L.G. , and Nott, J., 1978. Satiation amount, frequency of feeding and gastric emptying rate in Salmo gairdneri. J. Fish Biol. 12:507-516. Guthrie, D .M. , 1980. Neuroethology, Blackwell, Oxford. 221 pp. Guthrie, D .M. , 1986. Role of vision in fish behaviour. In: Pitcher, T J . (Ed.), The Behaviour of Teleost Fishes. Croom Helm, London. 553 pp. Guthrie, D .M. , and Muntz, W.R.A., 1993. Role of vision in fish behaviour. In: Pitcher, T.J. (Ed.), Behaviour of Teleost Fishes, 2nd. ed. Chapman and Hall, London. 715 pp. Hafez, E.S.E., 1975. The Behaviour of Domestic Animals. 3rd ed. Bailliere Tindall, London. 532 pp. Hart, B.L., 1985. The Behaviour of Domestic Animals. W.H. Freeman and Co., New York. 390 pp. Hart, P.J.B., 1993. Teleost foraging: facts and theories. In: Pitcher, T.J. (Ed.). Behaviour of Teleost Fishes, 2nd. ed., Chapman and Hall, London. 715 pp. Hasler, A.D., Scholz, A.T., and Horrall, A . M . , 1978. Olfactory imprinting and homing in salmon. Am. Sci. 66:347-355. Hassell, M.P:, 1976. The Dynamics of Competition and Predation. Edward Arnold, London. 289 pp. 217 APPENDICES Healey, M.C. (1991). Life History of Chinook Salmon (Oncorhynchus tshawytscha). In: Groot, C , and Margolis, L . (Eds.) Pacific salmon life histories. U B C Press, Vancouver, 564 pp. Heard, W.R., 1991. Life History of Pink Salmon (Oncorhynchus gorbuscha). In: Groot, C , and Margolis, L . (Eds.) Pacific salmon life histories. U B C Press, Vancouver, 564 pp. Higgins, P.J., and Talbot, C , 1985. Growth and feeding in juvenile Atlantic salmon, Salmo salar L. In: Cowey, C.B., Mackie, A . M . , and Bell, J.G. (Eds.), Nutrition and Feeding in Fish. Academic Press, London. 489 pp. Hoar, M . , 1942. Diurnal variation in feeding activity of young salmon and trout. J. Fish. Res. Bd. Can. 6:90-99. Hoar, W.S., 1958. The evolution of migratory behaviour among juvenile salmon of the genus Oncorhynchus. J. Fish. Res. Bd. Can. 15:391-428. Holm, M . , Huse, I., Doving, K.B. , and Aure, J., 1982. Behaviour of Atlantic salmon smolts during seaward migration. Int. Counc. Explor. Sea, C M . 1988/M:7. Howland, H . C , 1974. Optimal strategies for predator avoidance: The relative importance of speed and manoeuvrability. J. Theoret. Biol. 47:333-350. Huntingford, F., 1984. The Study of Animal Behaviour. Chapman and Hall, London. 411 pp. Huntingford, F.A., 1993. Development of behaviour in fish. In: Pitcher T.J. (Ed.), Behaviour of Teleost Fishes. 2nd ed., Chapman and Hall, London. 715 pp. Hutchinson, M.J. and Iwata, M . , 1997. A comparative analysis of aggression in migratory and non-migratory salmonids. Env. Biol. Fishes 50:209-215. Hyatt, K.D., 1979. Feeding strategy. In: Hoar W.S., Randall, D.J., and Brett, J.R.(Eds.), Fish Physiology. Academic Press, London. Vol. 8, pp. 71-119. Irvine, J.R. and Northcote, T.G., 1983. Selection of young rainbow trout (Salmo gairdneri) in simulated stream environments for live and dead prey of different sizes. Can. J. Fish. Aquat. Sci. 40:1745-1749. Jakobsen, P.J., Johnsen, G.H., and Holm, J .C, 1987. Increased growth rate in Atlantic salmon parr (Salmo salar) by using a two-coloured diet. Can. J. Fish. Aquat. Sci. 44:1079-1082. Jakubsstovu, J.L, 1988. Exploitation and migration of salmon in Faroese waters. In: Mills, D. and Piggins, D. (Eds.), Atlantic Salmon: Planning for the Future. Croom Helm, London. 587 pp. Jobling, M . , 1986. Mythical models of gastric emptying and implications for food consumption studies. Env. Biol. Fish. 16 (l-3):35-50. 218 APPENDICES Jobling, M . , Arnesen, A . M . , Baardvik, B . M . , Christiansen, J.S. and Jorgensen, E.H., 1995. Monitoring feeding behaviour and food intake: methods and applications. Aquacult. Nutr. 1:131-143. Johannessen, A. , 1978. Early stages of Lepeophtheirus salmonis (Copepoda Caligidae). Sarsia. 63:169-176. Johnson, P.C., 1984. The Scientific Angler. Charles Scribner's Sons, NewYork. 289 pp. Johnston, D.D., and Wildish, D.J., 1982. Effect of suspended sediment on feeding by larval herring (C/wpea harengus harengus L.). Bull. Env. Contam. Toxicol. 29:261-267. Juell, J.E., 1991. Hydroacoustic detection of food waste - A method to estimate maximum food intake of fish populations in sea cages. Aquacult. Eng. 10:207-217. Juell, J.E. and Westerberg, H. , 1993. A n ultrasonic telemetric system for Automatic positioning of individual fish used to track Atlantic salmon (Salmo salar L.) in sea cages. Aquacult. Eng. 12:1-18. Juell, J.E., Furevik, D .M. , and Bjordal, A. , 1993. Demand feeding in salmon farming by hydroacoustic food detection. Aquacult. Eng. 12:155-167. Juell, J.E., Bjordal, A. , Ferno, A . and Huse, I., 1994a. Effect of feeding intensity on food intake and growth of Atlantic salmon (Salmo salar L.) in sea cages. Aquacult. Fish. Mgmt. 25:453-464. Juell, J.E., Ferno, A. , Furevik, D., and Huse, I., 1994b. Influence of hunger level and food availability on the spatial distribution of Atlantic salmon, Salmo salar L. , in sea cages. Aquacult. Fish. Mgmt. 25:439-451. Kadri, S., Metcalfe, N.B. , Huntingford, F.A., and Thorpe, J.E., 1991. Daily feeding rhythms in Atlantic salmon in sea cages. Aquacult. 