SEARCH

SEARCH BY CITATION

Keywords:

  • behaviour;
  • brown shrimp;
  • climate change;
  • thermal preference;
  • thermoregulation;
  • Wadden Sea

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  1. Annular chambers represent a novel approach for thermal preference experiments in aquatic ectothermic organisms. Most approaches using annular chambers so far lack automation in data recording and analysis, making temperature preference experiments laborious and time consuming.
  2. Here, we describe the design and construction of a modified version of an annular chamber system. We conducted extensive tests to improve the systems' functionality and confirm accuracy of the thermal gradient. Additionally, we present an automated matlab routine for data recording and analysis of temperature preference experiments using the common brown shrimp (Crangon crangon, L.) as a test organism. Using this automated routine, we performed an in silico comparison of different thermal gradient representations with various complexities to test for the effect of temperature resolution on the accuracy of thermal preference estimates.
  3. The here presented annular chamber produced a stable thermal gradient of ∆23 °C, ranging between 3 and 25 °C. Automated recording and data analysis facilitated implementation of long-term experiments and allowed the collection of highly resolved preference data. The in silico comparison revealed a more accurate specification of the preference zone with increasing resolution of the temperature gradient. With regard to spatial resolution of the thermal gradient and assignment of position and temperature data, the in silico comparison demonstrated previous approaches to be inappropriate for benthic and passive species.
  4. We present guidelines for annular chamber construction and automation of data analysis in these systems, making annular chambers more handy and applicable for a wide range of preference studies. Besides its use for experiments in annular chambers, the principle of the here presented automated matlab routine can be applied to a wide range of behavioural and preference studies.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Climate change causing ocean temperatures to rise is currently postulated as one of the main drivers in aquatic ecosystems. Several studies highlighted the consequences of increased seawater temperature on the aquatic life, with changes in geographic distribution and abundance as well as aquatic organisms approaching their physiological limits being the most prominent ones (e.g. Parmesan & Yohe, 2003, Perry et al. 2005; Dulvy et al. 2008). Aquatic ectotherms may be especially affected by climatic-driven temperature increases, as ambient temperature directly controls body temperature in these organisms. By this, rising seawater temperatures instantaneously act on physiological rates, affecting most life-history traits as well as habitat utilization and distribution (Neill & Magnuson 1974; Perry et al. 2005; Dulvy et al. 2008; Bertolo et al. 2011). By means of behavioural thermoregulation, however, ectothermic organisms are able to actively control and modulate body temperature, in turn optimizing for physiological processes in a heterogeneous thermal environment (Fry 1947; Reynolds & Casterlin 1979a; Bicego, Barros & Branco 2007).

Laboratory-based temperature gradient experiments are an effective way to study thermoregulatory behaviour and preferred body temperature of aquatic ectotherms (McCauley 1977). Different methodological approaches have been used so far, with the majority of studies adopting rectangular troughs (Mathur, Schutsky & Purdy 1982; Lafrance et al. 2005; Tepler, Mach & Denny 2011) and shuttle-box systems (Neill, Magnuson & Chipman 1972; Reynolds & Casterlin 1979b; Staaks, Kirschbaum & Williot 1999; Mortensen, Ugedal & Lund 2007) on numerous vertebrate as well as invertebrate aquatic ectothermic species (McCauley 1977). However, these classical systems have certain drawbacks inherent to their design. The rectangular shape as well as the presence of corners can induce a site-specific bias towards corners or to the ends of the apparatus, especially in thigmotactic species (Badenhuizen 1967; Bevelhimer 1996; Dillon et al. 2009). In rectangular systems, currents along the trough might differ providing various points of rheotaxis (McCauley 1977). The presence of cover, differences in light intensity and pressure that comes along with vertical thermal gradients might affect temperature selection as well (McCauley 1977; Myrick, Folgner & Cech 2004). Additionally, shuttle boxes are not suitable for slow-moving species or organisms that may not be able to learn how to behaviourally control the temperature within the experimental system (Kivivuori 1994; Lagerspetz & Vainio 2006; Ohlberger et al. 2008).

In contrast to these classical systems, annular chambers represent a new methodological approach for aquatic ectotherm thermal preference studies (Myrick, Folgner & Cech 2004). Annular chambers are considered to be advantageous to more classical systems as they circumvent certain confounding variables present in the aforementioned set-ups (Myrick, Folgner & Cech 2004). In annular chambers, light intensity is even, water depth and flow rate are constant and most important is that, due to the annular shape of the swimming channel (SC), thigmotactic cues like corners are absent. In annular chambers, wide temperature gradients can be established, covering the temperature range of even eurythermal species (Myrick, Folgner & Cech 2004). Additionally, chamber design is rather flexible and can be individually modified and scaled to meet ones specific experimental requirements.

Custom-made annular-shaped preference chambers, either based on acrylic or PVC, have been used in several studies on fishes so far (Myrick, Folgner & Cech 2004; Chen et al. 2008; McMahon, Bear & Zale 2008; Gräns et al. 2010; Klimley et al. 2011; Behrens et al. 2012). Based on the original work from Myrick, Folgner & Cech (2004), systems ranging from 0·22 (Chen et al. 2008) to 3 m (Gräns et al. 2010; Klimley et al. 2011) in total diameter were utilized. Besides the annular shape of the apparatus, however, chamber features and functioning in these studies differed considerably. Temperature monitoring, spatial resolution of the gradient as well as allocation of the test organisms and respective temperature assignment were accomplished differently and at varying complexity. Indeed, the spatial resolution of temperature assignment might influence precision in thermal preference zone determination. This particularly applies to mobile benthic species that are closely associated with the substratum. In contrast to fish that continuously move in a thermal gradient, benthic organisms that gravitate to a respective temperature will remain at a certain position (Hesthagen 1979; Behrens et al. 2012). Even slight discrepancies in temperature allocation will therefore prompt blurred assignment of the frequented temperatures.

