This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Benchmarking Growth Performance and Feed Efficiency
of Commercial Rainbow Trout Farms in Ontario, Canada
4.2.3 - Initial Body Weight ................................................................................................................. 53
4.2.4 - Final Body Weight ................................................................................................................... 53
4.2.5 - Days ......................................................................................................................................... 54
4.2.6 - Temperature ............................................................................................................................. 55
Table 4.1: Descriptive statistics of various production parameters, analyzed across sites. ......................................... 48 Table 4.2: Average site values for various production parameters .............................................................................. 49 Table 4.3: Average site values for various production parameters, calculated for each cohort year (ie. year in which
cohort or series of common lotgroups was initially stocked) .............................................................................. 50 Table 4.4: Average values for various production parameters, calculated using total lot and lotgroup values (ie. from
stocking to completion of harvest). ..................................................................................................................... 51 Table 4.5: Results of least squares analysis of fixed effect (ie. site) on global lot values for various performance
parameters ........................................................................................................................................................... 57 Table 4.6: Results of least squares means for multiple comparisons of fixed effect levels (ie. sites) for global lot
thermal-unit growth coefficient values.. ............................................................................................................. 57 Table 4.7: Results of least squares means for multiple comparisons of fixed effect levels (ie. sites) for global lot
mortality rates (MORTRATEHATCH) .............................................................................................................. 61 Table 4.8: Commercial Feed IDs, percent crude protein and crude lipid for each of the feed types served across sites
during the study period. ...................................................................................................................................... 63 Table 4.9: Rankings of each commercial site for various production parameters ...................................................... 71 Table 4.10: Comparison of the residual sums of squares between producer-estimated average body weights and
corresponding model predictions, the latter performed using either traditional or modified thermal-unit growth
Figure 4.1: Average lot initial inventory for each site ................................................................................................. 53 Figure 4.2: Box plots of final average body weights from all lots of each site. ........................................................... 54 Figure 4.3: Histogram of thermal-unit growth coefficients from lots of Site A and all other commercial sites
combined. ............................................................................................................................................................ 58 Figure 4.4: Box plots of lot thermal-unit growth coefficients from each site ............................................................. 58 Figure 5.5: Average lot mortality rates for each site.................................................................................................... 61 Figure 4.6: Median values and 10
th and 90
th percentiles of live interval mortality rates (ie. aggregated across sites,
divided into 50 day periods of grow-out) ............................................................................................................ 62 Figure 4.7: Median values of live interval mortality rates and corresponding water temperatures (ie. aggregated
across sites, divided into 50 day periods of grow-
out)……………………………………………………………………… ........................................................... 62 Figure 4.8: Box plots of lot economic feed conversion ratios for each site ................................................................ 65 Figure 4.9: Box plots of lot biological feed conversion ratios for each site.. .............................................................. 67 Figure 4.10: Median values and 10
th and 90
th percentiles of live interval biological feed conversion ratios (ie.
aggregated across sites, divided into 50 g body weight stanzas) ........................................................................ 67 Figure 4.11: Live interval biological feed conversion ratios, aggregated across sites. ................................................ 68 Figure 4.12: Box plots of lot nitrogen retention efficiency values for each site. ......................................................... 70 Figure 4.13: Live interval nitrogen retention efficiency values, aggregated across sites............................................. 70 Figures 4.14a and 4.14b: Producer-estimated average body weights from Site A and expected corresponding size
distribution ranges, plotted relative to model-predicted average body weights .................................................. 74
1
1 - GENERAL INTRODUCTION
The global production of fed aquatic species has increased at an annual rate of approximately
eight percent since 1980 (FAO, 2012), now generating almost 50% of all fish and shellfish consumed
globally (FAO, 2012). The global aquaculture industry is incredibly diverse, with over 250 species being
produced to varying scales by more than an estimated 500,000 operations. Management strategies (eg.
feeds and feeding strategies, genetic backgrounds, etc.) and environmental conditions (eg. water
temperature, dissolved oxygen, etc.) tend to vary from one site to the next. While industry growth is
expected to continue throughout the coming decades, aquaculture operations continue to face a number of
common challenges which collectively threaten the industry’s economic sustainability. Feed and
production costs continue to rise while product prices stagnate, diminishing operational profit margins.
Public scrutiny towards the safety of consumed aquatic products and potential environmental impacts
continues to increase. To ensure the sustainability of the aquaculture sector, solutions are needed to
produce more efficient, faster-growing, and disease-resistant animals that result in high-quality, safe, and
marketable products for consumers.
Decades of experience in other terrestrial livestock sectors have demonstrated the value in the
systematic, standardized, industry-wide recording and evaluation of animal performance (ie. animal
recording systems). Standardized animal recording systems permit systematic, reliable comparisons of
performance achieved by animals within and across operations, a function referred to as benchmarking.
Through comparing various performance parameters of one operation against those of its competitors,
benchmarking facilitates identification of operational strengths and weaknesses as well as the levels of
performance achieved across an industry (Giacomini, 2008; McDougal, 2012). Benchmarking ultimately
provide producers with the context needed to understand the true extent of their operational performance.
