The skewness and/or bimodal nature of the data distribution is usually not considered. Also, statistics of "extreme values" are hardly investigated, and we have no idea to discuss statistically the "extreme values having physical significance". We shall introduce a try to handle extreme values statistically in the next section. Of course, we do not recommend to develop any special technique to handle such special data distribution nature and "extreme values" in a routine data management system. However, if we clarify their nature, we can understand much precise meaning of the basic quantifies such as "mean and standard deviation". We have lots of research items relating to data management, and this is the reason that the term "research" is included in the name of MIRC.
It should be noted that "quality control" does not mean simply to put quality flags to each data or each data set. We need to improve the quality of the data flowing into JODC and MIRC. Enlightening and popularization of marine knowledge is one of the main tasks of MIRC, by using various data products both for the general public and for scientists and engineers. Moreover, we need to establish the consulting and educating systems on management of the oceanic instruments, observation and data handling techniques, and quality control tactics, in order to improve data quality. Data checking programs which are easily used by researchers in not established organizations would be one of the first products of MIRC.
International co-operation is essential to manage the data as our targets are huge oceans. A high quality data set of the world oceans is required for climate change investigations. We need to set the standard techniques for quality control and related formats in order to improve data flow in the world, but these systems should be renewed according to the development in science and technology, and to changes of demands of various users. MIRC should join to and contribute for this international games. Enlightening or popularization business should be done also in an international sense. MIRC hopes to contribute to the education and training programs of oceanographically developing countries.
There are huge volumes of valuable oceanographic data which have not been collected by JODC. We should be involved in data rescue also. The present data management business is rather limited in physical quantifies. We need to expand our activity also for biological and chemical fields. There are also many research items relating to data management.
AN EXAMPLE OF EXTREME VALUE ANALYSIS
The water salinity of the North Pacific Ocean is much lower than that of the Atlantic Ocean, and saline water higher than 35.00 is scarcely observed in the vicinity of Japan. However, Takeuchi (1984) reported such water was found off the Kii Peninsula on August 2, 1984 on board R/V Wakayama, though he never observed such abnormal water thereafter. In order to check whether such high salinity water is real or not, we analysed the historical data compiled by JODC after 1975 in the region shown in Fig.2. We found 13 cases in total for the analysed area and period. Among these, R/V Soyo-maru of the Japan Fisheries Agency observed such water on the same day as R/V Wakayama. (In a preliminary survey, we found more than ten observation points where the abnormal water was observed in the Japan Sea. The water of the Japan Sea is much less saline as the water has been influenced by the coastal waters in the East China Sea. All of the vessels reporting such data belong to a non-research organization. This means that the analysis of abnormal waters can be used for judging quality of the reliability of organization involved.)
We selected salinity values from 34.70 through to 35.00 at intervals of 0.05. The occurrence frequency of observations where salinities was higher than each of these values was computed. The frequencies in % are shown in Fig.3 for each sub-domain shown in Fig.2 (Morikawa et al., 1997). We classified the decreasing manner of the curves into three groups; coastal water, offshore (subtropical) water and boundary water (these are different lines and different shades in Figs. 2 and 3). The decreasing manners in Fig.3 suggest that the salinity values lower than 34.90 would occur due to usual variability of the salinity, in this analysed area, but saline water higher than 34.95 appears to be abnormal, even though its existence appears to be real. The nature of these curves relates to the skewness of the data distribution. We are analyzing the relation between such abnormal values and the standard deviation of the data distribution. Anyhow, this is one of the examples that we can use the detailed distribution nature as a tool of quality control.
