Ecol Res (2005) 20: 497–501 DOI 10.1007/s11284-004-0034-5
NOTE AND COMMENT
P. A. Todd Æ R. J. Ladle Æ R. A. Briers Æ A. Brunton
Quantifying two-dimensional dichromatic patterns using a photographic technique: case study on the shore crab (Carcinus maenas L.)
Received: 13 September 2004 / Accepted: 29 November 2004 / Published online: 18 February 2005 The Ecological Society of Japan 2005
Abstract Contrasting patterns of pigmentation, such as those associated with crypsis and aposematism, are common in many taxa. In order to determine why patterning varies among individuals or populations, it is important to quantify how these patches of pigment are arranged. Here we present a simple technique for measuring areas of pigmentation as well as their spatial distribution, and demonstrate its application to the study of substrate-associated patterning in shore crabs (Carcinus maenas L.). The results, based on a virtual grid laid over digital images of crab carapaces, allow for correlations to be made among sample populations. The technique, or variations of it, can be applied to any situation where two-dimensional dichromatic patterns need to be quantiﬁed. Keywords Dichromatic pattern Æ Shore crab Æ Photographic technique Æ Crypsis Æ Aposematism
Introduction Polymorphisms and polyphenisms are widespread in many taxa and are frequently characterised by diﬀerP. A. Todd Æ R. A. Briers Æ A. Brunton School of Life Sciences, Napier University, Merchiston Campus, Edinburgh, EH10 5DT, Scotland, UK R. J. Ladle School of Geography and Environment, University of Oxford, Mansﬁeld Road, Oxford, OX1 3TB, England, UK Present address: P. A. Todd (&) Marine Biology Laboratory, National University of Singapore, 14, Science Drive 4, Blk S1, #02-05, 117543, Singapore E-mail: [email protected]
Tel.: +65-68741034 Fax: +65-67792486
ences in colour, dichromatic patterns such as spots and stripes, or a combination of both. These patterns are studied in various contexts, including mimicry, crypsis and aposematism (Mallet and Gilbert 1995; Merilaita 1998; Joron and Mallet 1998; Bedini 2002), selective processes (Jones et al. 1977; Forsman and Appelqvist 1999; Palma and Steneck 2001), genetic diﬀerentiation/ phenotypic plasticity (Ekendahl and Johannesson 1997; Wente and Phillips 2003) and habitat selection (Forsman and Shine 1995; Ravigne et al. 2004). To make diﬀerent morphs statistically amenable they are usually categorised into groups, thus enabling frequencies to be compared and associations to be made (e.g. Ekendahl and Johannesson 1997; Kark et al. 1997; Carretero 2002). In most organisms, however, such grouping is arbitrary as intermediate forms and outliers are common. In addition to the information lost via a classiﬁcation approach, researchers cannot be certain their deﬁnitions exactly match those of other studies. Accurately and simply measuring patterns on threedimensional (3-D) objects is likely to remain diﬃcult for some time (Bythel et al. 2001). But dichromatic variability in a 2-D appendage, for example an insect wing, or a plane that is relatively constant among individuals, such as the proﬁle of a ﬁsh, should be quantiﬁable—especially if an image of the appendage or organism can be captured and examined in detail. The combination of digital photography and image analysis software is a powerful tool that has been used in medicine and dentistry (e.g. Benson et al. 1998; Cochran et al. 2004; Hamza and Reddy 2004; Owen et al. 2004,), as well as for various aspects of plant (Hyder et al. 2003, Jia et al. 2004) and animal (Pech et al. 2004; Heide-Jorgensen 2004) biology and ecology. Todd et al. (2001; 2004) demonstrated that photography oﬀers the study of coral morphometrics various advantages including speed of sampling, non-destructiveness, an instant permanent record, and the opportunity for repeat measurements. These same four advantages should also apply to the study of 2-D dichromatic patterns using digital images.