The annular-shaped design has proved successful for thermal preference studies in fishes; however, its suitability for invertebrate species like crustaceans has not been evaluated, yet. The objective of the present study was to reproduce and improve an annular chamber based on the original work of Myrick, Folgner & Cech (2004), compiling the information on chamber construction and handling from subsequent studies (Chen et al. 2008; McMahon, Bear & Zale 2008; Gräns et al. 2010; Klimley et al. 2011; Behrens et al. 2012). As most of the previously used annular chambers lack automation in recording of animal position and temperatures (Myrick, Folgner & Cech 2004; Chen et al. 2008; McMahon, Bear & Zale 2008; Gräns et al. 2010; Klimley et al. 2011), we established an automated routine for data recording and analysis of thermal preference experiments in matlab. The aim of this automated routine was to reduce the presence of an observer for experimental monitoring and data acquisition as well as analysis of thermal preference experiments. Thus, thermal preference experiments should become less time consuming and laborious and result in prolonged and continuous periods of observation generating highly resolved data in time. We finally compared five different approaches on thermal gradient representation and temperature assignment in silico using the common brown shrimp (Crangon crangon, L.) as an invertebrate test organism.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Annular chamber system

The annular chamber system was a modified version of the set-up originally described by Myrick, Folgner & Cech (2004) with a total diameter of 145 cm and further dimensions as specified (Fig. 1, Table 1). In contrast to the original system, the chamber for the present study was made of concentric polyvinyl chloride (PVC) walls mounted on a glass base using SikaFlex (Sika Group GmbH, Stuttgart, Germany). Due to the sedentary and bottom-associated living of brown shrimp, the holes for water intake were placed below the prospective water surface rather than above as described in Myrick, Folgner & Cech (2004). The holes were drilled at 1·1 cm distance each, along four shifted rows over the whole respective height of the water column. By this, we intended to assure a smoothed water inflow into the SC and avoid thermal stratification. To allow for observation during day and night, the area below the SC was illuminated by 24 equally spaced infrared LEDs (SFH 485 P, 880 nm, OSRAM).

Table 1. Dimensions of the annular chamber system
 Diameter (cm)Height (cm)Channel width (cm)Water level (cm)
Circle a14515
Circle b12515
Circle c955·5
Circle d754·5
Reservoir channel (ab)107
Swimming channel (bc)155·5
Effluent channel (cd)105·5
image

Figure 1. Schematic illustration of the annular preference chamber set-up. (a) Top view of the annular preference chamber and water delivery system. (1) water inlet, (2) spherical valve, (3) on–off temperature control switch, (4) heater, (5) pump, (6) reservoir tank, (7) level sensor, (8) divider, (9) centre drain, (10) circle d, (11) circle c, (12) circle b, (13) circle a, (14) temperature sensor, (15) PID controller, (16) Pt100 temperature sensor, (17) three-way control valve, (18) heat exchanger water inlet, (19) heat exchanger coolant inlet, (20) heat exchanger coolant outlet, (21) heat exchanger. Light blue lines indicate water and dark blue lines coolant pipes. (b) Schematic 3D illustration of the annular preference chamber system. (1) temperature sensor, (2) swimming channel (SC) outflow holes, (3) SC inflow holes, (4) v-notch, (5) centre drain, (6) circle a, (7) circle b, (8) circle c, (9) circle d, (10) divider, (11) compartment wall, (12) glass base, (13) wooden base.

Download figure to PowerPoint

Temperated water was provided from five reservoirs containing water of 3, 9, 14, 19 and 26 °C. By these temperature steps, a thermal gradient of c. 5 °C increments between each of the eight compartments could be achieved. Hot water (14, 19 and 26 °C) was obtained by 3 kW immersion heaters (RY330; Redring Electric LTD, Peterborough, UK) and electrical titanium heating rods (600 W; Schego, Offenbach am Main, Germany). Heaters were regulated by temperature sensors (Pt100 RTD temperature probe; JUMO GmbH & Co. KG, Fulda, Germany) connected to an electronic thermostat (Jumo eTRON M; JUMO GmbH und Co KG) keeping temperatures at the respective set value ± 0·2 °C. Water was cooled (3 and 9 °C) via the central in-house cooling unit (EUWAB24KAZW1; DAIKIN Airconditioning Germany GmbH, Unterhaching, Germany) charging two titanium heat exchangers (VT04 CD16; GEA Ecoflex, Sarstedt, Germany). The outflow of the heat exchangers was controlled by Pt100 thermocouples (Pt100 Class B sensor; RS Components GmbH, Mörfelden-Walldorf, Germany) connected to a PID process controller (4100+; West Control Solutions, Kassel, Germany). The PID controller regulated a three-way control valve (three-way control valve type 323; Belimo Automation AG, Hinwil, Switzerland) via a modulating rotary actuator (LR24A-SR; Belimo Automation AG, Hinwil, Switzerland) to keep temperatures at the respective temperature ± 0·2 °C. Water was distributed at 3·5 L min−1 (2·0–5·0 L min−1 in the evaluation phase) to each of the eight compartments of the reservoir channel, resulting in c. 100% SC volume exchange per min of the SC. For chamber evaluation, test runs were conducted at six different flow rates, that is, 50%, 70%, 90%, 110%, 130% and 150% of SC volume exchange per min with the thermal gradient established. Dye tests were conducted at all flow rates to control for water flow throughout the SC. Cooled air was injected below the SC to avoid water condensation below the coldest compartment.

The temperature gradient in the SC was monitored by 32 equally spaced temperature sensors (DS1820-LC; B+B Thermo-Technik GmbH, Donaueschingen, Germany), attached to the outer wall of the SC at mid-water depth and connected to a digital USB thermometer (TLOG64-USB; B+B Thermo-Technik GmbH). Temperature was recorded every 15 s and visualized in real time using the PC-Datalogger Software (PC-Datalogger; B+B Thermo-Technik GmbH). Perpendicular to the centre of the annular chamber, a mirror was mounted at 45°, deflecting the SC to a camera (EcoLine TV7002; ABUS Security-Center GmbH & Co. KG, Affing, Germany) equipped with a daylight filter (SKR FIL 093; Joseph Schneider Optische Werke GmbH, Bad Kreuznach, Germany) and the CAT filter removed. The camera was connected to a video monitor to allow for continuous surveillance of the set-up. To achieve an even and diffuse illumination of the SC, eight cold cathode tube lights (350V, 2·4W, 6 mA; Conrad Electronics, Hirschau, Germany) were mounted in equal distances on a circular PVC frame suspended 1·5 m above the experimental chamber. The whole set-up was surrounded by a 2-m curtain to exclude outside light and avoid any disturbance during the experiments. We conducted initial test runs using brown shrimp without a thermal gradient to check for a potential tank bias.

