Research on the impact of climate change on our society is becoming more important in the current situation. Crops are not only particularly susceptible to climate change, but also they are the basis of people's lives, and from the perspective of food security, they are also important issues that can be said to be one of the most important elements in society. On the other hand, the studies on the relationship between climate change and crop productivity are insufficient despite the importance of analyzing it from a more elaborate statistical perspective. This paper aims to promote and help researchers with statistical and informatics backgrounds participate in studying the impact of climate change on crop productivity. First, we introduce the particularity of the study of climate change, and then point out that the current statistical analysis on crop productivity is inadequate in dealing with factors other than meteorology. We also briefly discuss the application of machine learning analyses and the relationship between process-based crop models and statistical models.

Toward assessing climate change impact on extreme meteorological events, database for policy decision making for future climate change (d4PDF) was developed and 3000-year climate reproduction data is provided for the present climate. This accelerated impact assessment studies on floods, initiating multi-basin simultaneous flood risk. However, there are few applications of multivariate extreme value theory on flood frequency analysis and no case for d4PDF. This study applied bivariate extreme value distribution to d4PDF annual maxima of flood peaks for two sets of pairs (the Tone and Ara Rivers from the Kanto region, the Kuma and Midori River from the Kyushu region). Nine distributions with different dependence function were tested. As a model selection is hard for bivariate cumulative probability or dependence function, the two sample Kolmogorov-Smirnov test was applied to the probability distribution of the angular component. Although d4PDF is annual maxima, targeting far longer block size such as 30-year or 60-year maxima, the sample of angular component is approximated by the original annual maxima. As the result, the models well fitted to the samples of the angular component showed consistent χ value at around 0.2 in the Kanto rivers and 0.3 to 0.4 in the Kyushu rivers, indicating higher dependence between the rivers in the Kyushu region.

This article discusses world grid square statistics and data, architecture design,and methods of empirical proofs of autonomous distributed data platform for world grid square statistics from four aspects of computational systems, organizational issues, relationships among data flows, and ecosystem to enable us to sustainably maintain them. Our trials for tackling SDGs by using the proposed distributed data and statistics platform in our project are addressed.