To successfully discriminate among diﬀerent pattern morphs a quantitative approach must be able to measure the relative amount of each colour as well at its shape and spatial orientation. For instance, when examining an eye-spot on a butterﬂy’s wing, the size and shape of the spot, as well as its position, are all important. Furthermore, issues regarding allometry must also be considered, as diﬀerent sized individuals will tend not to have the same absolute quantities of pigment. Here we demonstrate a photography-based technique for quantifying 2-D dichromatic patterns using variation in shore crab (Carcinus maenas) carapaces as a case study. The shore crab is an interesting study species because it exhibits a large range of carapace patterns (especially when young) of which little is known (Hogarth 1975, 1978). If these patterns represent some form of crypsis, then associations between carapace colouration and substrate is expected (Endler 1988; Ekendahl 1998). To help identify any such associations, we have quantiﬁed the carapace patterns of shore crabs from sites where the substrate composition has also been measured.
Materials and methods Due to their pelagic larval phase, shore crabs have high levels of gene ﬂow (Brian 2002); nevertheless, diﬀerent morphs have been associated with diﬀerent substrates (Hogarth 1978). To test whether carapace patterns vary among sites with diﬀering substrates in this study, crabs were sampled from three intertidal rocky shores along the Firth of Forth, Scotland, i.e. Ferny Ness (5559.215¢N, 00254.091¢W), Milsey Bay (5603.674¢N, 00243.016¢W) and Long Craigs (5600.317¢N, 00232.518¢W). At each site a sampling area of 50 · 120 m perpendicular to the shore was delimited. This large area was then subdivided into 15 smaller (10 · 40 m) quadrats, arranged so that there were ﬁve adjacent replicates at each of three shore positions: high, middle, and low. From each shore position, two of the ﬁve replicate quadrats were randomly selected (using
Substrate cover (%)
random number tables) and all crabs found therein were collected (45 min searching time). The substrate composition of these three sites, as determined using line intercept transect methodology (English et al. 1997), is provided in Fig. 1. All patterned crabs were photographed in the laboratory with a high-quality digital camera at 640·480 pixel resolution before being released at a location >20 km from the sampling sites. RGB colour images, as oppose to black and white ones, were used as they were slightly easier to interpret. A millimetre scale (an engineer’s ruler) and a code for sex and quadrat of origin (written on paper) were physically included in each photograph; when each image was downloaded onto computer this information was added as a label. Using a raster-based graphics software [Paint Shop Pro version 7.04 (Jasc Software 2000)] the images were straightened and cropped so that all four carapace extremities (anterior and posterior edge, left and right edge) were just touching the four sides of the frame (Fig. 2) with the rostrum to the top. In Microsoft PowerPoint a 20·20 cell, font size 10, table was created and placed over each imported image. Both table and image were then adjusted so that the two matched exactly. To quantify patterns zeros were allocated to areas of dull green carapace as well as the non-crab regions of the image, whereas ones were used to identify cells that (visually and subjectively) appeared to contain >50% white pigment (Fig. 2). After all 400 cells were ﬁlled, the curser was dragged over the entire table and the data copied and pasted into Microsoft Excel (in most versions of Excel the 400 data points will automatically form into one column) and appropriately labelled. All the patterned crabs from any one quadrat were treated in the same way, thus one spreadsheet contained columns representing individual crabs and 400 rows that represented presence and absence of white pigment. After all
mussel - bed
0 Ferny Ness
Fig. 1 Substrate composition of the three sampling sites. At each site a total of 1,200 m of line intercept transect data were collected
Fig. 2 The technique is based upon a 20 · 20 virtual grid laid over the image. Depending on the level of resolution required, grids with more, or fewer, cells can also be used. Here, for clarity, the noncarapace regions of the image have been blocked out and diﬀerent colours used for ones and zeros
the crab data from one quadrat had been entered, the mean value for each row was calculated. Thus, one column of 400 variables eﬀectively encapsulated the mean quantity, shape, and spatial orientation of white pigment for all patterned crabs from that quadrat. After all crabs from the 18 quadrats had been treated in the same way, a similarity matrix was formed using the correlation statistic in Excel (Pearson correlation coeﬃcient) applied to the 18 summary columns. Due to the high number of zeros and similar spatial arrangement of patterns, these correlation ﬁgures were very high and could not be interpreted in the usual manner. However, the required information was contained within the relative correlations among quadrats. This was analysed by ordinating the correlation matrix using non-metric multidimensional scaling (MDS). We chose to do this with Primer version 5 (Clarke and Gornley 2001) as this programme allows the user to input their own distance (in this case correlation) matrix. The signiﬁcance of differences in patterning among shores was assessed using the one-way analysis of similarity (ANOSIM; Clarke and Warwick 1994) procedure in Primer. To visualise the actual diﬀerences in carapace pattern among sites, the mean row values for each site were reassembled into 20·20 tables and used to create 3-D graphs.