Evaluation of the annular chamber

As the proportions of the present chamber deviated from the set-up originally described by Myrick, Folgner & Cech (2004), the system was thoroughly evaluated for evenness in flow rate (flow metre, mn 7·5, 0·04–10 m s−1; Höntzsch GmbH, Waiblingen, Germany), illumination (LI-250A light meter; LI-COR, Lincoln, NE, USA) and development of the thermal gradient. Flow rate was determined at mid-water depth at 3 points per compartment, whereas light intensity was measured in the SC centre of each compartment taking the 15 s average per measurement. Each measurement was repeated three times. Temperature measurements on the thermal gradient were conducted by a thermocouple (TS-NTC202 temperature sensor; B+B Thermo-Technik GmbH; calibrated with a Technoterm 9500; Testoterm KG, Lenzkirch, Germany) fixed to a vertically adjustable custom-made PVC rack. For temperature measurement, the SC was divided into 64 virtual segments in radial direction. Thirty-two of the transects were in line with the 32 mounted DS1820-LC temperature sensors, whereas the other 32 segments were located between two adjacent sensors. Temperature measurements were conducted according to a fixed, circular grid throughout the SC consisting of 64 × 3 × 3 nodes (eight compartments including eight transitions between the compartments × seven transects per compartment × three water depths × three positions in radial direction) and replicated three times. Temperature measurements with the TS-NTC202 thermocouple were synchronized to the measurements by the mounted DS1820-LC sensors, recording temperature every 2 s. This resulted in two temperature matrices (TS-NTC202 thermocouple grid and mounted DS1820-LC sensors) that were used to calculate the deviance between the mounted sensors and the grid measurements. The resultant matrix was then used to calculate a spatially resolved 64 × 3 × 3 temperature grid based on the temperature measurements by the mounted DS1820-LC sensors (Fig. 2b).

image

Figure 2. Schematic illustration of the temperature profile in the temperature preference chamber. (a) Temperature profile within the swimming channel at mid-channel position. (b) Top view of the spatial temperature distribution as determined by means of the temperature grid. Black dots in (b) indicate grid nodes. For illustration purposes, the thermal gradient is rotated clockwise by 90° compared to Fig. 1a.

Download figure to PowerPoint

Automation of data recording and analysis

Temperature and image data were recorded using a custom-made matlab program including the Image Acquisition Toolbox. Temperature data were retrieved from the PC-Datalogger software every 60 s synchronized to image acquisition. Compilation and assignment of image and temperature data was conducted offline by a second custom-made matlab program including the Image Processing Toolbox (see Supporting information, Data S1, for sample code). As a prerequisite for automated analysis, the outer and inner margins of SC were vectorized by means of two ellipses enclosing the annular-shaped SC, that is, the arena for observation (see Supporting information, lines 159–223). We initially assumed a circular shape of the SC in the acquired image; however, the deflection due to the huge size of the mirror added some amount of distortion to the image. Temperature sensor positions and the coordinates of the temperature gird were integrated into this vectorization as well (see Supporting information, lines 243–246 and 114–129).

For offline analysis of thermal preference experiments, an array of acquired image and temperature data from each experiment was loaded to the program successively (see Supporting information, lines 33–43). The background, recorded at light and dark prior to the experiment (see Supporting information, lines 134–157), was subtracted from the respective target image, obtaining an image that just contained the difference of both pictures, that is, the experimental animals. Upon conversion to a binary image, an image erosion and dilation was applied, and the resulting blobs were filtered based on their size to remove false-positive detections (see Supporting information, lines 253–318). The geometrical centre of gravity (approximately mid-body position) for each experimental animal was determined, and the position within the SC specified according to its xy coordinates (see Supporting information, lines 297–341). Subsequently, each single animal was assigned to the closest node of the circular temperature grid and the respective temperature ascribed accordingly (see Supporting information, lines 345–414). This procedure was conducted for the whole image and temperature data array, storing position and temperature information of the experimental animals into a consecutive data array that was exported to MS Excel (see Supporting information, lines 416–432).

In silico comparison of thermal gradient representations

We conducted an in silico comparison of four different approaches of thermal gradient representation and temperature assignment to test for the effect of spatial temperature resolution on thermal preference estimates. By this, we intended to potentially reduce the amount of effort for data monitoring and acquisition in prospective studies using annular chamber systems for thermal preference or preference testing on any other environmental factor. The four approaches were derived from previously published studies on annular chambers and temperature monitoring and assignment procedures therein (Myrick, Folgner & Cech 2004; McMahon, Bear & Zale 2008; Gräns et al. 2010; Klimley et al. 2011).

In approach (1), we used the temperature recordings provided by the evenly spaced DS1820-LC temperature sensors, mounted to the outer wall of the SC (sensor mode). The sensor-based mode has been repeatedly applied in annular thermal preference studies (Myrick, Folgner & Cech 2004; Gräns et al. 2010; Klimley et al. 2011; Behrens et al. 2012). For approach (2), we assumed a continuous and linear thermal gradient between the hottest and coldest temperature in the SC, representing the quasi-perfect state of the gradient (continuous mode). For this approach, the hottest and coldest temperatures were determined by means of the temperature sensors, and all further temperatures were interpolated at uniform intervals of 0·67 °C in 64 steps around the whole SC. In approach (3), the temperatures of the five reservoirs were used to create discrete temperature fields for each of the eight compartments throughout the SC (discrete mode). This approach provided an example for a low spatial resolution of the temperature gradient. All three approaches were compared to approach (4), that is, a temperature grid with 64 × 3 × 3 nodes covering the whole SC (grid mode). However, as the shrimp exclusively stayed at the bottom of the SC, we just used the bottom layer for temperature assignment, that is, 64 × 1 × 3 nodes. We assume the grid scenario to be the best representation of the real state of thermal gradient. The three other scenarios were compared against the grid mode in silico, to ultimately identify the most accurate approach.

Experimental animals and protocol

The common brown shrimp was used as test organism, as this species is a key component for the North Sea coastal ecosystem, a habitat with documented response attributed to climatic-driven changes and contemporary shifts in water temperature (Lotze et al. 2005; Perry et al. 2005; Dulvy et al. 2008; Reise & van Beusekom 2008). Besides its ecological importance, the common brown shrimp is a highly valuable fishery resource (ICES 2011). The brown shrimp represents an ideal test organism for the system, being both highly mobile and tolerating a wide thermal range (Campos & van der Veer 2008). As brown shrimp occur at high densities in the field, an approach where multiple animals could be tested within one experimental trial was needed (Richards, Reynolds & McCauley 1977).