92:219-224. Kadri, S., Huntingford, F.A., Metcalfe,-N.B., and Thorpe, J.E., 1996. Social interactions and the distribution of food among one sea-winter Atlantic salmon (Salmo salar L.) in a sea cage. Aquacult. 139:1-10. Kadri, S., Metcalfe, N . , Huntingford, F.A., and Thorpe, J.E., 1995. What controls the onset of anorexia in maturing adult female Atlantic salmon? Func. Ecol. 9:790-797. Kapoor, B.G., Smith, H. , and Verighina, I.A., 1975. The alimentary canal and digestion in teleosts. Adv. Mar. Biol. 13:109-239. 219 APPENDICES Kimura, R., Kawaguchi, S., and Sugitani, S., 1993. Automatic feeding control system for fish. In: Reinertsen, H., Dahle, L .A. , Jorgensen, L. , and Tvinnereim, K . (Eds.), Fish Farming Technology. Rotterdam:Balkema, 482 pp. Klontz, G.W., 1993. Producing a marketable fish. Part V . Inventory techniques. North. Aquacult. 10:21-25. Knights, B., 1985. Feeding behavoir and fish culture. In: Cowey, C.B., Mackie, A . M . , and Bell, J.G. (Eds.), Nutrition and feeding in fish. Academic Press, London. 489 pp. Koebele, B.P., 1985. Growth and the size hierarchy effect: an experimental assessment of three proposed mechanisms: activity differences, disproportional food acquisition, physiological stress. Env. Biol. Fish. 12:181-8. Krebs, J.R., 1978. Optimal Foraging: Decision Rules for Predators. In: Krebs. J.R., and Davies, N.B. (Eds.), Behavioural Ecology: an Evolutionary Approach, Blackwell, Oxford. 494 pp. Krebs, J.R. and Stephens, D.W., 1986. Foraging Theory. Princeton University Press, NJ. Laird, L . M . , 1996. History and applications of salmonid culture. In: Pennell, W., and Barton, B.A. (Eds.). Principles of Salmonid Culture, Developments in Aquaculture and Fisheries Science, Vol. 29, Elsevier, Amsterdam. 1039pp. Lehner, P.N., 1996. Handbook of Ethological Methods. Cambridge University Press, New York. 672 pp. Lemm, C.A., 1983. Growth and survival of Atlantic salmon fed semi-moist or dry starter diets. Prog. Fish-Cult. 45 (2):72-74. Lemm, C.A., and Hendrix, M.A. , 1981. Growth and survival of Atlantic salmon fed various starter diets. Prog. Fish-Cult. 43:195-199. Levine, J.S., Lobel, P.S., and MacNichol, E. F., 1980. Visual Communication in Fishes. In: A l i , M . A . (Ed.), Environmental Physiology of Fishes, Plenum, New York. 723 pp. Levy, D.A., 1990. Sensory mechanisms and selective advantage for diel vertical migration in juvenile sockeye salmon, Oncorhynchus nerka. Can. J. Fish. Aquat. Sci. 47:1796-1802. Loew, E.R., and McFarland, W.N., 1990. The underwater visual environment. In: Douglas, R.H., and Djamgoz, M.B.A. (Eds.), The Visual System of Fish, Chapman and Hall, London. 526 pp. Lythgoe, J.N., 1972. The adaptation of visual pigments to their photic environment. In: Dartnall, H J . A . (Ed.) Handbook of Sensory Physiology. Springer-Verlag, New York Vol . VII / l , pp. 566-603. 220 APPENDICES Lythgoe, J.N., 1975. The ecology, function and phylogeny of iridescent multilayers in fish corneas. In: Evans, G . C , Bainbridge, R., and Rackham, O. (Eds.), Light as an ecological Factor If Blackwell, Oxford. 616 pp. Lythgoe, J.N., 1979. The Ecology of Vision, Clarendon Press, Oxford. 244 pp. Lythgoe, J.N., 1988. Light and vision in the aquatic environment. In: Atema, J., Fay, R.R., Popper, A . N . , and Tavolga, W.N. (Eds.), Sensory Biology of Aquatic Animals. Springer-Verlag, New York. 936 pp. McBride, J.R., Idler, D.R., Jonas, R.E.E., and Tomlinson, N . , 1962. Olfactory perception in juvenile salmon. I. Observations on response of juvenile sockeye to extracts of food. J. Fish. Res. BdCan. 19:327-334. Maier, R.A., and Maier, B .M. , 1970. Comparative Animal Behaviour. Brooks/Cole, Belmont, California. 459 pp. Mearns, K.J. , 1985. Response of Atlantic salmon (Salmo salar L.) yearlings to individual 1-amino acids. Aquacult. 48:253-259. Mearns, K.J., 1986. Sensitivity of brown trout (Salmo trutta L.) and Atlantic salmon (Salmo salar L.) fry to amino acids at the start of exogenous feeding. Aquacult. 55:191-200. Metcalfe, N.B. , 1994. The role of behaviour in determining salmon growth and development. Aquacult. Fish. Mgmt. 25:67-76. Metcalfe, N.B. , 1990. Aquaculture. In: Monaghan, P. and Wood-Gush, D. (Eds.), Managing the Behaviour of Animals. Chapman and Hall, London. 256 pp. Milinski, M . , 1979. An evolutionary stable feeding strategy in sticklebacks. Z. Tierpsychoi, 51:36-40. Milinski, M . , 1982. Optimal foraging: the influence of intraspecific competitionon diet selection. Behav. Ecol. Sociobiol, 11:109-15. Milinski, M . , 1984a. A predator's costs of overcoming the confusion effect of swarming prey. Anim. Behav. 