November 30, 2010

Technical Question of the Month: Why use a 5 m gap in the tropical transect surveys?

\"\"

For convenience, we recommend using a single 100 m transect line to carry out Reef Check (RC) tropical surveys. It is simpler to roll out one 100 m line than to use multiple 10m or 20 m lines. In fact, the actual design of RC surveys is based on four 20 m long “segments” of the full transect line. So we sample 20 m, and then leave a 5 m gap, followed by a second 20 m that is also sampled and so on until we have completed all four “segments” for a total of 80 m along the 100 m line. We leave a 5 m gap in between each 20 m segment because we would like each segment to be independent or unaffected by the others. How does this help the data analysis?

A common question one might ask about a data set from a Marine Protected Area (MPA) is whether there are more grouper inside the MPA than outside. To test this hypothesis, we need to use statistics to compare the mean number of grouper inside and outside. When we carry out statistical analyses using what are known as parametric statistics (e.g. a t-test) there are certain requirements that must be met. One is that the samples (our segments) are independent of each other and that each sample is random.

For example, if we did not use a 5 m gap, then one large coral might grow across the boundary between the first and second 20 m segments. In that case, the data would not be independent because the same coral was found in both segments – the coral in the first segment “affects” the presence of the same coral in the second segment. So we use the 5 m gap to increase the independence of our samples. It is not a perfect solution but it helps prevent what is known as “pseudoreplication.”

While most scientific monitoring methods used in ecology “violate” some of the prerequisites of parametric statistics we can still analyze tropical data using parametric tests. In addition, there are many non-parametric tests available to analyze RC data. Why do ecologists not use a strictly random sampling design? A true random sampling design is very costly and may have its own biases when for example, by chance, many randomly chosen sample points cluster in one area. Thus we have chosen the most efficient method for Reef Check that also allows us to use parametric statistics. Of course, those who prefer to use shorter reels – 10 m or 20 m long transects may lay these on separate sections of reef. In some areas with patch reefs or spur and groove formations – this will be the logical choice.