A comparison of two techniques for the rapid assessment of marine habitat complexity



  1. Monitoring and assessment of the status and distribution of marine seabed habitats is needed to support existing and emerging environmental policy commitments. Traditional monitoring of habitats and associated species using grabs and trawls is costly and labour intensive and might usefully be complemented by cheaper and more readily automated methods that can be used at higher frequencies and/or on larger spatial scales.
  2. We develop and apply two methods to measure seabed habitat complexity and demonstrate how they can be used to describe impacts (e.g. fishing gear impacts) and monitor recovery. The first method relies on the analysis of deviations in a laser line projected on the seabed. The second method is based on the pixel value distribution in seabed photographs. We use both methods to quantify the complexity created by different substrates and habitat-forming species and to establish links between habitat complexity and faunal diversity (richness) and abundance.
  3. The habitat complexity index calculated with the laser line method provided a reliable index of complexity across a range of habitat types, showing a monotonic increase with coarseness of the substratum and the abundance of sessile epifauna. Pixel value distributions in the photographs did not reflect the increase in complexity due to sessile epifauna but only reflected substratum differences.
  4. Results suggested that the laser line method would be suitable for monitoring the effect of disturbance on habitats ranging from gravelly sands to rock, and their subsequent recovery. The photographic method would be better suited to assessing complexity and heterogeneity of the substratum. Both methods complement conventional biological sampling and can be used at higher frequencies and/or on larger spatial scales per unit cost.
  5. The laser line method has considerable potential to support demands for frequent monitoring of seabed habitats and human impacts at a range of spatial scales. It is less costly and labour intensive than existing approaches and can be deployed from vessels of many sizes.


The requirement to protect marine habitats and biodiversity is articulated in national, regional and international guidelines and policies [e.g. Habitat Directive (EC 1992), Magnuson-Stevens Fishery Conservation and Management Act in the US (U.S. Congress 1996) and the Convention on Biological Diversity (UNEP/CBD 2010)]. To support such guidelines and policies, both the distribution and status of different habitat types have to be described, monitored and reported. This information is used to assess human impacts and to determine the need for, and performance of, management measures.

The distribution and status of habitat can be described from the identity and abundance of component species, but the approach can be too costly and time-consuming to apply at large spatial scales. However, to meet many operational needs, information on habitat complexity may be sufficient to assess state and impact, based on the assumption that complexity will be linked to biodiversity and function (e.g. Crowder & Cooper 1982; Wildish & Kristmanson 1997; Bruno & Bertness 2001; Bolam, Fernandes & Huxham 2002). The structural complexity of a habitat depends on the substrate type and the types of sessile fauna that are present (Auster & Langton 1999). Sessile epifaunal communities, consisting, for example, of soft corals and sponges, on hard mineral substrates or biogenic reefs may create structurally complex environments which modify the environment, increase the area available for settlement and provide shelter for a variety of organisms such as fish recruits and small crustaceans (e.g. Connell & Jones 1991; Beck 1997; Kaiser, Rogers & Ellis 1999; Bradshaw, Collins & Brand 2003).

There have been several attempts to describe habitat complexity in general terms, instead of focusing on species identity and abundance (McCormick 1994 and references therein). These approaches have been adopted to reduce sampling and sample processing costs and to increase the frequency and scale of replication. Most methods for describing complexity have relied on direct measurement, for example, by using a profile gauge or comparing linear distances with distances across a habitat surface (Luckhurst & Luckhurst 1978; McCormick 1994). These measurements, when taken underwater, are usually made by divers and are, therefore, difficult to make over larger spatial scales and are also depth limited. Acoustic methods, such as multibeam sonar (Brown & Blondel 2009), can provide high-resolution descriptions of topographic complexity and substrate type at large spatial scales, but they do not consistently describe the contribution of fauna with soft tissues to habitat complexity.