32:1157-62. Milinski, M . , 1984b. Competitive resource sharing: an experimental test of a learning rule for ESSs. Anim. Behav., 32:233-42. Milinski, M . , 1986. A review of competitive resource sharing under constraints in sticklebacks. J. Fish Biol. 29 (Suppl. A): 1-14. 221 APPENDICES Miner, J.G., and Stein, R.A., 1993. Interactive influence of turbidity and light on larval bluegill (Lepomis macrochirus) foraging. Can. J. FishAquat. Sci. 50:781-788. Mogurran, A.E. , 1993. Individual differences and alternative behaviours. In: Pitcher, T J . (Ed.). Behaviour of Teleost Fishes, 2nd. ed., Chapman and Hall, London. 715 pp. Monaghan, P., 1984. Applied etholgy. Aram. Behav. 32:908-915. Monaghan, P., and Wood-Gush, D. (Eds.), 1990. Managing the Behaviour of Animals. Chapman and Hall, London. 256 pp. Muntz, W.R.A., 1990. Stimulus, environment and vision in fishes. In: Douglas, R.H., and Djamgoz, M.B.A. (Eds.), The Visual System of Fish. Chapman and Hall, London. 526 pp. Munz, F.W., and McFarland, W.N., 1977. Evolutionary adaptations of fishes to the photic environment. In: Crescitelli, F. (Ed.), The Handbook of Sensory Physiology, VII 15, Springer-Verlag, Berlin, pp. 193-274. Naiberg, A. , Petrell, R. J., Savage, C.R., and Neufeld, T.P., 1993. Stereo video technique to size fish in sea cages and tanks. Techniques For Modern Aquaculture: Proceedings of an Aquacultural Engineering Conference. American Society of Agricultural Engineers, St. Joseph, MI, pp. 393-402. Neilson, J.D., and Perry, R.J., 1990. Diel vertical migrations of marine fishes: an obligate or facultative process? Adv. Mar. Biol. 26:115-167. Nicol, J.A.C., 1989. The Eyes of Fishes. Oxford University Press, New York. 308 pp. Noakes, D.L.G., and Grant, J.W.A., 1990. Feeding and social behaviour of brook and lake charr. In: Thorpe, J.E., and Huntingford, F A . (Eds.), The importance of feeding behavior for the efficient culture of salmonid fishes. The World Aquaculture Society. Baton Rouge. 60 pp. Novotny, A.J. , and Pennell. W., 1996. Rearing fish to market size in marine waters. In: Pennell, W., and Barton, B.A. (Eds.). Principles of Salmonid Culture, Developments in Aquaculture and Fisheries Science, Vol. 29, Elsevier, Amsterdam. 1039pp. Ogura, M . , and Ishida, Y. , 1992. Swimming behaviour of coho salmon, Oncorhynchus kisutch, in the open sea as determined by ultrasonic telemetry. Can. J. Fish. Aquat. Sci. 49 : 453-457. Oikawa, C K . and March, B.E., 1997. A method for assessment of the efficacy of feed attractants for fish. Prog. Fish-Cult. 59 (3):213-217. Olsen, O A . , and Balchen, J.G., 1992. Structured modelling of Fish Physiology. Biosci. Math. 112:81-113. 222 APPENDICES Parsons, T.R., Takahasi, M . , and Hargrave, B., 1984. Biological Oceanographic Processes. 3rd ed. Pergammon Press, New York. 330 pp. Petrell, R.J., Shi, X . , Ward, R.K., Naiburg, N . , and Savage, C.R., 1997. Determining fish size and swimming speed in cages and tanks using simple video techniques. Aquacult. Eng. 16:63-84. Pickering, A.D. , 1992. Rainbow trout husbandry: management of the stress response. Aquacult. 100 (1992):125-139. Pitcher, T.J., 1986. Functions of shoaling behaviour in teleosts. In: Pitcher, T J . (Ed.), The Behaviour of Teleost Fishes, Croom Helm, London. 553 pp. Pitcher, T J . (Ed.), 1993. Behaviour of Teleost Fishes. Chapman and Hall, London. 715 pp. Price, E.O., 1984. Behavioural aspects of animal domestication. Q. Rev. Biol. 59:1-32. Puckett, K.J . , and Di l l , L . M . , 1985. The energetics of territoriality in juvenile coho salmon (Oncorhynchus kisutch). Beh. 92:97-111. Quinn, T.P., TerHart, B.A., and Groot, C , 1989. Migratory orientation and vertical movements of homing adult sockeye salmon, Oncorhynchus nerka, in coastal waters. Anim. Behav. 37:587-599. Reese, E.S., and Lighter, F.J. (Eds.), 1978. Contrasts in Behaviour: Adaptations in the aquatic and terrestrial environments. John Wiley and Sons, Inc., New York. 406 pp. Rimmer, D.M. , and Power, G., 1978. Feeding response of Atlantic salmon, Salmo salar L. alevins in flowing and still water. J. Fish. Res. Bd. Can. 35 (3):329-332. Ruggerone, G.T., Quinn, T.P., McGregor, I.A., and Wilkinson, T.D., 1990. Horizontal and vertical movements of adult steelhead trout, Oncorhynchus mykiss, in the Dean and Fisher channels, British Columbia. Can. J. Aquat. Sci. 47:1963-1969. Sachs, L., 1984. Applied Statistics: A Handbook of Techniques. 2nd ed. Springer-Verlag, New York. 707 pp. Seymour, E.A., and Bergheim, A. , 1991. Towards a reduction of pollution from intensive aquaculture with reference to the farming of salmonids in Norway. Aquacult. Eng. 10:73-88. Shaw, E., 1978. Schooling fishes. Am. Sci. 66:166-175. 223 APPENDICES Shaw, S.A., and Egan, D.N., 1996. Marketing. In: Pennell, W., and Barton, B.A. (Eds.). Principles of Salmonid Culture, Developments in Aquaculture and Fisheries Science, Vol. 29, Elsevier, Amsterdam. 1039pp. Shelton, P.M.J., 1987. How trout see color. Trout Fisherman. 