Résumé
A multidisciplinary ecological study is in progress in the Thau marine lagoon, on the Mediterranean coast of France. Sampling is being conducted in two phases. Phase 1 is a pre-sampling program (pilot study), space- and time-intensive, bearing on 10 variables only; it was conducted in 1986 and 1987. During phase 2, that began in 1988, more variables will be studied at fewer stations, and at the most appropriate time scales; the purpose is to increase our understanding of ecological processes through modelling. This paper examines the results of the pre-sampling program and attempts to determine how to distribute samples through space, and through time, in order to best sample the variability of the system. Through space, four methods are proposed to select 20 stations among 63. It is shown that none of the methods always performs better than all others, their power of reproducing the best part of the original variable's variability depending upon the shape of the spatial structure (gradient, patches, hole, etc.). It is also shown that all four methods are far more efficient at rendering the system's variability than either random or systematic sampling designs. Along the time axis, the hourly, daily and monthly sampling scales were compared as to their coefficients of variation for each variable, and the daily and monthly scales were selected as being, overall, the most informative for the processes under study.