Here, we explore the performance of two methods for assessing habitat complexity; both are easier to use, faster and cheaper than the manual analysis of biological samples or photographs. The first method is a modern analogue of profile gauge and chain methods and uses a laser line to provide rapid and repeated topographic measurements. The process of the laser line images was inspired by O'Neill, Summerbell & Breen (2009) who showed that a laser line could be used to detect the physical impact of an experimental fishing event. The second method uses seabed photographs to assess two-dimensional heterogeneity within habitats. An index of complexity is calculated from the layout of pixel values in each photograph (Proulx & Parrott 2008). This method has recently been used to describe the complexity of coral reef habitat at different spatial scales (Mellin et al. 2012). Both methods are cost effective as they do not require the collection and identification of fauna. We assessed the capacity of both methods to distinguish between habitats by comparing derived indices of habitat complexity with detailed information on substratum type and sessile epifaunal communities collected at the same spatial scale. To show the ecological relevance of the complexity indices, we tested whether they could be used to quantify ecological patterns such as changes in abundance and richness of associated mobile species.

Materials and methods

Survey Description

Seabed habitats around the Isle of Man, Irish Sea, were surveyed in August 2008. The survey encompassed 120 stations located on a regular grid with 5 km spacing inside the 12 nautical mile territorial limit and their depth ranging between 6 and 100 m. Of these 120 stations, 53 were chosen to be used in the present study, based on their habitat type and the quality of the data collected (as explained below) (Fig. 1). A sledge, fitted with a laser, a still camera (high-resolution Canon 400D digital camera) and a video camera (high-resolution Underwater Zoom Camera in a stainless steel housing from Rovtech Systems Ltd., Barrow-in-Furness, England), was towed on the seabed for 15–20 min at each station, covering an average distance of 300 m.

Figure 1.

Map of the study area – Isle of Man territorial waters (within 12 nautical miles), Irish Sea, UK. The triangles represent the stations for which complexity indices based on Hue, saturation, value (HSV) photographs have been estimated. The triangles in circles are the stations for which complexity indices based on both HSV photographs and laser lines have been estimated.

The laser was a Z-Laser model Z3A with a red wavelength of 635 nm and a line length of 1 m when mounted at 50 cm in the air. Z3A is a cable-less, battery powered, line laser. It was protected in a waterproof housing and fitted to the towed sledge so that it pointed perpendicularly towards the seabed from a height of 60 cm, at a distance of c. 50 cm in front of the video camera. The video camera pointed forward at an angle of 45° of the seabed and recorded throughout each tow (Fig. 2). The so-called laser line is more precisely described as a laser ‘plane’ that draws a line on the seabed which crosses the video screen from left to right (Fig. 3). Deviations of the seafloor from a flat surface could be observed on the video monitor as deviations of the line from a straight line. The position and deflections of the laser line were thus recorded on the video camera as the sled was towed. The best quality images of the red line were recorded if the lights on the sledge were off. Therefore, after recording videos for c. 15 min with lights on, where every 9 s, a 10 megapixel still photograph was taken with the still camera pointing perpendicularly towards the seabed (c. 100 pictures of 0·14 m2 per station), the sled was towed for 2–3 min with lights off to record the laser line with the video camera. During these 2–3 min, the still camera was still taking pictures every 9 s using its inbuilt flash lights, ensuring that the laser line videos we obtained sampled a similar seabed to what was previously filmed with lights on. Due to logistical constraints, the laser could not be used at all stations; therefore, only 20 of the 53 stations were sampled with the laser line system. The remaining 33 stations were only sampled with the still camera.

Figure 2.

Schematical explanation of the set-up of the laser line and the video camera to record and measure habitat complexity. L, laser; VC, video camera; LP, laser plane; IP, image plane.

Figure 3.

Extraction of a laser line and calibration. Left: laser line as recorded from the video, right: calibrated laser line used to calculate complexity indices.