1987 (Dec):37-39. Skjervold, P.O., 1993. Fish feeding station. In: Reinertsen, H. , Dahle, L .A. , J0rgensen, L. and Tvinnereim, K. (Eds.), Fish Farming Technology. Balkema, Rotterdam. 482 pp. Smith, L.S., 1989. Digestive functions in teleost fishes. In: Halver, J.E. (Ed.), Fish Nutrition, 2nd. ed. Academic Press, London. 798 pp. Smith, I. P., Metcalfe, N . B., Huntingford, F. A. , and Kadri, S., 1993. Daily and seasonal patterns in the feeding behaviour of Atlantic salmon (Salmo salar L.) in a sea cage. Aquacult. 117:165-178. Smith, LP., Metcalfe, N.B., and Huntingford, F.A., 1995. The effects of food pellet dimensions on feeding responses by Atlantic salmon (Salmo salar L.) in a marine net pen. Aquacult. 130 (1995):167-175. Snedecor, G. W., and Cochran, W. G., 1967. Statistical Methods. 6th ed. The Iowa State University Press, Iowa. 593 pp. Steel, R.G.D., and Torrie, J.H., 1960. Principles and Procedures of Statistics. McGraw-Hill, New York. 481 pp. Storebakken, T., and Austreng, E., 1987a. Ration level for salmonids. I. Growth, survival, body composition, and feed conversion in Atlantic salmon fry and fingerlings. Aquacult. 60:189-206. Storebakken, T., and Austreng, E., 1987b. Ration level for salmonids. II. Growth, feed intake, protein digestibility, body composition, and feed conversion in rainbow trout weighing 0.5-1.0 kg. Aquacult. 60:207-221. Stradmeyer, L . , Metcalfe, N . B., and Thorpe, J. E., 1988. Effect of food pellet shape and texture on the feeding response of juvenile Atlantic salmon. Aquacult. 73:217-228. Stradmeyer, L. , 1989. A behavioural method to test feeding responses offish to pelleted diets. Aquacult. 79:303-310. Stradmeyer, L., 1992. Appearance and taste of pellets influence feeding behaviour of Atlantic salmon. In: Thorpe, J.E., and Huntingford, F.A. (Eds.), The importance of feeding behaviour for the efficient culture of salmonid fishes. The World Aquaculture Society, Baton Rouge. 60 pp. 224 APPENDICES Stradmeyer, L. and Thorpe, J.E., 1987. The responses of hatchery-reared Atlantic salmon, Salmo salar L., parr to pelleted and wild prey. Aquacult. Fish. Mgmt. 18:51-61. Sutterlin, A . M . , and Sutterlin, N . , 1970. Taste responses in Atlantic salmon {Salmo salar). J. Fish. Res. Bd. Can. 27:1927'-1942. Sutterlin, A . M . , Jokola, K. J., and Holte, B., 1979. Swimming behaviour of salmonid fish in ocean pens. J. Fish. Res. Bd. Can. 36:948-954. Talbot, C , 1995. Feed Management Guidelines for Salmon and Trout. Report # 6/95. Trouw Aquaculture. Stavanger, Norway. 25 pp. Tang, M . , and Boisclair, D., 1993. Influence of the size of enclosures on the swimming characteristics of juvenile brook trout (Salvelinus fontinalis). Can. J. Fish. Aquat. Sci. 50:1786-1793. Tett, P., 1990. The photic zone. In: Herring, P.J., Campbell, A . K . , Whitefield, M . , and Maddock, L. (Eds.), Light and life in the sea. Cambridge University Press, Cambridge. 357 pp. Thomassen, J .M. and Lekang, O.I., 1993. Optimal distribution of feed in sea cages. In: Reinertsen, H. , Dahle, L .A. , Jorgensen, L . and Tvinnereim, K. (Eds.), Fish Farming Technology. Balkema, Rotterdam. 482 pp. Thorpe, J.E., 1989. Developmental variation in salmonid populations. J. Fish Biol. 35:295-303. Thorpe, J.E., Morgan, R.I.G., Prestwell, D., and Higgins, P.J., 1988. Movement rhythms in juvenile Atlantic salmon (Salmo salar L.). J. Fish Biol. 33:931-940. Thorpe, J.E., Talbot, C , Miles, M.S., Rawlings, C , and Keay, D.S., 1990. Food consumption in 24 hours by Atlantic salmon (Salmo salar L.) in sea cage. Aquacult. 90:41-47. Tytler, P. and Calow, P. (Eds.), 1985. Fish Energetics: New Perspectives. Croom Helm, London. 349 pp. Vinyard, G. L. , and O'Brien, W. J, 1976. Effects of light and turbidity on the reactive distance ofbluegill (Lepomis macrochirus). J. Fish. Res. Bd. Can. 33:2845-2849. Wankowski, J.W.J., 1979. Morphological limitations, prey size selectivity, and growth response of juvenile Atlantic salmon (Salmo salar L.). J. Fish Biol. : 89-100. Wankowski, J.W.J, and Thorpe, J.E., 1979. The role of food particle size in the growth of juvenile Atlantic salmon (Salmo salar L.). J. Fish Biol. 14:351-370. Ward, F., 1919. Animal Life Underwater, Cassell, London. 232 pp. 225 APPENDICES Weihs, D., 1973. Optimal cruising speed for migrating fish. Nature 245:28-50. Weihs, D., 1975. Some hydrodynamical aspects of fish schooling. Vol. 2. In: Wu, T.Y., Brockaw, C J . , and Brennan, C. (Eds.), Swimming and flying in nature. Plenum Press, New York, pp. 703-718. Weihs, D., and Webb, P.W., 1983. Optimization of Locomotion. In: Webb, P. W. and Weihs, D., (Eds), Fish Biomechanics. Praeger, New York. 398 pp. Westerberg, H. , 1982. Ultrasonic tracking of Atlantic salmon (Salmo salar L.) II. Swimming depth and temperature stratificatioa Rep. Inst. Fresh. Res. 60:102-115. Wood-Gush, D.G.M., 1983. Elements of ethology. A text book for agricultural and veterinary students. Chapman and Hall, London. 