Cory W. Child^[1] and Nicholas J. Jouett^[2] , Department of Marine Science, Eckerd College, Florida, and College of Environmental and Life Sciences, University of Rhode Island

Abstract

The Asian shore crab, Hemigrapsus sanguineus, is invasive to the northern east coast of the USA. The crab has established itself in its new environment successfully, inhabiting rocky shorelines from Maine to the Carolinas, sometimes as the most abundant crab present. Though H. sanguineus has been the subject of many invasion studies, its ecology has yet to be fully understood. The crab has, however, been found to be dependent on structurally complex, rocky shorelines for shelter and mobility, although the relationship between complexity and density is understudied due to the difficulties of measuring structural or habitat complexity in situ. The current study has found a preliminary proxy method for measuring the complexity of rocky shorelines to reliably predict H. sanguineus density. We used a photo-quadrat system with software analysis to identify individual two-dimensional rock metrics. Regression and ANOVA analyses revealed a statistically significant relationship that can be used to predict H. sanguineus population densities. Our findings have implications for identifying at-risk coastlines and guiding initial management strategies. Outside of this scope, our findings could be applied to predicting different ecological factors that are dependent on the structural complexity of rocky shorelines, and they can be paired with aerial photography in order to create a streamlined, high-throughput process for preliminary ecological analyses.

Keywords: Predictive modelling, invasive species, intertidal ecology, Hemigrapsus sanguineus, habitat complexity.

Introduction

The Asian shore crab, Hemigrapsus sanguineus (De Haan, 1835) is a brachyuran crab native to rocky coastlines of eastern Asia. This highly mobile and fecund species (Fukui, 1998; McDermott, 1998) has successfully invaded rocky shorelines of the USA and areas of Europe (Breton et al., 2002). It was first observed in the USA near Delaware Bay in 1988 (Ai-Yun and Yang, 1991; McDermott, 1991; Williams and McDermott, 1990). Recognising an imminent biological invasion, numerous investigators took the opportunity to research and document the invasion carefully. Since its arrival to the USA, which was estimated to be in the early 1980s via ballast water (McDermott, 1998: 289), the Asian shore crab has spread from Delaware to rocky shores as far north as Schoodic peninsula in Maine and as far south as Cape Hatteras (Delaney and Sperling, 2008: 118) to become the most abundant intertidal crab at many locales (Ahl and Moss, 1999; Lohrer and Whitlatch, 1997; Griffen, 2011; Kraemer et al., 2007), reaching densities of >300 crabs/m2 in extreme cases (McDermott, 1998: 297). Numerous investigations have sought to determine the ecological impact that this alien crab may have in both its current and potential range.

Hemigrapsus sanguineus has displaced native crabs at many locations in both Europe and the USA (Breton et al., 2002; Bourdeau and O'Connor, 2003; Grosholz et al., 2000; Lohrer et al., 2000). Lohrer and Whitlatch (2002) found a decline in green crab abundance with a likely related increase in H. sanguineus abundance. The species' diet overlaps extensively with those of other crabs (Tyrrell and Harris, 2001: 214), and it is also superior to other crabs at seizing intertidal shelter (see Jensen et al., 2002). Laboratory and field observations indicate that this generalist predator has the potential to affect several sympatric species of molluscs and crustaceans including commercially important shellfish species such as Mytilus edulis, Mya arenaria, and Crassostrea virginica (Brousseau et al., 2001; Brousseau and Baglivo, 2005; Griffen and Byers, 2009; Lohrer and Whitlatch, 2002; Seed, 1980). Hemigrapsus sanguineus has a large vertical range between the low- to high-tide mark, meaning it interacts with a wide range of species, from the aforementioned invertebrates to fish (Brousseau et al., 2000; Heinonen and Auster, 2012). As a whole, results and findings predict or have shown that H. sanguineus may significantly impact rocky intertidal communities (Epifanio, 2013), with potential impacts as both a predator and prey item (Kim and O'Connor, 2007, Brousseau et al., 2008).

