4. CONCLUSIONS
The hydrological data of flash flood events were obtained from the WH system developed in JRV. The statistical analysis quantified the correlations among the variables and their probability distributions. A strong correlation was not found among the variables, implicating that the spatial rainfall distribution was uneven and that the runoff process was nonlinear and highly stochastic. Researches on those physical phenomena shall be conducted in the follow-up studies, but we recommend direct monitoring of runoff discharges rather than rainfall depths for practical desing of WH systems. Probability distribution fitting was successful with LN3 or GEV, and the design of the WH system was verified. The spillway was designed with a return period of about 12 years but well functioned for the historical flash flood event with a return period of about 26 years. The design of a spillway based on the 200 years return period might be excessive in the context of cost-benefit analysis if the failure of the hydraulic structures would not give rise to significant consequences. The obtained knowledge of hydrology in JRV will be useful for designing and operating WH systems in the future.