Skip to main content
Dryad logo

Data from: Intertidal gobies acclimate rate of luminance change for background matching with shifts in seasonal temperature

Citation

da Silva, Carmen et al. (2020), Data from: Intertidal gobies acclimate rate of luminance change for background matching with shifts in seasonal temperature, Dryad, Dataset, https://doi.org/10.5061/dryad.zkh189371

Abstract

1. Rate of colour change and background matching capacity are important functional traits for avoiding predation and hiding from prey. Acute changes in environmental temperature are known to impact the rate at which animals change colour, and therefore may affect their survival. 2. Many ectotherms have the ability to acclimate performance traits such as locomotion, metabolic rate and growth rate with changes in seasonal temperature. However, it is unclear how other functional traits that are directly linked to behaviour and survival respond to long-term changes in temperature (within an individual’s lifetime). 3. We assessed whether the rate of colour change is altered by long-term changes in temperature (seasonal variation) and if rate of colour change can acclimate to seasonal thermal conditions. We used an intertidal rock-pool goby, Bathygobius cocosensis, to test this and exposed individuals to representative seasonal mean temperatures (16 ºC or 31 ºC, herein referred to cold and warm-exposed fish, respectively) for nine weeks and then tested their rate of luminance change when placed on white and black backgrounds at acute test temperatures 16 ºC and 31 ºC. We modelled rate of luminance change using the visual sensitives of a coral trout (Plectropmus leopardus) to determine how well gobies matched their backgrounds in terms of luminance contrast to a potential predator. 4. After exposure to long-term seasonal conditions, the warm-exposed fish had faster rates of luminance change and matched their background more closely when tested at 31 ºC than at 16 ºC. Similarly, the cold-exposed fish had faster rates of luminance change and matched their backgrounds more closely at 16 ºC than at 31 ºC. This demonstrates that rate of luminance change can be adjusted to compensate for long-term changes in seasonal temperature. 5. This is the first study to show that animals can acclimate rate of colour change for background matching to seasonal thermal conditions. We also show that rapid changes in acute temperature reduce background matching capabilities. Stochastic changes in climate are likely to affect the frequency of predator-prey interactions which may have substantial knock-on effects throughout ecosystems. 13-Mar-2020

Methods

Bathygobius cocosensis (n = 80) were collected from the rocky inter-tidal zone at Point Cartwright, SE Queensland, Australia (26.6804°S, 153.1390°E) in November 2016. All fish were collected using a battery-operated bilge-pump and hand nets. Fish (22 – 47 mm standard length – tip of nose to last vertebrae) were transported to The University of Queensland by vehicle in oxygen-saturated bags within an insulated container. Fish were anesthetized (0.3 x 10-3 mg L-1 of Aqui-S®) (Malard, McGuigan & Riginos 2016) and tagged dorsally between their head and first dorsal fin, with unique Visible Implant Elastomer (VIE) florescent subdermal tags (Northwest Marine Technologies®, Inc.) for individual recognition. VIE tags do not affect growth rates or survival in fish (FitzGerald, Sheehan & Kocik 2004). Fish were split into two treatments: warm fish (n = 40) were randomly allocated to 6 tanks (~ 7 fish per tank) set at 31 °C and cold fish (n = 40) were randomly allocated to 6 tanks (~ 7 fish per tank) set at 16 °C. Fish were housed in 60L glass tanks with shells provided for shelter. Fish were exposed to a 12:12hr 6am – 6pm light dark cycle in a controlled temperature room at The University of Queensland. Fish were held for nine weeks in treatment conditions before the start of luminance change testing, as it usually takes fish several weeks to acclimate to changed thermal conditions (Seebacher et al. 2005). Unhealthy fish were removed from the experiment.