Given the immense diversity present in global aquaculture operations, there exists tremendous potential
value for aquaculture producers in performance benchmarking.
2
The Ontario rainbow trout industry provides a good example of the operational diversity
characterizing aquaculture. Of the 4,060 total annual tonnes produced in Ontario, approximately eighty
percent is raised on a small number of open-water cage sites situated in the North Channel and Georgian
Bay of Lake Huron. Across these sites there is considerable diversity in management strategies and
environmental variables (eg. feeds and feeding strategies, stocking and harvest sizes, inventory
adjustment schemes, water temperature profiles, etc.). Performance measuring and recording techniques
currently used in Ontario are highly variable amongst producers, particularly with respect to body weight
and temperature sampling. As a result of this operational diversity and the resulting difficulty in merging
idiosyncratic datasets there has to date been no attempt to survey commercial performance data across
Ontario. In absence of information, there has been no systematic evaluation of trends in trout
performance, thus precluding any attempts at performance benchmarking. Producers thus lack
perspective as to the breadth of trout performance achieved across the industry, and thus their own
position along this spectrum. Therefore, there is an immediate need within the Ontario rainbow trout
industry for systematic surveying and evaluation of commercial performance data.
1.1 - Objectives
The general objective of this thesis was to carry out a preliminary, systematic survey of
performance data from Ontario trout cage culture operations and to examine variability within and across
production sites. In addition to this, preliminary analysis of commercial data with current mathematical
growth and linear mixed models will be performed. These exercises would also serve to highlight the
challenges associated with the collection and analysis of idiosyncratic commercial farm data. With an
inventory of the functional limitations of the collected data, this thesis would then be able to offer
recommendations to the industry for the refinement and standardization of recording methods to improve
upon future data quality and its suitability for evaluative models and benchmarking schemes.
3
2 - LITERATURE REVIEW
2.1 - Introduction
The global production of fed aquatic species has increased at an annual rate of approximately
eight percent since 1980, now generating approximately 50% of all fish and shellfish consumed globally
(FAO, 2012). Aquaculture is a young and globally diverse industry currently comprised of more than 250
species produced by more than 500,000 small, medium, and large operations around the World (Rana,
1997). Production strategies vary significantly within and amongst these species and operations, with
differences in feed types and feeding strategies, scales of production, culture practices, genetic
backgrounds, and environmental conditions. The industry is expected to continue its growth throughout
the coming decades (FAO, 2012). However, there are many common challenges to the profitability of the
varied operations that are increasingly compromising the industry’s economic sustainability. The costs of
feed and production for aquaculture operations are ever increasing, in most cases without proportionate
increases in product prices and/or productivity, restricting operational profitability. Meanwhile, there is
increasing public scrutiny as to the safety of consumed aquatic products and the environmental effects of
their production, limiting their market value and creating conflicts between multiple users of a given body
of water.
2.2 - Overview of Rainbow Trout and Its Production in Ontario
Rainbow trout is the most extensively cultivated cold freshwater species in the World, with global
2011 production estimated at approximately 770, 000 metric tons, or $3.84 billion (FAO, 2013). Since
the early 1980s, when the cage culture of trout was first introduced to Lake Huron, Ontario has produced
the majority of rainbow trout (Oncorhynchus mykiss) in Canada. While a variety of production systems
(ie. cages, tanks, raceways, and ponds) contribute to the sum of Ontario’s production, approximately 80%
4
of the 4,060 total annual tonnes is raised on a small number of cage farms (Statistics Canada, 2011),
where trout are raised from fingerlings to market size. Ontario cage systems provide superior water
quality and historically suitable water temperature profiles, permitting greater stocking densities and
ultimately greater farm profitability.
Rainbow trout are raised to a market size of approximately one to two kilograms, after which the
majority are processed and sold as fillets in Ontario and the US. Trout are fed mostly extruded pellets,
sometimes formulated to contain levels of nutrients specific to a given life stage or production strategy;
trout diets distributed in Ontario may contain anywhere from 34-52% protein and 16-26% lipids. Feed is
manufactured mostly by a company in Ontario and a small group of companies located on the East Coast.
Feed companies use an ever-changing blend of fisheries-, animal by-product-, and plant-based ingredients
(eg. fish meal, fish oil, poultry by-products, feather meal, wheat, corn gluten meal, etc.), the relative
proportions of each depending on a number of physical, biological, and economic considerations.
There is considerable diversity in management strategies and culture environments within and
across Ontario sites. The open-water cages in which Ontario rainbow trout are raised, situated in the
North Channel and Georgian Bay of Lake Huron, experience extreme fluctuations in water temperatures
and environmental conditions over time and at varying depths and locales. Fingerlings are purchased
from a number of private hatcheries which have, for the most part, independently developed their
breeding programs. Fingerlings are stocked and market fish harvested each over a wide range of average
body weights. Stocking densities and inventory adjustment schemes (eg. grading, splitting) exhibit
marked differences amongst sites. Distributed feeds tend to vary amongst sites in ingredient and nutrient
compositions, while feeding strategies also differ (eg. number of times per day, hand vs. mechanized
feeding, feeding guidelines, etc.). Such varying input mixes are likely to result in substantial differences
in trout performance (eg. growth rates, feed efficiency, and potentially mortality rates) within and
amongst sites.
