Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 181-188 (2013)

A Trial for Evaluating Effects of Climate Change on National Scale Carbon Dynamics Using a National-level System for Simulating Forest Carbon Dynamics

Yasushi Mitsuda
(Shikoku Research Center, Forestry and Forest Products Research Institute; Now at Faculty of Agriculture, University of Miyazaki)
Hidesato Kanomata
(Department of Forest Policy and Economics, Forestry and Forest Products Research Institute)
Mitsuo Matsumoto
(Principal Research Coordinator, Forestry and Forest Products Research Institute)

We simulated forest carbon dynamics on a national scale using a forest carbon dynamics simulation system. We developed a simulation system consisting of a national-level forest database and a stand-level forest growth model. We developed a forest database from information on species, age and stand structure for each 1-km grid for all of Japan based on the National Forest Resource Database. The stand structure was estimated using the National Forest Inventory. A matter-balance-based stand-level forest growth model was adopted for the system. Based on predicted and average climatic values, we simulated the carbon dynamics of sugi (Cryptomeria japonica) stands on a national scale from 2005 to 2050 using the system. Simulated carbon stocks and its flux with predicted climatic values were lower than that simulated using average climatic values.

Key words: Carbon stock, regulating services, National Forest Resources Database, National Forest Inventory, Cryptomeria japonica planted forest.

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 189-200 (2013)

Snow Damage Analysis by Discrete Regression Models

Ken-ichi Kamo
(Center for Medical Education, Sapporo Medical University)
Akio Kato
(Toyama Prefectural Agricultural, Forestry and Fisheries Research Center)
Atsushi Yoshimoto
(The Institute of Statistical Mathematics)

The objective of this paper is to evaluate the risks of snow damage to Toyama prefecture's forests. It is well known that the risk probability of snow damage is affected by several factors including climate, weather, geography and forest stand conditions. In order to evaluate the risk of snow damage, discrete regression models, which are logistic regression and multinomial regressions, are applied. The factors that essentially affect risk probability are specified throughout the model selection procedure. The results indicate that areas with little wind or those with thin forest stands may have a high-risk probability of snow damage. The conclusions from this evaluation regarding risk probability should have broad implications on management techniques such as species control or thinning that are presently used to minimize snow damage.

Key words: Logistic regression model, multinomial regression model, Akaike's information criterion, model selection, snow damage, risk probability.

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 201-216 (2013)

Sustainable and Adaptive Forest Management and Data Infrastructure under Stand-based Silvicultural System

Toshiaki Owari
(The University of Tokyo Hokkaido Forest)

The University of Tokyo Hokkaido Forest (UTHF) owns and manages approximately 20 thousand hectares of forestland in central Hokkaido, northern Japan, and has been implementing a management experiment called `Rinbun Segyo Ho' (stand-based silvicultural system) since 1958. This paper synthesizes type and contents of forest dynamics and management data accumulated at UTHF through more than 50 years' management practices in order to contribute to the development and arrangement of field information infrastructure, which is essential to risk assessment and research for forest ecosystem management. Under the forest management procedure of Rinbun Segyo Ho, the UTHF staff annually conduct 1) ground forest surveys for spatially-explicit stand classification, 2) field measurements of forest inventory plots, and 3) single-tree selection for harvest every year. The stand-scale data of forest resource management has been recorded from each investigation. In UTHF, the individual-scale permanent measurement plots including 1) natural forest management research plots and 2) long-term and large-scale ecological research plots have been established and periodically measured. Various types of geospatial data including aerial orthophotos and digital elevation models (DEM) have been prepared. Through multi-disciplinary researchers' collaboration with the use of this data infrastructure, possible risks in forest ecosystem management may be assessed quantitatively by properly modeling the stand dynamics and the responses to management practices.

Key words: Adaptive management, sustainable forestry, stand-based silvicultural system, permanent research plot, geospatial data.

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 217-231 (2013)

Environmental Sciences and BIG Data
—Current Status and Future Perspective on Biodiversity Informatics in Japan—

Takeshi Osawa
(Natural Resources Inventory Center, National Institute for Agro-Environmental Sciences; Japan Node of Global Biodiversity Information Facility)
Utsugi Jinbo
(Department of Zoology, National Museum of Nature and Science; Japan Node of Global Biodiversity Information Facility)

Recently many large databases have been developed for research in environmental science fields. To make meaningful use of such large databases, we need three components: 1) large databases, 2) data management techniques and 3) analyzing techniques. However, 2) data management techniques are often viewed as unimportant especially in biodiversity science fields. In this review we explain the technical aspects of data management as well as show the current status of information technology in biodiversity fields in Japan. Accordingly, we discussed future perspectives in developing an infrastructure on biodiversity informatics in Japan.