Before testing, fish were brought to test temperature at a rate of 5 °C per hour, as deemed an appropriate rate for intertidal fish (Schulte, Healy & Fangue 2011). Rate of change in dorsal body luminance was measured in a full factorial design where individuals from warm and cold-exposed treatment groups were tested against both white and black backgrounds at 16 °C and 31 °C. This resulted in four test groups: white background at 16 °C; white background at 31 °C; black background at 16 °C; black background at 31 °C. Fish from each treatment group were randomly assigned to test conditions (by temperature and colour) so order of test temperature or colour would not affect results (a fully crossed experimental design). Fish were tested in a controlled temperature room (16 °C or 31 °C) at The University of Queensland under LED lights (Arlec 9 Watt Slim Bar Lights) to ensure environmental lighting conditions were consistent across trials.

Experimental preparation

For testing, containers were created by spray-painting matt black or white (MMP industrial Pty Ltd, Mulgrave, NSW, AUS) onto PVC plastic sheets to line the inside of 8cm x 15cm x 4cm plastic testing containers. Once dry, the luminance of the paint colour was obtained by taking a photograph of the painted containers with a calibrated Samsung NX1000 with Nikkor EL 80mm lens camera. We used the Multispectral Image Analysis and Calibration (MICA) Toolbox plugin (Troscianko & Stevens 2015) in the program ImageJ (https://imagej.nih.gov/ij/) to normalise and linearise the photograph. Reflectance of the black and white backgrounds was measured as the sum of the red, green and blue channel of a given background relative to a hypothetical 100% white standard. We used a spectrophotometer (USB2000 Ocean Optics©, Largo, FL USA) (https://oceanoptics.com/) with a 180° cosine corrector, 400mm bifurcated optical fiber cable and a PX-2 light source  to determine the luminance of the grey standard (black and white Kodak colour squares), which were used to normalise and linearise each photograph, and to determine the illuminant light spectrum provided by our artificial light source  (Arclec 9 Watt Slim Bar Lights). For holding fish before they were tested, intermediate grey backgrounds were produced by printing a grey 50% luminance background onto Kodak printing paper using an HP laser jet printer (HP LaserJet Pro 400 colour M451dn) for the fish to be held in prior to testing. The intermediate background was then laminated with matt sheets to reduce reflection and to ensure they were waterproof.

 

Experimental protocol

Fish were first placed into the intermediate grey background for 10 minutes to ensure that all fish were exposed to the same conditions before testing. The fish were photographed at the end of the 10 minutes using the calibrated Samsung NX1000 with Nikkor EL 80mm lens camera, with a fixed aperture, manual white balance settings, and with a grey standard (Kodak colour squares) for image calibration within the photo frame. Photographs were taken in RAW format. Fish were swiftly (< 5 seconds) transferred with a small dip net into either the white or the black background container at either 16 °C or 31 °C and photographed immediately and once every 15 seconds for 10 minutes with the colour standard for image calibration within each photo. Each fish was tested separately to avoid stimulating a behavioural change in luminance. On the last day of testing, mass (g) and standard length (mm) of each fish was measured.

 

Image Analysis

We quantified dorsal body luminance of gobies through the eye of a potential predatory coral trout (Plectropmus leopardus). To do this, we used the MICA Toolbox plugin (Troscianko & Stevens 2015) in the program ImageJ (https://imagej.nih.gov/ij/). Images were calibrated and turned into 16-bit multispectral images using a grey standard with 73.3% and 5.1% reflectance (Figures 1 & 2). Colour vision modelling using calibrated digital photography relies on detailed knowledge about the spectral sensitivities of both the potential animal viewer, the camera settings and the illumination within the image. Coral trout have single cones that contain a short wavelength sensitive pigment (λmax = 455nm); and double cones that exhibit broad absorbance spectra ranging from 507 to 532 nm (mean λmax = 522 nm) (Cortesi et al. 2016). In fish, luminance is likely to be processed by double cones, as it is in birds and reptiles (Lythgoe, 1979); therefore, we used the mean λmax of double cone members (522 nm) to model luminance perception. Although coral trout are unlikely to be the main predator of B. cocosensis along Australia’s East coast, their visual capacity is likely to be similar to many other predatory teleosts (Losey et al. 2003).