Brown shrimp for the experiments were caught by the research vessel FFS Solea in January 2011, off the Isle of Helgoland (54°20′N, 007°22′E) at 37 m depth. On board of FFS Solea, animals were kept in an aerated tank with surface water flow-through until arrival in Cuxhaven, c. 5-h postcatch. Shrimp were transferred to continuously aerated tanks and transported to the laboratory facilities of the Institute of Hydrobiology and Fisheries Science, University of Hamburg, Germany. Here, animals were maintained in 1 m³ circular tanks at 8 ± 0·5 °C with aerated artificial seawater of 30 PSU. The tanks were connected to the in-house temperature-controlled recirculating water system with a foam fractionator and a moving bed biofilter. Upon 2 days of acclimation, shrimp were sorted to the nearest 5 mm total length and transferred to separate temperature-controlled circular tanks and maintained at 8 ± 0·5 °C. The brown shrimp were fed dry feed (Marico Advance; Coppens International, Helmond, Netherlands), live Artemia nauplii (SEPArt; Inve Aquaculture, Dendermonde, Belgium) and chopped herring and sprat pieces to apparent satiation every day. Twenty-four hours prior to each experiment, 10 animals from one respective size class were dip-netted from the holding units. Sex was then determined based on the appendices of the first and second endopodite (Tiews 1954), and the brown shrimp were transferred to a separate holding unit with temperature conditions as stated above to minimize handling stress before the experiment. Experiments were started at the following day between 7 and 8 am in the morning. Brown shrimp were released into the SC with the temperature gradient being established at that segment corresponding to the temperature the shrimp were maintained. Data acquisition was started, and animals were left undisturbed throughout the whole experiment. After 20 h of exposure to the thermal gradient, temperature preference was analysed from the last 3-h period. To avoid pseudoreplication, the preferred temperature of brown shrimp within one run was calculated as the mean of the median selected temperatures of each single shrimp in one experimental trial (Mathur & Silver 1980; Karlsson, Ekbohm & Steinholtz 1984). Three successive trials were conducted using 4·5-, 5·5- and 6·5-cm female brown shrimp to test the automated analysis procedure. For each trial, the four different in silico temperature allocation procedures were performed.

Data analysis

Thermal preferenda were calculated as the mean of the median selected temperatures as well as the first and third quartiles, representing the upper and lower limits of the thermal preference zone (Magnuson, Crowder & Medvick 1979). Data analysis was conducted in r (R Development Core Team 2011) using the car (Fox & Weisberg 2011) and pgirmess (Giraudoux 2011) packages. Assumptions of normality and homogeneity of variances were determined by means of Shapiro–Wilk's test and Levene's test, respectively. If assumptions were confirmed, an anova otherwise a Kruskal–Wallis anova was conducted. Tukey tests or multiple comparison tests were used for post hoc testing, respectively. During chamber evaluation, we tested whether the SC was illuminated evenly. Based on pretrials at different flow rates (50–150% SC volume exchange per min), we tested whether a vertical thermal stratification occurred in the SC and validated if the temperature gradient was homogenous in radial direction. We also tested whether brown shrimp showed any site preference in the absence of a thermal gradient.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Annular chamber and set-up evaluation

The presented version of an annular chamber system consistently produced a stable 3–25 °C temperature gradient (Fig. 2), covering the full thermal niche of the common brown shrimp. Temperatures at individual positions in the SC were considerably stable, varying ±0·2 °C. However, the wide thermal gradient of ∆23 °C throughout the SC evoked subsequent problems. In the initial phase of set-up evaluation, experimenting with different flow rates of 50–100% of SC volume exchange per min, we observed a considerable amount of thermal stratification in the SC (Kruskal–Wallis anova, P < 0·01). Warm water masses spread into surface layers of adjacent compartments, whereas the cold water expanded along the bottom. By increasing flow rates up to 150% SC volume exchange per min, thermal stratification could be eliminated (Kruskal–Wallis anova, P > 0·05). High flow rates, however, made the gradient unsteady and volatile, especially at the transition zones between adjacent reservoir compartments as reported previously (McMahon, Bear & Zale 2008). Additionally, brown shrimp used for successive test runs to determine a potential tank bias without a thermal gradient, appeared to be disturbed by flow rates >110% SC volume exchange per min, as they did not come to rest, continued to swim up and down and tried to escape from the set-up. Following Chen et al. (2008), we mounted small radial dividers (15 × 4 cm) between the eight compartments. In the present study, the dividers were immersed into the top 1 cm of the water column of the SC (Fig. 1). By this, we were able to block the proliferation of warm surface water, counterbalancing the thermally induced shearing forces due to the high temperature differences at narrow space. Successive dye tests confirmed an even and linear, radial flow through the SC, with the shrimp not showing any avoidance of dividers in subsequent test runs at 100% SC volume exchange per min. The small dividers promoted an even and smooth temperature transition throughout the SC (Fig. 2) and eliminated vertical thermal stratification at 100% SC volume exchange per min (Kruskal–Wallis anova, P > 0·05). At 100% SC volume exchange per min, current velocities were below detectable limits at all locations. Illumination throughout the SC ranged from 0·149 to 0·175 W m−2, but differed not significantly (one-way anova, P > 0·05).

Brown shrimp behaviour

Brown shrimp did not show any site preference in the SC when the thermal gradient was absent (one-way anova, P > 0·05). Shrimp dispersed equally throughout the set-up showing moderate activity and alternate times of rest. However, there was a slight preference towards the outer and inner walls of the SC indicating positive thigmotaxis in brown shrimp. At 100% SC volume exchange per min, shrimp exclusively remained at the bottom of the SC showing predominantly pacing locomotion. If swimming, shrimp stayed close to the bottom of the SC as well.

