The authors have developed a probabilistic risk assessment method on a forest fire as one of external hazards. A hazard curve by heat effect of a forest fire had been obtained by using a logic tree in our previous study. The main application target of the forest fire probabilistic risk assessment is for sodium-cooled fast reactor systems. Databases for a hazard curve evaluation were based on forest fire records, meteorological and vegetation data of a studied area which is near a typical sodium-cooled fast reactor in Japan. There are two intensity parameters of heat effect of a forest fire, namely, reaction intensity and frontal fireline intensity. The hazard curves of these two intensities obtained in our previous study were referred to as “reference case” where constant breakout frequency throughout a day, equal probability distribution for potential breakout points, and firefighting effect on a forest fire were assumed as a priori. The reference reaction intensity and the fireline intensity became 935 kW/m2 and 107 kW/m for the annual exceedance frequency of 10−4/year, respectively.
This paper describes a sensitivity study of the hazard curves on condition parameters where frequency/probability variables in the logic tree were varied within respective fluctuation ranges in order to evaluate quantitative effects on the frequency and/or intensity of the hazard curves. As for the forest fire breakout frequency and propagation probability, important variables are “fluctuation of breakout time”, “probability distribution fluctuation of breakout point”, and “firefighting effect on a probability of forest fire arrival at a nuclear power plant (NPP)”.
The intensities increase in daytime due to sunshine, and the breakout probability in daytime is statistically 2.8 times higher than a daily average, and that in nighttime is 1/9 of the average. As a result, the hazard curves of the reaction intensity and the fireline intensity increased around 4% and 14% respectively in intensity direction in comparison with those of the reference case. The “fluctuation of breakout time” only affects the intensities of the hazard curves, but not the frequency.
As for the “probability distribution fluctuation of breakout point”, one selected point is given higher probability than the other points. The hazard curves vary around +70% to −40% in frequency direction; each breakout point has different distance to the NPP and the forest fire arrival probability varies with a propagation duration. Namely, the longer duration, the higher probability of the extinguishment by firefighting, accordingly the lower probability of the arrival at the NPP. The “probability distribution fluctuation of breakout point” affects only the frequency of the hazard curves, but not the intensities.
“Firefighting effect on a probability of forest fire arrival at an NPP” was conservatively assumed for the sensitivity study in which there is no firefighting action outside the NPP, hence all potential forest fires arrive at the NPP. The hazard curves remarkably increase around 40 to 80 times in frequency direction in comparison with those of the reference case. This is because most of forest fires in Japan are extinguished within one to two hours by fire departments, and the conditional probability of a forest fire arrival at an NPP from a potential breakout point with kilometer range distance was evaluated to be very low (i.e. less than a few percent). The “firefighting effect on a probability of forest fire arrival at an NPP” only affects the frequency of the hazard curves, but not the intensity.
This study indicated that the most significant factor in the forest fire hazard curve is whether the firefighting action outside an NPP is expected before the arrival at an NPP.