Key words: Biodiversity informatics, cross-use, Darwin Core, data format, metadata, standardization.

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 233-246 (2013)

Derivation of Dose-response Relationships for Risk-risk Trade-off Assessment Regarding Substitution of Lead-free Solder for Lead Solder

Jun-ichi Takeshita
(Research Institute of Science for Safety and Sustainability (RISS), National Institute of Advanced Industrial Science and Technology (AIST))
Masashi Gamo
(Research Institute of Science for Safety and Sustainability (RISS), National Institute of Advanced Industrial Science and Technology (AIST))

In this paper, we investigate the substitution of lead-free solder for lead solder, and calculate the loss of QALY (quality adjusted life years) per unit increase in exposure from the baseline exposure of four metals: lead, copper, tin, and silver. In risk assessment regarding metals, we typically require information concerning the dose-response relationship for humans, with respect to the characteristic hazardousness of each metal. However, no such information is currently available on the four metals in question. Thus, in this study, we employ a methodology, proposed by ourselves, for calculating the loss of QALY through exposure to chemical substances. More precisely, we determine the dose-response relationship for humans for each internal organ, for each substance, based on (a) the dose-response relationship for humans of the reference substance, established for all principal internal organs, and (b) the toxicity value of the target substance relative to that of the reference substance. To calculate the relative toxicity values, we follow the following procedure: We first design a covariance structure modeling using the NOEL (No-Observable-Effect Level) values of the existing animal testing data, as the training set. Then, we infer the NOEL values of the endpoints, for which animal testing data is not available, by using the optimal predictive equation with the implied covariance structure. Finally, we calculate the relative toxicity value as a ratio of the NOEL value of each metal to one of the reference substances. We select vinyl chloride monomer and cadmium, respectively, as reference substances for liver and kidney effects.

Key words: Risk-risk trade-off assessment, substitution, relative toxicity value, loss of QALY, covariance structure analysis.} \end{Etitle

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 247-256 (2013)

Relevance of Degree of Radiation Contamination of Soil and Air Radiation Dose Rate in Neighborhood of Fukushima Daiichi Nuclear Power Plant

Megu Ohtaki
(Research Institute for Radiation Biology and Medicine, Hiroshima University)
Keiko Otani
(Research Institute for Radiation Biology and Medicine, Hiroshima University)
Tetsuji Imanaka
(Kyoto University Research Reactor Institute)
Satoru Endou
(Graduate School of Engineering, Hiroshima University)
Masaharu Hoshi
(Professor Emeritus, Hiroshima University)

The Ministry of Education, Culture, Sports, Science and Technology in Japan conducted a survey on the soil radioactive contamination and air radiation dose rate in the neighborhood of the Fukushima Daiichi nuclear power plant in June-July, 2011. We thus analyzed the association between the air dose rate and the soil radiation contamination. When precipitation occurred on the measurement date, the determination coefficient became 64%, and it reduced to 28% when there was no precipitation. According to this analysis, about 20% to 35% of air radiation dose rate cannot be explained by the degree of soil radioactive contamination directly beneath. It suggests that local decontamination may have a limited effect on reducing the air radiation dose rate.

Key words: Air dose rate of radiation, decontamination, Fukushima Daiichi nuclear power plant, radiation soil contamination, regression analysis.

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 257-270 (2013)

Data Assimilation System for Seismoacoustic Waves — Verification by a Twin Experiment —

Hiromichi Nagao
(The Institute of Statistical Mathematics; Now at Earthquake Research Institute, The University of Tokyo)
Tomoyuki Higuchi
(The Institute of Statistical Mathematics)

The off-Tohoku Pacific Ocean Earthquake and especially the accompanying giant tsunamis, which occurred on March~11, 2011, seriously damaged the Pacific coast in the Tohoku region. Since a Tokai-Tonankai-Nankai Earthquake and giant tsunamis are expected to occur in the future, researchers in various fields are making efforts to minimize damage by putting our experience to use. Large-scale oscillations in the Earth's atmosphere including the ionosphere probably excited by the tsunamis were observed by highly sensitive barometers and GNSS receivers. This is expected to realize a tsunami early warning system, which will predict the magnitude of an approaching tsunami and its arrival time by utilizing the feature that acoustic waves propagate faster than tsunamis. The present paper introduces a data assimilation system that infers model parameters related to the hypocenter and states of propagating seismoacoustic waves at every time point by integrating numerical simulation and observed microbarometer data, and verifies the system by a twin experiment, in which the hypocentral parameters of a synthetic earthquake that models the forthcoming Tonankai Earthquake are properly estimated.