Data Collection

To link descriptions of habitat complexity from the laser line and photographic methods to the ecology of the habitat, data on species identity and abundance and substratum type were required. The still photographs that were taken every 9 s throughout the whole tow were used to identify and quantify the benthic epifauna present at each station to the lowest possible taxonomic level. Species were identified and quantified in terms of abundance on 10–50 photographs per station; the number of photographs analysed per station was limited by the quality of the pictures and by time constraints. The photographs were taken in both light and dark periods because the camera used an inbuilt flash light when the sled lights were turned off for the laser work. The area sampled needed to be the same at each station to compare abundance and richness among stations. Therefore, 15 photographs were selected at random from all photos that were analysed for fauna for each station. The random sub-sampling of 15 analysed photographs was repeated 50 times with replacement to determine a more reliable average estimate of the abundance of mobile and sessile species and of the diversity (measured as richness) of mobile species. The species richness of sessile species was not estimated as most sessile animals need to be identified with a microscope. Fish and worms were excluded as our photographic and video methods are not suitable for estimating abundance of infaunal, tube-dwelling or highly mobile species. The photographs were also used to make a qualitative assessment of substratum type categorized as mud, sand, mixed sand (including shells), mixed gravel, mixed cobble or mixed rocks (when boulders were present). Substratum type was visually determined by one single observer at all stations to ensure consistency of the assessment. Mud and sand habitats were not considered further in the current study because their epifauna was very scarce and, therefore, not representatively sampled by the present survey design.

Calculation of Structural Complexity Indices

Two indices of structural habitat complexity were calculated. The first one, derived from the recorded laser line, was chosen to give information on the vertical structure of the surveyed seabed habitats, while the second one, based on the still photographs, was selected to give information on the two-dimensional heterogeneity of the same habitats.

Complexity indices derived from the laser lines

Still images that showed the laser lines were extracted from the videos at 20 stations with coarse substratum (i.e. excluding sand and mud). The number of images extracted, between 10 and 30, corresponding to c. 1 every 5 s during the 2–3 min when the lights of the sled were turned off, depended on the actual length of the tow and the quality of the video. On each extracted image, a red line was drawn over the laser line to remove the noise (such as air bubbles that would reflect the laser at random) and the coordinates of the red lines were extracted using imagej v.1.43 (National Institute of Mental Health) and R (R Development Core Team 2008, www.r-project.org). The laser lines coordinates were extracted as a series of points on the two-dimensional (x, y) plane. To extract the points, for each picture, each pixel value in red green blue (RGB) and its position were extracted in the form of a data set using imagej. The data set was then imported into R and only the pixels with the RGB values (255,0,0), that is, corresponding to the colour red, and their respective positions (x,y) were retained. This process separated the profile line that would be used to calculate topographic complexity.

Changes in the distance between the camera and the point at which the laser line illuminated the seabed (due to lifting of the sled from the seabed as a result of wave action) meant that the laser line was not in the same location on successive images. There was also some distortion of the image because the laser line was recorded underwater. To determine the true length of the laser line, extracted lines were calibrated from pixels to cm. The calibration technique was based on that developed in soil sciences by Darboux & Huang (2003) and applied to fisheries research by O'Neill, Summerbell & Breen (2009). A grid with 4 cm graduations was placed underwater in the laser plane in front of the sledge (Fig. 4). The real coordinates of each intersection point (in cm) were then related to the (x,y) coordinates of the pixels in the image of the grid using 4th degree polynomial equations (Figs 3 and 4):

display math(eqn 1)
display math(eqn 2)
Figure 4.

Position of the calibration points of the grid recorded by the video camera in the ‘x, y’ plan (left) and real position of the calibration points of the grid in the real system in cm (right).

(adapted from O'Neill, Summerbell & Breen 2009).

where Aij and Bij are the polynomial coefficients. More details of the calibration procedure can be found in Darboux & Huang (2003) and O'Neill, Summerbell & Breen (2009).