Shrimp released into the thermal gradient behaved considerably different and showed marked differences in individual behaviour. While some shrimp quickly found a target area in the SC and shuttled within a narrow temperature range, others continued to cruise throughout the SC for an extended time period while exploring the whole thermal gradient. Animals entering temperatures >18 °C performed U-turns or successively increased locomotor speed until favourable thermal conditions were reattained. However, following 3–5 h of exposure to the gradient, all animals ended up in a restricted temperature range of ±5 °C. The subsequent time was characterized by intermediate shuttling behaviour, with short excursions throughout the SC and subsequent return to the previously frequented temperature areas.

Automation of thermal preference experiments

The automated monitoring of the annular chamber experiments allowed us to record highly resolved temperature preference data as demonstrated by three test runs using 4·5-, 5·5- and 6·5-cm female brown shrimp (Fig. 3). During the experiments, observer time could be reduced to occasional controls of set-up functioning, without permanent presence to record temperatures and animal positions in the apparatus. Using approach (4), that is, the grid mode, these test runs revealed thermal preference zones for 4·5-cm female brown shrimp ranging from 4·7 to 7·9 °C and 6·2 to 8·2 °C for 5·5-cm females. Female brown shrimp of 6·5 cm selected 4·1–7·7 °C. In general, the preferred temperature zone at dark was slightly more variable than at light. However, the preference zone (median ± 1st and 3rd quartile) narrowed towards the end of the experiments in all groups.

image

Figure 3. Thermoregulatory behaviour of female brown shrimp of (a) 4·5, (b) 5·5 and (c) 6·5 cm within the annular chamber system. Boxes include 1st and 3rd quartiles. Whiskers denote 95% of the data range. Shaded areas correspond to observations during scotophase.

Download figure to PowerPoint

Automated analysis and data association of position and temperature records within matlab worked reliably for all three size classes. When testing the 4·5-cm size class, however, automated analysis was more error prone than for 5·5- and 6·5-cm shrimp, mainly due to size-related limitations in object detection. Error rates for automated object detection ranged from 2% (6·5 cm) to 10% (4·5 cm). We faced two major problems in object detection. First, brown shrimp showed a slight preference towards the outer and inner walls of the SC. By approaching the inner wall, shrimp became occluded by entering the dead angle of the system. Although the shrimps were still visible in the acquired image, the program parameters specified for image processing to avoid false-positive detection excluded these shrimp from the data array. This could be adjusted by proper parameterization to some part, but for 4·5-cm individuals, this procedure ultimately reached an end. Secondly, having a more or less narrow thermal preference zone, shrimp accumulated at certain areas of the annular chamber, causing distinct objects to merge. However, a manual inspection of the recorded thermal preference data could be achieved within minutes for each single trial to correct for these deficiencies in the automated analysis.

In silico comparison of thermal gradient representations

For in silico comparison, approach (4) (grid mode) represented the reference state as this scenario represented the thermal gradient most accurately. The in silico comparison of the four gradient representations (Fig. 4) revealed slight differences in the estimated median preferred temperatures of up to c. 1 °C between the scenarios (Fig. 5, Table 2). The 1st and 3rd quartiles confining the thermal preference zone differed with up to c. 1·6 °C (Fig. 5, Table 2). In general, the thermal preference zone became narrower with increasing resolution of the thermal gradient. However, approach (1) (sensor mode) with 32 equally spaced temperature sensors produced a similar thermal preference zone as approach (3) (discrete mode), using a low spatial resolution of the thermal gradient (Fig. 4, Table 2). In contrast, approach (2) (continuous mode), assuming a linear temperature gradient between the hottest and coldest temperature in the SC was comparable to approach (4) (grid mode), correcting each sensor value at a particular position within the SC. Here, the median temperature preferenda were considerably lower with narrower thermal preference zones than in approach (1) (sensor mode) and approach (3) (discrete mode).

Table 2. Thermal preference zone, median preferred temperature, and 25% and 75% quartiles for four different spatial temperature gradient representations as derived by the in silico analysis with brown shrimp (Crangon crangon) in a 3–25 °C gradient
 Median25%75%
Grid mode6·34·97·9
Continuous mode6·14·97·6
Discrete mode7·53·919·3
Sensor mode7·24·179·1
image

Figure 4. Schematic illustration of in silico temperature gradient resolution for four spatially resolved temperature scenarios. (a) discrete mode, (b) continuous mode, (c) sensor mode. Black dots in (c) indicate sensor positions. For illustration purposes, the thermal gradient is rotated clockwise by 90° compared to Fig. 1a.

Download figure to PowerPoint

image

Figure 5. Thermal preference zones for brown shrimp in a 3–25 °C gradient as derived from median selected temperatures for four different spatial temperature gradient representations (for illustration of the temperature scenarios, see Figs 2b and 4). Error bars denote 95% of the data range.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Annular chamber set-up and functioning

The general set-up of the annular chamber for the present study was straightforward. Basic craftsmanship and CAD software tools are sufficient to design and construct an annular chamber in c. 6–8 weeks of working time. However, we identified certain pitfalls that have to be considered and eliminated when constructing and using such systems. These problems refer to chamber dimensioning, thermal stratification and water flux through the SC.

In the first study using an annular chamber system, Myrick, Folgner & Cech (2004) highlighted the chamber design to be rather flexible and stated that chamber dimensions could be easily modified to meet ones specific experimental requirements. The total diameter could be increased or decreased and the SC width and depth be modified to account for experimental animal size (Chen et al. 2008; McMahon, Bear & Zale 2008; Gräns et al. 2010). Indeed, total diameter and chamber size are directly linked to the thermal gradient width that can be established within the SC (Chen et al. 2008). For eurythermic species, as in the present study, SC width has to be increased accordingly to provide the full temperature range of the thermal niche. In large set-ups, however, small specimens are hard to detect due to constraints in camera resolution and object detection. Additionally, size and total chamber volume are directly linked to heating and cooling capacity for the overall setting. This is of particular importance when continuously high flow rates of temperated water are required, especially for long-lasting gravitational thermal preference experiments. For the here presented system, a total of c. 12·3 kW was needed to create and maintain a thermal gradient as described.