Within the MICA toolbox we modelled double cone photoreceptor stimulation (fdbl) as a proxy for luminance. The visual (cone mapping) model included information on the type of camera (Samsung NX1000), lens (80mm Nikkor EL), the artificial lighting used to take the photographs (Arlec 9 Watt Slim Bar Lights), the coral trout visual sensitives (Cortesi et al. 2016), and the model illuminant (400-700nm daylight) (gobies are found in very shallow water so we modelled them being in a clear-sky daylight spectrum) (cone catch model in supplementary material). Examples (reconstructed RGB images) of cone catch images of gobies against black and white backgrounds through coral trout in comparison to human colour vision are shown in Figures 1 and 2.

Using the visual models for each photograph, we assessed the luminance of each goby at each time point as the median coral trout double cone stimulation across our region of interest (ROI). An ROI was drawn around the inside edge of each goby (a triangle from their pectoral fins down their body and tail to avoid the elastomer tag) and calculated the median luminance. As thousands of photos were generated for this experiment (n > 5000), we used an ImageJ based script to partially automate this process for each photograph (ImageJ script in supplementary data).

We calculated the achromatic perceptual distance (∆S) (how closely the goby matches its background through the eye of a coral trout in terms of perceived brightness), between each goby and its visual background using the receptor noise limited model (Vorobyev & Osorio 1998). The receptor noise limited model assumes the inherent noise in photoreceptors ultimately limits contrast perception. Animal-background contrast was measured at each time point and against white and black backgrounds to assess the achromatic contrast between the goby and its background throughout the duration of testing. Achromatic perceptual distance (∆S) was calculated according to equation 4 & 7 from Siddiqi et al. (2004) (eq. 1 & 2)

Δfdbl=lnl1-lnl2          (equation 1)

∆S=Δfdbl/ωdbl                 (equation 2)

 

where, l1 is the % double cone stimulation (fdbl) of the background (white or black background), l2 is the % double cone stimulation (fdbl) of the goby polygon (ROI), and ωdbl is the weber fraction of the double cone receptor channel as per Cortesi et al. 2006. Achromatic perceptual distance was measured at the start and end of each trial to measure how well matched a fish was to its background. The change in achromatic perceptual distance between those two times was calculated to determine if fish achieved a greater match during the trial. Threshold values that determine if a change in perceived achromatic distance are distinguishable (i.e. whether goby can be distinguished from its background by the coral trout) have been suggested to be above 1∆S (Siddiqi et al. 2004), however, we used a conservative threshold assumption of 3∆S as suggested by Abernathy et al. (2017). Therefore, the greater the achromatic perceptual distance, the more contrasting the goby is against its background, increasing the likelihood of detection by a coral trout. The lower the achromatic perceptual distance, the harder the goby is to detect against its background to the coral trout.

 

Statistical analysis

Rate of Luminance Change

We assessed rate of luminance change over the first four minutes of testing (rather than the full ten minutes test period) as the slopes were generally steepest and most linear over this time (Supplementary Figure 1). Using the first four minutes of luminance change allowed us to identify how an animal’s short-term background matching response might be affected by a change in water temperature (i.e. if an animal takes many minutes to change luminance they are more likely to be noticed than an individual that can match their background within a few seconds or minutes). When comparing rates of luminance change positive slopes indicate a brightening in fish skin pigment and negative slopes indicate darkening.

