Is the sampling strategy interfering with the study of spatial variability of zooplankton communities?

Abstract

Surveys at the whole-lake scale take some time to carry out: several hours or several days. For logistic reasons, the sites are not sampled simultaneously or in a random sequence. Traditional limnological sampling methods require an appreciable amount of time at each site. Any sampling strategy that is not random or simultaneous introduces dependencies among the observations, which must be taken into account during the analysis and interpretation of the data. What is the real nature of the variation measured using a given sampling design? This question is approached using sites sampled by two boat teams during two consecutive days. Statistical modelling was used to partition the variation of zooplankton size-class data into environmental and spatial components. The conclusions reached after an analysis that did not control for the sampling design are erroneous and quite different from those reached when the effect of the sampling design (factors Day, Boat, and Hour) was taken into account. Clearly, when a significant effect of the sampling design is found, one must control for it during the analysis and interpretation of ecological variation.