Setting a critical line that incorporates the concept of probability rainfall

Takeharu SATO and Hiroaki TAKEMOTO

We started using cooperation methods with the intention of providing sediment disaster warning information nearly10years ago. However, setting a critical line using this method is problematic in the case of heavy rains. Also, it has been pointed out that the meaning of any such critical line is unclear. In this study, we examined a range of values at a radius of4km from the Okayama Prefecture Tamano Observatory. We analyzed rainfall data to determine when the critical line was crossed, and found that some heavy rains occurred within the critical region. Due to this fact, and noting the shape of the functions using the RBFN method, which is central to the cooperation method, we proposed basing the response surface on a circle. When such a surface was considered, the critical line was not crossed during times of heavy rain at which no disaster occurred. Additionally, there is a problem with the proposed cooperation method that makes the RBFN line unclear. To solve this problem, we calculated the RBFN value for each rainfall incident. Then, we evaluated the RBFN value in terms of probabilities corresponding to the yearly probability of crossing the critical line. For example, it is possible to set the critical line in terms of the probability of it being exceeded over20years. Our proposed critical line differs to the previous version with respect to short]term rainfall measures. Our formulation enables us to calculate the probability of the critical line being exceeded. This calculation takes into account the soil conditions and the effect of60minutes of continuous rainfall.

Key wordsFcritical line (CL), radial basis function network (RBFN), probability year