Results and discussion A total of 1,772 crabs were collected of which 789 were patterned: 373 from Ferny Ness, 155 from Milsey Bay and 261 from Long Craigs. All patterned crabs were successfully photographed, and the resultant images analysed. For each crab, the entire process from downloading the image to entering the data into the spreadsheet took 2.5––3 min. Thus approximately 40 h of processing was needed for this, relatively large, data set. Multidimensional scaling ordinates our correlation matrix in two dimensions with a low stress value of 0.07. There was a signiﬁcant among-site diﬀerence in patterning (ANOSIM, global R=0.302, P=0.002). Pairwise comparisons between the sites indicated that there was a signiﬁcant diﬀerence between Ferny Ness and Milsey Bay (R=0.428, P=0.004) and Ferny Ness and Long Craigs (R=0.481, P=0.002), but not between Milsey Bay and Long Craigs (R=0.019, P=0.372). This separation is reﬂected in the MDS plot where Ferny Ness forms a distinct group, but there is a strong overlap between Milsey Bay and Long Craigs (Fig. 3). The 3-D representations of the data for each site indicate that Ferny Ness hosts crabs with more widely distributed white areas, especially in the central and anterior part of the carapace (Fig. 4). Using predator-exclusion experiments, Palma and Steneck (2001) demonstrated that polymorphic rock crabs (Cancer irroratus) possessing non-adult carapace colours had a higher survival rate in polychromatic habitats than monochromatic ones. As the substrate at Ferny Ness is characterised by mussel-
axis 1 = Ferny Ness
= Milsey Bay
= Long Craigs
Fig. 3 This multidimensional scaling plot, based on the carapace pattern correlation matrix, illustrates how the crabs from the Ferny Ness quadrats form a distinct group whereas those from Milsey Bay and Long Craigs overlap
bed and is thus highly polychromatic, it is possible the relationship between this site and strong carapace patterns is not simply stochastic (Ekendahl 1998). Exactly where on the carapace the white spots appear most frequently can be identiﬁed from the 3-D graphs, i.e. the hepatic and facial regions, plus some marking along the anterior–posterior axis (Fig. 4). This provides clues as to whether the colouration is some form of crypsis or mimicry. The high frequency of white spots touching the edge of the carapace is suggestive of disruptive colouration as this type of marking helps break up the outline of an organism (Merilaita 1998). The fact that most of the white pigmentation appears towards the anterior, rather than the posterior, of the carapace may be associated with shore crab behaviour, i.e. their tendency to peer out from under rocks or among mussel shells, thus only revealing their anterior region (P. Todd, personal observation). Finally, the hepatic patches could possibly be interpreted as eye-spots, and thus aposematic. The advantages of photographic techniques discussed in Todd et al. (2001), and outlined in the Introduction of this paper, also apply to the present study. Sampling is fast and non-destructive, only requiring a photograph of the carapace that includes a scale bar and some indicator of sex and site. Downloading images onto a computer or some other form of electronic memory creates a permanent record; and the technique would certainly be suitable for studies of pattern change with growth. The ensuing analysis is also straightforward, and while it may be time-consuming once specimen numbers become high, compared to manual scoring methods the time saving is substantial. The majority of the process can be conducted on a personal computer loaded with standard software (Excel, PowerPoint, Paint Shop Pro); the only specialist programme is the one selected to ordinate the similarity matrix. A virtual grid placed over a crab image means that direct comparisons of patterns can be made among individuals of varying size. If absolute measurements of
c) non-crab areas
Fig. 4 Three-dimensional graphs illustrating the relative mean frequency of white pigment on the carapaces of crabs from a Ferny Ness, b Milsey Bay, and c Long Craigs. The crabs from Milsey Bay and Long Craigs tend to have less white pigment in the central region of their carapaces, as indicated by the depressed areas to either side of the anterior–posterior axis, than those from Ferny Ness