Early on, it was noticed that a critical requirement for H. sanguineus was a certain degree of rocky, complex structure (Williams and McDermott, 1991: 108). In fact, it is thought that the crab's current southern range in the USA is limited due to the lack of rocky substrate (Epifanio, 2013: 35). Structurally complex environments provide protection from predators (Grabowski, 2004), thereby increasing foraging proficiency and mobility to adjacent areas (Lohrer et al., 2000). Small shelters serve as a buffer against temperature change (Taylor, 1981; Abele, 1986), to which H. sanguineus is sensitive (Epifanio et al., 1998). Numerous studies have investigated the role that substrate plays on metamorphic and settling cues for crab megalopae. Hemigrapsus sanguineus megalopae respond to an insoluble cue present in biofilms associated with rocky intertidal habitats (Anderson and Epifanio, 2009). The combination of biofilm and structural complexity was investigated, and it was physical aspects such as roughness and rock size that were determined to be the critical contributors to metamorphic cues (O'Connor, 2007). Additionally, O'Connor (2007) found a strong response from megalopae to conspecifics adults.

It is therefore clear that H. sanguineus heavily relies on rocky structural complexity, and this assertion has been investigated previously (Ledesma and O'Connor, 2001). Ledesma and O'Connor (2001: 68) linearly correlated rock cover, estimated via eye by two observers, to H. sanguineus abundance. The R2 value of their results was 0.53. While this is a strong relationship and does accurately describe what is found in situ to an extent, rock cover percentage is not a fully adequate metric, as it fails to take into account the number, size and shape of the rocks, which contributes greatly to structural complexity.

Structural or habitat complexity is an often understudied field which is usually limited to trees (MacArthur, 1964) or aquatic benthic habitats (Kohn and Leviten, 1976; Diehl, 1992), often coral reefs (Luckhurst and Luckhurst, 1978). Rocky intertidal environments have also received some attention (see Kovalenko et al., 2012). In most instances it has been observed that habitat complexity increases diversity (Kostylev et. al., 2005). For example, in the rocky intertidal, both gastropod diversity and abundance correlate positively with structural complexity (Beck, 2000: 45). Three-dimensional measurements of structural complexity offer great precision, but they are laborious and difficult to obtain. Fractals are a popular tool for measuring complexity in three dimensions (Tokeshi and Arekaki, 2012) and have been employed in the rocky intertidal to demonstrate the relationship with several macrofauna. Specifically, Kostylev et al. (2005) looked at how fractals dictated the abundance and diversity of several sessile organisms and some mobile gastropods in the intertidal, but there are no studies that look this in depth at highly mobile organisms, like H. sanguineus.

In light of these realisations, the current study seeks to combine several easily measured metrics to improve upon past findings. Previous work uses either methods too complex for large-scale applications or methods too simple to be reliable. This new approach seeks to find a middle ground, sacrificing some of the accuracy of more laborious methods in favor of large-scale applicability. Due to H. sanguineus' reliance on structural complexity, the many difficulties associated with accurately measuring this metric in the field, and the lack of high-throughput potential for using fractal geometry methods, our study seeks to provide an initial model for predicting H. sanguineus population densities based on rock metrics that can be easily extracted from photographs of rocky shorelines as a proxy for structural complexity.

Materials and methods

Field collection

Two sites, Gardner Island and Beach Island (41° 24' 22.83' N, 71° 30' 21.40' W), were sampled a total of eight times between July and August 2013 in Point Judith Pond, a small protected estuary in Rhode Island, USA. For comprehensive information concerning the salt pond, see Stout (2006). Both islands have rocky shores and are easily accessible by pontoon boat. Each collection was carried out as the tide was receding and at least below the mid-tidal line between the hours of 10:00 and 17:00. We conducted three collections at Beach Island and five collections at Plato Island over the course of 45 days. Each collection was carried out on a different day.

Each sampling consisted of twenty-one 0.5 m2 quadrats taken at equal intervals along 33m transects that ran parallel to the shoreline. The transects were laid 1-3m up from the waterline. Fully submerged crabs were therefore not collected, though upon investigation few, if any, were present (personal observation by Jouett). Hemigrapsus sanguineus is reported to be subtidal only during winter months, and therefore this sampling technique has been determined to be sufficient (Takahashi et al., 1985; McDermott, 1998).