5
The monitoring and evaluation of performance variables is inherently more challenging when
raising animals in open bodies of water. For instance, determining weight distributions within a cage is
complicated when only a small proportion of the animals can be seen or weighed at any given time.
Similarly, determining the amount of feed consumed within a cage is challenging when the amount of
uneaten feed cannot be observed or weighed. Water temperatures fluctuate constantly within an Ontario
trout cage, due for example to the presence of thermoclines and the movement and mixing of water that
occurs within such large bodies of fresh water as Georgian Bay. Temperature extremes also vary greatly
from one season to the next, with surface waters typically frozen in the winter and reaching maximum
temperatures of 20-25°C in the summer.
With such challenges to the accurate monitoring of relevant performance parameters, Ontario
trout producers are believed to have developed recording methods that function to serve their own needs
in site management and that are compatible with the tools and resources available to them. For example,
producers are suspected to sample water temperatures at varying depths and at varying times of day, as
permitted by their own schedules. There are also a variety of methods used within the industry for
estimating the average body weight of a cage, some of which are believed to result in selection bias. For
instance, some producers offer small amounts of feed into their cages during body weight sampling to
entice trout to the water’s surface, at which point the fish are captured with a dip net for subsequent
weighing. It is believed that feed enticement might select for more aggressive and thus larger fish (Gord
Cole, personal communication, July 2011).
2.3 - Challenges for the Ontario Trout Industry
Although market prices for trout have increased somewhat in recent years, they have for the most
part remained stagnant since the 1980s, in part due to competition from the extensively produced Atlantic
salmon. In fact, prices for trout were lower from 2001-2005 than from 1996-2000 (Cummings et al.,
6
2007). This coupled with rising feed and production costs have contributed to diminishing producer
profit margins. As a result, improvements to productivity (eg. growth rates) and feed efficiency are goals
common to all producers.
Optimizing management strategies to achieve maximum fish performance is complicated in open-
water cage settings by a range of dynamic biological and environmental variables and their interaction.
That is, a number of exogenous and endogenous factors (eg. water temperature, feed composition,
species, strain, life stage, etc.) and other management variables (eg. feeding strategies, stocking size,
stocking density, etc.) contribute and interact to influence fish performance at any one time. The effects
of such factors and management variables and their interactions on fish performance have thus been the
focus of significant research efforts across the World (Azevedo et al., 2004a; Dumas et al., 2007a; Dumas
et al., 2007b; Encarnação et al., 2006; McKenzie et al., 2012; Overturf et al., 2012; Wilkinson et al.,
2010).
Knowledge gleaned from these types of experimentally-based studies has helped global
aquaculture producers achieve substantial gains to the productivity and feed efficiency of their operations
(Asche et al., 1999; Muir, 2005; Naylor et al., 2009). However, producer application of these scientific
advancements is complicated in commercial settings by an array of biological and environmental factors
that interact to affect or limit the intended outcomes or benefits of such knowledge transfer. Ontario
producers thus need solutions that are tailored to the idiosyncrasies of their own production environments.
Ideally, commercial-scale trials comparing production systems, fish strains, feeds, and feeding strategies
could be performed locally as needed to better support farm management decisions. However,
commercial-scale studies would be impractical and cost-prohibitive (Parker, 1998). Therefore, solutions
are needed to continuously improve performance and productivity of trout culture operations in a manner
which is responsive to dynamic and fluctuating production environments and market demands.
The consumer market for Ontario trout demands a product of a certain size and quality, available
on a year-round basis. Producers tend to target specific market weights to meet the demands of their
7
buyers and to capture particular price points. For producers to provide processing plants with expected
dates of harvest, thus enabling processing plants to coordinate harvest schedules amongst cage culture
operations to ensure a steady flow of product to the marketplace, producers will often attempt to forecast
(ie. model) the growth of their fish throughout the course of grow-out periods. However, failure to
achieve modelled growth targets and anticipated harvest schedules often result in producers having to
prematurely harvest their fish in order to provide their processors, and ultimately the marketplace, with a
timely product. Early harvests might come at a cost to the producer through smaller trout sizes and thus
lower product prices. Therefore, there is a need for refinement of current growth models using
commercial data. Furthermore, while larger fish will typically sell at a higher price, the most profitable
harvest weight is by no means certain. Experimental evidence suggests that feed efficiency decreases as
trout approach maturation or market size (Azevedo et al., 2004), although the extent of this process in
commercial settings is not known. As such, exploration of longitudinal trends in commercial feed
efficiency data is now needed in order to support optimization of this trade-off between market prices and
feed efficiency.