pigment patch size were used, they would have to be adjusted to overall organism/appendage size before comparisons among individuals could be made. Using a grid means that every cell has a constant, isometric (sensu Bookstein 1989), relationship with its neighbouring cells, regardless of size. This constant relationship also means that, when the grid and image are re-sized to match each other, the information extracted from the grid is still an accurate reproduction of the pattern being studied. Issues regarding correlations between pattern and sex, or between pattern and organism size, need to be addressed just as they would if the patterns had been categorised into groups (Hoﬀman and Blouin 2000; Hull and Rollinson 2000). In cases of sexual dimorphism, treating sexes separately is appropriate. Where patterns change with size, organisms/appendages can be pooled by some representative mensuration and a semi-continuous gradient of pattern created. Alternative approaches to statistically explore polymorphic populations grouped by morph type include ANOVA (Palma and Steneck 2001), diversity indices and v2-test (Ekendahl and Johannesson 1997). All these analyses can be adversely aﬀected by low sample size and/or ‘‘empty cells’’. As the technique presented here does not split the populations into smaller components, it is not compromised by low or zero counts of particular morphs. Naturally, there remains a positive relationship between sample size and the information content of the data. Furthermore, a balanced design is preferable and, ideally, each replicate should comprise an equal number of specimens. The primary drawback of the technique is that it is only applicable to dichromatic patterns. For polychromatic patterns each colour has to either be treated separately or weighted. Treating the colours separately means diﬀerent distance matrices for every colour combination of interest, although the matrices themselves can be amalgamated to produce a single result. Alternatively, if some colours are considered more important
than others, they can be assigned a number higher than 0 or 1. This produces a distance matrix that accurately reﬂects the weighting, but the 3-D graphs become redundant as they can only represent two colours. The technique has successfully identiﬁed diﬀerences in shore crab carapace pattern among three sites—information that helps to address questions regarding crypsis, mimicry, and habitat selection. The combination of ordination plots and 3-D graphs provides an easily interpretable summary of large data sets — in the present case, >800,000 data points. Only requiring readily available hardware and software, the process is straightforward. This novel approach could be applied to the proﬁles of higher animals, but is probably best suited to the study of patterns on the appendages and bodies of invertebrates, or on plant leaves and petals. Acknowledgements This study was supported by Napier University, Edinburgh, UK. We would like to thank the numerous students who have assisted with ﬁeldwork and laboratory analysis.
References Bedini R (2002) Colour change and mimicry from juvenile to adult: Xantho poressa (Olivi, 1792) (Brachyura, Xanthidae) and Carcinus maenas (Linnaeus, 1758) (Brachyura, Portunidae). Crustaceana 75:703–710 Benson PE, Pender N, Higham SM, Edgar WM (1998) Morphometric assessment of enamel demineralisation from photographs. J Dent 26:669–677 Bookstein FL (1989) ‘‘Size and shape’’: a comment on semantics. Syst Zool 38:173–180 Brian JV (2002) Inter-population variability in the shore crab (Carcinus maenas) and its potential use as a biomarker of anthropogenic eﬀects. PhD thesis. Napier University, Edinburgh Bythell JC, Pan P, Lee J (2001) Three-dimensional morphometric measurements of reef corals using underwater photogrammetry techniques. Coral Reefs 20:193–199 Carretero MA (2002) Sources of colour pattern variation in Mediterranean Psammodromus algirus. Neth J Zool 52:43–60 Clarke KR, Gornley RN (2001) Primer v5: user manual/tutorial. PRIMER-E, Plymouth, UK Clarke KR, Warwick RM (1994) Similarity-based testing for community pattern—the two-way layout with no replication. Mar Biol 118:167–176 Cochran JA, Ketley CE, Sanches L, Mamai-Homata E, Oila AM, Arnadottir IB, van Loveren C, Whelton HP, O’Mullane DM (2004) A standardized photographic method for evaluating enamel opacities including ﬂuorosis. Community Dent Oral Epidemiol 32:19–27
