A statistical study for ecosystem modeling focused on marine mammals

Most of cetacean species are highly migratory and difficult to keep in captivity so that it is not easy to grasp precise status of the populations. On the other hand, the limitation of the management based on single species has been recognized and the demand on the ecosystem-based population assessment and management has been increasing over the world. It is important to know the status of the population precisely and accurately as far as we can in ecosystem-based approach. To extract information efficiently from the data in which a lot of uncertainty exist and grasp the status of the population precisely and accurately, the use of statistical models is essential. This paper focuses on population assessment of marine mammals and develops some statistical models. Furthermore, the competition between marine mammals and fishery is investigated using an ecosystem model. Marine mammals are one of the important top-predators in an offshore ecosystem. The information on the distribution, migration, and stock structure is therefore indispensable to the construction of the offshore ecosystem model. In addition, the degree that the biological resource is affected by the catch greatly varies with the assumption of the stock structure in the procedure developed by the International Whaling Commission (IWC) for the management of large baleen whales. The regression analyses based on the generalized linear and additive models are useful for modeling the complicated phenomena such as the temporal-spatial structure of biological resources. In this paper, the spatial distribution and seasonal changes of the density of the western North Pacific minke whales was estimated using the generalized additive model so that there was no evidence that multiple stocks exist in the western North Pacific. On the other hand, the seasonal change of the density reproduced the aspect of migration of the minke whales well. The investigation of temporal-spatial distribution of biological resources using the regression models can be extensively useful for other species other than marine mammals. Line transect sampling is a standard method for estimating the population size of marine mammals. One of the important assumptions in standard line transect sampling is that all animals on the trackline are detected without fail. However, the surfacing-diving behavior of marine mammals for respiration and feeding can lead to failures in detection even if they are on the trackline. As a result, the population size tends to be underestimated. Although the underestimation of the population size brings conservative effect on single-species management, it leads to another problem in ecosystem approach because the impact by consumption of predator is underestimated. Conventional methods for estimating the detection probability on the trackline have several problems. The flexible method developed in this paper modified and improved such problems. Simulation studies showed that the method is promising. In addition, it was confirmed that the present method is able to be applied to real data sets using Antarctic minke whales data. The prey-predator interaction is extremely important in ecosystem modeling. If we can know preference of animals in laboratory, we can estimate diet composition of animals non-lethally by putting the information on the preference and the availability of animals obtained from sighting survey together. Preference for multiple food items can be measured in the laboratory by choice tests like cafeteria trials, in which an animal is initially offered a variety of foods that are equally available at the same time. However, the results of the multiple-choice experiments can be affected by many factors such as placement, amounts of foods, and the length of time that the experimental animal is permitted to eat. An alternative way to determine multiple-preference is to repeat a series of paired comparisons. The paired comparisons have an advantage over the multiple-choice tests because the design of each trial is simpler so that the conditions of the experiment can be easily controlled. The data collected from paired comparisons can be analyzed with the Bradley-Terry model. However, conventional Bradley-Terry model has some defects; for instance, it is not able to deal with the amount of consumption. This paper proposes a modification of the estimation method used in the Bradley-terry model. The maximum likelihood estimation based on a binomial distribution in the original model is replaced with the nonlinear least squares method. Furthermore, some simple methods to estimate the precision of parameters and to test several hypotheses on preferences are provided. The methods were applied to the analysis of food preference of a captive northern fur seal. The comparison between the method and the conventional Bradley-Terry model indicated that the statistical power of the method can be higher than the conventional one. Finally, the ecosystem model off ...