Previous studies using annular chambers highlighted constant and radial flow-through providing a nonthermally stratified temperature gradient in the SC (Myrick, Folgner & Cech 2004; Gräns et al. 2010; Klimley et al. 2011; Behrens et al. 2012). We were therefore surprised by the high amount of thermal stratification in the initial test runs in our system. Thermal stratification was not omnipresent in the SC, but marked temperature differences between surface and bottom water developed especially within the coldest compartment, with up to ∆4 °C at no more than 5·5 cm water depth. For the initial test runs, we started with 50% flow rate, that is, 50% SC volume exchange per min, following McMahon, Bear & Zale (2008) and gradually increased flow rate in 10% steps with subsequent temperature measurements. High flow rates eliminated thermal stratification in our system, but produced volatile temperature gradients. Subsequent dye tests confirmed vast backwash and eddy formation as described by McMahon, Bear & Zale (2008). Additionally, we found flow rates generated at >110% SC volume exchange per min to disturb the shrimp as they did not come to rest, continued to swim up and down in the SC and even tried to escape from the set-up. In contrast, low flow rates increased stratification as residence time was increased and an insufficient amount of mixing occurred in the SC. In the present system with small dividers (Fig. 1) immersed into the SC, flow rate could be kept at c. 100% while thermal stratification was nonsignificant. In contrast to previous set-ups (Chen et al. 2008), these dividers were just immersed into the upper 1-cm water layer to block the shearing forces in the surface water. However, this also reduced the spread of the cold bottom water. Still, the test animals did not show any avoidance of these dividers while shuttling in the SC.

Automation of data recording and analysis

Automation of image and temperature data recording allowed for highly resolved preference data for an extended experimental period in the present study. Most set-ups used for thermal preference experiments so far, except for shuttle boxes, relied on an observer recording temperatures and positions of the test animals within the set-up (e.g. Myrick, Folgner & Cech 2004; Chen et al. 2008; Diaz et al. 2011). Video or image recording with subsequent data analysis facilitated the analysis of temperature preference experiments in previous studies, especially for prolonged gravitational preference tests (McMahon, Bear & Zale 2008; Gräns et al. 2010; Klimley et al. 2011). Still, data analysis and assignment of individual temperature and position data for large datasets are laborious and time consuming. Without automated analysis, huge amounts of video or image footage have to be inspected and position data assigned to the respective temperatures manually. By means of the here presented matlab program, data analysis of thermal preference experiments in annular chamber systems could be conducted automatically. Just minor adjustments of the presented matlab program are needed (see Supporting information) to adapt the automation routine to other annular chambers. Additionally, the basic principle of position and temperature assignment – or assignment of any other factor – can be transferred to other types of experimental systems as well.

Basically, the matlab program performed a simple object detection with subsequent position and temperature assignment based on an annular shape of the arena. We did not apply dynamic background subtraction but used two separate background images (day and night) instead. Additionally, we omitted a tracking module in our program. Due to this, mergence and occlusion of objects was of minor importance, as it was not necessary to detect and follow each individual separately but simple object detection was sufficient. Mergence and occlusion of objects could be partly counterbalanced by a proper parameterization of the automated routine (see Supporting information, lines 290–291). The remaining errors in object detection and temperature assignment could be corrected by a manual inspection of the data exported to MS EXCEL.

Apart from these technical reasons, tracking was omitted in the present approach as the whole group of animals within each trial was treated as one single experimental unit, and the mean of the individual medians calculated accordingly. Individual tracking of multiple test organisms generating multiple preference values within the same run would be problematic for statistical reasons (McCauley 1977). The number of degrees of freedom would be overestimated, resulting in type I errors as shown for statistical analysis of thermal preference experiments before (Mathur & Silver 1980; Karlsson, Ekbohm & Steinholtz 1984). However, our computer program can in fact be used for tracking applications in future studies, as the coordinates of the detected objects are extracted by means of the matlab routine offline (see Supporting information, lines 397–341). For a single object, a slight modification of the provided code would be necessary to accomplish tracking for future applications.

Brown shrimp behaviour

Shrimp showed typical behavioural patterns in the SC, with orthokinesis and klinokinesis while exploring the thermal gradient. Following a period of acclimation and exploratory movements, however, the shrimp settled within a restricted range of temperatures. Ortho- and klinokinetic behaviour revealed thermosensitivity of brown shrimp (Lagerspetz & Vainio 2006) and showed brown shrimp to be able to thermoregulate behaviourally.

The temperature the brown shrimp experienced during husbandry seemed to be of minor importance with regard to the thermal preference zone. On average, brown shrimp selected considerably colder temperatures compared to acclimation temperature. This is potentially due to the prolonged period the shrimp were exposed to the thermal gradient allowing for reacclimation and the shrimp to gravitate towards their final or ultimate thermal preferendum (Fry 1947; Reynolds 1978). The thermal preferenda just varied slightly between the three tested size groups, with the lower limit of the temperature preference zone differing as much as ∆1·1 °C between 4·5-, 5·5- and 6·5-cm female brown shrimp. The upper limit was more consistent, with a standard deviation of 0·25 °C. However, the presented preferenda should be treated with caution and should not be used to draw conclusions on thermal preference in the common brown shrimp, as we just tested three size classes without replication. Indeed, the annular shape of the present set-up can be considered as a prerequisite to study thermal behaviour in brown shrimp in future studies, as we observed a positive thigmotactic behaviour towards the outer and inner margin of the SC. In a rectangular set-up, this positive thigmotaxis could affect thermal selection by an end of tank bias as has been previously reported (Badenhuizen 1967; Bevelhimer 1996).

In silico comparison of thermal gradient representations

Temperature preference is most adequately described as a preference zone instead of a single temperature value (Reynolds 1978; Magnuson, Crowder & Medvick 1979). To account for exploratory and shuttling behaviour, the median with the 1st and 3rd quartiles is commonly used to describe the thermal preference zone including potential skewness of selected temperatures (Magnuson, Crowder & Medvick 1979; Martin & Huey 2008). Hence, spatial resolution of the thermal gradient and correct temperature assignment are fundamental tasks in thermal preference testing. This was evaluated and confirmed by means of different temperature gradient representations, using different temperature matrices at varying resolution and complexity. All four gradient approaches were found to represent the median adequately; however, the widths of the temperature preference zones differed considerably. The deviations among the four different approaches might partly be due to the specific behaviour of the shrimp. In contrast to demersal or pelagic fish, benthic organisms as shrimp once gravitated settle and do hardly move between shuttling phases. Even slight deficiencies in temperature assignments will therefore accumulate and cause an underestimation or overestimation of the selected temperature massively.

