Computing Optimal Stratified Random Sampling in Marine Environment

INTRODUCTION The study of the distribution of any variable in marine environment generally begins by a sampling experiment. Usually, the collection of samples follows a procedure, often fixed in advance, that takes into account the general distribution of the water masses for a given region. However, much attention may be devoted to the actual structure of the water column present below the ship at a given experimental station, and about previous informations on its spatial variability. The most significant information available in marine environment is related to the stratified structure of the vertical water column. This structure can be identified by measuring some environmental variables such as temperature or conductivity of the water. These variables, well known for its conservative properties, can be measured quickly with a high precision and recorded on magnetic tape or disk. The present study suggests a mini-computer program that permits the construction of strata on the basis of the environmental variables just after their recording. During an oceanographic survey, many other variables of chemical of biological nature must be evaluated for a given area. To do this, water samples are usually collected at fixed depths by means of sampling bottles. In contrast to this procedure, the second objective of the present program is to define an optimal allocation of a sample into the strata identified previously on the basis of the environmental variables. A more relevant allocation of a sample in the strata will be defined if previous estimations on the variability of the variables of interest have been entered in the program instead of the variability of the environmental variables. A compromise allocation is then provided. The last operation of the program is to select randomly the sampling depths in agreement with the limits of strata and with the number of sampling units allocated to each stratum. Hence, the application of the entire procedure involved in the program provides an optimal stratified random sample for estimating oceanographic parameters. Because the marine environment and the experimental location are continuously changing during an oceanographic cruise, it is suggested to use this sampling strategy on board of the ship with the support of a micro-computer. An illustrative example based on experimental data is provided in order to evaluate the program. A specific study is made on the vertical distribution of the oxygen and on the estimation of the associated parameters.