Qualitative measures

Detailed surveys of abundance and distribution of biota

Design

Throughout the history of rocky shore studies, successively more detailed methods have been devised for describing the distribution and abundance of organisms. Pioneers like Stephenson and Stephenson (1949) and Lewis (1964) gave broad scale qualitative descriptions. Later, semi-quantitative techniques, (Crisp and Southward 1958) allowed rapid estimates to be made of the abundances of organisms in particular localities. More recently, quantitative techniques involving prescribed sampling regimes and analytical methods have been used to provide very fine-scale descriptions of distribution and abundance. In addition to these techniques, experiments can be conducted to test specific hypotheses.

Each of the above methods has a role to play in surveillance and monitoring programmes for rocky shores. More quantitative techniques allow changes to be assessed and understood in more detail. However, qualitative techniques provide relatively quick and simple methods for broadscale surveillance which will meet many of the requirements of conservation workers. The level of detail required from any programme will depend of the questions being asked. Therefore, the aims of the study must be made absolutely clear before any descriptive work is undertaken. In parallel experimentation allows hypotheses on causes of change to be tested (Underwood, in press).

Qualitative measures.

A general, qualitative, description of the shore will be useful in all studies. This should include information on physical and biological factors. A good description of the physical features of the shore should include say whether it is exposed or sheltered, whether it is bed-rock or predominantly boulders, the size of the boulders, whether it is evenly sloping or highly broken and the rock type. A general description of the major communities should say, for example, how many zones are present what shore levels they occur on and what are the dominant space-occupying organisms in these zones. The MNCR biotopes classification (Connor et al., 1997) provides a standardised format for describing the communities and habitat features present (see Chapter I). This will allow for a degree of objective comparison of reports produced by different observers. However, fieldworkers should be aware that some of the community-habitat combinations they encounter may not be covered by the biotope classification. Users of this system should be careful to avoid misclassifying biotopes.

Photographs or video will provide an instant record of the gross features of the shore community. Photographs taken on different site visits can be compared with assess any large scale changes in the shore community. Fixed viewpoint photography, where an area is photographed from the same position at each site visit, allows unambiguous comparisons. Aerial photography allows unobstructed views of whole shores and for very rapid coverage of many shores in a single flight. Infra-red photographs can be used to distinguish chlorophyll bearing algae from a background of similarly coloured rock. The advantages of these methods are that they allow broad coverage and that data is easily obtained and stored as photographs. Gross changes can be easily detected. The disadvantages are that subtle changes will not be recognised and that comparison with other areas is impossible. Understorey and encrusting species can be obscured by taller organisms and only species which occupy large areas will be detected. Aerial photography has the obvious disadvantage of cost. Electronic cameras are rapidly becoming cheaper and may offer significant advantages over conventional photography with rapid digital archiving capabilities and potential for image analysis.

Correct interpretation of photographic records will rely on some degree of ground truthing. We strongly recommend that the time available during a site visit is used to collect numerical data on the community, which will provide more detail and facilitate statistical analysis.

Conducting a detailed survey therefore represents an efficient use of the time available during a site visit. The level of detail will depend on the information required. Numerous surveying methods have been developed, a few of which are described below.

All numerical methods aim to provide a description of the abundance of certain species at different positions on the shore. Both physical and biological information should be collected. In many cases, as with fixed transects and quadrats, most of the physical information needs to be collected only once.

Some species which are present on rocky shores will be too small, too rare or too well hidden to be detected whichever sampling method is used. Biological sampling therefore provides an index of the condition of the community but not a thorough description of its composition. Monitoring aims can often be achieved by restricting surveillance to a number of important species (Lewis, 1976). These species should be easy to identify with the minimum of disturbance and should play a key role in the community. That is, the abundance of these species should affect and be affected by the abundance of other key species. Recommended species are those which dominate available space such as macroalgae, barnacles and mussels, and grazers such as limpets and important predators. Dogwhelks are an important indicator species for the effects of TBT (see Chapter V) while various species of barnacles, limpets, winkles and topshells could be used for monitoring climate change. It should be possible to assess the abundance of 15 to 30 species during a single shore visit. We therefore recommend that the number of species used should be in this range with a high number being used where more detail is required. Sampling should be non destructive wherever possible. Destructive sampling will, however, be unavoidable when specimens are required for analysis elsewhere and when sampling the fauna associated with algae or sheets of sessile animals. Fieldworkers should select target species and sampling methods to minimise potential impacts of these methods on the shore community.

The key species approach is less time consuming than producing detailed inventories of all species present. It will not, however, provide information on species richness. When using the key species approach, a smaller number of species means that more sites can be sampled per unit time and vice versa. It is important to reach a balance between species and spatial information which allows the objectives of the monitoring or surveillance programme to be met.