240 pp. Wootton, R J . , 1990. Ecology of teleost fishes, Chapman and Hall, London. 404 pp. Zar, J.H., 1999. Biostatistical analysis. Prentice Hall, New Jersey. 663+pp. 226 APPENDICES Appendix A- 1. Two-sample / tests (a = 0.10) for SGR, BioFCR and Mortality rates between CMCs and control cages, for Experiments 1, 2 and 3 (Chapter 3). t, df calculated critical P Result Experiment 1 SGR, % d"1 4 0.6348 2.1318 0.5601 DNR BioFCR 4 -2.1760 2.1318 0.0952 Reject Mort., % 4 -2.2124 2.1318 0.0914 Reject Experiment 2 SGR, % d"1 4 2.7135 2.1318 0.0533 Reject BioFCR 4 -2.2508 2.1318 0.0876 Reject Mort., % 4 -2.5830 2.1318 0.0611 Reject Experiment 3 SGR, % d"1 4 3.4601 2.1318 0.0258 Reject BioFCR 4 -2.1620 2.1318 0.0967 Reject Mort., % 4 -0.1440 2.1318 0.9148 DNR Hypotheses tested (two-tailed) H 0 ; \L\ = \n; H A : \i\ * u 2; df: degrees of freedom; Reject: Reject H 0 ; DNR: Do not reject H 0 ; Appendix A- 2. F-test two-sample for variances (a = 0.05) for daily ration amounts in CMCs and control cages for Experiments 1, 2 and 3 (Chapter 3). F, F, Variance Variance V i v2 calculated Critical (CMC) (control) P Experiment 1 92 92 3.7194 1.4116 876.24 233.17 0.0000 Experiment 2 120 122 1.4784 1.3499 1314.96 889.48 0.0162 Experiment 3 59 59 1.7362 1.5400 4234.02 2438.69 0.0180 Hypotheses tested H 0 ; cu = o 2; H A : ai * a 2 ; v b v2: degrees of freedom. 227 APPENDICES Appendix B- 1. Two-way analysis of variance (using S-Plus)for data from Bamfield Marine Station, with visible distance as outcome and light level (depth) and pellet type as the factors, for clear water condition. With 18 different types of pellets, there were 153 comparisons. Significance is indicated by "****". Fl refers to fluorescent colours. Pellet type Estimate Std. Error Lower bound Upper bound Significance Blue-Brown 1. 45e-001 0. 358 -1. 17000 1. 460 Blue-Eggplant 1. 05e-001 0. 358 -1. 21000 1. 420 B l u e - F l yellow -2 . 37e+000 0 . 358 -3. 68000 -1. 050 •k ~k k ~k B l u e - F l blue -1. 92e+000 0. 358 -3. 23000 -0. 606 -k k k ~k B l u e - F l green -1. 73e+000 0. 358 -3. 04000 -0. 416 -k k k k B l u e - F l orange -2 . 20e+000 0. 358 -3. 51000 -0. 886 -k ~k -k -k B l u e - F l red -2. 26e+000 0. 358 -3. 57000 -0. 941 -k-k-k-k Blue-Green 1. 50e-001 0. 358 -1. 16000 1. 460 Blue-Grey -5. 00e-002 0 . 358 -1. 36000 1. 260 Blue-Nat guanine -9. 35e-001 0. 358 -2 . 25000 0. 379 Blue-Orange -3. 30e-001 0 . 358 -1. 64000 0 . 984 Blue-Red 1. 05e-001 0 . 358 -1. 21000 1. 420 B l u e - S i l v e r -1. 90e+000 0. 358 -3. 21000 -0. 586 -k-k-k-k Blue-Sparkle -1 50e-001 0. 358 -1. 46000 1. 160 Blue-Uncoated -1 50e-002 0 358 -1. 33000 1. 300 Blue-White -2 39e+000 0 358 -3. 70000 -1. 070 -k-k-k-k Blue-yellow -1 04e+000 0 358 -2 35000 0 274 Brown-Eggplant -4 00e-002 0 358 -1 35000 1 270 Brown-Fl yellow -2 51e+000 0 358 -3 82000 -1 200 -k -k -k -k Brown-Fl blue -2 07e+000 0 358 -3 38000 -0 751 •k-k-k-k Brown-Fl green -1 87e+000 0 358 -3 19000 -0 561 k k k k Brown-Fl orange -2 35e+000 0 358 -3 66000 -1 030 * k * k Brown-Fl red -2 40e+000 0 358 -3 71000 -1 090 * * -k -k Brown-Green 5 00e-003 0 358 -1 31000 1 320 Brown-Grey -1 95e-001 0 358 -1 51000 1 120 Brown-Nat guanine -1 08e+000 0 358 -2 39000 0 234 Brown-Orange -4 75e-001 0 358 -1 79000 0 839 Brown-Red -4 00e-002 0 358 -1 35000 1 270 Brown-Silver -2 04e+000 0 358 -3 36000 -0 731 -k-k k k Brown-Sparkle -2 95e-001 0 358 -1 61000 1 020 Brown-Uncoated -1 60e-001 0 358 -1 47000 1 150 Brown-White -2 53e+000 0 358 -3 84000 -1 220 -k-k-k-k Brown-yellow -1 .19e+000 0 358 -2 50000 0 129 Eggplant-Fl yellow -2 .47e+000 0 .358 -3 78000 -1 160 k k k k Eggplant-Fl blue -2 .03e+000 0 . 358 -3 . 34000 -0 711 •k "Jc -k ~k Eggplant-Fl green -1 .83e+000 0 . 358 -3 .15000 -0 . 521 k k k k Eggplant-Fl orange -2 . 31e+000 0 . 358 -3 .62000 -0 . 991 k k k k Eggplant-Fl red -2 . 36e + 000 0 . 358 -3 .67000 -1 .050 •k -k -k •k Eggplant-Green 4 .50e-002 0 . 358 -1 .27000 1 .360 Eggplant-Grey -1 .55e-001 0 . 358 -1 .47000 1 . 160 Eggplant-Nat guanine -1 .04e+000 0 . 358 -2 . 35000 0 . 274 Eggplant-Orange -4 .35e-001 0 .358 -1 . 75000 0 . 879 Eggplant-Red 2 .10e-016 0 .358 -1 . 31000 1 . 310 E g g p l a n t - S i l v e r -2 . 00e+000 0 . 358 -3 .32000 -0 . 691 •k-k-k-k Eggplant-Sparkle -2 . 55e-001 0 . 358 -1 . 57000 1 .060 Eggplant-Uncoated -1 .20e-001 0 . 358 -1 .43000 1 . 