Along the real x-axis, the real value of y was extracted every 1 mm. Four indices of complexity were then derived from the adjusted laser line: the ratio of the length of the laser line to a straight line (i.e. the Chain-and-Tape index) (McCormick 1994), the vector standard deviation (McCormick 1994), the fractal dimension (e.g. Johnson et al. 2003) and the deviation of the laser line from a straight line, calculated as the sum of the standard deviations of the residuals of a linear regression fitted to the laser line xy coordinates. Fractals are dimensionless numbers that express the complexity of the static geometry of a structure (Kostylev et al. 2005). The fractal dimension can distinguish between a surface that has many small irregularities and one that has a few large ones (Commito & Rusignuolo 2000).

We assessed the relationship between changes in complexity indices derived from the laser line and changes in habitat complexity (substratum type and sessile epifaunal densities). Complexity indices were calculated from images captured during the last few minutes of the tow, whereas epifaunal densities were estimated for the whole length of the tow. However, the difference between the sessile epifaunal densities estimated from the last seven images that were analysed (and that corresponded with the location of the laser line videos) and estimates based on the repeated sub-sampling of 15 images over the whole tow did not differ significantly (F1,131 = 0·863, P = 0·355).

Complexity indices derived from the still photographs

Complexity measurements were made from still photographs taken at 53 stations, including the 20 stations where laser line measurements were available. Complexity indices were calculated based on 30 randomly selected photographs wherever possible. Individual photographs were replaced by the next available photograph when numerous or large mobile species, such as scallops, fish or crabs (e.g. Cancer pagurus), obscured the seabed and, therefore, prevented the calculation of the index of habitat structural complexity.

Prior to analysis, photographic quality was enhanced using imagej software. To remove dark edges, photographs were cropped by removing 300 pixels from all edges (pictures cropped from 3888 × 2592 to 3288 × 1992 pixels). Then, the brightness and contrast of each image was optimized based on an analysis of the image's histogram. The optimization was performed using imagej by allowing a small percentage of pixels in the image to become saturated (displayed as black or white). Complexity indices derived from optimized vs. non-optimized images were compared to verify that this did not bias the results. There were no significant differences between them [Pearson's correlation coefficients for H-values of mean information gain (MIG) complexity r = 0·999, P < 0·001; for S-values of MIG complexity r = 0·998, P < 0·001; for V-values of MIG complexity r = 0·998, P < 0·001; see below for explanation on Hue, saturation and value (HSV) and MIG values]. Cropped and optimized RGB images were converted into HSV images, also called the Hue, saturation and brightness (HSB) colour space. Hue is the pure colour, saturation, is the chroma and value is the intensity component (Gonzalez & Woods 2007).

Following Proulx & Parrott (2008), an index of complexity, called MIG, was then calculated (Fig. 5). Proulx & Parrott (2008) explain that MIG determines the amount of spatial heterogeneity that excludes the fraction devoted to a spatial heterogeneity. In other words, MIG measures the information gained by looking at the value of pixels in the neighbourhood of a particular pixel (eqns (eqn 3), (eqn 4), (eqn 5)).

display math(eqn 3)
display math(eqn 4)
display math(eqn 5)
Figure 5.

Example of one photograph and its HSV transformation for the calculation of the mean information gain (MIG), the mean mutual information (MMI) and the Gamma complexity indices. (a) Original RGB photograph, (b) Hue colour band – MIG = 0·29, MMI = 0·66, Gamma= 0·19, (c), saturation colour band – MIG = 0·25, MMI = 0·48, Gamma = 0·12, (d) value colour band – MIG = 0·17, MMI = 0·80, Gamma = 0·14.

where pi) is the probability of observing a pixel value independently of its location on the picture, and N is the number of categories of pixel values; pi) is the probability of finding a specific 2 × 2 combination of pixel value categories in the picture, and N4 is the number of theoretical 2 × 2 combinations. In this study, a value of N = 10 was used (i.e. the 255 possible pixel values were binned into 10 categories before analysis).

If there is no information gained, that is, the distribution of pixels is uniform on the photograph, MIG equals 0, which represents order. However, if the distribution of pixels is random, MIG equals 1, representing disorder. Photographs presenting patterns, or clusters, will have intermediate values. Complex structural patterns are, therefore, expected to have values between 0 and 1. The value of MIG was calculated for the 30 pictures selected in each station in each colour band, H, S and V, and averaged to give one complexity index per colour band per station.