We chose approach (3) (discrete mode), that is, eight homogenous temperature fields within the SC assigned by the reservoir temperatures, as an example for a coarse spatial resolution of the thermal gradient. Interestingly, this coarse representation of the temperature gradient delivered comparable estimates of the temperature preference zone as approach (1) (sensor mode), which is commonly applied in annular chamber systems (Myrick, Folgner & Cech 2004; Gräns et al. 2010; Klimley et al. 2011; Behrens et al. 2012). Nevertheless, approach (1) (sensor mode) was found to represent the thermal gradient inadequately when compared to reference state for the thermal gradient, that is, approach (4) (grid mode) that represented the thermal gradient most accurately. This might be due to the slight inhomogeneity of the temperature field in radial direction that has been reported as a common problem in annular chamber systems (Myrick, Folgner & Cech 2004; McMahon, Bear & Zale 2008). Apart from that, too few sensors might have been used in the present study and sensor spacing might have been too extensive in relation to the test organism's size. Previous studies counterbalanced inhomogeneity of the temperature field in radial direction by means of two rows of temperature sensors (Gräns et al. 2010; Klimley et al. 2011). In these studies, the sensors were mounted on the outer as well as the inner wall of the SC, and the temperature per transect was calculated as the mean of the closest pair of radial sensor. However, using the same numbers of temperature sensors as in approach (1) (sensor mode), the aforementioned issues could be counterbalanced by means of approach (4) (grid mode) in the present study. To our opinion, the grid mode represented the thermal gradient most accurately. In approach (4) (grid mode), a temperature grid of 64 × 1 × 3 nodes was integrated into the SC allowing for precise temperature measurements as well as a high spatial resolution of the thermal gradient. The estimated thermal preference zones revealed equivalent results as in approach (2) (continuous mode), where a continuous and linear thermal gradient at uniform intervals of 0·67 °C in 32 steps between the hottest and coldest temperatures in the SC was interpolated. For future approaches in annular chamber systems, the approach presented here (4) (grid mode) is therefore recommended to assure a high resolution of the thermal gradient allowing for precise temperature assignment by means of a manageable amount of temperature sensors. However, once the thermal gradient is specified and its continuity and stability confirmed, approach (2) (continuous mode) can be used to avoid the extensive use of sensors.