Key words: Data assimilation, MCMC, the off-Tohoku Pacific Ocean Earthquake, infrasound, seismoacoustic wave, tsunami.

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 271-287 (2013)

Analysis of Data with Many Zero-valued Observations: Over-estimation of Temporal Trend by Negative Binomial Regression

Mihoko Minami
(Department of Mathematics, Keio University)
Cleridy E. Lennert-Cody
(Inter-American Tropical Tuna Commission)

In ecological and environmental studies, count data such as the number of animals per unit area or unit effort often contain many zero-valued observations. Such data unfortunately may be analyzed without any special consideration given to how the zeros arose. In particular, the negative binomial regression model has been a commonly used model for count data with overdispersion. However, we found that the negative binomial regression model over-estimated temporal trends in species relative abundance. Such over-estimation could be problematic, for example, for the development of management guidelines for conservation.

In this paper, we investigate this phenomena of over-estimation. We show that when the negative binomial regression model is fitted to data with excess zeros, the estimate of the size parameter becomes too small and the observations with small fitted values have more influence. This results in estimated coefficients of predictors that are too large in absolute value, and it produces exaggerated estimates of the marginal effects.

Key words: Over-dispersion, influence function, size parameter, zero-inflated negative binomial regression model, Cook's distance, leverage.

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 289-305 (2013)

Use of Directional Statistics in Environmental Science

Kunio Shimizu
(Department of Mathematics, Keio University; Visiting Professor, The Institute of Statistical Mathematics)
Minzhen Wang
(School of Fundamental Science and Technology, Keio University; Now at The Institute of Statistical Mathematics)

Wind direction is a typical and important angular variable in environmental science. Sometimes observations include wind speed and ozone concentration together with wind direction. This article is aimed at obtaining fundamental results in directional statistics that deal with modeling and analysis of data, including angular observations and a review of recent studies in this area.

The difference between the sample mean direction for angular data and the sample mean for linear data, and that between the mean resultant length and the variance, are emphasized. We use wind direction data available at the site of the National Institute for Environmental Studies to illustrate circular plots and calculations of basic statistics. Pewsey's test for symmetry, a flexible symmetric angular distribution by Jones and Pewsey, and the method of sine-skewing are presented, and an illustration of fitting angular distributions using the wind direction data is shown. Papakonstantinou and Batschelet distributions, which have the flat-topped and sharply-peaked properties, are introduced. The article also discusses circular-linear regression models derived from distributions on the cylinder, circular-circular regression models, and circular-circular structural models.

Key words: Angular regression, angular structural model, distributions on the cylinder, distributions on the torus, fitting angular distributions.

Proceedings of the Institute of Statistical Mathematics Vol. 61, No. 2, 307-322 (2013)

Parameterization of a Forest Stand Growth Model Using Long-term Field Survey Plot Data

Yasushi Mitsuda
(Shikoku Research Center, Forestry and Forest Products Research Institute;
Now at Faculty of Agriculture, University of Miyazaki)
Kazuo Hosoda
(Department of Forest Management, Forestry and Forest Products Research Institute)
Toshiro Iehara
(Department of Forest Management, Forestry and Forest Products Research Institute)

We developed a matter balance based stand-level forest growth model for Cryptomeria japonica planted forests using long-term field survey plot data. This model consists of six processes: (1) photosynthetically active radiation absorption; (2) conversion to gross primary production; (3) constraints on photosynthesis by environmental factors; (4) respiration; (5) litterfall and root turnover; and (6) biomass partitioning. For parameterization, we estimated time-series biomass and biomass growth data using repeated measurements of long-term field survey plots. We applied Bayesian calibration methodology to parameterization in our process-based model using time-series biomass and biomass growth data and time-series climatic values of solar radiation, mean temperature, and vapor pressure deficit. The results of model validation indicate that, although some large over-estimation errors were observed, our model could represent patterns of biomass growth measured in long-term field survey plots.

Key words: Pattern-oriented modeling, Bayesian calibration, Cryptomeria japonica planted forest.