For MIG to be a meaningful index of structural complexity in the present study, it was necessary to find out the critical value, between 0 and 1, at which complexity peaks. Another index of complexity, Gamma, has been developed by Proulx & Parrott (2008) to assess the value at which complexity peaks. Gamma is calculated as the product of MIG and another index called Mean Mutual Information (MMI). MMI varies between 0 and 1, with 0 representing random patterns and one uniform patterns, the opposite of MIG (eqn (eqn 1)).

display math(eqn 6)

As a result, Gamma peaks at maximum structural complexity and can be used to determine the MIG value which identifies the critical point between order and disorder in the data set, that is, the value of maximum structural complexity (see Proulx & Parrott 2008 for more details on the method) (Figs 5 and 6). Matlab code to calculate MIG, MMI and Gamma is available from the authors on request (lael.parrott@umontreal.ca).

Figure 6.

Relationships between the MIG and Gamma complexity indices for the three bands of the HSV-transformed images. Curves were smoothed using a locally weighted regression over a total of 83 stations where complexity indices were averaged from 30 pictures per station.

The relationships between the MIG and Gamma indices for each band of the HSV images are shown in Fig. 6. Only Gamma of the colour band V showed a peak, when MIG was c. 0·35. Gamma of H- and S- colour bands did not reach a maximum value at intermediate MIG levels but appeared to increase at a lower rate towards the end of the MIG-ranges they, respectively, covered, < 0·35 for H and > 0·35 for S. Therefore, 0·35 was defined as the critical value at which image complexity attained a maximum. Consequently, the absolute difference between 0·35 and MIG could be used as the index of structural complexity, here called derived-MIG index (dMIG) for dMIG. The closer MIG is to 0·35, the lower the value of dMIG and the higher the complexity of the image. As with the approach adopted for the laser line indices, we investigated whether dMIG was detecting changes in habitat complexity as indicated by changes in substratum type and sessile epifaunal densities.


Variation of Habitat Complexity Indices with Substratum Type and Density of Sessile Epifauna

The variations in dMIG of the three colour bands, H, S and V, were not related to the density of sessile epifauna (Table 1). Of the dMIG indices, only dMIG of the V- band (dMIG-V) showed a significant difference among substratum types (Fig. 7, Table 1). dMIG of the H and S bands showed some variability within substratum types but no significant difference among mixed sand, gravel, cobble and rock habitat types. dMIG-V increased with grain size, with values significantly lower in sand than in gravel, cobble or rock and lower in gravel than in cobble (Table 2). There was no significant difference in dMIG-V between rock and cobble or rock and gravel habitats.

Table 1. Results of the analysis of variance of the multiple linear regressions between habitat complexity indices and potential explanatory variables (substratum type, density of sessile species and their interaction)
ResponseVariabled.f.SSMSF- valueP-value
  1. d.f., degrees of freedom; SS, sum of squares; MS, mean squares.

  2. dMIG-H, dMIG-S and dMIG-V are the respective derived-MIG indices for the H, S and V colour bands of the images. CT is the Chain-and-Tape index, VD is the vector standard deviation index, FD is the fractal dimension, and LR is the standard deviation of residuals from linear regression calculated from the laser line method.

Density sessile sp.12·22E-042·22E-040·2400·627
Density sessile sp.12·13E-042·13E-040·2260·637
dMIG-VSubstratum31·91E-026·35E-039·769< 0·001
Density sessile sp.14·37E-044·37E-040·6720·417
CTSubstratum23·07E-011·54E-0129·302< 0·001
Density sessile sp.11·65E-031·65E-030·3180·582
VDSubstratum21·33E-016·67E-0292·463< 0·001
Density sessile sp.12·17E-032·17E-033·0080·105
Density sessile sp.17·84E-047·84E-040·0920·766
LRSubstratum34·01E-012·01E-0123·690< 0·001
Density sessile sp.16·21E-026·21E-027·3380·017
Table 2. Difference in habitat complexity indices, dMIG-V and LR, between substratum types, tested by analyses of variance and Tukey tests
Response anova Tukey test
  1. d.f., degrees of freedom.

