[Long-term ecological research]
Forest research requires very long period simply because of enormous lifespan of trees. In order to examine forest community structure and its time-development, we established permanent plots, marked all trees inside, and have repeatedly measured their growth, death and birth. This project analyzed such spatio-temporal forest data in Shiretoko (Hokkaido), Furano (Hokkaido), Hakoda (Aomori), Yatsugatake (Nagano), Ashiu (Kyoto), Tsushima (Nagasaki) and Yamubaru (Okinawa). Here let us briefly explain some of the results about 12-year monitoring in Tsushima Island, southern Japan.

[Size-dependent mortality]
Trees generally produce tons of seeds and there appear huge number of seedlings, although most of them die soon or later, and a very small part of them survive and grow to an adult stage. Hence, mortality rates are high when young and small, and should be decreasing along age or size. In addition, such patterns should differ depending on species. When we estimate mortality for each species, especially for a forest with abundant species, thousands of tree inventory are divided into species and size classes, then each category contains only a few trees. It is quite often when mortality of some class of some species is 100% because we observed only one dead tree, and 0% mortality because of one survivor. Then we would draw highly fluctuating mortality graph with missing parts like Fig. 1a. If size classes are broad, in contrast, we may miss some important traits (Fig. 1b).

[Akaike Bayesian Information Criterion]
Applying Bayesian binary analysis and Akaike Bayesian Information Criterion (ABIC), we can select a smooth curve that most appropriately reflects given data based on information theory. Fig. 2 illustrates size (diameter at breast height)-dependent mortality for three species in Tsushima forest, showing that the three each has each own life-history. Mortality of Quercus salicina (evergreen oak) is just decreasing along size while Castanopsis cuspidate (the most dominant evergreen species in this forest) exhibited an increase when it becomes very large. In contrast, Podocarpus macrophyllus (evergreen conifer) does not change the mortality regardless of its size.

[Maintenance of species diversity]
We found the diversity among mortality patterns. Probably, such diversity, together with other various ecological mechanisms, contributes to maintaining high species diversity of this warm-temperate forest, and biodiversity is a result of their spatio-temporal combinatorial effects. Therefore, we are investigating forest ecosystem in order to develop sustainable forest management and conservation by long-term field measurement and data analysis.

Members
SHIMATANI Kenichiro (The Institute of Statistical Mathematics), MANABE Toru (Kitakyushu Museum and Institute of Natural History), KAWARASAKI Satoko (Seikei University),
KUBOTA Yasuyuki (Kagoshima University), GOTO Susumu (Tokyo University), KATO Kyoko (Hokkaido University),
HIRAYAMA Kimiko (Forestry and Forest Products Institute), AIKAWA Shinichi (Ibaraki University)