190 228 APPENDICES Eggplant-White -2 49e+000 0 358 -3 80000 -1 180 •k-k -k-k Eggplant-yellow -1 15e+000 0 358 -2 46000 0 169 F l y e l l o w - F l blue 4 45e-001 0 358 -0 86900 1 760 F l y e l l o w - F l green 6 35e-001 0 358 -0 67900 1 950 F l y e l l o w - F l orange 1 65e-001 0 358 -1 15000 1 480 F l y e l l o w - F l red 1 lOe-001 0 358 -1 20000 1 420 F l yellow-Green 2 5'2e+000 0 358 1 20000 3 830 kkkk F l yellow-Grey 2 31e+000 0 358 1 00000 3 630 kkkk F l yellow-Natguanine 1 43e+000 0 358 0 11600 2 740 k k k k F l yellow-Orange 2 04e+000 0 358 0 72100 3 350 k k k k F l yellow-Red 2 47e+000 0 358 1 16000 3 780 k k k k F l • y e l l o w - S i l v e r 4 65e-001 0 358 -0 84900 1 780 F l yellow-Sparkle 2 21e+000 0 358 0 90100 3 530 k k k k F l yellow-Uncoated 2 35e+000 0 358 1 04000 3 660 k k k k F l yellow-White -2 00e-002 0 358 -1 33000 1 290 F l yellow-Yellow 1 32e+000 0 358 0 01150 2 640 kkkk F l b l u e - F l green 1 90e-001 0 358 -1 12000 1 500 F l b l u e - F l orange -2 80e-001 0 358 -1 59000 1 030 F l b l u e - F l red -3 35e-001 0 358 -1 65000 0 979 F l blue-Green 2 07e+000 0 358 0 75600 3 380 k k k k F l blue-Grey 1 87e+000 0 358 0 55600 3 180 kkkk F l blue-Nat guanine 9 85e-001 0 358 -0 32900 2 300 F l blue-Orange 1 59e+000 0 358 0 27600 2 900 kkkk F l blue-Red 2 03e+000 0 358 0 71100 3 340 kkkk F l b l u e - S i l v e r 2 00e-002 0 358 -1 29000 1 330 F l blue-Sparkle 1 77e+000 0 358 0 45600 3 080 kkkk F l blue-Uncoated 1 90e+000 0 358 0 59100 3 220 kkkk F l blue-White -4 65e-001 0 358 -1 78000 0 849 F l blue-Yellow 8 80e-001 0 358 -0 43400 2 190 F l green-Fl orange -4 70e-001 0 358 -1 78000 0 844 F l green-Fl red -5 25e-001 0 358 -1 84000 0 789 F l green-Green 1 88e+000 0 358 0 56600 3 190 kkkk F l green-Grey 1 68e+000 0 358 0 36600 2 990 kkkk F l green-Nat guanine 7 95e-001 0 358 -0 51900 2 110 F l green-Orange 1 40e+000 0 358 0 08650 2 710 kkkk F l green-Red 1 83e+000 0 358 0 52100 3 150 kkkk F l g r e e n - S i l v e r -1 70e-001 0 358 -1 48000 1 140 F l green-Sparkle 1 58e+000 0 358 0 26600 2 890 kkkk F l green-Uncoated 1 71e+000 0 358 0 40100 3 030 kkkk F l green-White -6 55e-001 0 358 -1 97000 0 659 F l green-Yellow 6 90e-001 0 358 -0 62400 2 000 F l orange-Fl red -5 50e-002 0 358 -1 37000 1 260 F l orange-Green 2 35e+000 0 358 1 04000 3 660 kkkk F l orange-Grey 2 15e+000 0 358 0 83600 3 460 kkkk Florange-Nat guanine 1 27e+000 0 358 -0 04850 2 580 F l orange-Orange 1 87e+000 0 358 0 55600 3 180 kkkk F l orange-Red 2 31e+000 0. 358 0 99100 3 620 kkkk F l o r a n g e - S i l v e r 3 00e-001 0. 358 -1 01000 1 610 F l orange-Sparkle 2 05e+000 0. 358 0 73600 3. 360 kkkk F l orange-Uncoated 2 18e+000 0. 358 0 87100 3 500 kkkk F l orange-White -1 85e-001 0 358 -1 50000 1 130 F l orange-Yellow 1 16e+000 0. 358 -0 15400 2 470 F l red-Green 2 41e+000 0. 358 1 09000 3. 720 kkkk F l red-Grey 2 20e+000 0. 358 0. 89100 3. 520 kkkk F l red-Nat guanine 1 32e+000 0. 358 0. 00649 2 . 630 kkkk 229 APPENDICES F l red-Orange 1 92e+000 0 358 0 61100 3 240 * Jr * * F l red-Red 2 36e+000 0 358 1 05000 3 670 Jr Jr Jr Jr F l r e d - S i l v e r 3 55e-001 0 358 -0 95900 1 670 F l red-Sparkle 2 10e+000 0 358 0 79100 3 420 Jr * Jr Jr F l red-Uncoated 2 24e+000 0 358 0 92600 3 550 * * * Jr F l red-White -1 30e-001 0 358 -1 44000 1 180 F l red-Yellow 1 21e+000 0 358 -0 09850 2 530 Green-Grey -2 00e-001 0 358 -1 51000 1 110 Green-Nat guanine -1 09e+000 0 358 -2 40000 0 229 Green-Orange -4 80e-001 0 358 -1 79000 0 834 Green-Red -4 50e-002 0 358 -1 36000 1 270 Green-Silver -2 05e+000 0 358 -3 36000 -0 736 Jr Jr Jr Jr Green-Sparkle -3 00e-001 0 358 -1 61000 1 010 Green-Uncoated -1 65e-001 0 358 -1 48000 1 150 Green-White -2 54e+000 0 358 -3 85000 -1 220 Jr Jr Jr Jr Green-Yellow -1 19e+000 0 358 -2 50000 0 124 Grey-Nat guanine -8 85e-001 0 358 -2 20000 0 429 Grey-Orange -2 80e-001 0 358 -1 59000 1 030 Grey-Red 1 55e-001 0 358 -1 16000 1 470 G r e y - S i l v e r -1 85e+000 0 358 -3 16000 -0 536 Jr Jr Jr Jr Grey-Sparkle -1 00e-001 0 358 -1 41000 1 210 Grey-Uncoated 3 50e-002 0 358 -1 28000 1 350 Grey-White -2 34e+000 0 358 -3 65000 -1 020 * Jr Jr Jr Grey-Yellow -9 90e-001 0 358 -2 30000 0 324 Nat guanine-Orange 6 05e-001 0 358 -0 70900 1 920 Nat guanine-Red 1 04e+000 0 358 -0 27400 2 350 Nat g u a n i n e - S i l v e r -9 65e-001 0 358 -2 28000 0 349 Nat guanine-Sparkle 7 85e-001 0 358 -0 52900 2 100 Nat guanine-Uncoated 9 20e-001 0 358 -0 39400 2 230 Nat guanine-White -1 45e+000 0 358 -2 76000 -0 136 Jr Jr Jr Jr Nat guanine-Yellow -1 05e-001 0 358 -1 42000 1 210 Orange-Red 4 35e-001 0 358 -0 87900 1 750 Orange-Silver -1 57e+000 0 358 -2 88000 -0 256 Jr Jr Jr Jr Orange-Sparkle 1 80e-001 0 358 -1 13000 1 490 Orange-Uncoated 3 15e-001 •0 358 -0 99900 1 630 Orange-White -2 06e+000 0 358 -3 37000 -0 741 Jr Jr Jr Jr Orange-Yellow -7 10e-001 0 358 -2 02000 0 604 Red-Silver -2 00e+000 0 358 -3 32000 -0 691 Jr Jr Jr Jr Red-Sparkle -2 55e-001 0 358 -1 57000 1 060 Red-Uncoated -1 20e-001 0 358 -1 43000 1 190 Red-White -2 4 9e+00O 0 358 -3 80000 -1 180 Jr Jr Jr Jr Red-Yellow -1 15e+000 0 358 -2 46000 0 169 S i l v e r - S p a r k l e 1 75e+000 0 358 0 43600 3 060 Jr Jr Jr Jr Silver-Uncoated 1 88e+000 0 358 0 57100 3 200 Jr Jr Jr Jr S i l v e r - W h i t e -4 85e-001 0 358 -1 80000 0 829 S i l v e r - Y e l l o w 8 60e-001 0 358 -0 45400 2 170 Sparkle-Uncoated 1 35e-001 0 358 -1 18000 1 450 Sparkle-White -2 24e+000 0 358 -3 55000 -0 921 Jr Jr Jr Jr Sparkle-Yellow -8 90e-001 0 358 -2 20000 0 424 Uncoated-White -2 37e+000 0 358 -3 68000 -1 060 Jr Jr Jr Jr