  2. dMIGV is the derived-MIG index for the V colour band of the images. CT is the Chain-and-Tape index, VD is the vector standard deviation index, and LR is the standard deviation of residuals from linear regression calculated from the laser line method.

dMIG-V311·478< 0·001Mixed gravel – Mixed sand0·0230·026
Mixed cobble – Mixed sand0·047< 0·001
Mixed rock – Mixed sand0·047< 0·001
Mixed cobble – Mixed gravel0·0240·077
Mixed rock – Mixed gravel0·0240·124
Mixed rock – Mixed cobble0·0011
CT232·408< 0·001Mixed gravel – Mixed sand0·0350·607
Mixed cobble – Mixed sand0·298< 0·001
Mixed cobble – Mixed gravel0·263< 0·001
VD276·067< 0·001Mixed gravel – Mixed sand0·0620·003
Mixed cobble – Mixed sand0·203< 0·001
Mixed cobble – Mixed gravel0·141< 0·001
LR218·605< 0·001Mixed gravel – Mixed sand0·0120·974
Mixed cobble – Mixed sand0·332< 0·001
Mixed cobble – Mixed gravel0·320< 0·001
Figure 7.

Complexity indices showing differences in substratum type among stations. The substrata are ordered from small to large grain sizes. (a) dMIG-V is the derived-MIG indices for the V colour band of the images, (b) CT is the Chain-and-Tape index, (c) VD is the vector standard deviation index, and (d) LR is the standard deviation of residuals from linear regression (see Tables 1 and 2 for statistics).

Cobble and rock substratum categories were collapsed into one group for the analyses of the laser line indices because there were only two stations with rock habitat type. Except for the fractal dimension index, all laser line indices detected differences in substratum type (Tables 1 and 2). The vector standard deviation index (VD) was the only one significantly detecting differences between all categories. The index based on the standard deviation of residuals from linear regression (LR) and the Chain-and-Tape index (CT) were not detecting changes in complexity between sand and gravel. However, only LR was showing a strong correlation with both substratum type and density of sessile epifauna (Fig. 8, Table 1).

Figure 8.

Relationship between density of sessile species and LR, complexity index derived from the laser line method. (slope = 4·801E-03, se = 1·46E-03, t-value = 2·606, P = 0·018, r2 = 0·28).

Figure 9.

Relationships between habitat complexity index LR and density (ind m−2) and species richness of mobile species. The solid lines represent the significant models (statistics given in the text).

The LR laser line index showed a monotonic increase with coarseness of the substratum and with abundance of sessile epifauna independently of substratum types. As the laser line index returned three-dimensional information, or complexity over heterogeneity, it was more sensitive to the presence of upright sessile epifauna than the MIG index. It could not differentiate between sand and gravel, as both laser signatures did not deviate much from a flat line, but it did differentiate between sand/gravel and cobble/rock better than the MIG index. Overall, the MIG index was not as sensitive as the laser line indices to sessile epifaunal changes. Therefore, the LR laser line index was preferred to quantify habitat complexity at the scale of the present study.

Example of Application: Relation Between Habitat Complexity Indices and Associated Mobile Species

The line laser index (LR) was selected to test the relationship between habitat complexity indices and species identity and abundance as it was the only index that detected changes in both substratum type and sessile epifauna. Both density and richness of associated mobile species increased along LR (density coefficient estimate = 39·830, se = 14·206, t(19) = 2·804, P = 0·012; richness coefficient estimate = 8·775, se = 3·612, t(19) = 2·430, P = 0·026) (Fig. 9). Explained variance in density and richness of mobile species was, respectively, 30% and 25%.


