Journal of the Marine Biological Association of India

Aspects of statistical methods in oceanography.

Volume 14 Issue 1

J. D. Kerr

Abstract

Statistical methods comprise design and analysis which together provide a sound scientific approach to the collection and analysis of the data and the scrutiny of hypotheses. Modern theory and techniques grew out of agricultural research and has spread to all branches of science. This paper looks at examples in which statistical methods have been applied to oceanographic problems, chiefly from the Indian Ocean. These have initially enabled a more objective but nevertheless still descriptive approach to the consideration of data by distinguishing between systematic and ' random ' variation. Here 'random' variation is that which cannot be related to such factors as latitude, depth and season.

Some of this results from measurement and estimation errors and small-scale sampling variation. Other variation is a systematic deviation from the simple model which a more adequate model based on causative factors would explain. While the descriptive approach does little to increase the understanding of basic causes and mechanisms, it aids the recognition of the important features of the data and reduces the likelihood of inferences based on unreliable data.

 

The great potential value of statistics lies in the development of methods for model building and testing, and in the suggestion of new hypotheses. The analysis essentially becomes a multivariate examination of the relations between factors because of the complex relationships between the physical, chemical and biological attributes. Such methods, whether new or adapted from other ecological studies, need to take account of the three dimensional structure of the ocean and of the dependence on time. Examples are given of some preliminary work in this field.

Design is of prime importance in the collection of data since it determines the inferences that may be drawn from it. In oceanography, sampling is essential and work in adapting suitable methods has hardly begun. Existing theory is well designed for land based problems

Such as economic surveys, but is of little use in oceanic work due to at least three major problems :

(1) the properties being measured may change markedly during the course of the survey ;

(2) the sampling must be conducted from one or at most a few ships, slow moving in

relation to ocean size ;

(3) special problems resulting from ocean movement which make almost impossible

many time and space studies quite practicable on land.

Approaches to this problem are tentatively explored. Since ship time is expensive and in short supply, the potential gains are great. However, an imaginative and intensive investigation is required.

Date : 30-06-1972