There are several methods for assessing and recording the abundance of chosen species within sampling stations. The semi-quantitative approach of Crisp and Southward (1958) has proved valuable, allowing a quick assessment of the abundance of organisms at a particular locality. Abundance is described by one of a limited number of terms, usually between five and eight. Such scales are usually based on a logarithmic or semi-logarithmic progression. In order to minimise between-observer variability, it is best to do a few counts in quadrats of various sizes before beginning a survey. This will give a good idea of the appearance of the various abundance levels.

One way to estimate an index of abundance for rare or inconspicuous species is to conduct a timed search within predefined sampling stations of fixed area. This is particularly appropriate for highly heterogeneous shores, especially those made up mainly of boulders. It is also appropriate for assessing microhabitats such as rock pools, crevices and the underside of stones within a predominantly bedrock shore. Personnel with experience in identifying the species of interest should be able to search an area of about 20m² within 15 minutes. This method is subject to between observer variation and at best it provides an index of abundance rather than an absolute measure.

The actual numbers of each species, or the area it covers can be recorded within quadrats. Cover can be estimated using the subdivisions of a quadrat as a guide. A better and more objective way of estimating cover is to use the percentage of "hits" underneath the cross-wires of a quadrat, or the dots or holes on a transparent overlay sheet. A double layer of sighting points avoids parallax errors. Alternatively, a pin-frame can be used, which is more precise but cumbersome. Good estimates can be made with 25 or more sighting points: the larger the number, the more accurate the estimate (30-50 seem appropriate for most cases). These sighting points can be arranged regularly or randomly. For most purposes regular arrays are probably best. Random or regular dot overlays can be used to measure percentage cover on photographs. In recent years, PC-based digitizers or image analysis systems have become much cheaper and are excellent for estimating cover from photographs (see Foster et al., 1991, Meese and Tomich, 1992). Data can be entered directly into statistical packages for analysis. Limited time available for field work will often preclude the use of detailed methods. A visual estimate of percentage cover is a very quick method. Tests have shown that there is strong correlation between visual estimates and actual cover. With a little practice, most observers should be able to estimate cover within 20% of the actual value. A recent detailed assessment of various approaches is given in Meese and Tomich (1992). The various kinds of cover: canopy cover of large seaweeds, understorey cover by turfs of algae and sessile animals, and encrustations on the rock surface itself should be considered separately. An estimate of cover can be converted into an estimate of abundance by multiplying the density of the species by the area it occupies.

The main disadvantage of the abundance scale approach is the limited scope for analysis. Usually, this method is used to estimate abundance for a whole site, in which case, no estimate of within-site variance is possible. However, it is possible to use this method for estimating abundances within quadrats at a site, although the main problem then becomes the limited detail. Analysis with non-parametric techniques is necessary despite their limited power. Surveillance and monitoring methods using abundance scales are less likely to detect subtle impacts than the more quantitative methods using counts of individuals or area covered. Quantitative methods are more amenable to statistical analysis. Standard parametric statistics requires that the variances are the same in the areas under comparison. Usually this is not the case and some form of transformation has to be applied to the data (Underwood, 1981; 1996). The power of any analysis is limited by the accuracy of the data. The collection of accurate data is time consuming. There is therefore a trade off to be made between the time available for sampling and the level of detail required.

A description of the zonation of organisms on the shore is a useful first step towards describing the community. The shore should, at the very least, be divided into areas representing different tidal heights.

Many sampling designs are based on abundance estimates from quadrats. The appropriate size of quadrat to be used depends on the size of the organisms of interest. A 1m x 1m quadrat is usually best for large seaweeds such as Ascophyllum or kelps. Smaller quadrats may be more appropriate for smaller organisms. For example, 50 x 50 cm is suitable for British limpet species and a 5 x 5 cm quadrat is recommended for sampling barnacles. As a general rule, the appropriate size of quadrat should enclose no more than about 100 individuals of each species at the densities normally encountered. Greater numbers will take too long to count. With too small a quadrat size there will be problems with zero counts. Sampling at any shore level will therefore normally involve the use of several different sized quadrats or subsampling within subdivisions of a large quadrat.

Shores consisting primarily of bedrock are amenable to quadrat sampling while boulder shores generally are not. Within-quadrat estimates of abundance should be made using either absolute numbers of individuals or percentage cover rather than using semi-quantitative abundance scales. Timed searches will generally rely on abundance scales although replicate 5 or 10 minute searches can be used. Quadrats, being generally smaller than the areas in which timed searches are conducted, allow easier use of designs involving repeat random sampling. Subject to these constraints, the design alternatives described below can apply equally to the areas within which timed searches are conducted as to quadrats.

