2002 年 55 巻 4 号 p. 23-28
Decision-making of evacuation from natural disasters, especially sediment-related disasters is surrounded by many uncertainties. A simple response system and a simple decision model are introduced in order to specifically analyze the uncertainty of forecast precipitation, which has a significant impact on the decision. The response system has precipitation as its input variables, geomorphologic and geological aspects as its parameters, and a signal of occur/not occur as its output. The decision model constitutes of a decision node (evacuate/not evacuate), uncertainty nodes (rain exceeding a critical amount), and consequence nodes (presumed loss of lives and cost of evacuation) . Operational forecast and actual precipitation used by Japan Meteorological Agency (JMA), which are called “Very Short-Range Forecast of Precipitation” and “Radar-Automated Meteorological Data Acquisition System (AmeDAS) Precipitation” respectively, are examined to verify their quantitative accuracy for the period from January 1998 to January 2001. Conditional probabilities of a precipitation amount exceeding a given critical amount when a certain amount of rain was forecast are obtained through the verification. The conditional probabilities have to go beyond the cost-loss ratio of evacuation decision-making in order for the evacuation decision to stand rational. The accuracy level of 1-hour operational forecast precipitation barely meets the required level indicated by cost-loss ratio for 5 by 5km square mesh, when a perfect response system is assumed. Importance of further research on (subjective) cost-loss ratios of actual decision-makers and sizes of areas in concern came to our attention.