Statistics and Physical Oceanography (1993)

Offered for the purpose of encouraging successful collaborations between statisticians and oceanographers, the following conclusions, observations, and suggestions are based on information that the Panel on Statistics and Oceanography gathered in this study, on the panel discussions that took place in preparing this report, and on the panelists’ own experience and knowledge concerning cross-disciplinary research and collaborative efforts. The panel believes understanding and appreciating these matters are as important to the encouragement and accomplishment of statistical research in physical oceanography as are the descriptions of statistical research opportunities discussed in Chapters 2 through 8.

1. There are many opportunities for statistical research in biological, chemical, geological, and physical oceanography, far more than this report can address (owing to constraints of time and resources). This report thus represents a first step, focusing on challenging statistical issues in physical oceanography. However, the statistical problems it describes are universal, and progress on them would benefit the other oceanographic disciplines and also contribute to a better understanding of the coupled ocean-atmosphere system, weather patterns, and global climate change.

2. Many sophisticated statistical techniques are used routinely in physical oceanography. Nevertheless, in numerous general areas collaboration between oceanographers and statisticians could contribute to improving currently used models, analysis techniques, data assimilation methods, visualization methods, and so on. Examples of such areas identified in this report include multiple-scale variability of oceanographic fields; use of Lagrangian data in descriptions of ocean circulation; ocean feature identification; pictorial representation of oceanographic data; interpolation, smoothing, filtering, and prediction in the context of oceanographic data; comparison of oceanographic models and data; and non-Gaussian, nonstationary random fields.

3. Identifying research areas of mutual interest and need is basic to achieving results of genuine value to all participants in cross-disciplinary projects; another crucial requirement is providing an environment that encourages and sustains individuals who embark on collaborative research. Although exploring this second issue was beyond the scope of this study, the panel became increasingly aware during its deliberations of just how difficult it can be to engage in truly collaborative, cross-disciplinary work. There are many possible reasons for such difficulties (see, e.g., NRC 1990a): different parties in a cross-disciplinary collaboration may have different motivations or different disciplinary imperatives; there may be institutional impediments due to the traditional organization of separate disciplines within

an institution; there may be inherent obstructions to peer-reviewed funding or publishing of cross-disciplinary research (for instance, in defining what constitutes a peer); and there may be contextual scientific obstacles (since the multifaceted system under study may not fit into traditional categories for scientific investigation).

Without attempting to specify particular remedies, the panel includes below a few generic observations and outlines some possible initial approaches to encouraging collaborative research, especially between statisticians and oceanographers. The recent publication of several excellent studies and reports addressing cross-disciplinary research in various contexts (e.g., NRC, 1987; Institute of Mathematical Statistics, 1988; NRC, 1990a; see also Goel et al., 1990; Gnanadesikan, 1990; Hoadley and Kettenring, 1990), together with heartening signs of an improving environment for such activities (Crank, 1993; Harris, 1993), suggests that attention to the value of collaborative research is increasing and that work toward facilitating it will be ongoing.

1. The need for clear communication and substantive interaction among collaborating researchers from different disciplines suggests the desirability of their working together at the same physical location for a significant period of time on specific problems to which both parties can contribute needed expertise. Funding agencies and research institutions could stimulate such interactions (a) by sponsoring workshops on well-delineated topics—drawn, for example, from the research areas discussed in this report—that are best addressed by a collaborative effort; (b) by providing for postdoctoral fellowships, senior research sabbaticals, and graduate student residencies that would enable statisticians to work with oceanographers at oceanographic research institutions; and (c) by sponsoring a series of one-or two-week short courses on oceanography for statisticians in which specialists would review selected topics and indicate open areas of research. It is much more likely that statistical research on one of the physical oceanographic challenges described in this report will produce valuable results if that research involves continuous interaction with an oceanographer who is versed both in the nuances of that challenge and in the practical oceanographic realities surrounding it.

   In all such considerations, the panel encourages active cooperation between statisticians and oceanographers at agencies that fund research in these disciplines.

2. Effectively communicating the results of successful collaborative research—and thereby increasing understanding of its value in addressing complex problems—includes having the results published in journals that are well regarded in the relevant disciplines. The panel suggests that, as an initial step, one or more of the major statistical journals could publish a special section or issue on statistics and oceanography designed to increase awareness of the research opportunities in that area. This would encourage interaction between statisticians and physical oceanographers, increase the visibility of the results of successful collaboration, and set a precedent that could stimulate other highly regarded disciplinary journals to publish statistics and oceanography cross-disciplinary papers.

3. Promoting and nurturing cross-disciplinary research in statistics and physical oceanography, which will likely involve broadening the educational base of prospective researchers as well as the criteria by which their later efforts are rewarded, can be fostered now (a) by university statistics departments that stimulate cross-disciplinary interactions and learning and encourage statistics undergraduate and graduate students to obtain an “applied” minor in some other area, with oceanography being but one possibility (others being physics, engineering, geology, and so on), and (b) by funding agencies that promote a broader orientation in graduate and undergraduate statistics education.

   It is likely that many people will be encouraged to undertake the significant efforts interdisciplinary statistics and oceanography research requires if funding agencies offer prospective cross-disciplinary collaborators some likelihood of obtaining research support, if recognized journals in an individual’s discipline offer sufficient flexibility in publishing such cross-disciplinary research papers, and if research institutions accord cross-disciplinary research the same level of professional recognition (in promotion and tenure considerations) as is currently given to research in the individual disciplines.

Many major national and global concerns involve scientific research challenges that are cross-disciplinary in nature, with weather prediction and global climate change being but two examples related to the focus of this report. Encouraging the pursuit of such cross-disciplinary research opportunities can benefit both science and society by focusing scientific attention on research issues relevant to societal concerns. Encouraging the pursuit of cross-disciplinary research opportunities in statistics and oceanography will certainly benefit both disciplines: application of sophisticated statistics techniques will lead to better descriptions and improved dynamical understanding of oceanographic phenomena, and the statistics research challenges presented by oceanographic issues will inspire the development of new statistical techniques.