With the diversity in management strategies and environmental conditions observed amongst
Ontario cage culture operations, there is expected to be similarly large variability in performance
parameters within and across sites. While producers track the performance of their own fish throughout
each production cycle, there has been little collaborative effort and sharing of data amongst producers in
this regard. As a result, there are no industry standards for the performance traits which they hope to
improve, and thus no understanding of their own performance relative to these standards. A systematic
survey of performance data from Ontario cage culture operations is now needed. Not only would this
permit the aforementioned investigations into longitudinal growth and feed efficiency data, but would
most importantly offer producers insight as to the variability in performance experienced across the
industry. With access to industry standards and ranges in performance parameter values across sites,
producers would have yardsticks against which they could compare their own production performance.
Such across-site performance comparisons (ie. benchmarking) would provide producers with immediate
8
perspective and context as to the true level of their performance, ultimately helping to guide and prioritize
management interventions and the efficient use of limited resources.
2.4 – Benchmarking
The comparison of organizational performance against that of competitors and/or accepted standards
and modelled estimations is a function known as benchmarking (Giacomini, 2008; McDougal, 2012).
Such comparative analyses provide important perspectives for organizations. That is, while most
producers will evaluate their own performance internally, their true performance might not be fully
understood or appreciated until compared with that of others outside of the organization (Bilbrey, 2012).
Benchmarking serves to highlight both the strengths and weaknesses of an organization, helping
decision-makers identify their operational areas most likely to gain from management intervention
(Giacomini, 2008). Typically, producers are provided with minimum and maximum levels for a range of
indicators (eg. biological, economic, etc.) and their position or ranking along this spectrum (Giacomini,
2008). Benchmarking thus serves as an unbiased analytical tool for identifying the facets of operation
with the greatest economic potential for improvement (Giacomini, 2008). It is a systematic data-yielding
process whereby producers continuously isolate and identify superior performance either from within or
outside the organization, striving to understand and adapt the practices contributing to this performance
(American Productivity and Quality Centre, 1997).
When applied in the context of established standardized recording and evaluation frameworks,
benchmarking can be an effective analytical tool for identifying an organization’s strengths and
weaknesses, and thus improving its performance incrementally over time (McDougal, 2012). A common
concern with benchmarking is that there is limited value in comparisons amongst firms with idiosyncratic
performance recording methods or with differing organizational structures and management strategies
9
(Parker, 1998). In these instances, the effectiveness of strategy analysis might be limited by differences
amongst organizations in the relative value of their inputs or by misrepresentation of relative values
through irregular or non-standardized recording and evaluation schemes (Parker, 1998). Thus, to enable
“apples to apples” comparisons and effective benchmarking, it is essential that not only the methods used
to measure and record performance parameters be consistent amongst organizations, but so too the
models and methods for evaluating and comparing performance (McDougal, 2012).
2.5 - Animal Recording Systems
There are many challenges to accurately monitoring and evaluating the performance of aquatic
animals and their highly variable environmental conditions. As a result, recording methods of such
industries as the Ontario trout cage culture industry comprise substantial differences amongst operations.
Differences amongst cage culture operations in recording methods and management practices (eg. genetic
backgrounds, stocking densities, feeding practices, timing of initial cage stocking, lengths of grow-out
periods, etc.) pose substantial challenges to systematic and effective performance benchmarking.
However, such challenges in irregular recording methods and non-standardized datasets are not unlike
those that once limited improvements to the performance and economic viability of terrestrial livestock
industries in the former half of the 20th century. The collection of early genetic information on dairy
animals began over 125 years ago with the establishment of breed associations (Agriculture and Agri-
Food Canada, 2009). While breed associations were aware of the potential for identifying superior
genetics, they were knowingly limited by a lack of established methods for measuring animal
characteristics that would permit proper ranking of animals over time (Black, 1936). A seminal paper on
the state of livestock was anonymously published in 1935, outlining the factors most limiting livestock
improvement and breeding, “paralyzing movement toward any practical goal” (Harris, 1998). These
factors are likely limiting the effective performance benchmarking of aquaculture operations today:
10
The use of standards that are incomplete and in some cases inaccurate.
Lack of yardsticks to supplement or revise existing standards.
Large gaps in the knowledge of animal genetics.
Factors limiting experimentation, namely costs.
Following WWII, the newly created Food and Agriculture Organization (FAO) sought to
overcome these limitations, capturing the potential value in global standards for animal recording
methods and the resulting wealth of data from subsequent collaboration amongst breeders. Shortly
thereafter, the FAO thus created the International Committee for Animal Recording (ICAR) to realize
these benefits. With a primary goal of comparing animal performance across regions and identifying
superior genetics, the committee, comprised of participating countries’ many breed organizations and
academic institutions, was to mobilize and exchange available expertise and address technical issues
limiting the adoption of standardized recording systems for dairy cattle (Rosati, 2011). ICAR has since
served as a platform for sharing the learned experiences of member organizations in performance
recording and evaluations (Boulesteix et al., 2004; Cromie et al., 2008; Wickham, 2008), benchmarking
(Baier, 2008; Giacomini, 2008), and genetic improvement services (Norman et al., 2008; Schaeffer, 2008;
Woodward and Van der Lende, 2008). By providing requirements to ensure a satisfactory level of
uniformity in recording and evaluation techniques, ICAR has enabled and supported decades of work in
performance comparisons and genetic evaluations that have been essential to the concurrent
improvements in animal performance and productivity.
