CONSTRUCCION 1. Materias Primas

In this section, I briefly give an overview of some important factors that need to be taken into consideration in the sampling design related to studying the impacts of fish farming on wild fish communities.

2.1.1.1 Overview of ecological experimentation

To test what impacts coastal aquaculture activities have on wild fish aggregations there needs to be clear hypothesis(es) and a sampling design that will have replication (measuring variability) and ensure results are not confounded (Underwood 1997; Kingsford 1999). Collecting data with no clear research goals will result in data that is useless (Underwood 1997). As the environment and habitat varies in times and space there needs to be a carefully planned sampling design to consider spatial and temporal variability and interactions between space and time, and logistics (Underwood 1997, 2009). Spatial variation includes differences of ecological processes in different places and temporal variation includes differences in biological processes related to seasonality, different ages or stages of development of an organism (Underwood 1997).

Coastal aquaculture activities have an impact on wild fish populations (see Chapter 1) which should be taken into consideration in the sampling design (see Kingsford 1999). To evaluate an impact there needs to be comparative studies. The simplest sampling design in detecting an impact on the environment is to collect data before and after an impact (Green 1979). BACI (Before, After, Control, Impact) design is common in assessing anthropogenic impacts on the environment (Underwood 1992; Kingsford 1999). As there is spatial and temporal variability in abundances in marine organisms between different locations it is a requirement in impact studies to compare the impact site(s) with several control (or reference) sites (Kingsford 1999). Different BACI designs have been developed to detect anthropogenic impacts in the environment. For example, beyond-BACI design allows an impact site to be compared to several control sites including before and after the impact (Underwood 1992; Kingsford 1999). A further development to the BACI design is the M-BACI (multiple before/after control/impact) design which takes into account several impacted sites and compares them to several control sites including also before and after impacts (Kingsford 1999). A number of studies have used before/after and/or control/impact designs to detect coastal aquaculture

impacts on wild fish populations (e.g. Tuya et al. 2006; Dempster et al. 2009, 2011; Tanner and Williams 2015).

The sampling design of a study should take into account the methods used to catch or count fish for the estimation of abundances (Kingsford 1999). Methods for sampling fish near artificial structures include non-destructive methods such as underwater visual censuses (e.g. direct observation by divers) and extractive methods such as hook and line, gill netting, seine netting (see Kingsford 1999; Lowry et al. 2012). Capture methods are often destructive and thus there is a widespread use of visual census techniques to observe fish around artificial structures (e.g. Lowry et al. 2012). Scuba (Self Contained Underwater Breathing Apparatus) diving is a relatively rapid, non-destructive method to observe fish and allows a number of variables to be measured such as number of fish and habitat characteristics (Lowry et al. 2012). However, diver-based techniques are restricted by depth, temperature, time, health safety issues (e.g. shark attacks in Australia) and can affect the behaviour of fish in response to divers (Tanner and Williams 2015). Alternatively, underwater video techniques which are not restricted by the physical limitations of divers, avoid the change in behaviour of fish that can be induced by the presence of divers, and provide information on habitat and species behaviour (see Tanner and Williams 2015 and references therein). All sampling techniques have advantages and disadvantages and depend on the research question, fish of interest, environmental conditions and habitat (Tanner and Williams 2015). Lowry et al. (2012) recommended the use of multi-method approach such as the use of diver techniques and underwater video techniques.

To count fish around coastal fish farms, previous studies have used various methods including non-destructive techniques such as diver-based techniques (Mediterranean Sea, Dempster et al. 2002; Canary Islands, Tuya et al. 2006), underwater video camera (Norway, Dempster et al. 2009), and baited remote underwater video (Australia, Tanner and Williams 2015).

Another factor to consider when choosing control sites to detect the impacts of artificial structures on wild fish populations is the spatial extent of the impact on the fishes (Kingsford 1999). Dempster et al. (2010) evaluated the spatial distribution of wild fish around salmon farms in Norway. The researchers reported highest fish abundance near the sea cages and the aggregation patterns of fish near the sea cages depended on the species.

Most marine organisms vary in time (e.g. days, months, seasons, years) and space (depth, location, distance from shore) and thus sampling design should take into account temporal and spatial variability (see Kingsford 1999). Factors such as spawning, recruitment and migration routes will cause variation in number of fish within a year and between years (Kingsford 1999). Fish can undergo vertical migrations within the day and horizontal migrations over long distances (e.g. for food and reproduction) which can result in temporal and spatial variation in abundances (Kingsford 1999).

To take account of natural variability replication should always be included in a sampling design to ensure any differences between experimental treatments are because of the treatment rather than natural variation (Underwood 1997; Kingsford 1999). In studies that are deficient in replication of control/impact sites the power to generalise the results are weaker (Kingsford 1999). About three decades ago, Hurlbert (1984) reviewed various ecological experiments and noted inadequate or no replication in a number of the studies. Hurlbert (1984) defined pseudoreplication as the "... use of inferential statistics to test for treatment effects with data from experiments where either treatments are not replicated (though samples maybe) or replicates are not statistically independent". Pseudoreplication can be avoided by clearly stating what the hypothesis is and planning an appropriate sampling design that would include controls, randomization and replication (Hulbert 1984; Underwood 1997).

In Chapter 3, I describe in detail the locations and sampling design for this thesis. I used static underwater video camera to observe fish around sea cages and hook and line to extract the fish of interest. During fieldwork conducted in 2013 and 2014, fish were extracted near and away from sea cages to investigate whether there were any differences in diet, condition, lipid and FA patterns in tissue between locations. The next few subsections describe in more details an overview of methodologies related to diet determination, condition and lipid and FA analysis in fish.