Quadrats were photographed first by a Nikon D40 (Nikon, Chiyoda, Tokyo), and then rocks were cleared by hand and H. sanguineus specimens collected. Large rocks that were partially (1/3 or more) in the quadrat were overturned in order to collect crabs that were still within the quadrat. Hemigrapsus sanguineus is reliably identifiable by the three anterolateral teeth on its square carapace and dark banding on its legs (see McDermott, 1991: 197).

photoQuad analysis

The programme photoQuad 'is a custom software for advanced image processing of 2D photographic quadrat samples, dedicated to ecological applications' (Trygonis and Sini, 2012). We used photoQuad to analyse photographs and determine the various rock metrics used in this study as a proxy for structural complexity. These are detailed as follows: the number of crabs was compared against the number of cobble-sized rocks (defined herein as rocks ≥68 cm2 according to the acquired photograph); the area of the largest rock in the quadrat; the area of the average rock size within the quadrat (out of rocks ≥16 cm2 according to the acquired photograph); and the total coverage of all rocks in the quadrat ≥16 cm2. It is important to note that we determined that these metrics were sufficient as proxies based on personal observations and that they serve only as a baseline in this pilot technique to synthesise a useful model.

Statistical analysis

Count data was not log transformed because there was a recurring instance of zero crabs within a quadrat. Log transformation of zeros therefore presented a problem, and thus this was the justification for eliminating a log transformation (O'Hara and Kotze, 2010). Additionally, due to the often high zero count within discrete response variables, these datasets are unlikely to have a normal distribution (Sileshi et al., 2009). Percent coverage data was arc-sine transformed in order to meet assumption of normality for ANOVA (Zar, 1999). Basic statistical analyses were carried out using Microsoft Excel and R.

Due to the excess of zero counts, a zero-inflated Poisson (ZIP) regression in R with the pscl package was used in order to model count zeros separate from excess zeros (see Zeileis et al., 2008). A Vuong test (Vuong, 1989) was used to compare the ZIP regression to an ordinary Poisson regression. The subsequent multiple linear regression that was employed based on the results of the Vuong test used the statistical software package PAST (version 3.02, Hammer et al., 2001). A total of 155 quadrats were surveyed, and these provided the inputs for the various statistical analyses.

Results

A Chi-square test demonstrated the overall significance of our model (Table 1). All individual metrics, through simple generalised linear modelling, correlated positively with H. sanguineus density (Table 2), but were found to be inferior when compared to the multivariate analysis (Table 3). These statistical methods revealed a significant relationship between the rock metrics as a predictor for the number of crabs.

Poisson Model	Estimate	Standard Error	Z	P
Intercept	-0.6933	0.2846	-2.435	0.0149
Total Coverage	4.472	0.8959	4.992	5.98E-7
# Cobble	-0.0896	0.0422	-2.123	0.0337
Avg. Rock	0.0081	0.0016	4.917	8.78E-7
Largest Rock	-0.0004	0.0007	-0.653	0.5138

ZIP Model	Estimate	Standard Error	Z	P
Intercept	3.243	1.0047	3.228	0.0012
Total Coverage	-8.532	3.323	-2.567	0.0102
# Cobble	0.1386	0.1645	0.843	0.3992
Avg. Rock	-0.0055	0.0145	-0.381	0.7029
Largest Rock	-0.0057	-0.0056	-1.019	0.3084

Table 1: Poisson model versus ZIP model. Via Vuong test, Poisson model > ZIP model, with p-value 2.22E-16. Chi-Square test- Degrees of freedom=10, p-value 1.44E-20.

Table 2	Coefficient	Std. Error	t	p	R2
Constant	-3.9675	0.76601	-5.1794	7.059E-07	-
Cobble	-0.4542	0.15735	-2.8865	0.0044711	0.080034
Largest	-0.00022921	0.0040204	-0.057012	0.95461	0.14788
Average	0.037772	0.011089	3.4063	0.00084531	0.12146
Cover	15.313	2.9913	5.1193	9.2683E-07	0.19508

Table 2: Individual rock metrics as they correlate with the number of crabs.

Dependent Variable	Crabs
N (#quadrats)	155
Multiple R	0.55385
Multiple R2	0.30675
Multiple R2 Adjusted ANOVA	0.28826
F	16.593
df1, df2	4,150
p	2.8001E-11

Table 3: Multivariate analyses of rock metrics extracted from photographs as they correlate with the number of crabs.

Figure 1: The degree of departure from observed crab abundance using the proposed model a priori. The bin 'More' includes values of ≥4 individuals. The greatest residual in our study was 12.

Figure 1 displays the residuals when our model is used to analyse our photographs a priori. 'Degree of departure' means by how many crabs the prediction was off, with negative values meaning there were fewer crabs than expected, and positive values meaning there were more crabs than expected. A value of zero means the model correctly predicted the number of crabs. Crabs are binned by values of 0.5 for the sake of relative accuracy and to remove rounding bias.