As summarized by the FAO (1998), “animal recording is a systematic process that leads to
outcomes that facilitate a comparison of production alternatives [ie. benchmarking], the availability of
baseline information on the performance of animals, animal management decisions and genetic
improvement that are beneficial to the governments and policy-makers, producers and by extension the
consumers.” The development of recording systems for extensive animal production sectors occurred
11
over decades in conjunction with that of other production technologies (Holst, 1999; Djemali, 2004). The
development of recording systems provided the foundation and structure for concurrent development and
dissemination of almost all high productive cattle and swine breeds raised today (Bougler, 1990; Harris,
1998). As a result of these successes, ICAR’s portfolio of species has since expanded to include beef
cattle, sheep, goat, and buffalo (Rosati, 2011); ICAR is now represented by 87 member organizations
(eg. breed associations, producer cooperatives, academic institutions, etc.) from 51 countries (Rosati,
2011).
An animal recording system is a critical tool for farm management, comprised of a series of
systematic processes involving the collection of information on animals, entry of data into a standardized
database, data processing and evaluation, and the interpretation and distribution of results (Flammant,
1998). Such interpretation of results is typically done in the context of comparing multiple operations and
their various production scenarios, improving management or optimization of production inputs,
identifying animals to be bred for more productive future generations, and/or other related extension
services (FAO, 1998; Flamant, 1998; Wasike et al., 2011). All such components of recording systems
can be performed by one or many organizations, with any combination of private, public, or academic
entities fulfilling these roles.
It is well established that animal recording systems are the primary technologies driving
extension services (eg. benchmarking, genetic improvement, etc.) for continuous and sustainable
improvements to the productivity and profitability of animal production sectors (Mackechnie, 1995).
There are many examples of recording systems and benchmarking programs established across the World
that have successfully demonstrated their usefulness in improving the economic viability of industries.
The Canadian dairy industry established its Record of Performance (R.O.P.) program in 1905 to permit
unbiased performance evaluations. This benchmarking program has since expanded in popularity to
include recording of over 70% of dairy cattle raised in Canada today (Agriculture and Agri-Food Canada,
2009). LAMPLAN is a genetic evaluation system involved with the recording of over 100,000 Australian
12
meatsheep per year (Banks and Kinghorn, 1997). The program began with a single district officer
measuring growth rate and backfat thickness, processing the data, and providing advice to producers
(Harris, 1985). Boer goat is bred for its meat in South Africa and distributed internationally, the
successes of its production tied to genetic improvement enabled by a standardized recording system and
database framework (Holst, 1999). The French dairy sheep industry employs ICAR procedures in its
improvement program, achieving annual gains in milk yield of up to 2.4%, values similar to those of
dairy cattle (Barillet et al., 1996; Holst, 1999).
With the widespread adoption and successes of animal recording systems and enabled extension
services, there is great incentive for producers to invest in such frameworks. However, economic
incentives have not always proven sufficient for such investments, as recording systems typically require
significant human, economic, and technical capital, and provide relatively minimal short-term economic
returns (Garrick and Golden, 2008; Geroski, 1995; Wasike et al., 2011). For example, employees must be
trained in any implemented measurement and recording techniques and indoctrinated in the culture of
continuous improvement. Upgrades to infrastructure are required to comply with established performance
recording protocols, and technicians must be hired to train and review on-farm recording techniques, to
update and maintain computer databases, and to process and evaluate commercial data for the purposes of
benchmarking and other such extension services (Baier, 2008).
The preliminary adoption and establishment of recording systems and evaluative frameworks
within an animal sector is further challenged by the distinct and unique characteristics amongst producer
methods for measuring and recording performance. In absence of an established animal recording
system, producers develop methods that are compatible with the resources available to them and that
serve their own objectives in day-to-day operations (Wasike et al., 2011). Such idiosyncrasies often result
in records appearing “incoherent” relative to one another, complicating not only the aggregation of
datasets into standardized databases, but also the systematic process of treating data and applying it to
evaluative models and frameworks (Wasike et al., 2011). As producers continue to function within their
13
own unique recording schemes in lieu of established protocols, differences in recording methods and
dataset structures are reinforced and added to over time, further challenging the eventual merging of past
records (Garrick and Golden, 2008; Mocquot et al., 2004; Parker, 1998; Wasike et al., 2011).
