The CCAMLR krill synoptic survey

Sampling strategy

The synoptic survey will be a stratified, randomized parallel transect survey design. The advantage of using such a design is that it will be possible to use both classical design-based statistical analyses (Jolly and Hampton 1990) and model-based geostatistical analyses (Petitgas 1993, Murray 1996). The basic requirement for a randomized survey is that any possible parallel transect line in the survey area shall have an equal probability of being  sampled. However, one problem with such a design is that if transect positions are chosen completely at random then it is possible for two adjacent transects to be very close to each other. To overcome this we suggest a two-stage randomization. Such a procedure has been used in previous surveys (see Brierley et al 1997) and the basic principles are shown in the diagram. The survey grid (black box) is split into a series of vertical zones; two zones for each transect in the grid. A transect is randomly placed within alternate zones (the vertical red arrows within the red-shaded zones). The white zones contain no transects and so act to keep transects a minimum distance apart. To comply with the requirement that any transect has an equal probability of being chosen, the location of the entire survey grid can be moved horizontally by the width of the white zone. The actual distance is chosen at random. The effect of moving the survey grid can be seen in the blue grid, where the blue transects are placed at random within the grey shaded zones.

Page last updated on 18 February 1999