To assess whether long-term exposure to warm or cold thermal conditions altered the rate at which gobies changed their luminance at different test temperatures, we ran linear mixed effect models using the package nlme (Pinheiro et al. 2018) for each background colour (black or white) in the statistical program R version 1.1453 (R Core Team 2018). Long-term exposure (warm or cold) was set as a categorical variable and test temperature (16 ºC or 31 ºC) was set as a continuous variable. Fish mass, standard length and test time were included as covariates, and fish identification number was nested within tank number and was set as a random variable. Fish length was eventually removed from the final model because it was correlated with fish mass. An interaction between acute test temperature and long-term thermal exposure was included in the full model. Model reduction was used to test if the interaction between acute test temperature and long-term thermal exposure had a significant effect on rate of luminance change to assess if B. cocosensis could acclimate rate of luminance change to long-term thermal exposure. Models were compared with one another using a likelihood ratio test using the lmtest library in R (Zeileis & Hothorn 2002). As we hypothesised that B. cocosensis would acclimate rate of luminance change to seasonal thermal conditions, we expected that there would be a significant interaction between long-term exposure temperature (warm or cold) and test temperature on rate of luminance change when the fish are tested against both black and white backgrounds. We also predicted the sign of the estimate of the interaction term to determine the directionality of the effect of long-term thermal exposure and acute test temperature on rate of luminance change against black and white backgrounds. Against a black background we would expect the estimate of the interaction term to be negative, as the cold-exposed fish are likely to turn darker faster (lower luminance value) at 16 ºC than 31 ºC (positive slope) and the warm-exposed fish to turn darker faster at 31 ºC than 16 ºC (negative slope). Therefore, the difference between the cold- and warm-exposed fish rates of luminance change across test temperature is likely to produce a negative estimate for the interaction term between test temperature and long-term thermal exposure group (cold-exposed group is always used as the reference group) (Faraway 2016). In contrast, we expect the estimate for the interaction term between long-term thermal exposure and acute test temperature to be positive when fish are tested against a white background. In this case, we expect the cold-exposed fish to turn lighter faster at 16 ºC than 31 ºC (negative slope), and the warm-exposed fish to turn lighter faster at 31 ºC than 16 ºC (positive slope). Therefore, the difference in slope (interaction term estimate) between the cold-exposed and warm-exposed fish is likely to be positive.

Long-term effect of temperature on luminance

We assessed if long-term thermal exposure affected the starting luminance of B. cocosensis against the white, black and intermediate grey backgrounds. We used separate linear models (black, white and intermediate grey background) to compare the starting luminance (first photograph taken during the 10 minute trial) of warm- and cold-exposed fish. Linear models included long-term exposure group as categorical variable and fish mass as a continuous variable. Like the statistical methods outlined above, we used model reduction and comparison using likelihood ratio tests to examine if long-term exposure group had an effect on starting luminance. 

Coral trout visual perception

We assessed if the gobies became more or less distinguishable from their backgrounds to the coral trout throughout the duration of the testing period by assessing their total change in achromatic perceptual distance (10 minutes). We calculated mean achromatic perceptual distance at the start (0 seconds) and end (10 minutes) of the testing period and calculated the total change in achromatic perceptual distance to determine the degree of ecologically significant luminance change the gobies underwent for each test temperature and background. We used the full 10-minute test period, rather than the first four minutes to capture how well the gobies could eventually match their background (slower background matching response). We used the same model structure as “rate of luminance change” above to assess how long-term thermal exposure and test temperature affected total change in achromatic perceptual distance over time. Greater positive changes in achromatic perceptual distance over time indicate that the gobies became less distinguishable against their background, negative values indicate that the goby became more distinguishable against its background. Again, we expect there to be a significant interaction between long-term thermal exposure group and test temperature on total change in achromatic perceptual distance indicating that B. cocosensis can acclimate total change in achromatic perceptual distance with long-term thermal exposure. We expect the warm-exposed fish to have greater total changes in achromatic perceptual distance at 31 ºC than 16 ºC against black and white backgrounds (positive slope), and the cold-exposed fish to have greater changes in achromatic perceptual distance at 16 ºC than at 31 ºC (negative slope). Therefore, we expect the estimate of the interaction term to be positive for both the black and the white background total change in achromatic perceptual distance models. All figures were produced using ggplot2 (Wickham 2016).