Inconsistencies or gaps in datasets may also lead to the loss of valuable data following their
integration into a standardized database, limiting the short-term value of extension services and
improvement programs (Garrick and Golden, 2008). For example, the National Beef Consortium
attempted to standardize database frameworks for across-herd evaluations amongst breed associations in
the early 2000s (Garrick and Golden, 2008). Its hope was to transition the sector’s “data-driven”
approach to improvement (ie. adding parameters to databases as they are needed) to a goal-driven one, for
example by including parentage information in all frameworks for across-breed EPD evaluations (Garrick
and Golden, 2008). Without standardized registration systems across breed associations, it was
determined that multiple identification codes for individual parents would preclude the complete use of
their data in evaluations, leading to wasted information (Garrick and Golden, 2008). Thus far, a lack of
leadership and common objectives amongst breed organizations has indeed prevented homogenization of
registration systems and complete use of information.
Such obstacles to the adoption of standardized animal recording systems have been partially
overcome in the past through collaborations between breeder associations and/or producer cooperatives
with academic institutions, beginning in the 1970s (Harris, 1998; Zimmerman, 2008). Such
collaborations provide mutual benefits in that researchers gain access to valuable swaths of commercial
data on which to apply and further develop their breeding techniques, statistical models, etc., while
producers in turn benefit from these academic advancements and their practical application. With such
functions being performed by academic institutions, and also with governments occasionally providing
funding for the costs of framework implementation (eg. animal recording infrastructure, integration of
datasets into standardized databases, commercial-scale growth trials, etc.), such collaborations often help
minimize the initial financial barriers that may discourage producers from participating in animal
14
recording systems (Garrick and Golden, 2008). In the past, swine breed associations have partnered with
federal research services and Purdue University (Stewart et al., 1991), while American dairy coops and
the US National Beef Consortium have collaborated with four separate land-grant universities (Garrick
and Golden, 2008; Zimmerman, 2008). The University of Guelph had for decades performed genetic
evaluations for the conformation traits of dairy cattle in coordination with the provincial government and
the Holstein Association of Canada (Lazenby and Stanley, 1997). This partnership enabled substantial
improvements to the methodologies and statistical procedures for genetic improvement strategies (eg.
linear mixed models, random regressions, test day models, etc.).
2.6 - Benchmarking and Animal Recording Systems in Aquaculture
The state of standardized performance recording and evaluation in aquaculture is not unlike that of
livestock in the former half of the 20th century. Indeed, the factors discussed as limiting livestock
improvement in 1935 (ie. the use of standards that are incomplete, lack of yardsticks to supplement or
revise existing standards, etc.) are today limiting that of global aquaculture. For most regions, species,
and/or production systems there has been little to no attempt to standardize performance recording and
evaluation methods. There exist few commercial yardsticks or benchmarks for any culture variety.
National or interregional databases for performance data do not exist, perhaps outside of the Federation of
European Aquaculture Producers (F.E.A.P.). Where performance recording systems do exist, they are
typically established by a company as a proprietary tool for internal benchmarking and production
planning.
The lack of benchmarking programs and animal recording systems in aquaculture is due largely
to a few main factors. First, aquaculture is a relatively young industry. For example, the Norwegian
salmon farming industry was one of the first to develop modern intensive cage culture operations, the
advancement of its practices initiated only in the 1970s (Coull, 1993). The animal recording systems used
15
in extensive terrestrial livestock sectors are, alternatively, the result of over a century of progress in the
field; progress paralleled by concurrent industry growth. Secondly, aquaculture is the most diverse form
of animal production in the World with over 250 species cultured (Rana, 1997). As has occurred with
terrestrial livestock sectors, local and/or regional differences in producer goals and objectives has limited
the sorts of interregional collaborations in aquaculture that are required for standardization of recording
systems, database frameworks, and evaluation schemes. Finally, the development of national breeding
programs has been limited in aquaculture, with only 8.2% of 2010 global production derived from
improved stocks (Gjedrem et al., 2012; Neira, 2010; Rye et al., 2010). While breeding programmes are
not the only motivation for establishing standardized recording systems, they indeed necessitate and are a
large impetus for such efforts (Garrick and Golden, 2008; Rosati, 2011).
The diversity characterizing aquaculture – both of cultured species and production environments – is
both a limiting factor and impetus for performance benchmarking and the adoption of standardized animal
recording systems. Similar to the present extent of diversity in genetics, culture techniques, and
environmental conditions within global aquaculture is the extent of diversity in its recording and
evaluation systems. This variety in recording methods not only complicates the merging of past datasets
into standardized databases, but also the process of systematically treating and applying data to
standardized models and frameworks for performance evaluations and benchmarking (Wasike et al.,
2011). However, the diversity amongst production and management environments also creates
tremendous potential value for performance benchmarking. For every cultured species there is
considerable global variety in production techniques and animal performance, providing vast assemblages
of benchmarks and alternative management scenarios against which producers can compare their own
performance, isolate superior management strategies, and eventually identify superior genetics. As was
captured long ago in other animal sectors, the potential value in recording systems and comparative
benchmarking must now be captured in aquaculture.
