Experimental setup for thermal preference determination

Selecting an appropriate experimental system for the thermal preference experiments comprised a crucial first step in this thesis. As a wide range of experimental setups has been used to determine thermal preferenda in laboratory based experiments so far (McCauley, 1977; Myrick et al., 2004) several methodological approaches were under consideration.

From the various types of setups (McCauley, 1977), linear gradient tanks, shuttleboxes and annular chamber systems were of particular interest as this thesis started.

Linear gradient tanks of rectangular shape were taken into consideration as this type of setup has been used in numerous thermal preference studies before (e.g., Mathur et al., 1982;

Kivivuori and Lagerspetz, 1990; Chen and Chen, 1991; Lafrance et al., 2005; Bates et al., 2010). Thus, a vast amount of information concerning construction, handling as well as strengths, weaknesses and potential pitfalls of this type of setup was available. In addition, rectangular gradient tanks are easy to construct and would allow for a contemporary start of data collection. However, the use of linear gradient tanks becomes problematic when thermal gradients of a wide range are being established as intended in this thesis. The length of the trough has to be adapted accordingly and as the length of the test apparatus increases, automated recording and object detection in automated analysis becomes problematic, especially when small individuals shall be investigated in the setup as well. Still, the biggest drawback of rectangular setups is the various points of thigmotactic cues that are associated with the different temperature levels inside the system. The presence of corners or the proximity to corners meaning potential cover might bias thermal selection as the test organisms orientate towards both ends of the setup (Badenhuizen, 1967; Bevelhimer 1996;

Dillon et al. 2009).

In contrast to linear gradient tanks, shuttleboxes use a temporal rather than a spatial temperature gradient allowing for a wide thermal range (Neill et al. 1972; Reynolds &

Casterlin 1979a; Staaks et al. 1999; Mortensen et al. 2007). Shuttleboxes further offer the advantage of thermal preference experiments being recorded and analysed automatically (Neill et al., 1972; McCauley, 1977). However, only one specimen can be tested at a time (McCauley, 1977) unless the animals show synchronous swarming behavior (Ohlberger et al., 2008) which does not apply to brown shrimp (Tiews, 1970). Non synchronous movements of numerous brown shrimp within the shuttlebox would thus prevent a temporal thermal gradient from being established in the system. It is further unknown whether brown shrimp are able to learn how to control the shuttlebox and thus their body temperature. As this thesis started, there were just few examples of thermal preference experiments with crustaceans in shuttleboxes (Casterlin and Reynolds, 1977; Reynolds and Casterlin, 1979a;

Tattersall et al., 2012). These studies, however, investigated solitary members of the Reptantia which usually walk rather than swim. Brown shrimp appear in groups and might perform differently as it belongs to the Natantia which predominantly move by swimming. The study by Tattersall et al. (2012) moreover showed that the thigmotactic cues in shuttleboxes affect the distribution within the chambers and might thus affect thermal preferenda of brown shrimp.

As both types of systems seemed to be unsuitable to determine thermal preference in the common brown shrimp, an annular chamber system (Myrick et al., 2004) was used for the experiments in the present thesis. Annular chambers counteract most of the problems in the aforementioned systems and are considered advantageous compared to more classical setups (Myrick et al., 2004). Annular chambers have been successfully employed in several thermal preference studies on fishes already (Myrick et al., 2004; Chen et al., 2008;

McMahon et al., 2008; Gräns et al., 2010, 2012; Klimley et al., 2011; Behrens et al., 2012;

Schram et al., 2013). However, the suitability of the advantageous design for experiments on the common brown shrimp had to be evaluated in the present thesis. Still, in addition to extensive evaluation and construction works that are associated with the establishment of a relatively new and technically complex experimental setup, an automation procedure for data recording and data analysis was not available at the beginning of this thesis. Myrick et al.

(2004) as well as Chen et al. (2008) used an observer to record position and temperature data, making the experiments with annular chambers laborious and time consuming. Indeed, permanent observation does not allow for highly resolved preference data over a prolonged experimental period. McMahon et al. (2008) therefore used a video camera for observation and recording of the animal’s positions within the annular chamber. Temperatures were

assigned subsequently based on temperature measurements that have been conducted prior and after each experiment. Gräns et al. (2010), Klimley et al. (2011) and Schram et al. (2013) extended this approach and provided the annular chamber with thermistors for automated temperature measurements throughout the swimming channel. Still, assignment of position and temperature data had to be conducted manually. Behrens et al. (2012), in a very recent study, used a custom made combination between the National Instrument Vision Builder and the LabView software. In their approach, a single experimental animal per run was tracked over the whole experimental period resulting in highly resolved thermal preference data.