Discussion

The results of this preliminary study show a statistically significant relationship between the selected rock metrics and the population density of H. sanguineus. Unlike fractals, which require complex fieldwork, our technique is synergistic with aerial image acquisition and could be used to predict, immediately and reliably, H. sanguineus populations at any given site through photoQuad or related software analysis. This removes the necessity of fieldwork, thereby increasing throughput. It therefore has value as a rapid preliminary assessment method. However, it is important to note that this is a preliminary model and that its use comes with certain assumptions, and by extension inaccuracies. Our approach, no matter how much it is improved, can only utilise two-dimensional metrics. The greatest shortcoming our model had was underestimating crabs/ quadrat. Residuals taken from our study have an approximately normal distribution, with 80% of the quadrats possessing accuracy within two H. sanguineus individuals. The only departure from a typical normal distribution is that 12 of the 155 quadrats underestimated the amount of crabs by four individuals or more. Conversely, in 72 of the 155 quadrats, crabs were underestimated, but never by four or more individuals.

The multiple linear regression analysis produced an R value of 0.55, which is stronger than any relationship generated with our univariate linear comparisons. Additionally, the ANOVA's p-value (2.8E-11) strongly suggests that the trends observed are not due to random sampling. Therefore, it appears that this study's analysis is not only objective (free of estimation), but also significant and may have applications to other intertidal species dependent on structural complexity (see Tokeshi and Arekaki, 2012; Kohn and Leviten, 1976; Beck, 1998; Beck, 2000; Kostylev et al., 2005). In terms of density, although potential estimates have put maximum H. sanguineus/m2 at 320 individuals (McDermott, 1998: 297), this seems far from the average. The current results are the same as previous results in both Japan and the USA (see Takahashi et al., 1985; McDermott, 1998).

Conclusion

The results herein have demonstrated a significant relationship between the selected rock metrics and the density of H. sanguineus in a given location. In the future, other rock metrics may be considered. For instance, future studies could employ the highest vertical height of a rock as an additional metric, which could strengthen the relationship. In order to test the relationship shown herein, we propose that an area be photographed beforehand, possibly by aerial means, and then surveyed by field investigators so that the actual results could be compared to initial predicted results. Structural complexity is an important and difficult-to-quantify aspect of intertidal environments, and there are many other ecological aspects with which it shares a relationship (Pianka, 1988). This study may therefore serve as a baseline for future investigations involving a range of intertidal organisms (see Underwood and Denley, 1984; Menge et al., 1985) and may have applications for studying other complex habitats, such as forests. Fractals are a useful technique, but they are laborious and lack the ability to become more streamlined. Our technique should be complimented by aerial image acquisition in order to create a high-throughput method that is reliable enough to find use in initial assessments, which would be valuable for management strategies of H. sanguineus.

---

Acknowledgements

The authors would like to thank YMCA Camp Fuller for its facilities and Prentice Stout for his mentoring and dedication to its marine program over the many years. We would also like to thank Nikki Sabatino for assistance in the field, Laura Young, and all the campers who participated in fieldwork. Last but not least, we would like to thank Jaydon Gianfrancesco, who has a bright future in this field and was immensely helpful during fieldwork.

List of figures

Figure 1: The degree of departure from observed crab abundance using the proposed model a priori.

List of tables

Table 1: Poisson model versus ZIP model. Via Vuong test, Poisson model > ZIP model, with p-value 2.22E-16. Chi-Square test- Degrees of freedom=10, p-value 1.44E-20.

Table 2: Individual rock metrics as they correlate with the number of crabs.

Table 3: Multivariate analyses of rock metrics extracted from photographs as they correlate with the number of crabs.

Notes

[1] Nicholas J. Jouett received his BSc at the University of Rhode Island in RI, USA in Marine Biology and Molecular Biology. He is currently working as a food microbiologist.

[2] Cory W. Child is finishing his undergraduate studies at Eckerd College in FL, USA. He is studying Marine Science with an emphasis in Biology and psychology.

References

 

To cite this paper please use the following details: Child, C.W. and Jouett, N.J (2015), 'Pilot Study Posits Predictive Model for Population Densities of an Invasive Intertidal Brachyuran Crab in New England, USA', Reinvention: an International Journal of Undergraduate Research, Volume 8, Issue 2, http://www.warwick.ac.uk/reinventionjournal/issues/volume8issue2/child Date accessed [insert date]. If you cite this article or use it in any teaching or other related activities please let us know by e-mailing us at Reinventionjournal at warwick dot ac dot uk.