16
There have to date been relatively few published accounts of performance benchmarking in
aquaculture. Bolton-Warberg and FitzGerald (2012) performed preliminary analyses on growth
trajectories of different North Atlantic cod strains to compare prospective value for producers. They
assessed and compared such simple parameters as days-post-hatch to harvest weight, and developed
growth models to project growth trajectories of the various cod strains in a range of temperature profiles.
Soares et al. (2011) developed a system for benchmarking weekly mortality rates on Scottish Atlantic
salmon farms, applying the parameter as an indicator of animal health. Production data has been used for
examination of economic functions and optimizations, risk analysis and related functions (Guttormsen,
2002), analysis of inefficiencies (eg. Asche et al., 2009), and for guiding of policy development (Asche,
1997), but little has been done in terms of practical benchmarking of performance criteria.
While the private sector has begun providing services and technologies in performance recording
and management (eg.Fishtalk™ by Akva (Europe), Aquabench® (Chile)), benchmarking services remain
limited or simplistic in their capabilities. Furthermore, corresponding methodologies and results of
benchmarking efforts are not made available to the public. Documentation of experiences and challenges
in the preliminary performance benchmarking of aquatic species is valuable to the global industry given
the ubiquitous variability in current recording and evaluation methods and the inevitable complications
arising from benchmarking of idiosyncratic datasets. Cataloguing of lessons and experiences amongst
ICAR members has been fundamental to their development of recording systems and methodologies for
benchmarking functions (Guellouz et al., 2004; Sattler, 2008), ensuring advancement of the most current
state of the art.
2.7 - Thesis Objectives
There is a need for performance benchmarking of the Ontario trout industry to provide producers
perspective as to the range of values achieved across the industry and their own performance relative to
17
this. As such, the initial objective of this thesis was to perform a preliminary, systematic survey of
commercial performance data for Ontario trout and to examine variability within and across operational
sites. This thesis would also initiate the application of data to current mathematical growth and
nutritional models in order to explore the potential for model refinement with commercial data. In
addition to this, preliminary evaluation of longitudinal (ie. time-series) data for various performance
parameters (eg. feed efficiency, mortality rates, etc.) would provide further insight into trends in
commercial data and the limitations of the dataset when applied to evaluative models. These exercises
would thus serve to illustrate the functional shortcomings of the current Ontario dataset and the highly
irregular sampling and recording methods from which it was derived. With a working knowledge of the
dataset’s practical limitations, this thesis would then conclude with recommendations to Ontario
producers as to the adjustment and standardization of their recording methods in order to enable
continuous improvement to the quality and value of the commercial dataset for purposes of performance
benchmarking.
18
3 – MATERIALS AND METHODS
*Glossary provided in Appendix 1.
3.1 - Data Survey
Data was gathered from five commercial open-water cage sites located in Georgian Bay, Ontario,
from September 2008 to June 2012. Included in the dataset were ancillary values reported by affiliated
hatcheries and processors/processing plants over the same period of time. Data from one experimental
cage site (Experimental Lakes Area (ELA)), in operation from 2003 to 2007, was also included in the
analysis for comparison.
3.2 - Database Structure and Organization
The preliminary surveying and evaluating of performance data from Ontario trout cage culture
operations was the primary goal of this thesis. This being a novel effort, it was necessary to first design
and construct a database framework in Microsoft Excel for the storage and evaluation of data, and to
develop associated methodologies for the systematic refinement and incorporation of data into this
framework. Methods for recording and evaluating trout performance in Ontario have never been
systematically coordinated for means of benchmarking and animal improvement. The survey’s resulting
dataset thus provides a valuable exercise in testing the adaptability of the database framework to variable
production settings and in applying the dataset to integrated mathematical growth and nutritional models
(eg. Fish-PrFEQ system (Cho and Bureau, 1998)), any associated challenges providing insight into
practical limitations of the dataset and current commercial recording methods. Protocols for treating and
19
incorporating into the database the many idiosyncrasies amongst producer sampling and reporting
methods will be discussed where appropriate with descriptions of performance variables below.
Ontario trout producers often report data at different time intervals (eg. daily, monthly). To reconcile
these differences and to facilitate evaluations at varying time scales (eg. descriptive summary statistics,
time-series models, etc.), the performance database was structured into three spreadsheet formats, each
representing production lots at a different time scale:
1) Daily Format (DAILYFORM), in which each day of grow-out (ie. stocking to harvest) is
designated its own row/entry. Each site is designated its own DAILYFORM spreadsheet.
2) Management Event Format (MANEVFORM), in which data is summarized for intervals of time
between average body weight sampling events, fish movement events (eg. transfers, size grades,
etc.), and hatchery- or processing plant-reported body weights. Each event is designated its own
row/entry and each site its own MANEVFORM spreadsheet.
3) Summary Format (SUMFORM), in which data is summarized in one row/entry for the complete
grow-out stage of a given lot (ie. “global” values, from cage stocking to harvest). Data from all
sites is compiled into one SUMFORM spreadsheet.
3.3 - Use of Excel Functions
A number of Excel functions were used to facilitate inter-format functioning and compatibility.