Overall, automation of data recording and image analysis facilitated thermal preference experiments and reduced the amount of effort considerably, allowing for more complex and comprehensive thermal preference studies in the future. The presented automated analysis can easily be projected to other environmental variables like salinity, ammonia or oxygen concentration, turbidity, etc. Different types of sensors can be integrated into this network to monitor gradients in annular chambers. Additionally, once the gradient is specified and its stability confirmed, a modified version of approach (2) (continuous mode) can be used to assign animal position to respective gradient values. By this, preferenda for a variety of environmental variables can be determined in annular chamber systems.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The authors thank Thomas Neudecker and the crew from the research vessel FFS Solea for providing life brown shrimp for this study. Carmen Czerwinski and Wiebke Bretting are thanked for data validation and laboratory assistance. We thank Marc Hufnagl for his helpful comments on the matlab script. This study was partly funded by the Cluster of Excellence ‘Integrated Climate System Analysis and Prediction’ (CliSAP) of the University of Hamburg. We also acknowledge the constructive comments of two anonymous reviewers on an earlier version of this manuscript.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Badenhuizen, T.R. (1967) Temperatures selected by Tilapia mossambica (Peters) in a test tank with a horizontal temperature gradient. Hydrobiologia, 30, 541554.
  • Behrens, J.W., Gräns, A., Therkildsen, N.O., Neuenfeldt, S. & Axelsson, M. (2012) Correlations between hemoglobin type and temperature preference of juvenile Atlantic cod Gadus morhua. Journal of Experimental Marine Biology and Ecology, 413, 7177.
  • Bertolo, A., Pépino, M., Adams, J. & Magnan, P. (2011) Behavioural thermoregulatory tactics in lacustrine brook charr, Salvelinus fontinalis. PLoS ONE, 6, e18603.
  • Bevelhimer, M.S. (1996) Relative importance of temperature, food, and physical structure to habitat choice by smallmouth bass in laboratory experiments. Transactions of the American Fisheries Society, 125, 274283.
  • Bicego, K.C., Barros, R.C.H. & Branco, L.G.S. (2007) Physiology of temperature regulation: comparative aspects. Comparative Biochemistry and Physiology. Part A, Molecular & Integrative Physiology, 147, 616639.
  • Campos, J. & van der Veer, H.W. (2008) Autecology of Crangon crangon (L.) with an emphasis on latitudinal trends. Oceanography and Marine Biology - An Annual Review, 46, 65104.
  • Chen, S.X., Hong, W.S., Su, Y.Q. & Zhang, Q.Y. (2008) Microhabitat selection in the early juvenile mudskipper Boleophthalmus pectinirostris (L.). Journal of Fish Biology, 72, 585593.
  • Diaz, F., Salas, A., Re, A.D., Gonzalez, M. & Reyes, I. (2011) Thermal preference and tolerance of Megastrea (Lithopoma) undosa (Wood, 1828; Gastropoda: Turbinidae). Journal of Thermal Biology, 36, 3437.
  • Dillon, M.E., Wang, G., Garrity, P. & Huey, R.B. (2009) Thermal preference in Drosophila. Journal of Thermal Biology, 34, 109119.
  • Dulvy, N.K., Rogers, S.I., Jennings, S., Stelzenmller, V., Dye, S.R. & Skjoldal, H.R. (2008) Climate change and deepening of the North Sea fish assemblage: a biotic indicator of warming seas. Journal of Applied Ecology, 45, 10291039.
  • Fox, J. & Weisberg, S. (2011) An R Companion to Applied Regression, 2nd edn. Sage, Thousand Oaks, California, USA.
  • Fry, F.E.J. (1947) Effects of the environment in animal activity. Publications of the Ontario Fisheries Research Laboratory, 68, 162.
  • Giraudoux, P. (2011) pgirmess: Data analysis in ecology. R package version 1.5.2.
  • Gräns, A., Olsson, C., Pitsillides, K., Nelson, H.E., Cech, J.J. & Axelsson, M. (2010) Effects of feeding on thermoregulatory behaviours and gut blood flow in white sturgeon (Acipenser transmontanus) using biotelemetry in combination with standard techniques. Journal of Experimental Biology, 213, 31983206.
  • Hesthagen, I.H. (1979) Temperature selection and avoidance in the sand goby Pomatoschistus minutus (Pallas), collected at different seasons. Environmental Biology of Fishes, 4, 369377.
  • ICES (2011) Report of the working group on Crangon fisheries and life history (WGCRAN). International Council for the Exploration of the Sea, ICES CM/2011 SSGEF 11.
  • Karlsson, L., Ekbohm, G. & Steinholtz, G. (1984) Comments on a study of the thermal behaviour of the American eel, (Anguilla rostrata) and some statistical suggestions for temperature preference studies. Hydrobiologia, 109, 7578.
  • Kivivuori, L.A. (1994) Temperature selection behaviour of cold- and warm-acclimated crayfish [Astacus astacus (L.)]. Journal of Thermal Biology, 19, 291297.
  • Klimley, A.P., Cech, J.J., Thomspon, L.C., Hamliton, S.A. & Cochrell, D.E. (2011) Experimental and field studies to assess pulsed water flow impacts on the behaviour and distribution of fishes in the South Fork of the American River, Second year. California Energy Commission, PIER Energy-Related Environmental Research Program. CEC-500-2009-067.
  • Lafrance, P., Castonguay, M., Chabot, D. & Audet, C. (2005) Ontogenetic changes in temperature preference of Atlantic cod. Journal of Fish Biology, 66, 553567.
  • Lagerspetz, K.Y.H. & Vainio, L.A. (2006) Thermal behaviour of crustaceans. Biological Reviews, 81, 237258.
  • Lotze, H.K., Reise, K., Worm, B., van Beusekom, J., Busch, M., Ehlers, A., Heinrich, D., Hoffmann, R.C., Holm, P., Jensen, C., Knottnerus, O.S., Langhanki, N., Prummel, W., Vollmer, M. & Wolff, W.J. (2005) Human transformations of the Wadden Sea ecosystem through time: a synthesis. Helgoland Marine Research, 59, 8495.
  • Magnuson, J.J., Crowder, L.B. & Medvick, P.A. (1979) Temperature as an ecological resource. American Zoologist, 19, 331343.
  • Martin, T.L. & Huey, R.B. (2008) Why “suboptimal” is optimal: Jensen's inequality and ectotherm thermal preferences. The American Naturalist, 171, E102E118.
  • Mathur, D., Schutsky, R.M. & Purdy, E.J. (1982) Temperature and avoidance responses of crayfish, Orconectes obscurus, and associated statistical problems. Canadian Journal of Fisheries and Aquatic Sciences, 39, 548553.
  • Mathur, D. & Silver, C.A. (1980) Statistical problems in studies of temperature preference of fishes. Canadian Journal of Fisheries and Aquatic Sciences, 37, 733737.
  • McCauley, R.W. (1977) Laboratory methods for determining temperature preference. Journal of the Fisheries Research Board of Canada, 34, 749752.
  • McMahon, T.E., Bear, E.A. & Zale, A.V. (2008) Use of an annular chamber for testing thermal preference of Westslope Cutthroat Trout and Rainbow Trout. Journal of Freshwater Ecology, 23, 5563.
  • Mortensen, A., Ugedal, O. & Lund, F. (2007) Seasonal variation in the temperature preference of Arctic charr (Salvelinus alpinus). Journal of Thermal Biology, 32, 314320.
  • Myrick, C.A., Folgner, D.K. & Cech, J.J. (2004) An annular chamber for aquatic animal preference studies. Transactions of the American Fisheries Society, 133, 427433.
  • Neill, W.H. & Magnuson, J.J. (1974) Distributional ecology and behavioral thermoregulation of fishes in relation to heated effluent from a power plant at Lake Monona, Wisconsin. Transactions of the American Fisheries Society, 103, 663710.
  • Neill, W.H., Magnuson, J.J. & Chipman, G.G. (1972) Behavioral thermoregulation by fishes: a new experimental approach. Science, 176, 14431445.
  • Ohlberger, J., Staaks, G., Petzoldt, T., Mehner, T. & Hölker, F. (2008) Physiological specialization by thermal adaptation drives ecological divergence in a sympatric fish species pair. Evolutionary Ecology Research, 10, 11731185.
  • Parmesan, C. & Yohe, G. (2003) A globally coherent fingerprint of climate change impacts across natural systems. Nature, 421, 3742.
  • Perry, A.L., Low, P.J., Ellis, J.R. & Reynolds, J.D. (2005) Climate change and distribution shifts in marine fishes. Science, 308, 19121915.
  • R Development Core Team (2011) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org.
  • Reise, K. & van Beusekom, J. (2008) Interactive effects of global and regional change on a coastal ecosystem. Helgoland Marine Research, 62, 8591.
  • Reynolds, W.W. (1978) The final thermal preferendum of fishes: shuttling behavior and acclimation overshoot. Hydrobiologia, 57, 123124.
  • Reynolds, W.W. & Casterlin, M.A. (1979a) Behavioral thermoregulation and the final preferendum paradigm. American Zoologist, 19, 211224.
  • Reynolds, W.W. & Casterlin, M.A. (1979b) Behavioral thermoregulation and activity in Homarus americanus. Comparative Biochemistry and Physiology. Part A, Physiology, 64, 2528.
  • Richards, F.P., Reynolds, W.W. & McCauley, R.W. (1977) Temperature preference studies and environmental impact assessment: an overview with procedural recommendations. Journal of the Fisheries Research Board of Canada, 34, 728761.
  • Staaks, G., Kirschbaum, F. & Williot, P. (1999) Experimental studies on thermal behaviour and diurnal activity rhythms of juvenile European sturgeon (Acipenser sturio). Journal of Applied Ichthyology, 15, 243247.
  • Tepler, S., Mach, K. & Denny, M. (2011) Preference versus performance: body temperature of the intertidal snail Chlorostoma funebralis. The Biological Bulletin, 220, 107117.
  • Tiews, K. (1954) Die biologischen Grundlagen der Büsumer Garnelenfischerei. Berichte der Deutschen Wissenschaftlichen Kommission fuer Meeresforschung, 13, 235269.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
mee312045-sup-0001-SuppInfo.txtplain text document28KData S1. Exemplary MATLAB code for automated analysis of annular chamber experiments.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.