501 Ekendahl A (1998) Colour polymorphic prey (Littorina saxatilis Olivi) and predatory eﬀects of a crab population (Carcinus maenas L.). J Exp Mar Biol Ecol 222:239–246 Ekendahl A, Johannesson K (1997) Shell colour variation in Littorina saxatilis Olivi (Prosobranchia: Littorinidae): a multifactor approach. Biol J Linn Soc 62:401–419 Endler JA (1988) Frequency-dependent predation, crypsis and aposematic coloration. Philos Trans R Soc Lond B Biol Sci 319:505–523 English S, Wilkinson C, Baker V (1997) Survey manual for tropical marine resources. Australian Institute of Marine Science, Townsville Forsman A, Appelqvist S (1999) Experimental manipulation reveals diﬀerential eﬀects of colour pattern on survival in male and female pygmy grasshoppers. J Evol Biol 12:391–401 Forsman A, Shine R (1995) The adaptive signiﬁcance of colour pattern polymorphism in the Australian scincid lizard Lampropholis delicata. Biol J Linn Soc 55:273–291 Hamza SH, Reddy VVB (2004) Digital image acquisition using consumer-type digital camera in the anatomic pathology setting. Adv Anat Pathol 11:94–100 Heide-Jorgensen MP (2004) Aerial digital photographic surveys of narwhals, Monodon monoceros, in northwest Greenland. Mar Mamm Sci 20:246–261 Hoﬀman EA, Blouin MS (2000) A review of colour and pattern polymorphisms in anurans. Biol J Linn Soc 70:633–665 Hogarth PJ (1975) Pattern polymorphism and predation in the shore crab, Carcinus maenas (L.). Crustaceana 28:316–319 Hogarth PJ (1978) Variation in the carapace pattern of juvenile Carcinus maenas. Mar Biol 44:337–343 Hull SL, Rollinson D (2000) Sex-biased colour polymorphism in the marine ostracod Paradoxostoma variabile (Crustacea). J Mar Biol Assoc UK 80:69–73 Hyder PW, Fredrickson EL, Remmenga MD, Estell RE, Pieper RD, Anderson DM (2003) A digital photographic technique for assessing forage utilization. J Range Manage 56:140–145 Jasc Software (2000) Paint Shop Pro version 7.04. Jasc Software, Minn.
Jia LL, Chen XP, Zhang F, Buerkert A, Romheld V (2004) Use of digital camera to assess nitrogen status of winter wheat in the northern China Plain. J Plant Nutr 27:441–450 Jones JS, Leith BH, Rawlings P (1977) Polymorphism in Cepaea: a problem with too many solutions? Annu Rev Ecol Syst 8:109– 143 Joron M, Mallet JLB (1998) Diversity in mimicry: paradox or paradigm? Trends Evol Ecol 13:461–466 Kark S, Warburg I, Werner YL (1997) Polymorphism in the snake Psammophis schokari on both sides of the desert edge in Israel and Sinai. J Arid Environ 37:513–527 Mallet J, Gilbert LE Jr (1995) Why are there so many mimicry rings? Correlations between habitat, behaviour and mimicry in Heliconius butterﬂies. Biol J Linn Soc 55:159–180 Merilaita S (1998) Crypsis through disruptive coloration in an isopod. Proc R Soc Lond B Biol Sci 265:1059–1064 Owen CG, Ellis TJ, Woodward EG (2004) A comparison of manual and automated methods of measuring conjunctival vessel widths from photographic and digital images. Ophthalmic Physiol Opt 24:74–81 Palma A, Steneck RS (2001) Does variable coloration in juvenile marine crabs reduce risk of visual predation? Ecology 82:2961– 2967 Pech D, Condal AR, Bourget E, Ardisson P (2004) Abundance estimation of rocky shore invertebrates at small spatial scale by high-resolution digital photography and digital image analysis. J Exp Mar Biol Ecol 299:185–199 Ravigne V, Olivieri I, Dieckmann U (2004) Implications of habitat choice for protected polymorphisms. Evol Ecol Res 6:125–145 Todd PA, Sanderson PG, Chou LM (2001) Photographic technique for the study of coral morphometrics. Raﬀ Bull Zool 49:191–195 Todd PA, Ladle RJ, Lewin-Koh N, Chou LM (2004) Flesh or bone? Quantifying small-scale coral morphology using withtissue and without-tissue techniques. Mar Biol 145:323–328 Wente WH, Phillips JB (2003) Fixed green and brown color morphs and a novel color-changing morph of the Paciﬁc tree frog Hyla regilla. Am Nat 162:461–473