For example, ROUND was used to maintain consistency amongst formats in number of decimals. Other
important functions will be discussed as appropriate with descriptions of performance variables below.
3.4 - Experimental Units (Lots and Lotgroups)
In Ontario cage culture, trout are typically stocked in relatively large numbers into one cage and
subsequently split or graded (ie. ordered into separate cages according to specific size class/grade) into
20
multiple other cages as the fish increase in size. There were thus two classes of production units in this
thesis considered as “experimental units” for purposes of performance evaluation: “lots” (ie. cages, from
initial stocking to harvest), for which fish are tracked and evaluated through a single cage over time, and
“lotgroups” that had once comprised a common singular cage at time of stocking and are evaluated as a
singular aggregate of lots as they are split and redistributed over time. To summarize certain variables by
lot from stocking to harvest, a number of corollary assumptions and data modifications were required to
utilize data from periods of time in which multiple separate lots shared a singular cage. For example, if a
cage was graded into two empty cages whose data was to be summarized on a per lot basis, feed and
mortality totals from the initial period of time as one shared cage had to be divided amongst its
succeeding two lots. Periods of time before or between fish movement events will subsequently be
referred to as “inter-movement periods.” To verify and/or validate results across lots, variables were thus
also assessed across aggregate lotgroups for comparison, the calculation of which did not require
accessory assumptions and data alterations.
3.5 - Commercial Lot ID Codes
Unlike with other animal production systems (beef, dairy, swine, sheep, etc.), fish cannot be cost-
effectively identified individually on a commercial scale. Fish were thus identified by a) the country in
which their grow-out cycle occurred, b) their species, c) their hatchery/strain, d) the site of their grow-out,
and e) the cage(s) or tank number(s) in which they were held during grow-out (ie. from stocking as
fingerlings to harvest at market size). To explain the commercial lot ID system, an example ID code from
the Ontario dataset will be provided and described sequentially in a stepwise manner, beginning with
identification of the country in which fingerlings were raised. Unless explained otherwise, all listed
hatchery and site alphanumeric codes were unique to this system, developed and assigned to
21
hatcheries/sites as part of this thesis.
Example ID Code:
124-OM-E2E3E1-C05093-2-5-6
Piecemeal description of ID code:
“124”
As is used by ICAR, the ISO 3166-1 numeric system was used by this system to identify country
(ie. of grow-out). “124” is the code for Canada.
“124-OM”
Species code was listed after country code, “OM” representing Oncorhynchus mykiss (rainbow
trout).
“124-OM-E2-C05093”
Fingerlings purchased for grow-out were sold by a hatchery designated as “E2,” “E” being the
letter assigned to the company name, “2” being the company’s specific hatchery site.
Fish were stocked into Site C in May of 2009 (C0509). Month and year were listed directly after
site letter(s).
Fish were stocked initially into cage 3, as recorded by the site technician (C05093).
Tank/lot/cage numbers assigned by site technicians were retained for use in this system to avoid
confusion. Initial cage number was listed following date of stocking.
“124-OM-E2E3-C05093-2”
A group of fish from 124-OM-E2-C05093 were moved into cage 2 (C05093-2). A hyphen was
used to signal the movement of fish, its new cage number listed thereafter.
A group of fish from a cage other than cage 3 was also stocked into cage 2 at the same time, its
hatchery code identified as E3 (E2E3). In instances such as these, where an empty cage was
stocked with fish from multiple other cages, or an existing group of fish was supplemented with
fish from another cage, the ID code of the fish group with the largest standing biomass was
retained and assigned to all fish sharing the same cage. If the smaller group of fish had a
22
different hatchery code, as is the case in this example, the hatchery code of the smaller group was
added to the retained ID sequence (E2E3).
“124-OM-E2E3E1-C05093-2-5”
A group of fish from 124-OM-E2E3-C05093-2 was moved into cage 5 (C05093-2-5).
At the same time, a second group of fish (ie. with lesser biomass than the first) was also moved
into cage 5, thus assigned the ID code of the first group.
The second group of fish was from a different hatchery than the first, its hatchery code thus added
to the existing ID sequence (E2E3E1).
3.6 - Interval (INT) and Cumulative (CUMUL) Values
Performance variables were evaluated for individual lots in MANEVFORM on an interval and/or
cumulative basis. Interval values in MANEVFORM were calculated from one management event (eg.
weight sampling event, fish transfer, etc.) to the next, whereas cumulative values were calculated relative
to Day 0 (day of stocking). All values in SUMFORM were inherently cumulative.
It was not possible to calculate cumulative values for lotgroups in MANEVFORM. That is,
cumulative values in MANEVFORM were calculated up to the date of a specific management event. As
the multiple lots of a particular lotgroup would have had differing management event schedules, there
were generally no common dates to which cumulative lotgroup totals could have been calculated.
Cumulative values in MANEVFORM were either simple running totals of a variable (eg. feed/fish,
degree days), or more elaborate calculations referencing the cumulative totals of other variables (eg.