Our results show how laser lines and the processing of digital images can be used to measure the structural complexity and heterogeneity of a range of seabed habitats. For a given amount of time or financial resources, both methods allow greater levels of replication in space and time than the direct sampling and identification of fauna and substratum types. Both methods can be used to quantify some differences in substratum type, while one of the line laser indices can be used to further quantify the differences in the densities of sessile epifauna.

Clearly, the practical value of methods for describing and monitoring marine habitat complexity depends on whether they meet the operational needs of scientists studying habitat distributions and the effects of pressures as well as the needs of those authorities responsible for habitat mapping, monitoring and assessment. No generic method is likely to meet all needs, but the methods we propose provide information on complexity that can be used to assess state and/or to describe changes in state. Likely, applications would be habitat mapping, monitoring responses of habitat to management measures and the large-scale assessment of human and environmental impacts. Given significant limitations on resources for monitoring and assessment of marine habitat, it is unlikely that detailed species-based analysis would be feasible and/or provide adequate statistical power to detect trends on equivalent space or time-scales. However, confidence in the relationships between habitat complexity and the biodiversity or function of habitat (in particular, those aspects of policy relevance) will likely determine the extent to which information on complexity, as derived with our methods, is deemed sufficient to guide assessment and management.

The dMIG, which was used to quantify heterogeneity on a two-dimensional scale, could differentiate between sand, gravel and larger stones. dMIG, unlike MIG, no longer differentiates between random and ordered images. However, dMIG was appropriate when we were focusing on complexity rather than order and randomness. Surprisingly, MIG appeared to decrease with substrate complexity when an increase was expected. This meant that dMIG increased from sand to gravel and from gravel to cobble and rock. One explanation might lie in the scale and resolution of the photographs. Because we were working at very small scales (0·14 m2 photographs), the presence of larger stones seems to have increased the uniformity (or order) in the picture rather than decreasing it towards more clustered patterns. The MIG index did not capture variations in sessile epifaunal coverage, probably due to similar scale effects.

The LR index provided a quantitative index of complexity that reflected both substratum type and sessile epifaunal abundance; the two components of structural complexity (Auster & Langton 1999). Such a measure of structural complexity has potential value when quantifying relationships between habitats, diversity and ecological processes. Indeed, in this study, the LR index was related to mobile benthic species abundance and richness. The response of abundance and richness to changes in habitat complexity or increases in surface area has generally been difficult to disentangle, because surface area usually increases with surface complexity. One way of dealing with this issue is to study the fractal dimension of the habitats (Johnson et al. 2003). The laser line data allow the calculation of an index of fractal dimensions, as when manual profile gauges or cast cross-sections have been used on rocky shores or mussel beds for instance (e.g. McCormick 1994; Beck 2000; Commito & Rusignuolo 2000; Johnson et al. 2003; Kostylev et al. 2005). However, as no relationship between substratum type or sessile epifauna and fractal dimension was observed here, we could not determine whether abundance and richness responded to complexity or increases in surface area.

The scale of complexity measurements may affect the strength of relationships between complexity, faunal composition and function. For example, Mellin et al. (2012) used the MIG method to look at how hierarchical habitat complexity in coral reefs reflected variations in abundance, richness and community structure in fish populations. They showed that 25% to 33% of the deviance in abundance, richness and community structure could be explained by the MIG index. Similarly, we found that 30% and 25% of the variance in abundance and richness of benthic mobile organisms could be explained by our index of structural complexity, the LR laser line index. Mellin et al. (2012)'s study suggests that a spatially structured design including replicates of laser line measurements and MIG indices from larger scale images could be used to improve the predictive power of our index of complexity in temperate sand, gravel, cobble and rock habitats.

The distribution and status of marine habitats is affected by human and environmental pressures. Bottom trawling is one of the most widespread human pressures (e.g. Eastwood et al. 2007) and can modify habitat structure and associated biodiversity (e.g. Auster & Langton 1999; Thrush & Dayton 2002). Changes in habitat owing to trawling pressure are usually quantified using grabs, dredges and trawls to provide species identity, abundance and/or body size data (e.g. Jennings et al. 2001; Blanchard et al. 2004) or by photographic methods that provide metrics of species' abundance (e.g. Collie, Escanero & Valentine 2000; Lambert et al. 2011). Both approaches are relatively costly and labour intensive, require specialist taxonomic skills and are, therefore, challenging to use for frequent monitoring over large spatial scales. The methods developed in the present study would support fishing impact assessments and monitoring on larger space and time-scales.