The design of a sampling scheme will depend on the information required from it. The design will be limited by the resources available. Where a single worker is required to survey one or more shores during a single low tide, the design options are limited by what can reasonably be accomplished in this time. It is also good practice to consider the analytical techniques to be used before deciding on the final design. Some idea of community structure and dynamics can be gained without formal analysis. However statistical methods, when properly used, have several advantages over the former approach. Not least among these is the possibility of extracting relevant information from large data sets. An enormous amount of data might be generated through the long term monitoring of up to 56 SACs. Thus, monitoring strategies which allow these data to be collected in a way which is amenable to statistical analysis will maximise the value of the resulting information. Statistical analysis also allows the identification of sometimes subtle changes and the scales at which they are occurring.

The use of fixed-location quadrats (Lewis, 1976) was widely adopted during the 1970s. Using this method, the abundance and distribution of organisms in a limited number of fixed sites can be regularly recorded. This approach has been used to study changes within such fixed stations (Hartnoll and Hawkins, 1985). However, the major disadvantage of fixed quadrats is that they are rarely replicated, and even when they are replicated, sequential measurements cannot be treated as independent of each other in statistical tests. It is therefore not possible to draw any conclusion about spatial scales larger than the sampling station. However, when species are rare, they may well escape detection using random sampling methods. It is therefore recommended that when rare species of conservation interest are identified, fixed stations should be established to monitor their condition. The abundance of these species on the rest of the shore can be estimated from random sampling.

Localised maps of species distribution and abundance can be produced by sampling at regular intervals within a grid (Johnson et al., 1997). However, this technique is not recommended for making comparisons with other areas as the quadrats are unlikely to be separated by a large enough gap to make them independent of each other. A comparable technique which is useful for mapping the absolute vertical distribution of organisms is to use a thin transect with contiguous quadrats. The limitations are similar to those for regular sampling because of a lack of independence. The problems which arise from this are discussed by Legendre (1993) and Schneider and Gorewich (1994). The choice of quadrat size is an important consideration when using thin transects to describe zonation. Extremely narrow transects might give a false idea of the abundance and vertical range of species. Very wide transects might include variation caused by the horizontal wave action gradient. Replicated thin transects may be the best option if the main interest is in zonation but are unlikely to be feasible in the time available for a shore survey.

The design that best facilitates statistical analysis is stratified random sampling. The strata are areas of interest which might be compared in the analysis. Within each of these strata, subsections are sampled at random. The distance between each successive quadrat is determined using tables of random numbers. It is likely that conservation workers and researchers will wish to compare communities at various levels on the shore, in which case the strata are these levels. Comparisons may be made at a variety of spatial scales using nested sampling. Nested sampling schemes have proved extremely useful in detecting the scales of spatial variation in abundance of rocky shore organisms (Caffey, 1985, Underwood and Chapman, 1996). A nested sampling design consists of sampling units repeated at a number of predefined spatial scales. An example of such a scheme appropriate to surveying SACs would consist of a sampling unit of a group of quadrats at a defined shore level (say mid tide level) repeated at three to four 100m intervals within a single location (a ‘shore’) in an SAC. Several shores could be surveyed in this way in a larger SAC, while a repeat of the design at different SACs would permit a rigorous analysis of the variation in dominant species among SACs. Recent studies adopting this procedure have shown that variation is most pronounced at scales of up to 2m (microhabitat patchiness) and at 10km and beyond (between shore variation: Underwood and Chapman, 1986).

It is likely that workers will also wish to monitor changes which occur. Two major design aspects of the sampling scheme are the number of shore levels to be considered and the number of samples to be taken in each.

The number of shore levels is determined by the resolution of the study and the time available for surveillance. Biogeographic mapping projects might concentrate on three major zones; high, mid and low shore or at the zone of maximal abundance of the species under study. Greater resolution might be required when assessing environmental impacts or producing detailed local information. There is a need to balance the number of species used with the number of sites surveyed. More sites allow greater spatial resolution while more species provide greater detail on community structure. However, a finite amount of data can be collected and analysed in the time available. It should normally be possible to sample at 5 or 6 shore levels during a single site visit. In monitoring programmes, sampling stations should initially be located by levelling along vertical transects. The position of these transects can be relocated either using detailed photography or markers fixed into the rock (See Appendix 7 in Hiscock, 1998)

The purpose of repeated sampling in each station is to gain the best estimate of the mean and variance of the abundance for each species or of diversity. The best number of samples to use should balance this need with the need to avoid excessive sampling effort. A pilot study at the outset of a sampling programme can be used to determine the correct number of samples. This pilot study should consist of a survey using a large number of quadrats; say, 20 per shore level. These should be placed in random order and the cumulative mean and its confidence limits plotted against the number of samples which gave that estimate. Eventually the fluctuations on the graph will damp down, showing the number of quadrats required to reach a reasonable estimate. It is unlikely that conservation workers will have time to do this for each site, especially when several quadrat sizes are being used. Where possible, a pilot study should be conducted at a representative site. When this is not possible, a rule of thumb is to use as many quadrats per station as possible within the time available. A minimum of three quadrats per station will facilitate analysis. More will probably provide better estimates.

Summary of recommended approaches to surveying rocky shores

Next Section                     References