Uncoated-Yellow -1 03e+000 0 358 -2 34000 0 289 White-Yellow 1 34e+000 0 358 0 03150 2 660 Jr Jr Jr Jr 230 APPENDICES Appendix B - 2. Two-way analysis of variance (using S-Plus) for data from Bamfield Marine Station, with visible distance as outcome and light level (depth) and pellet type as the factors, for turbid water condition. With 18 different types of pellets, there were 153 comparisons. Significance is indicated by "****". Fl refers to fluorescent colours. Pellet type Estimate Std. Error Lower bound Upper bound Significance Blue-Brown 6 OOe--002 0 0636 -0 17300 0 29300 Blue-Eggplant -5 50e--002 0 0636 -0 28800 0 17800 B l u e - F l yellow 1 05e--001 0 0636 -0 12800 0 33800 B l u e - F l blue 2 lOe--001 0 0636 -0 02330 0 44300 B l u e - F l green 2 65e--001 0 0636 0 03170 0 49800 B l u e - F l orange 3 OOe--001 0 0636 0 06670 0 53300 B l u e - F l red 2 85e--001 0 0636 0 05170 0 51800 Blue-Green 2 50e--002 0 0636 -0 20800 0 25800 Blue-Grey 4 OOe--002 0 0636 -0 19300 0 27300 Blue-Nat guanine 1 65e--001 0 0636 -0 06830 0 39800 Blue-Orange 1 70e--001 0 0636 -0 06330 0 40300 Blue-Red -6 50e--002 0 0636 -0 29800 0 16800 B l u e - S i l v e r 1 55e--001 0 0636 -0 07830 0 38800 Blue-Sparkle 5 OOe--002 0 0636 -0 18300 0 28300 Blue-Uncoated 1 05e--001 0 0636 -0 12800 0 33800 Blue-White 1 50e--002 0 0636 -0 21800 0 24800 Blue-Yellow 2 OOe--001 0 0636 -0 03330 0 43300 Brown-Eggplant -1 15e--001 0 0636 -0 34800 0 11800 Brown-Fl yellow 4 50e--002 0 0636 -0 18800 0 27800 Brown-Fl blue 1 50e--001 0 0636 -0 08330 0 38300 Brown-Fl green 2 05e--001 0 0636 -0 02830 0 43800 Brown-Fl orange 2 40e--001 0 0636 0 00671 0 47300 Brown-Fl red 2 25e--001 0 0636 -0 00829 0 45800 Brown-Green -3 50e--002 0 0636 -0 26800 0 19800 Brown-Grey -2 OOe--002 0 0636 -0 25300 0 21300 Brown-Nat guanine 1 05e--001 0 0636 -0 12800 0 33800 Brown-Orange 1 lOe--001 0 0636 -0 12300 0 34300 Brown-Red -1 25e--001 0 0636 -0 35800 0 10800 Brown-Silver 9 50e--002 0 0636 -0 13800 0 32800 Brown-Sparkle -1 OOe--002 0 0636 -0 24300 0 22300 Brown-Uncoated 4 50e--002 0 0636 -0 18800 0 27800 Brown-White -4 50e--002 0 0636 -0 27800 0 18800 Brown-Yellow 1 40e--001 0 0636 -0 09330 0 37300 Eggplant-Fl yellow 1 60e--001 0 0636 -0 07330 0 39300 Eggplant-Fl blue 2 65e--001 0 0636 0 03170 0 49800 Eggplant-Fl green 3 20e--001 0 0636 0 08670 0 55300 Eggplant-Fl orange 3 55e--001 0 0636 0 12200 0 58800 Eggplant-Fl red 3 40e--001 0 0636 0 10700 0 57300 Eggplant-Green 8 OOe--002 0 0636 -0 15300 0 31300 Eggplant-Grey 9 50e--002 0 0636 -0 13800 0 32800 Eggplant-Nat guanine 2 20e--001 0 0636 -0 01330 0 45300 Eggplant-Orange 2 25e--001 0 0636 -0 00829 0 45800 Eggplant-Red -1 OOe--002 0 0636 -0 24300 0 22300 E g g p l a n t - S i l v e r 2 lOe--001 0 0636 -0 02330 0 44300 Eggplant-Sparkle 1 05e--001 0 0636 -0 12800 0 33800 Eggplant-Uncoated 1 60e--001 0 0636 -0 07330 0 39300 231 APPENDICES Eggplant-White 7 OOe- 002 0 0636 -0 16300 0 30300 Eggplant-Yellow 2 55e- 001 0 0636 0 02170 0 48800 F l y e l l o w - F l blue 1 05e- 001 0 0636 -0 12800 0 33800 F l y e l l o w - F l green 1 60e- 001 0 0636 -0 07330 0 39300 F l y e l l o w - F l orange 1 95e- 001 0 0636 -0 03830 0 42800 F l y e l l o w - F l red 1 80e- 001 0 0636 -0 05330 0 41300 F l yellow-Green -8 OOe- 002 0 0636 -0 31300 0 15300 F l yellow-Grey -6 50e- 002 0 0636 -0 29800 0 16800 F l yellow-Nat guanine 6 OOe- 002 0 0636 -0 17300 0 29300 F l yellow-Orange 6 50e- 002 0 0636 -0 16800 0 29800 F l yellow-Red -1 70e- 001 0 0636 -0 40300 0 06330 F l y e l l o w - S i l v e r 5 OOe- 002 0 0636 -0 18300 0 28300 F l yellow-Sparkle -5 50e- 002 0 0636 -0 28800 0 17800 F l yellow-Uncoated 2 53e- 016 0 0636 -0 23300 0 23300 F l yellow-White -9 OOe- 002 0 0636 -0 32300 0 14300 F l yellow-Yellow 9 50e- 002 0 0636 -0 13800 0 32800 F l b l u e - F l green 5 50e- 002 0 0636 -0 17800 0 28800 F l b l u e - F l orange 9 OOe- 002 0 0636 -0 14300 0 32300 F l b l u e - F l red 7 50e- 002 0 0636 -0 15800 0 30800 F l blue-Green -1 85e- 001 0 0636 -0 41800 0 04830 F l blue-Grey -1 70e- 001 0 0636 -0 40300 0 06330 F l blue-Nat guanine -4 50e- 002 0 0636 -0 27800 0 18800 F l blue-Orange -4 OOe- 002 0 0636 -0 27300 0 19300 F l blue-Red -2 75e- 001 0 0636 -0 50800 -0 04170 F l b l u e - S i l v e r -5 50e- 002 0 0636 -0 28800 0 17800 F l blue-Sparkle -1 60e- 001 0 0636 -0 39300 0 07330 F l blue-Uncoated -1 05e- 001 0 0636 -0 33800 0 12800 F l blue-White -1 95e- 001 0 0636 -0 42800 0 03830 F l blue-Yellow -1 OOe- 002 0 0636 -0 24300 0 22300 F l green-Fl orange 3 50e- 002 0 0636 -0 19800 0 26800 F l green-Fl red 2 OOe- 002 0 0636 -0 21300 0 25300 F l green-Green -2 40e- 001 0 0636 -0 47300 -0 00671 F l green-Grey -2 25e- 001 0 0636 -0 45800 0 00829 F l green-Nat guanine -1 OOe- 001 0 0636 -0 33300 0 13300 F l green-Orange -9 50e- 002 0 0636 -0 32800 0 13800 F l green-Red -3 30e- 001 0 0636 -0 56300 -0 09670 F l g r e e n - S i l v e r -1 lOe- 001 0 0636 -0 34300 0 12300 F l green-Sparkle -2 15e- 001 0 0636 -0 44800 0 01830 F l green-Uncoated -1 60e- 001 0 0636 -0 39300 0 07330 F l green-White -2 50e- 001 0 0636 -0 48300 -0 01670 F l green-Yellow -6 50e- 002 0 0636 -0 29800 0 16800 F l orange-Fl red -1 50e- 002 0 0636 -0 24800 0 21800 F l orange-Green -2 75e- 001 0 0636 -0 50800 -0 04170 F l orange-Grey -2 60e- 001 0 0636 -0 49300 -0 02670 F l orange-Nat guanine -1 35e- 001 0 0636 -0 36800 0 09830 F l orange-Orange -1 30e- 001 0 0636 -0 36300 0 10300 F l orange-Red -3 65e- 001 0 0636 -0 59800 -0 13200 F l o r a n ge-Silver -1 45e- 001 0 0636 -0 37800 0 08830 F l orange-Sparkle -2 50e- 001 0 0636 -0 48300 -0 01670 F l orange-Uncoated -1 95e- 001 0 0636 -0 42800 0 03830 F l orange-White -2 85e- 001 0 0636 -0 51800 -0 05170 F l orange-Yellow -1 OOe- 001 0 0636 -0 33300 0 13300 F l red-Green -2 60e- 001 0 0636 -0 49300 -0 02670 F l red-Grey -2 45e- 001 0 0636 -0 47800 -0 01170 F l red-Nat guanine -1 20e- 001 0 0636 -0 35300 0 11300 232 APPENDICES F l red-Orange -1 . 15e -001 0 .0636 -0 .34800 0 .11800 F l red-Red -3 . 50e -001 0 .0636 -0 . 58300 -0 . 11700 k k k * F l r e d - S i l v e r -1 . 30e -001 0 . 0636 -0 .36300 0 .10300 F l red-Sparkle -2 . 35e -001 0 . 0636 -0 .46800 -0 . 00171 * * * * F l red-Uncoated -1 . 80e-001 0 . 0636 -0 .41300 0 . 05330 F l red-White -2 . 70e -001 0 . 0636 -0 . 50300 -0 .03670 ^ ~k -k k F l red-Yellow -8 . 50e-002 0 . 0636 -0 .31800 0 .14800 Green-Grey 1 . 50e-002 0 .0636 -0 .21800 0 .24800 Green-Nat guanine 1 . 40e-001 0 . 0636 -0 .09330 0 .37300 Green-Orange 1 . 45e-001 0 . 0636 -0 .08830 0 .37800 Green-Red -9 . OOe--002 0 . 0636 -0 .32300 0 .14300 Green-Silver 1 . 30e-001 0 .0636 -0 . 10300 0 .36300 Green-Sparkle 2 50e--002 0 .0636 -0 20800 0 . 25800 Green-Uncoated 8 OOe--002 0 . 0636 -0 . 15300 0 .31300 Green-White -1 OOe--002 0 . 0636 -0 24300 0 .22300 Green-Yellow 1 75e--001 0 . 0636 -0 05830 0 .40800 Grey-Nat guanine 1 25e--001 0 0636 -0 10800 0 35800 Grey-Orange 1 30e--001 0 0636 -0 10300 0 36300 Grey-Red -1 05e--001 0 0636 -0 33800 0 12800 G r e y - S i l v e r 1 15e--001 0 0636 -0 11800 0 34800 Grey-Sparkle 1 OOe--002 0 0636 -0 22300 0 24300 Grey-Uncoated 6 50e--002 0 0636 -0 16800 0 29800 Grey-White -2 50e--002 0 0636 -0 25800 0 20800 Grey-Yellow 1 60e--001 0 0636 -0 07330 0 39300 Nat guanine-Orange 5 OOe--003 0 0636 -0 22800 0 23800 Nat guanine-Red -2 30e--001 0 0636 -0 46300 0 00329 Nat g u a n i n e - S i l v e r -1 OOe--002 0 0636 -0 24300 0 22300 Nat guanine-Sparkle -1 15e--001 0 0636 -0 34800 0 11800 Nat guanine-Uncoated -6 OOe--002 0 0636 -0 29300 0 17300 Nat guanine-White -1 50e--001 0 0636 -0 38300 0 08330 Nat guanine-Yellow 3 50e--002 0 0636 -0 19800 0 26800 Orange-Red -2 35e--001 0 0636 -0 46800 -0 00171 kkkk Orange-Silver -1 50e--002 0 0636 -0 24800 0 21800 Orange-Sparkle -1 20e--001 0 0636 -0 35300 0 11300 Orange-Uncoated -6. 50e-•002 0 0636 -0. 29800 0 16800 Orange-White -1. 55e-•001 0 0636 -0. 38800 0 07830 Orange-Yellow 3. OOe- 002 0 0636 -0. 20300 0 26300 Red-Silver 2 . 20e- 001 0. 0636 -0. 01330 0 45300 Red-Sparkle 1. 15e- 001 0. 0636 -0. 11800 0. 34800 Red-Uncoated 1. 70e- 001 0. 0636 -0. 06330 0. 40300 Red-White 8. OOe- 002 0. 0636 -0. 15300 0. 31300 Red-Yellow 2 . 65e- 001 0. 0636 0. 03170 0 . 49800 •kkk* S i l v e r - S p a r k l e -1. 05e- 001 0. 0636 -0. 33800 0. 12800 Silver-Uncoated -5 . OOe- 002 0 . 0636 -0 . 28300 0. 18300 Si l v e r - W h i t e -1. 40e- 001 0. 0636 -0. 37300 0. 09330 S i l v e r - Y e l l o w 4 . 50e- 002 0. 0636 -0. 18800 0. 27800 Sparkle-Uncoated 5. 50e- 002 0. 0636 -0 . 17800 0. 28800 Sparkle-White -3. 50e- 002 0. 0636 -0. 26800 0 . 19800 Sparkle-Yellow 1. 50e- 001 0. 0636 -0. 08330 0. 38300 Uncoated-White -9. OOe- 002 0. 0636 -0. 32300 0. 14300 Uncoated-Yellow 9. 50e- 002 0. 0636 -0. 13800 0. 32800 White-Yellow 1. 85e- 001 0. 0636 -0. 04830 0. 41800 233 APPENDIX C Above: Overview of farm site selected for field trial. Below: Electronic meters (i.e. a YSI model salinity meter and a dissolved oxygen meter). 234 APPENDIX 235 APPENDIX Above: Pellets being discharged into a cage surface with a good (>80%) spread. Below: Feed pellets coated with 18 different colours and materials to enhance their contrasts. 236 APPENDIX Above: Print showing the black shading material for controlling light levels in the laboratory. Below: Close-up of a 150 L fibreglass holding tank for experimental fish (rainbow trout). 237 APPENDIX 238 APPENDIX Above: Close-up of aquarium showing fish pulling on a test pellet and tilting the A-beam. Below: Close-up of the same fish pulling a test pellet and executing the S-shape manoeuvre. 239 APPENDIX • A b o v e : R a i n b o w trout at tacking a s i lver pellet when offered a s i lver and orange pellet. B e l o w : R a i n b o w trout at tacking a s i lver pellet when offered a s i lver and whi te pellet. 240 

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.831.1-0089259/manifest

Comment

Related Items