As our approach was only semi-automated, it required some operator time to enhance the images. In some images, we also observed a shadow effect where part of the view of the laser line was blocked by the surface. This could largely be solved by using two lasers illuminating the surface from opposite incident angles (see Darboux & Huang 2003 for details). The image processing speed (30–40 min per station of 10–30 images) could readily be increased by developing software to extract the wavelength required at high frequencies so that the number of replicates would be high enough to appropriately represent complex surfaces (Frost et al. 2005). However, the laser line method is limited, just like the profile gauge method, by the incapacity to model features such as overhangs, which may be important as species refuges (Commito & Rusignuolo 2000).

Of the methods considered, the laser line gives more consistent information on habitat complexity. It has potential to support demands for frequent monitoring of habitats and fishing effects on large spatial scales (e.g. monitoring the performance of marine protected areas or the effects of changes in fishery management regulations). It is less costly and labour intensive than other approaches and can be deployed from vessels of many sizes. In comparison, the consultancy price for processing sediment samples, that is, sorting and identification of benthic invertebrates in grab samples, is c. £300 per sample in the UK, noting that several replicates would be required per station to measure habitat complexity based on the biomass of sessile epifauna. The laser model used in this particular study costs c. £150. The housing was not included and was manufactured for the present study but complete models, including housing, are generally more expensive. The sled was about £1500 and the video camera around £8000, including umbilical and lights. The equipment could also be deployed by divers on habitats where the impacts of towing a sled over the seabed were not acceptable, such as coral reefs. Since the 2008 survey, further development and improvement have been made to the method. For instance, we deemed our laser line method suitable although we expected the power of our analyses to be decreased by the lack of direct correspondence between laser line measurements and quantified sessile epifaunal densities (see methods). However, ideally there should be direct correspondence between the two. Therefore, we have now fitted the line laser and a smaller, cheaper, video camera with a wider angle and higher resolution to a beam trawl to relate habitat characteristics directly with catches and by-catches. Work is in progress to redesign the sled with this same camera and assess the impact of fishing on habitat complexity.

Compared with other methods for assessing complexity discussed by Frost et al. (2005): chains, profile gauges and stereo photography, the laser line is likely the most practical, precise and cost-effective instrument for measuring subtidal habitat complexity at large scales. Two of its major advantages are that each measurement can be replicated at high frequencies and with high precision on a particular habitat. Another potential benefit of the laser line method is that the video camera provides a continuous record of the laser line along the length of the transect. Thus, when post-processing, different numbers of ‘samples’ could be taken at random or even fixed intervals along the transect, depending on the objectives of a study. Further, the existence of a continuous record of changes in the laser line might be used to support image processing at very small distance intervals: to describe transitions in habitat complexity and impacts at different scales. While the capacity to describe habitat complexity on habitat at large spatial scales and at high resolution would add to existing monitoring and assessment, it is unlikely that it would entirely replace taxonomic assessment. For studies of human impacts and for habitat monitoring and assessment, we would envisage that the scale and frequency sampling might be increased using a laser line system, potentially combined with the calculation of MIG indices at different scales, but that interpretation would be supported by a lower frequency assessment based on conventional biological sampling or photography.


We thank the UK Department of Environment, Food and Rural Affairs (Defra Project M1001) for funding the research, and the Isle of Man Department of Environment, Food and Agriculture for permission to use these data and photographs from their habitat mapping survey conducted in August 2008. Collection and quantification of biological data were carried out with the assistance of numerous Bangor University students, the efforts of all of whom are gratefully acknowledged. Finally, we thank Moritz Staebler for his contribution to the definition and coding of laser line complexity indices.