The MATLAB routine developed for the present thesis (Chapter I, supplementary information) avoids the time consuming manual assignment of temperature data as this procedure is conducted automatically. By means of this program, multiple animals got detected in the setup and the respective temperatures were assigned accordingly. Extensive evaluations have been performed to assure proper functioning and correct assignment by the routine. This revealed a high precision and low error rate. Supposing some slight modifications and adjustments, the here presented program can be easily transferred to other annular chamber systems. Thus, the automated routine presented in this thesis will facilitate future thermal preference experiments in annular chambers considerably allowing for highly resolved preference data, even in prolonged trials. Besides this, the general principle underlying the routine can also be applied to other types of experimental systems where position data have to be assigned to any spatially resolved factor.

Preliminary tests of the annular chamber that were conducted without a thermal gradient in the swimming channel, revealed the suitability of the system for experiments on the common brown shrimp (Chapter I). However, a slight preference of the brown shrimp towards the outer and the inner walls of the swimming channel was observed. This confirmed the current setup as a prerequisite to obtain unbiased preferenda as the circular shape counteracts previously observed site-specific bias and preference towards the end of rectangular setups (Badenhuizen, 1967; Bevelhimer 1996; Dillon et al. 2009).

Apart from the numerous advantages of the annular chamber system (Myrick et al., 2004), this thesis also identified some drawbacks in this type of setup. As for any spatial, either horizontal or vertical thermal gradient, the test organisms can easily get access a relatively wide range of temperatures within a limited distance. Brown shrimp could thus shuttle within a temperature range that is considerably wider compared to natural conditions and the huge thermal range at narrow space might prompt the shrimp to explore extreme temperatures more frequently as discussed by Behrens et al. (2012) already. This experimental constrain, however, is inherent to almost all choice and preference experiments using setups holding a

gradient. However, the present thesis obtained thermal preferenda by means of the gravitational approach (Chapter II). In contrast to preferenda determined by the acute approach, gravitational preferenda are retrieved once shuttling stopped, meaning that brown shrimp select a relatively narrow temperature range. Spatial proximity of a wide range of temperatures might therefore be of less concern for the results of this thesis.

Another drawback of the system has already been identified in the original setup by Myrick et al. (2004). As in the original setup, the space occupied by the different temperatures was not equal in the system used in this thesis. Due to the annular design, the thermal gradient consisted of two semicircles of equal temperatures (Chapter I). The warmest and coldest temperatures were therefore less available compared to intermediate temperatures as these temperatures were available on both sides of the circular gradient. McMahon et al. (2008) increased the warmest and coldest section of the swimming channel accordingly counterbalancing this discrepancy. Indeed, McMahon et al. (2008) as well as Myrick et al.

(2004) investigated considerably larger animals in their annular chamber systems. In contrast, the annular chamber used of the present study was, first, considerably larger compared to the two aforementioned systems and, second, the experimental animals were much smaller. Thus, all temperatures should be provided in a sufficient spatial amount. Still, as preference temperature was calculated as the median preferred temperature among all individuals within one experimental trial (Mathur & Silver 1980; Karlsson et al. 1984) the upper thermal preference range might potentially be slightly underestimated.

Running an annular chamber turned out to be quite delicate and labour intensive. The complex setup with numerous water inlets to heating and cooling reservoirs as well as water lines into the annular chamber needed to be adjusted, controlled and readjusted on a regular basis. Especially when operated in seawater, fouling is of concern and the setup has to be cleaned thoroughly to avoid the formation of biofilms. Biofilms will especially arise in the warm sections of the annular chamber and might attract the test organisms and thus bias thermal preference. Fouling is also a great problem regarding the pipes and pumps for water supply as well as the perforated walls of the annular chamber as flow rate and thus the discharged water volume might be affected. Additionally, debris from the pipes might get detached and enter the swimming channel, potentially interfering with automated object detection. Thus, one day per week was exclusively provided for cleaning the setup and readjusting the installation to assure proper functioning throughout this thesis. Some of these issues, however, might be related to the present system being incorporated in the in-house recirculating water system. In a second annular chamber system constructed during this thesis (Mues, 2012), the chamber was provided with a separate water circulation making the

overall system more stable and independent from potential effects that may derive from the main recirculating system. In this closed setup, fouling was less of a problem and maintenance was by far less labour intensive, potentially as the reduced amount of nutrients delayed microbial growth.